diff options
author | Michael Stapelberg <michael@stapelberg.de> | 2014-04-26 10:08:46 +0200 |
---|---|---|
committer | Michael Stapelberg <michael@stapelberg.de> | 2014-04-26 10:08:46 +0200 |
commit | 663996b3bb3cdaa81ac830674fdda00e28a2485b (patch) | |
tree | ec5ccfae391a7cc7970f7618a58970c87dec770b /src/core | |
download | systemd-663996b3bb3cdaa81ac830674fdda00e28a2485b.tar.gz |
Imported Upstream version 204
Diffstat (limited to 'src/core')
130 files changed, 46053 insertions, 0 deletions
diff --git a/src/core/audit-fd.c b/src/core/audit-fd.c new file mode 100644 index 0000000000..5955bd846e --- /dev/null +++ b/src/core/audit-fd.c @@ -0,0 +1,73 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + + +#include <errno.h> +#include "audit-fd.h" + +#ifdef HAVE_AUDIT + +#include <stdbool.h> +#include <libaudit.h> + +#include "log.h" +#include "util.h" + +static bool initialized = false; +static int audit_fd; + +int get_audit_fd(void) { + + if (!initialized) { + audit_fd = audit_open(); + + if (audit_fd < 0) { + if (errno != EAFNOSUPPORT && errno != EPROTONOSUPPORT) + log_error("Failed to connect to audit log: %m"); + + audit_fd = errno ? -errno : -EINVAL; + } + + initialized = true; + } + + return audit_fd; +} + +void close_audit_fd(void) { + + if (initialized && audit_fd >= 0) + close_nointr_nofail(audit_fd); + + initialized = true; + audit_fd = -ECONNRESET; +} + +#else + +int get_audit_fd(void) { + return -EAFNOSUPPORT; +} + +void close_audit_fd(void) { +} + +#endif diff --git a/src/core/audit-fd.h b/src/core/audit-fd.h new file mode 100644 index 0000000000..8b58289dc5 --- /dev/null +++ b/src/core/audit-fd.h @@ -0,0 +1,25 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int get_audit_fd(void); +void close_audit_fd(void); diff --git a/src/core/automount.c b/src/core/automount.c new file mode 100644 index 0000000000..a20d5340f2 --- /dev/null +++ b/src/core/automount.c @@ -0,0 +1,920 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <limits.h> +#include <sys/mount.h> +#include <unistd.h> +#include <fcntl.h> +#include <sys/epoll.h> +#include <sys/stat.h> +#include <linux/auto_fs4.h> +#include <linux/auto_dev-ioctl.h> + +#include "unit.h" +#include "automount.h" +#include "mount.h" +#include "load-fragment.h" +#include "load-dropin.h" +#include "unit-name.h" +#include "dbus-automount.h" +#include "bus-errors.h" +#include "special.h" +#include "label.h" +#include "mkdir.h" +#include "path-util.h" +#include "dbus-common.h" + +static const UnitActiveState state_translation_table[_AUTOMOUNT_STATE_MAX] = { + [AUTOMOUNT_DEAD] = UNIT_INACTIVE, + [AUTOMOUNT_WAITING] = UNIT_ACTIVE, + [AUTOMOUNT_RUNNING] = UNIT_ACTIVE, + [AUTOMOUNT_FAILED] = UNIT_FAILED +}; + +static int open_dev_autofs(Manager *m); + +static void automount_init(Unit *u) { + Automount *a = AUTOMOUNT(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + a->pipe_watch.fd = a->pipe_fd = -1; + a->pipe_watch.type = WATCH_INVALID; + + a->directory_mode = 0755; + + UNIT(a)->ignore_on_isolate = true; +} + +static void repeat_unmout(const char *path) { + assert(path); + + for (;;) { + /* If there are multiple mounts on a mount point, this + * removes them all */ + + if (umount2(path, MNT_DETACH) >= 0) + continue; + + if (errno != EINVAL) + log_error("Failed to unmount: %m"); + + break; + } +} + +static void unmount_autofs(Automount *a) { + assert(a); + + if (a->pipe_fd < 0) + return; + + automount_send_ready(a, -EHOSTDOWN); + + unit_unwatch_fd(UNIT(a), &a->pipe_watch); + close_nointr_nofail(a->pipe_fd); + a->pipe_fd = -1; + + /* If we reload/reexecute things we keep the mount point + * around */ + if (a->where && + (UNIT(a)->manager->exit_code != MANAGER_RELOAD && + UNIT(a)->manager->exit_code != MANAGER_REEXECUTE)) + repeat_unmout(a->where); +} + +static void automount_done(Unit *u) { + Automount *a = AUTOMOUNT(u); + + assert(a); + + unmount_autofs(a); + + free(a->where); + a->where = NULL; + + set_free(a->tokens); + a->tokens = NULL; +} + +int automount_add_one_mount_link(Automount *a, Mount *m) { + int r; + + assert(a); + assert(m); + + if (UNIT(a)->load_state != UNIT_LOADED || + UNIT(m)->load_state != UNIT_LOADED) + return 0; + + if (!path_startswith(a->where, m->where)) + return 0; + + if (path_equal(a->where, m->where)) + return 0; + + r = unit_add_two_dependencies(UNIT(a), UNIT_AFTER, UNIT_REQUIRES, UNIT(m), true); + if (r < 0) + return r; + + return 0; +} + +static int automount_add_mount_links(Automount *a) { + Unit *other; + int r; + + assert(a); + + LIST_FOREACH(units_by_type, other, UNIT(a)->manager->units_by_type[UNIT_MOUNT]) { + r = automount_add_one_mount_link(a, MOUNT(other)); + if (r < 0) + return r; + } + + return 0; +} + +static int automount_add_default_dependencies(Automount *a) { + int r; + + assert(a); + + if (UNIT(a)->manager->running_as != SYSTEMD_SYSTEM) + return 0; + + r = unit_add_two_dependencies_by_name(UNIT(a), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_UMOUNT_TARGET, NULL, true); + if (r < 0) + return r; + + return 0; +} + +static int automount_verify(Automount *a) { + bool b; + char *e; + assert(a); + + if (UNIT(a)->load_state != UNIT_LOADED) + return 0; + + if (path_equal(a->where, "/")) { + log_error_unit(UNIT(a)->id, "Cannot have an automount unit for the root directory. Refusing."); + return -EINVAL; + } + + e = unit_name_from_path(a->where, ".automount"); + if (!e) + return -ENOMEM; + + b = unit_has_name(UNIT(a), e); + free(e); + + if (!b) { + log_error_unit(UNIT(a)->id, "%s's Where setting doesn't match unit name. Refusing.", UNIT(a)->id); + return -EINVAL; + } + + return 0; +} + +static int automount_load(Unit *u) { + Automount *a = AUTOMOUNT(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + /* Load a .automount file */ + r = unit_load_fragment_and_dropin_optional(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + Unit *x; + + if (!a->where) { + a->where = unit_name_to_path(u->id); + if (!a->where) + return -ENOMEM; + } + + path_kill_slashes(a->where); + + r = unit_load_related_unit(u, ".mount", &x); + if (r < 0) + return r; + + r = unit_add_two_dependencies(u, UNIT_BEFORE, UNIT_TRIGGERS, x, true); + if (r < 0) + return r; + + r = automount_add_mount_links(a); + if (r < 0) + return r; + + if (UNIT(a)->default_dependencies) { + r = automount_add_default_dependencies(a); + if (r < 0) + return r; + } + } + + return automount_verify(a); +} + +static void automount_set_state(Automount *a, AutomountState state) { + AutomountState old_state; + assert(a); + + old_state = a->state; + a->state = state; + + if (state != AUTOMOUNT_WAITING && + state != AUTOMOUNT_RUNNING) + unmount_autofs(a); + + if (state != old_state) + log_debug_unit(UNIT(a)->id, + "%s changed %s -> %s", + UNIT(a)->id, + automount_state_to_string(old_state), + automount_state_to_string(state)); + + unit_notify(UNIT(a), state_translation_table[old_state], state_translation_table[state], true); +} + +static int automount_coldplug(Unit *u) { + Automount *a = AUTOMOUNT(u); + int r; + + assert(a); + assert(a->state == AUTOMOUNT_DEAD); + + if (a->deserialized_state != a->state) { + + r = open_dev_autofs(u->manager); + if (r < 0) + return r; + + if (a->deserialized_state == AUTOMOUNT_WAITING || + a->deserialized_state == AUTOMOUNT_RUNNING) { + + assert(a->pipe_fd >= 0); + + r = unit_watch_fd(UNIT(a), a->pipe_fd, EPOLLIN, &a->pipe_watch); + if (r < 0) + return r; + } + + automount_set_state(a, a->deserialized_state); + } + + return 0; +} + +static void automount_dump(Unit *u, FILE *f, const char *prefix) { + Automount *a = AUTOMOUNT(u); + + assert(a); + + fprintf(f, + "%sAutomount State: %s\n" + "%sResult: %s\n" + "%sWhere: %s\n" + "%sDirectoryMode: %04o\n", + prefix, automount_state_to_string(a->state), + prefix, automount_result_to_string(a->result), + prefix, a->where, + prefix, a->directory_mode); +} + +static void automount_enter_dead(Automount *a, AutomountResult f) { + assert(a); + + if (f != AUTOMOUNT_SUCCESS) + a->result = f; + + automount_set_state(a, a->result != AUTOMOUNT_SUCCESS ? AUTOMOUNT_FAILED : AUTOMOUNT_DEAD); +} + +static int open_dev_autofs(Manager *m) { + struct autofs_dev_ioctl param; + + assert(m); + + if (m->dev_autofs_fd >= 0) + return m->dev_autofs_fd; + + label_fix("/dev/autofs", false, false); + + m->dev_autofs_fd = open("/dev/autofs", O_CLOEXEC|O_RDONLY); + if (m->dev_autofs_fd < 0) { + log_error("Failed to open /dev/autofs: %s", strerror(errno)); + return -errno; + } + + init_autofs_dev_ioctl(¶m); + if (ioctl(m->dev_autofs_fd, AUTOFS_DEV_IOCTL_VERSION, ¶m) < 0) { + close_nointr_nofail(m->dev_autofs_fd); + m->dev_autofs_fd = -1; + return -errno; + } + + log_debug("Autofs kernel version %i.%i", param.ver_major, param.ver_minor); + + return m->dev_autofs_fd; +} + +static int open_ioctl_fd(int dev_autofs_fd, const char *where, dev_t devid) { + struct autofs_dev_ioctl *param; + size_t l; + + assert(dev_autofs_fd >= 0); + assert(where); + + l = sizeof(struct autofs_dev_ioctl) + strlen(where) + 1; + param = alloca(l); + + init_autofs_dev_ioctl(param); + param->size = l; + param->ioctlfd = -1; + param->openmount.devid = devid; + strcpy(param->path, where); + + if (ioctl(dev_autofs_fd, AUTOFS_DEV_IOCTL_OPENMOUNT, param) < 0) + return -errno; + + if (param->ioctlfd < 0) + return -EIO; + + fd_cloexec(param->ioctlfd, true); + return param->ioctlfd; +} + +static int autofs_protocol(int dev_autofs_fd, int ioctl_fd) { + uint32_t major, minor; + struct autofs_dev_ioctl param; + + assert(dev_autofs_fd >= 0); + assert(ioctl_fd >= 0); + + init_autofs_dev_ioctl(¶m); + param.ioctlfd = ioctl_fd; + + if (ioctl(dev_autofs_fd, AUTOFS_DEV_IOCTL_PROTOVER, ¶m) < 0) + return -errno; + + major = param.protover.version; + + init_autofs_dev_ioctl(¶m); + param.ioctlfd = ioctl_fd; + + if (ioctl(dev_autofs_fd, AUTOFS_DEV_IOCTL_PROTOSUBVER, ¶m) < 0) + return -errno; + + minor = param.protosubver.sub_version; + + log_debug("Autofs protocol version %i.%i", major, minor); + return 0; +} + +static int autofs_set_timeout(int dev_autofs_fd, int ioctl_fd, time_t sec) { + struct autofs_dev_ioctl param; + + assert(dev_autofs_fd >= 0); + assert(ioctl_fd >= 0); + + init_autofs_dev_ioctl(¶m); + param.ioctlfd = ioctl_fd; + param.timeout.timeout = sec; + + if (ioctl(dev_autofs_fd, AUTOFS_DEV_IOCTL_TIMEOUT, ¶m) < 0) + return -errno; + + return 0; +} + +static int autofs_send_ready(int dev_autofs_fd, int ioctl_fd, uint32_t token, int status) { + struct autofs_dev_ioctl param; + + assert(dev_autofs_fd >= 0); + assert(ioctl_fd >= 0); + + init_autofs_dev_ioctl(¶m); + param.ioctlfd = ioctl_fd; + + if (status) { + param.fail.token = token; + param.fail.status = status; + } else + param.ready.token = token; + + if (ioctl(dev_autofs_fd, status ? AUTOFS_DEV_IOCTL_FAIL : AUTOFS_DEV_IOCTL_READY, ¶m) < 0) + return -errno; + + return 0; +} + +int automount_send_ready(Automount *a, int status) { + int ioctl_fd, r; + unsigned token; + + assert(a); + assert(status <= 0); + + if (set_isempty(a->tokens)) + return 0; + + ioctl_fd = open_ioctl_fd(UNIT(a)->manager->dev_autofs_fd, a->where, a->dev_id); + if (ioctl_fd < 0) { + r = ioctl_fd; + goto fail; + } + + if (status) + log_debug_unit(UNIT(a)->id, "Sending failure: %s", strerror(-status)); + else + log_debug_unit(UNIT(a)->id, "Sending success."); + + r = 0; + + /* Autofs thankfully does not hand out 0 as a token */ + while ((token = PTR_TO_UINT(set_steal_first(a->tokens)))) { + int k; + + /* Autofs fun fact II: + * + * if you pass a positive status code here, the kernel will + * freeze! Yay! */ + + k = autofs_send_ready(UNIT(a)->manager->dev_autofs_fd, + ioctl_fd, + token, + status); + if (k < 0) + r = k; + } + +fail: + if (ioctl_fd >= 0) + close_nointr_nofail(ioctl_fd); + + return r; +} + +static void automount_enter_waiting(Automount *a) { + int p[2] = { -1, -1 }; + char name[32], options[128]; + bool mounted = false; + int r, ioctl_fd = -1, dev_autofs_fd; + struct stat st; + + assert(a); + assert(a->pipe_fd < 0); + assert(a->where); + + if (a->tokens) + set_clear(a->tokens); + + dev_autofs_fd = open_dev_autofs(UNIT(a)->manager); + if (dev_autofs_fd < 0) { + r = dev_autofs_fd; + goto fail; + } + + /* We knowingly ignore the results of this call */ + mkdir_p_label(a->where, 0555); + + warn_if_dir_nonempty(a->meta.id, a->where); + + if (pipe2(p, O_NONBLOCK|O_CLOEXEC) < 0) { + r = -errno; + goto fail; + } + + snprintf(options, sizeof(options), "fd=%i,pgrp=%u,minproto=5,maxproto=5,direct", p[1], (unsigned) getpgrp()); + char_array_0(options); + + snprintf(name, sizeof(name), "systemd-%u", (unsigned) getpid()); + char_array_0(name); + + if (mount(name, a->where, "autofs", 0, options) < 0) { + r = -errno; + goto fail; + } + + mounted = true; + + close_nointr_nofail(p[1]); + p[1] = -1; + + if (stat(a->where, &st) < 0) { + r = -errno; + goto fail; + } + + ioctl_fd = open_ioctl_fd(dev_autofs_fd, a->where, st.st_dev); + if (ioctl_fd < 0) { + r = ioctl_fd; + goto fail; + } + + r = autofs_protocol(dev_autofs_fd, ioctl_fd); + if (r < 0) + goto fail; + + r = autofs_set_timeout(dev_autofs_fd, ioctl_fd, 300); + if (r < 0) + goto fail; + + /* Autofs fun fact: + * + * Unless we close the ioctl fd here, for some weird reason + * the direct mount will not receive events from the + * kernel. */ + + close_nointr_nofail(ioctl_fd); + ioctl_fd = -1; + + r = unit_watch_fd(UNIT(a), p[0], EPOLLIN, &a->pipe_watch); + if (r < 0) + goto fail; + + a->pipe_fd = p[0]; + a->dev_id = st.st_dev; + + automount_set_state(a, AUTOMOUNT_WAITING); + + return; + +fail: + assert_se(close_pipe(p) == 0); + + if (ioctl_fd >= 0) + close_nointr_nofail(ioctl_fd); + + if (mounted) + repeat_unmout(a->where); + + log_error_unit(UNIT(a)->id, + "Failed to initialize automounter: %s", strerror(-r)); + automount_enter_dead(a, AUTOMOUNT_FAILURE_RESOURCES); +} + +static void automount_enter_runnning(Automount *a) { + _cleanup_dbus_error_free_ DBusError error; + struct stat st; + int r; + + assert(a); + + dbus_error_init(&error); + + /* We don't take mount requests anymore if we are supposed to + * shut down anyway */ + if (unit_stop_pending(UNIT(a))) { + log_debug_unit(UNIT(a)->id, + "Suppressing automount request on %s since unit stop is scheduled.", UNIT(a)->id); + automount_send_ready(a, -EHOSTDOWN); + return; + } + + mkdir_p_label(a->where, a->directory_mode); + + /* Before we do anything, let's see if somebody is playing games with us? */ + if (lstat(a->where, &st) < 0) { + log_warning_unit(UNIT(a)->id, + "%s failed to stat automount point: %m", UNIT(a)->id); + goto fail; + } + + if (!S_ISDIR(st.st_mode) || st.st_dev != a->dev_id) + log_info_unit(UNIT(a)->id, + "%s's automount point already active?", UNIT(a)->id); + else { + r = manager_add_job(UNIT(a)->manager, JOB_START, UNIT_TRIGGER(UNIT(a)), + JOB_REPLACE, true, &error, NULL); + if (r < 0) { + log_warning_unit(UNIT(a)->id, + "%s failed to queue mount startup job: %s", + UNIT(a)->id, bus_error(&error, r)); + goto fail; + } + } + + automount_set_state(a, AUTOMOUNT_RUNNING); + return; + +fail: + automount_enter_dead(a, AUTOMOUNT_FAILURE_RESOURCES); +} + +static int automount_start(Unit *u) { + Automount *a = AUTOMOUNT(u); + + assert(a); + assert(a->state == AUTOMOUNT_DEAD || a->state == AUTOMOUNT_FAILED); + + if (path_is_mount_point(a->where, false)) { + log_error_unit(u->id, + "Path %s is already a mount point, refusing start for %s", + a->where, u->id); + return -EEXIST; + } + + if (UNIT_TRIGGER(u)->load_state != UNIT_LOADED) + return -ENOENT; + + a->result = AUTOMOUNT_SUCCESS; + automount_enter_waiting(a); + return 0; +} + +static int automount_stop(Unit *u) { + Automount *a = AUTOMOUNT(u); + + assert(a); + assert(a->state == AUTOMOUNT_WAITING || a->state == AUTOMOUNT_RUNNING); + + automount_enter_dead(a, AUTOMOUNT_SUCCESS); + return 0; +} + +static int automount_serialize(Unit *u, FILE *f, FDSet *fds) { + Automount *a = AUTOMOUNT(u); + void *p; + Iterator i; + + assert(a); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", automount_state_to_string(a->state)); + unit_serialize_item(u, f, "result", automount_result_to_string(a->result)); + unit_serialize_item_format(u, f, "dev-id", "%u", (unsigned) a->dev_id); + + SET_FOREACH(p, a->tokens, i) + unit_serialize_item_format(u, f, "token", "%u", PTR_TO_UINT(p)); + + if (a->pipe_fd >= 0) { + int copy; + + copy = fdset_put_dup(fds, a->pipe_fd); + if (copy < 0) + return copy; + + unit_serialize_item_format(u, f, "pipe-fd", "%i", copy); + } + + return 0; +} + +static int automount_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Automount *a = AUTOMOUNT(u); + int r; + + assert(a); + assert(fds); + + if (streq(key, "state")) { + AutomountState state; + + state = automount_state_from_string(value); + if (state < 0) + log_debug_unit(u->id, "Failed to parse state value %s", value); + else + a->deserialized_state = state; + } else if (streq(key, "result")) { + AutomountResult f; + + f = automount_result_from_string(value); + if (f < 0) + log_debug_unit(u->id, "Failed to parse result value %s", value); + else if (f != AUTOMOUNT_SUCCESS) + a->result = f; + + } else if (streq(key, "dev-id")) { + unsigned d; + + if (safe_atou(value, &d) < 0) + log_debug_unit(u->id, "Failed to parse dev-id value %s", value); + else + a->dev_id = (unsigned) d; + } else if (streq(key, "token")) { + unsigned token; + + if (safe_atou(value, &token) < 0) + log_debug_unit(u->id, "Failed to parse token value %s", value); + else { + if (!a->tokens) + if (!(a->tokens = set_new(trivial_hash_func, trivial_compare_func))) + return -ENOMEM; + + r = set_put(a->tokens, UINT_TO_PTR(token)); + if (r < 0) + return r; + } + } else if (streq(key, "pipe-fd")) { + int fd; + + if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_debug_unit(u->id, "Failed to parse pipe-fd value %s", value); + else { + if (a->pipe_fd >= 0) + close_nointr_nofail(a->pipe_fd); + + a->pipe_fd = fdset_remove(fds, fd); + } + } else + log_debug_unit(u->id, "Unknown serialization key '%s'", key); + + return 0; +} + +static UnitActiveState automount_active_state(Unit *u) { + assert(u); + + return state_translation_table[AUTOMOUNT(u)->state]; +} + +static const char *automount_sub_state_to_string(Unit *u) { + assert(u); + + return automount_state_to_string(AUTOMOUNT(u)->state); +} + +static bool automount_check_gc(Unit *u) { + assert(u); + + if (!UNIT_TRIGGER(u)) + return false; + + return UNIT_VTABLE(UNIT_TRIGGER(u))->check_gc(UNIT_TRIGGER(u)); +} + +static void automount_fd_event(Unit *u, int fd, uint32_t events, Watch *w) { + Automount *a = AUTOMOUNT(u); + union autofs_v5_packet_union packet; + ssize_t l; + int r; + + assert(a); + assert(fd == a->pipe_fd); + + if (events != EPOLLIN) { + log_error_unit(u->id, "Got invalid poll event on pipe."); + goto fail; + } + + l = loop_read(a->pipe_fd, &packet, sizeof(packet), true); + if (l != sizeof(packet)) { + log_error_unit(u->id, "Invalid read from pipe: %s", l < 0 ? strerror(-l) : "short read"); + goto fail; + } + + switch (packet.hdr.type) { + + case autofs_ptype_missing_direct: + + if (packet.v5_packet.pid > 0) { + _cleanup_free_ char *p = NULL; + + get_process_comm(packet.v5_packet.pid, &p); + log_debug_unit(u->id, + "Got direct mount request on %s, triggered by %lu (%s)", + a->where, (unsigned long) packet.v5_packet.pid, strna(p)); + } else + log_debug_unit(u->id, "Got direct mount request on %s", a->where); + + r = set_ensure_allocated(&a->tokens, trivial_hash_func, trivial_compare_func); + if (r < 0) { + log_error_unit(u->id, "Failed to allocate token set."); + goto fail; + } + + r = set_put(a->tokens, UINT_TO_PTR(packet.v5_packet.wait_queue_token)); + if (r < 0) { + log_error_unit(u->id, "Failed to remember token: %s", strerror(-r)); + goto fail; + } + + automount_enter_runnning(a); + break; + + default: + log_error_unit(u->id, "Received unknown automount request %i", packet.hdr.type); + break; + } + + return; + +fail: + automount_enter_dead(a, AUTOMOUNT_FAILURE_RESOURCES); +} + +static void automount_shutdown(Manager *m) { + assert(m); + + if (m->dev_autofs_fd >= 0) + close_nointr_nofail(m->dev_autofs_fd); +} + +static void automount_reset_failed(Unit *u) { + Automount *a = AUTOMOUNT(u); + + assert(a); + + if (a->state == AUTOMOUNT_FAILED) + automount_set_state(a, AUTOMOUNT_DEAD); + + a->result = AUTOMOUNT_SUCCESS; +} + +static const char* const automount_state_table[_AUTOMOUNT_STATE_MAX] = { + [AUTOMOUNT_DEAD] = "dead", + [AUTOMOUNT_WAITING] = "waiting", + [AUTOMOUNT_RUNNING] = "running", + [AUTOMOUNT_FAILED] = "failed" +}; + +DEFINE_STRING_TABLE_LOOKUP(automount_state, AutomountState); + +static const char* const automount_result_table[_AUTOMOUNT_RESULT_MAX] = { + [AUTOMOUNT_SUCCESS] = "success", + [AUTOMOUNT_FAILURE_RESOURCES] = "resources" +}; + +DEFINE_STRING_TABLE_LOOKUP(automount_result, AutomountResult); + +const UnitVTable automount_vtable = { + .object_size = sizeof(Automount), + .sections = + "Unit\0" + "Automount\0" + "Install\0", + + .no_alias = true, + .no_instances = true, + + .init = automount_init, + .load = automount_load, + .done = automount_done, + + .coldplug = automount_coldplug, + + .dump = automount_dump, + + .start = automount_start, + .stop = automount_stop, + + .serialize = automount_serialize, + .deserialize_item = automount_deserialize_item, + + .active_state = automount_active_state, + .sub_state_to_string = automount_sub_state_to_string, + + .check_gc = automount_check_gc, + + .fd_event = automount_fd_event, + + .reset_failed = automount_reset_failed, + + .bus_interface = "org.freedesktop.systemd1.Automount", + .bus_message_handler = bus_automount_message_handler, + .bus_invalidating_properties = bus_automount_invalidating_properties, + + .shutdown = automount_shutdown, + + .status_message_formats = { + .finished_start_job = { + [JOB_DONE] = "Set up automount %s.", + [JOB_FAILED] = "Failed to set up automount %s.", + [JOB_DEPENDENCY] = "Dependency failed for %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Unset automount %s.", + [JOB_FAILED] = "Failed to unset automount %s.", + }, + }, +}; diff --git a/src/core/automount.h b/src/core/automount.h new file mode 100644 index 0000000000..0c6b8a72e9 --- /dev/null +++ b/src/core/automount.h @@ -0,0 +1,71 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Automount Automount; + +#include "unit.h" + +typedef enum AutomountState { + AUTOMOUNT_DEAD, + AUTOMOUNT_WAITING, + AUTOMOUNT_RUNNING, + AUTOMOUNT_FAILED, + _AUTOMOUNT_STATE_MAX, + _AUTOMOUNT_STATE_INVALID = -1 +} AutomountState; + +typedef enum AutomountResult { + AUTOMOUNT_SUCCESS, + AUTOMOUNT_FAILURE_RESOURCES, + _AUTOMOUNT_RESULT_MAX, + _AUTOMOUNT_RESULT_INVALID = -1 +} AutomountResult; + +struct Automount { + Unit meta; + + AutomountState state, deserialized_state; + + char *where; + + int pipe_fd; + mode_t directory_mode; + Watch pipe_watch; + dev_t dev_id; + + Set *tokens; + + AutomountResult result; +}; + +extern const UnitVTable automount_vtable; + +int automount_send_ready(Automount *a, int status); + +int automount_add_one_mount_link(Automount *a, Mount *m); + +const char* automount_state_to_string(AutomountState i) _const_; +AutomountState automount_state_from_string(const char *s) _pure_; + +const char* automount_result_to_string(AutomountResult i) _const_; +AutomountResult automount_result_from_string(const char *s) _pure_; diff --git a/src/core/build.h b/src/core/build.h new file mode 100644 index 0000000000..4513a0bad7 --- /dev/null +++ b/src/core/build.h @@ -0,0 +1,84 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#ifdef HAVE_PAM +#define _PAM_FEATURE_ "+PAM" +#else +#define _PAM_FEATURE_ "-PAM" +#endif + +#ifdef HAVE_LIBWRAP +#define _LIBWRAP_FEATURE_ "+LIBWRAP" +#else +#define _LIBWRAP_FEATURE_ "-LIBWRAP" +#endif + +#ifdef HAVE_AUDIT +#define _AUDIT_FEATURE_ "+AUDIT" +#else +#define _AUDIT_FEATURE_ "-AUDIT" +#endif + +#ifdef HAVE_SELINUX +#define _SELINUX_FEATURE_ "+SELINUX" +#else +#define _SELINUX_FEATURE_ "-SELINUX" +#endif + +#ifdef HAVE_IMA +#define _IMA_FEATURE_ "+IMA" +#else +#define _IMA_FEATURE_ "-IMA" +#endif + +#ifdef HAVE_SYSV_COMPAT +#define _SYSVINIT_FEATURE_ "+SYSVINIT" +#else +#define _SYSVINIT_FEATURE_ "-SYSVINIT" +#endif + +#ifdef HAVE_LIBCRYPTSETUP +#define _LIBCRYPTSETUP_FEATURE_ "+LIBCRYPTSETUP" +#else +#define _LIBCRYPTSETUP_FEATURE_ "-LIBCRYPTSETUP" +#endif + +#ifdef HAVE_GCRYPT +#define _GCRYPT_FEATURE_ "+GCRYPT" +#else +#define _GCRYPT_FEATURE_ "-GCRYPT" +#endif + +#ifdef HAVE_ACL +#define _ACL_FEATURE_ "+ACL" +#else +#define _ACL_FEATURE_ "-ACL" +#endif + +#ifdef HAVE_XZ +#define _XZ_FEATURE_ "+XZ" +#else +#define _XZ_FEATURE_ "-XZ" +#endif + +#define SYSTEMD_FEATURES _PAM_FEATURE_ " " _LIBWRAP_FEATURE_ " " _AUDIT_FEATURE_ " " _SELINUX_FEATURE_ " " _IMA_FEATURE_ " " _SYSVINIT_FEATURE_ " " _LIBCRYPTSETUP_FEATURE_ " " _GCRYPT_FEATURE_ " " _ACL_FEATURE_ " " _XZ_FEATURE_ diff --git a/src/core/bus-errors.h b/src/core/bus-errors.h new file mode 100644 index 0000000000..7a4084ea15 --- /dev/null +++ b/src/core/bus-errors.h @@ -0,0 +1,42 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#define BUS_ERROR_NO_SUCH_UNIT "org.freedesktop.systemd1.NoSuchUnit" +#define BUS_ERROR_NO_SUCH_JOB "org.freedesktop.systemd1.NoSuchJob" +#define BUS_ERROR_NOT_SUBSCRIBED "org.freedesktop.systemd1.NotSubscribed" +#define BUS_ERROR_INVALID_PATH "org.freedesktop.systemd1.InvalidPath" +#define BUS_ERROR_INVALID_NAME "org.freedesktop.systemd1.InvalidName" +#define BUS_ERROR_UNIT_TYPE_MISMATCH "org.freedesktop.systemd1.UnitTypeMismatch" +#define BUS_ERROR_UNIT_EXISTS "org.freedesktop.systemd1.UnitExists" +#define BUS_ERROR_NOT_SUPPORTED "org.freedesktop.systemd1.NotSupported" +#define BUS_ERROR_INVALID_JOB_MODE "org.freedesktop.systemd1.InvalidJobMode" +#define BUS_ERROR_ONLY_BY_DEPENDENCY "org.freedesktop.systemd1.OnlyByDependency" +#define BUS_ERROR_NO_ISOLATION "org.freedesktop.systemd1.NoIsolation" +#define BUS_ERROR_LOAD_FAILED "org.freedesktop.systemd1.LoadFailed" +#define BUS_ERROR_MASKED "org.freedesktop.systemd1.Masked" +#define BUS_ERROR_JOB_TYPE_NOT_APPLICABLE "org.freedesktop.systemd1.JobTypeNotApplicable" +#define BUS_ERROR_TRANSACTION_IS_DESTRUCTIVE "org.freedesktop.systemd1.TransactionIsDestructive" +#define BUS_ERROR_TRANSACTION_JOBS_CONFLICTING "org.freedesktop.systemd1.TransactionJobsConflicting" +#define BUS_ERROR_TRANSACTION_ORDER_IS_CYCLIC "org.freedesktop.systemd1.TransactionOrderIsCyclic" +#define BUS_ERROR_SHUTTING_DOWN "org.freedesktop.systemd1.ShuttingDown" +#define BUS_ERROR_NO_SUCH_PROCESS "org.freedesktop.systemd1.NoSuchProcess" diff --git a/src/core/cgroup-attr.c b/src/core/cgroup-attr.c new file mode 100644 index 0000000000..7e3e08eabb --- /dev/null +++ b/src/core/cgroup-attr.c @@ -0,0 +1,132 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "cgroup-attr.h" +#include "cgroup-util.h" +#include "list.h" +#include "fileio.h" + +int cgroup_attribute_apply(CGroupAttribute *a, CGroupBonding *b) { + _cleanup_free_ char *path = NULL, *v = NULL; + int r; + + assert(a); + + b = cgroup_bonding_find_list(b, a->controller); + if (!b) + return 0; + + if (a->semantics && a->semantics->map_write) { + r = a->semantics->map_write(a->semantics, a->value, &v); + if (r < 0) + return r; + } + + r = cg_get_path(a->controller, b->path, a->name, &path); + if (r < 0) + return r; + + r = write_string_file(path, v ? v : a->value); + if (r < 0) + log_warning("Failed to write '%s' to %s: %s", v ? v : a->value, path, strerror(-r)); + + return r; +} + +int cgroup_attribute_apply_list(CGroupAttribute *first, CGroupBonding *b) { + CGroupAttribute *a; + int r = 0; + + LIST_FOREACH(by_unit, a, first) { + int k; + + k = cgroup_attribute_apply(a, b); + if (r == 0) + r = k; + } + + return r; +} + +bool cgroup_attribute_matches(CGroupAttribute *a, const char *controller, const char *name) { + assert(a); + + if (controller) { + if (streq(a->controller, controller) && (!name || streq(a->name, name))) + return true; + + } else if (!name) + return true; + else if (streq(a->name, name)) { + size_t x, y; + x = strlen(a->controller); + y = strlen(name); + + if (y > x && + memcmp(a->controller, name, x) == 0 && + name[x] == '.') + return true; + } + + return false; +} + +CGroupAttribute *cgroup_attribute_find_list( + CGroupAttribute *first, + const char *controller, + const char *name) { + CGroupAttribute *a; + + assert(name); + + LIST_FOREACH(by_unit, a, first) + if (cgroup_attribute_matches(a, controller, name)) + return a; + + return NULL; +} + +void cgroup_attribute_free(CGroupAttribute *a) { + assert(a); + + if (a->unit) + LIST_REMOVE(CGroupAttribute, by_unit, a->unit->cgroup_attributes, a); + + free(a->controller); + free(a->name); + free(a->value); + free(a); +} + +void cgroup_attribute_free_list(CGroupAttribute *first) { + CGroupAttribute *a, *n; + + LIST_FOREACH_SAFE(by_unit, a, n, first) + cgroup_attribute_free(a); +} + +void cgroup_attribute_free_some(CGroupAttribute *first, const char *controller, const char *name) { + CGroupAttribute *a, *n; + + LIST_FOREACH_SAFE(by_unit, a, n, first) + if (cgroup_attribute_matches(a, controller, name)) + cgroup_attribute_free(a); +} diff --git a/src/core/cgroup-attr.h b/src/core/cgroup-attr.h new file mode 100644 index 0000000000..3a13b7c92d --- /dev/null +++ b/src/core/cgroup-attr.h @@ -0,0 +1,50 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct CGroupAttribute CGroupAttribute; + +#include "unit.h" +#include "cgroup.h" +#include "cgroup-semantics.h" + +struct CGroupAttribute { + char *controller; + char *name; + char *value; + + Unit *unit; + + const CGroupSemantics *semantics; + + LIST_FIELDS(CGroupAttribute, by_unit); +}; + +int cgroup_attribute_apply(CGroupAttribute *a, CGroupBonding *b); +int cgroup_attribute_apply_list(CGroupAttribute *first, CGroupBonding *b); + +bool cgroup_attribute_matches(CGroupAttribute *a, const char *controller, const char *name) _pure_; +CGroupAttribute *cgroup_attribute_find_list(CGroupAttribute *first, const char *controller, const char *name) _pure_; + +void cgroup_attribute_free(CGroupAttribute *a); +void cgroup_attribute_free_list(CGroupAttribute *first); +void cgroup_attribute_free_some(CGroupAttribute *first, const char *controller, const char *name); diff --git a/src/core/cgroup-semantics.c b/src/core/cgroup-semantics.c new file mode 100644 index 0000000000..82b02bbd78 --- /dev/null +++ b/src/core/cgroup-semantics.c @@ -0,0 +1,333 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "util.h" +#include "strv.h" +#include "path-util.h" +#include "cgroup-util.h" + +#include "cgroup-semantics.h" + +static int parse_cpu_shares(const CGroupSemantics *s, const char *value, char **ret) { + unsigned long ul; + + assert(s); + assert(value); + assert(ret); + + if (safe_atolu(value, &ul) < 0 || ul < 1) + return -EINVAL; + + if (asprintf(ret, "%lu", ul) < 0) + return -ENOMEM; + + return 1; +} + +static int parse_memory_limit(const CGroupSemantics *s, const char *value, char **ret) { + off_t sz; + + assert(s); + assert(value); + assert(ret); + + if (parse_bytes(value, &sz) < 0 || sz <= 0) + return -EINVAL; + + if (asprintf(ret, "%llu", (unsigned long long) sz) < 0) + return -ENOMEM; + + return 1; +} + +static int parse_device(const CGroupSemantics *s, const char *value, char **ret) { + _cleanup_strv_free_ char **l = NULL; + char *x; + unsigned k; + + assert(s); + assert(value); + assert(ret); + + l = strv_split_quoted(value); + if (!l) + return -ENOMEM; + + k = strv_length(l); + if (k < 1 || k > 2) + return -EINVAL; + + if (!streq(l[0], "*") && !path_startswith(l[0], "/dev")) + return -EINVAL; + + if (!isempty(l[1]) && !in_charset(l[1], "rwm")) + return -EINVAL; + + x = strdup(value); + if (!x) + return -ENOMEM; + + *ret = x; + return 1; +} + +static int parse_blkio_weight(const CGroupSemantics *s, const char *value, char **ret) { + _cleanup_strv_free_ char **l = NULL; + unsigned long ul; + + assert(s); + assert(value); + assert(ret); + + l = strv_split_quoted(value); + if (!l) + return -ENOMEM; + + if (strv_length(l) != 1) + return 0; /* Returning 0 will cause parse_blkio_weight_device() be tried instead */ + + if (safe_atolu(l[0], &ul) < 0 || ul < 10 || ul > 1000) + return -EINVAL; + + if (asprintf(ret, "%lu", ul) < 0) + return -ENOMEM; + + return 1; +} + +static int parse_blkio_weight_device(const CGroupSemantics *s, const char *value, char **ret) { + _cleanup_strv_free_ char **l = NULL; + unsigned long ul; + + assert(s); + assert(value); + assert(ret); + + l = strv_split_quoted(value); + if (!l) + return -ENOMEM; + + if (strv_length(l) != 2) + return -EINVAL; + + if (!path_startswith(l[0], "/dev")) + return -EINVAL; + + if (safe_atolu(l[1], &ul) < 0 || ul < 10 || ul > 1000) + return -EINVAL; + + if (asprintf(ret, "%s %lu", l[0], ul) < 0) + return -ENOMEM; + + return 1; +} + +static int parse_blkio_bandwidth(const CGroupSemantics *s, const char *value, char **ret) { + _cleanup_strv_free_ char **l = NULL; + off_t bytes; + + assert(s); + assert(value); + assert(ret); + + l = strv_split_quoted(value); + if (!l) + return -ENOMEM; + + if (strv_length(l) != 2) + return -EINVAL; + + if (!path_startswith(l[0], "/dev")) { + return -EINVAL; + } + + if (parse_bytes(l[1], &bytes) < 0 || bytes <= 0) + return -EINVAL; + + if (asprintf(ret, "%s %llu", l[0], (unsigned long long) bytes) < 0) + return -ENOMEM; + + return 0; +} + +static int map_device(const CGroupSemantics *s, const char *value, char **ret) { + _cleanup_strv_free_ char **l = NULL; + unsigned k; + + assert(s); + assert(value); + assert(ret); + + l = strv_split_quoted(value); + if (!l) + return -ENOMEM; + + k = strv_length(l); + if (k < 1 || k > 2) + return -EINVAL; + + if (streq(l[0], "*")) { + + if (asprintf(ret, "a *:*%s%s", + isempty(l[1]) ? "" : " ", strempty(l[1])) < 0) + return -ENOMEM; + } else { + struct stat st; + + if (stat(l[0], &st) < 0) { + log_warning("Couldn't stat device %s", l[0]); + return -errno; + } + + if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) { + log_warning("%s is not a device.", l[0]); + return -ENODEV; + } + + if (asprintf(ret, "%c %u:%u%s%s", + S_ISCHR(st.st_mode) ? 'c' : 'b', + major(st.st_rdev), minor(st.st_rdev), + isempty(l[1]) ? "" : " ", strempty(l[1])) < 0) + return -ENOMEM; + } + + return 0; +} + +static int map_blkio(const CGroupSemantics *s, const char *value, char **ret) { + _cleanup_strv_free_ char **l = NULL; + struct stat st; + dev_t d; + + assert(s); + assert(value); + assert(ret); + + l = strv_split_quoted(value); + if (!l) + return log_oom(); + + if (strv_length(l) != 2) + return -EINVAL; + + if (stat(l[0], &st) < 0) { + log_warning("Couldn't stat device %s", l[0]); + return -errno; + } + + if (S_ISBLK(st.st_mode)) + d = st.st_rdev; + else if (major(st.st_dev) != 0) { + /* If this is not a device node then find the block + * device this file is stored on */ + d = st.st_dev; + + /* If this is a partition, try to get the originating + * block device */ + block_get_whole_disk(d, &d); + } else { + log_warning("%s is not a block device and file system block device cannot be determined or is not local.", l[0]); + return -ENODEV; + } + + if (asprintf(ret, "%u:%u %s", major(d), minor(d), l[1]) < 0) + return -ENOMEM; + + return 0; +} + +static const CGroupSemantics semantics[] = { + { "cpu", "cpu.shares", "CPUShare", false, parse_cpu_shares, NULL, NULL }, + { "memory", "memory.soft_limit_in_bytes", "MemorySoftLimit", false, parse_memory_limit, NULL, NULL }, + { "memory", "memory.limit_in_bytes", "MemoryLimit", false, parse_memory_limit, NULL, NULL }, + { "devices", "devices.allow", "DeviceAllow", true, parse_device, map_device, NULL }, + { "devices", "devices.deny", "DeviceDeny", true, parse_device, map_device, NULL }, + { "blkio", "blkio.weight", "BlockIOWeight", false, parse_blkio_weight, NULL, NULL }, + { "blkio", "blkio.weight_device", "BlockIOWeight", true, parse_blkio_weight_device, map_blkio, NULL }, + { "blkio", "blkio.read_bps_device", "BlockIOReadBandwidth", true, parse_blkio_bandwidth, map_blkio, NULL }, + { "blkio", "blkio.write_bps_device", "BlockIOWriteBandwidth", true, parse_blkio_bandwidth, map_blkio, NULL } +}; + +int cgroup_semantics_find( + const char *controller, + const char *name, + const char *value, + char **ret, + const CGroupSemantics **_s) { + + _cleanup_free_ char *c = NULL; + unsigned i; + int r; + + assert(name); + assert(_s); + assert(!value == !ret); + + if (!controller) { + r = cg_controller_from_attr(name, &c); + if (r < 0) + return r; + + controller = c; + } + + for (i = 0; i < ELEMENTSOF(semantics); i++) { + const CGroupSemantics *s = semantics + i; + bool matches_name, matches_pretty; + + if (controller && s->controller && !streq(s->controller, controller)) + continue; + + matches_name = s->name && streq(s->name, name); + matches_pretty = s->pretty && streq(s->pretty, name); + + if (!matches_name && !matches_pretty) + continue; + + if (value) { + if (matches_pretty && s->map_pretty) { + + r = s->map_pretty(s, value, ret); + if (r < 0) + return r; + + if (r == 0) + continue; + + } else { + char *x; + + x = strdup(value); + if (!x) + return -ENOMEM; + + *ret = x; + } + } + + *_s = s; + return 1; + } + + *ret = NULL; + *_s = NULL; + return 0; +} diff --git a/src/core/cgroup-semantics.h b/src/core/cgroup-semantics.h new file mode 100644 index 0000000000..4f848f4bb7 --- /dev/null +++ b/src/core/cgroup-semantics.h @@ -0,0 +1,43 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct CGroupSemantics CGroupSemantics; + +struct CGroupSemantics { + const char *controller; + const char *name; + const char *pretty; + + bool multiple; + + /* This call is used for parsing the pretty value to the actual attribute value */ + int (*map_pretty)(const CGroupSemantics *semantics, const char *value, char **ret); + + /* Right before writing this attribute the attribute value is converted to a low-level value */ + int (*map_write)(const CGroupSemantics *semantics, const char *value, char **ret); + + /* If this attribute takes a list, this call can be used to reset the list to empty */ + int (*reset)(const CGroupSemantics *semantics, const char *group); +}; + +int cgroup_semantics_find(const char *controller, const char *name, const char *value, char **ret, const CGroupSemantics **semantics); diff --git a/src/core/cgroup.c b/src/core/cgroup.c new file mode 100644 index 0000000000..83df0f3c9a --- /dev/null +++ b/src/core/cgroup.c @@ -0,0 +1,628 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <assert.h> +#include <unistd.h> +#include <sys/types.h> +#include <signal.h> +#include <sys/mount.h> +#include <fcntl.h> + +#include "cgroup.h" +#include "cgroup-util.h" +#include "log.h" +#include "strv.h" +#include "path-util.h" + +int cgroup_bonding_realize(CGroupBonding *b) { + int r; + + assert(b); + assert(b->path); + assert(b->controller); + + r = cg_create(b->controller, b->path, NULL); + if (r < 0) { + log_warning("Failed to create cgroup %s:%s: %s", b->controller, b->path, strerror(-r)); + return r; + } + + b->realized = true; + + return 0; +} + +int cgroup_bonding_realize_list(CGroupBonding *first) { + CGroupBonding *b; + int r; + + LIST_FOREACH(by_unit, b, first) + if ((r = cgroup_bonding_realize(b)) < 0 && b->essential) + return r; + + return 0; +} + +void cgroup_bonding_free(CGroupBonding *b, bool trim) { + assert(b); + + if (b->unit) { + CGroupBonding *f; + + LIST_REMOVE(CGroupBonding, by_unit, b->unit->cgroup_bondings, b); + + if (streq(b->controller, SYSTEMD_CGROUP_CONTROLLER)) { + assert_se(f = hashmap_get(b->unit->manager->cgroup_bondings, b->path)); + LIST_REMOVE(CGroupBonding, by_path, f, b); + + if (f) + hashmap_replace(b->unit->manager->cgroup_bondings, b->path, f); + else + hashmap_remove(b->unit->manager->cgroup_bondings, b->path); + } + } + + if (b->realized && b->ours && trim) + cg_trim(b->controller, b->path, false); + + free(b->controller); + free(b->path); + free(b); +} + +void cgroup_bonding_free_list(CGroupBonding *first, bool remove_or_trim) { + CGroupBonding *b, *n; + + LIST_FOREACH_SAFE(by_unit, b, n, first) + cgroup_bonding_free(b, remove_or_trim); +} + +void cgroup_bonding_trim(CGroupBonding *b, bool delete_root) { + assert(b); + + if (b->realized && b->ours) + cg_trim(b->controller, b->path, delete_root); +} + +void cgroup_bonding_trim_list(CGroupBonding *first, bool delete_root) { + CGroupBonding *b; + + LIST_FOREACH(by_unit, b, first) + cgroup_bonding_trim(b, delete_root); +} + +int cgroup_bonding_install(CGroupBonding *b, pid_t pid, const char *cgroup_suffix) { + _cleanup_free_ char *p = NULL; + const char *path; + int r; + + assert(b); + assert(pid >= 0); + + if (cgroup_suffix) { + p = strjoin(b->path, "/", cgroup_suffix, NULL); + if (!p) + return -ENOMEM; + + path = p; + } else + path = b->path; + + r = cg_create_and_attach(b->controller, path, pid); + if (r < 0) + return r; + + b->realized = true; + return 0; +} + +int cgroup_bonding_install_list(CGroupBonding *first, pid_t pid, const char *cgroup_suffix) { + CGroupBonding *b; + int r; + + LIST_FOREACH(by_unit, b, first) { + r = cgroup_bonding_install(b, pid, cgroup_suffix); + if (r < 0 && b->essential) + return r; + } + + return 0; +} + +int cgroup_bonding_migrate(CGroupBonding *b, CGroupBonding *list) { + CGroupBonding *q; + int ret = 0; + + LIST_FOREACH(by_unit, q, list) { + int r; + + if (q == b) + continue; + + if (!q->ours) + continue; + + r = cg_migrate_recursive(q->controller, q->path, b->controller, b->path, true, false); + if (r < 0 && ret == 0) + ret = r; + } + + return ret; +} + +int cgroup_bonding_migrate_to(CGroupBonding *b, const char *target, bool rem) { + assert(b); + assert(target); + + return cg_migrate_recursive(b->controller, b->path, b->controller, target, true, rem); +} + +int cgroup_bonding_set_group_access(CGroupBonding *b, mode_t mode, uid_t uid, gid_t gid) { + assert(b); + + if (!b->realized) + return -EINVAL; + + return cg_set_group_access(b->controller, b->path, mode, uid, gid); +} + +int cgroup_bonding_set_group_access_list(CGroupBonding *first, mode_t mode, uid_t uid, gid_t gid) { + CGroupBonding *b; + int r; + + LIST_FOREACH(by_unit, b, first) { + r = cgroup_bonding_set_group_access(b, mode, uid, gid); + if (r < 0) + return r; + } + + return 0; +} + +int cgroup_bonding_set_task_access(CGroupBonding *b, mode_t mode, uid_t uid, gid_t gid, int sticky) { + assert(b); + + if (!b->realized) + return -EINVAL; + + return cg_set_task_access(b->controller, b->path, mode, uid, gid, sticky); +} + +int cgroup_bonding_set_task_access_list(CGroupBonding *first, mode_t mode, uid_t uid, gid_t gid, int sticky) { + CGroupBonding *b; + int r; + + LIST_FOREACH(by_unit, b, first) { + r = cgroup_bonding_set_task_access(b, mode, uid, gid, sticky); + if (r < 0) + return r; + } + + return 0; +} + +int cgroup_bonding_kill(CGroupBonding *b, int sig, bool sigcont, bool rem, Set *s, const char *cgroup_suffix) { + char *p = NULL; + const char *path; + int r; + + assert(b); + assert(sig >= 0); + + /* Don't kill cgroups that aren't ours */ + if (!b->ours) + return 0; + + if (cgroup_suffix) { + p = strjoin(b->path, "/", cgroup_suffix, NULL); + if (!p) + return -ENOMEM; + + path = p; + } else + path = b->path; + + r = cg_kill_recursive(b->controller, path, sig, sigcont, true, rem, s); + free(p); + + return r; +} + +int cgroup_bonding_kill_list(CGroupBonding *first, int sig, bool sigcont, bool rem, Set *s, const char *cgroup_suffix) { + CGroupBonding *b; + Set *allocated_set = NULL; + int ret = -EAGAIN, r; + + if (!first) + return 0; + + if (!s) + if (!(s = allocated_set = set_new(trivial_hash_func, trivial_compare_func))) + return -ENOMEM; + + LIST_FOREACH(by_unit, b, first) { + r = cgroup_bonding_kill(b, sig, sigcont, rem, s, cgroup_suffix); + if (r < 0) { + if (r == -EAGAIN || r == -ESRCH) + continue; + + ret = r; + goto finish; + } + + if (ret < 0 || r > 0) + ret = r; + } + +finish: + if (allocated_set) + set_free(allocated_set); + + return ret; +} + +/* Returns 1 if the group is empty, 0 if it is not, -EAGAIN if we + * cannot know */ +int cgroup_bonding_is_empty(CGroupBonding *b) { + int r; + + assert(b); + + if ((r = cg_is_empty_recursive(b->controller, b->path, true)) < 0) + return r; + + /* If it is empty it is empty */ + if (r > 0) + return 1; + + /* It's not only us using this cgroup, so we just don't know */ + return b->ours ? 0 : -EAGAIN; +} + +int cgroup_bonding_is_empty_list(CGroupBonding *first) { + CGroupBonding *b; + + LIST_FOREACH(by_unit, b, first) { + int r; + + if ((r = cgroup_bonding_is_empty(b)) < 0) { + /* If this returned -EAGAIN, then we don't know if the + * group is empty, so let's see if another group can + * tell us */ + + if (r != -EAGAIN) + return r; + } else + return r; + } + + return -EAGAIN; +} + +int manager_setup_cgroup(Manager *m) { + _cleanup_free_ char *current = NULL, *path = NULL; + char suffix_buffer[sizeof("/systemd-") + DECIMAL_STR_MAX(pid_t)]; + const char *suffix; + int r; + + assert(m); + + /* 0. Be nice to Ingo Molnar #628004 */ + if (path_is_mount_point("/sys/fs/cgroup/systemd", false) <= 0) { + log_warning("No control group support available, not creating root group."); + return 0; + } + + /* 1. Determine hierarchy */ + r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, 0, ¤t); + if (r < 0) { + log_error("Cannot determine cgroup we are running in: %s", strerror(-r)); + return r; + } + + if (m->running_as == SYSTEMD_SYSTEM) + suffix = "/system"; + else { + sprintf(suffix_buffer, "/systemd-%lu", (unsigned long) getpid()); + suffix = suffix_buffer; + } + + free(m->cgroup_hierarchy); + if (endswith(current, suffix)) { + /* We probably got reexecuted and can continue to use our root cgroup */ + m->cgroup_hierarchy = current; + current = NULL; + } else { + /* We need a new root cgroup */ + if (streq(current, "/")) + m->cgroup_hierarchy = strdup(suffix); + else + m->cgroup_hierarchy = strappend(current, suffix); + + if (!m->cgroup_hierarchy) + return log_oom(); + } + + /* 2. Show data */ + r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_hierarchy, NULL, &path); + if (r < 0) { + log_error("Cannot find cgroup mount point: %s", strerror(-r)); + return r; + } + + log_debug("Using cgroup controller " SYSTEMD_CGROUP_CONTROLLER ". File system hierarchy is at %s.", path); + + /* 3. Install agent */ + if (m->running_as == SYSTEMD_SYSTEM) { + r = cg_install_release_agent(SYSTEMD_CGROUP_CONTROLLER, SYSTEMD_CGROUP_AGENT_PATH); + if (r < 0) + log_warning("Failed to install release agent, ignoring: %s", strerror(-r)); + else if (r > 0) + log_debug("Installed release agent."); + else + log_debug("Release agent already installed."); + } + + /* 4. Realize the group */ + r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_hierarchy, 0); + if (r < 0) { + log_error("Failed to create root cgroup hierarchy: %s", strerror(-r)); + return r; + } + + /* 5. And pin it, so that it cannot be unmounted */ + if (m->pin_cgroupfs_fd >= 0) + close_nointr_nofail(m->pin_cgroupfs_fd); + + m->pin_cgroupfs_fd = open(path, O_RDONLY|O_CLOEXEC|O_DIRECTORY|O_NOCTTY|O_NONBLOCK); + if (r < 0) { + log_error("Failed to open pin file: %m"); + return -errno; + } + + /* 6. Remove non-existing controllers from the default controllers list */ + cg_shorten_controllers(m->default_controllers); + + /* 7. Let's create the user and machine hierarchies + * right-away, so that people can inotify on them, if they + * wish, without this being racy. */ + if (m->running_as == SYSTEMD_SYSTEM) { + cg_create(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_hierarchy, "../user"); + cg_create(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_hierarchy, "../machine"); + } + + return 0; +} + +void manager_shutdown_cgroup(Manager *m, bool delete) { + assert(m); + + if (delete && m->cgroup_hierarchy) + cg_delete(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_hierarchy); + + if (m->pin_cgroupfs_fd >= 0) { + close_nointr_nofail(m->pin_cgroupfs_fd); + m->pin_cgroupfs_fd = -1; + } + + free(m->cgroup_hierarchy); + m->cgroup_hierarchy = NULL; +} + +int cgroup_bonding_get(Manager *m, const char *cgroup, CGroupBonding **bonding) { + CGroupBonding *b; + char *p; + + assert(m); + assert(cgroup); + assert(bonding); + + b = hashmap_get(m->cgroup_bondings, cgroup); + if (b) { + *bonding = b; + return 1; + } + + p = strdupa(cgroup); + if (!p) + return -ENOMEM; + + for (;;) { + char *e; + + e = strrchr(p, '/'); + if (e == p || !e) { + *bonding = NULL; + return 0; + } + + *e = 0; + + b = hashmap_get(m->cgroup_bondings, p); + if (b) { + *bonding = b; + return 1; + } + } +} + +int cgroup_notify_empty(Manager *m, const char *group) { + CGroupBonding *l, *b; + int r; + + assert(m); + assert(group); + + r = cgroup_bonding_get(m, group, &l); + if (r <= 0) + return r; + + LIST_FOREACH(by_path, b, l) { + int t; + + if (!b->unit) + continue; + + t = cgroup_bonding_is_empty_list(b); + if (t < 0) { + + /* If we don't know, we don't know */ + if (t != -EAGAIN) + log_warning("Failed to check whether cgroup is empty: %s", strerror(errno)); + + continue; + } + + if (t > 0) { + /* If it is empty, let's delete it */ + cgroup_bonding_trim_list(b->unit->cgroup_bondings, true); + + if (UNIT_VTABLE(b->unit)->cgroup_notify_empty) + UNIT_VTABLE(b->unit)->cgroup_notify_empty(b->unit); + } + } + + return 0; +} + +Unit* cgroup_unit_by_pid(Manager *m, pid_t pid) { + CGroupBonding *l, *b; + char *group = NULL; + + assert(m); + + if (pid <= 1) + return NULL; + + if (cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, pid, &group) < 0) + return NULL; + + l = hashmap_get(m->cgroup_bondings, group); + + if (!l) { + char *slash; + + while ((slash = strrchr(group, '/'))) { + if (slash == group) + break; + + *slash = 0; + + if ((l = hashmap_get(m->cgroup_bondings, group))) + break; + } + } + + free(group); + + LIST_FOREACH(by_path, b, l) { + + if (!b->unit) + continue; + + if (b->ours) + return b->unit; + } + + return NULL; +} + +CGroupBonding *cgroup_bonding_find_list(CGroupBonding *first, const char *controller) { + CGroupBonding *b; + + if (!controller) + controller = SYSTEMD_CGROUP_CONTROLLER; + + LIST_FOREACH(by_unit, b, first) + if (streq(b->controller, controller)) + return b; + + return NULL; +} + +char *cgroup_bonding_to_string(CGroupBonding *b) { + char *r; + + assert(b); + + if (asprintf(&r, "%s:%s", b->controller, b->path) < 0) + return NULL; + + return r; +} + +pid_t cgroup_bonding_search_main_pid(CGroupBonding *b) { + FILE *f; + pid_t pid = 0, npid, mypid; + + assert(b); + + if (!b->ours) + return 0; + + if (cg_enumerate_processes(b->controller, b->path, &f) < 0) + return 0; + + mypid = getpid(); + + while (cg_read_pid(f, &npid) > 0) { + pid_t ppid; + + if (npid == pid) + continue; + + /* Ignore processes that aren't our kids */ + if (get_parent_of_pid(npid, &ppid) >= 0 && ppid != mypid) + continue; + + if (pid != 0) { + /* Dang, there's more than one daemonized PID + in this group, so we don't know what process + is the main process. */ + pid = 0; + break; + } + + pid = npid; + } + + fclose(f); + + return pid; +} + +pid_t cgroup_bonding_search_main_pid_list(CGroupBonding *first) { + CGroupBonding *b; + pid_t pid; + + /* Try to find a main pid from this cgroup, but checking if + * there's only one PID in the cgroup and returning it. Later + * on we might want to add additional, smarter heuristics + * here. */ + + LIST_FOREACH(by_unit, b, first) + if ((pid = cgroup_bonding_search_main_pid(b)) != 0) + return pid; + + return 0; + +} diff --git a/src/core/cgroup.h b/src/core/cgroup.h new file mode 100644 index 0000000000..6555d89e37 --- /dev/null +++ b/src/core/cgroup.h @@ -0,0 +1,94 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct CGroupBonding CGroupBonding; + +#include "unit.h" + +/* Binds a cgroup to a name */ +struct CGroupBonding { + char *controller; + char *path; + + Unit *unit; + + /* For the Unit::cgroup_bondings list */ + LIST_FIELDS(CGroupBonding, by_unit); + + /* For the Manager::cgroup_bondings hashmap */ + LIST_FIELDS(CGroupBonding, by_path); + + /* When shutting down, remove cgroup? Are our own tasks the + * only ones in this group?*/ + bool ours:1; + + /* If we cannot create this group, or add a process to it, is this fatal? */ + bool essential:1; + + /* This cgroup is realized */ + bool realized:1; +}; + +int cgroup_bonding_realize(CGroupBonding *b); +int cgroup_bonding_realize_list(CGroupBonding *first); + +void cgroup_bonding_free(CGroupBonding *b, bool trim); +void cgroup_bonding_free_list(CGroupBonding *first, bool trim); + +int cgroup_bonding_install(CGroupBonding *b, pid_t pid, const char *suffix); +int cgroup_bonding_install_list(CGroupBonding *first, pid_t pid, const char *suffix); + +int cgroup_bonding_migrate(CGroupBonding *b, CGroupBonding *list); +int cgroup_bonding_migrate_to(CGroupBonding *b, const char *target, bool rem); + +int cgroup_bonding_set_group_access(CGroupBonding *b, mode_t mode, uid_t uid, gid_t gid); +int cgroup_bonding_set_group_access_list(CGroupBonding *b, mode_t mode, uid_t uid, gid_t gid); + +int cgroup_bonding_set_task_access(CGroupBonding *b, mode_t mode, uid_t uid, gid_t gid, int sticky); +int cgroup_bonding_set_task_access_list(CGroupBonding *b, mode_t mode, uid_t uid, gid_t gid, int sticky); + +int cgroup_bonding_kill(CGroupBonding *b, int sig, bool sigcont, bool rem, Set *s, const char *suffix); +int cgroup_bonding_kill_list(CGroupBonding *first, int sig, bool sigcont, bool rem, Set *s, const char *suffix); + +void cgroup_bonding_trim(CGroupBonding *first, bool delete_root); +void cgroup_bonding_trim_list(CGroupBonding *first, bool delete_root); + +int cgroup_bonding_is_empty(CGroupBonding *b); +int cgroup_bonding_is_empty_list(CGroupBonding *first); + +CGroupBonding *cgroup_bonding_find_list(CGroupBonding *first, const char *controller) _pure_; + +char *cgroup_bonding_to_string(CGroupBonding *b); + +pid_t cgroup_bonding_search_main_pid(CGroupBonding *b); +pid_t cgroup_bonding_search_main_pid_list(CGroupBonding *b); + +#include "manager.h" + +int manager_setup_cgroup(Manager *m); +void manager_shutdown_cgroup(Manager *m, bool delete); + +int cgroup_bonding_get(Manager *m, const char *cgroup, CGroupBonding **bonding); +int cgroup_notify_empty(Manager *m, const char *group); + +Unit* cgroup_unit_by_pid(Manager *m, pid_t pid); diff --git a/src/core/condition.c b/src/core/condition.c new file mode 100644 index 0000000000..16cae6d23b --- /dev/null +++ b/src/core/condition.c @@ -0,0 +1,389 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdlib.h> +#include <errno.h> +#include <string.h> +#include <unistd.h> +#include <sys/capability.h> +#include <sys/statvfs.h> +#include <fnmatch.h> + +#ifdef HAVE_SELINUX +#include <selinux/selinux.h> +#endif + +#include <systemd/sd-id128.h> +#include "util.h" +#include "condition.h" +#include "virt.h" +#include "path-util.h" +#include "fileio.h" + +Condition* condition_new(ConditionType type, const char *parameter, bool trigger, bool negate) { + Condition *c; + + assert(type < _CONDITION_TYPE_MAX); + + c = new0(Condition, 1); + if (!c) + return NULL; + + c->type = type; + c->trigger = trigger; + c->negate = negate; + + if (parameter) { + c->parameter = strdup(parameter); + if (!c->parameter) { + free(c); + return NULL; + } + } + + return c; +} + +void condition_free(Condition *c) { + assert(c); + + free(c->parameter); + free(c); +} + +void condition_free_list(Condition *first) { + Condition *c, *n; + + LIST_FOREACH_SAFE(conditions, c, n, first) + condition_free(c); +} + +static bool test_kernel_command_line(const char *parameter) { + char *line, *w, *state, *word = NULL; + bool equal; + int r; + size_t l, pl; + bool found = false; + + assert(parameter); + + if (detect_container(NULL) > 0) + return false; + + r = read_one_line_file("/proc/cmdline", &line); + if (r < 0) { + log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r)); + return false; + } + + equal = !!strchr(parameter, '='); + pl = strlen(parameter); + + FOREACH_WORD_QUOTED(w, l, line, state) { + + free(word); + word = strndup(w, l); + if (!word) + break; + + if (equal) { + if (streq(word, parameter)) { + found = true; + break; + } + } else { + if (startswith(word, parameter) && (word[pl] == '=' || word[pl] == 0)) { + found = true; + break; + } + } + + } + + free(word); + free(line); + + return found; +} + +static bool test_virtualization(const char *parameter) { + int b; + Virtualization v; + const char *id; + + assert(parameter); + + v = detect_virtualization(&id); + if (v < 0) { + log_warning("Failed to detect virtualization, ignoring: %s", strerror(-v)); + return false; + } + + /* First, compare with yes/no */ + b = parse_boolean(parameter); + + if (v > 0 && b > 0) + return true; + + if (v == 0 && b == 0) + return true; + + /* Then, compare categorization */ + if (v == VIRTUALIZATION_VM && streq(parameter, "vm")) + return true; + + if (v == VIRTUALIZATION_CONTAINER && streq(parameter, "container")) + return true; + + /* Finally compare id */ + return v > 0 && streq(parameter, id); +} + +static bool test_security(const char *parameter) { +#ifdef HAVE_SELINUX + if (streq(parameter, "selinux")) + return is_selinux_enabled() > 0; +#endif + if (streq(parameter, "apparmor")) + return access("/sys/kernel/security/apparmor/", F_OK) == 0; + if (streq(parameter, "smack")) + return access("/sys/fs/smackfs", F_OK) == 0; + return false; +} + +static bool test_capability(const char *parameter) { + cap_value_t value; + FILE *f; + char line[LINE_MAX]; + unsigned long long capabilities = (unsigned long long) -1; + + /* If it's an invalid capability, we don't have it */ + + if (cap_from_name(parameter, &value) < 0) + return false; + + /* If it's a valid capability we default to assume + * that we have it */ + + f = fopen("/proc/self/status", "re"); + if (!f) + return true; + + while (fgets(line, sizeof(line), f)) { + truncate_nl(line); + + if (startswith(line, "CapBnd:")) { + (void) sscanf(line+7, "%llx", &capabilities); + break; + } + } + + fclose(f); + + return !!(capabilities & (1ULL << value)); +} + +static bool test_host(const char *parameter) { + sd_id128_t x, y; + char *h; + int r; + bool b; + + if (sd_id128_from_string(parameter, &x) >= 0) { + + r = sd_id128_get_machine(&y); + if (r < 0) + return false; + + return sd_id128_equal(x, y); + } + + h = gethostname_malloc(); + if (!h) + return false; + + b = fnmatch(parameter, h, FNM_CASEFOLD) == 0; + free(h); + + return b; +} + +static bool test_ac_power(const char *parameter) { + int r; + + r = parse_boolean(parameter); + if (r < 0) + return true; + + return (on_ac_power() != 0) == !!r; +} + +bool condition_test(Condition *c) { + assert(c); + + switch(c->type) { + + case CONDITION_PATH_EXISTS: + return (access(c->parameter, F_OK) >= 0) == !c->negate; + + case CONDITION_PATH_EXISTS_GLOB: + return (glob_exists(c->parameter) > 0) == !c->negate; + + case CONDITION_PATH_IS_DIRECTORY: { + struct stat st; + + if (stat(c->parameter, &st) < 0) + return c->negate; + return S_ISDIR(st.st_mode) == !c->negate; + } + + case CONDITION_PATH_IS_SYMBOLIC_LINK: { + struct stat st; + + if (lstat(c->parameter, &st) < 0) + return c->negate; + return S_ISLNK(st.st_mode) == !c->negate; + } + + case CONDITION_PATH_IS_MOUNT_POINT: + return (path_is_mount_point(c->parameter, true) > 0) == !c->negate; + + case CONDITION_PATH_IS_READ_WRITE: + return (path_is_read_only_fs(c->parameter) > 0) == c->negate; + + case CONDITION_DIRECTORY_NOT_EMPTY: { + int k; + + k = dir_is_empty(c->parameter); + return !(k == -ENOENT || k > 0) == !c->negate; + } + + case CONDITION_FILE_NOT_EMPTY: { + struct stat st; + + if (stat(c->parameter, &st) < 0) + return c->negate; + + return (S_ISREG(st.st_mode) && st.st_size > 0) == !c->negate; + } + + case CONDITION_FILE_IS_EXECUTABLE: { + struct stat st; + + if (stat(c->parameter, &st) < 0) + return c->negate; + + return (S_ISREG(st.st_mode) && (st.st_mode & 0111)) == !c->negate; + } + + case CONDITION_KERNEL_COMMAND_LINE: + return test_kernel_command_line(c->parameter) == !c->negate; + + case CONDITION_VIRTUALIZATION: + return test_virtualization(c->parameter) == !c->negate; + + case CONDITION_SECURITY: + return test_security(c->parameter) == !c->negate; + + case CONDITION_CAPABILITY: + return test_capability(c->parameter) == !c->negate; + + case CONDITION_HOST: + return test_host(c->parameter) == !c->negate; + + case CONDITION_AC_POWER: + return test_ac_power(c->parameter) == !c->negate; + + case CONDITION_NULL: + return !c->negate; + + default: + assert_not_reached("Invalid condition type."); + } +} + +bool condition_test_list(Condition *first) { + Condition *c; + int triggered = -1; + + /* If the condition list is empty, then it is true */ + if (!first) + return true; + + /* Otherwise, if all of the non-trigger conditions apply and + * if any of the trigger conditions apply (unless there are + * none) we return true */ + LIST_FOREACH(conditions, c, first) { + bool b; + + b = condition_test(c); + + if (!c->trigger && !b) + return false; + + if (c->trigger && triggered <= 0) + triggered = b; + } + + return triggered != 0; +} + +void condition_dump(Condition *c, FILE *f, const char *prefix) { + assert(c); + assert(f); + + if (!prefix) + prefix = ""; + + fprintf(f, + "%s\t%s: %s%s%s\n", + prefix, + condition_type_to_string(c->type), + c->trigger ? "|" : "", + c->negate ? "!" : "", + c->parameter); +} + +void condition_dump_list(Condition *first, FILE *f, const char *prefix) { + Condition *c; + + LIST_FOREACH(conditions, c, first) + condition_dump(c, f, prefix); +} + +static const char* const condition_type_table[_CONDITION_TYPE_MAX] = { + [CONDITION_PATH_EXISTS] = "ConditionPathExists", + [CONDITION_PATH_EXISTS_GLOB] = "ConditionPathExistsGlob", + [CONDITION_PATH_IS_DIRECTORY] = "ConditionPathIsDirectory", + [CONDITION_PATH_IS_SYMBOLIC_LINK] = "ConditionPathIsSymbolicLink", + [CONDITION_PATH_IS_MOUNT_POINT] = "ConditionPathIsMountPoint", + [CONDITION_PATH_IS_READ_WRITE] = "ConditionPathIsReadWrite", + [CONDITION_DIRECTORY_NOT_EMPTY] = "ConditionDirectoryNotEmpty", + [CONDITION_FILE_NOT_EMPTY] = "ConditionFileNotEmpty", + [CONDITION_KERNEL_COMMAND_LINE] = "ConditionKernelCommandLine", + [CONDITION_VIRTUALIZATION] = "ConditionVirtualization", + [CONDITION_SECURITY] = "ConditionSecurity", + [CONDITION_HOST] = "ConditionHost", + [CONDITION_AC_POWER] = "ConditionACPower", + [CONDITION_NULL] = "ConditionNull" +}; + +DEFINE_STRING_TABLE_LOOKUP(condition_type, ConditionType); diff --git a/src/core/condition.h b/src/core/condition.h new file mode 100644 index 0000000000..50ed955af9 --- /dev/null +++ b/src/core/condition.h @@ -0,0 +1,70 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +#include "list.h" + +typedef enum ConditionType { + CONDITION_PATH_EXISTS, + CONDITION_PATH_EXISTS_GLOB, + CONDITION_PATH_IS_DIRECTORY, + CONDITION_PATH_IS_SYMBOLIC_LINK, + CONDITION_PATH_IS_MOUNT_POINT, + CONDITION_PATH_IS_READ_WRITE, + CONDITION_DIRECTORY_NOT_EMPTY, + CONDITION_FILE_NOT_EMPTY, + CONDITION_FILE_IS_EXECUTABLE, + CONDITION_KERNEL_COMMAND_LINE, + CONDITION_VIRTUALIZATION, + CONDITION_SECURITY, + CONDITION_CAPABILITY, + CONDITION_HOST, + CONDITION_AC_POWER, + CONDITION_NULL, + _CONDITION_TYPE_MAX, + _CONDITION_TYPE_INVALID = -1 +} ConditionType; + +typedef struct Condition { + ConditionType type; + char *parameter; + + bool trigger:1; + bool negate:1; + + LIST_FIELDS(struct Condition, conditions); +} Condition; + +Condition* condition_new(ConditionType type, const char *parameter, bool trigger, bool negate); +void condition_free(Condition *c); +void condition_free_list(Condition *c); + +bool condition_test(Condition *c); +bool condition_test_list(Condition *c); + +void condition_dump(Condition *c, FILE *f, const char *prefix); +void condition_dump_list(Condition *c, FILE *f, const char *prefix); + +const char* condition_type_to_string(ConditionType t) _const_; +int condition_type_from_string(const char *s) _pure_; diff --git a/src/core/dbus-automount.c b/src/core/dbus-automount.c new file mode 100644 index 0000000000..060cbf7707 --- /dev/null +++ b/src/core/dbus-automount.c @@ -0,0 +1,75 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus-unit.h" +#include "dbus-automount.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_AUTOMOUNT_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Automount\">\n" \ + " <property name=\"Where\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"DirectoryMode\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"Result\" type=\"s\" access=\"read\"/>\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_AUTOMOUNT_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Automount\0" + +const char bus_automount_interface[] _introspect_("Automount") = BUS_AUTOMOUNT_INTERFACE; + +const char bus_automount_invalidating_properties[] = + "Result\0"; + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_automount_append_automount_result, automount_result, AutomountResult); + +static const BusProperty bus_automount_properties[] = { + { "Where", bus_property_append_string, "s", offsetof(Automount, where), true }, + { "DirectoryMode", bus_property_append_mode, "u", offsetof(Automount, directory_mode) }, + { "Result", bus_automount_append_automount_result, "s", offsetof(Automount, result) }, + { NULL, } +}; + +DBusHandlerResult bus_automount_message_handler(Unit *u, DBusConnection *c, DBusMessage *message) { + Automount *am = AUTOMOUNT(u); + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { "org.freedesktop.systemd1.Automount", bus_automount_properties, am }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "status"); + + return bus_default_message_handler(c, message, INTROSPECTION, INTERFACES_LIST, bps); +} diff --git a/src/core/dbus-automount.h b/src/core/dbus-automount.h new file mode 100644 index 0000000000..b338e25fc1 --- /dev/null +++ b/src/core/dbus-automount.h @@ -0,0 +1,31 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_automount_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_automount_interface[]; +extern const char bus_automount_invalidating_properties[]; diff --git a/src/core/dbus-device.c b/src/core/dbus-device.c new file mode 100644 index 0000000000..dbd91fe3db --- /dev/null +++ b/src/core/dbus-device.c @@ -0,0 +1,68 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "dbus-unit.h" +#include "dbus-device.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_DEVICE_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Device\">\n" \ + " <property name=\"SysFSPath\" type=\"s\" access=\"read\"/>\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_DEVICE_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Device\0" + +const char bus_device_interface[] _introspect_("Device") = BUS_DEVICE_INTERFACE; + +const char bus_device_invalidating_properties[] = + "SysFSPath\0"; + +static const BusProperty bus_device_properties[] = { + { "SysFSPath", bus_property_append_string, "s", offsetof(Device, sysfs), true }, + { NULL, } +}; + + +DBusHandlerResult bus_device_message_handler(Unit *u, DBusConnection *c, DBusMessage *message) { + Device *d = DEVICE(u); + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { "org.freedesktop.systemd1.Device", bus_device_properties, d }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "status"); + + return bus_default_message_handler(c, message, INTROSPECTION, INTERFACES_LIST, bps); +} diff --git a/src/core/dbus-device.h b/src/core/dbus-device.h new file mode 100644 index 0000000000..311e0685d1 --- /dev/null +++ b/src/core/dbus-device.h @@ -0,0 +1,31 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_device_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_device_interface[]; +extern const char bus_device_invalidating_properties[]; diff --git a/src/core/dbus-execute.c b/src/core/dbus-execute.c new file mode 100644 index 0000000000..2a8a0e1ac5 --- /dev/null +++ b/src/core/dbus-execute.c @@ -0,0 +1,439 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <dbus/dbus.h> +#include <sys/prctl.h> + +#include "dbus-execute.h" +#include "missing.h" +#include "ioprio.h" +#include "strv.h" +#include "dbus-common.h" +#include "syscall-list.h" +#include "fileio.h" + +DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_execute_append_input, exec_input, ExecInput); +DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_execute_append_output, exec_output, ExecOutput); + +int bus_execute_append_env_files(DBusMessageIter *i, const char *property, void *data) { + char **env_files = data, **j; + DBusMessageIter sub, sub2; + + assert(i); + assert(property); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "(sb)", &sub)) + return -ENOMEM; + + STRV_FOREACH(j, env_files) { + dbus_bool_t b = false; + char *fn = *j; + + if (fn[0] == '-') { + b = true; + fn++; + } + + if (!dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &fn) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_BOOLEAN, &b) || + !dbus_message_iter_close_container(&sub, &sub2)) + return -ENOMEM; + } + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_oom_score_adjust(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + int32_t n; + + assert(i); + assert(property); + assert(c); + + if (c->oom_score_adjust_set) + n = c->oom_score_adjust; + else { + char *t; + + n = 0; + if (read_one_line_file("/proc/self/oom_score_adj", &t) >= 0) { + safe_atoi(t, &n); + free(t); + } + } + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_INT32, &n)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_nice(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + int32_t n; + + assert(i); + assert(property); + assert(c); + + if (c->nice_set) + n = c->nice; + else + n = getpriority(PRIO_PROCESS, 0); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_INT32, &n)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_ioprio(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + int32_t n; + + assert(i); + assert(property); + assert(c); + + if (c->ioprio_set) + n = c->ioprio; + else + n = ioprio_get(IOPRIO_WHO_PROCESS, 0); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_INT32, &n)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_cpu_sched_policy(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + int32_t n; + + assert(i); + assert(property); + assert(c); + + if (c->cpu_sched_set) + n = c->cpu_sched_policy; + else + n = sched_getscheduler(0); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_INT32, &n)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_cpu_sched_priority(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + int32_t n; + + assert(i); + assert(property); + assert(c); + + if (c->cpu_sched_set) + n = c->cpu_sched_priority; + else { + struct sched_param p = {}; + + if (sched_getparam(0, &p) >= 0) + n = p.sched_priority; + else + n = 0; + } + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_INT32, &n)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_affinity(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + dbus_bool_t b; + DBusMessageIter sub; + + assert(i); + assert(property); + assert(c); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "y", &sub)) + return -ENOMEM; + + if (c->cpuset) + b = dbus_message_iter_append_fixed_array(&sub, DBUS_TYPE_BYTE, &c->cpuset, CPU_ALLOC_SIZE(c->cpuset_ncpus)); + else + b = dbus_message_iter_append_fixed_array(&sub, DBUS_TYPE_BYTE, &c->cpuset, 0); + + if (!b) + return -ENOMEM; + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_timer_slack_nsec(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + uint64_t u; + + assert(i); + assert(property); + assert(c); + + if (c->timer_slack_nsec != (nsec_t) -1) + u = (uint64_t) c->timer_slack_nsec; + else + u = (uint64_t) prctl(PR_GET_TIMERSLACK); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_UINT64, &u)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_capability_bs(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + uint64_t normal, inverted; + + assert(i); + assert(property); + assert(c); + + /* We store this negated internally, to match the kernel, but + * we expose it normalized. */ + + normal = *(uint64_t*) data; + inverted = ~normal; + + return bus_property_append_uint64(i, property, &inverted); +} + +int bus_execute_append_capabilities(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + char *t = NULL; + const char *s; + dbus_bool_t b; + + assert(i); + assert(property); + assert(c); + + if (c->capabilities) + s = t = cap_to_text(c->capabilities, NULL); + else + s = ""; + + if (!s) + return -ENOMEM; + + b = dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &s); + + if (t) + cap_free(t); + + if (!b) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_rlimits(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + int r; + uint64_t u; + + assert(i); + assert(property); + assert(c); + + assert_se((r = rlimit_from_string(property)) >= 0); + + if (c->rlimit[r]) + u = (uint64_t) c->rlimit[r]->rlim_max; + else { + struct rlimit rl = {}; + + getrlimit(r, &rl); + + u = (uint64_t) rl.rlim_max; + } + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_UINT64, &u)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_command(DBusMessageIter *i, const char *property, void *data) { + ExecCommand *c = data; + DBusMessageIter sub, sub2, sub3; + + assert(i); + assert(property); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "(sasbttttuii)", &sub)) + return -ENOMEM; + + LIST_FOREACH(command, c, c) { + char **l; + uint32_t pid; + int32_t code, status; + dbus_bool_t b; + + if (!c->path) + continue; + + if (!dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &c->path) || + !dbus_message_iter_open_container(&sub2, DBUS_TYPE_ARRAY, "s", &sub3)) + return -ENOMEM; + + STRV_FOREACH(l, c->argv) + if (!dbus_message_iter_append_basic(&sub3, DBUS_TYPE_STRING, l)) + return -ENOMEM; + + pid = (uint32_t) c->exec_status.pid; + code = (int32_t) c->exec_status.code; + status = (int32_t) c->exec_status.status; + + b = !!c->ignore; + + if (!dbus_message_iter_close_container(&sub2, &sub3) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_BOOLEAN, &b) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT64, &c->exec_status.start_timestamp.realtime) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT64, &c->exec_status.start_timestamp.monotonic) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT64, &c->exec_status.exit_timestamp.realtime) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT64, &c->exec_status.exit_timestamp.monotonic) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT32, &pid) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_INT32, &code) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_INT32, &status)) + return -ENOMEM; + + if (!dbus_message_iter_close_container(&sub, &sub2)) + return -ENOMEM; + } + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +int bus_execute_append_syscall_filter(DBusMessageIter *i, const char *property, void *data) { + ExecContext *c = data; + dbus_bool_t b; + DBusMessageIter sub; + + assert(i); + assert(property); + assert(c); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "u", &sub)) + return -ENOMEM; + + if (c->syscall_filter) + b = dbus_message_iter_append_fixed_array(&sub, DBUS_TYPE_UINT32, &c->syscall_filter, (syscall_max() + 31) >> 4); + else + b = dbus_message_iter_append_fixed_array(&sub, DBUS_TYPE_UINT32, &c->syscall_filter, 0); + + if (!b) + return -ENOMEM; + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +const BusProperty bus_exec_context_properties[] = { + { "Environment", bus_property_append_strv, "as", offsetof(ExecContext, environment), true }, + { "EnvironmentFiles", bus_execute_append_env_files, "a(sb)", offsetof(ExecContext, environment_files), true }, + { "UMask", bus_property_append_mode, "u", offsetof(ExecContext, umask) }, + { "LimitCPU", bus_execute_append_rlimits, "t", 0 }, + { "LimitFSIZE", bus_execute_append_rlimits, "t", 0 }, + { "LimitDATA", bus_execute_append_rlimits, "t", 0 }, + { "LimitSTACK", bus_execute_append_rlimits, "t", 0 }, + { "LimitCORE", bus_execute_append_rlimits, "t", 0 }, + { "LimitRSS", bus_execute_append_rlimits, "t", 0 }, + { "LimitNOFILE", bus_execute_append_rlimits, "t", 0 }, + { "LimitAS", bus_execute_append_rlimits, "t", 0 }, + { "LimitNPROC", bus_execute_append_rlimits, "t", 0 }, + { "LimitMEMLOCK", bus_execute_append_rlimits, "t", 0 }, + { "LimitLOCKS", bus_execute_append_rlimits, "t", 0 }, + { "LimitSIGPENDING", bus_execute_append_rlimits, "t", 0 }, + { "LimitMSGQUEUE", bus_execute_append_rlimits, "t", 0 }, + { "LimitNICE", bus_execute_append_rlimits, "t", 0 }, + { "LimitRTPRIO", bus_execute_append_rlimits, "t", 0 }, + { "LimitRTTIME", bus_execute_append_rlimits, "t", 0 }, + { "WorkingDirectory", bus_property_append_string, "s", offsetof(ExecContext, working_directory), true }, + { "RootDirectory", bus_property_append_string, "s", offsetof(ExecContext, root_directory), true }, + { "OOMScoreAdjust", bus_execute_append_oom_score_adjust, "i", 0 }, + { "Nice", bus_execute_append_nice, "i", 0 }, + { "IOScheduling", bus_execute_append_ioprio, "i", 0 }, + { "CPUSchedulingPolicy", bus_execute_append_cpu_sched_policy, "i", 0 }, + { "CPUSchedulingPriority", bus_execute_append_cpu_sched_priority, "i", 0 }, + { "CPUAffinity", bus_execute_append_affinity, "ay", 0 }, + { "TimerSlackNSec", bus_execute_append_timer_slack_nsec, "t", 0 }, + { "CPUSchedulingResetOnFork", bus_property_append_bool, "b", offsetof(ExecContext, cpu_sched_reset_on_fork) }, + { "NonBlocking", bus_property_append_bool, "b", offsetof(ExecContext, non_blocking) }, + { "StandardInput", bus_execute_append_input, "s", offsetof(ExecContext, std_input) }, + { "StandardOutput", bus_execute_append_output, "s", offsetof(ExecContext, std_output) }, + { "StandardError", bus_execute_append_output, "s", offsetof(ExecContext, std_error) }, + { "TTYPath", bus_property_append_string, "s", offsetof(ExecContext, tty_path), true }, + { "TTYReset", bus_property_append_bool, "b", offsetof(ExecContext, tty_reset) }, + { "TTYVHangup", bus_property_append_bool, "b", offsetof(ExecContext, tty_vhangup) }, + { "TTYVTDisallocate", bus_property_append_bool, "b", offsetof(ExecContext, tty_vt_disallocate) }, + { "SyslogPriority", bus_property_append_int, "i", offsetof(ExecContext, syslog_priority) }, + { "SyslogIdentifier", bus_property_append_string, "s", offsetof(ExecContext, syslog_identifier), true }, + { "SyslogLevelPrefix", bus_property_append_bool, "b", offsetof(ExecContext, syslog_level_prefix) }, + { "Capabilities", bus_execute_append_capabilities, "s", 0 }, + { "SecureBits", bus_property_append_int, "i", offsetof(ExecContext, secure_bits) }, + { "CapabilityBoundingSet", bus_execute_append_capability_bs, "t", offsetof(ExecContext, capability_bounding_set_drop) }, + { "User", bus_property_append_string, "s", offsetof(ExecContext, user), true }, + { "Group", bus_property_append_string, "s", offsetof(ExecContext, group), true }, + { "SupplementaryGroups", bus_property_append_strv, "as", offsetof(ExecContext, supplementary_groups), true }, + { "TCPWrapName", bus_property_append_string, "s", offsetof(ExecContext, tcpwrap_name), true }, + { "PAMName", bus_property_append_string, "s", offsetof(ExecContext, pam_name), true }, + { "ReadWriteDirectories", bus_property_append_strv, "as", offsetof(ExecContext, read_write_dirs), true }, + { "ReadOnlyDirectories", bus_property_append_strv, "as", offsetof(ExecContext, read_only_dirs), true }, + { "InaccessibleDirectories", bus_property_append_strv, "as", offsetof(ExecContext, inaccessible_dirs), true }, + { "MountFlags", bus_property_append_ul, "t", offsetof(ExecContext, mount_flags) }, + { "PrivateTmp", bus_property_append_bool, "b", offsetof(ExecContext, private_tmp) }, + { "PrivateNetwork", bus_property_append_bool, "b", offsetof(ExecContext, private_network) }, + { "SameProcessGroup", bus_property_append_bool, "b", offsetof(ExecContext, same_pgrp) }, + { "UtmpIdentifier", bus_property_append_string, "s", offsetof(ExecContext, utmp_id), true }, + { "ControlGroupModify", bus_property_append_bool, "b", offsetof(ExecContext, control_group_modify) }, + { "ControlGroupPersistent", bus_property_append_tristate_false, "b", offsetof(ExecContext, control_group_persistent) }, + { "IgnoreSIGPIPE", bus_property_append_bool, "b", offsetof(ExecContext, ignore_sigpipe) }, + { "NoNewPrivileges", bus_property_append_bool, "b", offsetof(ExecContext, no_new_privileges) }, + { "SystemCallFilter", bus_execute_append_syscall_filter, "au", 0 }, + { NULL, } +}; diff --git a/src/core/dbus-execute.h b/src/core/dbus-execute.h new file mode 100644 index 0000000000..91d70e535f --- /dev/null +++ b/src/core/dbus-execute.h @@ -0,0 +1,123 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "manager.h" +#include "dbus-common.h" + +#define BUS_EXEC_STATUS_INTERFACE(prefix) \ + " <property name=\"" prefix "StartTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"" prefix "StartTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"" prefix "ExitTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"" prefix "ExitTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"" prefix "PID\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"" prefix "Code\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"" prefix "Status\" type=\"i\" access=\"read\"/>\n" + +#define BUS_EXEC_CONTEXT_INTERFACE \ + " <property name=\"Environment\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"EnvironmentFiles\" type=\"a(sb)\" access=\"read\"/>\n" \ + " <property name=\"UMask\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"LimitCPU\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitFSIZE\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitDATA\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitSTACK\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitCORE\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitRSS\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitNOFILE\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitAS\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitNPROC\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitMEMLOCK\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitLOCKS\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitSIGPENDING\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitMSGQUEUE\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitNICE\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitRTPRIO\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LimitRTTIME\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"WorkingDirectory\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"RootDirectory\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"OOMScoreAdjust\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"Nice\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"IOScheduling\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"CPUSchedulingPolicy\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"CPUSchedulingPriority\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"CPUAffinity\" type=\"ay\" access=\"read\"/>\n" \ + " <property name=\"TimerSlackNS\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"CPUSchedulingResetOnFork\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"NonBlocking\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"StandardInput\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"StandardOutput\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"StandardError\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"TTYPath\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"TTYReset\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"TTYVHangup\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"TTYVTDisallocate\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"SyslogPriority\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"SyslogIdentifier\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"SyslogLevelPrefix\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"Capabilities\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"SecureBits\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"CapabilityBoundingSet\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"User\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Group\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"SupplementaryGroups\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"TCPWrapName\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"PAMName\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"ReadWriteDirectories\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"ReadOnlyDirectories\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"InaccessibleDirectories\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"MountFlags\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"PrivateTmp\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"PrivateNetwork\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"SameProcessGroup\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"UtmpIdentifier\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"ControlGroupModify\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"ControlGroupPersistent\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"IgnoreSIGPIPE\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"NoNewPrivileges\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"SystemCallFilter\" type=\"au\" access=\"read\"/>\n" + +#define BUS_EXEC_COMMAND_INTERFACE(name) \ + " <property name=\"" name "\" type=\"a(sasbttuii)\" access=\"read\"/>\n" + +extern const BusProperty bus_exec_context_properties[]; + +#define BUS_EXEC_COMMAND_PROPERTY(name, command, indirect) \ + { name, bus_execute_append_command, "a(sasbttttuii)", (command), (indirect), NULL } + +int bus_execute_append_output(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_input(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_oom_score_adjust(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_nice(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_ioprio(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_cpu_sched_policy(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_cpu_sched_priority(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_affinity(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_timer_slack_nsec(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_capabilities(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_capability_bs(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_rlimits(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_command(DBusMessageIter *u, const char *property, void *data); +int bus_execute_append_env_files(DBusMessageIter *i, const char *property, void *data); +int bus_execute_append_syscall_filter(DBusMessageIter *i, const char *property, void *data); diff --git a/src/core/dbus-job.c b/src/core/dbus-job.c new file mode 100644 index 0000000000..98ccfa62ec --- /dev/null +++ b/src/core/dbus-job.c @@ -0,0 +1,373 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus.h" +#include "log.h" +#include "dbus-job.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_JOB_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Job\">\n" \ + " <method name=\"Cancel\"/>\n" \ + " <property name=\"Id\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"Unit\" type=\"(so)\" access=\"read\"/>\n" \ + " <property name=\"JobType\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"State\" type=\"s\" access=\"read\"/>\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_JOB_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +const char bus_job_interface[] _introspect_("Job") = BUS_JOB_INTERFACE; + +#define INTERFACES_LIST \ + BUS_GENERIC_INTERFACES_LIST \ + "org.freedesktop.systemd1.Job\0" + +#define INVALIDATING_PROPERTIES \ + "State\0" + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_job_append_state, job_state, JobState); +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_job_append_type, job_type, JobType); + +static int bus_job_append_unit(DBusMessageIter *i, const char *property, void *data) { + Job *j = data; + DBusMessageIter sub; + char *p; + + assert(i); + assert(property); + assert(j); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_STRUCT, NULL, &sub)) + return -ENOMEM; + + p = unit_dbus_path(j->unit); + if (!p) + return -ENOMEM; + + if (!dbus_message_iter_append_basic(&sub, DBUS_TYPE_STRING, &j->unit->id) || + !dbus_message_iter_append_basic(&sub, DBUS_TYPE_OBJECT_PATH, &p)) { + free(p); + return -ENOMEM; + } + + free(p); + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static const BusProperty bus_job_properties[] = { + { "Id", bus_property_append_uint32, "u", offsetof(Job, id) }, + { "State", bus_job_append_state, "s", offsetof(Job, state) }, + { "JobType", bus_job_append_type, "s", offsetof(Job, type) }, + { "Unit", bus_job_append_unit, "(so)", 0 }, + { NULL, } +}; + +static DBusHandlerResult bus_job_message_dispatch(Job *j, DBusConnection *connection, DBusMessage *message) { + _cleanup_dbus_message_unref_ DBusMessage *reply = NULL; + + if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Job", "Cancel")) { + + SELINUX_UNIT_ACCESS_CHECK(j->unit, connection, message, "stop"); + job_finish_and_invalidate(j, JOB_CANCELED, true); + + reply = dbus_message_new_method_return(message); + if (!reply) + return DBUS_HANDLER_RESULT_NEED_MEMORY; + } else { + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Job", bus_job_properties, j }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(j->unit, connection, message, "status"); + return bus_default_message_handler(connection, message, INTROSPECTION, INTERFACES_LIST, bps); + } + + if (!bus_maybe_send_reply(connection, message, reply)) + return DBUS_HANDLER_RESULT_NEED_MEMORY; + + return DBUS_HANDLER_RESULT_HANDLED; +} + +static DBusHandlerResult bus_job_message_handler(DBusConnection *connection, DBusMessage *message, void *data) { + Manager *m = data; + Job *j; + int r; + _cleanup_dbus_message_unref_ DBusMessage *reply = NULL; + + assert(connection); + assert(message); + assert(m); + + if (streq(dbus_message_get_path(message), "/org/freedesktop/systemd1/job")) { + /* Be nice to gdbus and return introspection data for our mid-level paths */ + + if (dbus_message_is_method_call(message, "org.freedesktop.DBus.Introspectable", "Introspect")) { + char *introspection = NULL; + FILE *f; + Iterator i; + size_t size; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + /* We roll our own introspection code here, instead of + * relying on bus_default_message_handler() because we + * need to generate our introspection string + * dynamically. */ + + f = open_memstream(&introspection, &size); + if (!f) + goto oom; + + fputs(DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE + "<node>\n", f); + + fputs(BUS_INTROSPECTABLE_INTERFACE, f); + fputs(BUS_PEER_INTERFACE, f); + + HASHMAP_FOREACH(j, m->jobs, i) + fprintf(f, "<node name=\"job/%lu\"/>", (unsigned long) j->id); + + fputs("</node>\n", f); + + if (ferror(f)) { + fclose(f); + free(introspection); + goto oom; + } + + fclose(f); + + if (!introspection) + goto oom; + + if (!dbus_message_append_args(reply, DBUS_TYPE_STRING, &introspection, DBUS_TYPE_INVALID)) { + free(introspection); + goto oom; + } + + free(introspection); + + if (!bus_maybe_send_reply(connection, message, reply)) + goto oom; + + return DBUS_HANDLER_RESULT_HANDLED; + } + + return DBUS_HANDLER_RESULT_NOT_YET_HANDLED; + } + + r = manager_get_job_from_dbus_path(m, dbus_message_get_path(message), &j); + if (r == -ENOMEM) + goto oom; + if (r == -ENOENT) { + DBusError e; + + dbus_error_init(&e); + dbus_set_error_const(&e, DBUS_ERROR_UNKNOWN_OBJECT, "Unknown job"); + return bus_send_error_reply(connection, message, &e, r); + } + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + return bus_job_message_dispatch(j, connection, message); + +oom: + return DBUS_HANDLER_RESULT_NEED_MEMORY; +} + +const DBusObjectPathVTable bus_job_vtable = { + .message_function = bus_job_message_handler +}; + +static int job_send_message(Job *j, DBusMessage* (*new_message)(Job *j)) { + _cleanup_dbus_message_unref_ DBusMessage *m = NULL; + int r; + + assert(j); + assert(new_message); + + if (bus_has_subscriber(j->manager) || j->forgot_bus_clients) { + m = new_message(j); + if (!m) + return -ENOMEM; + + r = bus_broadcast(j->manager, m); + if (r < 0) + return r; + + } else { + /* If nobody is subscribed, we just send the message + * to the client(s) which created the job */ + JobBusClient *cl; + assert(j->bus_client_list); + + LIST_FOREACH(client, cl, j->bus_client_list) { + assert(cl->bus); + + m = new_message(j); + if (!m) + return -ENOMEM; + + if (!dbus_message_set_destination(m, cl->name)) + return -ENOMEM; + + if (!dbus_connection_send(cl->bus, m, NULL)) + return -ENOMEM; + + dbus_message_unref(m); + m = NULL; + } + } + + return 0; +} + +static DBusMessage* new_change_signal_message(Job *j) { + DBusMessage *m = NULL; + char *p = NULL; + + p = job_dbus_path(j); + if (!p) + goto oom; + + if (j->sent_dbus_new_signal) { + /* Send a properties changed signal */ + m = bus_properties_changed_new(p, "org.freedesktop.systemd1.Job", INVALIDATING_PROPERTIES); + if (!m) + goto oom; + + } else { + /* Send a new signal */ + + m = dbus_message_new_signal("/org/freedesktop/systemd1", "org.freedesktop.systemd1.Manager", "JobNew"); + if (!m) + goto oom; + + if (!dbus_message_append_args(m, + DBUS_TYPE_UINT32, &j->id, + DBUS_TYPE_OBJECT_PATH, &p, + DBUS_TYPE_STRING, &j->unit->id, + DBUS_TYPE_INVALID)) + goto oom; + } + + return m; + +oom: + if (m) + dbus_message_unref(m); + free(p); + return NULL; +} + +static DBusMessage* new_removed_signal_message(Job *j) { + DBusMessage *m = NULL; + char *p = NULL; + const char *r; + + p = job_dbus_path(j); + if (!p) + goto oom; + + m = dbus_message_new_signal("/org/freedesktop/systemd1", "org.freedesktop.systemd1.Manager", "JobRemoved"); + if (!m) + goto oom; + + r = job_result_to_string(j->result); + + if (!dbus_message_append_args(m, + DBUS_TYPE_UINT32, &j->id, + DBUS_TYPE_OBJECT_PATH, &p, + DBUS_TYPE_STRING, &j->unit->id, + DBUS_TYPE_STRING, &r, + DBUS_TYPE_INVALID)) + goto oom; + + return m; + +oom: + if (m) + dbus_message_unref(m); + free(p); + return NULL; +} + +void bus_job_send_change_signal(Job *j) { + assert(j); + + if (j->in_dbus_queue) { + LIST_REMOVE(Job, dbus_queue, j->manager->dbus_job_queue, j); + j->in_dbus_queue = false; + } + + if (!bus_has_subscriber(j->manager) && !j->bus_client_list && !j->forgot_bus_clients) { + j->sent_dbus_new_signal = true; + return; + } + + if (job_send_message(j, new_change_signal_message) < 0) + goto oom; + + j->sent_dbus_new_signal = true; + + return; + +oom: + log_error("Failed to allocate job change signal."); +} + +void bus_job_send_removed_signal(Job *j) { + assert(j); + + if (!bus_has_subscriber(j->manager) && !j->bus_client_list && !j->forgot_bus_clients) + return; + + if (!j->sent_dbus_new_signal) + bus_job_send_change_signal(j); + + if (job_send_message(j, new_removed_signal_message) < 0) + goto oom; + + return; + +oom: + log_error("Failed to allocate job remove signal."); +} diff --git a/src/core/dbus-job.h b/src/core/dbus-job.h new file mode 100644 index 0000000000..a1b928fb16 --- /dev/null +++ b/src/core/dbus-job.h @@ -0,0 +1,33 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "job.h" + +void bus_job_send_change_signal(Job *j); +void bus_job_send_removed_signal(Job *j); + +extern const DBusObjectPathVTable bus_job_vtable; + +extern const char bus_job_interface[]; diff --git a/src/core/dbus-kill.c b/src/core/dbus-kill.c new file mode 100644 index 0000000000..165f63074b --- /dev/null +++ b/src/core/dbus-kill.c @@ -0,0 +1,35 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <dbus/dbus.h> + +#include "dbus-kill.h" +#include "dbus-common.h" + +DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_kill_append_mode, kill_mode, KillMode); + +const BusProperty bus_kill_context_properties[] = { + { "KillMode", bus_kill_append_mode, "s", offsetof(KillContext, kill_mode) }, + { "KillSignal", bus_property_append_int, "i", offsetof(KillContext, kill_signal) }, + { "SendSIGKILL", bus_property_append_bool, "b", offsetof(KillContext, send_sigkill) }, + { NULL, } +}; diff --git a/src/core/dbus-kill.h b/src/core/dbus-kill.h new file mode 100644 index 0000000000..238fbd36d6 --- /dev/null +++ b/src/core/dbus-kill.h @@ -0,0 +1,39 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "manager.h" +#include "dbus-common.h" + +#define BUS_KILL_CONTEXT_INTERFACE \ + " <property name=\"KillMode\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"KillSignal\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"SendSIGKILL\" type=\"b\" access=\"read\"/>\n" + +#define BUS_KILL_COMMAND_INTERFACE(name) \ + " <property name=\"" name "\" type=\"a(sasbttuii)\" access=\"read\"/>\n" + +extern const BusProperty bus_kill_context_properties[]; + +int bus_kill_append_mode(DBusMessageIter *i, const char *property, void *data); diff --git a/src/core/dbus-manager.c b/src/core/dbus-manager.c new file mode 100644 index 0000000000..56b02a1cf5 --- /dev/null +++ b/src/core/dbus-manager.c @@ -0,0 +1,1912 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <unistd.h> + +#include "dbus.h" +#include "log.h" +#include "dbus-manager.h" +#include "strv.h" +#include "bus-errors.h" +#include "build.h" +#include "dbus-common.h" +#include "install.h" +#include "selinux-access.h" +#include "watchdog.h" +#include "hwclock.h" +#include "path-util.h" +#include "dbus-unit.h" +#include "virt.h" +#include "env-util.h" + +#define BUS_MANAGER_INTERFACE_BEGIN \ + " <interface name=\"org.freedesktop.systemd1.Manager\">\n" + +#define BUS_MANAGER_INTERFACE_METHODS \ + " <method name=\"GetUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"unit\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"GetUnitByPID\">\n" \ + " <arg name=\"pid\" type=\"u\" direction=\"in\"/>\n" \ + " <arg name=\"unit\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"LoadUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"unit\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"StartUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"StartUnitReplace\">\n" \ + " <arg name=\"old_unit\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"new_unit\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"StopUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"ReloadUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"RestartUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"TryRestartUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"ReloadOrRestartUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"ReloadOrTryRestartUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"KillUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"who\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"signal\" type=\"i\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"ResetFailedUnit\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"SetUnitControlGroup\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"group\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"UnsetUnitControlGroup\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"group\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"\n/>" \ + " </method>\n" \ + " <method name=\"GetUnitControlGroupAttribute\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"attribute\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"values\" type=\"as\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"SetUnitControlGroupAttribute\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"attribute\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"values\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"\n/>" \ + " </method>\n" \ + " <method name=\"UnsetUnitControlGroupAttributes\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"attribute\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"GetJob\">\n" \ + " <arg name=\"id\" type=\"u\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"CancelJob\">\n" \ + " <arg name=\"id\" type=\"u\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"ClearJobs\"/>\n" \ + " <method name=\"ResetFailed\"/>\n" \ + " <method name=\"ListUnits\">\n" \ + " <arg name=\"units\" type=\"a(ssssssouso)\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"ListJobs\">\n" \ + " <arg name=\"jobs\" type=\"a(usssoo)\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"Subscribe\"/>\n" \ + " <method name=\"Unsubscribe\"/>\n" \ + " <method name=\"Dump\">\n" \ + " <arg name=\"dump\" type=\"s\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"CreateSnapshot\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"cleanup\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"unit\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"RemoveSnapshot\">\n" \ + " <arg name=\"name\" type=\"s\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"Reload\"/>\n" \ + " <method name=\"Reexecute\"/>\n" \ + " <method name=\"Exit\"/>\n" \ + " <method name=\"Reboot\"/>\n" \ + " <method name=\"PowerOff\"/>\n" \ + " <method name=\"Halt\"/>\n" \ + " <method name=\"KExec\"/>\n" \ + " <method name=\"SwitchRoot\">\n" \ + " <arg name=\"new_root\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"init\" type=\"s\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"SetEnvironment\">\n" \ + " <arg name=\"names\" type=\"as\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"UnsetEnvironment\">\n" \ + " <arg name=\"names\" type=\"as\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"UnsetAndSetEnvironment\">\n" \ + " <arg name=\"unset\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"set\" type=\"as\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"ListUnitFiles\">\n" \ + " <arg name=\"files\" type=\"a(ss)\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"GetUnitFileState\">\n" \ + " <arg name=\"file\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"state\" type=\"s\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"EnableUnitFiles\">\n" \ + " <arg name=\"files\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"runtime\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"force\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"carries_install_info\" type=\"b\" direction=\"out\"/>\n" \ + " <arg name=\"changes\" type=\"a(sss)\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"DisableUnitFiles\">\n" \ + " <arg name=\"files\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"runtime\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"changes\" type=\"a(sss)\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"ReenableUnitFiles\">\n" \ + " <arg name=\"files\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"runtime\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"force\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"carries_install_info\" type=\"b\" direction=\"out\"/>\n" \ + " <arg name=\"changes\" type=\"a(sss)\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"LinkUnitFiles\">\n" \ + " <arg name=\"files\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"runtime\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"force\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"changes\" type=\"a(sss)\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"PresetUnitFiles\">\n" \ + " <arg name=\"files\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"runtime\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"force\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"carries_install_info\" type=\"b\" direction=\"out\"/>\n" \ + " <arg name=\"changes\" type=\"a(sss)\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"MaskUnitFiles\">\n" \ + " <arg name=\"files\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"runtime\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"force\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"changes\" type=\"a(sss)\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"UnmaskUnitFiles\">\n" \ + " <arg name=\"files\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"runtime\" type=\"b\" direction=\"in\"/>\n" \ + " <arg name=\"changes\" type=\"a(sss)\" direction=\"out\"/>\n" \ + " </method>\n" + +#define BUS_MANAGER_INTERFACE_SIGNALS \ + " <signal name=\"UnitNew\">\n" \ + " <arg name=\"id\" type=\"s\"/>\n" \ + " <arg name=\"unit\" type=\"o\"/>\n" \ + " </signal>\n" \ + " <signal name=\"UnitRemoved\">\n" \ + " <arg name=\"id\" type=\"s\"/>\n" \ + " <arg name=\"unit\" type=\"o\"/>\n" \ + " </signal>\n" \ + " <signal name=\"JobNew\">\n" \ + " <arg name=\"id\" type=\"u\"/>\n" \ + " <arg name=\"job\" type=\"o\"/>\n" \ + " <arg name=\"unit\" type=\"s\"/>\n" \ + " </signal>\n" \ + " <signal name=\"JobRemoved\">\n" \ + " <arg name=\"id\" type=\"u\"/>\n" \ + " <arg name=\"job\" type=\"o\"/>\n" \ + " <arg name=\"unit\" type=\"s\"/>\n" \ + " <arg name=\"result\" type=\"s\"/>\n" \ + " </signal>" \ + " <signal name=\"StartupFinished\">\n" \ + " <arg name=\"firmware\" type=\"t\"/>\n" \ + " <arg name=\"loader\" type=\"t\"/>\n" \ + " <arg name=\"kernel\" type=\"t\"/>\n" \ + " <arg name=\"initrd\" type=\"t\"/>\n" \ + " <arg name=\"userspace\" type=\"t\"/>\n" \ + " <arg name=\"total\" type=\"t\"/>\n" \ + " </signal>" \ + " <signal name=\"UnitFilesChanged\"/>\n" + +#define BUS_MANAGER_INTERFACE_PROPERTIES_GENERAL \ + " <property name=\"Version\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Features\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Tainted\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"FirmwareTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"FirmwareTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LoaderTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LoaderTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"KernelTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"KernelTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"InitRDTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"InitRDTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"UserspaceTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"UserspaceTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"FinishTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"FinishTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"LogLevel\" type=\"s\" access=\"readwrite\"/>\n" \ + " <property name=\"LogTarget\" type=\"s\" access=\"readwrite\"/>\n" \ + " <property name=\"NNames\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"NJobs\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"NInstalledJobs\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"NFailedJobs\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"Progress\" type=\"d\" access=\"read\"/>\n" \ + " <property name=\"Environment\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"ConfirmSpawn\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"ShowStatus\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"UnitPath\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"ControlGroupHierarchy\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"DefaultControllers\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"DefaultStandardOutput\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"DefaultStandardError\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"RuntimeWatchdogUSec\" type=\"t\" access=\"readwrite\"/>\n" \ + " <property name=\"ShutdownWatchdogUSec\" type=\"t\" access=\"readwrite\"/>\n" \ + " <property name=\"Virtualization\" type=\"s\" access=\"read\"/>\n" + +#define BUS_MANAGER_INTERFACE_END \ + " </interface>\n" + +#define BUS_MANAGER_INTERFACE \ + BUS_MANAGER_INTERFACE_BEGIN \ + BUS_MANAGER_INTERFACE_METHODS \ + BUS_MANAGER_INTERFACE_SIGNALS \ + BUS_MANAGER_INTERFACE_PROPERTIES_GENERAL \ + BUS_MANAGER_INTERFACE_END + +#define INTROSPECTION_BEGIN \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_MANAGER_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE + +#define INTROSPECTION_END \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_GENERIC_INTERFACES_LIST \ + "org.freedesktop.systemd1.Manager\0" + +const char bus_manager_interface[] _introspect_("Manager") = BUS_MANAGER_INTERFACE; + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_manager_append_exec_output, exec_output, ExecOutput); + +static int bus_manager_append_tainted(DBusMessageIter *i, const char *property, void *data) { + const char *t; + Manager *m = data; + char buf[LINE_MAX] = "", *e = buf, *p = NULL; + + assert(i); + assert(property); + assert(m); + + if (m->taint_usr) + e = stpcpy(e, "split-usr:"); + + if (readlink_malloc("/etc/mtab", &p) < 0) + e = stpcpy(e, "mtab-not-symlink:"); + else + free(p); + + if (access("/proc/cgroups", F_OK) < 0) + e = stpcpy(e, "cgroups-missing:"); + + if (hwclock_is_localtime() > 0) + e = stpcpy(e, "local-hwclock:"); + + /* remove the last ':' */ + if (e != buf) + e[-1] = 0; + + t = buf; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &t)) + return -ENOMEM; + + return 0; +} + +static int bus_manager_append_log_target(DBusMessageIter *i, const char *property, void *data) { + const char *t; + + assert(i); + assert(property); + + t = log_target_to_string(log_get_target()); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &t)) + return -ENOMEM; + + return 0; +} + +static int bus_manager_set_log_target(DBusMessageIter *i, const char *property, void *data) { + const char *t; + + assert(i); + assert(property); + + dbus_message_iter_get_basic(i, &t); + + return log_set_target_from_string(t); +} + +static int bus_manager_append_log_level(DBusMessageIter *i, const char *property, void *data) { + char *t; + int r; + + assert(i); + assert(property); + + r = log_level_to_string_alloc(log_get_max_level(), &t); + if (r < 0) + return r; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &t)) + r = -ENOMEM; + + free(t); + return r; +} + +static int bus_manager_set_log_level(DBusMessageIter *i, const char *property, void *data) { + const char *t; + + assert(i); + assert(property); + + dbus_message_iter_get_basic(i, &t); + + return log_set_max_level_from_string(t); +} + +static int bus_manager_append_n_names(DBusMessageIter *i, const char *property, void *data) { + Manager *m = data; + uint32_t u; + + assert(i); + assert(property); + assert(m); + + u = hashmap_size(m->units); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_UINT32, &u)) + return -ENOMEM; + + return 0; +} + +static int bus_manager_append_n_jobs(DBusMessageIter *i, const char *property, void *data) { + Manager *m = data; + uint32_t u; + + assert(i); + assert(property); + assert(m); + + u = hashmap_size(m->jobs); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_UINT32, &u)) + return -ENOMEM; + + return 0; +} + +static int bus_manager_append_progress(DBusMessageIter *i, const char *property, void *data) { + double d; + Manager *m = data; + + assert(i); + assert(property); + assert(m); + + if (dual_timestamp_is_set(&m->finish_timestamp)) + d = 1.0; + else + d = 1.0 - ((double) hashmap_size(m->jobs) / (double) m->n_installed_jobs); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_DOUBLE, &d)) + return -ENOMEM; + + return 0; +} + +static int bus_manager_append_virt(DBusMessageIter *i, const char *property, void *data) { + Manager *m = data; + const char *id = ""; + + assert(i); + assert(property); + assert(m); + + detect_virtualization(&id); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &id)) + return -ENOMEM; + + return 0; +} + +static DBusMessage *message_from_file_changes( + DBusMessage *m, + UnitFileChange *changes, + unsigned n_changes, + int carries_install_info) { + + DBusMessageIter iter, sub, sub2; + DBusMessage *reply; + unsigned i; + + reply = dbus_message_new_method_return(m); + if (!reply) + return NULL; + + dbus_message_iter_init_append(reply, &iter); + + if (carries_install_info >= 0) { + dbus_bool_t b; + + b = !!carries_install_info; + if (!dbus_message_iter_append_basic(&iter, DBUS_TYPE_BOOLEAN, &b)) + goto oom; + } + + if (!dbus_message_iter_open_container(&iter, DBUS_TYPE_ARRAY, "(sss)", &sub)) + goto oom; + + for (i = 0; i < n_changes; i++) { + const char *type, *path, *source; + + type = unit_file_change_type_to_string(changes[i].type); + path = strempty(changes[i].path); + source = strempty(changes[i].source); + + if (!dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &type) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &path) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &source) || + !dbus_message_iter_close_container(&sub, &sub2)) + goto oom; + } + + if (!dbus_message_iter_close_container(&iter, &sub)) + goto oom; + + return reply; + +oom: + dbus_message_unref(reply); + return NULL; +} + +static int bus_manager_send_unit_files_changed(Manager *m) { + DBusMessage *s; + int r; + + s = dbus_message_new_signal("/org/freedesktop/systemd1", "org.freedesktop.systemd1.Manager", "UnitFilesChanged"); + if (!s) + return -ENOMEM; + + r = bus_broadcast(m, s); + dbus_message_unref(s); + + return r; +} + +static int bus_manager_set_runtime_watchdog_usec(DBusMessageIter *i, const char *property, void *data) { + uint64_t *t = data; + + assert(i); + assert(property); + + dbus_message_iter_get_basic(i, t); + + return watchdog_set_timeout(t); +} + +static const char systemd_property_string[] = + PACKAGE_STRING "\0" + SYSTEMD_FEATURES; + +static const BusProperty bus_systemd_properties[] = { + { "Version", bus_property_append_string, "s", 0 }, + { "Features", bus_property_append_string, "s", sizeof(PACKAGE_STRING) }, + { NULL, } +}; + +static const BusProperty bus_manager_properties[] = { + { "Tainted", bus_manager_append_tainted, "s", 0 }, + { "FirmwareTimestamp", bus_property_append_uint64, "t", offsetof(Manager, firmware_timestamp.realtime) }, + { "FirmwareTimestampMonotonic", bus_property_append_uint64, "t", offsetof(Manager, firmware_timestamp.monotonic) }, + { "LoaderTimestamp", bus_property_append_uint64, "t", offsetof(Manager, loader_timestamp.realtime) }, + { "LoaderTimestampMonotonic", bus_property_append_uint64, "t", offsetof(Manager, loader_timestamp.monotonic) }, + { "KernelTimestamp", bus_property_append_uint64, "t", offsetof(Manager, kernel_timestamp.realtime) }, + { "KernelTimestampMonotonic", bus_property_append_uint64, "t", offsetof(Manager, kernel_timestamp.monotonic) }, + { "InitRDTimestamp", bus_property_append_uint64, "t", offsetof(Manager, initrd_timestamp.realtime) }, + { "InitRDTimestampMonotonic", bus_property_append_uint64, "t", offsetof(Manager, initrd_timestamp.monotonic) }, + { "UserspaceTimestamp", bus_property_append_uint64, "t", offsetof(Manager, userspace_timestamp.realtime) }, + { "UserspaceTimestampMonotonic", bus_property_append_uint64, "t", offsetof(Manager, userspace_timestamp.monotonic) }, + { "FinishTimestamp", bus_property_append_uint64, "t", offsetof(Manager, finish_timestamp.realtime) }, + { "FinishTimestampMonotonic", bus_property_append_uint64, "t", offsetof(Manager, finish_timestamp.monotonic) }, + { "LogLevel", bus_manager_append_log_level, "s", 0, false, bus_manager_set_log_level }, + { "LogTarget", bus_manager_append_log_target, "s", 0, false, bus_manager_set_log_target }, + { "NNames", bus_manager_append_n_names, "u", 0 }, + { "NJobs", bus_manager_append_n_jobs, "u", 0 }, + { "NInstalledJobs", bus_property_append_uint32, "u", offsetof(Manager, n_installed_jobs) }, + { "NFailedJobs", bus_property_append_uint32, "u", offsetof(Manager, n_failed_jobs) }, + { "Progress", bus_manager_append_progress, "d", 0 }, + { "Environment", bus_property_append_strv, "as", offsetof(Manager, environment), true }, + { "ConfirmSpawn", bus_property_append_bool, "b", offsetof(Manager, confirm_spawn) }, + { "ShowStatus", bus_property_append_bool, "b", offsetof(Manager, show_status) }, + { "UnitPath", bus_property_append_strv, "as", offsetof(Manager, lookup_paths.unit_path), true }, + { "ControlGroupHierarchy", bus_property_append_string, "s", offsetof(Manager, cgroup_hierarchy), true }, + { "DefaultControllers", bus_property_append_strv, "as", offsetof(Manager, default_controllers), true }, + { "DefaultStandardOutput", bus_manager_append_exec_output, "s", offsetof(Manager, default_std_output) }, + { "DefaultStandardError", bus_manager_append_exec_output, "s", offsetof(Manager, default_std_error) }, + { "RuntimeWatchdogUSec", bus_property_append_usec, "t", offsetof(Manager, runtime_watchdog), false, bus_manager_set_runtime_watchdog_usec }, + { "ShutdownWatchdogUSec", bus_property_append_usec, "t", offsetof(Manager, shutdown_watchdog), false, bus_property_set_usec }, + { "Virtualization", bus_manager_append_virt, "s", 0, }, + { NULL, } +}; + +static DBusHandlerResult bus_manager_message_handler(DBusConnection *connection, DBusMessage *message, void *data) { + _cleanup_dbus_message_unref_ DBusMessage *reply = NULL; + _cleanup_free_ char * path = NULL; + Manager *m = data; + int r; + DBusError error; + JobType job_type = _JOB_TYPE_INVALID; + bool reload_if_possible = false; + const char *member; + + assert(connection); + assert(message); + assert(m); + + dbus_error_init(&error); + + member = dbus_message_get_member(message); + + if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "GetUnit")) { + const char *name; + Unit *u; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + u = manager_get_unit(m, name); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + path = unit_dbus_path(u); + if (!path) + goto oom; + + if (!dbus_message_append_args( + reply, + DBUS_TYPE_OBJECT_PATH, &path, + DBUS_TYPE_INVALID)) + goto oom; + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "GetUnitByPID")) { + Unit *u; + uint32_t pid; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_UINT32, &pid, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + u = cgroup_unit_by_pid(m, (pid_t) pid); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "No unit for PID %lu is loaded.", (unsigned long) pid); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + path = unit_dbus_path(u); + if (!path) + goto oom; + + if (!dbus_message_append_args( + reply, + DBUS_TYPE_OBJECT_PATH, &path, + DBUS_TYPE_INVALID)) + goto oom; + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "LoadUnit")) { + const char *name; + Unit *u; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + r = manager_load_unit(m, name, NULL, &error, &u); + if (r < 0) + return bus_send_error_reply(connection, message, &error, r); + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + path = unit_dbus_path(u); + if (!path) + goto oom; + + if (!dbus_message_append_args( + reply, + DBUS_TYPE_OBJECT_PATH, &path, + DBUS_TYPE_INVALID)) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "StartUnit")) + job_type = JOB_START; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "StartUnitReplace")) + job_type = JOB_START; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "StopUnit")) + job_type = JOB_STOP; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ReloadUnit")) + job_type = JOB_RELOAD; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "RestartUnit")) + job_type = JOB_RESTART; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "TryRestartUnit")) + job_type = JOB_TRY_RESTART; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ReloadOrRestartUnit")) { + reload_if_possible = true; + job_type = JOB_RESTART; + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ReloadOrTryRestartUnit")) { + reload_if_possible = true; + job_type = JOB_TRY_RESTART; + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "KillUnit")) { + const char *name, *swho; + int32_t signo; + Unit *u; + KillWho who; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_STRING, &swho, + DBUS_TYPE_INT32, &signo, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + if (isempty(swho)) + who = KILL_ALL; + else { + who = kill_who_from_string(swho); + if (who < 0) + return bus_send_error_reply(connection, message, &error, -EINVAL); + } + + if (signo <= 0 || signo >= _NSIG) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + u = manager_get_unit(m, name); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "stop"); + + r = unit_kill(u, who, signo, &error); + if (r < 0) + return bus_send_error_reply(connection, message, &error, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "GetJob")) { + uint32_t id; + Job *j; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_UINT32, &id, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + j = manager_get_job(m, id); + if (!j) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_JOB, "Job %u does not exist.", (unsigned) id); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(j->unit, connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + path = job_dbus_path(j); + if (!path) + goto oom; + + if (!dbus_message_append_args( + reply, + DBUS_TYPE_OBJECT_PATH, &path, + DBUS_TYPE_INVALID)) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "CancelJob")) { + uint32_t id; + Job *j; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_UINT32, &id, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + j = manager_get_job(m, id); + if (!j) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_JOB, "Job %u does not exist.", (unsigned) id); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(j->unit, connection, message, "stop"); + job_finish_and_invalidate(j, JOB_CANCELED, true); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ClearJobs")) { + + SELINUX_ACCESS_CHECK(connection, message, "reboot"); + manager_clear_jobs(m); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ResetFailed")) { + + SELINUX_ACCESS_CHECK(connection, message, "reload"); + + manager_reset_failed(m); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ResetFailedUnit")) { + const char *name; + Unit *u; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + u = manager_get_unit(m, name); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "reload"); + + unit_reset_failed(u); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "SetUnitControlGroup")) { + const char *name; + Unit *u; + DBusMessageIter iter; + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_iter_get_basic_and_next(&iter, DBUS_TYPE_STRING, &name, true); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + u = manager_get_unit(m, name); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "start"); + + r = bus_unit_cgroup_set(u, &iter); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "UnsetUnitControlGroup")) { + const char *name; + Unit *u; + DBusMessageIter iter; + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_iter_get_basic_and_next(&iter, DBUS_TYPE_STRING, &name, true); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + u = manager_get_unit(m, name); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "stop"); + + r = bus_unit_cgroup_unset(u, &iter); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "SetUnitControlGroupAttribute")) { + const char *name; + Unit *u; + DBusMessageIter iter; + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_iter_get_basic_and_next(&iter, DBUS_TYPE_STRING, &name, true); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + u = manager_get_unit(m, name); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "start"); + + r = bus_unit_cgroup_attribute_set(u, &iter); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "UnsetUnitControlGroupAttribute")) { + const char *name; + Unit *u; + DBusMessageIter iter; + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_iter_get_basic_and_next(&iter, DBUS_TYPE_STRING, &name, true); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + u = manager_get_unit(m, name); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "stop"); + + r = bus_unit_cgroup_attribute_unset(u, &iter); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "GetUnitControlGroupAttribute")) { + const char *name; + Unit *u; + DBusMessageIter iter; + _cleanup_strv_free_ char **list = NULL; + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_iter_get_basic_and_next(&iter, DBUS_TYPE_STRING, &name, true); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + u = manager_get_unit(m, name); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "status"); + + r = bus_unit_cgroup_attribute_get(u, &iter, &list); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + dbus_message_iter_init_append(reply, &iter); + if (bus_append_strv_iter(&iter, list) < 0) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ListUnits")) { + DBusMessageIter iter, sub; + Iterator i; + Unit *u; + const char *k; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + dbus_message_iter_init_append(reply, &iter); + + if (!dbus_message_iter_open_container(&iter, DBUS_TYPE_ARRAY, "(ssssssouso)", &sub)) + goto oom; + + HASHMAP_FOREACH_KEY(u, k, m->units, i) { + char *u_path, *j_path; + const char *description, *load_state, *active_state, *sub_state, *sjob_type, *following; + DBusMessageIter sub2; + uint32_t job_id; + Unit *f; + + if (k != u->id) + continue; + + if (!dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2)) + goto oom; + + description = unit_description(u); + load_state = unit_load_state_to_string(u->load_state); + active_state = unit_active_state_to_string(unit_active_state(u)); + sub_state = unit_sub_state_to_string(u); + + f = unit_following(u); + following = f ? f->id : ""; + + u_path = unit_dbus_path(u); + if (!u_path) + goto oom; + + if (u->job) { + job_id = (uint32_t) u->job->id; + + if (!(j_path = job_dbus_path(u->job))) { + free(u_path); + goto oom; + } + + sjob_type = job_type_to_string(u->job->type); + } else { + job_id = 0; + j_path = u_path; + sjob_type = ""; + } + + if (!dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &u->id) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &description) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &load_state) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &active_state) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &sub_state) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &following) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_OBJECT_PATH, &u_path) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT32, &job_id) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &sjob_type) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_OBJECT_PATH, &j_path)) { + free(u_path); + if (u->job) + free(j_path); + goto oom; + } + + free(u_path); + if (u->job) + free(j_path); + + if (!dbus_message_iter_close_container(&sub, &sub2)) + goto oom; + } + + if (!dbus_message_iter_close_container(&iter, &sub)) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ListJobs")) { + DBusMessageIter iter, sub; + Iterator i; + Job *j; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + dbus_message_iter_init_append(reply, &iter); + + if (!dbus_message_iter_open_container(&iter, DBUS_TYPE_ARRAY, "(usssoo)", &sub)) + goto oom; + + HASHMAP_FOREACH(j, m->jobs, i) { + char *u_path, *j_path; + const char *state, *type; + uint32_t id; + DBusMessageIter sub2; + + if (!dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2)) + goto oom; + + id = (uint32_t) j->id; + state = job_state_to_string(j->state); + type = job_type_to_string(j->type); + + j_path = job_dbus_path(j); + if (!j_path) + goto oom; + + u_path = unit_dbus_path(j->unit); + if (!u_path) { + free(j_path); + goto oom; + } + + if (!dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT32, &id) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &j->unit->id) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &type) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &state) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_OBJECT_PATH, &j_path) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_OBJECT_PATH, &u_path)) { + free(j_path); + free(u_path); + goto oom; + } + + free(j_path); + free(u_path); + + if (!dbus_message_iter_close_container(&sub, &sub2)) + goto oom; + } + + if (!dbus_message_iter_close_container(&iter, &sub)) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "Subscribe")) { + char *client; + Set *s; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + s = BUS_CONNECTION_SUBSCRIBED(m, connection); + if (!s) { + s = set_new(string_hash_func, string_compare_func); + if (!s) + goto oom; + + if (!dbus_connection_set_data(connection, m->subscribed_data_slot, s, NULL)) { + set_free(s); + goto oom; + } + } + + client = strdup(bus_message_get_sender_with_fallback(message)); + if (!client) + goto oom; + + r = set_consume(s, client); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "Unsubscribe")) { + char *client; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + client = set_remove(BUS_CONNECTION_SUBSCRIBED(m, connection), (char*) bus_message_get_sender_with_fallback(message)); + if (!client) { + dbus_set_error(&error, BUS_ERROR_NOT_SUBSCRIBED, "Client is not subscribed."); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + free(client); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "Dump")) { + FILE *f; + char *dump = NULL; + size_t size; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + f = open_memstream(&dump, &size); + if (!f) + goto oom; + + manager_dump_units(m, f, NULL); + manager_dump_jobs(m, f, NULL); + + if (ferror(f)) { + fclose(f); + free(dump); + goto oom; + } + + fclose(f); + + if (!dbus_message_append_args(reply, DBUS_TYPE_STRING, &dump, DBUS_TYPE_INVALID)) { + free(dump); + goto oom; + } + + free(dump); + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "CreateSnapshot")) { + const char *name; + dbus_bool_t cleanup; + Snapshot *s; + + SELINUX_ACCESS_CHECK(connection, message, "start"); + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_BOOLEAN, &cleanup, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + if (isempty(name)) + name = NULL; + + r = snapshot_create(m, name, cleanup, &error, &s); + if (r < 0) + return bus_send_error_reply(connection, message, &error, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + path = unit_dbus_path(UNIT(s)); + if (!path) + goto oom; + + if (!dbus_message_append_args( + reply, + DBUS_TYPE_OBJECT_PATH, &path, + DBUS_TYPE_INVALID)) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "RemoveSnapshot")) { + const char *name; + Unit *u; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + u = manager_get_unit(m, name); + if (!u) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s does not exist.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + if (u->type != UNIT_SNAPSHOT) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not a snapshot.", name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "stop"); + snapshot_remove(SNAPSHOT(u)); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.DBus.Introspectable", "Introspect")) { + char *introspection = NULL; + FILE *f; + Iterator i; + Unit *u; + Job *j; + const char *k; + size_t size; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + /* We roll our own introspection code here, instead of + * relying on bus_default_message_handler() because we + * need to generate our introspection string + * dynamically. */ + + f = open_memstream(&introspection, &size); + if (!f) + goto oom; + + fputs(INTROSPECTION_BEGIN, f); + + HASHMAP_FOREACH_KEY(u, k, m->units, i) { + char *p; + + if (k != u->id) + continue; + + p = bus_path_escape(k); + if (!p) { + fclose(f); + free(introspection); + goto oom; + } + + fprintf(f, "<node name=\"unit/%s\"/>", p); + free(p); + } + + HASHMAP_FOREACH(j, m->jobs, i) + fprintf(f, "<node name=\"job/%lu\"/>", (unsigned long) j->id); + + fputs(INTROSPECTION_END, f); + + if (ferror(f)) { + fclose(f); + free(introspection); + goto oom; + } + + fclose(f); + + if (!introspection) + goto oom; + + if (!dbus_message_append_args(reply, DBUS_TYPE_STRING, &introspection, DBUS_TYPE_INVALID)) { + free(introspection); + goto oom; + } + + free(introspection); + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "Reload")) { + + SELINUX_ACCESS_CHECK(connection, message, "reload"); + + assert(!m->queued_message); + + /* Instead of sending the reply back right away, we + * just remember that we need to and then send it + * after the reload is finished. That way the caller + * knows when the reload finished. */ + + m->queued_message = dbus_message_new_method_return(message); + if (!m->queued_message) + goto oom; + + m->queued_message_connection = connection; + m->exit_code = MANAGER_RELOAD; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "Reexecute")) { + + SELINUX_ACCESS_CHECK(connection, message, "reload"); + + /* We don't send a reply back here, the client should + * just wait for us disconnecting. */ + + m->exit_code = MANAGER_REEXECUTE; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "Exit")) { + + SELINUX_ACCESS_CHECK(connection, message, "halt"); + + if (m->running_as == SYSTEMD_SYSTEM) { + dbus_set_error(&error, BUS_ERROR_NOT_SUPPORTED, "Exit is only supported for user service managers."); + return bus_send_error_reply(connection, message, &error, -ENOTSUP); + } + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + m->exit_code = MANAGER_EXIT; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "Reboot")) { + + SELINUX_ACCESS_CHECK(connection, message, "reboot"); + + if (m->running_as != SYSTEMD_SYSTEM) { + dbus_set_error(&error, BUS_ERROR_NOT_SUPPORTED, "Reboot is only supported for system managers."); + return bus_send_error_reply(connection, message, &error, -ENOTSUP); + } + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + m->exit_code = MANAGER_REBOOT; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "PowerOff")) { + + SELINUX_ACCESS_CHECK(connection, message, "halt"); + + if (m->running_as != SYSTEMD_SYSTEM) { + dbus_set_error(&error, BUS_ERROR_NOT_SUPPORTED, "Powering off is only supported for system managers."); + return bus_send_error_reply(connection, message, &error, -ENOTSUP); + } + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + m->exit_code = MANAGER_POWEROFF; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "Halt")) { + + SELINUX_ACCESS_CHECK(connection, message, "halt"); + + if (m->running_as != SYSTEMD_SYSTEM) { + dbus_set_error(&error, BUS_ERROR_NOT_SUPPORTED, "Halting is only supported for system managers."); + return bus_send_error_reply(connection, message, &error, -ENOTSUP); + } + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + m->exit_code = MANAGER_HALT; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "KExec")) { + + SELINUX_ACCESS_CHECK(connection, message, "reboot"); + + if (m->running_as != SYSTEMD_SYSTEM) { + dbus_set_error(&error, BUS_ERROR_NOT_SUPPORTED, "kexec is only supported for system managers."); + return bus_send_error_reply(connection, message, &error, -ENOTSUP); + } + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + m->exit_code = MANAGER_KEXEC; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "SwitchRoot")) { + const char *switch_root, *switch_root_init; + char *u, *v; + bool good; + + SELINUX_ACCESS_CHECK(connection, message, "reboot"); + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &switch_root, + DBUS_TYPE_STRING, &switch_root_init, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + if (path_equal(switch_root, "/") || !path_is_absolute(switch_root)) + return bus_send_error_reply(connection, message, NULL, -EINVAL); + + if (!isempty(switch_root_init) && !path_is_absolute(switch_root_init)) + return bus_send_error_reply(connection, message, NULL, -EINVAL); + + if (m->running_as != SYSTEMD_SYSTEM) { + dbus_set_error(&error, BUS_ERROR_NOT_SUPPORTED, "Switching root is only supported for system managers."); + return bus_send_error_reply(connection, message, &error, -ENOTSUP); + } + + /* Safety check */ + if (isempty(switch_root_init)) { + good = path_is_os_tree(switch_root); + if (!good) + log_error("Not switching root: %s does not seem to be an OS tree. /etc/os-release is missing.", switch_root); + } + else { + _cleanup_free_ char *p = NULL; + + p = strjoin(switch_root, "/", switch_root_init, NULL); + if (!p) + goto oom; + + good = access(p, X_OK) >= 0; + if (!good) + log_error("Not switching root: cannot execute new init %s", p); + } + if (!good) + return bus_send_error_reply(connection, message, NULL, -EINVAL); + + u = strdup(switch_root); + if (!u) + goto oom; + + if (!isempty(switch_root_init)) { + v = strdup(switch_root_init); + if (!v) { + free(u); + goto oom; + } + } else + v = NULL; + + free(m->switch_root); + free(m->switch_root_init); + m->switch_root = u; + m->switch_root_init = v; + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + m->exit_code = MANAGER_SWITCH_ROOT; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "SetEnvironment")) { + _cleanup_strv_free_ char **l = NULL; + char **e = NULL; + + SELINUX_ACCESS_CHECK(connection, message, "reboot"); + + r = bus_parse_strv(message, &l); + if (r == -ENOMEM) + goto oom; + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + if (!strv_env_is_valid(l)) + return bus_send_error_reply(connection, message, NULL, -EINVAL); + + e = strv_env_merge(2, m->environment, l); + if (!e) + goto oom; + + reply = dbus_message_new_method_return(message); + if (!reply) { + strv_free(e); + goto oom; + } + + strv_free(m->environment); + m->environment = e; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "UnsetEnvironment")) { + _cleanup_strv_free_ char **l = NULL; + char **e = NULL; + + SELINUX_ACCESS_CHECK(connection, message, "reboot"); + + r = bus_parse_strv(message, &l); + if (r == -ENOMEM) + goto oom; + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + if (!strv_env_name_or_assignment_is_valid(l)) + return bus_send_error_reply(connection, message, NULL, -EINVAL); + + e = strv_env_delete(m->environment, 1, l); + if (!e) + goto oom; + + reply = dbus_message_new_method_return(message); + if (!reply) { + strv_free(e); + goto oom; + } + + strv_free(m->environment); + m->environment = e; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "UnsetAndSetEnvironment")) { + _cleanup_strv_free_ char **l_set = NULL, **l_unset = NULL, **e = NULL; + char **f = NULL; + DBusMessageIter iter; + + SELINUX_ACCESS_CHECK(connection, message, "reboot"); + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_parse_strv_iter(&iter, &l_unset); + if (r == -ENOMEM) + goto oom; + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + if (!strv_env_name_or_assignment_is_valid(l_unset)) + return bus_send_error_reply(connection, message, NULL, -EINVAL); + + if (!dbus_message_iter_next(&iter)) + return bus_send_error_reply(connection, message, NULL, -EINVAL); + + r = bus_parse_strv_iter(&iter, &l_set); + if (r == -ENOMEM) + goto oom; + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + if (!strv_env_is_valid(l_set)) + return bus_send_error_reply(connection, message, NULL, -EINVAL); + + e = strv_env_delete(m->environment, 1, l_unset); + if (!e) + goto oom; + + f = strv_env_merge(2, e, l_set); + if (!f) + goto oom; + + reply = dbus_message_new_method_return(message); + if (!reply) { + strv_free(f); + goto oom; + } + + strv_free(m->environment); + m->environment = f; + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ListUnitFiles")) { + DBusMessageIter iter, sub, sub2; + Hashmap *h; + Iterator i; + UnitFileList *item; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + h = hashmap_new(string_hash_func, string_compare_func); + if (!h) + goto oom; + + r = unit_file_get_list(m->running_as == SYSTEMD_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, NULL, h); + if (r < 0) { + unit_file_list_free(h); + return bus_send_error_reply(connection, message, NULL, r); + } + + dbus_message_iter_init_append(reply, &iter); + + if (!dbus_message_iter_open_container(&iter, DBUS_TYPE_ARRAY, "(ss)", &sub)) { + unit_file_list_free(h); + goto oom; + } + + HASHMAP_FOREACH(item, h, i) { + const char *state; + + state = unit_file_state_to_string(item->state); + assert(state); + + if (!dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &item->path) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &state) || + !dbus_message_iter_close_container(&sub, &sub2)) { + unit_file_list_free(h); + goto oom; + } + } + + unit_file_list_free(h); + + if (!dbus_message_iter_close_container(&iter, &sub)) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "GetUnitFileState")) { + const char *name; + UnitFileState state; + const char *s; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + state = unit_file_get_state(m->running_as == SYSTEMD_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, NULL, name); + if (state < 0) + return bus_send_error_reply(connection, message, NULL, state); + + s = unit_file_state_to_string(state); + assert(s); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + if (!dbus_message_append_args( + reply, + DBUS_TYPE_STRING, &s, + DBUS_TYPE_INVALID)) + goto oom; + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "EnableUnitFiles") || + dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "ReenableUnitFiles") || + dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "LinkUnitFiles") || + dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "PresetUnitFiles") || + dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "MaskUnitFiles")) { + + char **l = NULL; + DBusMessageIter iter; + UnitFileScope scope = m->running_as == SYSTEMD_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + dbus_bool_t runtime, force; + int carries_install_info = -1; + + SELINUX_ACCESS_CHECK(connection, message, streq(member, "MaskUnitFiles") ? "disable" : "enable"); + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_parse_strv_iter(&iter, &l); + if (r < 0) { + if (r == -ENOMEM) + goto oom; + + return bus_send_error_reply(connection, message, NULL, r); + } + + if (!dbus_message_iter_next(&iter) || + bus_iter_get_basic_and_next(&iter, DBUS_TYPE_BOOLEAN, &runtime, true) < 0 || + bus_iter_get_basic_and_next(&iter, DBUS_TYPE_BOOLEAN, &force, false) < 0) { + strv_free(l); + return bus_send_error_reply(connection, message, NULL, -EIO); + } + + if (streq(member, "EnableUnitFiles")) { + r = unit_file_enable(scope, runtime, NULL, l, force, &changes, &n_changes); + carries_install_info = r; + } else if (streq(member, "ReenableUnitFiles")) { + r = unit_file_reenable(scope, runtime, NULL, l, force, &changes, &n_changes); + carries_install_info = r; + } else if (streq(member, "LinkUnitFiles")) + r = unit_file_link(scope, runtime, NULL, l, force, &changes, &n_changes); + else if (streq(member, "PresetUnitFiles")) { + r = unit_file_preset(scope, runtime, NULL, l, force, &changes, &n_changes); + carries_install_info = r; + } else if (streq(member, "MaskUnitFiles")) + r = unit_file_mask(scope, runtime, NULL, l, force, &changes, &n_changes); + else + assert_not_reached("Uh? Wrong method"); + + strv_free(l); + bus_manager_send_unit_files_changed(m); + + if (r < 0) { + unit_file_changes_free(changes, n_changes); + return bus_send_error_reply(connection, message, NULL, r); + } + + reply = message_from_file_changes(message, changes, n_changes, carries_install_info); + unit_file_changes_free(changes, n_changes); + + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "DisableUnitFiles") || + dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "UnmaskUnitFiles")) { + + char **l = NULL; + DBusMessageIter iter; + UnitFileScope scope = m->running_as == SYSTEMD_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + dbus_bool_t runtime; + + SELINUX_ACCESS_CHECK(connection, message, streq(member, "UnmaskUnitFiles") ? "enable" : "disable"); + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_parse_strv_iter(&iter, &l); + if (r < 0) { + if (r == -ENOMEM) + goto oom; + + return bus_send_error_reply(connection, message, NULL, r); + } + + if (!dbus_message_iter_next(&iter) || + bus_iter_get_basic_and_next(&iter, DBUS_TYPE_BOOLEAN, &runtime, false) < 0) { + strv_free(l); + return bus_send_error_reply(connection, message, NULL, -EIO); + } + + if (streq(member, "DisableUnitFiles")) + r = unit_file_disable(scope, runtime, NULL, l, &changes, &n_changes); + else if (streq(member, "UnmaskUnitFiles")) + r = unit_file_unmask(scope, runtime, NULL, l, &changes, &n_changes); + else + assert_not_reached("Uh? Wrong method"); + + strv_free(l); + bus_manager_send_unit_files_changed(m); + + if (r < 0) { + unit_file_changes_free(changes, n_changes); + return bus_send_error_reply(connection, message, NULL, r); + } + + reply = message_from_file_changes(message, changes, n_changes, -1); + unit_file_changes_free(changes, n_changes); + + if (!reply) + goto oom; + + } else { + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Manager", bus_systemd_properties, systemd_property_string }, + { "org.freedesktop.systemd1.Manager", bus_manager_properties, m }, + { NULL, } + }; + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + return bus_default_message_handler(connection, message, NULL, INTERFACES_LIST, bps); + } + + if (job_type != _JOB_TYPE_INVALID) { + const char *name, *smode, *old_name = NULL; + JobMode mode; + Unit *u; + dbus_bool_t b; + + if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Manager", "StartUnitReplace")) + b = dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &old_name, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_STRING, &smode, + DBUS_TYPE_INVALID); + else + b = dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_STRING, &smode, + DBUS_TYPE_INVALID); + if (!b) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + if (old_name) { + u = manager_get_unit(m, old_name); + if (!u || !u->job || u->job->type != JOB_START) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_JOB, "No job queued for unit %s", old_name); + return bus_send_error_reply(connection, message, &error, -ENOENT); + } + } + + mode = job_mode_from_string(smode); + if (mode < 0) { + dbus_set_error(&error, BUS_ERROR_INVALID_JOB_MODE, "Job mode %s is invalid.", smode); + return bus_send_error_reply(connection, message, &error, -EINVAL); + } + + r = manager_load_unit(m, name, NULL, &error, &u); + if (r < 0) + return bus_send_error_reply(connection, message, &error, r); + + return bus_unit_queue_job(connection, message, u, job_type, mode, reload_if_possible); + } + + if (reply) + if (!bus_maybe_send_reply(connection, message, reply)) + goto oom; + + return DBUS_HANDLER_RESULT_HANDLED; + +oom: + dbus_error_free(&error); + + return DBUS_HANDLER_RESULT_NEED_MEMORY; +} + +const DBusObjectPathVTable bus_manager_vtable = { + .message_function = bus_manager_message_handler +}; diff --git a/src/core/dbus-manager.h b/src/core/dbus-manager.h new file mode 100644 index 0000000000..f0dce5a2e9 --- /dev/null +++ b/src/core/dbus-manager.h @@ -0,0 +1,28 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +extern const DBusObjectPathVTable bus_manager_vtable; + +extern const char bus_manager_interface[]; diff --git a/src/core/dbus-mount.c b/src/core/dbus-mount.c new file mode 100644 index 0000000000..0fcceb500d --- /dev/null +++ b/src/core/dbus-mount.c @@ -0,0 +1,170 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus-unit.h" +#include "dbus-mount.h" +#include "dbus-kill.h" +#include "dbus-execute.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_MOUNT_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Mount\">\n" \ + " <property name=\"Where\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"What\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Options\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Type\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"TimeoutUSec\" type=\"t\" access=\"read\"/>\n" \ + BUS_EXEC_COMMAND_INTERFACE("ExecMount") \ + BUS_EXEC_COMMAND_INTERFACE("ExecUnmount") \ + BUS_EXEC_COMMAND_INTERFACE("ExecRemount") \ + BUS_EXEC_CONTEXT_INTERFACE \ + BUS_KILL_CONTEXT_INTERFACE \ + BUS_UNIT_CGROUP_INTERFACE \ + " <property name=\"ControlPID\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"DirectoryMode\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"Result\" type=\"s\" access=\"read\"/>\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_MOUNT_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Mount\0" + +const char bus_mount_interface[] _introspect_("Mount") = BUS_MOUNT_INTERFACE; + +const char bus_mount_invalidating_properties[] = + "What\0" + "Options\0" + "Type\0" + "ExecMount\0" + "ExecUnmount\0" + "ExecRemount\0" + "ControlPID\0" + "Result\0"; + +static int bus_mount_append_what(DBusMessageIter *i, const char *property, void *data) { + Mount *m = data; + const char *d; + + assert(i); + assert(property); + assert(m); + + if (m->from_proc_self_mountinfo && m->parameters_proc_self_mountinfo.what) + d = m->parameters_proc_self_mountinfo.what; + else if (m->from_fragment && m->parameters_fragment.what) + d = m->parameters_fragment.what; + else + d = ""; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &d)) + return -ENOMEM; + + return 0; +} + +static int bus_mount_append_options(DBusMessageIter *i, const char *property, void *data) { + Mount *m = data; + const char *d; + + assert(i); + assert(property); + assert(m); + + if (m->from_proc_self_mountinfo && m->parameters_proc_self_mountinfo.options) + d = m->parameters_proc_self_mountinfo.options; + else if (m->from_fragment && m->parameters_fragment.options) + d = m->parameters_fragment.options; + else + d = ""; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &d)) + return -ENOMEM; + + return 0; +} + +static int bus_mount_append_type(DBusMessageIter *i, const char *property, void *data) { + Mount *m = data; + const char *d; + + assert(i); + assert(property); + assert(m); + + if (m->from_proc_self_mountinfo && m->parameters_proc_self_mountinfo.fstype) + d = m->parameters_proc_self_mountinfo.fstype; + else if (m->from_fragment && m->parameters_fragment.fstype) + d = m->parameters_fragment.fstype; + else + d = ""; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &d)) + return -ENOMEM; + + return 0; +} + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_mount_append_mount_result, mount_result, MountResult); + +static const BusProperty bus_mount_properties[] = { + { "Where", bus_property_append_string, "s", offsetof(Mount, where), true }, + { "What", bus_mount_append_what, "s", 0 }, + { "Options", bus_mount_append_options, "s", 0 }, + { "Type", bus_mount_append_type, "s", 0 }, + { "TimeoutUSec", bus_property_append_usec, "t", offsetof(Mount, timeout_usec) }, + BUS_EXEC_COMMAND_PROPERTY("ExecMount", offsetof(Mount, exec_command[MOUNT_EXEC_MOUNT]), false), + BUS_EXEC_COMMAND_PROPERTY("ExecUnmount", offsetof(Mount, exec_command[MOUNT_EXEC_UNMOUNT]), false), + BUS_EXEC_COMMAND_PROPERTY("ExecRemount", offsetof(Mount, exec_command[MOUNT_EXEC_REMOUNT]), false), + { "ControlPID", bus_property_append_pid, "u", offsetof(Mount, control_pid) }, + { "DirectoryMode", bus_property_append_mode, "u", offsetof(Mount, directory_mode) }, + { "Result", bus_mount_append_mount_result, "s", offsetof(Mount, result) }, + { NULL, } +}; + +DBusHandlerResult bus_mount_message_handler(Unit *u, DBusConnection *c, DBusMessage *message) { + Mount *m = MOUNT(u); + + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { "org.freedesktop.systemd1.Mount", bus_mount_properties, m }, + { "org.freedesktop.systemd1.Mount", bus_exec_context_properties, &m->exec_context }, + { "org.freedesktop.systemd1.Mount", bus_kill_context_properties, &m->kill_context }, + { "org.freedesktop.systemd1.Mount", bus_unit_cgroup_properties, u }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "status"); + + return bus_default_message_handler(c, message, INTROSPECTION, INTERFACES_LIST, bps ); +} diff --git a/src/core/dbus-mount.h b/src/core/dbus-mount.h new file mode 100644 index 0000000000..8597394373 --- /dev/null +++ b/src/core/dbus-mount.h @@ -0,0 +1,31 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_mount_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_mount_interface[]; +extern const char bus_mount_invalidating_properties[]; diff --git a/src/core/dbus-path.c b/src/core/dbus-path.c new file mode 100644 index 0000000000..1e62083d9b --- /dev/null +++ b/src/core/dbus-path.c @@ -0,0 +1,122 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus-unit.h" +#include "dbus-path.h" +#include "dbus-execute.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_PATH_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Path\">\n" \ + " <property name=\"Unit\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Paths\" type=\"a(ss)\" access=\"read\"/>\n" \ + " <property name=\"MakeDirectory\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"DirectoryMode\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"Result\" type=\"s\" access=\"read\"/>\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_PATH_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Path\0" + +const char bus_path_interface[] _introspect_("Path") = BUS_PATH_INTERFACE; + +const char bus_path_invalidating_properties[] = + "Result\0"; + +static int bus_path_append_paths(DBusMessageIter *i, const char *property, void *data) { + Path *p = data; + DBusMessageIter sub, sub2; + PathSpec *k; + + assert(i); + assert(property); + assert(p); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "(ss)", &sub)) + return -ENOMEM; + + LIST_FOREACH(spec, k, p->specs) { + const char *t = path_type_to_string(k->type); + + if (!dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &t) || + !dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &k->path) || + !dbus_message_iter_close_container(&sub, &sub2)) + return -ENOMEM; + } + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static int bus_path_append_unit(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data, *trigger; + const char *t; + + assert(i); + assert(property); + assert(u); + + trigger = UNIT_TRIGGER(u); + t = trigger ? trigger->id : ""; + + return dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &t) ? 0 : -ENOMEM; +} + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_path_append_path_result, path_result, PathResult); + +static const BusProperty bus_path_properties[] = { + { "Unit", bus_path_append_unit, "s", 0 }, + { "Paths", bus_path_append_paths, "a(ss)", 0 }, + { "MakeDirectory", bus_property_append_bool, "b", offsetof(Path, make_directory) }, + { "DirectoryMode", bus_property_append_mode, "u", offsetof(Path, directory_mode) }, + { "Result", bus_path_append_path_result, "s", offsetof(Path, result) }, + { NULL, } +}; + +DBusHandlerResult bus_path_message_handler(Unit *u, DBusConnection *c, DBusMessage *message) { + Path *p = PATH(u); + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { "org.freedesktop.systemd1.Path", bus_path_properties, p }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "status"); + + return bus_default_message_handler(c, message, INTROSPECTION, INTERFACES_LIST, bps); +} diff --git a/src/core/dbus-path.h b/src/core/dbus-path.h new file mode 100644 index 0000000000..c945f7d588 --- /dev/null +++ b/src/core/dbus-path.h @@ -0,0 +1,32 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_path_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_path_interface[]; + +extern const char bus_path_invalidating_properties[]; diff --git a/src/core/dbus-service.c b/src/core/dbus-service.c new file mode 100644 index 0000000000..e06a5dce97 --- /dev/null +++ b/src/core/dbus-service.c @@ -0,0 +1,163 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus-unit.h" +#include "dbus-execute.h" +#include "dbus-kill.h" +#include "dbus-service.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_SERVICE_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Service\">\n" \ + " <property name=\"Type\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Restart\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"PIDFile\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"NotifyAccess\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"RestartUSec\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"TimeoutUSec\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"WatchdogUSec\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"WatchdogTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"WatchdogTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"StartLimitInterval\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"StartLimitBurst\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"StartLimitAction\" type=\"s\" access=\"readwrite\"/>\n" \ + BUS_EXEC_COMMAND_INTERFACE("ExecStartPre") \ + BUS_EXEC_COMMAND_INTERFACE("ExecStart") \ + BUS_EXEC_COMMAND_INTERFACE("ExecStartPost") \ + BUS_EXEC_COMMAND_INTERFACE("ExecReload") \ + BUS_EXEC_COMMAND_INTERFACE("ExecStop") \ + BUS_EXEC_COMMAND_INTERFACE("ExecStopPost") \ + BUS_EXEC_CONTEXT_INTERFACE \ + BUS_KILL_CONTEXT_INTERFACE \ + BUS_UNIT_CGROUP_INTERFACE \ + " <property name=\"PermissionsStartOnly\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"RootDirectoryStartOnly\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"RemainAfterExit\" type=\"b\" access=\"read\"/>\n" \ + BUS_EXEC_STATUS_INTERFACE("ExecMain") \ + " <property name=\"MainPID\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"ControlPID\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"BusName\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"StatusText\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Result\" type=\"s\" access=\"read\"/>\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_SERVICE_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Service\0" + +const char bus_service_interface[] _introspect_("Service") = BUS_SERVICE_INTERFACE; + +const char bus_service_invalidating_properties[] = + "ExecStartPre\0" + "ExecStart\0" + "ExecStartPost\0" + "ExecReload\0" + "ExecStop\0" + "ExecStopPost\0" + "ExecMain\0" + "WatchdogTimestamp\0" + "WatchdogTimestampMonotonic\0" + "MainPID\0" + "ControlPID\0" + "StatusText\0" + "Result\0"; + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_service_append_type, service_type, ServiceType); +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_service_append_restart, service_restart, ServiceRestart); +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_service_append_notify_access, notify_access, NotifyAccess); +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_service_append_service_result, service_result, ServiceResult); +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_service_append_start_limit_action, start_limit_action, StartLimitAction); +static DEFINE_BUS_PROPERTY_SET_ENUM(bus_service_set_start_limit_action, start_limit_action, StartLimitAction); + +static const BusProperty bus_exec_main_status_properties[] = { + { "ExecMainStartTimestamp", bus_property_append_usec, "t", offsetof(ExecStatus, start_timestamp.realtime) }, + { "ExecMainStartTimestampMonotonic",bus_property_append_usec, "t", offsetof(ExecStatus, start_timestamp.monotonic) }, + { "ExecMainExitTimestamp", bus_property_append_usec, "t", offsetof(ExecStatus, start_timestamp.realtime) }, + { "ExecMainExitTimestampMonotonic", bus_property_append_usec, "t", offsetof(ExecStatus, start_timestamp.monotonic) }, + { "ExecMainPID", bus_property_append_pid, "u", offsetof(ExecStatus, pid) }, + { "ExecMainCode", bus_property_append_int, "i", offsetof(ExecStatus, code) }, + { "ExecMainStatus", bus_property_append_int, "i", offsetof(ExecStatus, status) }, + { NULL, } +}; + +static const BusProperty bus_service_properties[] = { + { "Type", bus_service_append_type, "s", offsetof(Service, type) }, + { "Restart", bus_service_append_restart, "s", offsetof(Service, restart) }, + { "PIDFile", bus_property_append_string, "s", offsetof(Service, pid_file), true }, + { "NotifyAccess", bus_service_append_notify_access, "s", offsetof(Service, notify_access) }, + { "RestartUSec", bus_property_append_usec, "t", offsetof(Service, restart_usec) }, + { "TimeoutUSec", bus_property_append_usec, "t", offsetof(Service, timeout_start_usec) }, + { "TimeoutStartUSec", bus_property_append_usec, "t", offsetof(Service, timeout_start_usec) }, + { "TimeoutStopUSec", bus_property_append_usec, "t", offsetof(Service, timeout_stop_usec) }, + { "WatchdogUSec", bus_property_append_usec, "t", offsetof(Service, watchdog_usec) }, + { "WatchdogTimestamp", bus_property_append_usec, "t", offsetof(Service, watchdog_timestamp.realtime) }, + { "WatchdogTimestampMonotonic",bus_property_append_usec, "t", offsetof(Service, watchdog_timestamp.monotonic) }, + { "StartLimitInterval", bus_property_append_usec, "t", offsetof(Service, start_limit.interval) }, + { "StartLimitBurst", bus_property_append_uint32, "u", offsetof(Service, start_limit.burst) }, + { "StartLimitAction", bus_service_append_start_limit_action,"s", offsetof(Service, start_limit_action), false, bus_service_set_start_limit_action}, + BUS_EXEC_COMMAND_PROPERTY("ExecStartPre", offsetof(Service, exec_command[SERVICE_EXEC_START_PRE]), true ), + BUS_EXEC_COMMAND_PROPERTY("ExecStart", offsetof(Service, exec_command[SERVICE_EXEC_START]), true ), + BUS_EXEC_COMMAND_PROPERTY("ExecStartPost", offsetof(Service, exec_command[SERVICE_EXEC_START_POST]), true ), + BUS_EXEC_COMMAND_PROPERTY("ExecReload", offsetof(Service, exec_command[SERVICE_EXEC_RELOAD]), true ), + BUS_EXEC_COMMAND_PROPERTY("ExecStop", offsetof(Service, exec_command[SERVICE_EXEC_STOP]), true ), + BUS_EXEC_COMMAND_PROPERTY("ExecStopPost", offsetof(Service, exec_command[SERVICE_EXEC_STOP_POST]), true ), + { "PermissionsStartOnly", bus_property_append_bool, "b", offsetof(Service, permissions_start_only) }, + { "RootDirectoryStartOnly", bus_property_append_bool, "b", offsetof(Service, root_directory_start_only) }, + { "RemainAfterExit", bus_property_append_bool, "b", offsetof(Service, remain_after_exit) }, + { "GuessMainPID", bus_property_append_bool, "b", offsetof(Service, guess_main_pid) }, + { "MainPID", bus_property_append_pid, "u", offsetof(Service, main_pid) }, + { "ControlPID", bus_property_append_pid, "u", offsetof(Service, control_pid) }, + { "BusName", bus_property_append_string, "s", offsetof(Service, bus_name), true }, + { "StatusText", bus_property_append_string, "s", offsetof(Service, status_text), true }, + { "Result", bus_service_append_service_result,"s", offsetof(Service, result) }, + { NULL, } +}; + +DBusHandlerResult bus_service_message_handler(Unit *u, DBusConnection *connection, DBusMessage *message) { + Service *s = SERVICE(u); + + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { "org.freedesktop.systemd1.Service", bus_service_properties, s }, + { "org.freedesktop.systemd1.Service", bus_exec_context_properties, &s->exec_context }, + { "org.freedesktop.systemd1.Service", bus_kill_context_properties, &s->kill_context }, + { "org.freedesktop.systemd1.Service", bus_exec_main_status_properties, &s->main_exec_status }, + { "org.freedesktop.systemd1.Service", bus_unit_cgroup_properties, u }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "status"); + + return bus_default_message_handler(connection, message, INTROSPECTION, INTERFACES_LIST, bps); +} diff --git a/src/core/dbus-service.h b/src/core/dbus-service.h new file mode 100644 index 0000000000..143aed7ae5 --- /dev/null +++ b/src/core/dbus-service.h @@ -0,0 +1,31 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_service_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_service_interface[]; +extern const char bus_service_invalidating_properties[]; diff --git a/src/core/dbus-snapshot.c b/src/core/dbus-snapshot.c new file mode 100644 index 0000000000..2ae8574f59 --- /dev/null +++ b/src/core/dbus-snapshot.c @@ -0,0 +1,84 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "dbus-unit.h" +#include "dbus-snapshot.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_SNAPSHOT_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Snapshot\">\n" \ + " <method name=\"Remove\"/>\n" \ + " <property name=\"Cleanup\" type=\"b\" access=\"read\"/>\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_SNAPSHOT_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Snapshot\0" + +const char bus_snapshot_interface[] _introspect_("Snapshot") = BUS_SNAPSHOT_INTERFACE; + +static const BusProperty bus_snapshot_properties[] = { + { "Cleanup", bus_property_append_bool, "b", offsetof(Snapshot, cleanup) }, + { NULL, } +}; + +DBusHandlerResult bus_snapshot_message_handler(Unit *u, DBusConnection *c, DBusMessage *message) { + Snapshot *s = SNAPSHOT(u); + _cleanup_dbus_message_unref_ DBusMessage *reply = NULL; + + if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Snapshot", "Remove")) { + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "stop"); + + reply = dbus_message_new_method_return(message); + if (!reply) + return DBUS_HANDLER_RESULT_NEED_MEMORY; + + snapshot_remove(SNAPSHOT(u)); + + } else { + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { "org.freedesktop.systemd1.Snapshot", bus_snapshot_properties, s }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "status"); + + return bus_default_message_handler(c, message, INTROSPECTION, INTERFACES_LIST, bps); + } + + if (!bus_maybe_send_reply(c, message, reply)) + return DBUS_HANDLER_RESULT_NEED_MEMORY; + + return DBUS_HANDLER_RESULT_HANDLED; +} diff --git a/src/core/dbus-snapshot.h b/src/core/dbus-snapshot.h new file mode 100644 index 0000000000..1208aafff6 --- /dev/null +++ b/src/core/dbus-snapshot.h @@ -0,0 +1,30 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_snapshot_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_snapshot_interface[]; diff --git a/src/core/dbus-socket.c b/src/core/dbus-socket.c new file mode 100644 index 0000000000..77d98ea0fd --- /dev/null +++ b/src/core/dbus-socket.c @@ -0,0 +1,215 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus-unit.h" +#include "dbus-socket.h" +#include "dbus-execute.h" +#include "dbus-kill.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_SOCKET_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Socket\">\n" \ + " <property name=\"BindIPv6Only\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"Backlog\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"TimeoutUSec\" type=\"t\" access=\"read\"/>\n" \ + BUS_EXEC_COMMAND_INTERFACE("ExecStartPre") \ + BUS_EXEC_COMMAND_INTERFACE("ExecStartPost") \ + BUS_EXEC_COMMAND_INTERFACE("ExecStopPre") \ + BUS_EXEC_COMMAND_INTERFACE("ExecStopPost") \ + BUS_EXEC_CONTEXT_INTERFACE \ + BUS_KILL_CONTEXT_INTERFACE \ + BUS_UNIT_CGROUP_INTERFACE \ + " <property name=\"ControlPID\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"BindToDevice\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"DirectoryMode\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"SocketMode\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"Accept\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"KeepAlive\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"Priority\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"ReceiveBuffer\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"SendBuffer\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"IPTOS\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"IPTTL\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"PipeSize\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"FreeBind\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"Transparent\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"Broadcast\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"PassCredentials\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"PassSecurity\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"Mark\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"MaxConnections\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"NAccepted\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"NConnections\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"MessageQueueMaxMessages\" type=\"x\" access=\"read\"/>\n" \ + " <property name=\"MessageQueueMessageSize\" type=\"x\" access=\"read\"/>\n" \ + " <property name=\"Listen\" type=\"a(ss)\" access=\"read\"/>\n" \ + " <property name=\"Result\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"SmackLabel\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"SmackLabelIPIn\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"SmackLabelIPOut\" type=\"s\" access=\"read\"/>\n" \ + " </interface>\n" \ + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_SOCKET_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Socket\0" + +const char bus_socket_interface[] _introspect_("Socket") = BUS_SOCKET_INTERFACE; + +const char bus_socket_invalidating_properties[] = + "ExecStartPre\0" + "ExecStartPost\0" + "ExecStopPre\0" + "ExecStopPost\0" + "ControlPID\0" + "NAccepted\0" + "NConnections\0" + "Result\0"; + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_socket_append_bind_ipv6_only, socket_address_bind_ipv6_only, SocketAddressBindIPv6Only); +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_socket_append_socket_result, socket_result, SocketResult); + +static int bus_socket_append_listen(DBusMessageIter *i, const char *property, void *data) { + + Socket *s = SOCKET(data); + SocketPort *p; + DBusMessageIter array, stru; + + assert(data); + assert(property); + assert(s); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "(ss)", &array)) + return log_oom(); + + LIST_FOREACH(port, p, s->ports) { + const char *type = socket_port_type_to_string(p); + _cleanup_free_ char *address = NULL; + const char *a; + + if (!dbus_message_iter_open_container(&array, DBUS_TYPE_STRUCT, NULL, &stru)) + return log_oom(); + + if (!dbus_message_iter_append_basic(&stru, DBUS_TYPE_STRING, &type)) + return log_oom(); + + switch (p->type) { + case SOCKET_SOCKET: { + int r; + + r = socket_address_print(&p->address, &address); + if (r) { + log_error("socket_address_print failed: %s", strerror(-r)); + return r; + } + a = address; + break; + } + + case SOCKET_SPECIAL: + case SOCKET_MQUEUE: + case SOCKET_FIFO: + a = p->path; + break; + + default: + a = type; + } + + if (!dbus_message_iter_append_basic(&stru, DBUS_TYPE_STRING, &a)) + return -ENOMEM; + + if (!dbus_message_iter_close_container(&array, &stru)) + return -ENOMEM; + } + + if (!dbus_message_iter_close_container(i, &array)) + return -ENOMEM; + + return 0; +} + +static const BusProperty bus_socket_properties[] = { + { "BindIPv6Only", bus_socket_append_bind_ipv6_only, "s", offsetof(Socket, bind_ipv6_only) }, + { "Backlog", bus_property_append_unsigned, "u", offsetof(Socket, backlog) }, + { "TimeoutUSec", bus_property_append_usec, "t", offsetof(Socket, timeout_usec) }, + BUS_EXEC_COMMAND_PROPERTY("ExecStartPre", offsetof(Socket, exec_command[SOCKET_EXEC_START_PRE]), true ), + BUS_EXEC_COMMAND_PROPERTY("ExecStartPost", offsetof(Socket, exec_command[SOCKET_EXEC_START_POST]), true ), + BUS_EXEC_COMMAND_PROPERTY("ExecStopPre", offsetof(Socket, exec_command[SOCKET_EXEC_STOP_PRE]), true ), + BUS_EXEC_COMMAND_PROPERTY("ExecStopPost", offsetof(Socket, exec_command[SOCKET_EXEC_STOP_POST]), true ), + { "ControlPID", bus_property_append_pid, "u", offsetof(Socket, control_pid) }, + { "BindToDevice", bus_property_append_string, "s", offsetof(Socket, bind_to_device), true }, + { "DirectoryMode", bus_property_append_mode, "u", offsetof(Socket, directory_mode) }, + { "SocketMode", bus_property_append_mode, "u", offsetof(Socket, socket_mode) }, + { "Accept", bus_property_append_bool, "b", offsetof(Socket, accept) }, + { "KeepAlive", bus_property_append_bool, "b", offsetof(Socket, keep_alive) }, + { "Priority", bus_property_append_int, "i", offsetof(Socket, priority) }, + { "ReceiveBuffer", bus_property_append_size, "t", offsetof(Socket, receive_buffer) }, + { "SendBuffer", bus_property_append_size, "t", offsetof(Socket, send_buffer) }, + { "IPTOS", bus_property_append_int, "i", offsetof(Socket, ip_tos) }, + { "IPTTL", bus_property_append_int, "i", offsetof(Socket, ip_ttl) }, + { "PipeSize", bus_property_append_size, "t", offsetof(Socket, pipe_size) }, + { "FreeBind", bus_property_append_bool, "b", offsetof(Socket, free_bind) }, + { "Transparent", bus_property_append_bool, "b", offsetof(Socket, transparent) }, + { "Broadcast", bus_property_append_bool, "b", offsetof(Socket, broadcast) }, + { "PassCredentials",bus_property_append_bool, "b", offsetof(Socket, pass_cred) }, + { "PassSecurity", bus_property_append_bool, "b", offsetof(Socket, pass_sec) }, + { "Listen", bus_socket_append_listen, "a(ss)", 0, }, + { "Mark", bus_property_append_int, "i", offsetof(Socket, mark) }, + { "MaxConnections", bus_property_append_unsigned, "u", offsetof(Socket, max_connections) }, + { "NConnections", bus_property_append_unsigned, "u", offsetof(Socket, n_connections) }, + { "NAccepted", bus_property_append_unsigned, "u", offsetof(Socket, n_accepted) }, + { "MessageQueueMaxMessages", bus_property_append_long, "x", offsetof(Socket, mq_maxmsg) }, + { "MessageQueueMessageSize", bus_property_append_long, "x", offsetof(Socket, mq_msgsize) }, + { "Result", bus_socket_append_socket_result, "s", offsetof(Socket, result) }, + { "SmackLabel", bus_property_append_string, "s", offsetof(Socket, smack), true }, + { "SmackLabelIPIn", bus_property_append_string, "s", offsetof(Socket, smack_ip_in), true }, + { "SmackLabelIPOut",bus_property_append_string, "s", offsetof(Socket, smack_ip_out), true }, + { NULL, } +}; + +DBusHandlerResult bus_socket_message_handler(Unit *u, DBusConnection *c, DBusMessage *message) { + Socket *s = SOCKET(u); + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { "org.freedesktop.systemd1.Socket", bus_socket_properties, s }, + { "org.freedesktop.systemd1.Socket", bus_exec_context_properties, &s->exec_context }, + { "org.freedesktop.systemd1.Socket", bus_kill_context_properties, &s->kill_context }, + { "org.freedesktop.systemd1.Socket", bus_unit_properties, u }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "status"); + + return bus_default_message_handler(c, message, INTROSPECTION, INTERFACES_LIST, bps); +} diff --git a/src/core/dbus-socket.h b/src/core/dbus-socket.h new file mode 100644 index 0000000000..5369b22e5e --- /dev/null +++ b/src/core/dbus-socket.h @@ -0,0 +1,31 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_socket_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_socket_interface[]; +extern const char bus_socket_invalidating_properties[]; diff --git a/src/core/dbus-swap.c b/src/core/dbus-swap.c new file mode 100644 index 0000000000..2e99fba7db --- /dev/null +++ b/src/core/dbus-swap.c @@ -0,0 +1,117 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2010 Maarten Lankhorst + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus-unit.h" +#include "dbus-swap.h" +#include "dbus-execute.h" +#include "dbus-kill.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_SWAP_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Swap\">\n" \ + " <property name=\"What\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Priority\" type=\"i\" access=\"read\"/>\n" \ + " <property name=\"TimeoutUSec\" type=\"t\" access=\"read\"/>\n" \ + BUS_EXEC_COMMAND_INTERFACE("ExecActivate") \ + BUS_EXEC_COMMAND_INTERFACE("ExecDeactivate") \ + BUS_EXEC_CONTEXT_INTERFACE \ + BUS_KILL_CONTEXT_INTERFACE \ + BUS_UNIT_CGROUP_INTERFACE \ + " <property name=\"ControlPID\" type=\"u\" access=\"read\"/>\n" \ + " <property name=\"Result\" type=\"s\" access=\"read\"/>\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_SWAP_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Swap\0" + +const char bus_swap_interface[] _introspect_("Swap") = BUS_SWAP_INTERFACE; + +const char bus_swap_invalidating_properties[] = + "What\0" + "Priority\0" + "ExecActivate\0" + "ExecDeactivate\0" + "ControlPID\0" + "Result\0"; + +static int bus_swap_append_priority(DBusMessageIter *i, const char *property, void *data) { + Swap *s = data; + dbus_int32_t j; + + assert(i); + assert(property); + assert(s); + + if (s->from_proc_swaps) + j = s->parameters_proc_swaps.priority; + else if (s->from_fragment) + j = s->parameters_fragment.priority; + else + j = -1; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_INT32, &j)) + return -ENOMEM; + + return 0; +} + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_swap_append_swap_result, swap_result, SwapResult); + +static const BusProperty bus_swap_properties[] = { + { "What", bus_property_append_string, "s", offsetof(Swap, what), true }, + { "Priority", bus_swap_append_priority, "i", 0 }, + BUS_EXEC_COMMAND_PROPERTY("ExecActivate", offsetof(Swap, exec_command[SWAP_EXEC_ACTIVATE]), false), + BUS_EXEC_COMMAND_PROPERTY("ExecDeactivate", offsetof(Swap, exec_command[SWAP_EXEC_DEACTIVATE]), false), + { "ControlPID", bus_property_append_pid, "u", offsetof(Swap, control_pid) }, + { "Result", bus_swap_append_swap_result,"s", offsetof(Swap, result) }, + { NULL, } +}; + +DBusHandlerResult bus_swap_message_handler(Unit *u, DBusConnection *c, DBusMessage *message) { + Swap *s = SWAP(u); + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { "org.freedesktop.systemd1.Swap", bus_swap_properties, s }, + { "org.freedesktop.systemd1.Swap", bus_exec_context_properties, &s->exec_context }, + { "org.freedesktop.systemd1.Swap", bus_kill_context_properties, &s->kill_context }, + { "org.freedesktop.systemd1.Swap", bus_unit_cgroup_properties, u }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "status"); + + return bus_default_message_handler(c, message, INTROSPECTION, INTERFACES_LIST, bps); +} diff --git a/src/core/dbus-swap.h b/src/core/dbus-swap.h new file mode 100644 index 0000000000..41fe4447ff --- /dev/null +++ b/src/core/dbus-swap.h @@ -0,0 +1,32 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2010 Maarten Lankhorst + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_swap_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_swap_interface[]; +extern const char bus_swap_invalidating_properties[]; diff --git a/src/core/dbus-target.c b/src/core/dbus-target.c new file mode 100644 index 0000000000..6a775506cc --- /dev/null +++ b/src/core/dbus-target.c @@ -0,0 +1,58 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus-unit.h" +#include "dbus-target.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_TARGET_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Target\">\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_TARGET_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Target\0" + +const char bus_target_interface[] _introspect_("Target") = BUS_TARGET_INTERFACE; + +DBusHandlerResult bus_target_message_handler(Unit *u, DBusConnection *c, DBusMessage *message) { + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "status"); + + return bus_default_message_handler(c, message, INTROSPECTION, INTERFACES_LIST, bps); +} diff --git a/src/core/dbus-target.h b/src/core/dbus-target.h new file mode 100644 index 0000000000..a8a0304c75 --- /dev/null +++ b/src/core/dbus-target.h @@ -0,0 +1,30 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_target_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_target_interface[]; diff --git a/src/core/dbus-timer.c b/src/core/dbus-timer.c new file mode 100644 index 0000000000..4082f7f9b9 --- /dev/null +++ b/src/core/dbus-timer.c @@ -0,0 +1,191 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus-unit.h" +#include "dbus-timer.h" +#include "dbus-execute.h" +#include "dbus-common.h" +#include "selinux-access.h" + +#define BUS_TIMER_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Timer\">\n" \ + " <property name=\"Unit\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"TimersMonotonic\" type=\"a(stt)\" access=\"read\"/>\n" \ + " <property name=\"TimersCalendar\" type=\"a(sst)\" access=\"read\"/>\n" \ + " <property name=\"NextElapseUSecRealtime\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"NextElapseUSecMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"Result\" type=\"s\" access=\"read\"/>\n" \ + " </interface>\n" + +#define INTROSPECTION \ + DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE \ + "<node>\n" \ + BUS_UNIT_INTERFACE \ + BUS_TIMER_INTERFACE \ + BUS_PROPERTIES_INTERFACE \ + BUS_PEER_INTERFACE \ + BUS_INTROSPECTABLE_INTERFACE \ + "</node>\n" + +#define INTERFACES_LIST \ + BUS_UNIT_INTERFACES_LIST \ + "org.freedesktop.systemd1.Timer\0" + +const char bus_timer_interface[] _introspect_("Timer") = BUS_TIMER_INTERFACE; + +const char bus_timer_invalidating_properties[] = + "TimersMonotonic\0" + "TimersRealtime\0" + "NextElapseUSecRealtime\0" + "NextElapseUSecMonotonic\0" + "Result\0"; + +static int bus_timer_append_monotonic_timers(DBusMessageIter *i, const char *property, void *data) { + Timer *p = data; + DBusMessageIter sub, sub2; + TimerValue *k; + + assert(i); + assert(property); + assert(p); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "(stt)", &sub)) + return -ENOMEM; + + LIST_FOREACH(value, k, p->values) { + _cleanup_free_ char *buf = NULL; + const char *t; + size_t l; + bool b; + + if (k->base == TIMER_CALENDAR) + continue; + + t = timer_base_to_string(k->base); + assert(endswith(t, "Sec")); + + /* s/Sec/USec/ */ + l = strlen(t); + buf = new(char, l+2); + if (!buf) + return -ENOMEM; + + memcpy(buf, t, l-3); + memcpy(buf+l-3, "USec", 5); + + b = dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2) && + dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &buf) && + dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT64, &k->value) && + dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT64, &k->next_elapse) && + dbus_message_iter_close_container(&sub, &sub2); + + if (!b) + return -ENOMEM; + } + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static int bus_timer_append_calendar_timers(DBusMessageIter *i, const char *property, void *data) { + Timer *p = data; + DBusMessageIter sub, sub2; + TimerValue *k; + + assert(i); + assert(property); + assert(p); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "(sst)", &sub)) + return -ENOMEM; + + LIST_FOREACH(value, k, p->values) { + _cleanup_free_ char *buf = NULL; + const char *t; + bool b; + int j; + + if (k->base != TIMER_CALENDAR) + continue; + + t = timer_base_to_string(k->base); + j = calendar_spec_to_string(k->calendar_spec, &buf); + if (j < 0) + return j; + + b = dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2) && + dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &t) && + dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &buf) && + dbus_message_iter_append_basic(&sub2, DBUS_TYPE_UINT64, &k->next_elapse) && + dbus_message_iter_close_container(&sub, &sub2); + + if (!b) + return -ENOMEM; + } + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static int bus_timer_append_unit(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data, *trigger; + const char *t; + + assert(i); + assert(property); + assert(u); + + trigger = UNIT_TRIGGER(u); + t = trigger ? trigger->id : ""; + + return dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &t) ? 0 : -ENOMEM; +} + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_timer_append_timer_result, timer_result, TimerResult); + +static const BusProperty bus_timer_properties[] = { + { "Unit", bus_timer_append_unit, "s", 0 }, + { "TimersMonotonic", bus_timer_append_monotonic_timers, "a(stt)", 0 }, + { "TimersCalendar", bus_timer_append_calendar_timers, "a(sst)", 0 }, + { "NextElapseUSecMonotonic", bus_property_append_usec, "t", offsetof(Timer, next_elapse_monotonic) }, + { "NextElapseUSecRealtime", bus_property_append_usec, "t", offsetof(Timer, next_elapse_realtime) }, + { "Result", bus_timer_append_timer_result, "s", offsetof(Timer, result) }, + { NULL, } +}; + +DBusHandlerResult bus_timer_message_handler(Unit *u, DBusConnection *c, DBusMessage *message) { + Timer *t = TIMER(u); + const BusBoundProperties bps[] = { + { "org.freedesktop.systemd1.Unit", bus_unit_properties, u }, + { "org.freedesktop.systemd1.Timer", bus_timer_properties, t }, + { NULL, } + }; + + SELINUX_UNIT_ACCESS_CHECK(u, c, message, "status"); + + return bus_default_message_handler(c, message, INTROSPECTION, INTERFACES_LIST, bps); +} diff --git a/src/core/dbus-timer.h b/src/core/dbus-timer.h new file mode 100644 index 0000000000..9ac30501d1 --- /dev/null +++ b/src/core/dbus-timer.h @@ -0,0 +1,31 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "unit.h" + +DBusHandlerResult bus_timer_message_handler(Unit *u, DBusConnection *c, DBusMessage *message); + +extern const char bus_timer_interface[]; +extern const char bus_timer_invalidating_properties[]; diff --git a/src/core/dbus-unit.c b/src/core/dbus-unit.c new file mode 100644 index 0000000000..575f8eb36a --- /dev/null +++ b/src/core/dbus-unit.c @@ -0,0 +1,1324 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "dbus.h" +#include "log.h" +#include "dbus-unit.h" +#include "bus-errors.h" +#include "dbus-common.h" +#include "selinux-access.h" +#include "cgroup-util.h" +#include "strv.h" +#include "path-util.h" +#include "fileio.h" + +const char bus_unit_interface[] _introspect_("Unit") = BUS_UNIT_INTERFACE; + +#define INVALIDATING_PROPERTIES \ + "LoadState\0" \ + "ActiveState\0" \ + "SubState\0" \ + "InactiveExitTimestamp\0" \ + "ActiveEnterTimestamp\0" \ + "ActiveExitTimestamp\0" \ + "InactiveEnterTimestamp\0" \ + "Job\0" \ + "NeedDaemonReload\0" + +static int bus_unit_append_names(DBusMessageIter *i, const char *property, void *data) { + char *t; + Iterator j; + DBusMessageIter sub; + Unit *u = data; + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "s", &sub)) + return -ENOMEM; + + SET_FOREACH(t, u->names, j) + if (!dbus_message_iter_append_basic(&sub, DBUS_TYPE_STRING, &t)) + return -ENOMEM; + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_following(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data, *f; + const char *d; + + assert(i); + assert(property); + assert(u); + + f = unit_following(u); + d = f ? f->id : ""; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &d)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_dependencies(DBusMessageIter *i, const char *property, void *data) { + Unit *u; + Iterator j; + DBusMessageIter sub; + Set *s = data; + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "s", &sub)) + return -ENOMEM; + + SET_FOREACH(u, s, j) + if (!dbus_message_iter_append_basic(&sub, DBUS_TYPE_STRING, &u->id)) + return -ENOMEM; + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_description(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + const char *d; + + assert(i); + assert(property); + assert(u); + + d = unit_description(u); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &d)) + return -ENOMEM; + + return 0; +} + +static DEFINE_BUS_PROPERTY_APPEND_ENUM(bus_unit_append_load_state, unit_load_state, UnitLoadState); + +static int bus_unit_append_active_state(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + const char *state; + + assert(i); + assert(property); + assert(u); + + state = unit_active_state_to_string(unit_active_state(u)); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &state)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_sub_state(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + const char *state; + + assert(i); + assert(property); + assert(u); + + state = unit_sub_state_to_string(u); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &state)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_file_state(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + const char *state; + + assert(i); + assert(property); + assert(u); + + state = strempty(unit_file_state_to_string(unit_get_unit_file_state(u))); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &state)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_can_start(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + dbus_bool_t b; + + assert(i); + assert(property); + assert(u); + + b = unit_can_start(u) && + !u->refuse_manual_start; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_BOOLEAN, &b)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_can_stop(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + dbus_bool_t b; + + assert(i); + assert(property); + assert(u); + + /* On the lower levels we assume that every unit we can start + * we can also stop */ + + b = unit_can_start(u) && + !u->refuse_manual_stop; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_BOOLEAN, &b)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_can_reload(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + dbus_bool_t b; + + assert(i); + assert(property); + assert(u); + + b = unit_can_reload(u); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_BOOLEAN, &b)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_can_isolate(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + dbus_bool_t b; + + assert(i); + assert(property); + assert(u); + + b = unit_can_isolate(u) && + !u->refuse_manual_start; + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_BOOLEAN, &b)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_job(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + DBusMessageIter sub; + _cleanup_free_ char *p = NULL; + + assert(i); + assert(property); + assert(u); + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_STRUCT, NULL, &sub)) + return -ENOMEM; + + if (u->job) { + + p = job_dbus_path(u->job); + if (!p) + return -ENOMEM; + + if (!dbus_message_iter_append_basic(&sub, DBUS_TYPE_UINT32, &u->job->id) || + !dbus_message_iter_append_basic(&sub, DBUS_TYPE_OBJECT_PATH, &p)) + return -ENOMEM; + } else { + uint32_t id = 0; + + /* No job, so let's fill in some placeholder + * data. Since we need to fill in a valid path we + * simple point to ourselves. */ + + p = unit_dbus_path(u); + if (!p) + return -ENOMEM; + + if (!dbus_message_iter_append_basic(&sub, DBUS_TYPE_UINT32, &id) || + !dbus_message_iter_append_basic(&sub, DBUS_TYPE_OBJECT_PATH, &p)) + return -ENOMEM; + } + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_default_cgroup(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + char *t; + CGroupBonding *cgb; + bool success; + + assert(i); + assert(property); + assert(u); + + cgb = unit_get_default_cgroup(u); + if (cgb) { + t = cgroup_bonding_to_string(cgb); + if (!t) + return -ENOMEM; + } else + t = (char*) ""; + + success = dbus_message_iter_append_basic(i, DBUS_TYPE_STRING, &t); + + if (cgb) + free(t); + + return success ? 0 : -ENOMEM; +} + +static int bus_unit_append_cgroups(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + CGroupBonding *cgb; + DBusMessageIter sub; + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "s", &sub)) + return -ENOMEM; + + LIST_FOREACH(by_unit, cgb, u->cgroup_bondings) { + _cleanup_free_ char *t = NULL; + bool success; + + t = cgroup_bonding_to_string(cgb); + if (!t) + return -ENOMEM; + + success = dbus_message_iter_append_basic(&sub, DBUS_TYPE_STRING, &t); + if (!success) + return -ENOMEM; + } + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_cgroup_attrs(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + CGroupAttribute *a; + DBusMessageIter sub, sub2; + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_ARRAY, "(sss)", &sub)) + return -ENOMEM; + + LIST_FOREACH(by_unit, a, u->cgroup_attributes) { + _cleanup_free_ char *v = NULL; + bool success; + + if (a->semantics && a->semantics->map_write) + a->semantics->map_write(a->semantics, a->value, &v); + + success = + dbus_message_iter_open_container(&sub, DBUS_TYPE_STRUCT, NULL, &sub2) && + dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &a->controller) && + dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, &a->name) && + dbus_message_iter_append_basic(&sub2, DBUS_TYPE_STRING, v ? &v : &a->value) && + dbus_message_iter_close_container(&sub, &sub2); + if (!success) + return -ENOMEM; + } + + if (!dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_need_daemon_reload(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + dbus_bool_t b; + + assert(i); + assert(property); + assert(u); + + b = unit_need_daemon_reload(u); + + if (!dbus_message_iter_append_basic(i, DBUS_TYPE_BOOLEAN, &b)) + return -ENOMEM; + + return 0; +} + +static int bus_unit_append_load_error(DBusMessageIter *i, const char *property, void *data) { + Unit *u = data; + const char *name, *message; + DBusMessageIter sub; + + assert(i); + assert(property); + assert(u); + + if (u->load_error != 0) { + name = bus_errno_to_dbus(u->load_error); + message = strempty(strerror(-u->load_error)); + } else + name = message = ""; + + if (!dbus_message_iter_open_container(i, DBUS_TYPE_STRUCT, NULL, &sub) || + !dbus_message_iter_append_basic(&sub, DBUS_TYPE_STRING, &name) || + !dbus_message_iter_append_basic(&sub, DBUS_TYPE_STRING, &message) || + !dbus_message_iter_close_container(i, &sub)) + return -ENOMEM; + + return 0; +} + +static DBusHandlerResult bus_unit_message_dispatch(Unit *u, DBusConnection *connection, DBusMessage *message) { + _cleanup_dbus_message_unref_ DBusMessage *reply = NULL; + DBusError error; + JobType job_type = _JOB_TYPE_INVALID; + bool reload_if_possible = false; + int r; + + dbus_error_init(&error); + + if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Unit", "Start")) + job_type = JOB_START; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Unit", "Stop")) + job_type = JOB_STOP; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Unit", "Reload")) + job_type = JOB_RELOAD; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Unit", "Restart")) + job_type = JOB_RESTART; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Unit", "TryRestart")) + job_type = JOB_TRY_RESTART; + else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Unit", "ReloadOrRestart")) { + reload_if_possible = true; + job_type = JOB_RESTART; + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Unit", "ReloadOrTryRestart")) { + reload_if_possible = true; + job_type = JOB_TRY_RESTART; + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Unit", "Kill")) { + const char *swho; + int32_t signo; + KillWho who; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &swho, + DBUS_TYPE_INT32, &signo, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + if (isempty(swho)) + who = KILL_ALL; + else { + who = kill_who_from_string(swho); + if (who < 0) + return bus_send_error_reply(connection, message, &error, -EINVAL); + } + + if (signo <= 0 || signo >= _NSIG) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "stop"); + + r = unit_kill(u, who, signo, &error); + if (r < 0) + return bus_send_error_reply(connection, message, &error, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (dbus_message_is_method_call(message, "org.freedesktop.systemd1.Unit", "ResetFailed")) { + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "reload"); + + unit_reset_failed(u); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (streq_ptr(dbus_message_get_member(message), "SetControlGroup")) { + DBusMessageIter iter; + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "start"); + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_unit_cgroup_set(u, &iter); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (streq_ptr(dbus_message_get_member(message), "UnsetControlGroup")) { + DBusMessageIter iter; + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "stop"); + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_unit_cgroup_unset(u, &iter); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + } else if (streq_ptr(dbus_message_get_member(message), "GetControlGroupAttribute")) { + DBusMessageIter iter; + _cleanup_strv_free_ char **list = NULL; + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "status"); + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_unit_cgroup_attribute_get(u, &iter, &list); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + dbus_message_iter_init_append(reply, &iter); + if (bus_append_strv_iter(&iter, list) < 0) + goto oom; + + } else if (streq_ptr(dbus_message_get_member(message), "SetControlGroupAttribute")) { + DBusMessageIter iter; + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "start"); + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_unit_cgroup_attribute_set(u, &iter); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (streq_ptr(dbus_message_get_member(message), "UnsetControlGroupAttribute")) { + DBusMessageIter iter; + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, "stop"); + + if (!dbus_message_iter_init(message, &iter)) + goto oom; + + r = bus_unit_cgroup_attribute_unset(u, &iter); + if (r < 0) + return bus_send_error_reply(connection, message, NULL, r); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + } else if (UNIT_VTABLE(u)->bus_message_handler) + return UNIT_VTABLE(u)->bus_message_handler(u, connection, message); + else + return DBUS_HANDLER_RESULT_NOT_YET_HANDLED; + + if (job_type != _JOB_TYPE_INVALID) { + const char *smode; + JobMode mode; + + if (!dbus_message_get_args( + message, + &error, + DBUS_TYPE_STRING, &smode, + DBUS_TYPE_INVALID)) + return bus_send_error_reply(connection, message, &error, -EINVAL); + + mode = job_mode_from_string(smode); + if (mode < 0) { + dbus_set_error(&error, BUS_ERROR_INVALID_JOB_MODE, "Job mode %s is invalid.", smode); + return bus_send_error_reply(connection, message, &error, -EINVAL); + } + + return bus_unit_queue_job(connection, message, u, job_type, mode, reload_if_possible); + } + + if (reply) + if (!bus_maybe_send_reply(connection, message, reply)) + goto oom; + + return DBUS_HANDLER_RESULT_HANDLED; + +oom: + dbus_error_free(&error); + return DBUS_HANDLER_RESULT_NEED_MEMORY; +} + +static DBusHandlerResult bus_unit_message_handler(DBusConnection *connection, DBusMessage *message, void *data) { + Manager *m = data; + Unit *u; + int r; + _cleanup_dbus_message_unref_ DBusMessage *reply = NULL; + DBusError error; + + assert(connection); + assert(message); + assert(m); + + dbus_error_init(&error); + + if (streq(dbus_message_get_path(message), "/org/freedesktop/systemd1/unit")) { + /* Be nice to gdbus and return introspection data for our mid-level paths */ + + SELINUX_ACCESS_CHECK(connection, message, "status"); + + if (dbus_message_is_method_call(message, "org.freedesktop.DBus.Introspectable", "Introspect")) { + char *introspection = NULL; + FILE *f; + Iterator i; + const char *k; + size_t size; + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + /* We roll our own introspection code here, instead of + * relying on bus_default_message_handler() because we + * need to generate our introspection string + * dynamically. */ + + f = open_memstream(&introspection, &size); + if (!f) + goto oom; + + fputs(DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE + "<node>\n", f); + + fputs(BUS_INTROSPECTABLE_INTERFACE, f); + fputs(BUS_PEER_INTERFACE, f); + + HASHMAP_FOREACH_KEY(u, k, m->units, i) { + char *p; + + if (k != u->id) + continue; + + p = bus_path_escape(k); + if (!p) { + fclose(f); + free(introspection); + goto oom; + } + + fprintf(f, "<node name=\"%s\"/>", p); + free(p); + } + + fputs("</node>\n", f); + + if (ferror(f)) { + fclose(f); + free(introspection); + goto oom; + } + + fclose(f); + + if (!introspection) + goto oom; + + if (!dbus_message_append_args(reply, DBUS_TYPE_STRING, &introspection, DBUS_TYPE_INVALID)) { + free(introspection); + goto oom; + } + + free(introspection); + + if (!bus_maybe_send_reply(connection, message, reply)) + goto oom; + + return DBUS_HANDLER_RESULT_HANDLED; + } + + return DBUS_HANDLER_RESULT_NOT_YET_HANDLED; + } + + r = manager_load_unit_from_dbus_path(m, dbus_message_get_path(message), &error, &u); + if (r == -ENOMEM) + goto oom; + if (r < 0) + return bus_send_error_reply(connection, message, &error, r); + + return bus_unit_message_dispatch(u, connection, message); + +oom: + dbus_error_free(&error); + + return DBUS_HANDLER_RESULT_NEED_MEMORY; +} + +const DBusObjectPathVTable bus_unit_vtable = { + .message_function = bus_unit_message_handler +}; + +void bus_unit_send_change_signal(Unit *u) { + _cleanup_free_ char *p = NULL; + _cleanup_dbus_message_unref_ DBusMessage *m = NULL; + + assert(u); + + if (u->in_dbus_queue) { + LIST_REMOVE(Unit, dbus_queue, u->manager->dbus_unit_queue, u); + u->in_dbus_queue = false; + } + + if (!u->id) + return; + + if (!bus_has_subscriber(u->manager)) { + u->sent_dbus_new_signal = true; + return; + } + + p = unit_dbus_path(u); + if (!p) + goto oom; + + if (u->sent_dbus_new_signal) { + /* Send a properties changed signal. First for the + * specific type, then for the generic unit. The + * clients may rely on this order to get atomic + * behavior if needed. */ + + if (UNIT_VTABLE(u)->bus_invalidating_properties) { + + m = bus_properties_changed_new(p, + UNIT_VTABLE(u)->bus_interface, + UNIT_VTABLE(u)->bus_invalidating_properties); + if (!m) + goto oom; + + if (bus_broadcast(u->manager, m) < 0) + goto oom; + + dbus_message_unref(m); + } + + m = bus_properties_changed_new(p, "org.freedesktop.systemd1.Unit", + INVALIDATING_PROPERTIES); + if (!m) + goto oom; + + } else { + /* Send a new signal */ + + m = dbus_message_new_signal("/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "UnitNew"); + if (!m) + goto oom; + + if (!dbus_message_append_args(m, + DBUS_TYPE_STRING, &u->id, + DBUS_TYPE_OBJECT_PATH, &p, + DBUS_TYPE_INVALID)) + goto oom; + } + + if (bus_broadcast(u->manager, m) < 0) + goto oom; + + u->sent_dbus_new_signal = true; + + return; + +oom: + log_oom(); +} + +void bus_unit_send_removed_signal(Unit *u) { + _cleanup_free_ char *p = NULL; + _cleanup_dbus_message_unref_ DBusMessage *m = NULL; + + assert(u); + + if (!bus_has_subscriber(u->manager)) + return; + + if (!u->sent_dbus_new_signal) + bus_unit_send_change_signal(u); + + if (!u->id) + return; + + p = unit_dbus_path(u); + if (!p) + goto oom; + + m = dbus_message_new_signal("/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "UnitRemoved"); + if (!m) + goto oom; + + if (!dbus_message_append_args(m, + DBUS_TYPE_STRING, &u->id, + DBUS_TYPE_OBJECT_PATH, &p, + DBUS_TYPE_INVALID)) + goto oom; + + if (bus_broadcast(u->manager, m) < 0) + goto oom; + + return; + +oom: + log_oom(); +} + +DBusHandlerResult bus_unit_queue_job( + DBusConnection *connection, + DBusMessage *message, + Unit *u, + JobType type, + JobMode mode, + bool reload_if_possible) { + + _cleanup_dbus_message_unref_ DBusMessage *reply = NULL; + _cleanup_free_ char *path = NULL; + Job *j; + JobBusClient *cl; + DBusError error; + int r; + + assert(connection); + assert(message); + assert(u); + assert(type >= 0 && type < _JOB_TYPE_MAX); + assert(mode >= 0 && mode < _JOB_MODE_MAX); + + dbus_error_init(&error); + + if (reload_if_possible && unit_can_reload(u)) { + if (type == JOB_RESTART) + type = JOB_RELOAD_OR_START; + else if (type == JOB_TRY_RESTART) + type = JOB_RELOAD; + } + + SELINUX_UNIT_ACCESS_CHECK(u, connection, message, + (type == JOB_START || type == JOB_RESTART || type == JOB_TRY_RESTART) ? "start" : + type == JOB_STOP ? "stop" : "reload"); + + if (type == JOB_STOP && u->load_state == UNIT_ERROR && unit_active_state(u) == UNIT_INACTIVE) { + dbus_set_error(&error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s not loaded.", u->id); + return bus_send_error_reply(connection, message, &error, -EPERM); + } + + if ((type == JOB_START && u->refuse_manual_start) || + (type == JOB_STOP && u->refuse_manual_stop) || + ((type == JOB_RESTART || type == JOB_TRY_RESTART) && (u->refuse_manual_start || u->refuse_manual_stop))) { + dbus_set_error(&error, BUS_ERROR_ONLY_BY_DEPENDENCY, + "Operation refused, unit %s may be requested by dependency only.", u->id); + return bus_send_error_reply(connection, message, &error, -EPERM); + } + + r = manager_add_job(u->manager, type, u, mode, true, &error, &j); + if (r < 0) + return bus_send_error_reply(connection, message, &error, r); + + cl = job_bus_client_new(connection, bus_message_get_sender_with_fallback(message)); + if (!cl) + goto oom; + + LIST_PREPEND(JobBusClient, client, j->bus_client_list, cl); + + reply = dbus_message_new_method_return(message); + if (!reply) + goto oom; + + path = job_dbus_path(j); + if (!path) + goto oom; + + if (!dbus_message_append_args( + reply, + DBUS_TYPE_OBJECT_PATH, &path, + DBUS_TYPE_INVALID)) + goto oom; + + if (!bus_maybe_send_reply(connection, message, reply)) + goto oom; + + return DBUS_HANDLER_RESULT_HANDLED; + +oom: + dbus_error_free(&error); + + return DBUS_HANDLER_RESULT_NEED_MEMORY; +} + +static int parse_mode(DBusMessageIter *iter, bool *runtime, bool next) { + const char *mode; + int r; + + assert(iter); + assert(runtime); + + r = bus_iter_get_basic_and_next(iter, DBUS_TYPE_STRING, &mode, next); + if (r < 0) + return r; + + if (streq(mode, "runtime")) + *runtime = true; + else if (streq(mode, "persistent")) + *runtime = false; + else + return -EINVAL; + + return 0; +} + +int bus_unit_cgroup_set(Unit *u, DBusMessageIter *iter) { + _cleanup_free_ char *controller = NULL, *old_path = NULL, *new_path = NULL, *contents = NULL; + const char *name; + CGroupBonding *b; + bool runtime; + int r; + + assert(u); + assert(iter); + + if (!unit_get_exec_context(u)) + return -EINVAL; + + r = bus_iter_get_basic_and_next(iter, DBUS_TYPE_STRING, &name, true); + if (r < 0) + return r; + + r = parse_mode(iter, &runtime, false); + if (r < 0) + return r; + + r = cg_split_spec(name, &controller, &new_path); + if (r < 0) + return r; + + if (!new_path) { + new_path = unit_default_cgroup_path(u); + if (!new_path) + return -ENOMEM; + } + + if (!controller || streq(controller, SYSTEMD_CGROUP_CONTROLLER)) + return -EINVAL; + + b = cgroup_bonding_find_list(u->cgroup_bondings, controller); + if (b) { + if (streq(b->path, new_path)) + return 0; + + if (b->essential) + return -EINVAL; + + old_path = strdup(b->path); + if (!old_path) + return -ENOMEM; + } + + r = unit_add_cgroup_from_text(u, name, true, &b); + if (r < 0) + return r; + if (r > 0) { + CGroupAttribute *a; + + /* Try to move things to the new place, and clean up the old place */ + cgroup_bonding_realize(b); + cgroup_bonding_migrate(b, u->cgroup_bondings); + + if (old_path) + cg_trim(controller, old_path, true); + + /* Apply the attributes to the new group */ + LIST_FOREACH(by_unit, a, u->cgroup_attributes) + if (streq(a->controller, controller)) + cgroup_attribute_apply(a, b); + } + + contents = strjoin("[", UNIT_VTABLE(u)->exec_section, "]\n" + "ControlGroup=", name, "\n", NULL); + if (!contents) + return -ENOMEM; + + return unit_write_drop_in(u, runtime, controller, contents); +} + +int bus_unit_cgroup_unset(Unit *u, DBusMessageIter *iter) { + _cleanup_free_ char *controller = NULL, *path = NULL, *target = NULL; + const char *name; + CGroupAttribute *a, *n; + CGroupBonding *b; + bool runtime; + int r; + + assert(u); + assert(iter); + + if (!unit_get_exec_context(u)) + return -EINVAL; + + r = bus_iter_get_basic_and_next(iter, DBUS_TYPE_STRING, &name, true); + if (r < 0) + return r; + + r = parse_mode(iter, &runtime, false); + if (r < 0) + return r; + + r = cg_split_spec(name, &controller, &path); + if (r < 0) + return r; + + if (!controller || streq(controller, SYSTEMD_CGROUP_CONTROLLER)) + return -EINVAL; + + b = cgroup_bonding_find_list(u->cgroup_bondings, controller); + if (!b) + return -ENOENT; + + if (path && !path_equal(path, b->path)) + return -ENOENT; + + if (b->essential) + return -EINVAL; + + unit_remove_drop_in(u, runtime, controller); + + /* Try to migrate the old group away */ + if (cg_pid_get_path(controller, 0, &target) >= 0) + cgroup_bonding_migrate_to(u->cgroup_bondings, target, false); + + cgroup_bonding_free(b, true); + + /* Drop all attributes of this controller */ + LIST_FOREACH_SAFE(by_unit, a, n, u->cgroup_attributes) { + if (!streq(a->controller, controller)) + continue; + + unit_remove_drop_in(u, runtime, a->name); + cgroup_attribute_free(a); + } + + return 0; +} + +int bus_unit_cgroup_attribute_get(Unit *u, DBusMessageIter *iter, char ***_result) { + _cleanup_free_ char *controller = NULL; + CGroupAttribute *a; + CGroupBonding *b; + const char *name; + char **l = NULL; + int r; + + assert(u); + assert(iter); + assert(_result); + + if (!unit_get_exec_context(u)) + return -EINVAL; + + r = bus_iter_get_basic_and_next(iter, DBUS_TYPE_STRING, &name, false); + if (r < 0) + return r; + + r = cg_controller_from_attr(name, &controller); + if (r < 0) + return r; + + /* First attempt, read the value from the kernel */ + b = cgroup_bonding_find_list(u->cgroup_bondings, controller); + if (b) { + _cleanup_free_ char *p = NULL, *v = NULL; + + r = cg_get_path(b->controller, b->path, name, &p); + if (r < 0) + return r; + + r = read_full_file(p, &v, NULL); + if (r >= 0) { + /* Split on new lines */ + l = strv_split_newlines(v); + if (!l) + return -ENOMEM; + + *_result = l; + return 0; + + } + } + + /* If that didn't work, read our cached value */ + LIST_FOREACH(by_unit, a, u->cgroup_attributes) { + + if (!cgroup_attribute_matches(a, controller, name)) + continue; + + r = strv_extend(&l, a->value); + if (r < 0) { + strv_free(l); + return r; + } + } + + if (!l) + return -ENOENT; + + *_result = l; + return 0; +} + +static int update_attribute_drop_in(Unit *u, bool runtime, const char *name) { + _cleanup_free_ char *buf = NULL; + CGroupAttribute *a; + + assert(u); + assert(name); + + LIST_FOREACH(by_unit, a, u->cgroup_attributes) { + if (!cgroup_attribute_matches(a, NULL, name)) + continue; + + if (!buf) { + buf = strjoin("[", UNIT_VTABLE(u)->exec_section, "]\n" + "ControlGroupAttribute=", a->name, " ", a->value, "\n", NULL); + + if (!buf) + return -ENOMEM; + } else { + char *b; + + b = strjoin(buf, + "ControlGroupAttribute=", a->name, " ", a->value, "\n", NULL); + + if (!b) + return -ENOMEM; + + free(buf); + buf = b; + } + } + + if (buf) + return unit_write_drop_in(u, runtime, name, buf); + else + return unit_remove_drop_in(u, runtime, name); +} + +int bus_unit_cgroup_attribute_set(Unit *u, DBusMessageIter *iter) { + _cleanup_strv_free_ char **l = NULL; + int r; + bool runtime = false; + char **value; + const char *name; + + assert(u); + assert(iter); + + if (!unit_get_exec_context(u)) + return -EINVAL; + + r = bus_iter_get_basic_and_next(iter, DBUS_TYPE_STRING, &name, true); + if (r < 0) + return r; + + r = bus_parse_strv_iter(iter, &l); + if (r < 0) + return r; + + if (!dbus_message_iter_next(iter)) + return -EINVAL; + + r = parse_mode(iter, &runtime, false); + if (r < 0) + return r; + + STRV_FOREACH(value, l) { + _cleanup_free_ char *v = NULL; + CGroupAttribute *a; + const CGroupSemantics *s; + + r = cgroup_semantics_find(NULL, name, *value, &v, &s); + if (r < 0) + return r; + + if (s && !s->multiple && l[1]) + return -EINVAL; + + r = unit_add_cgroup_attribute(u, s, NULL, name, v ? v : *value, &a); + if (r < 0) + return r; + + if (r > 0) { + CGroupBonding *b; + + b = cgroup_bonding_find_list(u->cgroup_bondings, a->controller); + if (!b) { + /* Doesn't exist yet? Then let's add it */ + r = unit_add_cgroup_from_text(u, a->controller, false, &b); + if (r < 0) + return r; + + if (r > 0) { + cgroup_bonding_realize(b); + cgroup_bonding_migrate(b, u->cgroup_bondings); + } + } + + /* Make it count */ + cgroup_attribute_apply(a, u->cgroup_bondings); + } + + } + + r = update_attribute_drop_in(u, runtime, name); + if (r < 0) + return r; + + return 0; +} + +int bus_unit_cgroup_attribute_unset(Unit *u, DBusMessageIter *iter) { + const char *name; + bool runtime; + int r; + + assert(u); + assert(iter); + + if (!unit_get_exec_context(u)) + return -EINVAL; + + r = bus_iter_get_basic_and_next(iter, DBUS_TYPE_STRING, &name, true); + if (r < 0) + return r; + + r = parse_mode(iter, &runtime, false); + if (r < 0) + return r; + + cgroup_attribute_free_some(u->cgroup_attributes, NULL, name); + update_attribute_drop_in(u, runtime, name); + + return 0; +} + +const BusProperty bus_unit_properties[] = { + { "Id", bus_property_append_string, "s", offsetof(Unit, id), true }, + { "Names", bus_unit_append_names, "as", 0 }, + { "Following", bus_unit_append_following, "s", 0 }, + { "Requires", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_REQUIRES]), true }, + { "RequiresOverridable", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_REQUIRES_OVERRIDABLE]), true }, + { "Requisite", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_REQUISITE]), true }, + { "RequisiteOverridable", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_REQUISITE_OVERRIDABLE]), true }, + { "Wants", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_WANTS]), true }, + { "BindsTo", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_BINDS_TO]), true }, + { "PartOf", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_PART_OF]), true }, + { "RequiredBy", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_REQUIRED_BY]), true }, + { "RequiredByOverridable",bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_REQUIRED_BY_OVERRIDABLE]), true }, + { "WantedBy", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_WANTED_BY]), true }, + { "BoundBy", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_BOUND_BY]), true }, + { "ConsistsOf", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_CONSISTS_OF]), true }, + { "Conflicts", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_CONFLICTS]), true }, + { "ConflictedBy", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_CONFLICTED_BY]), true }, + { "Before", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_BEFORE]), true }, + { "After", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_AFTER]), true }, + { "OnFailure", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_ON_FAILURE]), true }, + { "Triggers", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_TRIGGERS]), true }, + { "TriggeredBy", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_TRIGGERED_BY]), true }, + { "PropagatesReloadTo", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_PROPAGATES_RELOAD_TO]), true }, + { "ReloadPropagatedFrom", bus_unit_append_dependencies, "as", offsetof(Unit, dependencies[UNIT_RELOAD_PROPAGATED_FROM]), true }, + { "RequiresMountsFor", bus_property_append_strv, "as", offsetof(Unit, requires_mounts_for), true }, + { "Documentation", bus_property_append_strv, "as", offsetof(Unit, documentation), true }, + { "Description", bus_unit_append_description, "s", 0 }, + { "LoadState", bus_unit_append_load_state, "s", offsetof(Unit, load_state) }, + { "ActiveState", bus_unit_append_active_state, "s", 0 }, + { "SubState", bus_unit_append_sub_state, "s", 0 }, + { "FragmentPath", bus_property_append_string, "s", offsetof(Unit, fragment_path), true }, + { "SourcePath", bus_property_append_string, "s", offsetof(Unit, source_path), true }, + { "DropInPaths", bus_property_append_strv, "as", offsetof(Unit, dropin_paths), true }, + { "UnitFileState", bus_unit_append_file_state, "s", 0 }, + { "InactiveExitTimestamp",bus_property_append_usec, "t", offsetof(Unit, inactive_exit_timestamp.realtime) }, + { "InactiveExitTimestampMonotonic", bus_property_append_usec, "t", offsetof(Unit, inactive_exit_timestamp.monotonic) }, + { "ActiveEnterTimestamp", bus_property_append_usec, "t", offsetof(Unit, active_enter_timestamp.realtime) }, + { "ActiveEnterTimestampMonotonic", bus_property_append_usec, "t", offsetof(Unit, active_enter_timestamp.monotonic) }, + { "ActiveExitTimestamp", bus_property_append_usec, "t", offsetof(Unit, active_exit_timestamp.realtime) }, + { "ActiveExitTimestampMonotonic", bus_property_append_usec, "t", offsetof(Unit, active_exit_timestamp.monotonic) }, + { "InactiveEnterTimestamp", bus_property_append_usec, "t", offsetof(Unit, inactive_enter_timestamp.realtime) }, + { "InactiveEnterTimestampMonotonic",bus_property_append_usec, "t", offsetof(Unit, inactive_enter_timestamp.monotonic) }, + { "CanStart", bus_unit_append_can_start, "b", 0 }, + { "CanStop", bus_unit_append_can_stop, "b", 0 }, + { "CanReload", bus_unit_append_can_reload, "b", 0 }, + { "CanIsolate", bus_unit_append_can_isolate, "b", 0 }, + { "Job", bus_unit_append_job, "(uo)", 0 }, + { "StopWhenUnneeded", bus_property_append_bool, "b", offsetof(Unit, stop_when_unneeded) }, + { "RefuseManualStart", bus_property_append_bool, "b", offsetof(Unit, refuse_manual_start) }, + { "RefuseManualStop", bus_property_append_bool, "b", offsetof(Unit, refuse_manual_stop) }, + { "AllowIsolate", bus_property_append_bool, "b", offsetof(Unit, allow_isolate) }, + { "DefaultDependencies", bus_property_append_bool, "b", offsetof(Unit, default_dependencies) }, + { "OnFailureIsolate", bus_property_append_bool, "b", offsetof(Unit, on_failure_isolate) }, + { "IgnoreOnIsolate", bus_property_append_bool, "b", offsetof(Unit, ignore_on_isolate) }, + { "IgnoreOnSnapshot", bus_property_append_bool, "b", offsetof(Unit, ignore_on_snapshot) }, + { "NeedDaemonReload", bus_unit_append_need_daemon_reload, "b", 0 }, + { "JobTimeoutUSec", bus_property_append_usec, "t", offsetof(Unit, job_timeout) }, + { "ConditionTimestamp", bus_property_append_usec, "t", offsetof(Unit, condition_timestamp.realtime) }, + { "ConditionTimestampMonotonic", bus_property_append_usec, "t", offsetof(Unit, condition_timestamp.monotonic) }, + { "ConditionResult", bus_property_append_bool, "b", offsetof(Unit, condition_result) }, + { "LoadError", bus_unit_append_load_error, "(ss)", 0 }, + { NULL, } +}; + +const BusProperty bus_unit_cgroup_properties[] = { + { "DefaultControlGroup", bus_unit_append_default_cgroup, "s", 0 }, + { "ControlGroups", bus_unit_append_cgroups, "as", 0 }, + { "ControlGroupAttributes", bus_unit_append_cgroup_attrs, "a(sss)", 0 }, + { NULL, } +}; diff --git a/src/core/dbus-unit.h b/src/core/dbus-unit.h new file mode 100644 index 0000000000..acd1ddbe78 --- /dev/null +++ b/src/core/dbus-unit.h @@ -0,0 +1,179 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "manager.h" +#include "dbus-common.h" + +#define BUS_UNIT_INTERFACE \ + " <interface name=\"org.freedesktop.systemd1.Unit\">\n" \ + " <method name=\"Start\">\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"Stop\">\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"Reload\">\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"Restart\">\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"TryRestart\">\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"ReloadOrRestart\">\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"ReloadOrTryRestart\">\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"job\" type=\"o\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"Kill\">\n" \ + " <arg name=\"who\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"signal\" type=\"i\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"ResetFailed\"/>\n" \ + " <property name=\"Id\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Names\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"Following\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"Requires\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"RequiresOverridable\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"Requisite\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"RequisiteOverridable\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"Wants\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"BindsTo\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"PartOf\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"RequiredBy\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"RequiredByOverridable\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"WantedBy\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"BoundBy\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"ConsistsOf\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"Conflicts\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"ConflictedBy\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"Before\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"After\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"OnFailure\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"Triggers\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"TriggeredBy\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"PropagatesReloadTo\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"ReloadPropagatedFrom\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"RequiresMountsFor\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"Description\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"SourcePath\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"DropInPaths\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"Documentation\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"LoadState\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"ActiveState\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"SubState\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"FragmentPath\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"UnitFileState\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"InactiveExitTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"InactiveExitTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"ActiveEnterTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"ActiveEnterTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"ActiveExitTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"ActiveExitTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"InactiveEnterTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"InactiveEnterTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"CanStart\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"CanStop\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"CanReload\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"CanIsolate\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"Job\" type=\"(uo)\" access=\"read\"/>\n" \ + " <property name=\"StopWhenUnneeded\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"RefuseManualStart\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"RefuseManualStop\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"AllowIsolate\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"DefaultDependencies\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"OnFailureIsolate\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"IgnoreOnIsolate\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"IgnoreOnSnapshot\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"NeedDaemonReload\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"JobTimeoutUSec\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"ConditionTimestamp\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"ConditionTimestampMonotonic\" type=\"t\" access=\"read\"/>\n" \ + " <property name=\"ConditionResult\" type=\"b\" access=\"read\"/>\n" \ + " <property name=\"LoadError\" type=\"(ss)\" access=\"read\"/>\n" \ + " </interface>\n" + +#define BUS_UNIT_CGROUP_INTERFACE \ + " <property name=\"DefaultControlGroup\" type=\"s\" access=\"read\"/>\n" \ + " <property name=\"ControlGroups\" type=\"as\" access=\"read\"/>\n" \ + " <property name=\"ControlGroupAttributes\" type=\"a(sss)\" access=\"read\"/>\n" \ + " <method name=\"SetControlGroup\">\n" \ + " <arg name=\"group\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"UnsetControlGroup\">\n" \ + " <arg name=\"group\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"GetControlGroupAttribute\">\n" \ + " <arg name=\"attribute\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"values\" type=\"as\" direction=\"out\"/>\n" \ + " </method>\n" \ + " <method name=\"SetControlGroupAttribute\">\n" \ + " <arg name=\"attribute\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"values\" type=\"as\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " </method>\n" \ + " <method name=\"UnsetControlGroupAttribute\">\n" \ + " <arg name=\"attribute\" type=\"s\" direction=\"in\"/>\n" \ + " <arg name=\"mode\" type=\"s\" direction=\"in\"/>\n" \ + " </method>\n" + +#define BUS_UNIT_INTERFACES_LIST \ + BUS_GENERIC_INTERFACES_LIST \ + "org.freedesktop.systemd1.Unit\0" + +extern const BusProperty bus_unit_properties[]; +extern const BusProperty bus_unit_cgroup_properties[]; + +void bus_unit_send_change_signal(Unit *u); +void bus_unit_send_removed_signal(Unit *u); + +DBusHandlerResult bus_unit_queue_job( + DBusConnection *connection, + DBusMessage *message, + Unit *u, + JobType type, + JobMode mode, + bool reload_if_possible); + +int bus_unit_cgroup_set(Unit *u, DBusMessageIter *iter); +int bus_unit_cgroup_unset(Unit *u, DBusMessageIter *iter); +int bus_unit_cgroup_attribute_get(Unit *u, DBusMessageIter *iter, char ***_result); +int bus_unit_cgroup_attribute_set(Unit *u, DBusMessageIter *iter); +int bus_unit_cgroup_attribute_unset(Unit *u, DBusMessageIter *iter); + +extern const DBusObjectPathVTable bus_unit_vtable; + +extern const char bus_unit_interface[]; diff --git a/src/core/dbus.c b/src/core/dbus.c new file mode 100644 index 0000000000..1272c938cf --- /dev/null +++ b/src/core/dbus.c @@ -0,0 +1,1491 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/epoll.h> +#include <sys/timerfd.h> +#include <errno.h> +#include <unistd.h> +#include <dbus/dbus.h> + +#include "dbus.h" +#include "log.h" +#include "strv.h" +#include "cgroup.h" +#include "mkdir.h" +#include "missing.h" +#include "dbus-unit.h" +#include "dbus-job.h" +#include "dbus-manager.h" +#include "dbus-service.h" +#include "dbus-socket.h" +#include "dbus-target.h" +#include "dbus-device.h" +#include "dbus-mount.h" +#include "dbus-automount.h" +#include "dbus-snapshot.h" +#include "dbus-swap.h" +#include "dbus-timer.h" +#include "dbus-path.h" +#include "bus-errors.h" +#include "special.h" +#include "dbus-common.h" + +#define CONNECTIONS_MAX 512 + +/* Well-known address (http://dbus.freedesktop.org/doc/dbus-specification.html#message-bus-types) */ +#define DBUS_SYSTEM_BUS_DEFAULT_ADDRESS "unix:path=/var/run/dbus/system_bus_socket" +/* Only used as a fallback */ +#define DBUS_SESSION_BUS_DEFAULT_ADDRESS "autolaunch:" + +static const char bus_properties_interface[] = BUS_PROPERTIES_INTERFACE; +static const char bus_introspectable_interface[] = BUS_INTROSPECTABLE_INTERFACE; + +const char *const bus_interface_table[] = { + "org.freedesktop.DBus.Properties", bus_properties_interface, + "org.freedesktop.DBus.Introspectable", bus_introspectable_interface, + "org.freedesktop.systemd1.Manager", bus_manager_interface, + "org.freedesktop.systemd1.Job", bus_job_interface, + "org.freedesktop.systemd1.Unit", bus_unit_interface, + "org.freedesktop.systemd1.Service", bus_service_interface, + "org.freedesktop.systemd1.Socket", bus_socket_interface, + "org.freedesktop.systemd1.Target", bus_target_interface, + "org.freedesktop.systemd1.Device", bus_device_interface, + "org.freedesktop.systemd1.Mount", bus_mount_interface, + "org.freedesktop.systemd1.Automount", bus_automount_interface, + "org.freedesktop.systemd1.Snapshot", bus_snapshot_interface, + "org.freedesktop.systemd1.Swap", bus_swap_interface, + "org.freedesktop.systemd1.Timer", bus_timer_interface, + "org.freedesktop.systemd1.Path", bus_path_interface, + NULL +}; + +static void bus_done_api(Manager *m); +static void bus_done_system(Manager *m); +static void bus_done_private(Manager *m); +static void shutdown_connection(Manager *m, DBusConnection *c); + +static void bus_dispatch_status(DBusConnection *bus, DBusDispatchStatus status, void *data) { + Manager *m = data; + + assert(bus); + assert(m); + + /* We maintain two sets, one for those connections where we + * requested a dispatch, and another where we didn't. And then, + * we move the connections between the two sets. */ + + if (status == DBUS_DISPATCH_COMPLETE) + set_move_one(m->bus_connections, m->bus_connections_for_dispatch, bus); + else + set_move_one(m->bus_connections_for_dispatch, m->bus_connections, bus); +} + +void bus_watch_event(Manager *m, Watch *w, int events) { + assert(m); + assert(w); + + /* This is called by the event loop whenever there is + * something happening on D-Bus' file handles. */ + + if (!dbus_watch_get_enabled(w->data.bus_watch)) + return; + + dbus_watch_handle(w->data.bus_watch, bus_events_to_flags(events)); +} + +static dbus_bool_t bus_add_watch(DBusWatch *bus_watch, void *data) { + Manager *m = data; + Watch *w; + struct epoll_event ev; + + assert(bus_watch); + assert(m); + + if (!(w = new0(Watch, 1))) + return FALSE; + + w->fd = dbus_watch_get_unix_fd(bus_watch); + w->type = WATCH_DBUS_WATCH; + w->data.bus_watch = bus_watch; + + zero(ev); + ev.events = bus_flags_to_events(bus_watch); + ev.data.ptr = w; + + if (epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, w->fd, &ev) < 0) { + + if (errno != EEXIST) { + free(w); + return FALSE; + } + + /* Hmm, bloody D-Bus creates multiple watches on the + * same fd. epoll() does not like that. As a dirty + * hack we simply dup() the fd and hence get a second + * one we can safely add to the epoll(). */ + + if ((w->fd = dup(w->fd)) < 0) { + free(w); + return FALSE; + } + + if (epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, w->fd, &ev) < 0) { + close_nointr_nofail(w->fd); + free(w); + return FALSE; + } + + w->fd_is_dupped = true; + } + + dbus_watch_set_data(bus_watch, w, NULL); + + return TRUE; +} + +static void bus_remove_watch(DBusWatch *bus_watch, void *data) { + Manager *m = data; + Watch *w; + + assert(bus_watch); + assert(m); + + w = dbus_watch_get_data(bus_watch); + if (!w) + return; + + assert(w->type == WATCH_DBUS_WATCH); + assert_se(epoll_ctl(m->epoll_fd, EPOLL_CTL_DEL, w->fd, NULL) >= 0); + + if (w->fd_is_dupped) + close_nointr_nofail(w->fd); + + free(w); +} + +static void bus_toggle_watch(DBusWatch *bus_watch, void *data) { + Manager *m = data; + Watch *w; + struct epoll_event ev; + + assert(bus_watch); + assert(m); + + w = dbus_watch_get_data(bus_watch); + if (!w) + return; + + assert(w->type == WATCH_DBUS_WATCH); + + zero(ev); + ev.events = bus_flags_to_events(bus_watch); + ev.data.ptr = w; + + assert_se(epoll_ctl(m->epoll_fd, EPOLL_CTL_MOD, w->fd, &ev) == 0); +} + +static int bus_timeout_arm(Manager *m, Watch *w) { + struct itimerspec its = {}; + + assert(m); + assert(w); + + if (dbus_timeout_get_enabled(w->data.bus_timeout)) { + timespec_store(&its.it_value, dbus_timeout_get_interval(w->data.bus_timeout) * USEC_PER_MSEC); + its.it_interval = its.it_value; + } + + if (timerfd_settime(w->fd, 0, &its, NULL) < 0) + return -errno; + + return 0; +} + +void bus_timeout_event(Manager *m, Watch *w, int events) { + assert(m); + assert(w); + + /* This is called by the event loop whenever there is + * something happening on D-Bus' file handles. */ + + if (!(dbus_timeout_get_enabled(w->data.bus_timeout))) + return; + + dbus_timeout_handle(w->data.bus_timeout); +} + +static dbus_bool_t bus_add_timeout(DBusTimeout *timeout, void *data) { + Manager *m = data; + Watch *w; + struct epoll_event ev; + + assert(timeout); + assert(m); + + if (!(w = new0(Watch, 1))) + return FALSE; + + if ((w->fd = timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK|TFD_CLOEXEC)) < 0) + goto fail; + + w->type = WATCH_DBUS_TIMEOUT; + w->data.bus_timeout = timeout; + + if (bus_timeout_arm(m, w) < 0) + goto fail; + + zero(ev); + ev.events = EPOLLIN; + ev.data.ptr = w; + + if (epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, w->fd, &ev) < 0) + goto fail; + + dbus_timeout_set_data(timeout, w, NULL); + + return TRUE; + +fail: + if (w->fd >= 0) + close_nointr_nofail(w->fd); + + free(w); + return FALSE; +} + +static void bus_remove_timeout(DBusTimeout *timeout, void *data) { + Manager *m = data; + Watch *w; + + assert(timeout); + assert(m); + + w = dbus_timeout_get_data(timeout); + if (!w) + return; + + assert(w->type == WATCH_DBUS_TIMEOUT); + + assert_se(epoll_ctl(m->epoll_fd, EPOLL_CTL_DEL, w->fd, NULL) >= 0); + close_nointr_nofail(w->fd); + free(w); +} + +static void bus_toggle_timeout(DBusTimeout *timeout, void *data) { + Manager *m = data; + Watch *w; + int r; + + assert(timeout); + assert(m); + + w = dbus_timeout_get_data(timeout); + if (!w) + return; + + assert(w->type == WATCH_DBUS_TIMEOUT); + + if ((r = bus_timeout_arm(m, w)) < 0) + log_error("Failed to rearm timer: %s", strerror(-r)); +} + +static DBusHandlerResult api_bus_message_filter(DBusConnection *connection, DBusMessage *message, void *data) { + Manager *m = data; + DBusError error; + DBusMessage *reply = NULL; + + assert(connection); + assert(message); + assert(m); + + dbus_error_init(&error); + + if (dbus_message_get_type(message) == DBUS_MESSAGE_TYPE_METHOD_CALL || + dbus_message_get_type(message) == DBUS_MESSAGE_TYPE_SIGNAL) + log_debug("Got D-Bus request: %s.%s() on %s", + dbus_message_get_interface(message), + dbus_message_get_member(message), + dbus_message_get_path(message)); + + if (dbus_message_is_signal(message, DBUS_INTERFACE_LOCAL, "Disconnected")) { + log_debug("API D-Bus connection terminated."); + bus_done_api(m); + + } else if (dbus_message_is_signal(message, DBUS_INTERFACE_DBUS, "NameOwnerChanged")) { + const char *name, *old_owner, *new_owner; + + if (!dbus_message_get_args(message, &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_STRING, &old_owner, + DBUS_TYPE_STRING, &new_owner, + DBUS_TYPE_INVALID)) + log_error("Failed to parse NameOwnerChanged message: %s", bus_error_message(&error)); + else { + if (set_remove(BUS_CONNECTION_SUBSCRIBED(m, connection), (char*) name)) + log_debug("Subscription client vanished: %s (left: %u)", name, set_size(BUS_CONNECTION_SUBSCRIBED(m, connection))); + + if (old_owner[0] == 0) + old_owner = NULL; + + if (new_owner[0] == 0) + new_owner = NULL; + + manager_dispatch_bus_name_owner_changed(m, name, old_owner, new_owner); + } + } else if (dbus_message_is_signal(message, "org.freedesktop.systemd1.Activator", "ActivationRequest")) { + const char *name; + + if (!dbus_message_get_args(message, &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_INVALID)) + log_error("Failed to parse ActivationRequest message: %s", bus_error_message(&error)); + else { + int r; + Unit *u; + + log_debug("Got D-Bus activation request for %s", name); + + if (manager_unit_inactive_or_pending(m, SPECIAL_DBUS_SERVICE) || + manager_unit_inactive_or_pending(m, SPECIAL_DBUS_SOCKET)) { + r = -EADDRNOTAVAIL; + dbus_set_error(&error, BUS_ERROR_SHUTTING_DOWN, "Refusing activation, D-Bus is shutting down."); + } else { + r = manager_load_unit(m, name, NULL, &error, &u); + + if (r >= 0 && u->refuse_manual_start) + r = -EPERM; + + if (r >= 0) + r = manager_add_job(m, JOB_START, u, JOB_REPLACE, true, &error, NULL); + } + + if (r < 0) { + const char *id, *text; + + log_debug("D-Bus activation failed for %s: %s", name, strerror(-r)); + + if (!(reply = dbus_message_new_signal("/org/freedesktop/systemd1", "org.freedesktop.systemd1.Activator", "ActivationFailure"))) + goto oom; + + id = error.name ? error.name : bus_errno_to_dbus(r); + text = bus_error(&error, r); + + if (!dbus_message_set_destination(reply, DBUS_SERVICE_DBUS) || + !dbus_message_append_args(reply, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_STRING, &id, + DBUS_TYPE_STRING, &text, + DBUS_TYPE_INVALID)) + goto oom; + } + + /* On success we don't do anything, the service will be spawned now */ + } + } + + dbus_error_free(&error); + + if (reply) { + if (!bus_maybe_send_reply(connection, message, reply)) + goto oom; + + dbus_message_unref(reply); + } + + return DBUS_HANDLER_RESULT_NOT_YET_HANDLED; + +oom: + if (reply) + dbus_message_unref(reply); + + dbus_error_free(&error); + + return DBUS_HANDLER_RESULT_NEED_MEMORY; +} + +static DBusHandlerResult system_bus_message_filter(DBusConnection *connection, DBusMessage *message, void *data) { + Manager *m = data; + DBusError error; + + assert(connection); + assert(message); + assert(m); + + dbus_error_init(&error); + + if (m->api_bus != m->system_bus && + (dbus_message_get_type(message) == DBUS_MESSAGE_TYPE_METHOD_CALL || + dbus_message_get_type(message) == DBUS_MESSAGE_TYPE_SIGNAL)) + log_debug("Got D-Bus request on system bus: %s.%s() on %s", + dbus_message_get_interface(message), + dbus_message_get_member(message), + dbus_message_get_path(message)); + + if (dbus_message_is_signal(message, DBUS_INTERFACE_LOCAL, "Disconnected")) { + log_debug("System D-Bus connection terminated."); + bus_done_system(m); + + } else if (m->running_as != SYSTEMD_SYSTEM && + dbus_message_is_signal(message, "org.freedesktop.systemd1.Agent", "Released")) { + + const char *cgroup; + + if (!dbus_message_get_args(message, &error, + DBUS_TYPE_STRING, &cgroup, + DBUS_TYPE_INVALID)) + log_error("Failed to parse Released message: %s", bus_error_message(&error)); + else + cgroup_notify_empty(m, cgroup); + } + + dbus_error_free(&error); + return DBUS_HANDLER_RESULT_NOT_YET_HANDLED; +} + +static DBusHandlerResult private_bus_message_filter(DBusConnection *connection, DBusMessage *message, void *data) { + Manager *m = data; + DBusError error; + + assert(connection); + assert(message); + assert(m); + + dbus_error_init(&error); + + if (dbus_message_get_type(message) == DBUS_MESSAGE_TYPE_METHOD_CALL || + dbus_message_get_type(message) == DBUS_MESSAGE_TYPE_SIGNAL) + log_debug("Got D-Bus request: %s.%s() on %s", + dbus_message_get_interface(message), + dbus_message_get_member(message), + dbus_message_get_path(message)); + + if (dbus_message_is_signal(message, DBUS_INTERFACE_LOCAL, "Disconnected")) + shutdown_connection(m, connection); + else if (m->running_as == SYSTEMD_SYSTEM && + dbus_message_is_signal(message, "org.freedesktop.systemd1.Agent", "Released")) { + + const char *cgroup; + + if (!dbus_message_get_args(message, &error, + DBUS_TYPE_STRING, &cgroup, + DBUS_TYPE_INVALID)) + log_error("Failed to parse Released message: %s", bus_error_message(&error)); + else + cgroup_notify_empty(m, cgroup); + + /* Forward the message to the system bus, so that user + * instances are notified as well */ + + if (m->system_bus) + dbus_connection_send(m->system_bus, message, NULL); + } + + dbus_error_free(&error); + + return DBUS_HANDLER_RESULT_NOT_YET_HANDLED; +} + +unsigned bus_dispatch(Manager *m) { + DBusConnection *c; + + assert(m); + + if (m->queued_message) { + /* If we cannot get rid of this message we won't + * dispatch any D-Bus messages, so that we won't end + * up wanting to queue another message. */ + + if (m->queued_message_connection) + if (!dbus_connection_send(m->queued_message_connection, m->queued_message, NULL)) + return 0; + + dbus_message_unref(m->queued_message); + m->queued_message = NULL; + m->queued_message_connection = NULL; + } + + if ((c = set_first(m->bus_connections_for_dispatch))) { + if (dbus_connection_dispatch(c) == DBUS_DISPATCH_COMPLETE) + set_move_one(m->bus_connections, m->bus_connections_for_dispatch, c); + + return 1; + } + + return 0; +} + +static void request_name_pending_cb(DBusPendingCall *pending, void *userdata) { + DBusMessage *reply; + DBusError error; + + dbus_error_init(&error); + + assert_se(reply = dbus_pending_call_steal_reply(pending)); + + switch (dbus_message_get_type(reply)) { + + case DBUS_MESSAGE_TYPE_ERROR: + + assert_se(dbus_set_error_from_message(&error, reply)); + log_warning("RequestName() failed: %s", bus_error_message(&error)); + break; + + case DBUS_MESSAGE_TYPE_METHOD_RETURN: { + uint32_t r; + + if (!dbus_message_get_args(reply, + &error, + DBUS_TYPE_UINT32, &r, + DBUS_TYPE_INVALID)) { + log_error("Failed to parse RequestName() reply: %s", bus_error_message(&error)); + break; + } + + if (r == 1) + log_debug("Successfully acquired name."); + else + log_error("Name already owned."); + + break; + } + + default: + assert_not_reached("Invalid reply message"); + } + + dbus_message_unref(reply); + dbus_error_free(&error); +} + +static int request_name(Manager *m) { + const char *name = "org.freedesktop.systemd1"; + /* Allow replacing of our name, to ease implementation of + * reexecution, where we keep the old connection open until + * after the new connection is set up and the name installed + * to allow clients to synchronously wait for reexecution to + * finish */ + uint32_t flags = DBUS_NAME_FLAG_ALLOW_REPLACEMENT|DBUS_NAME_FLAG_REPLACE_EXISTING; + DBusMessage *message = NULL; + DBusPendingCall *pending = NULL; + + if (!(message = dbus_message_new_method_call( + DBUS_SERVICE_DBUS, + DBUS_PATH_DBUS, + DBUS_INTERFACE_DBUS, + "RequestName"))) + goto oom; + + if (!dbus_message_append_args( + message, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_UINT32, &flags, + DBUS_TYPE_INVALID)) + goto oom; + + if (!dbus_connection_send_with_reply(m->api_bus, message, &pending, -1)) + goto oom; + + if (!dbus_pending_call_set_notify(pending, request_name_pending_cb, m, NULL)) + goto oom; + + dbus_message_unref(message); + dbus_pending_call_unref(pending); + + /* We simple ask for the name and don't wait for it. Sooner or + * later we'll have it. */ + + return 0; + +oom: + if (pending) { + dbus_pending_call_cancel(pending); + dbus_pending_call_unref(pending); + } + + if (message) + dbus_message_unref(message); + + return -ENOMEM; +} + +static void query_name_list_pending_cb(DBusPendingCall *pending, void *userdata) { + DBusMessage *reply; + DBusError error; + Manager *m = userdata; + + assert(m); + + dbus_error_init(&error); + + assert_se(reply = dbus_pending_call_steal_reply(pending)); + + switch (dbus_message_get_type(reply)) { + + case DBUS_MESSAGE_TYPE_ERROR: + + assert_se(dbus_set_error_from_message(&error, reply)); + log_warning("ListNames() failed: %s", bus_error_message(&error)); + break; + + case DBUS_MESSAGE_TYPE_METHOD_RETURN: { + int r; + char **l; + + if ((r = bus_parse_strv(reply, &l)) < 0) + log_warning("Failed to parse ListNames() reply: %s", strerror(-r)); + else { + char **t; + + STRV_FOREACH(t, l) + /* This is a bit hacky, we say the + * owner of the name is the name + * itself, because we don't want the + * extra traffic to figure out the + * real owner. */ + manager_dispatch_bus_name_owner_changed(m, *t, NULL, *t); + + strv_free(l); + } + + break; + } + + default: + assert_not_reached("Invalid reply message"); + } + + dbus_message_unref(reply); + dbus_error_free(&error); +} + +static int query_name_list(Manager *m) { + DBusMessage *message = NULL; + DBusPendingCall *pending = NULL; + + /* Asks for the currently installed bus names */ + + if (!(message = dbus_message_new_method_call( + DBUS_SERVICE_DBUS, + DBUS_PATH_DBUS, + DBUS_INTERFACE_DBUS, + "ListNames"))) + goto oom; + + if (!dbus_connection_send_with_reply(m->api_bus, message, &pending, -1)) + goto oom; + + if (!dbus_pending_call_set_notify(pending, query_name_list_pending_cb, m, NULL)) + goto oom; + + dbus_message_unref(message); + dbus_pending_call_unref(pending); + + /* We simple ask for the list and don't wait for it. Sooner or + * later we'll get it. */ + + return 0; + +oom: + if (pending) { + dbus_pending_call_cancel(pending); + dbus_pending_call_unref(pending); + } + + if (message) + dbus_message_unref(message); + + return -ENOMEM; +} + +static int bus_setup_loop(Manager *m, DBusConnection *bus) { + assert(m); + assert(bus); + + dbus_connection_set_exit_on_disconnect(bus, FALSE); + + if (!dbus_connection_set_watch_functions(bus, bus_add_watch, bus_remove_watch, bus_toggle_watch, m, NULL) || + !dbus_connection_set_timeout_functions(bus, bus_add_timeout, bus_remove_timeout, bus_toggle_timeout, m, NULL)) + return log_oom(); + + if (set_put(m->bus_connections_for_dispatch, bus) < 0) + return log_oom(); + + dbus_connection_set_dispatch_status_function(bus, bus_dispatch_status, m, NULL); + return 0; +} + +static dbus_bool_t allow_only_same_user(DBusConnection *connection, unsigned long uid, void *data) { + return uid == 0 || uid == geteuid(); +} + +static void bus_new_connection( + DBusServer *server, + DBusConnection *new_connection, + void *data) { + + Manager *m = data; + + assert(m); + + if (set_size(m->bus_connections) >= CONNECTIONS_MAX) { + log_error("Too many concurrent connections."); + return; + } + + dbus_connection_set_unix_user_function(new_connection, allow_only_same_user, NULL, NULL); + + if (bus_setup_loop(m, new_connection) < 0) + return; + + if (!dbus_connection_register_object_path(new_connection, "/org/freedesktop/systemd1", &bus_manager_vtable, m) || + !dbus_connection_register_fallback(new_connection, "/org/freedesktop/systemd1/unit", &bus_unit_vtable, m) || + !dbus_connection_register_fallback(new_connection, "/org/freedesktop/systemd1/job", &bus_job_vtable, m) || + !dbus_connection_add_filter(new_connection, private_bus_message_filter, m, NULL)) { + log_oom(); + return; + } + + log_debug("Accepted connection on private bus."); + + dbus_connection_ref(new_connection); +} + +static int init_registered_system_bus(Manager *m) { + char *id; + + if (!dbus_connection_add_filter(m->system_bus, system_bus_message_filter, m, NULL)) + return log_oom(); + + if (m->running_as != SYSTEMD_SYSTEM) { + DBusError error; + + dbus_error_init(&error); + + dbus_bus_add_match(m->system_bus, + "type='signal'," + "interface='org.freedesktop.systemd1.Agent'," + "member='Released'," + "path='/org/freedesktop/systemd1/agent'", + &error); + + if (dbus_error_is_set(&error)) { + log_error("Failed to register match: %s", bus_error_message(&error)); + dbus_error_free(&error); + return -1; + } + } + + log_debug("Successfully connected to system D-Bus bus %s as %s", + strnull((id = dbus_connection_get_server_id(m->system_bus))), + strnull(dbus_bus_get_unique_name(m->system_bus))); + dbus_free(id); + + return 0; +} + +static int init_registered_api_bus(Manager *m) { + int r; + + if (!dbus_connection_register_object_path(m->api_bus, "/org/freedesktop/systemd1", &bus_manager_vtable, m) || + !dbus_connection_register_fallback(m->api_bus, "/org/freedesktop/systemd1/unit", &bus_unit_vtable, m) || + !dbus_connection_register_fallback(m->api_bus, "/org/freedesktop/systemd1/job", &bus_job_vtable, m) || + !dbus_connection_add_filter(m->api_bus, api_bus_message_filter, m, NULL)) + return log_oom(); + + /* Get NameOwnerChange messages */ + dbus_bus_add_match(m->api_bus, + "type='signal'," + "sender='"DBUS_SERVICE_DBUS"'," + "interface='"DBUS_INTERFACE_DBUS"'," + "member='NameOwnerChanged'," + "path='"DBUS_PATH_DBUS"'", + NULL); + + /* Get activation requests */ + dbus_bus_add_match(m->api_bus, + "type='signal'," + "sender='"DBUS_SERVICE_DBUS"'," + "interface='org.freedesktop.systemd1.Activator'," + "member='ActivationRequest'," + "path='"DBUS_PATH_DBUS"'", + NULL); + + r = request_name(m); + if (r < 0) + return r; + + r = query_name_list(m); + if (r < 0) + return r; + + if (m->running_as == SYSTEMD_USER) { + char *id; + log_debug("Successfully connected to API D-Bus bus %s as %s", + strnull((id = dbus_connection_get_server_id(m->api_bus))), + strnull(dbus_bus_get_unique_name(m->api_bus))); + dbus_free(id); + } else + log_debug("Successfully initialized API on the system bus"); + + return 0; +} + +static void bus_register_cb(DBusPendingCall *pending, void *userdata) { + Manager *m = userdata; + DBusConnection **conn; + DBusMessage *reply; + DBusError error; + const char *name; + int r = 0; + + dbus_error_init(&error); + + conn = dbus_pending_call_get_data(pending, m->conn_data_slot); + assert(conn == &m->system_bus || conn == &m->api_bus); + + reply = dbus_pending_call_steal_reply(pending); + + switch (dbus_message_get_type(reply)) { + case DBUS_MESSAGE_TYPE_ERROR: + assert_se(dbus_set_error_from_message(&error, reply)); + log_warning("Failed to register to bus: %s", bus_error_message(&error)); + r = -1; + break; + case DBUS_MESSAGE_TYPE_METHOD_RETURN: + if (!dbus_message_get_args(reply, &error, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_INVALID)) { + log_error("Failed to parse Hello reply: %s", bus_error_message(&error)); + r = -1; + break; + } + + log_debug("Received name %s in reply to Hello", name); + if (!dbus_bus_set_unique_name(*conn, name)) { + log_error("Failed to set unique name"); + r = -1; + break; + } + + if (conn == &m->system_bus) { + r = init_registered_system_bus(m); + if (r == 0 && m->running_as == SYSTEMD_SYSTEM) + r = init_registered_api_bus(m); + } else + r = init_registered_api_bus(m); + + break; + default: + assert_not_reached("Invalid reply message"); + } + + dbus_message_unref(reply); + dbus_error_free(&error); + + if (r < 0) { + if (conn == &m->system_bus) { + log_debug("Failed setting up the system bus"); + bus_done_system(m); + } else { + log_debug("Failed setting up the API bus"); + bus_done_api(m); + } + } +} + +static int manager_bus_async_register(Manager *m, DBusConnection **conn) { + DBusMessage *message = NULL; + DBusPendingCall *pending = NULL; + + message = dbus_message_new_method_call(DBUS_SERVICE_DBUS, + DBUS_PATH_DBUS, + DBUS_INTERFACE_DBUS, + "Hello"); + if (!message) + goto oom; + + if (!dbus_connection_send_with_reply(*conn, message, &pending, -1)) + goto oom; + + if (!dbus_pending_call_set_data(pending, m->conn_data_slot, conn, NULL)) + goto oom; + + if (!dbus_pending_call_set_notify(pending, bus_register_cb, m, NULL)) + goto oom; + + dbus_message_unref(message); + dbus_pending_call_unref(pending); + + return 0; +oom: + if (pending) { + dbus_pending_call_cancel(pending); + dbus_pending_call_unref(pending); + } + + if (message) + dbus_message_unref(message); + + return -ENOMEM; +} + +static DBusConnection* manager_bus_connect_private(Manager *m, DBusBusType type) { + const char *address; + DBusConnection *connection; + DBusError error; + + switch (type) { + case DBUS_BUS_SYSTEM: + address = secure_getenv("DBUS_SYSTEM_BUS_ADDRESS"); + if (!address || !address[0]) + address = DBUS_SYSTEM_BUS_DEFAULT_ADDRESS; + break; + case DBUS_BUS_SESSION: + address = secure_getenv("DBUS_SESSION_BUS_ADDRESS"); + if (!address || !address[0]) + address = DBUS_SESSION_BUS_DEFAULT_ADDRESS; + break; + default: + assert_not_reached("Invalid bus type"); + } + + dbus_error_init(&error); + + connection = dbus_connection_open_private(address, &error); + if (!connection) { + log_warning("Failed to open private bus connection: %s", bus_error_message(&error)); + goto fail; + } + + return connection; + +fail: + dbus_error_free(&error); + return NULL; +} + +static int bus_init_system(Manager *m) { + int r; + + if (m->system_bus) + return 0; + + m->system_bus = manager_bus_connect_private(m, DBUS_BUS_SYSTEM); + if (!m->system_bus) { + log_debug("Failed to connect to system D-Bus, retrying later"); + r = 0; + goto fail; + } + + r = bus_setup_loop(m, m->system_bus); + if (r < 0) + goto fail; + + r = manager_bus_async_register(m, &m->system_bus); + if (r < 0) + goto fail; + + return 0; +fail: + bus_done_system(m); + + return r; +} + +static int bus_init_api(Manager *m) { + int r; + + if (m->api_bus) + return 0; + + if (m->running_as == SYSTEMD_SYSTEM) { + m->api_bus = m->system_bus; + /* In this mode there is no distinct connection to the API bus, + * the API is published on the system bus. + * bus_register_cb() is aware of that and will init the API + * when the system bus gets registered. + * No need to setup anything here. */ + return 0; + } + + m->api_bus = manager_bus_connect_private(m, DBUS_BUS_SESSION); + if (!m->api_bus) { + log_debug("Failed to connect to API D-Bus, retrying later"); + r = 0; + goto fail; + } + + r = bus_setup_loop(m, m->api_bus); + if (r < 0) + goto fail; + + r = manager_bus_async_register(m, &m->api_bus); + if (r < 0) + goto fail; + + return 0; +fail: + bus_done_api(m); + + return r; +} + +static int bus_init_private(Manager *m) { + DBusError error; + int r; + static const char *const external_only[] = { + "EXTERNAL", + NULL + }; + + assert(m); + + dbus_error_init(&error); + + if (m->private_bus) + return 0; + + if (m->running_as == SYSTEMD_SYSTEM) { + + /* We want the private bus only when running as init */ + if (getpid() != 1) + return 0; + + unlink("/run/systemd/private"); + m->private_bus = dbus_server_listen("unix:path=/run/systemd/private", &error); + } else { + const char *e; + char *p; + char *escaped; + + e = secure_getenv("XDG_RUNTIME_DIR"); + if (!e) + return 0; + + if (asprintf(&p, "%s/systemd/private", e) < 0) { + r = log_oom(); + goto fail; + } + + mkdir_parents_label(p, 0755); + unlink(p); + free(p); + + escaped = dbus_address_escape_value(e); + if (!escaped) { + r = log_oom(); + goto fail; + } + if (asprintf(&p, "unix:path=%s/systemd/private", escaped) < 0) { + dbus_free(escaped); + r = log_oom(); + goto fail; + } + dbus_free(escaped); + + m->private_bus = dbus_server_listen(p, &error); + free(p); + } + + if (!m->private_bus) { + log_error("Failed to create private D-Bus server: %s", bus_error_message(&error)); + r = -EIO; + goto fail; + } + + if (!dbus_server_set_auth_mechanisms(m->private_bus, (const char**) external_only) || + !dbus_server_set_watch_functions(m->private_bus, bus_add_watch, bus_remove_watch, bus_toggle_watch, m, NULL) || + !dbus_server_set_timeout_functions(m->private_bus, bus_add_timeout, bus_remove_timeout, bus_toggle_timeout, m, NULL)) { + r = log_oom(); + goto fail; + } + + dbus_server_set_new_connection_function(m->private_bus, bus_new_connection, m, NULL); + + log_debug("Successfully created private D-Bus server."); + + return 0; + +fail: + bus_done_private(m); + dbus_error_free(&error); + + return r; +} + +int bus_init(Manager *m, bool try_bus_connect) { + int r; + + if (set_ensure_allocated(&m->bus_connections, trivial_hash_func, trivial_compare_func) < 0 || + set_ensure_allocated(&m->bus_connections_for_dispatch, trivial_hash_func, trivial_compare_func) < 0) + goto oom; + + if (m->name_data_slot < 0) + if (!dbus_pending_call_allocate_data_slot(&m->name_data_slot)) + goto oom; + + if (m->conn_data_slot < 0) + if (!dbus_pending_call_allocate_data_slot(&m->conn_data_slot)) + goto oom; + + if (m->subscribed_data_slot < 0) + if (!dbus_connection_allocate_data_slot(&m->subscribed_data_slot)) + goto oom; + + if (try_bus_connect) { + if ((r = bus_init_system(m)) < 0 || + (r = bus_init_api(m)) < 0) + return r; + } + + if ((r = bus_init_private(m)) < 0) + return r; + + return 0; +oom: + return log_oom(); +} + +static void shutdown_connection(Manager *m, DBusConnection *c) { + Set *s; + Job *j; + Iterator i; + + HASHMAP_FOREACH(j, m->jobs, i) { + JobBusClient *cl, *nextcl; + LIST_FOREACH_SAFE(client, cl, nextcl, j->bus_client_list) { + if (cl->bus == c) { + LIST_REMOVE(JobBusClient, client, j->bus_client_list, cl); + free(cl); + } + } + } + + set_remove(m->bus_connections, c); + set_remove(m->bus_connections_for_dispatch, c); + + if ((s = BUS_CONNECTION_SUBSCRIBED(m, c))) { + char *t; + + while ((t = set_steal_first(s))) + free(t); + + set_free(s); + } + + if (m->queued_message_connection == c) { + m->queued_message_connection = NULL; + + if (m->queued_message) { + dbus_message_unref(m->queued_message); + m->queued_message = NULL; + } + } + + dbus_connection_set_dispatch_status_function(c, NULL, NULL, NULL); + /* system manager cannot afford to block on DBus */ + if (m->running_as != SYSTEMD_SYSTEM) + dbus_connection_flush(c); + dbus_connection_close(c); + dbus_connection_unref(c); +} + +static void bus_done_api(Manager *m) { + if (!m->api_bus) + return; + + if (m->running_as == SYSTEMD_USER) + shutdown_connection(m, m->api_bus); + + m->api_bus = NULL; + + if (m->queued_message) { + dbus_message_unref(m->queued_message); + m->queued_message = NULL; + } +} + +static void bus_done_system(Manager *m) { + if (!m->system_bus) + return; + + if (m->running_as == SYSTEMD_SYSTEM) + bus_done_api(m); + + shutdown_connection(m, m->system_bus); + m->system_bus = NULL; +} + +static void bus_done_private(Manager *m) { + if (!m->private_bus) + return; + + dbus_server_disconnect(m->private_bus); + dbus_server_unref(m->private_bus); + m->private_bus = NULL; +} + +void bus_done(Manager *m) { + DBusConnection *c; + + bus_done_api(m); + bus_done_system(m); + bus_done_private(m); + + while ((c = set_steal_first(m->bus_connections))) + shutdown_connection(m, c); + + while ((c = set_steal_first(m->bus_connections_for_dispatch))) + shutdown_connection(m, c); + + set_free(m->bus_connections); + set_free(m->bus_connections_for_dispatch); + + if (m->name_data_slot >= 0) + dbus_pending_call_free_data_slot(&m->name_data_slot); + + if (m->conn_data_slot >= 0) + dbus_pending_call_free_data_slot(&m->conn_data_slot); + + if (m->subscribed_data_slot >= 0) + dbus_connection_free_data_slot(&m->subscribed_data_slot); +} + +static void query_pid_pending_cb(DBusPendingCall *pending, void *userdata) { + Manager *m = userdata; + DBusMessage *reply; + DBusError error; + const char *name; + + dbus_error_init(&error); + + assert_se(name = BUS_PENDING_CALL_NAME(m, pending)); + assert_se(reply = dbus_pending_call_steal_reply(pending)); + + switch (dbus_message_get_type(reply)) { + + case DBUS_MESSAGE_TYPE_ERROR: + + assert_se(dbus_set_error_from_message(&error, reply)); + log_warning("GetConnectionUnixProcessID() failed: %s", bus_error_message(&error)); + break; + + case DBUS_MESSAGE_TYPE_METHOD_RETURN: { + uint32_t r; + + if (!dbus_message_get_args(reply, + &error, + DBUS_TYPE_UINT32, &r, + DBUS_TYPE_INVALID)) { + log_error("Failed to parse GetConnectionUnixProcessID() reply: %s", bus_error_message(&error)); + break; + } + + manager_dispatch_bus_query_pid_done(m, name, (pid_t) r); + break; + } + + default: + assert_not_reached("Invalid reply message"); + } + + dbus_message_unref(reply); + dbus_error_free(&error); +} + +int bus_query_pid(Manager *m, const char *name) { + DBusMessage *message = NULL; + DBusPendingCall *pending = NULL; + char *n = NULL; + + assert(m); + assert(name); + + if (!(message = dbus_message_new_method_call( + DBUS_SERVICE_DBUS, + DBUS_PATH_DBUS, + DBUS_INTERFACE_DBUS, + "GetConnectionUnixProcessID"))) + goto oom; + + if (!(dbus_message_append_args( + message, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_INVALID))) + goto oom; + + if (!dbus_connection_send_with_reply(m->api_bus, message, &pending, -1)) + goto oom; + + if (!(n = strdup(name))) + goto oom; + + if (!dbus_pending_call_set_data(pending, m->name_data_slot, n, free)) + goto oom; + + n = NULL; + + if (!dbus_pending_call_set_notify(pending, query_pid_pending_cb, m, NULL)) + goto oom; + + dbus_message_unref(message); + dbus_pending_call_unref(pending); + + return 0; + +oom: + free(n); + + if (pending) { + dbus_pending_call_cancel(pending); + dbus_pending_call_unref(pending); + } + + if (message) + dbus_message_unref(message); + + return -ENOMEM; +} + +int bus_broadcast(Manager *m, DBusMessage *message) { + bool oom = false; + Iterator i; + DBusConnection *c; + + assert(m); + assert(message); + + SET_FOREACH(c, m->bus_connections_for_dispatch, i) + if (c != m->system_bus || m->running_as == SYSTEMD_SYSTEM) + oom = !dbus_connection_send(c, message, NULL); + + SET_FOREACH(c, m->bus_connections, i) + if (c != m->system_bus || m->running_as == SYSTEMD_SYSTEM) + oom = !dbus_connection_send(c, message, NULL); + + return oom ? -ENOMEM : 0; +} + +bool bus_has_subscriber(Manager *m) { + Iterator i; + DBusConnection *c; + + assert(m); + + SET_FOREACH(c, m->bus_connections_for_dispatch, i) + if (bus_connection_has_subscriber(m, c)) + return true; + + SET_FOREACH(c, m->bus_connections, i) + if (bus_connection_has_subscriber(m, c)) + return true; + + return false; +} + +bool bus_connection_has_subscriber(Manager *m, DBusConnection *c) { + assert(m); + assert(c); + + return !set_isempty(BUS_CONNECTION_SUBSCRIBED(m, c)); +} + +int bus_fdset_add_all(Manager *m, FDSet *fds) { + Iterator i; + DBusConnection *c; + + assert(m); + assert(fds); + + /* When we are about to reexecute we add all D-Bus fds to the + * set to pass over to the newly executed systemd. They won't + * be used there however, except that they are closed at the + * very end of deserialization, those making it possible for + * clients to synchronously wait for systemd to reexec by + * simply waiting for disconnection */ + + SET_FOREACH(c, m->bus_connections_for_dispatch, i) { + int fd; + + if (dbus_connection_get_unix_fd(c, &fd)) { + fd = fdset_put_dup(fds, fd); + + if (fd < 0) + return fd; + } + } + + SET_FOREACH(c, m->bus_connections, i) { + int fd; + + if (dbus_connection_get_unix_fd(c, &fd)) { + fd = fdset_put_dup(fds, fd); + + if (fd < 0) + return fd; + } + } + + return 0; +} + +void bus_broadcast_finished( + Manager *m, + usec_t firmware_usec, + usec_t loader_usec, + usec_t kernel_usec, + usec_t initrd_usec, + usec_t userspace_usec, + usec_t total_usec) { + + DBusMessage *message; + + assert(m); + + message = dbus_message_new_signal("/org/freedesktop/systemd1", "org.freedesktop.systemd1.Manager", "StartupFinished"); + if (!message) { + log_oom(); + return; + } + + assert_cc(sizeof(usec_t) == sizeof(uint64_t)); + if (!dbus_message_append_args(message, + DBUS_TYPE_UINT64, &firmware_usec, + DBUS_TYPE_UINT64, &loader_usec, + DBUS_TYPE_UINT64, &kernel_usec, + DBUS_TYPE_UINT64, &initrd_usec, + DBUS_TYPE_UINT64, &userspace_usec, + DBUS_TYPE_UINT64, &total_usec, + DBUS_TYPE_INVALID)) { + log_oom(); + goto finish; + } + + + if (bus_broadcast(m, message) < 0) { + log_oom(); + goto finish; + } + +finish: + if (message) + dbus_message_unref(message); +} diff --git a/src/core/dbus.h b/src/core/dbus.h new file mode 100644 index 0000000000..c7a058e198 --- /dev/null +++ b/src/core/dbus.h @@ -0,0 +1,50 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include "manager.h" + +int bus_init(Manager *m, bool try_bus_connect); +void bus_done(Manager *m); + +unsigned bus_dispatch(Manager *m); + +void bus_watch_event(Manager *m, Watch *w, int events); +void bus_timeout_event(Manager *m, Watch *w, int events); + +int bus_query_pid(Manager *m, const char *name); + +int bus_broadcast(Manager *m, DBusMessage *message); + +bool bus_has_subscriber(Manager *m); +bool bus_connection_has_subscriber(Manager *m, DBusConnection *c); + +int bus_fdset_add_all(Manager *m, FDSet *fds); + +void bus_broadcast_finished(Manager *m, usec_t firmware_usec, usec_t loader_usec, usec_t kernel_usec, usec_t initrd_usec, usec_t userspace_usec, usec_t total_usec); + +#define BUS_CONNECTION_SUBSCRIBED(m, c) dbus_connection_get_data((c), (m)->subscribed_data_slot) +#define BUS_PENDING_CALL_NAME(m, p) dbus_pending_call_get_data((p), (m)->name_data_slot) + +extern const char * const bus_interface_table[]; diff --git a/src/core/device.c b/src/core/device.c new file mode 100644 index 0000000000..9fca82ab16 --- /dev/null +++ b/src/core/device.c @@ -0,0 +1,654 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <sys/epoll.h> +#include <libudev.h> + +#include "unit.h" +#include "device.h" +#include "strv.h" +#include "log.h" +#include "unit-name.h" +#include "dbus-device.h" +#include "def.h" +#include "path-util.h" + +static const UnitActiveState state_translation_table[_DEVICE_STATE_MAX] = { + [DEVICE_DEAD] = UNIT_INACTIVE, + [DEVICE_PLUGGED] = UNIT_ACTIVE +}; + +static void device_unset_sysfs(Device *d) { + Device *first; + + assert(d); + + if (!d->sysfs) + return; + + /* Remove this unit from the chain of devices which share the + * same sysfs path. */ + first = hashmap_get(UNIT(d)->manager->devices_by_sysfs, d->sysfs); + LIST_REMOVE(Device, same_sysfs, first, d); + + if (first) + hashmap_remove_and_replace(UNIT(d)->manager->devices_by_sysfs, d->sysfs, first->sysfs, first); + else + hashmap_remove(UNIT(d)->manager->devices_by_sysfs, d->sysfs); + + free(d->sysfs); + d->sysfs = NULL; +} + +static void device_init(Unit *u) { + Device *d = DEVICE(u); + + assert(d); + assert(UNIT(d)->load_state == UNIT_STUB); + + /* In contrast to all other unit types we timeout jobs waiting + * for devices by default. This is because they otherwise wait + * indefinitely for plugged in devices, something which cannot + * happen for the other units since their operations time out + * anyway. */ + UNIT(d)->job_timeout = DEFAULT_TIMEOUT_USEC; + + UNIT(d)->ignore_on_isolate = true; + UNIT(d)->ignore_on_snapshot = true; +} + +static void device_done(Unit *u) { + Device *d = DEVICE(u); + + assert(d); + + device_unset_sysfs(d); +} + +static void device_set_state(Device *d, DeviceState state) { + DeviceState old_state; + assert(d); + + old_state = d->state; + d->state = state; + + if (state != old_state) + log_debug_unit(UNIT(d)->id, + "%s changed %s -> %s", UNIT(d)->id, + device_state_to_string(old_state), + device_state_to_string(state)); + + unit_notify(UNIT(d), state_translation_table[old_state], state_translation_table[state], true); +} + +static int device_coldplug(Unit *u) { + Device *d = DEVICE(u); + + assert(d); + assert(d->state == DEVICE_DEAD); + + if (d->sysfs) + device_set_state(d, DEVICE_PLUGGED); + + return 0; +} + +static void device_dump(Unit *u, FILE *f, const char *prefix) { + Device *d = DEVICE(u); + + assert(d); + + fprintf(f, + "%sDevice State: %s\n" + "%sSysfs Path: %s\n", + prefix, device_state_to_string(d->state), + prefix, strna(d->sysfs)); +} + +_pure_ static UnitActiveState device_active_state(Unit *u) { + assert(u); + + return state_translation_table[DEVICE(u)->state]; +} + +_pure_ static const char *device_sub_state_to_string(Unit *u) { + assert(u); + + return device_state_to_string(DEVICE(u)->state); +} + +static int device_add_escaped_name(Unit *u, const char *dn) { + char *e; + int r; + + assert(u); + assert(dn); + assert(dn[0] == '/'); + + e = unit_name_from_path(dn, ".device"); + if (!e) + return -ENOMEM; + + r = unit_add_name(u, e); + free(e); + + if (r < 0 && r != -EEXIST) + return r; + + return 0; +} + +static int device_find_escape_name(Manager *m, const char *dn, Unit **_u) { + char *e; + Unit *u; + + assert(m); + assert(dn); + assert(dn[0] == '/'); + assert(_u); + + e = unit_name_from_path(dn, ".device"); + if (!e) + return -ENOMEM; + + u = manager_get_unit(m, e); + free(e); + + if (u) { + *_u = u; + return 1; + } + + return 0; +} + +static int device_update_unit(Manager *m, struct udev_device *dev, const char *path, bool main) { + const char *sysfs, *model; + Unit *u = NULL; + int r; + bool delete; + + assert(m); + + if (!(sysfs = udev_device_get_syspath(dev))) + return -ENOMEM; + + if ((r = device_find_escape_name(m, path, &u)) < 0) + return r; + + if (u && DEVICE(u)->sysfs && !path_equal(DEVICE(u)->sysfs, sysfs)) + return -EEXIST; + + if (!u) { + delete = true; + + u = unit_new(m, sizeof(Device)); + if (!u) + return -ENOMEM; + + r = device_add_escaped_name(u, path); + if (r < 0) + goto fail; + + unit_add_to_load_queue(u); + } else + delete = false; + + /* If this was created via some dependency and has not + * actually been seen yet ->sysfs will not be + * initialized. Hence initialize it if necessary. */ + + if (!DEVICE(u)->sysfs) { + Device *first; + + if (!(DEVICE(u)->sysfs = strdup(sysfs))) { + r = -ENOMEM; + goto fail; + } + + if (!m->devices_by_sysfs) + if (!(m->devices_by_sysfs = hashmap_new(string_hash_func, string_compare_func))) { + r = -ENOMEM; + goto fail; + } + + first = hashmap_get(m->devices_by_sysfs, sysfs); + LIST_PREPEND(Device, same_sysfs, first, DEVICE(u)); + + if ((r = hashmap_replace(m->devices_by_sysfs, DEVICE(u)->sysfs, first)) < 0) + goto fail; + } + + if ((model = udev_device_get_property_value(dev, "ID_MODEL_FROM_DATABASE")) || + (model = udev_device_get_property_value(dev, "ID_MODEL"))) { + if ((r = unit_set_description(u, model)) < 0) + goto fail; + } else + if ((r = unit_set_description(u, path)) < 0) + goto fail; + + if (main) { + /* The additional systemd udev properties we only + * interpret for the main object */ + const char *wants, *alias; + + alias = udev_device_get_property_value(dev, "SYSTEMD_ALIAS"); + if (alias) { + char *state, *w; + size_t l; + + FOREACH_WORD_QUOTED(w, l, alias, state) { + char *e; + + e = strndup(w, l); + if (!e) { + r = -ENOMEM; + goto fail; + } + + if (!is_path(e)) { + log_warning("SYSTEMD_ALIAS for %s is not a path, ignoring: %s", sysfs, e); + free(e); + } else { + device_update_unit(m, dev, e, false); + free(e); + } + } + } + + wants = udev_device_get_property_value(dev, "SYSTEMD_WANTS"); + if (wants) { + char *state, *w; + size_t l; + + FOREACH_WORD_QUOTED(w, l, wants, state) { + char *e, *n; + + e = strndup(w, l); + if (!e) { + r = -ENOMEM; + goto fail; + } + n = unit_name_mangle(e); + if (!n) { + r = -ENOMEM; + goto fail; + } + free(e); + + r = unit_add_dependency_by_name(u, UNIT_WANTS, n, NULL, true); + free(n); + if (r < 0) + goto fail; + } + } + } + + unit_add_to_dbus_queue(u); + return 0; + +fail: + log_warning("Failed to load device unit: %s", strerror(-r)); + + if (delete && u) + unit_free(u); + + return r; +} + +static int device_process_new_device(Manager *m, struct udev_device *dev, bool update_state) { + const char *sysfs, *dn; + struct udev_list_entry *item = NULL, *first = NULL; + int r; + + assert(m); + + if (!(sysfs = udev_device_get_syspath(dev))) + return -ENOMEM; + + /* Add the main unit named after the sysfs path */ + r = device_update_unit(m, dev, sysfs, true); + if (r < 0) + return r; + + /* Add an additional unit for the device node */ + if ((dn = udev_device_get_devnode(dev))) + device_update_unit(m, dev, dn, false); + + /* Add additional units for all symlinks */ + first = udev_device_get_devlinks_list_entry(dev); + udev_list_entry_foreach(item, first) { + const char *p; + struct stat st; + + /* Don't bother with the /dev/block links */ + p = udev_list_entry_get_name(item); + + if (path_startswith(p, "/dev/block/") || + path_startswith(p, "/dev/char/")) + continue; + + /* Verify that the symlink in the FS actually belongs + * to this device. This is useful to deal with + * conflicting devices, e.g. when two disks want the + * same /dev/disk/by-label/xxx link because they have + * the same label. We want to make sure that the same + * device that won the symlink wins in systemd, so we + * check the device node major/minor*/ + if (stat(p, &st) >= 0) + if ((!S_ISBLK(st.st_mode) && !S_ISCHR(st.st_mode)) || + st.st_rdev != udev_device_get_devnum(dev)) + continue; + + device_update_unit(m, dev, p, false); + } + + if (update_state) { + Device *d, *l; + + manager_dispatch_load_queue(m); + + l = hashmap_get(m->devices_by_sysfs, sysfs); + LIST_FOREACH(same_sysfs, d, l) + device_set_state(d, DEVICE_PLUGGED); + } + + return 0; +} + +static int device_process_path(Manager *m, const char *path, bool update_state) { + int r; + struct udev_device *dev; + + assert(m); + assert(path); + + if (!(dev = udev_device_new_from_syspath(m->udev, path))) { + log_warning("Failed to get udev device object from udev for path %s.", path); + return -ENOMEM; + } + + r = device_process_new_device(m, dev, update_state); + udev_device_unref(dev); + return r; +} + +static int device_process_removed_device(Manager *m, struct udev_device *dev) { + const char *sysfs; + Device *d; + + assert(m); + assert(dev); + + if (!(sysfs = udev_device_get_syspath(dev))) + return -ENOMEM; + + /* Remove all units of this sysfs path */ + while ((d = hashmap_get(m->devices_by_sysfs, sysfs))) { + device_unset_sysfs(d); + device_set_state(d, DEVICE_DEAD); + } + + return 0; +} + +static Unit *device_following(Unit *u) { + Device *d = DEVICE(u); + Device *other, *first = NULL; + + assert(d); + + if (startswith(u->id, "sys-")) + return NULL; + + /* Make everybody follow the unit that's named after the sysfs path */ + for (other = d->same_sysfs_next; other; other = other->same_sysfs_next) + if (startswith(UNIT(other)->id, "sys-")) + return UNIT(other); + + for (other = d->same_sysfs_prev; other; other = other->same_sysfs_prev) { + if (startswith(UNIT(other)->id, "sys-")) + return UNIT(other); + + first = other; + } + + return UNIT(first); +} + +static int device_following_set(Unit *u, Set **_s) { + Device *d = DEVICE(u); + Device *other; + Set *s; + int r; + + assert(d); + assert(_s); + + if (!d->same_sysfs_prev && !d->same_sysfs_next) { + *_s = NULL; + return 0; + } + + if (!(s = set_new(NULL, NULL))) + return -ENOMEM; + + for (other = d->same_sysfs_next; other; other = other->same_sysfs_next) + if ((r = set_put(s, other)) < 0) + goto fail; + + for (other = d->same_sysfs_prev; other; other = other->same_sysfs_prev) + if ((r = set_put(s, other)) < 0) + goto fail; + + *_s = s; + return 1; + +fail: + set_free(s); + return r; +} + +static void device_shutdown(Manager *m) { + assert(m); + + if (m->udev_monitor) { + udev_monitor_unref(m->udev_monitor); + m->udev_monitor = NULL; + } + + if (m->udev) { + udev_unref(m->udev); + m->udev = NULL; + } + + hashmap_free(m->devices_by_sysfs); + m->devices_by_sysfs = NULL; +} + +static int device_enumerate(Manager *m) { + int r; + struct udev_enumerate *e = NULL; + struct udev_list_entry *item = NULL, *first = NULL; + + assert(m); + + if (!m->udev) { + struct epoll_event ev; + + if (!(m->udev = udev_new())) + return -ENOMEM; + + if (!(m->udev_monitor = udev_monitor_new_from_netlink(m->udev, "udev"))) { + r = -ENOMEM; + goto fail; + } + + /* This will fail if we are unprivileged, but that + * should not matter much, as user instances won't run + * during boot. */ + udev_monitor_set_receive_buffer_size(m->udev_monitor, 128*1024*1024); + + if (udev_monitor_filter_add_match_tag(m->udev_monitor, "systemd") < 0) { + r = -ENOMEM; + goto fail; + } + + if (udev_monitor_enable_receiving(m->udev_monitor) < 0) { + r = -EIO; + goto fail; + } + + m->udev_watch.type = WATCH_UDEV; + m->udev_watch.fd = udev_monitor_get_fd(m->udev_monitor); + + zero(ev); + ev.events = EPOLLIN; + ev.data.ptr = &m->udev_watch; + + if (epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, m->udev_watch.fd, &ev) < 0) + return -errno; + } + + if (!(e = udev_enumerate_new(m->udev))) { + r = -ENOMEM; + goto fail; + } + if (udev_enumerate_add_match_tag(e, "systemd") < 0) { + r = -EIO; + goto fail; + } + + if (udev_enumerate_scan_devices(e) < 0) { + r = -EIO; + goto fail; + } + + first = udev_enumerate_get_list_entry(e); + udev_list_entry_foreach(item, first) + device_process_path(m, udev_list_entry_get_name(item), false); + + udev_enumerate_unref(e); + return 0; + +fail: + if (e) + udev_enumerate_unref(e); + + device_shutdown(m); + return r; +} + +void device_fd_event(Manager *m, int events) { + struct udev_device *dev; + int r; + const char *action, *ready; + + assert(m); + + if (events != EPOLLIN) { + static RATELIMIT_DEFINE(limit, 10*USEC_PER_SEC, 5); + + if (!ratelimit_test(&limit)) + log_error("Failed to get udev event: %m"); + if (!(events & EPOLLIN)) + return; + } + + if (!(dev = udev_monitor_receive_device(m->udev_monitor))) { + /* + * libudev might filter-out devices which pass the bloom filter, + * so getting NULL here is not necessarily an error + */ + return; + } + + if (!(action = udev_device_get_action(dev))) { + log_error("Failed to get udev action string."); + goto fail; + } + + ready = udev_device_get_property_value(dev, "SYSTEMD_READY"); + + if (streq(action, "remove") || (ready && parse_boolean(ready) == 0)) { + if ((r = device_process_removed_device(m, dev)) < 0) { + log_error("Failed to process udev device event: %s", strerror(-r)); + goto fail; + } + } else { + if ((r = device_process_new_device(m, dev, true)) < 0) { + log_error("Failed to process udev device event: %s", strerror(-r)); + goto fail; + } + } + +fail: + udev_device_unref(dev); +} + +static const char* const device_state_table[_DEVICE_STATE_MAX] = { + [DEVICE_DEAD] = "dead", + [DEVICE_PLUGGED] = "plugged" +}; + +DEFINE_STRING_TABLE_LOOKUP(device_state, DeviceState); + +const UnitVTable device_vtable = { + .object_size = sizeof(Device), + .sections = + "Unit\0" + "Device\0" + "Install\0", + + .no_instances = true, + + .init = device_init, + + .load = unit_load_fragment_and_dropin_optional, + .done = device_done, + .coldplug = device_coldplug, + + .dump = device_dump, + + .active_state = device_active_state, + .sub_state_to_string = device_sub_state_to_string, + + .bus_interface = "org.freedesktop.systemd1.Device", + .bus_message_handler = bus_device_message_handler, + .bus_invalidating_properties = bus_device_invalidating_properties, + + .following = device_following, + .following_set = device_following_set, + + .enumerate = device_enumerate, + .shutdown = device_shutdown, + + .status_message_formats = { + .starting_stopping = { + [0] = "Expecting device %s...", + }, + .finished_start_job = { + [JOB_DONE] = "Found device %s.", + [JOB_TIMEOUT] = "Timed out waiting for device %s.", + }, + }, +}; diff --git a/src/core/device.h b/src/core/device.h new file mode 100644 index 0000000000..41e8de9703 --- /dev/null +++ b/src/core/device.h @@ -0,0 +1,56 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Device Device; + +#include "unit.h" + +/* We simply watch devices, we cannot plug/unplug them. That + * simplifies the state engine greatly */ +typedef enum DeviceState { + DEVICE_DEAD, + DEVICE_PLUGGED, + _DEVICE_STATE_MAX, + _DEVICE_STATE_INVALID = -1 +} DeviceState; + +struct Device { + Unit meta; + + char *sysfs; + + /* In order to be able to distinguish dependencies on + different device nodes we might end up creating multiple + devices for the same sysfs path. We chain them up here. */ + + LIST_FIELDS(struct Device, same_sysfs); + + DeviceState state; +}; + +extern const UnitVTable device_vtable; + +void device_fd_event(Manager *m, int events); + +const char* device_state_to_string(DeviceState i) _const_; +DeviceState device_state_from_string(const char *s) _pure_; diff --git a/src/core/execute.c b/src/core/execute.c new file mode 100644 index 0000000000..3959ef9623 --- /dev/null +++ b/src/core/execute.c @@ -0,0 +1,2254 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <assert.h> +#include <dirent.h> +#include <errno.h> +#include <fcntl.h> +#include <unistd.h> +#include <string.h> +#include <signal.h> +#include <sys/socket.h> +#include <sys/un.h> +#include <sys/prctl.h> +#include <linux/sched.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <grp.h> +#include <pwd.h> +#include <sys/mount.h> +#include <linux/fs.h> +#include <linux/oom.h> +#include <sys/poll.h> +#include <linux/seccomp-bpf.h> +#include <glob.h> +#include <libgen.h> + +#ifdef HAVE_PAM +#include <security/pam_appl.h> +#endif + +#include "execute.h" +#include "strv.h" +#include "macro.h" +#include "capability.h" +#include "util.h" +#include "log.h" +#include "sd-messages.h" +#include "ioprio.h" +#include "securebits.h" +#include "cgroup.h" +#include "namespace.h" +#include "tcpwrap.h" +#include "exit-status.h" +#include "missing.h" +#include "utmp-wtmp.h" +#include "def.h" +#include "loopback-setup.h" +#include "path-util.h" +#include "syscall-list.h" +#include "env-util.h" +#include "fileio.h" + +#define IDLE_TIMEOUT_USEC (5*USEC_PER_SEC) + +/* This assumes there is a 'tty' group */ +#define TTY_MODE 0620 + +static int shift_fds(int fds[], unsigned n_fds) { + int start, restart_from; + + if (n_fds <= 0) + return 0; + + /* Modifies the fds array! (sorts it) */ + + assert(fds); + + start = 0; + for (;;) { + int i; + + restart_from = -1; + + for (i = start; i < (int) n_fds; i++) { + int nfd; + + /* Already at right index? */ + if (fds[i] == i+3) + continue; + + if ((nfd = fcntl(fds[i], F_DUPFD, i+3)) < 0) + return -errno; + + close_nointr_nofail(fds[i]); + fds[i] = nfd; + + /* Hmm, the fd we wanted isn't free? Then + * let's remember that and try again from here*/ + if (nfd != i+3 && restart_from < 0) + restart_from = i; + } + + if (restart_from < 0) + break; + + start = restart_from; + } + + return 0; +} + +static int flags_fds(const int fds[], unsigned n_fds, bool nonblock) { + unsigned i; + int r; + + if (n_fds <= 0) + return 0; + + assert(fds); + + /* Drops/Sets O_NONBLOCK and FD_CLOEXEC from the file flags */ + + for (i = 0; i < n_fds; i++) { + + if ((r = fd_nonblock(fds[i], nonblock)) < 0) + return r; + + /* We unconditionally drop FD_CLOEXEC from the fds, + * since after all we want to pass these fds to our + * children */ + + if ((r = fd_cloexec(fds[i], false)) < 0) + return r; + } + + return 0; +} + +_pure_ static const char *tty_path(const ExecContext *context) { + assert(context); + + if (context->tty_path) + return context->tty_path; + + return "/dev/console"; +} + +void exec_context_tty_reset(const ExecContext *context) { + assert(context); + + if (context->tty_vhangup) + terminal_vhangup(tty_path(context)); + + if (context->tty_reset) + reset_terminal(tty_path(context)); + + if (context->tty_vt_disallocate && context->tty_path) + vt_disallocate(context->tty_path); +} + +static bool is_terminal_output(ExecOutput o) { + return + o == EXEC_OUTPUT_TTY || + o == EXEC_OUTPUT_SYSLOG_AND_CONSOLE || + o == EXEC_OUTPUT_KMSG_AND_CONSOLE || + o == EXEC_OUTPUT_JOURNAL_AND_CONSOLE; +} + +void exec_context_serialize(const ExecContext *context, Unit *u, FILE *f) { + assert(context); + assert(u); + assert(f); + + if (context->tmp_dir) + unit_serialize_item(u, f, "tmp-dir", context->tmp_dir); + + if (context->var_tmp_dir) + unit_serialize_item(u, f, "var-tmp-dir", context->var_tmp_dir); +} + +static int open_null_as(int flags, int nfd) { + int fd, r; + + assert(nfd >= 0); + + if ((fd = open("/dev/null", flags|O_NOCTTY)) < 0) + return -errno; + + if (fd != nfd) { + r = dup2(fd, nfd) < 0 ? -errno : nfd; + close_nointr_nofail(fd); + } else + r = nfd; + + return r; +} + +static int connect_logger_as(const ExecContext *context, ExecOutput output, const char *ident, const char *unit_id, int nfd) { + int fd, r; + union sockaddr_union sa = { + .un.sun_family = AF_UNIX, + .un.sun_path = "/run/systemd/journal/stdout", + }; + + assert(context); + assert(output < _EXEC_OUTPUT_MAX); + assert(ident); + assert(nfd >= 0); + + fd = socket(AF_UNIX, SOCK_STREAM, 0); + if (fd < 0) + return -errno; + + r = connect(fd, &sa.sa, offsetof(struct sockaddr_un, sun_path) + strlen(sa.un.sun_path)); + if (r < 0) { + close_nointr_nofail(fd); + return -errno; + } + + if (shutdown(fd, SHUT_RD) < 0) { + close_nointr_nofail(fd); + return -errno; + } + + dprintf(fd, + "%s\n" + "%s\n" + "%i\n" + "%i\n" + "%i\n" + "%i\n" + "%i\n", + context->syslog_identifier ? context->syslog_identifier : ident, + unit_id, + context->syslog_priority, + !!context->syslog_level_prefix, + output == EXEC_OUTPUT_SYSLOG || output == EXEC_OUTPUT_SYSLOG_AND_CONSOLE, + output == EXEC_OUTPUT_KMSG || output == EXEC_OUTPUT_KMSG_AND_CONSOLE, + is_terminal_output(output)); + + if (fd != nfd) { + r = dup2(fd, nfd) < 0 ? -errno : nfd; + close_nointr_nofail(fd); + } else + r = nfd; + + return r; +} +static int open_terminal_as(const char *path, mode_t mode, int nfd) { + int fd, r; + + assert(path); + assert(nfd >= 0); + + if ((fd = open_terminal(path, mode | O_NOCTTY)) < 0) + return fd; + + if (fd != nfd) { + r = dup2(fd, nfd) < 0 ? -errno : nfd; + close_nointr_nofail(fd); + } else + r = nfd; + + return r; +} + +static bool is_terminal_input(ExecInput i) { + return + i == EXEC_INPUT_TTY || + i == EXEC_INPUT_TTY_FORCE || + i == EXEC_INPUT_TTY_FAIL; +} + +static int fixup_input(ExecInput std_input, int socket_fd, bool apply_tty_stdin) { + + if (is_terminal_input(std_input) && !apply_tty_stdin) + return EXEC_INPUT_NULL; + + if (std_input == EXEC_INPUT_SOCKET && socket_fd < 0) + return EXEC_INPUT_NULL; + + return std_input; +} + +static int fixup_output(ExecOutput std_output, int socket_fd) { + + if (std_output == EXEC_OUTPUT_SOCKET && socket_fd < 0) + return EXEC_OUTPUT_INHERIT; + + return std_output; +} + +static int setup_input(const ExecContext *context, int socket_fd, bool apply_tty_stdin) { + ExecInput i; + + assert(context); + + i = fixup_input(context->std_input, socket_fd, apply_tty_stdin); + + switch (i) { + + case EXEC_INPUT_NULL: + return open_null_as(O_RDONLY, STDIN_FILENO); + + case EXEC_INPUT_TTY: + case EXEC_INPUT_TTY_FORCE: + case EXEC_INPUT_TTY_FAIL: { + int fd, r; + + if ((fd = acquire_terminal( + tty_path(context), + i == EXEC_INPUT_TTY_FAIL, + i == EXEC_INPUT_TTY_FORCE, + false, + (usec_t) -1)) < 0) + return fd; + + if (fd != STDIN_FILENO) { + r = dup2(fd, STDIN_FILENO) < 0 ? -errno : STDIN_FILENO; + close_nointr_nofail(fd); + } else + r = STDIN_FILENO; + + return r; + } + + case EXEC_INPUT_SOCKET: + return dup2(socket_fd, STDIN_FILENO) < 0 ? -errno : STDIN_FILENO; + + default: + assert_not_reached("Unknown input type"); + } +} + +static int setup_output(const ExecContext *context, int fileno, int socket_fd, const char *ident, const char *unit_id, bool apply_tty_stdin) { + ExecOutput o; + ExecInput i; + int r; + + assert(context); + assert(ident); + + i = fixup_input(context->std_input, socket_fd, apply_tty_stdin); + o = fixup_output(context->std_output, socket_fd); + + if (fileno == STDERR_FILENO) { + ExecOutput e; + e = fixup_output(context->std_error, socket_fd); + + /* This expects the input and output are already set up */ + + /* Don't change the stderr file descriptor if we inherit all + * the way and are not on a tty */ + if (e == EXEC_OUTPUT_INHERIT && + o == EXEC_OUTPUT_INHERIT && + i == EXEC_INPUT_NULL && + !is_terminal_input(context->std_input) && + getppid () != 1) + return fileno; + + /* Duplicate from stdout if possible */ + if (e == o || e == EXEC_OUTPUT_INHERIT) + return dup2(STDOUT_FILENO, fileno) < 0 ? -errno : fileno; + + o = e; + + } else if (o == EXEC_OUTPUT_INHERIT) { + /* If input got downgraded, inherit the original value */ + if (i == EXEC_INPUT_NULL && is_terminal_input(context->std_input)) + return open_terminal_as(tty_path(context), O_WRONLY, fileno); + + /* If the input is connected to anything that's not a /dev/null, inherit that... */ + if (i != EXEC_INPUT_NULL) + return dup2(STDIN_FILENO, fileno) < 0 ? -errno : fileno; + + /* If we are not started from PID 1 we just inherit STDOUT from our parent process. */ + if (getppid() != 1) + return fileno; + + /* We need to open /dev/null here anew, to get the right access mode. */ + return open_null_as(O_WRONLY, fileno); + } + + switch (o) { + + case EXEC_OUTPUT_NULL: + return open_null_as(O_WRONLY, fileno); + + case EXEC_OUTPUT_TTY: + if (is_terminal_input(i)) + return dup2(STDIN_FILENO, fileno) < 0 ? -errno : fileno; + + /* We don't reset the terminal if this is just about output */ + return open_terminal_as(tty_path(context), O_WRONLY, fileno); + + case EXEC_OUTPUT_SYSLOG: + case EXEC_OUTPUT_SYSLOG_AND_CONSOLE: + case EXEC_OUTPUT_KMSG: + case EXEC_OUTPUT_KMSG_AND_CONSOLE: + case EXEC_OUTPUT_JOURNAL: + case EXEC_OUTPUT_JOURNAL_AND_CONSOLE: + r = connect_logger_as(context, o, ident, unit_id, fileno); + if (r < 0) { + log_struct_unit(LOG_CRIT, unit_id, + "MESSAGE=Failed to connect std%s of %s to the journal socket: %s", + fileno == STDOUT_FILENO ? "out" : "err", + unit_id, strerror(-r), + "ERRNO=%d", -r, + NULL); + r = open_null_as(O_WRONLY, fileno); + } + return r; + + case EXEC_OUTPUT_SOCKET: + assert(socket_fd >= 0); + return dup2(socket_fd, fileno) < 0 ? -errno : fileno; + + default: + assert_not_reached("Unknown error type"); + } +} + +static int chown_terminal(int fd, uid_t uid) { + struct stat st; + + assert(fd >= 0); + + /* This might fail. What matters are the results. */ + (void) fchown(fd, uid, -1); + (void) fchmod(fd, TTY_MODE); + + if (fstat(fd, &st) < 0) + return -errno; + + if (st.st_uid != uid || (st.st_mode & 0777) != TTY_MODE) + return -EPERM; + + return 0; +} + +static int setup_confirm_stdio(int *_saved_stdin, + int *_saved_stdout) { + int fd = -1, saved_stdin, saved_stdout = -1, r; + + assert(_saved_stdin); + assert(_saved_stdout); + + saved_stdin = fcntl(STDIN_FILENO, F_DUPFD, 3); + if (saved_stdin < 0) + return -errno; + + saved_stdout = fcntl(STDOUT_FILENO, F_DUPFD, 3); + if (saved_stdout < 0) { + r = errno; + goto fail; + } + + fd = acquire_terminal( + "/dev/console", + false, + false, + false, + DEFAULT_CONFIRM_USEC); + if (fd < 0) { + r = fd; + goto fail; + } + + r = chown_terminal(fd, getuid()); + if (r < 0) + goto fail; + + if (dup2(fd, STDIN_FILENO) < 0) { + r = -errno; + goto fail; + } + + if (dup2(fd, STDOUT_FILENO) < 0) { + r = -errno; + goto fail; + } + + if (fd >= 2) + close_nointr_nofail(fd); + + *_saved_stdin = saved_stdin; + *_saved_stdout = saved_stdout; + + return 0; + +fail: + if (saved_stdout >= 0) + close_nointr_nofail(saved_stdout); + + if (saved_stdin >= 0) + close_nointr_nofail(saved_stdin); + + if (fd >= 0) + close_nointr_nofail(fd); + + return r; +} + +_printf_attr_(1, 2) static int write_confirm_message(const char *format, ...) { + int fd; + va_list ap; + + assert(format); + + fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC); + if (fd < 0) + return fd; + + va_start(ap, format); + vdprintf(fd, format, ap); + va_end(ap); + + close_nointr_nofail(fd); + + return 0; +} + +static int restore_confirm_stdio(int *saved_stdin, + int *saved_stdout) { + + int r = 0; + + assert(saved_stdin); + assert(saved_stdout); + + release_terminal(); + + if (*saved_stdin >= 0) + if (dup2(*saved_stdin, STDIN_FILENO) < 0) + r = -errno; + + if (*saved_stdout >= 0) + if (dup2(*saved_stdout, STDOUT_FILENO) < 0) + r = -errno; + + if (*saved_stdin >= 0) + close_nointr_nofail(*saved_stdin); + + if (*saved_stdout >= 0) + close_nointr_nofail(*saved_stdout); + + return r; +} + +static int ask_for_confirmation(char *response, char **argv) { + int saved_stdout = -1, saved_stdin = -1, r; + char *line; + + r = setup_confirm_stdio(&saved_stdin, &saved_stdout); + if (r < 0) + return r; + + line = exec_command_line(argv); + if (!line) + return -ENOMEM; + + r = ask(response, "yns", "Execute %s? [Yes, No, Skip] ", line); + free(line); + + restore_confirm_stdio(&saved_stdin, &saved_stdout); + + return r; +} + +static int enforce_groups(const ExecContext *context, const char *username, gid_t gid) { + bool keep_groups = false; + int r; + + assert(context); + + /* Lookup and set GID and supplementary group list. Here too + * we avoid NSS lookups for gid=0. */ + + if (context->group || username) { + + if (context->group) { + const char *g = context->group; + + if ((r = get_group_creds(&g, &gid)) < 0) + return r; + } + + /* First step, initialize groups from /etc/groups */ + if (username && gid != 0) { + if (initgroups(username, gid) < 0) + return -errno; + + keep_groups = true; + } + + /* Second step, set our gids */ + if (setresgid(gid, gid, gid) < 0) + return -errno; + } + + if (context->supplementary_groups) { + int ngroups_max, k; + gid_t *gids; + char **i; + + /* Final step, initialize any manually set supplementary groups */ + assert_se((ngroups_max = (int) sysconf(_SC_NGROUPS_MAX)) > 0); + + if (!(gids = new(gid_t, ngroups_max))) + return -ENOMEM; + + if (keep_groups) { + if ((k = getgroups(ngroups_max, gids)) < 0) { + free(gids); + return -errno; + } + } else + k = 0; + + STRV_FOREACH(i, context->supplementary_groups) { + const char *g; + + if (k >= ngroups_max) { + free(gids); + return -E2BIG; + } + + g = *i; + r = get_group_creds(&g, gids+k); + if (r < 0) { + free(gids); + return r; + } + + k++; + } + + if (setgroups(k, gids) < 0) { + free(gids); + return -errno; + } + + free(gids); + } + + return 0; +} + +static int enforce_user(const ExecContext *context, uid_t uid) { + int r; + assert(context); + + /* Sets (but doesn't lookup) the uid and make sure we keep the + * capabilities while doing so. */ + + if (context->capabilities) { + cap_t d; + static const cap_value_t bits[] = { + CAP_SETUID, /* Necessary so that we can run setresuid() below */ + CAP_SETPCAP /* Necessary so that we can set PR_SET_SECUREBITS later on */ + }; + + /* First step: If we need to keep capabilities but + * drop privileges we need to make sure we keep our + * caps, while we drop privileges. */ + if (uid != 0) { + int sb = context->secure_bits | 1<<SECURE_KEEP_CAPS; + + if (prctl(PR_GET_SECUREBITS) != sb) + if (prctl(PR_SET_SECUREBITS, sb) < 0) + return -errno; + } + + /* Second step: set the capabilities. This will reduce + * the capabilities to the minimum we need. */ + + if (!(d = cap_dup(context->capabilities))) + return -errno; + + if (cap_set_flag(d, CAP_EFFECTIVE, ELEMENTSOF(bits), bits, CAP_SET) < 0 || + cap_set_flag(d, CAP_PERMITTED, ELEMENTSOF(bits), bits, CAP_SET) < 0) { + r = -errno; + cap_free(d); + return r; + } + + if (cap_set_proc(d) < 0) { + r = -errno; + cap_free(d); + return r; + } + + cap_free(d); + } + + /* Third step: actually set the uids */ + if (setresuid(uid, uid, uid) < 0) + return -errno; + + /* At this point we should have all necessary capabilities but + are otherwise a normal user. However, the caps might got + corrupted due to the setresuid() so we need clean them up + later. This is done outside of this call. */ + + return 0; +} + +#ifdef HAVE_PAM + +static int null_conv( + int num_msg, + const struct pam_message **msg, + struct pam_response **resp, + void *appdata_ptr) { + + /* We don't support conversations */ + + return PAM_CONV_ERR; +} + +static int setup_pam( + const char *name, + const char *user, + uid_t uid, + const char *tty, + char ***pam_env, + int fds[], unsigned n_fds) { + + static const struct pam_conv conv = { + .conv = null_conv, + .appdata_ptr = NULL + }; + + pam_handle_t *handle = NULL; + sigset_t ss, old_ss; + int pam_code = PAM_SUCCESS; + int err; + char **e = NULL; + bool close_session = false; + pid_t pam_pid = 0, parent_pid; + + assert(name); + assert(user); + assert(pam_env); + + /* We set up PAM in the parent process, then fork. The child + * will then stay around until killed via PR_GET_PDEATHSIG or + * systemd via the cgroup logic. It will then remove the PAM + * session again. The parent process will exec() the actual + * daemon. We do things this way to ensure that the main PID + * of the daemon is the one we initially fork()ed. */ + + if ((pam_code = pam_start(name, user, &conv, &handle)) != PAM_SUCCESS) { + handle = NULL; + goto fail; + } + + if (tty) + if ((pam_code = pam_set_item(handle, PAM_TTY, tty)) != PAM_SUCCESS) + goto fail; + + if ((pam_code = pam_acct_mgmt(handle, PAM_SILENT)) != PAM_SUCCESS) + goto fail; + + if ((pam_code = pam_open_session(handle, PAM_SILENT)) != PAM_SUCCESS) + goto fail; + + close_session = true; + + if ((!(e = pam_getenvlist(handle)))) { + pam_code = PAM_BUF_ERR; + goto fail; + } + + /* Block SIGTERM, so that we know that it won't get lost in + * the child */ + if (sigemptyset(&ss) < 0 || + sigaddset(&ss, SIGTERM) < 0 || + sigprocmask(SIG_BLOCK, &ss, &old_ss) < 0) + goto fail; + + parent_pid = getpid(); + + if ((pam_pid = fork()) < 0) + goto fail; + + if (pam_pid == 0) { + int sig; + int r = EXIT_PAM; + + /* The child's job is to reset the PAM session on + * termination */ + + /* This string must fit in 10 chars (i.e. the length + * of "/sbin/init"), to look pretty in /bin/ps */ + rename_process("(sd-pam)"); + + /* Make sure we don't keep open the passed fds in this + child. We assume that otherwise only those fds are + open here that have been opened by PAM. */ + close_many(fds, n_fds); + + /* Drop privileges - we don't need any to pam_close_session + * and this will make PR_SET_PDEATHSIG work in most cases. + * If this fails, ignore the error - but expect sd-pam threads + * to fail to exit normally */ + if (setresuid(uid, uid, uid) < 0) + log_error("Error: Failed to setresuid() in sd-pam: %s", strerror(-r)); + + /* Wait until our parent died. This will only work if + * the above setresuid() succeeds, otherwise the kernel + * will not allow unprivileged parents kill their privileged + * children this way. We rely on the control groups kill logic + * to do the rest for us. */ + if (prctl(PR_SET_PDEATHSIG, SIGTERM) < 0) + goto child_finish; + + /* Check if our parent process might already have + * died? */ + if (getppid() == parent_pid) { + for (;;) { + if (sigwait(&ss, &sig) < 0) { + if (errno == EINTR) + continue; + + goto child_finish; + } + + assert(sig == SIGTERM); + break; + } + } + + /* If our parent died we'll end the session */ + if (getppid() != parent_pid) + if ((pam_code = pam_close_session(handle, PAM_DATA_SILENT)) != PAM_SUCCESS) + goto child_finish; + + r = 0; + + child_finish: + pam_end(handle, pam_code | PAM_DATA_SILENT); + _exit(r); + } + + /* If the child was forked off successfully it will do all the + * cleanups, so forget about the handle here. */ + handle = NULL; + + /* Unblock SIGTERM again in the parent */ + if (sigprocmask(SIG_SETMASK, &old_ss, NULL) < 0) + goto fail; + + /* We close the log explicitly here, since the PAM modules + * might have opened it, but we don't want this fd around. */ + closelog(); + + *pam_env = e; + e = NULL; + + return 0; + +fail: + if (pam_code != PAM_SUCCESS) + err = -EPERM; /* PAM errors do not map to errno */ + else + err = -errno; + + if (handle) { + if (close_session) + pam_code = pam_close_session(handle, PAM_DATA_SILENT); + + pam_end(handle, pam_code | PAM_DATA_SILENT); + } + + strv_free(e); + + closelog(); + + if (pam_pid > 1) { + kill(pam_pid, SIGTERM); + kill(pam_pid, SIGCONT); + } + + return err; +} +#endif + +static void rename_process_from_path(const char *path) { + char process_name[11]; + const char *p; + size_t l; + + /* This resulting string must fit in 10 chars (i.e. the length + * of "/sbin/init") to look pretty in /bin/ps */ + + p = path_get_file_name(path); + if (isempty(p)) { + rename_process("(...)"); + return; + } + + l = strlen(p); + if (l > 8) { + /* The end of the process name is usually more + * interesting, since the first bit might just be + * "systemd-" */ + p = p + l - 8; + l = 8; + } + + process_name[0] = '('; + memcpy(process_name+1, p, l); + process_name[1+l] = ')'; + process_name[1+l+1] = 0; + + rename_process(process_name); +} + +static int apply_seccomp(uint32_t *syscall_filter) { + static const struct sock_filter header[] = { + VALIDATE_ARCHITECTURE, + EXAMINE_SYSCALL + }; + static const struct sock_filter footer[] = { + _KILL_PROCESS + }; + + int i; + unsigned n; + struct sock_filter *f; + struct sock_fprog prog = {}; + + assert(syscall_filter); + + /* First: count the syscalls to check for */ + for (i = 0, n = 0; i < syscall_max(); i++) + if (syscall_filter[i >> 4] & (1 << (i & 31))) + n++; + + /* Second: build the filter program from a header the syscall + * matches and the footer */ + f = alloca(sizeof(struct sock_filter) * (ELEMENTSOF(header) + 2*n + ELEMENTSOF(footer))); + memcpy(f, header, sizeof(header)); + + for (i = 0, n = 0; i < syscall_max(); i++) + if (syscall_filter[i >> 4] & (1 << (i & 31))) { + struct sock_filter item[] = { + BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, INDEX_TO_SYSCALL(i), 0, 1), + BPF_STMT(BPF_RET+BPF_K, SECCOMP_RET_ALLOW) + }; + + assert_cc(ELEMENTSOF(item) == 2); + + f[ELEMENTSOF(header) + 2*n] = item[0]; + f[ELEMENTSOF(header) + 2*n+1] = item[1]; + + n++; + } + + memcpy(f + (ELEMENTSOF(header) + 2*n), footer, sizeof(footer)); + + /* Third: install the filter */ + prog.len = ELEMENTSOF(header) + ELEMENTSOF(footer) + 2*n; + prog.filter = f; + if (prctl(PR_SET_SECCOMP, SECCOMP_MODE_FILTER, &prog) < 0) + return -errno; + + return 0; +} + +int exec_spawn(ExecCommand *command, + char **argv, + ExecContext *context, + int fds[], unsigned n_fds, + char **environment, + bool apply_permissions, + bool apply_chroot, + bool apply_tty_stdin, + bool confirm_spawn, + CGroupBonding *cgroup_bondings, + CGroupAttribute *cgroup_attributes, + const char *cgroup_suffix, + const char *unit_id, + int idle_pipe[2], + pid_t *ret) { + + pid_t pid; + int r; + char *line; + int socket_fd; + _cleanup_strv_free_ char **files_env = NULL; + + assert(command); + assert(context); + assert(ret); + assert(fds || n_fds <= 0); + + if (context->std_input == EXEC_INPUT_SOCKET || + context->std_output == EXEC_OUTPUT_SOCKET || + context->std_error == EXEC_OUTPUT_SOCKET) { + + if (n_fds != 1) + return -EINVAL; + + socket_fd = fds[0]; + + fds = NULL; + n_fds = 0; + } else + socket_fd = -1; + + r = exec_context_load_environment(context, &files_env); + if (r < 0) { + log_struct_unit(LOG_ERR, + unit_id, + "MESSAGE=Failed to load environment files: %s", strerror(-r), + "ERRNO=%d", -r, + NULL); + return r; + } + + if (!argv) + argv = command->argv; + + line = exec_command_line(argv); + if (!line) + return log_oom(); + + log_struct_unit(LOG_DEBUG, + unit_id, + "EXECUTABLE=%s", command->path, + "MESSAGE=About to execute: %s", line, + NULL); + free(line); + + r = cgroup_bonding_realize_list(cgroup_bondings); + if (r < 0) + return r; + + /* We must initialize the attributes in the parent, before we + fork, because we really need them initialized before making + the process a member of the group (which we do in both the + child and the parent), and we cannot really apply them twice + (due to 'append' style attributes) */ + cgroup_attribute_apply_list(cgroup_attributes, cgroup_bondings); + + if (context->private_tmp && !context->tmp_dir && !context->var_tmp_dir) { + r = setup_tmpdirs(&context->tmp_dir, &context->var_tmp_dir); + if (r < 0) + return r; + } + + pid = fork(); + if (pid < 0) + return -errno; + + if (pid == 0) { + int i, err; + sigset_t ss; + const char *username = NULL, *home = NULL; + uid_t uid = (uid_t) -1; + gid_t gid = (gid_t) -1; + _cleanup_strv_free_ char **our_env = NULL, **pam_env = NULL, + **final_env = NULL, **final_argv = NULL; + unsigned n_env = 0; + bool set_access = false; + + /* child */ + + rename_process_from_path(command->path); + + /* We reset exactly these signals, since they are the + * only ones we set to SIG_IGN in the main daemon. All + * others we leave untouched because we set them to + * SIG_DFL or a valid handler initially, both of which + * will be demoted to SIG_DFL. */ + default_signals(SIGNALS_CRASH_HANDLER, + SIGNALS_IGNORE, -1); + + if (context->ignore_sigpipe) + ignore_signals(SIGPIPE, -1); + + assert_se(sigemptyset(&ss) == 0); + if (sigprocmask(SIG_SETMASK, &ss, NULL) < 0) { + err = -errno; + r = EXIT_SIGNAL_MASK; + goto fail_child; + } + + if (idle_pipe) { + if (idle_pipe[1] >= 0) + close_nointr_nofail(idle_pipe[1]); + if (idle_pipe[0] >= 0) { + fd_wait_for_event(idle_pipe[0], POLLHUP, IDLE_TIMEOUT_USEC); + close_nointr_nofail(idle_pipe[0]); + } + } + + /* Close sockets very early to make sure we don't + * block init reexecution because it cannot bind its + * sockets */ + log_forget_fds(); + err = close_all_fds(socket_fd >= 0 ? &socket_fd : fds, + socket_fd >= 0 ? 1 : n_fds); + if (err < 0) { + r = EXIT_FDS; + goto fail_child; + } + + if (!context->same_pgrp) + if (setsid() < 0) { + err = -errno; + r = EXIT_SETSID; + goto fail_child; + } + + if (context->tcpwrap_name) { + if (socket_fd >= 0) + if (!socket_tcpwrap(socket_fd, context->tcpwrap_name)) { + err = -EACCES; + r = EXIT_TCPWRAP; + goto fail_child; + } + + for (i = 0; i < (int) n_fds; i++) { + if (!socket_tcpwrap(fds[i], context->tcpwrap_name)) { + err = -EACCES; + r = EXIT_TCPWRAP; + goto fail_child; + } + } + } + + exec_context_tty_reset(context); + + if (confirm_spawn) { + char response; + + err = ask_for_confirmation(&response, argv); + if (err == -ETIMEDOUT) + write_confirm_message("Confirmation question timed out, assuming positive response.\n"); + else if (err < 0) + write_confirm_message("Couldn't ask confirmation question, assuming positive response: %s\n", strerror(-err)); + else if (response == 's') { + write_confirm_message("Skipping execution.\n"); + err = -ECANCELED; + r = EXIT_CONFIRM; + goto fail_child; + } else if (response == 'n') { + write_confirm_message("Failing execution.\n"); + err = r = 0; + goto fail_child; + } + } + + /* If a socket is connected to STDIN/STDOUT/STDERR, we + * must sure to drop O_NONBLOCK */ + if (socket_fd >= 0) + fd_nonblock(socket_fd, false); + + err = setup_input(context, socket_fd, apply_tty_stdin); + if (err < 0) { + r = EXIT_STDIN; + goto fail_child; + } + + err = setup_output(context, STDOUT_FILENO, socket_fd, path_get_file_name(command->path), unit_id, apply_tty_stdin); + if (err < 0) { + r = EXIT_STDOUT; + goto fail_child; + } + + err = setup_output(context, STDERR_FILENO, socket_fd, path_get_file_name(command->path), unit_id, apply_tty_stdin); + if (err < 0) { + r = EXIT_STDERR; + goto fail_child; + } + + if (cgroup_bondings) { + err = cgroup_bonding_install_list(cgroup_bondings, 0, cgroup_suffix); + if (err < 0) { + r = EXIT_CGROUP; + goto fail_child; + } + } + + if (context->oom_score_adjust_set) { + char t[16]; + + snprintf(t, sizeof(t), "%i", context->oom_score_adjust); + char_array_0(t); + + if (write_string_file("/proc/self/oom_score_adj", t) < 0) { + err = -errno; + r = EXIT_OOM_ADJUST; + goto fail_child; + } + } + + if (context->nice_set) + if (setpriority(PRIO_PROCESS, 0, context->nice) < 0) { + err = -errno; + r = EXIT_NICE; + goto fail_child; + } + + if (context->cpu_sched_set) { + struct sched_param param = { + .sched_priority = context->cpu_sched_priority, + }; + + r = sched_setscheduler(0, + context->cpu_sched_policy | + (context->cpu_sched_reset_on_fork ? + SCHED_RESET_ON_FORK : 0), + ¶m); + if (r < 0) { + err = -errno; + r = EXIT_SETSCHEDULER; + goto fail_child; + } + } + + if (context->cpuset) + if (sched_setaffinity(0, CPU_ALLOC_SIZE(context->cpuset_ncpus), context->cpuset) < 0) { + err = -errno; + r = EXIT_CPUAFFINITY; + goto fail_child; + } + + if (context->ioprio_set) + if (ioprio_set(IOPRIO_WHO_PROCESS, 0, context->ioprio) < 0) { + err = -errno; + r = EXIT_IOPRIO; + goto fail_child; + } + + if (context->timer_slack_nsec != (nsec_t) -1) + if (prctl(PR_SET_TIMERSLACK, context->timer_slack_nsec) < 0) { + err = -errno; + r = EXIT_TIMERSLACK; + goto fail_child; + } + + if (context->utmp_id) + utmp_put_init_process(context->utmp_id, getpid(), getsid(0), context->tty_path); + + if (context->user) { + username = context->user; + err = get_user_creds(&username, &uid, &gid, &home, NULL); + if (err < 0) { + r = EXIT_USER; + goto fail_child; + } + + if (is_terminal_input(context->std_input)) { + err = chown_terminal(STDIN_FILENO, uid); + if (err < 0) { + r = EXIT_STDIN; + goto fail_child; + } + } + + if (cgroup_bondings && context->control_group_modify) { + err = cgroup_bonding_set_group_access_list(cgroup_bondings, 0755, uid, gid); + if (err >= 0) + err = cgroup_bonding_set_task_access_list( + cgroup_bondings, + 0644, + uid, + gid, + context->control_group_persistent); + if (err < 0) { + r = EXIT_CGROUP; + goto fail_child; + } + + set_access = true; + } + } + + if (cgroup_bondings && !set_access && context->control_group_persistent >= 0) { + err = cgroup_bonding_set_task_access_list( + cgroup_bondings, + (mode_t) -1, + (uid_t) -1, + (uid_t) -1, + context->control_group_persistent); + if (err < 0) { + r = EXIT_CGROUP; + goto fail_child; + } + } + + if (apply_permissions) { + err = enforce_groups(context, username, gid); + if (err < 0) { + r = EXIT_GROUP; + goto fail_child; + } + } + + umask(context->umask); + +#ifdef HAVE_PAM + if (apply_permissions && context->pam_name && username) { + err = setup_pam(context->pam_name, username, uid, context->tty_path, &pam_env, fds, n_fds); + if (err < 0) { + r = EXIT_PAM; + goto fail_child; + } + } +#endif + if (context->private_network) { + if (unshare(CLONE_NEWNET) < 0) { + err = -errno; + r = EXIT_NETWORK; + goto fail_child; + } + + loopback_setup(); + } + + if (strv_length(context->read_write_dirs) > 0 || + strv_length(context->read_only_dirs) > 0 || + strv_length(context->inaccessible_dirs) > 0 || + context->mount_flags != 0 || + context->private_tmp) { + err = setup_namespace(context->read_write_dirs, + context->read_only_dirs, + context->inaccessible_dirs, + context->tmp_dir, + context->var_tmp_dir, + context->private_tmp, + context->mount_flags); + if (err < 0) { + r = EXIT_NAMESPACE; + goto fail_child; + } + } + + if (apply_chroot) { + if (context->root_directory) + if (chroot(context->root_directory) < 0) { + err = -errno; + r = EXIT_CHROOT; + goto fail_child; + } + + if (chdir(context->working_directory ? context->working_directory : "/") < 0) { + err = -errno; + r = EXIT_CHDIR; + goto fail_child; + } + } else { + _cleanup_free_ char *d = NULL; + + if (asprintf(&d, "%s/%s", + context->root_directory ? context->root_directory : "", + context->working_directory ? context->working_directory : "") < 0) { + err = -ENOMEM; + r = EXIT_MEMORY; + goto fail_child; + } + + if (chdir(d) < 0) { + err = -errno; + r = EXIT_CHDIR; + goto fail_child; + } + } + + /* We repeat the fd closing here, to make sure that + * nothing is leaked from the PAM modules */ + err = close_all_fds(fds, n_fds); + if (err >= 0) + err = shift_fds(fds, n_fds); + if (err >= 0) + err = flags_fds(fds, n_fds, context->non_blocking); + if (err < 0) { + r = EXIT_FDS; + goto fail_child; + } + + if (apply_permissions) { + + for (i = 0; i < RLIMIT_NLIMITS; i++) { + if (!context->rlimit[i]) + continue; + + if (setrlimit_closest(i, context->rlimit[i]) < 0) { + err = -errno; + r = EXIT_LIMITS; + goto fail_child; + } + } + + if (context->capability_bounding_set_drop) { + err = capability_bounding_set_drop(context->capability_bounding_set_drop, false); + if (err < 0) { + r = EXIT_CAPABILITIES; + goto fail_child; + } + } + + if (context->user) { + err = enforce_user(context, uid); + if (err < 0) { + r = EXIT_USER; + goto fail_child; + } + } + + /* PR_GET_SECUREBITS is not privileged, while + * PR_SET_SECUREBITS is. So to suppress + * potential EPERMs we'll try not to call + * PR_SET_SECUREBITS unless necessary. */ + if (prctl(PR_GET_SECUREBITS) != context->secure_bits) + if (prctl(PR_SET_SECUREBITS, context->secure_bits) < 0) { + err = -errno; + r = EXIT_SECUREBITS; + goto fail_child; + } + + if (context->capabilities) + if (cap_set_proc(context->capabilities) < 0) { + err = -errno; + r = EXIT_CAPABILITIES; + goto fail_child; + } + + if (context->no_new_privileges) + if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) { + err = -errno; + r = EXIT_NO_NEW_PRIVILEGES; + goto fail_child; + } + + if (context->syscall_filter) { + err = apply_seccomp(context->syscall_filter); + if (err < 0) { + r = EXIT_SECCOMP; + goto fail_child; + } + } + } + + our_env = new0(char*, 7); + if (!our_env) { + err = -ENOMEM; + r = EXIT_MEMORY; + goto fail_child; + } + + if (n_fds > 0) + if (asprintf(our_env + n_env++, "LISTEN_PID=%lu", (unsigned long) getpid()) < 0 || + asprintf(our_env + n_env++, "LISTEN_FDS=%u", n_fds) < 0) { + err = -ENOMEM; + r = EXIT_MEMORY; + goto fail_child; + } + + if (home) + if (asprintf(our_env + n_env++, "HOME=%s", home) < 0) { + err = -ENOMEM; + r = EXIT_MEMORY; + goto fail_child; + } + + if (username) + if (asprintf(our_env + n_env++, "LOGNAME=%s", username) < 0 || + asprintf(our_env + n_env++, "USER=%s", username) < 0) { + err = -ENOMEM; + r = EXIT_MEMORY; + goto fail_child; + } + + if (is_terminal_input(context->std_input) || + context->std_output == EXEC_OUTPUT_TTY || + context->std_error == EXEC_OUTPUT_TTY) + if (!(our_env[n_env++] = strdup(default_term_for_tty(tty_path(context))))) { + err = -ENOMEM; + r = EXIT_MEMORY; + goto fail_child; + } + + assert(n_env <= 7); + + final_env = strv_env_merge(5, + environment, + our_env, + context->environment, + files_env, + pam_env, + NULL); + if (!final_env) { + err = -ENOMEM; + r = EXIT_MEMORY; + goto fail_child; + } + + final_argv = replace_env_argv(argv, final_env); + if (!final_argv) { + err = -ENOMEM; + r = EXIT_MEMORY; + goto fail_child; + } + + final_env = strv_env_clean(final_env); + + if (_unlikely_(log_get_max_level() >= LOG_PRI(LOG_DEBUG))) { + line = exec_command_line(final_argv); + if (line) { + log_open(); + log_struct_unit(LOG_DEBUG, + unit_id, + "EXECUTABLE=%s", command->path, + "MESSAGE=Executing: %s", line, + NULL); + log_close(); + free(line); + line = NULL; + } + } + execve(command->path, final_argv, final_env); + err = -errno; + r = EXIT_EXEC; + + fail_child: + if (r != 0) { + log_open(); + log_struct(LOG_ERR, MESSAGE_ID(SD_MESSAGE_SPAWN_FAILED), + "EXECUTABLE=%s", command->path, + "MESSAGE=Failed at step %s spawning %s: %s", + exit_status_to_string(r, EXIT_STATUS_SYSTEMD), + command->path, strerror(-err), + "ERRNO=%d", -err, + NULL); + log_close(); + } + + _exit(r); + } + + log_struct_unit(LOG_DEBUG, + unit_id, + "MESSAGE=Forked %s as %lu", + command->path, (unsigned long) pid, + NULL); + + /* We add the new process to the cgroup both in the child (so + * that we can be sure that no user code is ever executed + * outside of the cgroup) and in the parent (so that we can be + * sure that when we kill the cgroup the process will be + * killed too). */ + cgroup_bonding_install_list(cgroup_bondings, pid, cgroup_suffix); + + exec_status_start(&command->exec_status, pid); + + *ret = pid; + return 0; +} + +void exec_context_init(ExecContext *c) { + assert(c); + + c->umask = 0022; + c->ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 0); + c->cpu_sched_policy = SCHED_OTHER; + c->syslog_priority = LOG_DAEMON|LOG_INFO; + c->syslog_level_prefix = true; + c->control_group_persistent = -1; + c->ignore_sigpipe = true; + c->timer_slack_nsec = (nsec_t) -1; +} + +void exec_context_tmp_dirs_done(ExecContext *c) { + char* dirs[] = {c->tmp_dir ? c->tmp_dir : c->var_tmp_dir, + c->tmp_dir ? c->var_tmp_dir : NULL, + NULL}; + char **dirp; + + for(dirp = dirs; *dirp; dirp++) { + char *dir; + int r; + + r = rm_rf_dangerous(*dirp, false, true, false); + dir = dirname(*dirp); + if (r < 0) + log_warning("Failed to remove content of temporary directory %s: %s", + dir, strerror(-r)); + else { + r = rmdir(dir); + if (r < 0) + log_warning("Failed to remove temporary directory %s: %s", + dir, strerror(-r)); + } + + free(*dirp); + } + + c->tmp_dir = c->var_tmp_dir = NULL; +} + +void exec_context_done(ExecContext *c, bool reloading_or_reexecuting) { + unsigned l; + + assert(c); + + strv_free(c->environment); + c->environment = NULL; + + strv_free(c->environment_files); + c->environment_files = NULL; + + for (l = 0; l < ELEMENTSOF(c->rlimit); l++) { + free(c->rlimit[l]); + c->rlimit[l] = NULL; + } + + free(c->working_directory); + c->working_directory = NULL; + free(c->root_directory); + c->root_directory = NULL; + + free(c->tty_path); + c->tty_path = NULL; + + free(c->tcpwrap_name); + c->tcpwrap_name = NULL; + + free(c->syslog_identifier); + c->syslog_identifier = NULL; + + free(c->user); + c->user = NULL; + + free(c->group); + c->group = NULL; + + strv_free(c->supplementary_groups); + c->supplementary_groups = NULL; + + free(c->pam_name); + c->pam_name = NULL; + + if (c->capabilities) { + cap_free(c->capabilities); + c->capabilities = NULL; + } + + strv_free(c->read_only_dirs); + c->read_only_dirs = NULL; + + strv_free(c->read_write_dirs); + c->read_write_dirs = NULL; + + strv_free(c->inaccessible_dirs); + c->inaccessible_dirs = NULL; + + if (c->cpuset) + CPU_FREE(c->cpuset); + + free(c->utmp_id); + c->utmp_id = NULL; + + free(c->syscall_filter); + c->syscall_filter = NULL; + + if (!reloading_or_reexecuting) + exec_context_tmp_dirs_done(c); +} + +void exec_command_done(ExecCommand *c) { + assert(c); + + free(c->path); + c->path = NULL; + + strv_free(c->argv); + c->argv = NULL; +} + +void exec_command_done_array(ExecCommand *c, unsigned n) { + unsigned i; + + for (i = 0; i < n; i++) + exec_command_done(c+i); +} + +void exec_command_free_list(ExecCommand *c) { + ExecCommand *i; + + while ((i = c)) { + LIST_REMOVE(ExecCommand, command, c, i); + exec_command_done(i); + free(i); + } +} + +void exec_command_free_array(ExecCommand **c, unsigned n) { + unsigned i; + + for (i = 0; i < n; i++) { + exec_command_free_list(c[i]); + c[i] = NULL; + } +} + +int exec_context_load_environment(const ExecContext *c, char ***l) { + char **i, **r = NULL; + + assert(c); + assert(l); + + STRV_FOREACH(i, c->environment_files) { + char *fn; + int k; + bool ignore = false; + char **p; + _cleanup_globfree_ glob_t pglob = {}; + int count, n; + + fn = *i; + + if (fn[0] == '-') { + ignore = true; + fn ++; + } + + if (!path_is_absolute(fn)) { + if (ignore) + continue; + + strv_free(r); + return -EINVAL; + } + + /* Filename supports globbing, take all matching files */ + errno = 0; + if (glob(fn, 0, NULL, &pglob) != 0) { + if (ignore) + continue; + + strv_free(r); + return errno ? -errno : -EINVAL; + } + count = pglob.gl_pathc; + if (count == 0) { + if (ignore) + continue; + + strv_free(r); + return -EINVAL; + } + for (n = 0; n < count; n++) { + k = load_env_file(pglob.gl_pathv[n], NULL, &p); + if (k < 0) { + if (ignore) + continue; + + strv_free(r); + return k; + } + /* Log invalid environment variables with filename */ + if (p) + p = strv_env_clean_log(p, pglob.gl_pathv[n]); + + if (r == NULL) + r = p; + else { + char **m; + + m = strv_env_merge(2, r, p); + strv_free(r); + strv_free(p); + if (!m) + return -ENOMEM; + + r = m; + } + } + } + + *l = r; + + return 0; +} + +static bool tty_may_match_dev_console(const char *tty) { + char *active = NULL, *console; + bool b; + + if (startswith(tty, "/dev/")) + tty += 5; + + /* trivial identity? */ + if (streq(tty, "console")) + return true; + + console = resolve_dev_console(&active); + /* if we could not resolve, assume it may */ + if (!console) + return true; + + /* "tty0" means the active VC, so it may be the same sometimes */ + b = streq(console, tty) || (streq(console, "tty0") && tty_is_vc(tty)); + free(active); + + return b; +} + +bool exec_context_may_touch_console(ExecContext *ec) { + return (ec->tty_reset || ec->tty_vhangup || ec->tty_vt_disallocate || + is_terminal_input(ec->std_input) || + is_terminal_output(ec->std_output) || + is_terminal_output(ec->std_error)) && + tty_may_match_dev_console(tty_path(ec)); +} + +static void strv_fprintf(FILE *f, char **l) { + char **g; + + assert(f); + + STRV_FOREACH(g, l) + fprintf(f, " %s", *g); +} + +void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { + char ** e; + unsigned i; + + assert(c); + assert(f); + + if (!prefix) + prefix = ""; + + fprintf(f, + "%sUMask: %04o\n" + "%sWorkingDirectory: %s\n" + "%sRootDirectory: %s\n" + "%sNonBlocking: %s\n" + "%sPrivateTmp: %s\n" + "%sControlGroupModify: %s\n" + "%sControlGroupPersistent: %s\n" + "%sPrivateNetwork: %s\n" + "%sIgnoreSIGPIPE: %s\n", + prefix, c->umask, + prefix, c->working_directory ? c->working_directory : "/", + prefix, c->root_directory ? c->root_directory : "/", + prefix, yes_no(c->non_blocking), + prefix, yes_no(c->private_tmp), + prefix, yes_no(c->control_group_modify), + prefix, yes_no(c->control_group_persistent), + prefix, yes_no(c->private_network), + prefix, yes_no(c->ignore_sigpipe)); + + STRV_FOREACH(e, c->environment) + fprintf(f, "%sEnvironment: %s\n", prefix, *e); + + STRV_FOREACH(e, c->environment_files) + fprintf(f, "%sEnvironmentFile: %s\n", prefix, *e); + + if (c->tcpwrap_name) + fprintf(f, + "%sTCPWrapName: %s\n", + prefix, c->tcpwrap_name); + + if (c->nice_set) + fprintf(f, + "%sNice: %i\n", + prefix, c->nice); + + if (c->oom_score_adjust_set) + fprintf(f, + "%sOOMScoreAdjust: %i\n", + prefix, c->oom_score_adjust); + + for (i = 0; i < RLIM_NLIMITS; i++) + if (c->rlimit[i]) + fprintf(f, "%s%s: %llu\n", prefix, rlimit_to_string(i), (unsigned long long) c->rlimit[i]->rlim_max); + + if (c->ioprio_set) { + char *class_str; + int r; + + r = ioprio_class_to_string_alloc(IOPRIO_PRIO_CLASS(c->ioprio), &class_str); + if (r < 0) + class_str = NULL; + fprintf(f, + "%sIOSchedulingClass: %s\n" + "%sIOPriority: %i\n", + prefix, strna(class_str), + prefix, (int) IOPRIO_PRIO_DATA(c->ioprio)); + free(class_str); + } + + if (c->cpu_sched_set) { + char *policy_str; + int r; + + r = sched_policy_to_string_alloc(c->cpu_sched_policy, &policy_str); + if (r < 0) + policy_str = NULL; + fprintf(f, + "%sCPUSchedulingPolicy: %s\n" + "%sCPUSchedulingPriority: %i\n" + "%sCPUSchedulingResetOnFork: %s\n", + prefix, strna(policy_str), + prefix, c->cpu_sched_priority, + prefix, yes_no(c->cpu_sched_reset_on_fork)); + free(policy_str); + } + + if (c->cpuset) { + fprintf(f, "%sCPUAffinity:", prefix); + for (i = 0; i < c->cpuset_ncpus; i++) + if (CPU_ISSET_S(i, CPU_ALLOC_SIZE(c->cpuset_ncpus), c->cpuset)) + fprintf(f, " %i", i); + fputs("\n", f); + } + + if (c->timer_slack_nsec != (nsec_t) -1) + fprintf(f, "%sTimerSlackNSec: %lu\n", prefix, (unsigned long)c->timer_slack_nsec); + + fprintf(f, + "%sStandardInput: %s\n" + "%sStandardOutput: %s\n" + "%sStandardError: %s\n", + prefix, exec_input_to_string(c->std_input), + prefix, exec_output_to_string(c->std_output), + prefix, exec_output_to_string(c->std_error)); + + if (c->tty_path) + fprintf(f, + "%sTTYPath: %s\n" + "%sTTYReset: %s\n" + "%sTTYVHangup: %s\n" + "%sTTYVTDisallocate: %s\n", + prefix, c->tty_path, + prefix, yes_no(c->tty_reset), + prefix, yes_no(c->tty_vhangup), + prefix, yes_no(c->tty_vt_disallocate)); + + if (c->std_output == EXEC_OUTPUT_SYSLOG || c->std_output == EXEC_OUTPUT_KMSG || c->std_output == EXEC_OUTPUT_JOURNAL || + c->std_output == EXEC_OUTPUT_SYSLOG_AND_CONSOLE || c->std_output == EXEC_OUTPUT_KMSG_AND_CONSOLE || c->std_output == EXEC_OUTPUT_JOURNAL_AND_CONSOLE || + c->std_error == EXEC_OUTPUT_SYSLOG || c->std_error == EXEC_OUTPUT_KMSG || c->std_error == EXEC_OUTPUT_JOURNAL || + c->std_error == EXEC_OUTPUT_SYSLOG_AND_CONSOLE || c->std_error == EXEC_OUTPUT_KMSG_AND_CONSOLE || c->std_error == EXEC_OUTPUT_JOURNAL_AND_CONSOLE) { + char *fac_str, *lvl_str; + int r; + + r = log_facility_unshifted_to_string_alloc(c->syslog_priority >> 3, &fac_str); + if (r < 0) + fac_str = NULL; + + r = log_level_to_string_alloc(LOG_PRI(c->syslog_priority), &lvl_str); + if (r < 0) + lvl_str = NULL; + + fprintf(f, + "%sSyslogFacility: %s\n" + "%sSyslogLevel: %s\n", + prefix, strna(fac_str), + prefix, strna(lvl_str)); + free(lvl_str); + free(fac_str); + } + + if (c->capabilities) { + char *t; + if ((t = cap_to_text(c->capabilities, NULL))) { + fprintf(f, "%sCapabilities: %s\n", + prefix, t); + cap_free(t); + } + } + + if (c->secure_bits) + fprintf(f, "%sSecure Bits:%s%s%s%s%s%s\n", + prefix, + (c->secure_bits & 1<<SECURE_KEEP_CAPS) ? " keep-caps" : "", + (c->secure_bits & 1<<SECURE_KEEP_CAPS_LOCKED) ? " keep-caps-locked" : "", + (c->secure_bits & 1<<SECURE_NO_SETUID_FIXUP) ? " no-setuid-fixup" : "", + (c->secure_bits & 1<<SECURE_NO_SETUID_FIXUP_LOCKED) ? " no-setuid-fixup-locked" : "", + (c->secure_bits & 1<<SECURE_NOROOT) ? " noroot" : "", + (c->secure_bits & 1<<SECURE_NOROOT_LOCKED) ? "noroot-locked" : ""); + + if (c->capability_bounding_set_drop) { + unsigned long l; + fprintf(f, "%sCapabilityBoundingSet:", prefix); + + for (l = 0; l <= cap_last_cap(); l++) + if (!(c->capability_bounding_set_drop & ((uint64_t) 1ULL << (uint64_t) l))) { + char *t; + + if ((t = cap_to_name(l))) { + fprintf(f, " %s", t); + cap_free(t); + } + } + + fputs("\n", f); + } + + if (c->user) + fprintf(f, "%sUser: %s\n", prefix, c->user); + if (c->group) + fprintf(f, "%sGroup: %s\n", prefix, c->group); + + if (strv_length(c->supplementary_groups) > 0) { + fprintf(f, "%sSupplementaryGroups:", prefix); + strv_fprintf(f, c->supplementary_groups); + fputs("\n", f); + } + + if (c->pam_name) + fprintf(f, "%sPAMName: %s\n", prefix, c->pam_name); + + if (strv_length(c->read_write_dirs) > 0) { + fprintf(f, "%sReadWriteDirs:", prefix); + strv_fprintf(f, c->read_write_dirs); + fputs("\n", f); + } + + if (strv_length(c->read_only_dirs) > 0) { + fprintf(f, "%sReadOnlyDirs:", prefix); + strv_fprintf(f, c->read_only_dirs); + fputs("\n", f); + } + + if (strv_length(c->inaccessible_dirs) > 0) { + fprintf(f, "%sInaccessibleDirs:", prefix); + strv_fprintf(f, c->inaccessible_dirs); + fputs("\n", f); + } + + if (c->utmp_id) + fprintf(f, + "%sUtmpIdentifier: %s\n", + prefix, c->utmp_id); +} + +void exec_status_start(ExecStatus *s, pid_t pid) { + assert(s); + + zero(*s); + s->pid = pid; + dual_timestamp_get(&s->start_timestamp); +} + +void exec_status_exit(ExecStatus *s, ExecContext *context, pid_t pid, int code, int status) { + assert(s); + + if (s->pid && s->pid != pid) + zero(*s); + + s->pid = pid; + dual_timestamp_get(&s->exit_timestamp); + + s->code = code; + s->status = status; + + if (context) { + if (context->utmp_id) + utmp_put_dead_process(context->utmp_id, pid, code, status); + + exec_context_tty_reset(context); + } +} + +void exec_status_dump(ExecStatus *s, FILE *f, const char *prefix) { + char buf[FORMAT_TIMESTAMP_MAX]; + + assert(s); + assert(f); + + if (!prefix) + prefix = ""; + + if (s->pid <= 0) + return; + + fprintf(f, + "%sPID: %lu\n", + prefix, (unsigned long) s->pid); + + if (s->start_timestamp.realtime > 0) + fprintf(f, + "%sStart Timestamp: %s\n", + prefix, format_timestamp(buf, sizeof(buf), s->start_timestamp.realtime)); + + if (s->exit_timestamp.realtime > 0) + fprintf(f, + "%sExit Timestamp: %s\n" + "%sExit Code: %s\n" + "%sExit Status: %i\n", + prefix, format_timestamp(buf, sizeof(buf), s->exit_timestamp.realtime), + prefix, sigchld_code_to_string(s->code), + prefix, s->status); +} + +char *exec_command_line(char **argv) { + size_t k; + char *n, *p, **a; + bool first = true; + + assert(argv); + + k = 1; + STRV_FOREACH(a, argv) + k += strlen(*a)+3; + + if (!(n = new(char, k))) + return NULL; + + p = n; + STRV_FOREACH(a, argv) { + + if (!first) + *(p++) = ' '; + else + first = false; + + if (strpbrk(*a, WHITESPACE)) { + *(p++) = '\''; + p = stpcpy(p, *a); + *(p++) = '\''; + } else + p = stpcpy(p, *a); + + } + + *p = 0; + + /* FIXME: this doesn't really handle arguments that have + * spaces and ticks in them */ + + return n; +} + +void exec_command_dump(ExecCommand *c, FILE *f, const char *prefix) { + char *p2; + const char *prefix2; + + char *cmd; + + assert(c); + assert(f); + + if (!prefix) + prefix = ""; + p2 = strappend(prefix, "\t"); + prefix2 = p2 ? p2 : prefix; + + cmd = exec_command_line(c->argv); + + fprintf(f, + "%sCommand Line: %s\n", + prefix, cmd ? cmd : strerror(ENOMEM)); + + free(cmd); + + exec_status_dump(&c->exec_status, f, prefix2); + + free(p2); +} + +void exec_command_dump_list(ExecCommand *c, FILE *f, const char *prefix) { + assert(f); + + if (!prefix) + prefix = ""; + + LIST_FOREACH(command, c, c) + exec_command_dump(c, f, prefix); +} + +void exec_command_append_list(ExecCommand **l, ExecCommand *e) { + ExecCommand *end; + + assert(l); + assert(e); + + if (*l) { + /* It's kind of important, that we keep the order here */ + LIST_FIND_TAIL(ExecCommand, command, *l, end); + LIST_INSERT_AFTER(ExecCommand, command, *l, end, e); + } else + *l = e; +} + +int exec_command_set(ExecCommand *c, const char *path, ...) { + va_list ap; + char **l, *p; + + assert(c); + assert(path); + + va_start(ap, path); + l = strv_new_ap(path, ap); + va_end(ap); + + if (!l) + return -ENOMEM; + + if (!(p = strdup(path))) { + strv_free(l); + return -ENOMEM; + } + + free(c->path); + c->path = p; + + strv_free(c->argv); + c->argv = l; + + return 0; +} + +static const char* const exec_input_table[_EXEC_INPUT_MAX] = { + [EXEC_INPUT_NULL] = "null", + [EXEC_INPUT_TTY] = "tty", + [EXEC_INPUT_TTY_FORCE] = "tty-force", + [EXEC_INPUT_TTY_FAIL] = "tty-fail", + [EXEC_INPUT_SOCKET] = "socket" +}; + +DEFINE_STRING_TABLE_LOOKUP(exec_input, ExecInput); + +static const char* const exec_output_table[_EXEC_OUTPUT_MAX] = { + [EXEC_OUTPUT_INHERIT] = "inherit", + [EXEC_OUTPUT_NULL] = "null", + [EXEC_OUTPUT_TTY] = "tty", + [EXEC_OUTPUT_SYSLOG] = "syslog", + [EXEC_OUTPUT_SYSLOG_AND_CONSOLE] = "syslog+console", + [EXEC_OUTPUT_KMSG] = "kmsg", + [EXEC_OUTPUT_KMSG_AND_CONSOLE] = "kmsg+console", + [EXEC_OUTPUT_JOURNAL] = "journal", + [EXEC_OUTPUT_JOURNAL_AND_CONSOLE] = "journal+console", + [EXEC_OUTPUT_SOCKET] = "socket" +}; + +DEFINE_STRING_TABLE_LOOKUP(exec_output, ExecOutput); diff --git a/src/core/execute.h b/src/core/execute.h new file mode 100644 index 0000000000..15574dc97e --- /dev/null +++ b/src/core/execute.h @@ -0,0 +1,217 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct ExecStatus ExecStatus; +typedef struct ExecCommand ExecCommand; +typedef struct ExecContext ExecContext; + +#include <linux/types.h> +#include <sys/time.h> +#include <sys/resource.h> +#include <sys/capability.h> +#include <stdbool.h> +#include <stdio.h> +#include <sched.h> + +struct CGroupBonding; +struct CGroupAttribute; + +typedef struct Unit Unit; + +#include "list.h" +#include "util.h" + +typedef enum ExecInput { + EXEC_INPUT_NULL, + EXEC_INPUT_TTY, + EXEC_INPUT_TTY_FORCE, + EXEC_INPUT_TTY_FAIL, + EXEC_INPUT_SOCKET, + _EXEC_INPUT_MAX, + _EXEC_INPUT_INVALID = -1 +} ExecInput; + +typedef enum ExecOutput { + EXEC_OUTPUT_INHERIT, + EXEC_OUTPUT_NULL, + EXEC_OUTPUT_TTY, + EXEC_OUTPUT_SYSLOG, + EXEC_OUTPUT_SYSLOG_AND_CONSOLE, + EXEC_OUTPUT_KMSG, + EXEC_OUTPUT_KMSG_AND_CONSOLE, + EXEC_OUTPUT_JOURNAL, + EXEC_OUTPUT_JOURNAL_AND_CONSOLE, + EXEC_OUTPUT_SOCKET, + _EXEC_OUTPUT_MAX, + _EXEC_OUTPUT_INVALID = -1 +} ExecOutput; + +struct ExecStatus { + dual_timestamp start_timestamp; + dual_timestamp exit_timestamp; + pid_t pid; + int code; /* as in siginfo_t::si_code */ + int status; /* as in sigingo_t::si_status */ +}; + +struct ExecCommand { + char *path; + char **argv; + ExecStatus exec_status; + LIST_FIELDS(ExecCommand, command); /* useful for chaining commands */ + bool ignore; +}; + +struct ExecContext { + char **environment; + char **environment_files; + + struct rlimit *rlimit[RLIMIT_NLIMITS]; + char *working_directory, *root_directory; + + mode_t umask; + int oom_score_adjust; + int nice; + int ioprio; + int cpu_sched_policy; + int cpu_sched_priority; + + cpu_set_t *cpuset; + unsigned cpuset_ncpus; + + ExecInput std_input; + ExecOutput std_output; + ExecOutput std_error; + + nsec_t timer_slack_nsec; + + char *tcpwrap_name; + + char *tty_path; + + bool tty_reset; + bool tty_vhangup; + bool tty_vt_disallocate; + + bool ignore_sigpipe; + + /* Since resolving these names might might involve socket + * connections and we don't want to deadlock ourselves these + * names are resolved on execution only and in the child + * process. */ + char *user; + char *group; + char **supplementary_groups; + + char *pam_name; + + char *utmp_id; + + char **read_write_dirs, **read_only_dirs, **inaccessible_dirs; + unsigned long mount_flags; + + uint64_t capability_bounding_set_drop; + + cap_t capabilities; + int secure_bits; + + int syslog_priority; + char *syslog_identifier; + bool syslog_level_prefix; + + bool cpu_sched_reset_on_fork; + bool non_blocking; + bool private_tmp; + bool private_network; + char *tmp_dir; + char *var_tmp_dir; + + bool no_new_privileges; + + bool control_group_modify; + int control_group_persistent; + + /* This is not exposed to the user but available + * internally. We need it to make sure that whenever we spawn + * /bin/mount it is run in the same process group as us so + * that the autofs logic detects that it belongs to us and we + * don't enter a trigger loop. */ + bool same_pgrp; + + uint32_t *syscall_filter; + + bool oom_score_adjust_set:1; + bool nice_set:1; + bool ioprio_set:1; + bool cpu_sched_set:1; +}; + +int exec_spawn(ExecCommand *command, + char **argv, + ExecContext *context, + int fds[], unsigned n_fds, + char **environment, + bool apply_permissions, + bool apply_chroot, + bool apply_tty_stdin, + bool confirm_spawn, + struct CGroupBonding *cgroup_bondings, + struct CGroupAttribute *cgroup_attributes, + const char *cgroup_suffix, + const char *unit_id, + int pipe_fd[2], + pid_t *ret); + +void exec_command_done(ExecCommand *c); +void exec_command_done_array(ExecCommand *c, unsigned n); + +void exec_command_free_list(ExecCommand *c); +void exec_command_free_array(ExecCommand **c, unsigned n); + +char *exec_command_line(char **argv); + +void exec_command_dump(ExecCommand *c, FILE *f, const char *prefix); +void exec_command_dump_list(ExecCommand *c, FILE *f, const char *prefix); +void exec_command_append_list(ExecCommand **l, ExecCommand *e); +int exec_command_set(ExecCommand *c, const char *path, ...); + +void exec_context_init(ExecContext *c); +void exec_context_done(ExecContext *c, bool reloading_or_reexecuting); +void exec_context_tmp_dirs_done(ExecContext *c); +void exec_context_dump(ExecContext *c, FILE* f, const char *prefix); +void exec_context_tty_reset(const ExecContext *context); + +int exec_context_load_environment(const ExecContext *c, char ***l); + +bool exec_context_may_touch_console(ExecContext *c); +void exec_context_serialize(const ExecContext *c, Unit *u, FILE *f); + +void exec_status_start(ExecStatus *s, pid_t pid); +void exec_status_exit(ExecStatus *s, ExecContext *context, pid_t pid, int code, int status); +void exec_status_dump(ExecStatus *s, FILE *f, const char *prefix); + +const char* exec_output_to_string(ExecOutput i) _const_; +ExecOutput exec_output_from_string(const char *s) _pure_; + +const char* exec_input_to_string(ExecInput i) _const_; +ExecInput exec_input_from_string(const char *s) _pure_; diff --git a/src/core/hostname-setup.c b/src/core/hostname-setup.c new file mode 100644 index 0000000000..8aa1cff1d3 --- /dev/null +++ b/src/core/hostname-setup.c @@ -0,0 +1,92 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <unistd.h> +#include <stdio.h> +#include <errno.h> +#include <string.h> +#include <stdlib.h> + +#include "hostname-setup.h" +#include "macro.h" +#include "util.h" +#include "log.h" +#include "fileio.h" + +static int read_and_strip_hostname(const char *path, char **hn) { + char *s; + int r; + + assert(path); + assert(hn); + + r = read_one_line_file(path, &s); + if (r < 0) + return r; + + hostname_cleanup(s, false); + + if (isempty(s)) { + free(s); + return -ENOENT; + } + + *hn = s; + return 0; +} + +int hostname_setup(void) { + int r; + _cleanup_free_ char *b = NULL; + const char *hn; + bool enoent = false; + + r = read_and_strip_hostname("/etc/hostname", &b); + if (r < 0) { + if (r == -ENOENT) + enoent = true; + else + log_warning("Failed to read configured hostname: %s", strerror(-r)); + + hn = NULL; + } else + hn = b; + + if (isempty(hn)) { + /* Don't override the hostname if it is already set + * and not explicitly configured */ + if (hostname_is_set()) + return 0; + + if (enoent) + log_info("No hostname configured."); + + hn = "localhost"; + } + + if (sethostname(hn, strlen(hn)) < 0) { + log_warning("Failed to set hostname to <%s>: %m", hn); + return -errno; + } + + log_info("Set hostname to <%s>.", hn); + return 0; +} diff --git a/src/core/hostname-setup.h b/src/core/hostname-setup.h new file mode 100644 index 0000000000..8dc3a9e1d8 --- /dev/null +++ b/src/core/hostname-setup.h @@ -0,0 +1,24 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int hostname_setup(void); diff --git a/src/core/ima-setup.c b/src/core/ima-setup.c new file mode 100644 index 0000000000..7f8ec23d58 --- /dev/null +++ b/src/core/ima-setup.c @@ -0,0 +1,110 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright (C) 2012 Roberto Sassu - Politecnico di Torino, Italy + TORSEC group -- http://security.polito.it + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <unistd.h> +#include <stdio.h> +#include <errno.h> +#include <string.h> +#include <stdlib.h> +#include <fcntl.h> +#include <sys/stat.h> +#include <sys/mman.h> + +#include "ima-setup.h" +#include "mount-setup.h" +#include "macro.h" +#include "util.h" +#include "log.h" +#include "label.h" + +#define IMA_SECFS_DIR "/sys/kernel/security/ima" +#define IMA_SECFS_POLICY IMA_SECFS_DIR "/policy" +#define IMA_POLICY_PATH "/etc/ima/ima-policy" + +int ima_setup(void) { + +#ifdef HAVE_IMA + struct stat st; + ssize_t policy_size = 0, written = 0; + char *policy; + int policyfd = -1, imafd = -1; + int result = 0; + + if (stat(IMA_POLICY_PATH, &st) < 0) + return 0; + + policy_size = st.st_size; + if (stat(IMA_SECFS_DIR, &st) < 0) { + log_debug("IMA support is disabled in the kernel, ignoring."); + return 0; + } + + if (stat(IMA_SECFS_POLICY, &st) < 0) { + log_error("Another IMA custom policy has already been loaded, " + "ignoring."); + return 0; + } + + policyfd = open(IMA_POLICY_PATH, O_RDONLY|O_CLOEXEC); + if (policyfd < 0) { + log_error("Failed to open the IMA custom policy file %s (%m), " + "ignoring.", IMA_POLICY_PATH); + return 0; + } + + imafd = open(IMA_SECFS_POLICY, O_WRONLY|O_CLOEXEC); + if (imafd < 0) { + log_error("Failed to open the IMA kernel interface %s (%m), " + "ignoring.", IMA_SECFS_POLICY); + goto out; + } + + policy = mmap(NULL, policy_size, PROT_READ, MAP_PRIVATE, policyfd, 0); + if (policy == MAP_FAILED) { + log_error("mmap() failed (%m), freezing"); + result = -errno; + goto out; + } + + written = loop_write(imafd, policy, (size_t)policy_size, false); + if (written != policy_size) { + log_error("Failed to load the IMA custom policy file %s (%m), " + "ignoring.", IMA_POLICY_PATH); + goto out_mmap; + } + + log_info("Successfully loaded the IMA custom policy %s.", + IMA_POLICY_PATH); +out_mmap: + munmap(policy, policy_size); +out: + if (policyfd >= 0) + close_nointr_nofail(policyfd); + if (imafd >= 0) + close_nointr_nofail(imafd); + if (result) + return result; +#endif /* HAVE_IMA */ + + return 0; +} diff --git a/src/core/ima-setup.h b/src/core/ima-setup.h new file mode 100644 index 0000000000..14b56d1fc2 --- /dev/null +++ b/src/core/ima-setup.h @@ -0,0 +1,26 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright (C) 2012 Roberto Sassu - Politecnico di Torino, Italy + TORSEC group -- http://security.polito.it + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int ima_setup(void); diff --git a/src/core/initreq.h b/src/core/initreq.h new file mode 100644 index 0000000000..859042ce42 --- /dev/null +++ b/src/core/initreq.h @@ -0,0 +1,77 @@ +/* + * initreq.h Interface to talk to init through /dev/initctl. + * + * Copyright (C) 1995-2004 Miquel van Smoorenburg + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * Version: @(#)initreq.h 1.28 31-Mar-2004 MvS + * + */ +#ifndef _INITREQ_H +#define _INITREQ_H + +#include <sys/param.h> + +#if defined(__FreeBSD_kernel__) +# define INIT_FIFO "/etc/.initctl" +#else +# define INIT_FIFO "/dev/initctl" +#endif + +#define INIT_MAGIC 0x03091969 +#define INIT_CMD_START 0 +#define INIT_CMD_RUNLVL 1 +#define INIT_CMD_POWERFAIL 2 +#define INIT_CMD_POWERFAILNOW 3 +#define INIT_CMD_POWEROK 4 +#define INIT_CMD_BSD 5 +#define INIT_CMD_SETENV 6 +#define INIT_CMD_UNSETENV 7 + +#define INIT_CMD_CHANGECONS 12345 + +#ifdef MAXHOSTNAMELEN +# define INITRQ_HLEN MAXHOSTNAMELEN +#else +# define INITRQ_HLEN 64 +#endif + +/* + * This is what BSD 4.4 uses when talking to init. + * Linux doesn't use this right now. + */ +struct init_request_bsd { + char gen_id[8]; /* Beats me.. telnetd uses "fe" */ + char tty_id[16]; /* Tty name minus /dev/tty */ + char host[INITRQ_HLEN]; /* Hostname */ + char term_type[16]; /* Terminal type */ + int signal; /* Signal to send */ + int pid; /* Process to send to */ + char exec_name[128]; /* Program to execute */ + char reserved[128]; /* For future expansion. */ +}; + + +/* + * Because of legacy interfaces, "runlevel" and "sleeptime" + * aren't in a separate struct in the union. + * + * The weird sizes are because init expects the whole + * struct to be 384 bytes. + */ +struct init_request { + int magic; /* Magic number */ + int cmd; /* What kind of request */ + int runlevel; /* Runlevel to change to */ + int sleeptime; /* Time between TERM and KILL */ + union { + struct init_request_bsd bsd; + char data[368]; + } i; +}; + +#endif diff --git a/src/core/job.c b/src/core/job.c new file mode 100644 index 0000000000..d304a16d06 --- /dev/null +++ b/src/core/job.c @@ -0,0 +1,1143 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <assert.h> +#include <errno.h> +#include <sys/timerfd.h> +#include <sys/epoll.h> + +#include "systemd/sd-id128.h" +#include "systemd/sd-messages.h" +#include "set.h" +#include "unit.h" +#include "macro.h" +#include "strv.h" +#include "load-fragment.h" +#include "load-dropin.h" +#include "log.h" +#include "dbus-job.h" +#include "special.h" +#include "sync.h" +#include "virt.h" + +JobBusClient* job_bus_client_new(DBusConnection *connection, const char *name) { + JobBusClient *cl; + size_t name_len; + + name_len = strlen(name); + cl = malloc0(sizeof(JobBusClient) + name_len + 1); + if (!cl) + return NULL; + + cl->bus = connection; + memcpy(cl->name, name, name_len + 1); + return cl; +} + +Job* job_new_raw(Unit *unit) { + Job *j; + + /* used for deserialization */ + + assert(unit); + + j = new0(Job, 1); + if (!j) + return NULL; + + j->manager = unit->manager; + j->unit = unit; + j->type = _JOB_TYPE_INVALID; + j->timer_watch.type = WATCH_INVALID; + + return j; +} + +Job* job_new(Unit *unit, JobType type) { + Job *j; + + assert(type < _JOB_TYPE_MAX); + + j = job_new_raw(unit); + if (!j) + return NULL; + + j->id = j->manager->current_job_id++; + j->type = type; + + /* We don't link it here, that's what job_dependency() is for */ + + return j; +} + +void job_free(Job *j) { + JobBusClient *cl; + + assert(j); + assert(!j->installed); + assert(!j->transaction_prev); + assert(!j->transaction_next); + assert(!j->subject_list); + assert(!j->object_list); + + if (j->in_run_queue) + LIST_REMOVE(Job, run_queue, j->manager->run_queue, j); + + if (j->in_dbus_queue) + LIST_REMOVE(Job, dbus_queue, j->manager->dbus_job_queue, j); + + if (j->timer_watch.type != WATCH_INVALID) { + assert(j->timer_watch.type == WATCH_JOB_TIMER); + assert(j->timer_watch.data.job == j); + assert(j->timer_watch.fd >= 0); + + assert_se(epoll_ctl(j->manager->epoll_fd, EPOLL_CTL_DEL, j->timer_watch.fd, NULL) >= 0); + close_nointr_nofail(j->timer_watch.fd); + } + + while ((cl = j->bus_client_list)) { + LIST_REMOVE(JobBusClient, client, j->bus_client_list, cl); + free(cl); + } + free(j); +} + +void job_uninstall(Job *j) { + Job **pj; + + assert(j->installed); + + pj = (j->type == JOB_NOP) ? &j->unit->nop_job : &j->unit->job; + assert(*pj == j); + + /* Detach from next 'bigger' objects */ + + /* daemon-reload should be transparent to job observers */ + if (j->manager->n_reloading <= 0) + bus_job_send_removed_signal(j); + + *pj = NULL; + + unit_add_to_gc_queue(j->unit); + + hashmap_remove(j->manager->jobs, UINT32_TO_PTR(j->id)); + j->installed = false; +} + +static bool job_type_allows_late_merge(JobType t) { + /* Tells whether it is OK to merge a job of type 't' with an already + * running job. + * Reloads cannot be merged this way. Think of the sequence: + * 1. Reload of a daemon is in progress; the daemon has already loaded + * its config file, but hasn't completed the reload operation yet. + * 2. Edit foo's config file. + * 3. Trigger another reload to have the daemon use the new config. + * Should the second reload job be merged into the first one, the daemon + * would not know about the new config. + * JOB_RESTART jobs on the other hand can be merged, because they get + * patched into JOB_START after stopping the unit. So if we see a + * JOB_RESTART running, it means the unit hasn't stopped yet and at + * this time the merge is still allowed. */ + return t != JOB_RELOAD; +} + +static void job_merge_into_installed(Job *j, Job *other) { + assert(j->installed); + assert(j->unit == other->unit); + + if (j->type != JOB_NOP) + job_type_merge_and_collapse(&j->type, other->type, j->unit); + else + assert(other->type == JOB_NOP); + + j->override = j->override || other->override; + j->irreversible = j->irreversible || other->irreversible; + j->ignore_order = j->ignore_order || other->ignore_order; +} + +Job* job_install(Job *j) { + Job **pj; + Job *uj; + + assert(!j->installed); + assert(j->type < _JOB_TYPE_MAX_IN_TRANSACTION); + + pj = (j->type == JOB_NOP) ? &j->unit->nop_job : &j->unit->job; + uj = *pj; + + if (uj) { + if (j->type != JOB_NOP && job_type_is_conflicting(uj->type, j->type)) + job_finish_and_invalidate(uj, JOB_CANCELED, false); + else { + /* not conflicting, i.e. mergeable */ + + if (j->type == JOB_NOP || uj->state == JOB_WAITING || + (job_type_allows_late_merge(j->type) && job_type_is_superset(uj->type, j->type))) { + job_merge_into_installed(uj, j); + log_debug_unit(uj->unit->id, + "Merged into installed job %s/%s as %u", + uj->unit->id, job_type_to_string(uj->type), (unsigned) uj->id); + return uj; + } else { + /* already running and not safe to merge into */ + /* Patch uj to become a merged job and re-run it. */ + /* XXX It should be safer to queue j to run after uj finishes, but it is + * not currently possible to have more than one installed job per unit. */ + job_merge_into_installed(uj, j); + log_debug_unit(uj->unit->id, + "Merged into running job, re-running: %s/%s as %u", + uj->unit->id, job_type_to_string(uj->type), (unsigned) uj->id); + uj->state = JOB_WAITING; + uj->manager->n_running_jobs--; + return uj; + } + } + } + + /* Install the job */ + *pj = j; + j->installed = true; + j->manager->n_installed_jobs ++; + log_debug_unit(j->unit->id, + "Installed new job %s/%s as %u", + j->unit->id, job_type_to_string(j->type), (unsigned) j->id); + return j; +} + +int job_install_deserialized(Job *j) { + Job **pj; + + assert(!j->installed); + + if (j->type < 0 || j->type >= _JOB_TYPE_MAX_IN_TRANSACTION) { + log_debug("Invalid job type %s in deserialization.", strna(job_type_to_string(j->type))); + return -EINVAL; + } + + pj = (j->type == JOB_NOP) ? &j->unit->nop_job : &j->unit->job; + + if (*pj) { + log_debug_unit(j->unit->id, + "Unit %s already has a job installed. Not installing deserialized job.", + j->unit->id); + return -EEXIST; + } + *pj = j; + j->installed = true; + log_debug_unit(j->unit->id, + "Reinstalled deserialized job %s/%s as %u", + j->unit->id, job_type_to_string(j->type), (unsigned) j->id); + return 0; +} + +JobDependency* job_dependency_new(Job *subject, Job *object, bool matters, bool conflicts) { + JobDependency *l; + + assert(object); + + /* Adds a new job link, which encodes that the 'subject' job + * needs the 'object' job in some way. If 'subject' is NULL + * this means the 'anchor' job (i.e. the one the user + * explicitly asked for) is the requester. */ + + if (!(l = new0(JobDependency, 1))) + return NULL; + + l->subject = subject; + l->object = object; + l->matters = matters; + l->conflicts = conflicts; + + if (subject) + LIST_PREPEND(JobDependency, subject, subject->subject_list, l); + + LIST_PREPEND(JobDependency, object, object->object_list, l); + + return l; +} + +void job_dependency_free(JobDependency *l) { + assert(l); + + if (l->subject) + LIST_REMOVE(JobDependency, subject, l->subject->subject_list, l); + + LIST_REMOVE(JobDependency, object, l->object->object_list, l); + + free(l); +} + +void job_dump(Job *j, FILE*f, const char *prefix) { + assert(j); + assert(f); + + if (!prefix) + prefix = ""; + + fprintf(f, + "%s-> Job %u:\n" + "%s\tAction: %s -> %s\n" + "%s\tState: %s\n" + "%s\tForced: %s\n" + "%s\tIrreversible: %s\n", + prefix, j->id, + prefix, j->unit->id, job_type_to_string(j->type), + prefix, job_state_to_string(j->state), + prefix, yes_no(j->override), + prefix, yes_no(j->irreversible)); +} + +/* + * Merging is commutative, so imagine the matrix as symmetric. We store only + * its lower triangle to avoid duplication. We don't store the main diagonal, + * because A merged with A is simply A. + * + * If the resulting type is collapsed immediately afterwards (to get rid of + * the JOB_RELOAD_OR_START, which lies outside the lookup function's domain), + * the following properties hold: + * + * Merging is associative! A merged with B merged with C is the same as + * A merged with C merged with B. + * + * Mergeability is transitive! If A can be merged with B and B with C then + * A also with C. + * + * Also, if A merged with B cannot be merged with C, then either A or B cannot + * be merged with C either. + */ +static const JobType job_merging_table[] = { +/* What \ With * JOB_START JOB_VERIFY_ACTIVE JOB_STOP JOB_RELOAD */ +/*********************************************************************************/ +/*JOB_START */ +/*JOB_VERIFY_ACTIVE */ JOB_START, +/*JOB_STOP */ -1, -1, +/*JOB_RELOAD */ JOB_RELOAD_OR_START, JOB_RELOAD, -1, +/*JOB_RESTART */ JOB_RESTART, JOB_RESTART, -1, JOB_RESTART, +}; + +JobType job_type_lookup_merge(JobType a, JobType b) { + assert_cc(ELEMENTSOF(job_merging_table) == _JOB_TYPE_MAX_MERGING * (_JOB_TYPE_MAX_MERGING - 1) / 2); + assert(a >= 0 && a < _JOB_TYPE_MAX_MERGING); + assert(b >= 0 && b < _JOB_TYPE_MAX_MERGING); + + if (a == b) + return a; + + if (a < b) { + JobType tmp = a; + a = b; + b = tmp; + } + + return job_merging_table[(a - 1) * a / 2 + b]; +} + +bool job_type_is_redundant(JobType a, UnitActiveState b) { + switch (a) { + + case JOB_START: + return + b == UNIT_ACTIVE || + b == UNIT_RELOADING; + + case JOB_STOP: + return + b == UNIT_INACTIVE || + b == UNIT_FAILED; + + case JOB_VERIFY_ACTIVE: + return + b == UNIT_ACTIVE || + b == UNIT_RELOADING; + + case JOB_RELOAD: + return + b == UNIT_RELOADING; + + case JOB_RESTART: + return + b == UNIT_ACTIVATING; + + default: + assert_not_reached("Invalid job type"); + } +} + +void job_type_collapse(JobType *t, Unit *u) { + UnitActiveState s; + + switch (*t) { + + case JOB_TRY_RESTART: + s = unit_active_state(u); + if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s)) + *t = JOB_NOP; + else + *t = JOB_RESTART; + break; + + case JOB_RELOAD_OR_START: + s = unit_active_state(u); + if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s)) + *t = JOB_START; + else + *t = JOB_RELOAD; + break; + + default: + ; + } +} + +int job_type_merge_and_collapse(JobType *a, JobType b, Unit *u) { + JobType t = job_type_lookup_merge(*a, b); + if (t < 0) + return -EEXIST; + *a = t; + job_type_collapse(a, u); + return 0; +} + +bool job_is_runnable(Job *j) { + Iterator i; + Unit *other; + + assert(j); + assert(j->installed); + + /* Checks whether there is any job running for the units this + * job needs to be running after (in the case of a 'positive' + * job type) or before (in the case of a 'negative' job + * type. */ + + /* First check if there is an override */ + if (j->ignore_order) + return true; + + if (j->type == JOB_NOP) + return true; + + if (j->type == JOB_START || + j->type == JOB_VERIFY_ACTIVE || + j->type == JOB_RELOAD) { + + /* Immediate result is that the job is or might be + * started. In this case lets wait for the + * dependencies, regardless whether they are + * starting or stopping something. */ + + SET_FOREACH(other, j->unit->dependencies[UNIT_AFTER], i) + if (other->job) + return false; + } + + /* Also, if something else is being stopped and we should + * change state after it, then lets wait. */ + + SET_FOREACH(other, j->unit->dependencies[UNIT_BEFORE], i) + if (other->job && + (other->job->type == JOB_STOP || + other->job->type == JOB_RESTART)) + return false; + + /* This means that for a service a and a service b where b + * shall be started after a: + * + * start a + start b → 1st step start a, 2nd step start b + * start a + stop b → 1st step stop b, 2nd step start a + * stop a + start b → 1st step stop a, 2nd step start b + * stop a + stop b → 1st step stop b, 2nd step stop a + * + * This has the side effect that restarts are properly + * synchronized too. */ + + return true; +} + +static void job_change_type(Job *j, JobType newtype) { + log_debug_unit(j->unit->id, + "Converting job %s/%s -> %s/%s", + j->unit->id, job_type_to_string(j->type), + j->unit->id, job_type_to_string(newtype)); + + j->type = newtype; +} + +int job_run_and_invalidate(Job *j) { + int r; + uint32_t id; + Manager *m = j->manager; + + assert(j); + assert(j->installed); + assert(j->type < _JOB_TYPE_MAX_IN_TRANSACTION); + assert(j->in_run_queue); + + LIST_REMOVE(Job, run_queue, j->manager->run_queue, j); + j->in_run_queue = false; + + if (j->state != JOB_WAITING) + return 0; + + if (!job_is_runnable(j)) + return -EAGAIN; + + j->state = JOB_RUNNING; + m->n_running_jobs++; + job_add_to_dbus_queue(j); + + /* While we execute this operation the job might go away (for + * example: because it is replaced by a new, conflicting + * job.) To make sure we don't access a freed job later on we + * store the id here, so that we can verify the job is still + * valid. */ + id = j->id; + + switch (j->type) { + + case JOB_START: + r = unit_start(j->unit); + + /* If this unit cannot be started, then simply wait */ + if (r == -EBADR) + r = 0; + break; + + case JOB_VERIFY_ACTIVE: { + UnitActiveState t = unit_active_state(j->unit); + if (UNIT_IS_ACTIVE_OR_RELOADING(t)) + r = -EALREADY; + else if (t == UNIT_ACTIVATING) + r = -EAGAIN; + else + r = -ENOEXEC; + break; + } + + case JOB_STOP: + case JOB_RESTART: + r = unit_stop(j->unit); + + /* If this unit cannot stopped, then simply wait. */ + if (r == -EBADR) + r = 0; + break; + + case JOB_RELOAD: + r = unit_reload(j->unit); + break; + + case JOB_NOP: + r = -EALREADY; + break; + + default: + assert_not_reached("Unknown job type"); + } + + j = manager_get_job(m, id); + if (j) { + if (r == -EALREADY) + r = job_finish_and_invalidate(j, JOB_DONE, true); + else if (r == -ENOEXEC) + r = job_finish_and_invalidate(j, JOB_SKIPPED, true); + else if (r == -EAGAIN) { + j->state = JOB_WAITING; + m->n_running_jobs--; + } else if (r < 0) + r = job_finish_and_invalidate(j, JOB_FAILED, true); + } + + return r; +} + +_pure_ static const char *job_get_status_message_format(Unit *u, JobType t, JobResult result) { + const UnitStatusMessageFormats *format_table; + + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + format_table = &UNIT_VTABLE(u)->status_message_formats; + if (!format_table) + return NULL; + + if (t == JOB_START) + return format_table->finished_start_job[result]; + else if (t == JOB_STOP || t == JOB_RESTART) + return format_table->finished_stop_job[result]; + + return NULL; +} + +_pure_ static const char *job_get_status_message_format_try_harder(Unit *u, JobType t, JobResult result) { + const char *format; + + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + format = job_get_status_message_format(u, t, result); + if (format) + return format; + + /* Return generic strings */ + if (t == JOB_START) { + if (result == JOB_DONE) + return "Started %s."; + else if (result == JOB_FAILED) + return "Failed to start %s."; + else if (result == JOB_DEPENDENCY) + return "Dependency failed for %s."; + else if (result == JOB_TIMEOUT) + return "Timed out starting %s."; + } else if (t == JOB_STOP || t == JOB_RESTART) { + if (result == JOB_DONE) + return "Stopped %s."; + else if (result == JOB_FAILED) + return "Stopped (with error) %s."; + else if (result == JOB_TIMEOUT) + return "Timed out stoppping %s."; + } else if (t == JOB_RELOAD) { + if (result == JOB_DONE) + return "Reloaded %s."; + else if (result == JOB_FAILED) + return "Reload failed for %s."; + else if (result == JOB_TIMEOUT) + return "Timed out reloading %s."; + } + + return NULL; +} + +static void job_print_status_message(Unit *u, JobType t, JobResult result) { + const char *format; + + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + if (t == JOB_START) { + format = job_get_status_message_format(u, t, result); + if (!format) + return; + + switch (result) { + + case JOB_DONE: + if (u->condition_result) + unit_status_printf(u, ANSI_GREEN_ON " OK " ANSI_HIGHLIGHT_OFF, format); + break; + + case JOB_FAILED: + unit_status_printf(u, ANSI_HIGHLIGHT_RED_ON "FAILED" ANSI_HIGHLIGHT_OFF, format); + manager_status_printf(u->manager, false, NULL, "See 'systemctl status %s' for details.", u->id); + break; + + case JOB_DEPENDENCY: + unit_status_printf(u, ANSI_HIGHLIGHT_YELLOW_ON "DEPEND" ANSI_HIGHLIGHT_OFF, format); + break; + + case JOB_TIMEOUT: + unit_status_printf(u, ANSI_HIGHLIGHT_RED_ON " TIME " ANSI_HIGHLIGHT_OFF, format); + break; + + default: + ; + } + + } else if (t == JOB_STOP || t == JOB_RESTART) { + + format = job_get_status_message_format(u, t, result); + if (!format) + return; + + switch (result) { + + case JOB_TIMEOUT: + unit_status_printf(u, ANSI_HIGHLIGHT_RED_ON " TIME " ANSI_HIGHLIGHT_OFF, format); + break; + + case JOB_DONE: + case JOB_FAILED: + unit_status_printf(u, ANSI_GREEN_ON " OK " ANSI_HIGHLIGHT_OFF, format); + break; + + default: + ; + } + + } else if (t == JOB_VERIFY_ACTIVE) { + + /* When verify-active detects the unit is inactive, report it. + * Most likely a DEPEND warning from a requisiting unit will + * occur next and it's nice to see what was requisited. */ + if (result == JOB_SKIPPED) + unit_status_printf(u, ANSI_HIGHLIGHT_ON " INFO " ANSI_HIGHLIGHT_OFF, "%s is not active."); + } +} + +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wformat-nonliteral" +static void job_log_status_message(Unit *u, JobType t, JobResult result) { + const char *format; + char buf[LINE_MAX]; + + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + /* Skip this if it goes to the console. since we already print + * to the console anyway... */ + + if (log_on_console()) + return; + + format = job_get_status_message_format_try_harder(u, t, result); + if (!format) + return; + + snprintf(buf, sizeof(buf), format, unit_description(u)); + char_array_0(buf); + + if (t == JOB_START) { + sd_id128_t mid; + + mid = result == JOB_DONE ? SD_MESSAGE_UNIT_STARTED : SD_MESSAGE_UNIT_FAILED; + log_struct_unit(result == JOB_DONE ? LOG_INFO : LOG_ERR, + u->id, + MESSAGE_ID(mid), + "RESULT=%s", job_result_to_string(result), + "MESSAGE=%s", buf, + NULL); + + } else if (t == JOB_STOP) + log_struct_unit(result == JOB_DONE ? LOG_INFO : LOG_ERR, + u->id, + MESSAGE_ID(SD_MESSAGE_UNIT_STOPPED), + "RESULT=%s", job_result_to_string(result), + "MESSAGE=%s", buf, + NULL); + + else if (t == JOB_RELOAD) + log_struct_unit(result == JOB_DONE ? LOG_INFO : LOG_ERR, + u->id, + MESSAGE_ID(SD_MESSAGE_UNIT_RELOADED), + "RESULT=%s", job_result_to_string(result), + "MESSAGE=%s", buf, + NULL); +} +#pragma GCC diagnostic pop + +int job_finish_and_invalidate(Job *j, JobResult result, bool recursive) { + Unit *u; + Unit *other; + JobType t; + Iterator i; + + assert(j); + assert(j->installed); + assert(j->type < _JOB_TYPE_MAX_IN_TRANSACTION); + + u = j->unit; + t = j->type; + + j->result = result; + + if (j->state == JOB_RUNNING) + j->manager->n_running_jobs--; + + log_debug_unit(u->id, "Job %s/%s finished, result=%s", + u->id, job_type_to_string(t), job_result_to_string(result)); + + job_print_status_message(u, t, result); + job_log_status_message(u, t, result); + + job_add_to_dbus_queue(j); + + /* Patch restart jobs so that they become normal start jobs */ + if (result == JOB_DONE && t == JOB_RESTART) { + + job_change_type(j, JOB_START); + j->state = JOB_WAITING; + + job_add_to_run_queue(j); + + goto finish; + } + + if (result == JOB_FAILED) + j->manager->n_failed_jobs ++; + + job_uninstall(j); + job_free(j); + + /* Fail depending jobs on failure */ + if (result != JOB_DONE && recursive) { + + if (t == JOB_START || + t == JOB_VERIFY_ACTIVE) { + + SET_FOREACH(other, u->dependencies[UNIT_REQUIRED_BY], i) + if (other->job && + (other->job->type == JOB_START || + other->job->type == JOB_VERIFY_ACTIVE)) + job_finish_and_invalidate(other->job, JOB_DEPENDENCY, true); + + SET_FOREACH(other, u->dependencies[UNIT_BOUND_BY], i) + if (other->job && + (other->job->type == JOB_START || + other->job->type == JOB_VERIFY_ACTIVE)) + job_finish_and_invalidate(other->job, JOB_DEPENDENCY, true); + + SET_FOREACH(other, u->dependencies[UNIT_REQUIRED_BY_OVERRIDABLE], i) + if (other->job && + !other->job->override && + (other->job->type == JOB_START || + other->job->type == JOB_VERIFY_ACTIVE)) + job_finish_and_invalidate(other->job, JOB_DEPENDENCY, true); + + } else if (t == JOB_STOP) { + + SET_FOREACH(other, u->dependencies[UNIT_CONFLICTED_BY], i) + if (other->job && + (other->job->type == JOB_START || + other->job->type == JOB_VERIFY_ACTIVE)) + job_finish_and_invalidate(other->job, JOB_DEPENDENCY, true); + } + } + + /* Trigger OnFailure dependencies that are not generated by + * the unit itself. We don't treat JOB_CANCELED as failure in + * this context. And JOB_FAILURE is already handled by the + * unit itself. */ + if (result == JOB_TIMEOUT || result == JOB_DEPENDENCY) { + log_struct_unit(LOG_NOTICE, + u->id, + "JOB_TYPE=%s", job_type_to_string(t), + "JOB_RESULT=%s", job_result_to_string(result), + "Job %s/%s failed with result '%s'.", + u->id, + job_type_to_string(t), + job_result_to_string(result), + NULL); + + unit_start_on_failure(u); + } + + unit_trigger_notify(u); + +finish: + /* Try to start the next jobs that can be started */ + SET_FOREACH(other, u->dependencies[UNIT_AFTER], i) + if (other->job) + job_add_to_run_queue(other->job); + SET_FOREACH(other, u->dependencies[UNIT_BEFORE], i) + if (other->job) + job_add_to_run_queue(other->job); + + manager_check_finished(u->manager); + + return 0; +} + +int job_start_timer(Job *j) { + struct itimerspec its = {}; + struct epoll_event ev = { + .data.ptr = &j->timer_watch, + .events = EPOLLIN, + }; + int fd, r; + + if (j->unit->job_timeout <= 0 || + j->timer_watch.type == WATCH_JOB_TIMER) + return 0; + + assert(j->timer_watch.type == WATCH_INVALID); + + if ((fd = timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK|TFD_CLOEXEC)) < 0) { + r = -errno; + goto fail; + } + + timespec_store(&its.it_value, j->unit->job_timeout); + + if (timerfd_settime(fd, 0, &its, NULL) < 0) { + r = -errno; + goto fail; + } + + if (epoll_ctl(j->manager->epoll_fd, EPOLL_CTL_ADD, fd, &ev) < 0) { + r = -errno; + goto fail; + } + + j->timer_watch.type = WATCH_JOB_TIMER; + j->timer_watch.fd = fd; + j->timer_watch.data.job = j; + + return 0; + +fail: + if (fd >= 0) + close_nointr_nofail(fd); + + return r; +} + +void job_add_to_run_queue(Job *j) { + assert(j); + assert(j->installed); + + if (j->in_run_queue) + return; + + LIST_PREPEND(Job, run_queue, j->manager->run_queue, j); + j->in_run_queue = true; +} + +void job_add_to_dbus_queue(Job *j) { + assert(j); + assert(j->installed); + + if (j->in_dbus_queue) + return; + + /* We don't check if anybody is subscribed here, since this + * job might just have been created and not yet assigned to a + * connection/client. */ + + LIST_PREPEND(Job, dbus_queue, j->manager->dbus_job_queue, j); + j->in_dbus_queue = true; +} + +char *job_dbus_path(Job *j) { + char *p; + + assert(j); + + if (asprintf(&p, "/org/freedesktop/systemd1/job/%lu", (unsigned long) j->id) < 0) + return NULL; + + return p; +} + +void job_timer_event(Job *j, uint64_t n_elapsed, Watch *w) { + assert(j); + assert(w == &j->timer_watch); + + log_warning_unit(j->unit->id, "Job %s/%s timed out.", + j->unit->id, job_type_to_string(j->type)); + job_finish_and_invalidate(j, JOB_TIMEOUT, true); +} + +int job_serialize(Job *j, FILE *f, FDSet *fds) { + fprintf(f, "job-id=%u\n", j->id); + fprintf(f, "job-type=%s\n", job_type_to_string(j->type)); + fprintf(f, "job-state=%s\n", job_state_to_string(j->state)); + fprintf(f, "job-override=%s\n", yes_no(j->override)); + fprintf(f, "job-irreversible=%s\n", yes_no(j->irreversible)); + fprintf(f, "job-sent-dbus-new-signal=%s\n", yes_no(j->sent_dbus_new_signal)); + fprintf(f, "job-ignore-order=%s\n", yes_no(j->ignore_order)); + /* Cannot save bus clients. Just note the fact that we're losing + * them. job_send_message() will fallback to broadcasting. */ + fprintf(f, "job-forgot-bus-clients=%s\n", + yes_no(j->forgot_bus_clients || j->bus_client_list)); + if (j->timer_watch.type == WATCH_JOB_TIMER) { + int copy = fdset_put_dup(fds, j->timer_watch.fd); + if (copy < 0) + return copy; + fprintf(f, "job-timer-watch-fd=%d\n", copy); + } + + /* End marker */ + fputc('\n', f); + return 0; +} + +int job_deserialize(Job *j, FILE *f, FDSet *fds) { + for (;;) { + char line[LINE_MAX], *l, *v; + size_t k; + + if (!fgets(line, sizeof(line), f)) { + if (feof(f)) + return 0; + return -errno; + } + + char_array_0(line); + l = strstrip(line); + + /* End marker */ + if (l[0] == 0) + return 0; + + k = strcspn(l, "="); + + if (l[k] == '=') { + l[k] = 0; + v = l+k+1; + } else + v = l+k; + + if (streq(l, "job-id")) { + if (safe_atou32(v, &j->id) < 0) + log_debug("Failed to parse job id value %s", v); + } else if (streq(l, "job-type")) { + JobType t = job_type_from_string(v); + if (t < 0) + log_debug("Failed to parse job type %s", v); + else if (t >= _JOB_TYPE_MAX_IN_TRANSACTION) + log_debug("Cannot deserialize job of type %s", v); + else + j->type = t; + } else if (streq(l, "job-state")) { + JobState s = job_state_from_string(v); + if (s < 0) + log_debug("Failed to parse job state %s", v); + else + j->state = s; + } else if (streq(l, "job-override")) { + int b = parse_boolean(v); + if (b < 0) + log_debug("Failed to parse job override flag %s", v); + else + j->override = j->override || b; + } else if (streq(l, "job-irreversible")) { + int b = parse_boolean(v); + if (b < 0) + log_debug("Failed to parse job irreversible flag %s", v); + else + j->irreversible = j->irreversible || b; + } else if (streq(l, "job-sent-dbus-new-signal")) { + int b = parse_boolean(v); + if (b < 0) + log_debug("Failed to parse job sent_dbus_new_signal flag %s", v); + else + j->sent_dbus_new_signal = j->sent_dbus_new_signal || b; + } else if (streq(l, "job-ignore-order")) { + int b = parse_boolean(v); + if (b < 0) + log_debug("Failed to parse job ignore_order flag %s", v); + else + j->ignore_order = j->ignore_order || b; + } else if (streq(l, "job-forgot-bus-clients")) { + int b = parse_boolean(v); + if (b < 0) + log_debug("Failed to parse job forgot_bus_clients flag %s", v); + else + j->forgot_bus_clients = j->forgot_bus_clients || b; + } else if (streq(l, "job-timer-watch-fd")) { + int fd; + if (safe_atoi(v, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_debug("Failed to parse job-timer-watch-fd value %s", v); + else { + if (j->timer_watch.type == WATCH_JOB_TIMER) + close_nointr_nofail(j->timer_watch.fd); + + j->timer_watch.type = WATCH_JOB_TIMER; + j->timer_watch.fd = fdset_remove(fds, fd); + j->timer_watch.data.job = j; + } + } + } +} + +int job_coldplug(Job *j) { + struct epoll_event ev = { + .data.ptr = &j->timer_watch, + .events = EPOLLIN, + }; + + if (j->timer_watch.type != WATCH_JOB_TIMER) + return 0; + + if (epoll_ctl(j->manager->epoll_fd, EPOLL_CTL_ADD, j->timer_watch.fd, &ev) < 0) + return -errno; + + return 0; +} + +void job_shutdown_magic(Job *j) { + assert(j); + + /* The shutdown target gets some special treatment here: we + * tell the kernel to begin with flushing its disk caches, to + * optimize shutdown time a bit. Ideally we wouldn't hardcode + * this magic into PID 1. However all other processes aren't + * options either since they'd exit much sooner than PID 1 and + * asynchronous sync() would cause their exit to be + * delayed. */ + + if (!unit_has_name(j->unit, SPECIAL_SHUTDOWN_TARGET)) + return; + + if (j->type != JOB_START) + return; + + if (detect_container(NULL) > 0) + return; + + asynchronous_sync(); +} + +static const char* const job_state_table[_JOB_STATE_MAX] = { + [JOB_WAITING] = "waiting", + [JOB_RUNNING] = "running" +}; + +DEFINE_STRING_TABLE_LOOKUP(job_state, JobState); + +static const char* const job_type_table[_JOB_TYPE_MAX] = { + [JOB_START] = "start", + [JOB_VERIFY_ACTIVE] = "verify-active", + [JOB_STOP] = "stop", + [JOB_RELOAD] = "reload", + [JOB_RELOAD_OR_START] = "reload-or-start", + [JOB_RESTART] = "restart", + [JOB_TRY_RESTART] = "try-restart", + [JOB_NOP] = "nop", +}; + +DEFINE_STRING_TABLE_LOOKUP(job_type, JobType); + +static const char* const job_mode_table[_JOB_MODE_MAX] = { + [JOB_FAIL] = "fail", + [JOB_REPLACE] = "replace", + [JOB_REPLACE_IRREVERSIBLY] = "replace-irreversibly", + [JOB_ISOLATE] = "isolate", + [JOB_IGNORE_DEPENDENCIES] = "ignore-dependencies", + [JOB_IGNORE_REQUIREMENTS] = "ignore-requirements" +}; + +DEFINE_STRING_TABLE_LOOKUP(job_mode, JobMode); + +static const char* const job_result_table[_JOB_RESULT_MAX] = { + [JOB_DONE] = "done", + [JOB_CANCELED] = "canceled", + [JOB_TIMEOUT] = "timeout", + [JOB_FAILED] = "failed", + [JOB_DEPENDENCY] = "dependency", + [JOB_SKIPPED] = "skipped" +}; + +DEFINE_STRING_TABLE_LOOKUP(job_result, JobResult); diff --git a/src/core/job.h b/src/core/job.h new file mode 100644 index 0000000000..d90bc96b76 --- /dev/null +++ b/src/core/job.h @@ -0,0 +1,234 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> +#include <inttypes.h> +#include <errno.h> + +typedef struct Job Job; +typedef struct JobDependency JobDependency; +typedef struct JobBusClient JobBusClient; +typedef enum JobType JobType; +typedef enum JobState JobState; +typedef enum JobMode JobMode; +typedef enum JobResult JobResult; + +/* Be careful when changing the job types! Adjust job_merging_table[] accordingly! */ +enum JobType { + JOB_START, /* if a unit does not support being started, we'll just wait until it becomes active */ + JOB_VERIFY_ACTIVE, + + JOB_STOP, + + JOB_RELOAD, /* if running, reload */ + + /* Note that restarts are first treated like JOB_STOP, but + * then instead of finishing are patched to become + * JOB_START. */ + JOB_RESTART, /* If running, stop. Then start unconditionally. */ + + _JOB_TYPE_MAX_MERGING, + + /* JOB_NOP can enter into a transaction, but as it won't pull in + * any dependencies, it won't have to merge with anything. + * job_install() avoids the problem of merging JOB_NOP too (it's + * special-cased, only merges with other JOB_NOPs). */ + JOB_NOP = _JOB_TYPE_MAX_MERGING, /* do nothing */ + + _JOB_TYPE_MAX_IN_TRANSACTION, + + /* JOB_TRY_RESTART can never appear in a transaction, because + * it always collapses into JOB_RESTART or JOB_NOP before entering. + * Thus we never need to merge it with anything. */ + JOB_TRY_RESTART = _JOB_TYPE_MAX_IN_TRANSACTION, /* if running, stop and then start */ + + /* JOB_RELOAD_OR_START won't enter into a transaction and cannot result + * from transaction merging (there's no way for JOB_RELOAD and + * JOB_START to meet in one transaction). It can result from a merge + * during job installation, but then it will immediately collapse into + * one of the two simpler types. */ + JOB_RELOAD_OR_START, /* if running, reload, otherwise start */ + + _JOB_TYPE_MAX, + _JOB_TYPE_INVALID = -1 +}; + +enum JobState { + JOB_WAITING, + JOB_RUNNING, + _JOB_STATE_MAX, + _JOB_STATE_INVALID = -1 +}; + +enum JobMode { + JOB_FAIL, /* Fail if a conflicting job is already queued */ + JOB_REPLACE, /* Replace an existing conflicting job */ + JOB_REPLACE_IRREVERSIBLY, /* Like JOB_REPLACE + produce irreversible jobs */ + JOB_ISOLATE, /* Start a unit, and stop all others */ + JOB_IGNORE_DEPENDENCIES, /* Ignore both requirement and ordering dependencies */ + JOB_IGNORE_REQUIREMENTS, /* Ignore requirement dependencies */ + _JOB_MODE_MAX, + _JOB_MODE_INVALID = -1 +}; + +enum JobResult { + JOB_DONE, /* Job completed successfully */ + JOB_CANCELED, /* Job canceled by a conflicting job installation or by explicit cancel request */ + JOB_TIMEOUT, /* JobTimeout elapsed */ + JOB_FAILED, /* Job failed */ + JOB_DEPENDENCY, /* A required dependency job did not result in JOB_DONE */ + JOB_SKIPPED, /* JOB_RELOAD of inactive unit; negative result of JOB_VERIFY_ACTIVE */ + _JOB_RESULT_MAX, + _JOB_RESULT_INVALID = -1 +}; + +#include "manager.h" +#include "unit.h" +#include "hashmap.h" +#include "list.h" + +struct JobDependency { + /* Encodes that the 'subject' job needs the 'object' job in + * some way. This structure is used only while building a transaction. */ + Job *subject; + Job *object; + + LIST_FIELDS(JobDependency, subject); + LIST_FIELDS(JobDependency, object); + + bool matters; + bool conflicts; +}; + +struct JobBusClient { + LIST_FIELDS(JobBusClient, client); + /* Note that this bus object is not ref counted here. */ + DBusConnection *bus; + char name[0]; +}; + +struct Job { + Manager *manager; + Unit *unit; + + LIST_FIELDS(Job, transaction); + LIST_FIELDS(Job, run_queue); + LIST_FIELDS(Job, dbus_queue); + + LIST_HEAD(JobDependency, subject_list); + LIST_HEAD(JobDependency, object_list); + + /* Used for graph algs as a "I have been here" marker */ + Job* marker; + unsigned generation; + + uint32_t id; + + JobType type; + JobState state; + + Watch timer_watch; + + /* There can be more than one client, because of job merging. */ + LIST_HEAD(JobBusClient, bus_client_list); + + JobResult result; + + bool installed:1; + bool in_run_queue:1; + bool matters_to_anchor:1; + bool override:1; + bool in_dbus_queue:1; + bool sent_dbus_new_signal:1; + bool ignore_order:1; + bool forgot_bus_clients:1; + bool irreversible:1; +}; + +JobBusClient* job_bus_client_new(DBusConnection *connection, const char *name); + +Job* job_new(Unit *unit, JobType type); +Job* job_new_raw(Unit *unit); +void job_free(Job *job); +Job* job_install(Job *j); +int job_install_deserialized(Job *j); +void job_uninstall(Job *j); +void job_dump(Job *j, FILE*f, const char *prefix); +int job_serialize(Job *j, FILE *f, FDSet *fds); +int job_deserialize(Job *j, FILE *f, FDSet *fds); +int job_coldplug(Job *j); + +JobDependency* job_dependency_new(Job *subject, Job *object, bool matters, bool conflicts); +void job_dependency_free(JobDependency *l); + +int job_merge(Job *j, Job *other); + +JobType job_type_lookup_merge(JobType a, JobType b) _pure_; + +_pure_ static inline bool job_type_is_mergeable(JobType a, JobType b) { + return job_type_lookup_merge(a, b) >= 0; +} + +_pure_ static inline bool job_type_is_conflicting(JobType a, JobType b) { + return !job_type_is_mergeable(a, b); +} + +_pure_ static inline bool job_type_is_superset(JobType a, JobType b) { + /* Checks whether operation a is a "superset" of b in its actions */ + return a == job_type_lookup_merge(a, b); +} + +bool job_type_is_redundant(JobType a, UnitActiveState b) _pure_; + +/* Collapses a state-dependent job type into a simpler type by observing + * the state of the unit which it is going to be applied to. */ +void job_type_collapse(JobType *t, Unit *u); + +int job_type_merge_and_collapse(JobType *a, JobType b, Unit *u); + +bool job_is_runnable(Job *j); + +void job_add_to_run_queue(Job *j); +void job_add_to_dbus_queue(Job *j); + +int job_start_timer(Job *j); +void job_timer_event(Job *j, uint64_t n_elapsed, Watch *w); + +int job_run_and_invalidate(Job *j); +int job_finish_and_invalidate(Job *j, JobResult result, bool recursive); + +char *job_dbus_path(Job *j); + +void job_shutdown_magic(Job *j); + +const char* job_type_to_string(JobType t) _const_; +JobType job_type_from_string(const char *s) _pure_; + +const char* job_state_to_string(JobState t) _const_; +JobState job_state_from_string(const char *s) _pure_; + +const char* job_mode_to_string(JobMode t) _const_; +JobMode job_mode_from_string(const char *s) _pure_; + +const char* job_result_to_string(JobResult t) _const_; +JobResult job_result_from_string(const char *s) _pure_; diff --git a/src/core/kill.c b/src/core/kill.c new file mode 100644 index 0000000000..0775653f73 --- /dev/null +++ b/src/core/kill.c @@ -0,0 +1,63 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <string.h> + +#include "kill.h" +#include "util.h" + +void kill_context_init(KillContext *c) { + assert(c); + + c->kill_signal = SIGTERM; + c->send_sigkill = true; +} + +void kill_context_dump(KillContext *c, FILE *f, const char *prefix) { + assert(c); + + if (!prefix) + prefix = ""; + + fprintf(f, + "%sKillMode: %s\n" + "%sKillSignal: SIG%s\n" + "%sSendSIGKILL: %s\n", + prefix, kill_mode_to_string(c->kill_mode), + prefix, signal_to_string(c->kill_signal), + prefix, yes_no(c->send_sigkill)); +} + +static const char* const kill_mode_table[_KILL_MODE_MAX] = { + [KILL_CONTROL_GROUP] = "control-group", + [KILL_PROCESS] = "process", + [KILL_NONE] = "none" +}; + +DEFINE_STRING_TABLE_LOOKUP(kill_mode, KillMode); + +static const char* const kill_who_table[_KILL_WHO_MAX] = { + [KILL_MAIN] = "main", + [KILL_CONTROL] = "control", + [KILL_ALL] = "all" +}; + +DEFINE_STRING_TABLE_LOOKUP(kill_who, KillWho); diff --git a/src/core/kill.h b/src/core/kill.h new file mode 100644 index 0000000000..71a0513e84 --- /dev/null +++ b/src/core/kill.h @@ -0,0 +1,62 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct KillContext KillContext; + +#include <stdbool.h> +#include <stdio.h> + +#include "macro.h" + +typedef enum KillMode { + /* The kill mode is a property of a unit. */ + KILL_CONTROL_GROUP = 0, + KILL_PROCESS, + KILL_NONE, + _KILL_MODE_MAX, + _KILL_MODE_INVALID = -1 +} KillMode; + +struct KillContext { + KillMode kill_mode; + int kill_signal; + bool send_sigkill; +}; + +typedef enum KillWho { + /* Kill who is a property of an operation */ + KILL_MAIN, + KILL_CONTROL, + KILL_ALL, + _KILL_WHO_MAX, + _KILL_WHO_INVALID = -1 +} KillWho; + +void kill_context_init(KillContext *c); +void kill_context_dump(KillContext *c, FILE *f, const char *prefix); + +const char *kill_mode_to_string(KillMode k) _const_; +KillMode kill_mode_from_string(const char *s) _pure_; + +const char *kill_who_to_string(KillWho k) _const_; +KillWho kill_who_from_string(const char *s) _pure_; diff --git a/src/core/killall.c b/src/core/killall.c new file mode 100644 index 0000000000..a0f57455fb --- /dev/null +++ b/src/core/killall.c @@ -0,0 +1,210 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 ProFUSION embedded systems + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/wait.h> +#include <signal.h> +#include <errno.h> +#include <unistd.h> + +#include "util.h" +#include "def.h" +#include "killall.h" +#include "set.h" + +#define TIMEOUT_USEC (10 * USEC_PER_SEC) + +static bool ignore_proc(pid_t pid) { + _cleanup_fclose_ FILE *f = NULL; + char c; + size_t count; + uid_t uid; + int r; + + /* We are PID 1, let's not commit suicide */ + if (pid == 1) + return true; + + r = get_process_uid(pid, &uid); + if (r < 0) + return true; /* not really, but better safe than sorry */ + + /* Non-root processes otherwise are always subject to be killed */ + if (uid != 0) + return false; + + f = fopen(procfs_file_alloca(pid, "cmdline"), "re"); + if (!f) + return true; /* not really, but has the desired effect */ + + count = fread(&c, 1, 1, f); + + /* Kernel threads have an empty cmdline */ + if (count <= 0) + return true; + + /* Processes with argv[0][0] = '@' we ignore from the killing + * spree. + * + * http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons */ + if (count == 1 && c == '@') + return true; + + return false; +} + +static void wait_for_children(Set *pids, sigset_t *mask) { + usec_t until; + + assert(mask); + + if (set_isempty(pids)) + return; + + until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC; + for (;;) { + struct timespec ts; + int k; + usec_t n; + void *p; + Iterator i; + + /* First, let the kernel inform us about killed + * children. Most processes will probably be our + * children, but some are not (might be our + * grandchildren instead...). */ + for (;;) { + pid_t pid; + + pid = waitpid(-1, NULL, WNOHANG); + if (pid == 0) + break; + if (pid < 0) { + if (errno == ECHILD) + break; + + log_error("waitpid() failed: %m"); + return; + } + + set_remove(pids, ULONG_TO_PTR(pid)); + } + + /* Now explicitly check who might be remaining, who + * might not be our child. */ + SET_FOREACH(p, pids, i) { + + /* We misuse getpgid as a check whether a + * process still exists. */ + if (getpgid((pid_t) PTR_TO_ULONG(p)) >= 0) + continue; + + if (errno != ESRCH) + continue; + + set_remove(pids, p); + } + + if (set_isempty(pids)) + return; + + n = now(CLOCK_MONOTONIC); + if (n >= until) + return; + + timespec_store(&ts, until - n); + k = sigtimedwait(mask, NULL, &ts); + if (k != SIGCHLD) { + + if (k < 0 && errno != EAGAIN) { + log_error("sigtimedwait() failed: %m"); + return; + } + + if (k >= 0) + log_warning("sigtimedwait() returned unexpected signal."); + } + } +} + +static int killall(int sig, Set *pids) { + _cleanup_closedir_ DIR *dir = NULL; + struct dirent *d; + + dir = opendir("/proc"); + if (!dir) + return -errno; + + while ((d = readdir(dir))) { + pid_t pid; + + if (d->d_type != DT_DIR && + d->d_type != DT_UNKNOWN) + continue; + + if (parse_pid(d->d_name, &pid) < 0) + continue; + + if (ignore_proc(pid)) + continue; + + if (sig == SIGKILL) { + _cleanup_free_ char *s; + + get_process_comm(pid, &s); + log_notice("Sending SIGKILL to PID %lu (%s).", (unsigned long) pid, strna(s)); + } + + if (kill(pid, sig) >= 0) { + if (pids) + set_put(pids, ULONG_TO_PTR((unsigned long) pid)); + } else if (errno != ENOENT) + log_warning("Could not kill %d: %m", pid); + } + + return set_size(pids); +} + +void broadcast_signal(int sig, bool wait_for_exit) { + sigset_t mask, oldmask; + Set *pids = NULL; + + if (wait_for_exit) + pids = set_new(trivial_hash_func, trivial_compare_func); + + assert_se(sigemptyset(&mask) == 0); + assert_se(sigaddset(&mask, SIGCHLD) == 0); + assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0); + + if (kill(-1, SIGSTOP) < 0 && errno != ESRCH) + log_warning("kill(-1, SIGSTOP) failed: %m"); + + killall(sig, pids); + + if (kill(-1, SIGCONT) < 0 && errno != ESRCH) + log_warning("kill(-1, SIGCONT) failed: %m"); + + if (wait_for_exit) + wait_for_children(pids, &mask); + + assert_se(sigprocmask(SIG_SETMASK, &oldmask, NULL) == 0); + + set_free(pids); +} diff --git a/src/core/killall.h b/src/core/killall.h new file mode 100644 index 0000000000..d08ac142f0 --- /dev/null +++ b/src/core/killall.h @@ -0,0 +1,27 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#ifndef fookillallhfoo +#define fookillallhfoo + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +void broadcast_signal(int sig, bool wait); + +#endif diff --git a/src/core/kmod-setup.c b/src/core/kmod-setup.c new file mode 100644 index 0000000000..e4885fb212 --- /dev/null +++ b/src/core/kmod-setup.c @@ -0,0 +1,110 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/wait.h> +#include <unistd.h> +#include <string.h> +#include <errno.h> +#include <libkmod.h> + +#include "macro.h" +#include "execute.h" + +#include "kmod-setup.h" + +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wformat-nonliteral" + +static void systemd_kmod_log( + void *data, + int priority, + const char *file, int line, + const char *fn, + const char *format, + va_list args) { + + /* library logging is enabled at debug only */ + log_metav(LOG_DEBUG, file, line, fn, format, args); +} + +#pragma GCC diagnostic pop + +int kmod_setup(void) { + + static const char kmod_table[] = + /* This one we need to load explicitly, since + * auto-loading on use doesn't work before udev + * created the ghost device nodes, and we need it + * earlier than that. */ + "autofs4\0" "/sys/class/misc/autofs\0" + + /* This one we need to load explicitly, since + * auto-loading of IPv6 is not done when we try to + * configure ::1 on the loopback device. */ + "ipv6\0" "/sys/module/ipv6\0" + + "unix\0" "/proc/net/unix\0"; + + struct kmod_ctx *ctx = NULL; + const char *name, *path; + int r; + + NULSTR_FOREACH_PAIR(name, path, kmod_table) { + struct kmod_module *mod; + + if (access(path, F_OK) >= 0) + continue; + + log_debug("Your kernel apparently lacks built-in %s support. Might be a good idea to compile it in. " + "We'll now try to work around this by loading the module...", + name); + + if (!ctx) { + ctx = kmod_new(NULL, NULL); + if (!ctx) + return log_oom(); + + kmod_set_log_fn(ctx, systemd_kmod_log, NULL); + kmod_load_resources(ctx); + } + + r = kmod_module_new_from_name(ctx, name, &mod); + if (r < 0) { + log_error("Failed to lookup module '%s'", name); + continue; + } + + r = kmod_module_probe_insert_module(mod, KMOD_PROBE_APPLY_BLACKLIST, NULL, NULL, NULL, NULL); + if (r == 0) + log_info("Inserted module '%s'", kmod_module_get_name(mod)); + else if (r == KMOD_PROBE_APPLY_BLACKLIST) + log_info("Module '%s' is blacklisted", kmod_module_get_name(mod)); + else + log_error("Failed to insert module '%s'", kmod_module_get_name(mod)); + + kmod_module_unref(mod); + } + + if (ctx) + kmod_unref(ctx); + + return 0; +} diff --git a/src/core/kmod-setup.h b/src/core/kmod-setup.h new file mode 100644 index 0000000000..24dcdddfa4 --- /dev/null +++ b/src/core/kmod-setup.h @@ -0,0 +1,24 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int kmod_setup(void); diff --git a/src/core/load-dropin.c b/src/core/load-dropin.c new file mode 100644 index 0000000000..a877e66098 --- /dev/null +++ b/src/core/load-dropin.c @@ -0,0 +1,223 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dirent.h> +#include <errno.h> + +#include "unit.h" +#include "load-dropin.h" +#include "log.h" +#include "strv.h" +#include "unit-name.h" +#include "conf-parser.h" +#include "load-fragment.h" +#include "conf-files.h" + +static int iterate_dir( + Unit *u, + const char *path, + UnitDependency dependency, + char ***strv) { + + _cleanup_closedir_ DIR *d = NULL; + int r; + + assert(u); + assert(path); + + /* The config directories are special, since the order of the + * drop-ins matters */ + if (dependency < 0) { + r = strv_extend(strv, path); + if (r < 0) + return log_oom(); + + return 0; + } + + d = opendir(path); + if (!d) { + if (errno == ENOENT) + return 0; + + return -errno; + } + + for (;;) { + struct dirent *de; + union dirent_storage buf; + _cleanup_free_ char *f = NULL; + int k; + + k = readdir_r(d, &buf.de, &de); + if (k != 0) { + log_error("Failed to read directory %s: %s", path, strerror(k)); + return -k; + } + + if (!de) + break; + + if (ignore_file(de->d_name)) + continue; + + f = strjoin(path, "/", de->d_name, NULL); + if (!f) + return log_oom(); + + r = unit_add_dependency_by_name(u, dependency, de->d_name, f, true); + if (r < 0) + log_error("Cannot add dependency %s to %s, ignoring: %s", de->d_name, u->id, strerror(-r)); + } + + return 0; +} + +static int process_dir( + Unit *u, + const char *unit_path, + const char *name, + const char *suffix, + UnitDependency dependency, + char ***strv) { + + int r; + char *path; + + assert(u); + assert(unit_path); + assert(name); + assert(suffix); + + path = strjoin(unit_path, "/", name, suffix, NULL); + if (!path) + return log_oom(); + + if (u->manager->unit_path_cache && + !set_get(u->manager->unit_path_cache, path)) + r = 0; + else + r = iterate_dir(u, path, dependency, strv); + free(path); + + if (r < 0) + return r; + + if (u->instance) { + char *template; + /* Also try the template dir */ + + template = unit_name_template(name); + if (!template) + return log_oom(); + + path = strjoin(unit_path, "/", template, suffix, NULL); + free(template); + + if (!path) + return log_oom(); + + if (u->manager->unit_path_cache && + !set_get(u->manager->unit_path_cache, path)) + r = 0; + else + r = iterate_dir(u, path, dependency, strv); + free(path); + + if (r < 0) + return r; + } + + return 0; +} + +char **unit_find_dropin_paths(Unit *u) { + _cleanup_strv_free_ char **strv = NULL; + char **configs = NULL; + Iterator i; + char *t; + int r; + + assert(u); + + SET_FOREACH(t, u->names, i) { + char **p; + + STRV_FOREACH(p, u->manager->lookup_paths.unit_path) { + /* This loads the drop-in config snippets */ + r = process_dir(u, *p, t, ".d", _UNIT_DEPENDENCY_INVALID, &strv); + if (r < 0) + return NULL; + } + } + + if (strv_isempty(strv)) + return NULL; + + r = conf_files_list_strv(&configs, ".conf", NULL, (const char**) strv); + if (r < 0) { + log_error("Failed to get list of configuration files: %s", strerror(-r)); + strv_free(configs); + return NULL; + } + + return configs; +} + +int unit_load_dropin(Unit *u) { + Iterator i; + char *t, **f; + int r; + + assert(u); + + /* Load dependencies from supplementary drop-in directories */ + + SET_FOREACH(t, u->names, i) { + char **p; + + STRV_FOREACH(p, u->manager->lookup_paths.unit_path) { + r = process_dir(u, *p, t, ".wants", UNIT_WANTS, NULL); + if (r < 0) + return r; + + r = process_dir(u, *p, t, ".requires", UNIT_REQUIRES, NULL); + if (r < 0) + return r; + } + } + + u->dropin_paths = unit_find_dropin_paths(u); + if (! u->dropin_paths) + return 0; + + STRV_FOREACH(f, u->dropin_paths) { + r = config_parse(u->id, *f, NULL, + UNIT_VTABLE(u)->sections, config_item_perf_lookup, + (void*) load_fragment_gperf_lookup, false, false, u); + if (r < 0) + return r; + } + + u->dropin_mtime = now(CLOCK_REALTIME); + + return 0; +} diff --git a/src/core/load-dropin.h b/src/core/load-dropin.h new file mode 100644 index 0000000000..fd551179e2 --- /dev/null +++ b/src/core/load-dropin.h @@ -0,0 +1,29 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "unit.h" + +/* Read service data supplementary drop-in directories */ + +char **unit_find_dropin_paths(Unit *u); +int unit_load_dropin(Unit *u); diff --git a/src/core/load-fragment-gperf.gperf.m4 b/src/core/load-fragment-gperf.gperf.m4 new file mode 100644 index 0000000000..4e1454ee6c --- /dev/null +++ b/src/core/load-fragment-gperf.gperf.m4 @@ -0,0 +1,258 @@ +%{ +#include <stddef.h> +#include "conf-parser.h" +#include "load-fragment.h" +#include "missing.h" +%} +struct ConfigPerfItem; +%null_strings +%language=ANSI-C +%define slot-name section_and_lvalue +%define hash-function-name load_fragment_gperf_hash +%define lookup-function-name load_fragment_gperf_lookup +%readonly-tables +%omit-struct-type +%struct-type +%includes +%% +m4_dnl Define the context options only once +m4_define(`EXEC_CONTEXT_CONFIG_ITEMS', +`$1.WorkingDirectory, config_parse_unit_path_printf, 0, offsetof($1, exec_context.working_directory) +$1.RootDirectory, config_parse_unit_path_printf, 0, offsetof($1, exec_context.root_directory) +$1.User, config_parse_unit_string_printf, 0, offsetof($1, exec_context.user) +$1.Group, config_parse_unit_string_printf, 0, offsetof($1, exec_context.group) +$1.SupplementaryGroups, config_parse_strv, 0, offsetof($1, exec_context.supplementary_groups) +$1.Nice, config_parse_exec_nice, 0, offsetof($1, exec_context) +$1.OOMScoreAdjust, config_parse_exec_oom_score_adjust, 0, offsetof($1, exec_context) +$1.IOSchedulingClass, config_parse_exec_io_class, 0, offsetof($1, exec_context) +$1.IOSchedulingPriority, config_parse_exec_io_priority, 0, offsetof($1, exec_context) +$1.CPUSchedulingPolicy, config_parse_exec_cpu_sched_policy, 0, offsetof($1, exec_context) +$1.CPUSchedulingPriority, config_parse_exec_cpu_sched_prio, 0, offsetof($1, exec_context) +$1.CPUSchedulingResetOnFork, config_parse_bool, 0, offsetof($1, exec_context.cpu_sched_reset_on_fork) +$1.CPUAffinity, config_parse_exec_cpu_affinity, 0, offsetof($1, exec_context) +$1.UMask, config_parse_mode, 0, offsetof($1, exec_context.umask) +$1.Environment, config_parse_environ, 0, offsetof($1, exec_context.environment) +$1.EnvironmentFile, config_parse_unit_env_file, 0, offsetof($1, exec_context.environment_files) +$1.StandardInput, config_parse_input, 0, offsetof($1, exec_context.std_input) +$1.StandardOutput, config_parse_output, 0, offsetof($1, exec_context.std_output) +$1.StandardError, config_parse_output, 0, offsetof($1, exec_context.std_error) +$1.TTYPath, config_parse_unit_path_printf, 0, offsetof($1, exec_context.tty_path) +$1.TTYReset, config_parse_bool, 0, offsetof($1, exec_context.tty_reset) +$1.TTYVHangup, config_parse_bool, 0, offsetof($1, exec_context.tty_vhangup) +$1.TTYVTDisallocate, config_parse_bool, 0, offsetof($1, exec_context.tty_vt_disallocate) +$1.SyslogIdentifier, config_parse_unit_string_printf, 0, offsetof($1, exec_context.syslog_identifier) +$1.SyslogFacility, config_parse_facility, 0, offsetof($1, exec_context.syslog_priority) +$1.SyslogLevel, config_parse_level, 0, offsetof($1, exec_context.syslog_priority) +$1.SyslogLevelPrefix, config_parse_bool, 0, offsetof($1, exec_context.syslog_level_prefix) +$1.Capabilities, config_parse_exec_capabilities, 0, offsetof($1, exec_context) +$1.SecureBits, config_parse_exec_secure_bits, 0, offsetof($1, exec_context) +$1.CapabilityBoundingSet, config_parse_bounding_set, 0, offsetof($1, exec_context.capability_bounding_set_drop) +$1.TimerSlackNSec, config_parse_nsec, 0, offsetof($1, exec_context.timer_slack_nsec) +$1.NoNewPrivileges, config_parse_bool, 0, offsetof($1, exec_context.no_new_privileges) +$1.SystemCallFilter, config_parse_syscall_filter, 0, offsetof($1, exec_context) +$1.LimitCPU, config_parse_limit, RLIMIT_CPU, offsetof($1, exec_context.rlimit) +$1.LimitFSIZE, config_parse_limit, RLIMIT_FSIZE, offsetof($1, exec_context.rlimit) +$1.LimitDATA, config_parse_limit, RLIMIT_DATA, offsetof($1, exec_context.rlimit) +$1.LimitSTACK, config_parse_limit, RLIMIT_STACK, offsetof($1, exec_context.rlimit) +$1.LimitCORE, config_parse_limit, RLIMIT_CORE, offsetof($1, exec_context.rlimit) +$1.LimitRSS, config_parse_limit, RLIMIT_RSS, offsetof($1, exec_context.rlimit) +$1.LimitNOFILE, config_parse_limit, RLIMIT_NOFILE, offsetof($1, exec_context.rlimit) +$1.LimitAS, config_parse_limit, RLIMIT_AS, offsetof($1, exec_context.rlimit) +$1.LimitNPROC, config_parse_limit, RLIMIT_NPROC, offsetof($1, exec_context.rlimit) +$1.LimitMEMLOCK, config_parse_limit, RLIMIT_MEMLOCK, offsetof($1, exec_context.rlimit) +$1.LimitLOCKS, config_parse_limit, RLIMIT_LOCKS, offsetof($1, exec_context.rlimit) +$1.LimitSIGPENDING, config_parse_limit, RLIMIT_SIGPENDING, offsetof($1, exec_context.rlimit) +$1.LimitMSGQUEUE, config_parse_limit, RLIMIT_MSGQUEUE, offsetof($1, exec_context.rlimit) +$1.LimitNICE, config_parse_limit, RLIMIT_NICE, offsetof($1, exec_context.rlimit) +$1.LimitRTPRIO, config_parse_limit, RLIMIT_RTPRIO, offsetof($1, exec_context.rlimit) +$1.LimitRTTIME, config_parse_limit, RLIMIT_RTTIME, offsetof($1, exec_context.rlimit) +$1.ControlGroup, config_parse_unit_cgroup, 0, 0 +$1.ControlGroupAttribute, config_parse_unit_cgroup_attr, 0, 0 +$1.CPUShares, config_parse_unit_cgroup_attr_pretty, 0, 0 +$1.MemoryLimit, config_parse_unit_cgroup_attr_pretty, 0, 0 +$1.MemorySoftLimit, config_parse_unit_cgroup_attr_pretty, 0, 0 +$1.DeviceAllow, config_parse_unit_cgroup_attr_pretty, 0, 0 +$1.DeviceDeny, config_parse_unit_cgroup_attr_pretty, 0, 0 +$1.BlockIOWeight, config_parse_unit_cgroup_attr_pretty, 0, 0 +$1.BlockIOReadBandwidth, config_parse_unit_cgroup_attr_pretty, 0, 0 +$1.BlockIOWriteBandwidth, config_parse_unit_cgroup_attr_pretty, 0, 0 +$1.ReadWriteDirectories, config_parse_path_strv, 0, offsetof($1, exec_context.read_write_dirs) +$1.ReadOnlyDirectories, config_parse_path_strv, 0, offsetof($1, exec_context.read_only_dirs) +$1.InaccessibleDirectories, config_parse_path_strv, 0, offsetof($1, exec_context.inaccessible_dirs) +$1.PrivateTmp, config_parse_bool, 0, offsetof($1, exec_context.private_tmp) +$1.PrivateNetwork, config_parse_bool, 0, offsetof($1, exec_context.private_network) +$1.MountFlags, config_parse_exec_mount_flags, 0, offsetof($1, exec_context) +$1.TCPWrapName, config_parse_unit_string_printf, 0, offsetof($1, exec_context.tcpwrap_name) +$1.PAMName, config_parse_unit_string_printf, 0, offsetof($1, exec_context.pam_name) +$1.IgnoreSIGPIPE, config_parse_bool, 0, offsetof($1, exec_context.ignore_sigpipe) +$1.UtmpIdentifier, config_parse_unit_string_printf, 0, offsetof($1, exec_context.utmp_id) +$1.ControlGroupModify, config_parse_bool, 0, offsetof($1, exec_context.control_group_modify) +$1.ControlGroupPersistent, config_parse_tristate, 0, offsetof($1, exec_context.control_group_persistent)' +)m4_dnl +m4_define(`KILL_CONTEXT_CONFIG_ITEMS', +`$1.SendSIGKILL, config_parse_bool, 0, offsetof($1, kill_context.send_sigkill) +$1.KillMode, config_parse_kill_mode, 0, offsetof($1, kill_context.kill_mode) +$1.KillSignal, config_parse_kill_signal, 0, offsetof($1, kill_context.kill_signal)' +)m4_dnl +Unit.Description, config_parse_unit_string_printf, 0, offsetof(Unit, description) +Unit.Documentation, config_parse_documentation, 0, offsetof(Unit, documentation) +Unit.SourcePath, config_parse_path, 0, offsetof(Unit, source_path) +Unit.Requires, config_parse_unit_deps, UNIT_REQUIRES, 0 +Unit.RequiresOverridable, config_parse_unit_deps, UNIT_REQUIRES_OVERRIDABLE, 0 +Unit.Requisite, config_parse_unit_deps, UNIT_REQUISITE, 0 +Unit.RequisiteOverridable, config_parse_unit_deps, UNIT_REQUISITE_OVERRIDABLE, 0 +Unit.Wants, config_parse_unit_deps, UNIT_WANTS, 0 +Unit.BindsTo, config_parse_unit_deps, UNIT_BINDS_TO, 0 +Unit.BindTo, config_parse_unit_deps, UNIT_BINDS_TO, 0 +Unit.Conflicts, config_parse_unit_deps, UNIT_CONFLICTS, 0 +Unit.Before, config_parse_unit_deps, UNIT_BEFORE, 0 +Unit.After, config_parse_unit_deps, UNIT_AFTER, 0 +Unit.OnFailure, config_parse_unit_deps, UNIT_ON_FAILURE, 0 +Unit.PropagatesReloadTo, config_parse_unit_deps, UNIT_PROPAGATES_RELOAD_TO, 0 +Unit.PropagateReloadTo, config_parse_unit_deps, UNIT_PROPAGATES_RELOAD_TO, 0 +Unit.ReloadPropagatedFrom, config_parse_unit_deps, UNIT_RELOAD_PROPAGATED_FROM, 0 +Unit.PropagateReloadFrom, config_parse_unit_deps, UNIT_RELOAD_PROPAGATED_FROM, 0 +Unit.PartOf, config_parse_unit_deps, UNIT_PART_OF, 0 +Unit.RequiresMountsFor, config_parse_unit_requires_mounts_for, 0, offsetof(Unit, requires_mounts_for) +Unit.StopWhenUnneeded, config_parse_bool, 0, offsetof(Unit, stop_when_unneeded) +Unit.RefuseManualStart, config_parse_bool, 0, offsetof(Unit, refuse_manual_start) +Unit.RefuseManualStop, config_parse_bool, 0, offsetof(Unit, refuse_manual_stop) +Unit.AllowIsolate, config_parse_bool, 0, offsetof(Unit, allow_isolate) +Unit.DefaultDependencies, config_parse_bool, 0, offsetof(Unit, default_dependencies) +Unit.OnFailureIsolate, config_parse_bool, 0, offsetof(Unit, on_failure_isolate) +Unit.IgnoreOnIsolate, config_parse_bool, 0, offsetof(Unit, ignore_on_isolate) +Unit.IgnoreOnSnapshot, config_parse_bool, 0, offsetof(Unit, ignore_on_snapshot) +Unit.JobTimeoutSec, config_parse_sec, 0, offsetof(Unit, job_timeout) +Unit.ConditionPathExists, config_parse_unit_condition_path, CONDITION_PATH_EXISTS, 0 +Unit.ConditionPathExistsGlob, config_parse_unit_condition_path, CONDITION_PATH_EXISTS_GLOB, 0 +Unit.ConditionPathIsDirectory, config_parse_unit_condition_path, CONDITION_PATH_IS_DIRECTORY, 0 +Unit.ConditionPathIsSymbolicLink,config_parse_unit_condition_path, CONDITION_PATH_IS_SYMBOLIC_LINK,0 +Unit.ConditionPathIsMountPoint, config_parse_unit_condition_path, CONDITION_PATH_IS_MOUNT_POINT, 0 +Unit.ConditionPathIsReadWrite, config_parse_unit_condition_path, CONDITION_PATH_IS_READ_WRITE, 0 +Unit.ConditionDirectoryNotEmpty, config_parse_unit_condition_path, CONDITION_DIRECTORY_NOT_EMPTY, 0 +Unit.ConditionFileNotEmpty, config_parse_unit_condition_path, CONDITION_FILE_NOT_EMPTY, 0 +Unit.ConditionFileIsExecutable, config_parse_unit_condition_path, CONDITION_FILE_IS_EXECUTABLE, 0 +Unit.ConditionKernelCommandLine, config_parse_unit_condition_string, CONDITION_KERNEL_COMMAND_LINE, 0 +Unit.ConditionVirtualization, config_parse_unit_condition_string, CONDITION_VIRTUALIZATION, 0 +Unit.ConditionSecurity, config_parse_unit_condition_string, CONDITION_SECURITY, 0 +Unit.ConditionCapability, config_parse_unit_condition_string, CONDITION_CAPABILITY, 0 +Unit.ConditionHost, config_parse_unit_condition_string, CONDITION_HOST, 0 +Unit.ConditionACPower, config_parse_unit_condition_string, CONDITION_AC_POWER, 0 +Unit.ConditionNull, config_parse_unit_condition_null, 0, 0 +m4_dnl +Service.PIDFile, config_parse_unit_path_printf, 0, offsetof(Service, pid_file) +Service.ExecStartPre, config_parse_exec, SERVICE_EXEC_START_PRE, offsetof(Service, exec_command) +Service.ExecStart, config_parse_exec, SERVICE_EXEC_START, offsetof(Service, exec_command) +Service.ExecStartPost, config_parse_exec, SERVICE_EXEC_START_POST, offsetof(Service, exec_command) +Service.ExecReload, config_parse_exec, SERVICE_EXEC_RELOAD, offsetof(Service, exec_command) +Service.ExecStop, config_parse_exec, SERVICE_EXEC_STOP, offsetof(Service, exec_command) +Service.ExecStopPost, config_parse_exec, SERVICE_EXEC_STOP_POST, offsetof(Service, exec_command) +Service.RestartSec, config_parse_sec, 0, offsetof(Service, restart_usec) +Service.TimeoutSec, config_parse_service_timeout, 0, offsetof(Service, timeout_start_usec) +Service.TimeoutStartSec, config_parse_service_timeout, 0, offsetof(Service, timeout_start_usec) +Service.TimeoutStopSec, config_parse_service_timeout, 0, offsetof(Service, timeout_stop_usec) +Service.WatchdogSec, config_parse_sec, 0, offsetof(Service, watchdog_usec) +Service.StartLimitInterval, config_parse_sec, 0, offsetof(Service, start_limit.interval) +Service.StartLimitBurst, config_parse_unsigned, 0, offsetof(Service, start_limit.burst) +Service.StartLimitAction, config_parse_start_limit_action, 0, offsetof(Service, start_limit_action) +Service.Type, config_parse_service_type, 0, offsetof(Service, type) +Service.Restart, config_parse_service_restart, 0, offsetof(Service, restart) +Service.PermissionsStartOnly, config_parse_bool, 0, offsetof(Service, permissions_start_only) +Service.RootDirectoryStartOnly, config_parse_bool, 0, offsetof(Service, root_directory_start_only) +Service.RemainAfterExit, config_parse_bool, 0, offsetof(Service, remain_after_exit) +Service.GuessMainPID, config_parse_bool, 0, offsetof(Service, guess_main_pid) +Service.RestartPreventExitStatus, config_parse_set_status, 0, offsetof(Service, restart_ignore_status) +Service.SuccessExitStatus, config_parse_set_status, 0, offsetof(Service, success_status) +m4_ifdef(`HAVE_SYSV_COMPAT', +`Service.SysVStartPriority, config_parse_sysv_priority, 0, offsetof(Service, sysv_start_priority)', +`Service.SysVStartPriority, config_parse_warn_compat, 0, 0') +Service.NonBlocking, config_parse_bool, 0, offsetof(Service, exec_context.non_blocking) +Service.BusName, config_parse_unit_string_printf, 0, offsetof(Service, bus_name) +Service.NotifyAccess, config_parse_notify_access, 0, offsetof(Service, notify_access) +Service.Sockets, config_parse_service_sockets, 0, 0 +Service.FsckPassNo, config_parse_fsck_passno, 0, offsetof(Service, fsck_passno) +EXEC_CONTEXT_CONFIG_ITEMS(Service)m4_dnl +KILL_CONTEXT_CONFIG_ITEMS(Service)m4_dnl +m4_dnl +Socket.ListenStream, config_parse_socket_listen, SOCKET_SOCKET, 0 +Socket.ListenDatagram, config_parse_socket_listen, SOCKET_SOCKET, 0 +Socket.ListenSequentialPacket, config_parse_socket_listen, SOCKET_SOCKET, 0 +Socket.ListenFIFO, config_parse_socket_listen, SOCKET_FIFO, 0 +Socket.ListenNetlink, config_parse_socket_listen, SOCKET_SOCKET, 0 +Socket.ListenSpecial, config_parse_socket_listen, SOCKET_SPECIAL, 0 +Socket.ListenMessageQueue, config_parse_socket_listen, SOCKET_MQUEUE, 0 +Socket.BindIPv6Only, config_parse_socket_bind, 0, 0, +Socket.Backlog, config_parse_unsigned, 0, offsetof(Socket, backlog) +Socket.BindToDevice, config_parse_socket_bindtodevice, 0, 0 +Socket.ExecStartPre, config_parse_exec, SOCKET_EXEC_START_PRE, offsetof(Socket, exec_command) +Socket.ExecStartPost, config_parse_exec, SOCKET_EXEC_START_POST, offsetof(Socket, exec_command) +Socket.ExecStopPre, config_parse_exec, SOCKET_EXEC_STOP_PRE, offsetof(Socket, exec_command) +Socket.ExecStopPost, config_parse_exec, SOCKET_EXEC_STOP_POST, offsetof(Socket, exec_command) +Socket.TimeoutSec, config_parse_sec, 0, offsetof(Socket, timeout_usec) +Socket.DirectoryMode, config_parse_mode, 0, offsetof(Socket, directory_mode) +Socket.SocketMode, config_parse_mode, 0, offsetof(Socket, socket_mode) +Socket.Accept, config_parse_bool, 0, offsetof(Socket, accept) +Socket.MaxConnections, config_parse_unsigned, 0, offsetof(Socket, max_connections) +Socket.KeepAlive, config_parse_bool, 0, offsetof(Socket, keep_alive) +Socket.Priority, config_parse_int, 0, offsetof(Socket, priority) +Socket.ReceiveBuffer, config_parse_bytes_size, 0, offsetof(Socket, receive_buffer) +Socket.SendBuffer, config_parse_bytes_size, 0, offsetof(Socket, send_buffer) +Socket.IPTOS, config_parse_ip_tos, 0, offsetof(Socket, ip_tos) +Socket.IPTTL, config_parse_int, 0, offsetof(Socket, ip_ttl) +Socket.Mark, config_parse_int, 0, offsetof(Socket, mark) +Socket.PipeSize, config_parse_bytes_size, 0, offsetof(Socket, pipe_size) +Socket.FreeBind, config_parse_bool, 0, offsetof(Socket, free_bind) +Socket.Transparent, config_parse_bool, 0, offsetof(Socket, transparent) +Socket.Broadcast, config_parse_bool, 0, offsetof(Socket, broadcast) +Socket.PassCredentials, config_parse_bool, 0, offsetof(Socket, pass_cred) +Socket.PassSecurity, config_parse_bool, 0, offsetof(Socket, pass_sec) +Socket.TCPCongestion, config_parse_string, 0, offsetof(Socket, tcp_congestion) +Socket.MessageQueueMaxMessages, config_parse_long, 0, offsetof(Socket, mq_maxmsg) +Socket.MessageQueueMessageSize, config_parse_long, 0, offsetof(Socket, mq_msgsize) +Socket.Service, config_parse_socket_service, 0, 0 +Socket.SmackLabel, config_parse_string, 0, offsetof(Socket, smack) +Socket.SmackLabelIPIn, config_parse_string, 0, offsetof(Socket, smack_ip_in) +Socket.SmackLabelIPOut, config_parse_string, 0, offsetof(Socket, smack_ip_out) +EXEC_CONTEXT_CONFIG_ITEMS(Socket)m4_dnl +KILL_CONTEXT_CONFIG_ITEMS(Socket)m4_dnl +m4_dnl +Mount.What, config_parse_string, 0, offsetof(Mount, parameters_fragment.what) +Mount.Where, config_parse_path, 0, offsetof(Mount, where) +Mount.Options, config_parse_string, 0, offsetof(Mount, parameters_fragment.options) +Mount.Type, config_parse_string, 0, offsetof(Mount, parameters_fragment.fstype) +Mount.FsckPassNo, config_parse_fsck_passno, 0, offsetof(Mount, parameters_fragment.passno) +Mount.TimeoutSec, config_parse_sec, 0, offsetof(Mount, timeout_usec) +Mount.DirectoryMode, config_parse_mode, 0, offsetof(Mount, directory_mode) +EXEC_CONTEXT_CONFIG_ITEMS(Mount)m4_dnl +KILL_CONTEXT_CONFIG_ITEMS(Mount)m4_dnl +m4_dnl +Automount.Where, config_parse_path, 0, offsetof(Automount, where) +Automount.DirectoryMode, config_parse_mode, 0, offsetof(Automount, directory_mode) +m4_dnl +Swap.What, config_parse_path, 0, offsetof(Swap, parameters_fragment.what) +Swap.Priority, config_parse_int, 0, offsetof(Swap, parameters_fragment.priority) +Swap.TimeoutSec, config_parse_sec, 0, offsetof(Swap, timeout_usec) +EXEC_CONTEXT_CONFIG_ITEMS(Swap)m4_dnl +KILL_CONTEXT_CONFIG_ITEMS(Swap)m4_dnl +m4_dnl +Timer.OnCalendar, config_parse_timer, 0, 0 +Timer.OnActiveSec, config_parse_timer, 0, 0 +Timer.OnBootSec, config_parse_timer, 0, 0 +Timer.OnStartupSec, config_parse_timer, 0, 0 +Timer.OnUnitActiveSec, config_parse_timer, 0, 0 +Timer.OnUnitInactiveSec, config_parse_timer, 0, 0 +Timer.Unit, config_parse_trigger_unit, 0, 0 +m4_dnl +Path.PathExists, config_parse_path_spec, 0, 0 +Path.PathExistsGlob, config_parse_path_spec, 0, 0 +Path.PathChanged, config_parse_path_spec, 0, 0 +Path.PathModified, config_parse_path_spec, 0, 0 +Path.DirectoryNotEmpty, config_parse_path_spec, 0, 0 +Path.Unit, config_parse_trigger_unit, 0, 0 +Path.MakeDirectory, config_parse_bool, 0, offsetof(Path, make_directory) +Path.DirectoryMode, config_parse_mode, 0, offsetof(Path, directory_mode) +m4_dnl The [Install] section is ignored here. +Install.Alias, NULL, 0, 0 +Install.WantedBy, NULL, 0, 0 +Install.RequiredBy, NULL, 0, 0 +Install.Also, NULL, 0, 0 diff --git a/src/core/load-fragment.c b/src/core/load-fragment.c new file mode 100644 index 0000000000..e2015ed58f --- /dev/null +++ b/src/core/load-fragment.c @@ -0,0 +1,2486 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2012 Holger Hans Peter Freyther + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <linux/oom.h> +#include <assert.h> +#include <errno.h> +#include <string.h> +#include <unistd.h> +#include <fcntl.h> +#include <sched.h> +#include <sys/prctl.h> +#include <sys/mount.h> +#include <linux/fs.h> +#include <sys/stat.h> +#include <sys/time.h> +#include <sys/resource.h> + +#include <systemd/sd-messages.h> + +#include "unit.h" +#include "strv.h" +#include "conf-parser.h" +#include "load-fragment.h" +#include "log.h" +#include "ioprio.h" +#include "securebits.h" +#include "missing.h" +#include "unit-name.h" +#include "unit-printf.h" +#include "dbus-common.h" +#include "utf8.h" +#include "path-util.h" +#include "syscall-list.h" +#include "env-util.h" + +#ifndef HAVE_SYSV_COMPAT +int config_parse_warn_compat(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + log_syntax(unit, LOG_DEBUG, filename, line, EINVAL, + "Support for option %s= has been disabled at compile time and is ignored", + lvalue); + return 0; +} +#endif + +int config_parse_unit_deps(const char* unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + UnitDependency d = ltype; + Unit *u = userdata; + char *w; + size_t l; + char *state; + + assert(filename); + assert(lvalue); + assert(rvalue); + + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + _cleanup_free_ char *t = NULL, *k = NULL; + int r; + + t = strndup(w, l); + if (!t) + return log_oom(); + + k = unit_name_printf(u, t); + if (!k) + return log_oom(); + + r = unit_add_dependency_by_name(u, d, k, NULL, true); + if (r < 0) + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to add dependency on %s, ignoring: %s", k, strerror(-r)); + } + + return 0; +} + +int config_parse_unit_string_printf(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + _cleanup_free_ char *k = NULL; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + k = unit_full_printf(u, rvalue); + if (!k) + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to resolve unit specifiers on %s. Ignoring.", rvalue); + + return config_parse_string(unit, filename, line, section, lvalue, ltype, + k ? k : rvalue, data, userdata); +} + +int config_parse_unit_strv_printf(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + _cleanup_free_ char *k = NULL; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + k = unit_full_printf(u, rvalue); + if (!k) + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to resolve unit specifiers on %s. Ignoring.", rvalue); + + return config_parse_strv(unit, filename, line, section, lvalue, ltype, + k ? k : rvalue, data, userdata); +} + +int config_parse_unit_path_printf(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + _cleanup_free_ char *k = NULL; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + k = unit_full_printf(u, rvalue); + if (!k) + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to resolve unit specifiers on %s. Ignoring.", rvalue); + + return config_parse_path(unit, filename, line, section, lvalue, ltype, + k ? k : rvalue, data, userdata); +} + +int config_parse_socket_listen(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + SocketPort *p, *tail; + Socket *s; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + s = SOCKET(data); + + if (isempty(rvalue)) { + /* An empty assignment removes all ports */ + socket_free_ports(s); + return 0; + } + + p = new0(SocketPort, 1); + if (!p) + return log_oom(); + + if (ltype != SOCKET_SOCKET) { + + p->type = ltype; + p->path = unit_full_printf(UNIT(s), rvalue); + if (!p->path) { + p->path = strdup(rvalue); + if (!p->path) { + free(p); + return log_oom(); + } else + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to resolve unit specifiers on %s. Ignoring.", rvalue); + } + + path_kill_slashes(p->path); + + } else if (streq(lvalue, "ListenNetlink")) { + _cleanup_free_ char *k = NULL; + int r; + + p->type = SOCKET_SOCKET; + k = unit_full_printf(UNIT(s), rvalue); + if (!k) + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to resolve unit specifiers on %s. Ignoring.", rvalue); + + r = socket_address_parse_netlink(&p->address, k ? k : rvalue); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse address value, ignoring: %s", rvalue); + free(p); + return 0; + } + + } else { + _cleanup_free_ char *k = NULL; + int r; + + p->type = SOCKET_SOCKET; + k = unit_full_printf(UNIT(s), rvalue); + if (!k) + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to resolve unit specifiers on %s. Ignoring.", rvalue); + + r = socket_address_parse(&p->address, k ? k : rvalue); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse address value, ignoring: %s", rvalue); + free(p); + return 0; + } + + if (streq(lvalue, "ListenStream")) + p->address.type = SOCK_STREAM; + else if (streq(lvalue, "ListenDatagram")) + p->address.type = SOCK_DGRAM; + else { + assert(streq(lvalue, "ListenSequentialPacket")); + p->address.type = SOCK_SEQPACKET; + } + + if (socket_address_family(&p->address) != AF_LOCAL && p->address.type == SOCK_SEQPACKET) { + log_syntax(unit, LOG_ERR, filename, line, ENOTSUP, + "Address family not supported, ignoring: %s", rvalue); + free(p); + return 0; + } + } + + p->fd = -1; + + if (s->ports) { + LIST_FIND_TAIL(SocketPort, port, s->ports, tail); + LIST_INSERT_AFTER(SocketPort, port, s->ports, tail, p); + } else + LIST_PREPEND(SocketPort, port, s->ports, p); + + return 0; +} + +int config_parse_socket_bind(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Socket *s; + SocketAddressBindIPv6Only b; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + s = SOCKET(data); + + b = socket_address_bind_ipv6_only_from_string(rvalue); + if (b < 0) { + int r; + + r = parse_boolean(rvalue); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse bind IPv6 only value, ignoring: %s", rvalue); + return 0; + } + + s->bind_ipv6_only = r ? SOCKET_ADDRESS_IPV6_ONLY : SOCKET_ADDRESS_BOTH; + } else + s->bind_ipv6_only = b; + + return 0; +} + +int config_parse_exec_nice(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int priority, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &priority); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse nice priority, ignoring: %s. ", rvalue); + return 0; + } + + if (priority < PRIO_MIN || priority >= PRIO_MAX) { + log_syntax(unit, LOG_ERR, filename, line, ERANGE, + "Nice priority out of range, ignoring: %s", rvalue); + return 0; + } + + c->nice = priority; + c->nice_set = true; + + return 0; +} + +int config_parse_exec_oom_score_adjust(const char* unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int oa, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &oa); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse the OOM score adjust value, ignoring: %s", rvalue); + return 0; + } + + if (oa < OOM_SCORE_ADJ_MIN || oa > OOM_SCORE_ADJ_MAX) { + log_syntax(unit, LOG_ERR, filename, line, ERANGE, + "OOM score adjust value out of range, ignoring: %s", rvalue); + return 0; + } + + c->oom_score_adjust = oa; + c->oom_score_adjust_set = true; + + return 0; +} + +int config_parse_exec(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecCommand **e = data, *nce; + char *path, **n; + unsigned k; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(e); + + e += ltype; + + if (isempty(rvalue)) { + /* An empty assignment resets the list */ + exec_command_free_list(*e); + *e = NULL; + return 0; + } + + /* We accept an absolute path as first argument, or + * alternatively an absolute prefixed with @ to allow + * overriding of argv[0]. */ + for (;;) { + int i; + char *w; + size_t l; + char *state; + bool honour_argv0 = false, ignore = false; + + path = NULL; + nce = NULL; + n = NULL; + + rvalue += strspn(rvalue, WHITESPACE); + + if (rvalue[0] == 0) + break; + + for (i = 0; i < 2; i++) { + if (rvalue[0] == '-' && !ignore) { + ignore = true; + rvalue ++; + } + + if (rvalue[0] == '@' && !honour_argv0) { + honour_argv0 = true; + rvalue ++; + } + } + + if (*rvalue != '/') { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Executable path is not absolute, ignoring: %s", rvalue); + return 0; + } + + k = 0; + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + if (strneq(w, ";", MAX(l, 1U))) + break; + + k++; + } + + n = new(char*, k + !honour_argv0); + if (!n) + return log_oom(); + + k = 0; + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + if (strneq(w, ";", MAX(l, 1U))) + break; + else if (strneq(w, "\\;", MAX(l, 1U))) + w ++; + + if (honour_argv0 && w == rvalue) { + assert(!path); + + path = strndup(w, l); + if (!path) { + r = log_oom(); + goto fail; + } + + if (!utf8_is_valid(path)) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Path is not UTF-8 clean, ignoring assignment: %s", + rvalue); + r = 0; + goto fail; + } + + } else { + char *c; + + c = n[k++] = cunescape_length(w, l); + if (!c) { + r = log_oom(); + goto fail; + } + + if (!utf8_is_valid(c)) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Path is not UTF-8 clean, ignoring assignment: %s", + rvalue); + r = 0; + goto fail; + } + } + } + + n[k] = NULL; + + if (!n[0]) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Invalid command line, ignoring: %s", rvalue); + r = 0; + goto fail; + } + + if (!path) { + path = strdup(n[0]); + if (!path) { + r = log_oom(); + goto fail; + } + } + + assert(path_is_absolute(path)); + + nce = new0(ExecCommand, 1); + if (!nce) { + r = log_oom(); + goto fail; + } + + nce->argv = n; + nce->path = path; + nce->ignore = ignore; + + path_kill_slashes(nce->path); + + exec_command_append_list(e, nce); + + rvalue = state; + } + + return 0; + +fail: + n[k] = NULL; + strv_free(n); + free(path); + free(nce); + + return r; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_service_type, service_type, ServiceType, "Failed to parse service type"); +DEFINE_CONFIG_PARSE_ENUM(config_parse_service_restart, service_restart, ServiceRestart, "Failed to parse service restart specifier"); + +int config_parse_socket_bindtodevice(const char* unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Socket *s = data; + char *n; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (rvalue[0] && !streq(rvalue, "*")) { + n = strdup(rvalue); + if (!n) + return log_oom(); + } else + n = NULL; + + free(s->bind_to_device); + s->bind_to_device = n; + + return 0; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_output, exec_output, ExecOutput, "Failed to parse output specifier"); +DEFINE_CONFIG_PARSE_ENUM(config_parse_input, exec_input, ExecInput, "Failed to parse input specifier"); + +int config_parse_exec_io_class(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int x; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + x = ioprio_class_from_string(rvalue); + if (x < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse IO scheduling class, ignoring: %s", rvalue); + return 0; + } + + c->ioprio = IOPRIO_PRIO_VALUE(x, IOPRIO_PRIO_DATA(c->ioprio)); + c->ioprio_set = true; + + return 0; +} + +int config_parse_exec_io_priority(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int i, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &i); + if (r < 0 || i < 0 || i >= IOPRIO_BE_NR) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse IO priority, ignoring: %s", rvalue); + return 0; + } + + c->ioprio = IOPRIO_PRIO_VALUE(IOPRIO_PRIO_CLASS(c->ioprio), i); + c->ioprio_set = true; + + return 0; +} + +int config_parse_exec_cpu_sched_policy(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + + ExecContext *c = data; + int x; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + x = sched_policy_from_string(rvalue); + if (x < 0) { + log_syntax(unit, LOG_ERR, filename, line, -x, + "Failed to parse CPU scheduling policy, ignoring: %s", rvalue); + return 0; + } + + c->cpu_sched_policy = x; + /* Moving to or from real-time policy? We need to adjust the priority */ + c->cpu_sched_priority = CLAMP(c->cpu_sched_priority, sched_get_priority_min(x), sched_get_priority_max(x)); + c->cpu_sched_set = true; + + return 0; +} + +int config_parse_exec_cpu_sched_prio(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int i, min, max, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &i); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse CPU scheduling policy, ignoring: %s", rvalue); + return 0; + } + + /* On Linux RR/FIFO range from 1 to 99 and OTHER/BATCH may only be 0 */ + min = sched_get_priority_min(c->cpu_sched_policy); + max = sched_get_priority_max(c->cpu_sched_policy); + + if (i < min || i > max) { + log_syntax(unit, LOG_ERR, filename, line, ERANGE, + "CPU scheduling priority is out of range, ignoring: %s", rvalue); + return 0; + } + + c->cpu_sched_priority = i; + c->cpu_sched_set = true; + + return 0; +} + +int config_parse_exec_cpu_affinity(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + char *w; + size_t l; + char *state; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* An empty assignment resets the CPU list */ + if (c->cpuset) + CPU_FREE(c->cpuset); + c->cpuset = NULL; + return 0; + } + + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + _cleanup_free_ char *t = NULL; + int r; + unsigned cpu; + + t = strndup(w, l); + if (!t) + return log_oom(); + + r = safe_atou(t, &cpu); + + if (!c->cpuset) { + c->cpuset = cpu_set_malloc(&c->cpuset_ncpus); + if (!c->cpuset) + return log_oom(); + } + + if (r < 0 || cpu >= c->cpuset_ncpus) { + log_syntax(unit, LOG_ERR, filename, line, ERANGE, + "Failed to parse CPU affinity '%s', ignoring: %s", t, rvalue); + return 0; + } + + CPU_SET_S(cpu, CPU_ALLOC_SIZE(c->cpuset_ncpus), c->cpuset); + } + + return 0; +} + +int config_parse_exec_capabilities(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + cap_t cap; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + cap = cap_from_text(rvalue); + if (!cap) { + log_syntax(unit, LOG_ERR, filename, line, errno, + "Failed to parse capabilities, ignoring: %s", rvalue); + return 0; + } + + if (c->capabilities) + cap_free(c->capabilities); + c->capabilities = cap; + + return 0; +} + +int config_parse_exec_secure_bits(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + char *w; + size_t l; + char *state; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* An empty assignment resets the field */ + c->secure_bits = 0; + return 0; + } + + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + if (first_word(w, "keep-caps")) + c->secure_bits |= 1<<SECURE_KEEP_CAPS; + else if (first_word(w, "keep-caps-locked")) + c->secure_bits |= 1<<SECURE_KEEP_CAPS_LOCKED; + else if (first_word(w, "no-setuid-fixup")) + c->secure_bits |= 1<<SECURE_NO_SETUID_FIXUP; + else if (first_word(w, "no-setuid-fixup-locked")) + c->secure_bits |= 1<<SECURE_NO_SETUID_FIXUP_LOCKED; + else if (first_word(w, "noroot")) + c->secure_bits |= 1<<SECURE_NOROOT; + else if (first_word(w, "noroot-locked")) + c->secure_bits |= 1<<SECURE_NOROOT_LOCKED; + else { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse secure bits, ignoring: %s", rvalue); + return 0; + } + } + + return 0; +} + +int config_parse_bounding_set(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + uint64_t *capability_bounding_set_drop = data; + char *w; + size_t l; + char *state; + bool invert = false; + uint64_t sum = 0; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (rvalue[0] == '~') { + invert = true; + rvalue++; + } + + /* Note that we store this inverted internally, since the + * kernel wants it like this. But we actually expose it + * non-inverted everywhere to have a fully normalized + * interface. */ + + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + _cleanup_free_ char *t = NULL; + int r; + cap_value_t cap; + + t = strndup(w, l); + if (!t) + return log_oom(); + + r = cap_from_name(t, &cap); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, errno, + "Failed to parse capability in bounding set, ignoring: %s", t); + continue; + } + + sum |= ((uint64_t) 1ULL) << (uint64_t) cap; + } + + if (invert) + *capability_bounding_set_drop |= sum; + else + *capability_bounding_set_drop |= ~sum; + + return 0; +} + +int config_parse_limit(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + struct rlimit **rl = data; + unsigned long long u; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + rl += ltype; + + if (streq(rvalue, "infinity")) + u = (unsigned long long) RLIM_INFINITY; + else { + int r; + + r = safe_atollu(rvalue, &u); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse resource value, ignoring: %s", rvalue); + return 0; + } + } + + if (!*rl) { + *rl = new(struct rlimit, 1); + if (!*rl) + return log_oom(); + } + + (*rl)->rlim_cur = (*rl)->rlim_max = (rlim_t) u; + return 0; +} + +int config_parse_unit_cgroup(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + char *w; + size_t l; + char *state; + + if (isempty(rvalue)) { + /* An empty assignment resets the list */ + cgroup_bonding_free_list(u->cgroup_bondings, false); + u->cgroup_bondings = NULL; + return 0; + } + + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + _cleanup_free_ char *t = NULL, *k = NULL, *ku = NULL; + int r; + + t = strndup(w, l); + if (!t) + return log_oom(); + + k = unit_full_printf(u, t); + if (!k) + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to resolve unit specifiers on %s. Ignoring.", + t); + + ku = cunescape(k ? k : t); + if (!ku) + return log_oom(); + + r = unit_add_cgroup_from_text(u, ku, true, NULL); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse cgroup value %s, ignoring: %s", + k, rvalue); + return 0; + } + } + + return 0; +} + +#ifdef HAVE_SYSV_COMPAT +int config_parse_sysv_priority(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + int *priority = data; + int i, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &i); + if (r < 0 || i < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse SysV start priority, ignoring: %s", rvalue); + return 0; + } + + *priority = (int) i; + return 0; +} +#endif + +int config_parse_fsck_passno(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + int *passno = data; + int i, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &i); + if (r || i < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse fsck pass number, ignoring: %s", rvalue); + return 0; + } + + *passno = (int) i; + return 0; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_kill_mode, kill_mode, KillMode, "Failed to parse kill mode"); + +int config_parse_kill_signal(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + int *sig = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(sig); + + r = signal_from_string_try_harder(rvalue); + if (r <= 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse kill signal, ignoring: %s", rvalue); + return 0; + } + + *sig = r; + return 0; +} + +int config_parse_exec_mount_flags(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + char *w; + size_t l; + char *state; + unsigned long flags = 0; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + FOREACH_WORD_SEPARATOR(w, l, rvalue, ", ", state) { + _cleanup_free_ char *t; + + t = strndup(w, l); + if (!t) + return log_oom(); + + if (streq(t, "shared")) + flags |= MS_SHARED; + else if (streq(t, "slave")) + flags |= MS_SLAVE; + else if (streq(w, "private")) + flags |= MS_PRIVATE; + else { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse mount flag %s, ignoring: %s", + t, rvalue); + return 0; + } + } + + c->mount_flags = flags; + return 0; +} + +int config_parse_timer(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Timer *t = data; + usec_t u = 0; + TimerValue *v; + TimerBase b; + CalendarSpec *c = NULL; + clockid_t id; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets list */ + timer_free_values(t); + return 0; + } + + b = timer_base_from_string(lvalue); + if (b < 0) { + log_syntax(unit, LOG_ERR, filename, line, -b, + "Failed to parse timer base, ignoring: %s", lvalue); + return 0; + } + + if (b == TIMER_CALENDAR) { + if (calendar_spec_from_string(rvalue, &c) < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse calendar specification, ignoring: %s", + rvalue); + return 0; + } + + id = CLOCK_REALTIME; + } else { + if (parse_sec(rvalue, &u) < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse timer value, ignoring: %s", + rvalue); + return 0; + } + + id = CLOCK_MONOTONIC; + } + + v = new0(TimerValue, 1); + if (!v) + return log_oom(); + + v->base = b; + v->clock_id = id; + v->value = u; + v->calendar_spec = c; + + LIST_PREPEND(TimerValue, value, t->values, v); + + return 0; +} + +int config_parse_trigger_unit( + const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *p = NULL; + Unit *u = data; + UnitType type; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (!set_isempty(u->dependencies[UNIT_TRIGGERS])) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Multiple units to trigger specified, ignoring: %s", rvalue); + return 0; + } + + p = unit_name_printf(u, rvalue); + if (!p) + return log_oom(); + + type = unit_name_to_type(p); + if (type < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Unit type not valid, ignoring: %s", rvalue); + return 0; + } + + if (type == u->type) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Trigger cannot be of same type, ignoring: %s", rvalue); + return 0; + } + + r = unit_add_two_dependencies_by_name(u, UNIT_BEFORE, UNIT_TRIGGERS, p, NULL, true); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to add trigger on %s, ignoring: %s", p, strerror(-r)); + return 0; + } + + return 0; +} + +int config_parse_path_spec(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Path *p = data; + PathSpec *s; + PathType b; + _cleanup_free_ char *k = NULL; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment clears list */ + path_free_specs(p); + return 0; + } + + b = path_type_from_string(lvalue); + if (b < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse path type, ignoring: %s", lvalue); + return 0; + } + + k = unit_full_printf(UNIT(p), rvalue); + if (!k) { + k = strdup(rvalue); + if (!k) + return log_oom(); + else + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to resolve unit specifiers on %s. Ignoring.", + rvalue); + } + + if (!path_is_absolute(k)) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Path is not absolute, ignoring: %s", k); + return 0; + } + + s = new0(PathSpec, 1); + if (!s) + return log_oom(); + + s->path = path_kill_slashes(k); + k = NULL; + s->type = b; + s->inotify_fd = -1; + + LIST_PREPEND(PathSpec, spec, p->specs, s); + + return 0; +} + +int config_parse_socket_service(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Socket *s = data; + int r; + DBusError error; + Unit *x; + _cleanup_free_ char *p = NULL; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + dbus_error_init(&error); + + p = unit_name_printf(UNIT(s), rvalue); + if (!p) + return log_oom(); + + if (!endswith(p, ".service")) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Unit must be of type service, ignoring: %s", rvalue); + return 0; + } + + r = manager_load_unit(UNIT(s)->manager, p, NULL, &error, &x); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to load unit %s, ignoring: %s", + rvalue, bus_error(&error, r)); + dbus_error_free(&error); + return 0; + } + + unit_ref_set(&s->service, x); + + return 0; +} + +int config_parse_service_sockets(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Service *s = data; + int r; + char *state, *w; + size_t l; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + _cleanup_free_ char *t = NULL, *k = NULL; + + t = strndup(w, l); + if (!t) + return log_oom(); + + k = unit_name_printf(UNIT(s), t); + if (!k) + return log_oom(); + + if (!endswith(k, ".socket")) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Unit must be of type socket, ignoring: %s", k); + continue; + } + + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_WANTS, UNIT_AFTER, k, NULL, true); + if (r < 0) + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to add dependency on %s, ignoring: %s", + k, strerror(-r)); + + r = unit_add_dependency_by_name(UNIT(s), UNIT_TRIGGERED_BY, k, NULL, true); + if (r < 0) + return r; + } + + return 0; +} + +int config_parse_service_timeout(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Service *s = userdata; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(s); + + r = config_parse_sec(unit, filename, line, section, lvalue, ltype, + rvalue, data, userdata); + if (r < 0) + return r; + + if (streq(lvalue, "TimeoutSec")) { + s->start_timeout_defined = true; + s->timeout_stop_usec = s->timeout_start_usec; + } else if (streq(lvalue, "TimeoutStartSec")) + s->start_timeout_defined = true; + + return 0; +} + +int config_parse_unit_env_file(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + char ***env = data; + Unit *u = userdata; + _cleanup_free_ char *s = NULL; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment frees the list */ + strv_free(*env); + *env = NULL; + return 0; + } + + s = unit_full_printf(u, rvalue); + if (!s) + return log_oom(); + + if (!path_is_absolute(s[0] == '-' ? s + 1 : s)) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Path '%s' is not absolute, ignoring.", s); + return 0; + } + + r = strv_extend(env, s); + if (r < 0) + return log_oom(); + + return 0; +} + +int config_parse_environ(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + char*** env = data, *w, *state; + size_t l; + _cleanup_free_ char *k = NULL; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + strv_free(*env); + *env = NULL; + return 0; + } + + k = unit_full_printf(u, rvalue); + if (!k) + return log_oom(); + + FOREACH_WORD_QUOTED(w, l, k, state) { + _cleanup_free_ char *n; + char **x; + + n = cunescape_length(w, l); + if (!n) + return log_oom(); + + if (!env_assignment_is_valid(n)) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Invalid environment assignment, ignoring: %s", rvalue); + continue; + } + + x = strv_env_set(*env, n); + if (!x) + return log_oom(); + + strv_free(*env); + *env = x; + } + + return 0; +} + +int config_parse_ip_tos(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + int *ip_tos = data, x; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + x = ip_tos_from_string(rvalue); + if (x < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse IP TOS value, ignoring: %s", rvalue); + return 0; + } + + *ip_tos = x; + return 0; +} + +int config_parse_unit_condition_path(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ConditionType cond = ltype; + Unit *u = data; + bool trigger, negate; + Condition *c; + _cleanup_free_ char *p = NULL; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + condition_free_list(u->conditions); + u->conditions = NULL; + return 0; + } + + trigger = rvalue[0] == '|'; + if (trigger) + rvalue++; + + negate = rvalue[0] == '!'; + if (negate) + rvalue++; + + p = unit_full_printf(u, rvalue); + if (!p) + return log_oom(); + + if (!path_is_absolute(p)) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Path in condition not absolute, ignoring: %s", p); + return 0; + } + + c = condition_new(cond, p, trigger, negate); + if (!c) + return log_oom(); + + LIST_PREPEND(Condition, conditions, u->conditions, c); + return 0; +} + +int config_parse_unit_condition_string(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ConditionType cond = ltype; + Unit *u = data; + bool trigger, negate; + Condition *c; + _cleanup_free_ char *s = NULL; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + condition_free_list(u->conditions); + u->conditions = NULL; + return 0; + } + + trigger = rvalue[0] == '|'; + if (trigger) + rvalue++; + + negate = rvalue[0] == '!'; + if (negate) + rvalue++; + + s = unit_full_printf(u, rvalue); + if (!s) + return log_oom(); + + c = condition_new(cond, s, trigger, negate); + if (!c) + return log_oom(); + + LIST_PREPEND(Condition, conditions, u->conditions, c); + return 0; +} + +int config_parse_unit_condition_null(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = data; + Condition *c; + bool trigger, negate; + int b; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + condition_free_list(u->conditions); + u->conditions = NULL; + return 0; + } + + trigger = rvalue[0] == '|'; + if (trigger) + rvalue++; + + negate = rvalue[0] == '!'; + if (negate) + rvalue++; + + b = parse_boolean(rvalue); + if (b < 0) { + log_syntax(unit, LOG_ERR, filename, line, -b, + "Failed to parse boolean value in condition, ignoring: %s", + rvalue); + return 0; + } + + if (!b) + negate = !negate; + + c = condition_new(CONDITION_NULL, NULL, trigger, negate); + if (!c) + return log_oom(); + + LIST_PREPEND(Condition, conditions, u->conditions, c); + return 0; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_notify_access, notify_access, NotifyAccess, "Failed to parse notify access specifier"); +DEFINE_CONFIG_PARSE_ENUM(config_parse_start_limit_action, start_limit_action, StartLimitAction, "Failed to parse start limit action specifier"); + +int config_parse_unit_cgroup_attr(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = data; + size_t a, b; + _cleanup_free_ char *n = NULL, *v = NULL; + const CGroupSemantics *s; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment clears the list */ + cgroup_attribute_free_list(u->cgroup_attributes); + u->cgroup_attributes = NULL; + return 0; + } + + a = strcspn(rvalue, WHITESPACE); + b = strspn(rvalue + a, WHITESPACE); + if (a <= 0 || b <= 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse cgroup attribute value, ignoring: %s", + rvalue); + return 0; + } + + n = strndup(rvalue, a); + if (!n) + return log_oom(); + + r = cgroup_semantics_find(NULL, n, rvalue + a + b, &v, &s); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse cgroup attribute value, ignoring: %s", + rvalue); + return 0; + } + + r = unit_add_cgroup_attribute(u, s, NULL, n, v ? v : rvalue + a + b, NULL); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to add cgroup attribute value, ignoring: %s", rvalue); + return 0; + } + + return 0; +} + +int config_parse_unit_cgroup_attr_pretty(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = data; + _cleanup_free_ char *v = NULL; + const CGroupSemantics *s; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = cgroup_semantics_find(NULL, lvalue, rvalue, &v, &s); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse cgroup attribute value, ignoring: %s", + rvalue); + return 0; + } else if (r == 0) { + log_syntax(unit, LOG_ERR, filename, line, ENOTSUP, + "Unknown or unsupported cgroup attribute %s, ignoring: %s", + lvalue, rvalue); + return 0; + } + + r = unit_add_cgroup_attribute(u, s, NULL, NULL, v, NULL); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to add cgroup attribute value, ignoring: %s", rvalue); + return 0; + } + + return 0; +} + +int config_parse_unit_requires_mounts_for(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + int r; + bool empty_before; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + empty_before = !u->requires_mounts_for; + + r = config_parse_path_strv(unit, filename, line, section, lvalue, ltype, + rvalue, data, userdata); + + /* Make it easy to find units with requires_mounts set */ + if (empty_before && u->requires_mounts_for) + LIST_PREPEND(Unit, has_requires_mounts_for, u->manager->has_requires_mounts_for, u); + + return r; +} + +int config_parse_documentation(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + int r; + char **a, **b; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + strv_free(u->documentation); + u->documentation = NULL; + return 0; + } + + r = config_parse_unit_strv_printf(unit, filename, line, section, lvalue, ltype, + rvalue, data, userdata); + if (r < 0) + return r; + + for (a = b = u->documentation; a && *a; a++) { + + if (is_valid_documentation_url(*a)) + *(b++) = *a; + else { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Invalid URL, ignoring: %s", *a); + free(*a); + } + } + *b = NULL; + + return r; +} + +static void syscall_set(uint32_t *p, int nr) { + nr = SYSCALL_TO_INDEX(nr); + p[nr >> 4] |= 1 << (nr & 31); +} + +static void syscall_unset(uint32_t *p, int nr) { + nr = SYSCALL_TO_INDEX(nr); + p[nr >> 4] &= ~(1 << (nr & 31)); +} + +int config_parse_syscall_filter(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + Unit *u = userdata; + bool invert = false; + char *w; + size_t l; + char *state; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + free(c->syscall_filter); + c->syscall_filter = NULL; + return 0; + } + + if (rvalue[0] == '~') { + invert = true; + rvalue++; + } + + if (!c->syscall_filter) { + size_t n; + + n = (syscall_max() + 31) >> 4; + c->syscall_filter = new(uint32_t, n); + if (!c->syscall_filter) + return log_oom(); + + memset(c->syscall_filter, invert ? 0xFF : 0, n * sizeof(uint32_t)); + + /* Add these by default */ + syscall_set(c->syscall_filter, __NR_execve); + syscall_set(c->syscall_filter, __NR_rt_sigreturn); +#ifdef __NR_sigreturn + syscall_set(c->syscall_filter, __NR_sigreturn); +#endif + syscall_set(c->syscall_filter, __NR_exit_group); + syscall_set(c->syscall_filter, __NR_exit); + } + + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + int id; + _cleanup_free_ char *t = NULL; + + t = strndup(w, l); + if (!t) + return log_oom(); + + id = syscall_from_name(t); + if (id < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Failed to parse syscall, ignoring: %s", t); + continue; + } + + if (invert) + syscall_unset(c->syscall_filter, id); + else + syscall_set(c->syscall_filter, id); + } + + c->no_new_privileges = true; + + return 0; +} + +#define FOLLOW_MAX 8 + +static int open_follow(char **filename, FILE **_f, Set *names, char **_final) { + unsigned c = 0; + int fd, r; + FILE *f; + char *id = NULL; + + assert(filename); + assert(*filename); + assert(_f); + assert(names); + + /* This will update the filename pointer if the loaded file is + * reached by a symlink. The old string will be freed. */ + + for (;;) { + char *target, *name; + + if (c++ >= FOLLOW_MAX) + return -ELOOP; + + path_kill_slashes(*filename); + + /* Add the file name we are currently looking at to + * the names of this unit, but only if it is a valid + * unit name. */ + name = path_get_file_name(*filename); + + if (unit_name_is_valid(name, true)) { + + id = set_get(names, name); + if (!id) { + id = strdup(name); + if (!id) + return -ENOMEM; + + r = set_consume(names, id); + if (r < 0) + return r; + } + } + + /* Try to open the file name, but don't if its a symlink */ + fd = open(*filename, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); + if (fd >= 0) + break; + + if (errno != ELOOP) + return -errno; + + /* Hmm, so this is a symlink. Let's read the name, and follow it manually */ + r = readlink_and_make_absolute(*filename, &target); + if (r < 0) + return r; + + free(*filename); + *filename = target; + } + + f = fdopen(fd, "re"); + if (!f) { + r = -errno; + close_nointr_nofail(fd); + return r; + } + + *_f = f; + *_final = id; + return 0; +} + +static int merge_by_names(Unit **u, Set *names, const char *id) { + char *k; + int r; + + assert(u); + assert(*u); + assert(names); + + /* Let's try to add in all symlink names we found */ + while ((k = set_steal_first(names))) { + + /* First try to merge in the other name into our + * unit */ + r = unit_merge_by_name(*u, k); + if (r < 0) { + Unit *other; + + /* Hmm, we couldn't merge the other unit into + * ours? Then let's try it the other way + * round */ + + other = manager_get_unit((*u)->manager, k); + free(k); + + if (other) { + r = unit_merge(other, *u); + if (r >= 0) { + *u = other; + return merge_by_names(u, names, NULL); + } + } + + return r; + } + + if (id == k) + unit_choose_id(*u, id); + + free(k); + } + + return 0; +} + +static int load_from_path(Unit *u, const char *path) { + int r; + Set *symlink_names; + FILE *f = NULL; + char *filename = NULL, *id = NULL; + Unit *merged; + struct stat st; + + assert(u); + assert(path); + + symlink_names = set_new(string_hash_func, string_compare_func); + if (!symlink_names) + return -ENOMEM; + + if (path_is_absolute(path)) { + + filename = strdup(path); + if (!filename) { + r = -ENOMEM; + goto finish; + } + + r = open_follow(&filename, &f, symlink_names, &id); + if (r < 0) { + free(filename); + filename = NULL; + + if (r != -ENOENT) + goto finish; + } + + } else { + char **p; + + STRV_FOREACH(p, u->manager->lookup_paths.unit_path) { + + /* Instead of opening the path right away, we manually + * follow all symlinks and add their name to our unit + * name set while doing so */ + filename = path_make_absolute(path, *p); + if (!filename) { + r = -ENOMEM; + goto finish; + } + + if (u->manager->unit_path_cache && + !set_get(u->manager->unit_path_cache, filename)) + r = -ENOENT; + else + r = open_follow(&filename, &f, symlink_names, &id); + + if (r < 0) { + free(filename); + filename = NULL; + + if (r != -ENOENT) + goto finish; + + /* Empty the symlink names for the next run */ + set_clear_free(symlink_names); + continue; + } + + break; + } + } + + if (!filename) { + /* Hmm, no suitable file found? */ + r = 0; + goto finish; + } + + merged = u; + r = merge_by_names(&merged, symlink_names, id); + if (r < 0) + goto finish; + + if (merged != u) { + u->load_state = UNIT_MERGED; + r = 0; + goto finish; + } + + if (fstat(fileno(f), &st) < 0) { + r = -errno; + goto finish; + } + + if (null_or_empty(&st)) + u->load_state = UNIT_MASKED; + else { + /* Now, parse the file contents */ + r = config_parse(u->id, filename, f, UNIT_VTABLE(u)->sections, + config_item_perf_lookup, + (void*) load_fragment_gperf_lookup, false, true, u); + if (r < 0) + goto finish; + + u->load_state = UNIT_LOADED; + } + + free(u->fragment_path); + u->fragment_path = filename; + filename = NULL; + + u->fragment_mtime = timespec_load(&st.st_mtim); + + if (u->source_path) { + if (stat(u->source_path, &st) >= 0) + u->source_mtime = timespec_load(&st.st_mtim); + else + u->source_mtime = 0; + } + + r = 0; + +finish: + set_free_free(symlink_names); + free(filename); + + if (f) + fclose(f); + + return r; +} + +int unit_load_fragment(Unit *u) { + int r; + Iterator i; + const char *t; + + assert(u); + assert(u->load_state == UNIT_STUB); + assert(u->id); + + /* First, try to find the unit under its id. We always look + * for unit files in the default directories, to make it easy + * to override things by placing things in /etc/systemd/system */ + r = load_from_path(u, u->id); + if (r < 0) + return r; + + /* Try to find an alias we can load this with */ + if (u->load_state == UNIT_STUB) + SET_FOREACH(t, u->names, i) { + + if (t == u->id) + continue; + + r = load_from_path(u, t); + if (r < 0) + return r; + + if (u->load_state != UNIT_STUB) + break; + } + + /* And now, try looking for it under the suggested (originally linked) path */ + if (u->load_state == UNIT_STUB && u->fragment_path) { + + r = load_from_path(u, u->fragment_path); + if (r < 0) + return r; + + if (u->load_state == UNIT_STUB) { + /* Hmm, this didn't work? Then let's get rid + * of the fragment path stored for us, so that + * we don't point to an invalid location. */ + free(u->fragment_path); + u->fragment_path = NULL; + } + } + + /* Look for a template */ + if (u->load_state == UNIT_STUB && u->instance) { + char *k; + + k = unit_name_template(u->id); + if (!k) + return -ENOMEM; + + r = load_from_path(u, k); + free(k); + + if (r < 0) + return r; + + if (u->load_state == UNIT_STUB) + SET_FOREACH(t, u->names, i) { + + if (t == u->id) + continue; + + k = unit_name_template(t); + if (!k) + return -ENOMEM; + + r = load_from_path(u, k); + free(k); + + if (r < 0) + return r; + + if (u->load_state != UNIT_STUB) + break; + } + } + + return 0; +} + +void unit_dump_config_items(FILE *f) { + static const struct { + const ConfigParserCallback callback; + const char *rvalue; + } table[] = { + { config_parse_int, "INTEGER" }, + { config_parse_unsigned, "UNSIGNED" }, + { config_parse_bytes_size, "SIZE" }, + { config_parse_bool, "BOOLEAN" }, + { config_parse_string, "STRING" }, + { config_parse_path, "PATH" }, + { config_parse_unit_path_printf, "PATH" }, + { config_parse_strv, "STRING [...]" }, + { config_parse_exec_nice, "NICE" }, + { config_parse_exec_oom_score_adjust, "OOMSCOREADJUST" }, + { config_parse_exec_io_class, "IOCLASS" }, + { config_parse_exec_io_priority, "IOPRIORITY" }, + { config_parse_exec_cpu_sched_policy, "CPUSCHEDPOLICY" }, + { config_parse_exec_cpu_sched_prio, "CPUSCHEDPRIO" }, + { config_parse_exec_cpu_affinity, "CPUAFFINITY" }, + { config_parse_mode, "MODE" }, + { config_parse_unit_env_file, "FILE" }, + { config_parse_output, "OUTPUT" }, + { config_parse_input, "INPUT" }, + { config_parse_facility, "FACILITY" }, + { config_parse_level, "LEVEL" }, + { config_parse_exec_capabilities, "CAPABILITIES" }, + { config_parse_exec_secure_bits, "SECUREBITS" }, + { config_parse_bounding_set, "BOUNDINGSET" }, + { config_parse_limit, "LIMIT" }, + { config_parse_unit_cgroup, "CGROUP [...]" }, + { config_parse_unit_deps, "UNIT [...]" }, + { config_parse_exec, "PATH [ARGUMENT [...]]" }, + { config_parse_service_type, "SERVICETYPE" }, + { config_parse_service_restart, "SERVICERESTART" }, +#ifdef HAVE_SYSV_COMPAT + { config_parse_sysv_priority, "SYSVPRIORITY" }, +#else + { config_parse_warn_compat, "NOTSUPPORTED" }, +#endif + { config_parse_kill_mode, "KILLMODE" }, + { config_parse_kill_signal, "SIGNAL" }, + { config_parse_socket_listen, "SOCKET [...]" }, + { config_parse_socket_bind, "SOCKETBIND" }, + { config_parse_socket_bindtodevice, "NETWORKINTERFACE" }, + { config_parse_sec, "SECONDS" }, + { config_parse_nsec, "NANOSECONDS" }, + { config_parse_path_strv, "PATH [...]" }, + { config_parse_unit_requires_mounts_for, "PATH [...]" }, + { config_parse_exec_mount_flags, "MOUNTFLAG [...]" }, + { config_parse_unit_string_printf, "STRING" }, + { config_parse_trigger_unit, "UNIT" }, + { config_parse_timer, "TIMER" }, + { config_parse_path_spec, "PATH" }, + { config_parse_notify_access, "ACCESS" }, + { config_parse_ip_tos, "TOS" }, + { config_parse_unit_condition_path, "CONDITION" }, + { config_parse_unit_condition_string, "CONDITION" }, + { config_parse_unit_condition_null, "CONDITION" }, + }; + + const char *prev = NULL; + const char *i; + + assert(f); + + NULSTR_FOREACH(i, load_fragment_gperf_nulstr) { + const char *rvalue = "OTHER", *lvalue; + unsigned j; + size_t prefix_len; + const char *dot; + const ConfigPerfItem *p; + + assert_se(p = load_fragment_gperf_lookup(i, strlen(i))); + + dot = strchr(i, '.'); + lvalue = dot ? dot + 1 : i; + prefix_len = dot-i; + + if (dot) + if (!prev || !strneq(prev, i, prefix_len+1)) { + if (prev) + fputc('\n', f); + + fprintf(f, "[%.*s]\n", (int) prefix_len, i); + } + + for (j = 0; j < ELEMENTSOF(table); j++) + if (p->parse == table[j].callback) { + rvalue = table[j].rvalue; + break; + } + + fprintf(f, "%s=%s\n", lvalue, rvalue); + prev = i; + } +} diff --git a/src/core/load-fragment.h b/src/core/load-fragment.h new file mode 100644 index 0000000000..ff7f22a6f0 --- /dev/null +++ b/src/core/load-fragment.h @@ -0,0 +1,84 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "unit.h" + +/* Read service data from .desktop file style configuration fragments */ + +int unit_load_fragment(Unit *u); + +void unit_dump_config_items(FILE *f); + +int config_parse_warn_compat(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_deps(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_string_printf(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_strv_printf(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_path_printf(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_documentation(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_socket_listen(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_socket_bind(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_nice(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_oom_score_adjust(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_service_timeout(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_service_type(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_service_restart(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_socket_bindtodevice(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_output(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_input(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_io_class(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_io_priority(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_cpu_sched_policy(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_cpu_sched_prio(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_cpu_affinity(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_capabilities(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_secure_bits(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_bounding_set(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_limit(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_cgroup(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_sysv_priority(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_fsck_passno(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_kill_signal(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_mount_flags(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_timer(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_trigger_unit(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_path_spec(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_socket_service(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_service_sockets(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_env_file(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_ip_tos(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_condition_path(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_condition_string(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_condition_null(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_kill_mode(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_notify_access(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_start_limit_action(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_cgroup_attr(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_cgroup_attr_pretty(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_requires_mounts_for(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_syscall_filter(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_environ(const char *unit, const char *filename, unsigned line, const char *section, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); + +/* gperf prototypes */ +const struct ConfigPerfItem* load_fragment_gperf_lookup(const char *key, unsigned length); +extern const char load_fragment_gperf_nulstr[]; diff --git a/src/core/locale-setup.c b/src/core/locale-setup.c new file mode 100644 index 0000000000..d7113b9795 --- /dev/null +++ b/src/core/locale-setup.c @@ -0,0 +1,145 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <string.h> +#include <stdlib.h> +#include <errno.h> + +#include "locale-setup.h" +#include "util.h" +#include "macro.h" +#include "virt.h" +#include "fileio.h" + +enum { + /* We don't list LC_ALL here on purpose. People should be + * using LANG instead. */ + + VARIABLE_LANG, + VARIABLE_LANGUAGE, + VARIABLE_LC_CTYPE, + VARIABLE_LC_NUMERIC, + VARIABLE_LC_TIME, + VARIABLE_LC_COLLATE, + VARIABLE_LC_MONETARY, + VARIABLE_LC_MESSAGES, + VARIABLE_LC_PAPER, + VARIABLE_LC_NAME, + VARIABLE_LC_ADDRESS, + VARIABLE_LC_TELEPHONE, + VARIABLE_LC_MEASUREMENT, + VARIABLE_LC_IDENTIFICATION, + _VARIABLE_MAX +}; + +static const char * const variable_names[_VARIABLE_MAX] = { + [VARIABLE_LANG] = "LANG", + [VARIABLE_LANGUAGE] = "LANGUAGE", + [VARIABLE_LC_CTYPE] = "LC_CTYPE", + [VARIABLE_LC_NUMERIC] = "LC_NUMERIC", + [VARIABLE_LC_TIME] = "LC_TIME", + [VARIABLE_LC_COLLATE] = "LC_COLLATE", + [VARIABLE_LC_MONETARY] = "LC_MONETARY", + [VARIABLE_LC_MESSAGES] = "LC_MESSAGES", + [VARIABLE_LC_PAPER] = "LC_PAPER", + [VARIABLE_LC_NAME] = "LC_NAME", + [VARIABLE_LC_ADDRESS] = "LC_ADDRESS", + [VARIABLE_LC_TELEPHONE] = "LC_TELEPHONE", + [VARIABLE_LC_MEASUREMENT] = "LC_MEASUREMENT", + [VARIABLE_LC_IDENTIFICATION] = "LC_IDENTIFICATION" +}; + +int locale_setup(void) { + char *variables[_VARIABLE_MAX] = {}; + int r = 0, i; + + if (detect_container(NULL) <= 0) { + r = parse_env_file("/proc/cmdline", WHITESPACE, + "locale.LANG", &variables[VARIABLE_LANG], + "locale.LANGUAGE", &variables[VARIABLE_LANGUAGE], + "locale.LC_CTYPE", &variables[VARIABLE_LC_CTYPE], + "locale.LC_NUMERIC", &variables[VARIABLE_LC_NUMERIC], + "locale.LC_TIME", &variables[VARIABLE_LC_TIME], + "locale.LC_COLLATE", &variables[VARIABLE_LC_COLLATE], + "locale.LC_MONETARY", &variables[VARIABLE_LC_MONETARY], + "locale.LC_MESSAGES", &variables[VARIABLE_LC_MESSAGES], + "locale.LC_PAPER", &variables[VARIABLE_LC_PAPER], + "locale.LC_NAME", &variables[VARIABLE_LC_NAME], + "locale.LC_ADDRESS", &variables[VARIABLE_LC_ADDRESS], + "locale.LC_TELEPHONE", &variables[VARIABLE_LC_TELEPHONE], + "locale.LC_MEASUREMENT", &variables[VARIABLE_LC_MEASUREMENT], + "locale.LC_IDENTIFICATION", &variables[VARIABLE_LC_IDENTIFICATION], + NULL); + + if (r < 0 && r != -ENOENT) + log_warning("Failed to read /proc/cmdline: %s", strerror(-r)); + } + + /* Hmm, nothing set on the kernel cmd line? Then let's + * try /etc/locale.conf */ + if (r <= 0) { + r = parse_env_file("/etc/locale.conf", NEWLINE, + "LANG", &variables[VARIABLE_LANG], + "LANGUAGE", &variables[VARIABLE_LANGUAGE], + "LC_CTYPE", &variables[VARIABLE_LC_CTYPE], + "LC_NUMERIC", &variables[VARIABLE_LC_NUMERIC], + "LC_TIME", &variables[VARIABLE_LC_TIME], + "LC_COLLATE", &variables[VARIABLE_LC_COLLATE], + "LC_MONETARY", &variables[VARIABLE_LC_MONETARY], + "LC_MESSAGES", &variables[VARIABLE_LC_MESSAGES], + "LC_PAPER", &variables[VARIABLE_LC_PAPER], + "LC_NAME", &variables[VARIABLE_LC_NAME], + "LC_ADDRESS", &variables[VARIABLE_LC_ADDRESS], + "LC_TELEPHONE", &variables[VARIABLE_LC_TELEPHONE], + "LC_MEASUREMENT", &variables[VARIABLE_LC_MEASUREMENT], + "LC_IDENTIFICATION", &variables[VARIABLE_LC_IDENTIFICATION], + NULL); + + if (r < 0 && r != -ENOENT) + log_warning("Failed to read /etc/locale.conf: %s", strerror(-r)); + } + + if (!variables[VARIABLE_LANG]) { + variables[VARIABLE_LANG] = strdup("C"); + if (!variables[VARIABLE_LANG]) { + r = -ENOMEM; + goto finish; + } + } + + for (i = 0; i < _VARIABLE_MAX; i++) { + if (variables[i]) { + if (setenv(variable_names[i], variables[i], 1) < 0) { + r = -errno; + goto finish; + } + } else + unsetenv(variable_names[i]); + } + + r = 0; + +finish: + for (i = 0; i < _VARIABLE_MAX; i++) + free(variables[i]); + + return r; +} diff --git a/src/core/locale-setup.h b/src/core/locale-setup.h new file mode 100644 index 0000000000..5a0f2f7888 --- /dev/null +++ b/src/core/locale-setup.h @@ -0,0 +1,24 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int locale_setup(void); diff --git a/src/core/loopback-setup.c b/src/core/loopback-setup.c new file mode 100644 index 0000000000..aff24fa642 --- /dev/null +++ b/src/core/loopback-setup.c @@ -0,0 +1,308 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <sys/socket.h> +#include <net/if.h> +#include <asm/types.h> +#include <netinet/in.h> +#include <string.h> +#include <stdlib.h> +#include <unistd.h> +#include <linux/netlink.h> +#include <linux/rtnetlink.h> + +#include "util.h" +#include "macro.h" +#include "loopback-setup.h" +#include "socket-util.h" + +#define NLMSG_TAIL(nmsg) \ + ((struct rtattr *) (((uint8_t*) (nmsg)) + NLMSG_ALIGN((nmsg)->nlmsg_len))) + +static int add_rtattr(struct nlmsghdr *n, size_t max_length, int type, const void *data, size_t data_length) { + size_t length; + struct rtattr *rta; + + length = RTA_LENGTH(data_length); + + if (NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(length) > max_length) + return -E2BIG; + + rta = NLMSG_TAIL(n); + rta->rta_type = type; + rta->rta_len = length; + memcpy(RTA_DATA(rta), data, data_length); + n->nlmsg_len = NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(length); + + return 0; +} + +static ssize_t sendto_loop(int fd, const void *buf, size_t buf_len, int flags, const struct sockaddr *sa, socklen_t sa_len) { + + for (;;) { + ssize_t l; + + l = sendto(fd, buf, buf_len, flags, sa, sa_len); + if (l >= 0) + return l; + + if (errno != EINTR) + return -errno; + } +} + +static ssize_t recvfrom_loop(int fd, void *buf, size_t buf_len, int flags, struct sockaddr *sa, socklen_t *sa_len) { + + for (;;) { + ssize_t l; + + l = recvfrom(fd, buf, buf_len, flags, sa, sa_len); + if (l >= 0) + return l; + + if (errno != EINTR) + return -errno; + } +} + +static int add_adresses(int fd, int if_loopback, unsigned *requests) { + union { + struct sockaddr sa; + struct sockaddr_nl nl; + } sa = { + .nl.nl_family = AF_NETLINK, + }; + + union { + struct nlmsghdr header; + uint8_t buf[NLMSG_ALIGN(sizeof(struct nlmsghdr)) + + NLMSG_ALIGN(sizeof(struct ifaddrmsg)) + + RTA_LENGTH(sizeof(struct in6_addr))]; + } request = { + .header.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg)), + .header.nlmsg_type = RTM_NEWADDR, + .header.nlmsg_flags = NLM_F_REQUEST|NLM_F_CREATE|NLM_F_ACK, + .header.nlmsg_seq = *requests + 1, + }; + + struct ifaddrmsg *ifaddrmsg; + uint32_t ipv4_address = htonl(INADDR_LOOPBACK); + int r; + + ifaddrmsg = NLMSG_DATA(&request.header); + ifaddrmsg->ifa_family = AF_INET; + ifaddrmsg->ifa_prefixlen = 8; + ifaddrmsg->ifa_flags = IFA_F_PERMANENT; + ifaddrmsg->ifa_scope = RT_SCOPE_HOST; + ifaddrmsg->ifa_index = if_loopback; + + r = add_rtattr(&request.header, sizeof(request), IFA_LOCAL, + &ipv4_address, sizeof(ipv4_address)); + if (r < 0) + return r; + + if (sendto_loop(fd, &request, request.header.nlmsg_len, 0, &sa.sa, sizeof(sa)) < 0) + return -errno; + (*requests)++; + + if (!socket_ipv6_is_supported()) + return 0; + + request.header.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg)); + request.header.nlmsg_seq = *requests + 1; + + ifaddrmsg->ifa_family = AF_INET6; + ifaddrmsg->ifa_prefixlen = 128; + + r = add_rtattr(&request.header, sizeof(request), IFA_LOCAL, + &in6addr_loopback, sizeof(in6addr_loopback)); + if (r < 0) + return r; + + if (sendto_loop(fd, &request, request.header.nlmsg_len, 0, &sa.sa, sizeof(sa)) < 0) + return -errno; + (*requests)++; + + return 0; +} + +static int start_interface(int fd, int if_loopback, unsigned *requests) { + union { + struct sockaddr sa; + struct sockaddr_nl nl; + } sa = { + .nl.nl_family = AF_NETLINK, + }; + + union { + struct nlmsghdr header; + uint8_t buf[NLMSG_ALIGN(sizeof(struct nlmsghdr)) + + NLMSG_ALIGN(sizeof(struct ifinfomsg))]; + } request = { + .header.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg)), + .header.nlmsg_type = RTM_NEWLINK, + .header.nlmsg_flags = NLM_F_REQUEST|NLM_F_ACK, + .header.nlmsg_seq = *requests + 1, + }; + + struct ifinfomsg *ifinfomsg; + + ifinfomsg = NLMSG_DATA(&request.header); + ifinfomsg->ifi_family = AF_UNSPEC; + ifinfomsg->ifi_index = if_loopback; + ifinfomsg->ifi_flags = IFF_UP; + ifinfomsg->ifi_change = IFF_UP; + + if (sendto_loop(fd, &request, request.header.nlmsg_len, 0, &sa.sa, sizeof(sa)) < 0) + return -errno; + + (*requests)++; + + return 0; +} + +static int read_response(int fd, unsigned requests_max) { + union { + struct sockaddr sa; + struct sockaddr_nl nl; + } sa; + union { + struct nlmsghdr header; + uint8_t buf[NLMSG_ALIGN(sizeof(struct nlmsghdr)) + + NLMSG_ALIGN(sizeof(struct nlmsgerr))]; + } response; + + ssize_t l; + socklen_t sa_len = sizeof(sa); + struct nlmsgerr *nlmsgerr; + + l = recvfrom_loop(fd, &response, sizeof(response), 0, &sa.sa, &sa_len); + if (l < 0) + return -errno; + + if (sa_len != sizeof(sa.nl) || + sa.nl.nl_family != AF_NETLINK) + return -EIO; + + if (sa.nl.nl_pid != 0) + return 0; + + if ((size_t) l < sizeof(struct nlmsghdr)) + return -EIO; + + if (response.header.nlmsg_type != NLMSG_ERROR || + (pid_t) response.header.nlmsg_pid != getpid() || + response.header.nlmsg_seq >= requests_max) + return 0; + + if ((size_t) l < NLMSG_LENGTH(sizeof(struct nlmsgerr)) || + response.header.nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) + return -EIO; + + nlmsgerr = NLMSG_DATA(&response.header); + + if (nlmsgerr->error < 0 && nlmsgerr->error != -EEXIST) + return nlmsgerr->error; + + return response.header.nlmsg_seq; +} + +static int check_loopback(void) { + int r; + _cleanup_close_ int fd; + union { + struct sockaddr sa; + struct sockaddr_in in; + } sa = { + .in.sin_family = AF_INET, + .in.sin_addr.s_addr = INADDR_LOOPBACK, + }; + + /* If we failed to set up the loop back device, check whether + * it might already be set up */ + + fd = socket(AF_INET, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0); + if (fd < 0) + return -errno; + + if (bind(fd, &sa.sa, sizeof(sa.in)) >= 0) + r = 1; + else + r = errno == EADDRNOTAVAIL ? 0 : -errno; + + return r; +} + +int loopback_setup(void) { + int r, if_loopback; + union { + struct sockaddr sa; + struct sockaddr_nl nl; + } sa = { + .nl.nl_family = AF_NETLINK, + }; + unsigned requests = 0, i; + _cleanup_close_ int fd = -1; + bool eperm = false; + + errno = 0; + if_loopback = (int) if_nametoindex("lo"); + if (if_loopback <= 0) + return errno ? -errno : -ENODEV; + + fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE); + if (fd < 0) + return -errno; + + if (bind(fd, &sa.sa, sizeof(sa)) < 0) { + r = -errno; + goto error; + } + + r = add_adresses(fd, if_loopback, &requests); + if (r < 0) + goto error; + + r = start_interface(fd, if_loopback, &requests); + if (r < 0) + goto error; + + for (i = 0; i < requests; i++) { + r = read_response(fd, requests); + + if (r == -EPERM) + eperm = true; + else if (r < 0) + goto error; + } + + if (eperm && check_loopback() < 0) { + r = -EPERM; + goto error; + } + + return 0; + +error: + log_warning("Failed to configure loopback device: %s", strerror(-r)); + return r; +} diff --git a/src/core/loopback-setup.h b/src/core/loopback-setup.h new file mode 100644 index 0000000000..dd83cf13a7 --- /dev/null +++ b/src/core/loopback-setup.h @@ -0,0 +1,24 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int loopback_setup(void); diff --git a/src/core/machine-id-setup.c b/src/core/machine-id-setup.c new file mode 100644 index 0000000000..18e015fe7f --- /dev/null +++ b/src/core/machine-id-setup.c @@ -0,0 +1,237 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <unistd.h> +#include <stdio.h> +#include <errno.h> +#include <string.h> +#include <stdlib.h> +#include <fcntl.h> +#include <sys/mount.h> + +#include <systemd/sd-id128.h> + +#include "machine-id-setup.h" +#include "macro.h" +#include "util.h" +#include "mkdir.h" +#include "log.h" +#include "virt.h" +#include "fileio.h" + +static int shorten_uuid(char destination[36], const char *source) { + unsigned i, j; + + for (i = 0, j = 0; i < 36 && j < 32; i++) { + int t; + + t = unhexchar(source[i]); + if (t < 0) + continue; + + destination[j++] = hexchar(t); + } + + if (i == 36 && j == 32) { + destination[32] = '\n'; + destination[33] = 0; + return 0; + } + + return -EINVAL; +} + +static int generate(char id[34]) { + int fd, r; + unsigned char *p; + sd_id128_t buf; + char *q; + ssize_t k; + const char *vm_id; + + assert(id); + + /* First, try reading the D-Bus machine id, unless it is a symlink */ + fd = open("/var/lib/dbus/machine-id", O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); + if (fd >= 0) { + k = loop_read(fd, id, 33, false); + close_nointr_nofail(fd); + + if (k == 33 && id[32] == '\n') { + + id[32] = 0; + if (id128_is_valid(id)) { + id[32] = '\n'; + id[33] = 0; + + log_info("Initializing machine ID from D-Bus machine ID."); + return 0; + } + } + } + + /* If that didn't work, see if we are running in qemu/kvm and a + * machine ID was passed in via -uuid on the qemu/kvm command + * line */ + + r = detect_vm(&vm_id); + if (r > 0 && streq(vm_id, "kvm")) { + char uuid[37]; + + fd = open("/sys/class/dmi/id/product_uuid", O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); + if (fd >= 0) { + k = loop_read(fd, uuid, 36, false); + close_nointr_nofail(fd); + + if (k >= 36) { + r = shorten_uuid(id, uuid); + if (r >= 0) { + log_info("Initializing machine ID from KVM UUID."); + return 0; + } + } + } + } + + /* If that didn't work either, see if we are running in a + * container, and a machine ID was passed in via + * $container_uuid the way libvirt/LXC does it */ + r = detect_container(NULL); + if (r > 0) { + _cleanup_free_ char *e = NULL; + + r = getenv_for_pid(1, "container_uuid", &e); + if (r > 0) { + if (strlen(e) >= 36) { + r = shorten_uuid(id, e); + if (r >= 0) { + log_info("Initializing machine ID from container UUID."); + return 0; + } + } + } + } + + /* If that didn't work, generate a random machine id */ + r = sd_id128_randomize(&buf); + if (r < 0) { + log_error("Failed to open /dev/urandom: %s", strerror(-r)); + return r; + } + + for (p = buf.bytes, q = id; p < buf.bytes + sizeof(buf); p++, q += 2) { + q[0] = hexchar(*p >> 4); + q[1] = hexchar(*p & 15); + } + + id[32] = '\n'; + id[33] = 0; + + log_info("Initializing machine ID from random generator."); + + return 0; +} + +int machine_id_setup(void) { + _cleanup_close_ int fd = -1; + int r; + bool writable; + struct stat st; + char id[34]; /* 32 + \n + \0 */ + + RUN_WITH_UMASK(0000) { + /* We create this 0444, to indicate that this isn't really + * something you should ever modify. Of course, since the file + * will be owned by root it doesn't matter much, but maybe + * people look. */ + + fd = open("/etc/machine-id", O_RDWR|O_CREAT|O_CLOEXEC|O_NOCTTY, 0444); + if (fd >= 0) + writable = true; + else { + fd = open("/etc/machine-id", O_RDONLY|O_CLOEXEC|O_NOCTTY); + if (fd < 0) { + log_error("Cannot open /etc/machine-id: %m"); + return -errno; + } + + writable = false; + } + } + + if (fstat(fd, &st) < 0) { + log_error("fstat() failed: %m"); + return -errno; + } + + if (S_ISREG(st.st_mode)) + if (loop_read(fd, id, 33, false) == 33 && id[32] == '\n') { + id[32] = 0; + + if (id128_is_valid(id)) + return 0; + } + + /* Hmm, so, the id currently stored is not useful, then let's + * generate one */ + + r = generate(id); + if (r < 0) + return r; + + if (S_ISREG(st.st_mode) && writable) { + lseek(fd, 0, SEEK_SET); + + if (loop_write(fd, id, 33, false) == 33) + return 0; + } + + close_nointr_nofail(fd); + fd = -1; + + /* Hmm, we couldn't write it? So let's write it to + * /run/machine-id as a replacement */ + + RUN_WITH_UMASK(0022) { + r = write_string_file("/run/machine-id", id); + } + if (r < 0) { + log_error("Cannot write /run/machine-id: %s", strerror(-r)); + unlink("/run/machine-id"); + return r; + } + + /* And now, let's mount it over */ + r = mount("/run/machine-id", "/etc/machine-id", NULL, MS_BIND, NULL); + if (r < 0) { + log_error("Failed to mount /etc/machine-id: %m"); + unlink_noerrno("/run/machine-id"); + return -errno; + } + + log_info("Installed transient /etc/machine-id file."); + + /* Mark the mount read-only */ + if (mount(NULL, "/etc/machine-id", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, NULL) < 0) + log_warning("Failed to make transient /etc/machine-id read-only: %m"); + + return 0; +} diff --git a/src/core/machine-id-setup.h b/src/core/machine-id-setup.h new file mode 100644 index 0000000000..b9e6b4d674 --- /dev/null +++ b/src/core/machine-id-setup.h @@ -0,0 +1,24 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int machine_id_setup(void); diff --git a/src/core/macros.systemd.in b/src/core/macros.systemd.in new file mode 100644 index 0000000000..f77082c2db --- /dev/null +++ b/src/core/macros.systemd.in @@ -0,0 +1,73 @@ +# -*- Mode: makefile; indent-tabs-mode: t -*- */ +# +# This file is part of systemd. +# +# Copyright 2012 Lennart Poettering +# +# systemd is free software; you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published by +# the Free Software Foundation; either version 2.1 of the License, or +# (at your option) any later version. +# +# systemd is distributed in the hope that it will be useful, but +# WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public License +# along with systemd; If not, see <http://www.gnu.org/licenses/>. + +# RPM macros for packages installing systemd unit files + +%_unitdir @systemunitdir@ +%_presetdir @systempresetdir@ +%_udevhwdbdir @udevhwdbdir@ +%_udevrulesdir @udevrulesdir@ +%_journalcatalogdir @catalogdir@ +%_tmpfilesdir @tmpfilesdir@ +%_sysctldir @sysctldir@ + +%systemd_requires \ +Requires(post): systemd \ +Requires(preun): systemd \ +Requires(postun): systemd \ +%{nil} + +%systemd_post() \ +if [ $1 -eq 1 ] ; then \ + # Initial installation \ + @rootbindir@/systemctl preset %{?*} >/dev/null 2>&1 || : \ +fi \ +%{nil} + +%systemd_preun() \ +if [ $1 -eq 0 ] ; then \ + # Package removal, not upgrade \ + @rootbindir@/systemctl --no-reload disable %{?*} > /dev/null 2>&1 || : \ + @rootbindir@/systemctl stop %{?*} > /dev/null 2>&1 || : \ +fi \ +%{nil} + +%systemd_postun() \ +@rootbindir@/systemctl daemon-reload >/dev/null 2>&1 || : \ +%{nil} + +%systemd_postun_with_restart() \ +@rootbindir@/systemctl daemon-reload >/dev/null 2>&1 || : \ +if [ $1 -ge 1 ] ; then \ + # Package upgrade, not uninstall \ + @rootbindir@/systemctl try-restart %{?*} >/dev/null 2>&1 || : \ +fi \ +%{nil} + +%udev_hwdb_update() \ +@rootbindir@/udevadm hwdb --update >/dev/null 2>&1 || : \ +%{nil} + +%udev_rules_update() \ +@rootbindir@/udevadm control --reload >/dev/null 2>&1 || : \ +%{nil} + +%journal_catalog_update() \ +@rootbindir@/journalctl --update-catalog >/dev/null 2>&1 || : \ +%{nil} diff --git a/src/core/main.c b/src/core/main.c new file mode 100644 index 0000000000..7fc06bea05 --- /dev/null +++ b/src/core/main.c @@ -0,0 +1,1958 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus/dbus.h> + +#include <stdio.h> +#include <errno.h> +#include <string.h> +#include <unistd.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <getopt.h> +#include <signal.h> +#include <sys/wait.h> +#include <fcntl.h> +#include <sys/prctl.h> +#include <sys/mount.h> + +#include "manager.h" +#include "log.h" +#include "load-fragment.h" +#include "fdset.h" +#include "special.h" +#include "conf-parser.h" +#include "dbus-common.h" +#include "missing.h" +#include "label.h" +#include "build.h" +#include "strv.h" +#include "def.h" +#include "virt.h" +#include "watchdog.h" +#include "path-util.h" +#include "switch-root.h" +#include "capability.h" +#include "killall.h" +#include "env-util.h" +#include "hwclock.h" +#include "sd-daemon.h" +#include "sd-messages.h" + +#include "mount-setup.h" +#include "loopback-setup.h" +#ifdef HAVE_KMOD +#include "kmod-setup.h" +#endif +#include "hostname-setup.h" +#include "machine-id-setup.h" +#include "locale-setup.h" +#include "selinux-setup.h" +#include "ima-setup.h" +#include "fileio.h" +#include "smack-setup.h" +#include "efivars.h" + +static enum { + ACTION_RUN, + ACTION_HELP, + ACTION_VERSION, + ACTION_TEST, + ACTION_DUMP_CONFIGURATION_ITEMS, + ACTION_DONE +} arg_action = ACTION_RUN; + +static char *arg_default_unit = NULL; +static SystemdRunningAs arg_running_as = _SYSTEMD_RUNNING_AS_INVALID; + +static bool arg_dump_core = true; +static bool arg_crash_shell = false; +static int arg_crash_chvt = -1; +static bool arg_confirm_spawn = false; +static bool arg_show_status = true; +static bool arg_switched_root = false; +static char **arg_default_controllers = NULL; +static char ***arg_join_controllers = NULL; +static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL; +static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT; +static usec_t arg_runtime_watchdog = 0; +static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE; +static struct rlimit *arg_default_rlimit[RLIMIT_NLIMITS] = {}; +static uint64_t arg_capability_bounding_set_drop = 0; +static nsec_t arg_timer_slack_nsec = (nsec_t) -1; + +static FILE* serialization = NULL; + +static void nop_handler(int sig) { +} + +_noreturn_ static void crash(int sig) { + + if (!arg_dump_core) + log_error("Caught <%s>, not dumping core.", signal_to_string(sig)); + else { + struct sigaction sa = { + .sa_handler = nop_handler, + .sa_flags = SA_NOCLDSTOP|SA_RESTART, + }; + pid_t pid; + + /* We want to wait for the core process, hence let's enable SIGCHLD */ + assert_se(sigaction(SIGCHLD, &sa, NULL) == 0); + + pid = fork(); + if (pid < 0) + log_error("Caught <%s>, cannot fork for core dump: %s", signal_to_string(sig), strerror(errno)); + + else if (pid == 0) { + struct rlimit rl = {}; + + /* Enable default signal handler for core dump */ + zero(sa); + sa.sa_handler = SIG_DFL; + assert_se(sigaction(sig, &sa, NULL) == 0); + + /* Don't limit the core dump size */ + rl.rlim_cur = RLIM_INFINITY; + rl.rlim_max = RLIM_INFINITY; + setrlimit(RLIMIT_CORE, &rl); + + /* Just to be sure... */ + assert_se(chdir("/") == 0); + + /* Raise the signal again */ + raise(sig); + + assert_not_reached("We shouldn't be here..."); + _exit(1); + + } else { + siginfo_t status; + int r; + + /* Order things nicely. */ + r = wait_for_terminate(pid, &status); + if (r < 0) + log_error("Caught <%s>, waitpid() failed: %s", signal_to_string(sig), strerror(-r)); + else if (status.si_code != CLD_DUMPED) + log_error("Caught <%s>, core dump failed.", signal_to_string(sig)); + else + log_error("Caught <%s>, dumped core as pid %lu.", signal_to_string(sig), (unsigned long) pid); + } + } + + if (arg_crash_chvt) + chvt(arg_crash_chvt); + + if (arg_crash_shell) { + struct sigaction sa = { + .sa_handler = SIG_IGN, + .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART, + }; + pid_t pid; + + log_info("Executing crash shell in 10s..."); + sleep(10); + + /* Let the kernel reap children for us */ + assert_se(sigaction(SIGCHLD, &sa, NULL) == 0); + + pid = fork(); + if (pid < 0) + log_error("Failed to fork off crash shell: %m"); + else if (pid == 0) { + make_console_stdio(); + execl("/bin/sh", "/bin/sh", NULL); + + log_error("execl() failed: %m"); + _exit(1); + } + + log_info("Successfully spawned crash shell as pid %lu.", (unsigned long) pid); + } + + log_info("Freezing execution."); + freeze(); +} + +static void install_crash_handler(void) { + struct sigaction sa = { + .sa_handler = crash, + .sa_flags = SA_NODEFER, + }; + + sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1); +} + +static int console_setup(bool do_reset) { + int tty_fd, r; + + /* If we are init, we connect stdin/stdout/stderr to /dev/null + * and make sure we don't have a controlling tty. */ + + release_terminal(); + + if (!do_reset) + return 0; + + tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC); + if (tty_fd < 0) { + log_error("Failed to open /dev/console: %s", strerror(-tty_fd)); + return -tty_fd; + } + + /* We don't want to force text mode. + * plymouth may be showing pictures already from initrd. */ + r = reset_terminal_fd(tty_fd, false); + if (r < 0) + log_error("Failed to reset /dev/console: %s", strerror(-r)); + + close_nointr_nofail(tty_fd); + return r; +} + +static int set_default_unit(const char *u) { + char *c; + + assert(u); + + c = strdup(u); + if (!c) + return -ENOMEM; + + free(arg_default_unit); + arg_default_unit = c; + + return 0; +} + +static int parse_proc_cmdline_word(const char *word) { + + static const char * const rlmap[] = { + "emergency", SPECIAL_EMERGENCY_TARGET, + "-b", SPECIAL_EMERGENCY_TARGET, + "single", SPECIAL_RESCUE_TARGET, + "-s", SPECIAL_RESCUE_TARGET, + "s", SPECIAL_RESCUE_TARGET, + "S", SPECIAL_RESCUE_TARGET, + "1", SPECIAL_RESCUE_TARGET, + "2", SPECIAL_RUNLEVEL2_TARGET, + "3", SPECIAL_RUNLEVEL3_TARGET, + "4", SPECIAL_RUNLEVEL4_TARGET, + "5", SPECIAL_RUNLEVEL5_TARGET, + }; + + assert(word); + + if (startswith(word, "systemd.unit=")) { + + if (!in_initrd()) + return set_default_unit(word + 13); + + } else if (startswith(word, "rd.systemd.unit=")) { + + if (in_initrd()) + return set_default_unit(word + 16); + + } else if (startswith(word, "systemd.log_target=")) { + + if (log_set_target_from_string(word + 19) < 0) + log_warning("Failed to parse log target %s. Ignoring.", word + 19); + + } else if (startswith(word, "systemd.log_level=")) { + + if (log_set_max_level_from_string(word + 18) < 0) + log_warning("Failed to parse log level %s. Ignoring.", word + 18); + + } else if (startswith(word, "systemd.log_color=")) { + + if (log_show_color_from_string(word + 18) < 0) + log_warning("Failed to parse log color setting %s. Ignoring.", word + 18); + + } else if (startswith(word, "systemd.log_location=")) { + + if (log_show_location_from_string(word + 21) < 0) + log_warning("Failed to parse log location setting %s. Ignoring.", word + 21); + + } else if (startswith(word, "systemd.dump_core=")) { + int r; + + if ((r = parse_boolean(word + 18)) < 0) + log_warning("Failed to parse dump core switch %s. Ignoring.", word + 18); + else + arg_dump_core = r; + + } else if (startswith(word, "systemd.crash_shell=")) { + int r; + + if ((r = parse_boolean(word + 20)) < 0) + log_warning("Failed to parse crash shell switch %s. Ignoring.", word + 20); + else + arg_crash_shell = r; + + } else if (startswith(word, "systemd.confirm_spawn=")) { + int r; + + if ((r = parse_boolean(word + 22)) < 0) + log_warning("Failed to parse confirm spawn switch %s. Ignoring.", word + 22); + else + arg_confirm_spawn = r; + + } else if (startswith(word, "systemd.crash_chvt=")) { + int k; + + if (safe_atoi(word + 19, &k) < 0) + log_warning("Failed to parse crash chvt switch %s. Ignoring.", word + 19); + else + arg_crash_chvt = k; + + } else if (startswith(word, "systemd.show_status=")) { + int r; + + if ((r = parse_boolean(word + 20)) < 0) + log_warning("Failed to parse show status switch %s. Ignoring.", word + 20); + else + arg_show_status = r; + } else if (startswith(word, "systemd.default_standard_output=")) { + int r; + + if ((r = exec_output_from_string(word + 32)) < 0) + log_warning("Failed to parse default standard output switch %s. Ignoring.", word + 32); + else + arg_default_std_output = r; + } else if (startswith(word, "systemd.default_standard_error=")) { + int r; + + if ((r = exec_output_from_string(word + 31)) < 0) + log_warning("Failed to parse default standard error switch %s. Ignoring.", word + 31); + else + arg_default_std_error = r; + } else if (startswith(word, "systemd.setenv=")) { + _cleanup_free_ char *cenv = NULL; + char *eq; + int r; + + cenv = strdup(word + 15); + if (!cenv) + return -ENOMEM; + + eq = strchr(cenv, '='); + if (!eq) { + if (!env_name_is_valid(cenv)) + log_warning("Environment variable name '%s' is not valid. Ignoring.", cenv); + else { + r = unsetenv(cenv); + if (r < 0) + log_warning("Unsetting environment variable '%s' failed, ignoring: %m", cenv); + } + } else { + if (!env_assignment_is_valid(cenv)) + log_warning("Environment variable assignment '%s' is not valid. Ignoring.", cenv); + else { + *eq = 0; + r = setenv(cenv, eq + 1, 1); + if (r < 0) + log_warning("Setting environment variable '%s=%s' failed, ignoring: %m", cenv, eq + 1); + } + } + + } else if (startswith(word, "systemd.") || + (in_initrd() && startswith(word, "rd.systemd."))) { + + const char *c; + + /* Ignore systemd.journald.xyz and friends */ + c = word; + if (startswith(c, "rd.")) + c += 3; + if (startswith(c, "systemd.")) + c += 8; + if (c[strcspn(c, ".=")] != '.') { + + log_warning("Unknown kernel switch %s. Ignoring.", word); + + log_info("Supported kernel switches:\n" + "systemd.unit=UNIT Default unit to start\n" + "rd.systemd.unit=UNIT Default unit to start when run in initrd\n" + "systemd.dump_core=0|1 Dump core on crash\n" + "systemd.crash_shell=0|1 Run shell on crash\n" + "systemd.crash_chvt=N Change to VT #N on crash\n" + "systemd.confirm_spawn=0|1 Confirm every process spawn\n" + "systemd.show_status=0|1 Show status updates on the console during bootup\n" + "systemd.log_target=console|kmsg|journal|journal-or-kmsg|syslog|syslog-or-kmsg|null\n" + " Log target\n" + "systemd.log_level=LEVEL Log level\n" + "systemd.log_color=0|1 Highlight important log messages\n" + "systemd.log_location=0|1 Include code location in log messages\n" + "systemd.default_standard_output=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n" + " Set default log output for services\n" + "systemd.default_standard_error=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n" + " Set default log error output for services\n" + "systemd.setenv=ASSIGNMENT Set an environment variable for all spawned processes\n"); + } + + } else if (streq(word, "quiet")) + arg_show_status = false; + else if (!in_initrd()) { + unsigned i; + + /* SysV compatibility */ + for (i = 0; i < ELEMENTSOF(rlmap); i += 2) + if (streq(word, rlmap[i])) + return set_default_unit(rlmap[i+1]); + } + + return 0; +} + +#define DEFINE_SETTER(name, func, descr) \ + static int name(const char *unit, \ + const char *filename, \ + unsigned line, \ + const char *section, \ + const char *lvalue, \ + int ltype, \ + const char *rvalue, \ + void *data, \ + void *userdata) { \ + \ + int r; \ + \ + assert(filename); \ + assert(lvalue); \ + assert(rvalue); \ + \ + r = func(rvalue); \ + if (r < 0) \ + log_syntax(unit, LOG_ERR, filename, line, -r, \ + "Invalid " descr "'%s': %s", \ + rvalue, strerror(-r)); \ + \ + return 0; \ + } + +DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level") +DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target") +DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" ) +DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location") + + +static int config_parse_cpu_affinity2(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + char *w; + size_t l; + char *state; + cpu_set_t *c = NULL; + unsigned ncpus = 0; + + assert(filename); + assert(lvalue); + assert(rvalue); + + FOREACH_WORD_QUOTED(w, l, rvalue, state) { + char *t; + int r; + unsigned cpu; + + if (!(t = strndup(w, l))) + return log_oom(); + + r = safe_atou(t, &cpu); + free(t); + + if (!c) + if (!(c = cpu_set_malloc(&ncpus))) + return log_oom(); + + if (r < 0 || cpu >= ncpus) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to parse CPU affinity '%s'", rvalue); + CPU_FREE(c); + return -EBADMSG; + } + + CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c); + } + + if (c) { + if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0) + log_warning_unit(unit, "Failed to set CPU affinity: %m"); + + CPU_FREE(c); + } + + return 0; +} + +static void strv_free_free(char ***l) { + char ***i; + + if (!l) + return; + + for (i = l; *i; i++) + strv_free(*i); + + free(l); +} + +static void free_join_controllers(void) { + strv_free_free(arg_join_controllers); + arg_join_controllers = NULL; +} + +static int config_parse_join_controllers(const char *unit, + const char *filename, + unsigned line, + const char *section, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + unsigned n = 0; + char *state, *w; + size_t length; + + assert(filename); + assert(lvalue); + assert(rvalue); + + free_join_controllers(); + + FOREACH_WORD_QUOTED(w, length, rvalue, state) { + char *s, **l; + + s = strndup(w, length); + if (!s) + return log_oom(); + + l = strv_split(s, ","); + free(s); + + strv_uniq(l); + + if (strv_length(l) <= 1) { + strv_free(l); + continue; + } + + if (!arg_join_controllers) { + arg_join_controllers = new(char**, 2); + if (!arg_join_controllers) { + strv_free(l); + return log_oom(); + } + + arg_join_controllers[0] = l; + arg_join_controllers[1] = NULL; + + n = 1; + } else { + char ***a; + char ***t; + + t = new0(char**, n+2); + if (!t) { + strv_free(l); + return log_oom(); + } + + n = 0; + + for (a = arg_join_controllers; *a; a++) { + + if (strv_overlap(*a, l)) { + char **c; + + c = strv_merge(*a, l); + if (!c) { + strv_free(l); + strv_free_free(t); + return log_oom(); + } + + strv_free(l); + l = c; + } else { + char **c; + + c = strv_copy(*a); + if (!c) { + strv_free(l); + strv_free_free(t); + return log_oom(); + } + + t[n++] = c; + } + } + + t[n++] = strv_uniq(l); + + strv_free_free(arg_join_controllers); + arg_join_controllers = t; + } + } + + return 0; +} + +static int parse_config_file(void) { + + const ConfigTableItem items[] = { + { "Manager", "LogLevel", config_parse_level2, 0, NULL }, + { "Manager", "LogTarget", config_parse_target, 0, NULL }, + { "Manager", "LogColor", config_parse_color, 0, NULL }, + { "Manager", "LogLocation", config_parse_location, 0, NULL }, + { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core }, + { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell }, + { "Manager", "ShowStatus", config_parse_bool, 0, &arg_show_status }, + { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt }, + { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL }, + { "Manager", "DefaultControllers", config_parse_strv, 0, &arg_default_controllers }, + { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output }, + { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error }, + { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers }, + { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog }, + { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog }, + { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop }, + { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec }, + { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU]}, + { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE]}, + { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA]}, + { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK]}, + { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE]}, + { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS]}, + { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE]}, + { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS]}, + { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC]}, + { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK]}, + { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS]}, + { "Manager", "DefaultLimitSIGPENDING",config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING]}, + { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE]}, + { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE]}, + { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO]}, + { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME]}, + { NULL, NULL, NULL, 0, NULL } + }; + + _cleanup_fclose_ FILE *f; + const char *fn; + int r; + + fn = arg_running_as == SYSTEMD_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf"; + f = fopen(fn, "re"); + if (!f) { + if (errno == ENOENT) + return 0; + + log_warning("Failed to open configuration file '%s': %m", fn); + return 0; + } + + r = config_parse(NULL, fn, f, "Manager\0", config_item_table_lookup, (void*) items, false, false, NULL); + if (r < 0) + log_warning("Failed to parse configuration file: %s", strerror(-r)); + + return 0; +} + +static int parse_proc_cmdline(void) { + _cleanup_free_ char *line = NULL; + char *w, *state; + int r; + size_t l; + + /* Don't read /proc/cmdline if we are in a container, since + * that is only relevant for the host system */ + if (detect_container(NULL) > 0) + return 0; + + r = read_one_line_file("/proc/cmdline", &line); + if (r < 0) { + log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r)); + return 0; + } + + FOREACH_WORD_QUOTED(w, l, line, state) { + _cleanup_free_ char *word; + + word = strndup(w, l); + if (!word) + return log_oom(); + + r = parse_proc_cmdline_word(word); + if (r < 0) { + log_error("Failed on cmdline argument %s: %s", word, strerror(-r)); + return r; + } + } + + return 0; +} + +static int parse_argv(int argc, char *argv[]) { + + enum { + ARG_LOG_LEVEL = 0x100, + ARG_LOG_TARGET, + ARG_LOG_COLOR, + ARG_LOG_LOCATION, + ARG_UNIT, + ARG_SYSTEM, + ARG_USER, + ARG_TEST, + ARG_VERSION, + ARG_DUMP_CONFIGURATION_ITEMS, + ARG_DUMP_CORE, + ARG_CRASH_SHELL, + ARG_CONFIRM_SPAWN, + ARG_SHOW_STATUS, + ARG_DESERIALIZE, + ARG_SWITCHED_ROOT, + ARG_INTROSPECT, + ARG_DEFAULT_STD_OUTPUT, + ARG_DEFAULT_STD_ERROR + }; + + static const struct option options[] = { + { "log-level", required_argument, NULL, ARG_LOG_LEVEL }, + { "log-target", required_argument, NULL, ARG_LOG_TARGET }, + { "log-color", optional_argument, NULL, ARG_LOG_COLOR }, + { "log-location", optional_argument, NULL, ARG_LOG_LOCATION }, + { "unit", required_argument, NULL, ARG_UNIT }, + { "system", no_argument, NULL, ARG_SYSTEM }, + { "user", no_argument, NULL, ARG_USER }, + { "test", no_argument, NULL, ARG_TEST }, + { "help", no_argument, NULL, 'h' }, + { "version", no_argument, NULL, ARG_VERSION }, + { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS }, + { "dump-core", optional_argument, NULL, ARG_DUMP_CORE }, + { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL }, + { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN }, + { "show-status", optional_argument, NULL, ARG_SHOW_STATUS }, + { "deserialize", required_argument, NULL, ARG_DESERIALIZE }, + { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT }, + { "introspect", optional_argument, NULL, ARG_INTROSPECT }, + { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, }, + { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, }, + { NULL, 0, NULL, 0 } + }; + + int c, r; + + assert(argc >= 1); + assert(argv); + + if (getpid() == 1) + opterr = 0; + + while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0) + + switch (c) { + + case ARG_LOG_LEVEL: + if ((r = log_set_max_level_from_string(optarg)) < 0) { + log_error("Failed to parse log level %s.", optarg); + return r; + } + + break; + + case ARG_LOG_TARGET: + + if ((r = log_set_target_from_string(optarg)) < 0) { + log_error("Failed to parse log target %s.", optarg); + return r; + } + + break; + + case ARG_LOG_COLOR: + + if (optarg) { + if ((r = log_show_color_from_string(optarg)) < 0) { + log_error("Failed to parse log color setting %s.", optarg); + return r; + } + } else + log_show_color(true); + + break; + + case ARG_LOG_LOCATION: + + if (optarg) { + if ((r = log_show_location_from_string(optarg)) < 0) { + log_error("Failed to parse log location setting %s.", optarg); + return r; + } + } else + log_show_location(true); + + break; + + case ARG_DEFAULT_STD_OUTPUT: + + if ((r = exec_output_from_string(optarg)) < 0) { + log_error("Failed to parse default standard output setting %s.", optarg); + return r; + } else + arg_default_std_output = r; + break; + + case ARG_DEFAULT_STD_ERROR: + + if ((r = exec_output_from_string(optarg)) < 0) { + log_error("Failed to parse default standard error output setting %s.", optarg); + return r; + } else + arg_default_std_error = r; + break; + + case ARG_UNIT: + + if ((r = set_default_unit(optarg)) < 0) { + log_error("Failed to set default unit %s: %s", optarg, strerror(-r)); + return r; + } + + break; + + case ARG_SYSTEM: + arg_running_as = SYSTEMD_SYSTEM; + break; + + case ARG_USER: + arg_running_as = SYSTEMD_USER; + break; + + case ARG_TEST: + arg_action = ACTION_TEST; + break; + + case ARG_VERSION: + arg_action = ACTION_VERSION; + break; + + case ARG_DUMP_CONFIGURATION_ITEMS: + arg_action = ACTION_DUMP_CONFIGURATION_ITEMS; + break; + + case ARG_DUMP_CORE: + r = optarg ? parse_boolean(optarg) : 1; + if (r < 0) { + log_error("Failed to parse dump core boolean %s.", optarg); + return r; + } + arg_dump_core = r; + break; + + case ARG_CRASH_SHELL: + r = optarg ? parse_boolean(optarg) : 1; + if (r < 0) { + log_error("Failed to parse crash shell boolean %s.", optarg); + return r; + } + arg_crash_shell = r; + break; + + case ARG_CONFIRM_SPAWN: + r = optarg ? parse_boolean(optarg) : 1; + if (r < 0) { + log_error("Failed to parse confirm spawn boolean %s.", optarg); + return r; + } + arg_confirm_spawn = r; + break; + + case ARG_SHOW_STATUS: + r = optarg ? parse_boolean(optarg) : 1; + if (r < 0) { + log_error("Failed to parse show status boolean %s.", optarg); + return r; + } + arg_show_status = r; + break; + + case ARG_DESERIALIZE: { + int fd; + FILE *f; + + r = safe_atoi(optarg, &fd); + if (r < 0 || fd < 0) { + log_error("Failed to parse deserialize option %s.", optarg); + return r < 0 ? r : -EINVAL; + } + + fd_cloexec(fd, true); + + f = fdopen(fd, "r"); + if (!f) { + log_error("Failed to open serialization fd: %m"); + return -errno; + } + + if (serialization) + fclose(serialization); + + serialization = f; + + break; + } + + case ARG_SWITCHED_ROOT: + arg_switched_root = true; + break; + + case ARG_INTROSPECT: { + const char * const * i = NULL; + + for (i = bus_interface_table; *i; i += 2) + if (!optarg || streq(i[0], optarg)) { + fputs(DBUS_INTROSPECT_1_0_XML_DOCTYPE_DECL_NODE + "<node>\n", stdout); + fputs(i[1], stdout); + fputs("</node>\n", stdout); + + if (optarg) + break; + } + + if (!i[0] && optarg) + log_error("Unknown interface %s.", optarg); + + arg_action = ACTION_DONE; + break; + } + + case 'h': + arg_action = ACTION_HELP; + break; + + case 'D': + log_set_max_level(LOG_DEBUG); + break; + + case 'b': + case 's': + case 'z': + /* Just to eat away the sysvinit kernel + * cmdline args without getopt() error + * messages that we'll parse in + * parse_proc_cmdline_word() or ignore. */ + + case '?': + default: + if (getpid() != 1) { + log_error("Unknown option code %c", c); + return -EINVAL; + } + + break; + } + + if (optind < argc && getpid() != 1) { + /* Hmm, when we aren't run as init system + * let's complain about excess arguments */ + + log_error("Excess arguments."); + return -EINVAL; + } + + if (detect_container(NULL) > 0) { + char **a; + + /* All /proc/cmdline arguments the kernel didn't + * understand it passed to us. We're not really + * interested in that usually since /proc/cmdline is + * more interesting and complete. With one exception: + * if we are run in a container /proc/cmdline is not + * relevant for the container, hence we rely on argv[] + * instead. */ + + for (a = argv; a < argv + argc; a++) + if ((r = parse_proc_cmdline_word(*a)) < 0) { + log_error("Failed on cmdline argument %s: %s", *a, strerror(-r)); + return r; + } + } + + return 0; +} + +static int help(void) { + + printf("%s [OPTIONS...]\n\n" + "Starts up and maintains the system or user services.\n\n" + " -h --help Show this help\n" + " --test Determine startup sequence, dump it and exit\n" + " --dump-configuration-items Dump understood unit configuration items\n" + " --introspect[=INTERFACE] Extract D-Bus interface data\n" + " --unit=UNIT Set default unit\n" + " --system Run a system instance, even if PID != 1\n" + " --user Run a user instance\n" + " --dump-core[=0|1] Dump core on crash\n" + " --crash-shell[=0|1] Run shell on crash\n" + " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n" + " --show-status[=0|1] Show status updates on the console during bootup\n" + " --log-target=TARGET Set log target (console, journal, syslog, kmsg, journal-or-kmsg, syslog-or-kmsg, null)\n" + " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n" + " --log-color[=0|1] Highlight important log messages\n" + " --log-location[=0|1] Include code location in log messages\n" + " --default-standard-output= Set default standard output for services\n" + " --default-standard-error= Set default standard error output for services\n", + program_invocation_short_name); + + return 0; +} + +static int version(void) { + puts(PACKAGE_STRING); + puts(SYSTEMD_FEATURES); + + return 0; +} + +static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) { + FILE *f = NULL; + FDSet *fds = NULL; + int r; + + assert(m); + assert(_f); + assert(_fds); + + /* Make sure nothing is really destructed when we shut down */ + m->n_reloading ++; + + r = manager_open_serialization(m, &f); + if (r < 0) { + log_error("Failed to create serialization file: %s", strerror(-r)); + goto fail; + } + + fds = fdset_new(); + if (!fds) { + r = -ENOMEM; + log_error("Failed to allocate fd set: %s", strerror(-r)); + goto fail; + } + + r = manager_serialize(m, f, fds, switching_root); + if (r < 0) { + log_error("Failed to serialize state: %s", strerror(-r)); + goto fail; + } + + if (fseeko(f, 0, SEEK_SET) < 0) { + log_error("Failed to rewind serialization fd: %m"); + goto fail; + } + + r = fd_cloexec(fileno(f), false); + if (r < 0) { + log_error("Failed to disable O_CLOEXEC for serialization: %s", strerror(-r)); + goto fail; + } + + r = fdset_cloexec(fds, false); + if (r < 0) { + log_error("Failed to disable O_CLOEXEC for serialization fds: %s", strerror(-r)); + goto fail; + } + + *_f = f; + *_fds = fds; + + return 0; + +fail: + fdset_free(fds); + + if (f) + fclose(f); + + return r; +} + +static int bump_rlimit_nofile(struct rlimit *saved_rlimit) { + struct rlimit nl; + int r; + + assert(saved_rlimit); + + /* Save the original RLIMIT_NOFILE so that we can reset it + * later when transitioning from the initrd to the main + * systemd or suchlike. */ + if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) { + log_error("Reading RLIMIT_NOFILE failed: %m"); + return -errno; + } + + /* Make sure forked processes get the default kernel setting */ + if (!arg_default_rlimit[RLIMIT_NOFILE]) { + struct rlimit *rl; + + rl = newdup(struct rlimit, saved_rlimit, 1); + if (!rl) + return log_oom(); + + arg_default_rlimit[RLIMIT_NOFILE] = rl; + } + + /* Bump up the resource limit for ourselves substantially */ + nl.rlim_cur = nl.rlim_max = 64*1024; + r = setrlimit_closest(RLIMIT_NOFILE, &nl); + if (r < 0) { + log_error("Setting RLIMIT_NOFILE failed: %s", strerror(-r)); + return r; + } + + return 0; +} + +static struct dual_timestamp* parse_initrd_timestamp(struct dual_timestamp *t) { + const char *e; + unsigned long long a, b; + + assert(t); + + e = getenv("RD_TIMESTAMP"); + if (!e) + return NULL; + + if (sscanf(e, "%llu %llu", &a, &b) != 2) + return NULL; + + t->realtime = (usec_t) a; + t->monotonic = (usec_t) b; + + return t; +} + +static void test_mtab(void) { + char *p; + + /* Check that /etc/mtab is a symlink */ + + if (readlink_malloc("/etc/mtab", &p) >= 0) { + bool b; + + b = streq(p, "/proc/self/mounts") || streq(p, "/proc/mounts"); + free(p); + + if (b) + return; + } + + log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. " + "This is not supported anymore. " + "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output."); +} + +static void test_usr(void) { + + /* Check that /usr is not a separate fs */ + + if (dir_is_empty("/usr") <= 0) + return; + + log_warning("/usr appears to be on its own filesytem and is not already mounted. This is not a supported setup. " + "Some things will probably break (sometimes even silently) in mysterious ways. " + "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information."); +} + +static void test_cgroups(void) { + + if (access("/proc/cgroups", F_OK) >= 0) + return; + + log_warning("CONFIG_CGROUPS was not set when your kernel was compiled. " + "Systems without control groups are not supported. " + "We will now sleep for 10s, and then continue boot-up. " + "Expect breakage and please do not file bugs. " + "Instead fix your kernel and enable CONFIG_CGROUPS. " + "Consult http://0pointer.de/blog/projects/cgroups-vs-cgroups.html for more information."); + + sleep(10); +} + +static int initialize_join_controllers(void) { + /* By default, mount "cpu" + "cpuacct" together, and "net_cls" + * + "net_prio". We'd like to add "cpuset" to the mix, but + * "cpuset" does't really work for groups with no initialized + * attributes. */ + + arg_join_controllers = new(char**, 3); + if (!arg_join_controllers) + return -ENOMEM; + + arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL); + arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL); + arg_join_controllers[2] = NULL; + + if (!arg_join_controllers[0] || !arg_join_controllers[1]) { + free_join_controllers(); + return -ENOMEM; + } + + return 0; +} + +int main(int argc, char *argv[]) { + Manager *m = NULL; + int r, retval = EXIT_FAILURE; + usec_t before_startup, after_startup; + char timespan[FORMAT_TIMESPAN_MAX]; + FDSet *fds = NULL; + bool reexecute = false; + const char *shutdown_verb = NULL; + dual_timestamp initrd_timestamp = { 0ULL, 0ULL }; + dual_timestamp userspace_timestamp = { 0ULL, 0ULL }; + dual_timestamp kernel_timestamp = { 0ULL, 0ULL }; + dual_timestamp firmware_timestamp = { 0ULL, 0ULL }; + dual_timestamp loader_timestamp = { 0ULL, 0ULL }; + static char systemd[] = "systemd"; + bool skip_setup = false; + int j; + bool loaded_policy = false; + bool arm_reboot_watchdog = false; + bool queue_default_job = false; + char *switch_root_dir = NULL, *switch_root_init = NULL; + static struct rlimit saved_rlimit_nofile = { 0, 0 }; + +#ifdef HAVE_SYSV_COMPAT + if (getpid() != 1 && strstr(program_invocation_short_name, "init")) { + /* This is compatibility support for SysV, where + * calling init as a user is identical to telinit. */ + + errno = -ENOENT; + execv(SYSTEMCTL_BINARY_PATH, argv); + log_error("Failed to exec " SYSTEMCTL_BINARY_PATH ": %m"); + return 1; + } +#endif + + dual_timestamp_from_monotonic(&kernel_timestamp, 0); + dual_timestamp_get(&userspace_timestamp); + + /* Determine if this is a reexecution or normal bootup. We do + * the full command line parsing much later, so let's just + * have a quick peek here. */ + if (strv_find(argv+1, "--deserialize")) + skip_setup = true; + + /* If we have switched root, do all the special setup + * things */ + if (strv_find(argv+1, "--switched-root")) + skip_setup = false; + + /* If we get started via the /sbin/init symlink then we are + called 'init'. After a subsequent reexecution we are then + called 'systemd'. That is confusing, hence let's call us + systemd right-away. */ + program_invocation_short_name = systemd; + prctl(PR_SET_NAME, systemd); + + saved_argv = argv; + saved_argc = argc; + + log_show_color(isatty(STDERR_FILENO) > 0); + + if (getpid() == 1 && detect_container(NULL) <= 0) { +#ifdef ENABLE_EFI + efi_get_boot_timestamps(&userspace_timestamp, &firmware_timestamp, &loader_timestamp); +#endif + /* Running outside of a container as PID 1 */ + arg_running_as = SYSTEMD_SYSTEM; + make_null_stdio(); + log_set_target(LOG_TARGET_KMSG); + log_open(); + + if (in_initrd()) { + char *rd_timestamp = NULL; + + initrd_timestamp = userspace_timestamp; + asprintf(&rd_timestamp, "%llu %llu", + (unsigned long long) initrd_timestamp.realtime, + (unsigned long long) initrd_timestamp.monotonic); + if (rd_timestamp) { + setenv("RD_TIMESTAMP", rd_timestamp, 1); + free(rd_timestamp); + } + } + + if (!skip_setup) { + mount_setup_early(); + if (selinux_setup(&loaded_policy) < 0) + goto finish; + if (ima_setup() < 0) + goto finish; + if (smack_setup() < 0) + goto finish; + } + + if (label_init(NULL) < 0) + goto finish; + + if (!skip_setup) { + if (hwclock_is_localtime() > 0) { + int min; + + /* The first-time call to settimeofday() does a time warp in the kernel */ + r = hwclock_set_timezone(&min); + if (r < 0) + log_error("Failed to apply local time delta, ignoring: %s", strerror(-r)); + else + log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min); + } else if (!in_initrd()) { + /* + * Do dummy first-time call to seal the kernel's time warp magic + * + * Do not call this this from inside the initrd. The initrd might not + * carry /etc/adjtime with LOCAL, but the real system could be set up + * that way. In such case, we need to delay the time-warp or the sealing + * until we reach the real system. + */ + hwclock_reset_timezone(); + + /* Tell the kernel our time zone */ + r = hwclock_set_timezone(NULL); + if (r < 0) + log_error("Failed to set the kernel's time zone, ignoring: %s", strerror(-r)); + } + } + + /* Set the default for later on, but don't actually + * open the logs like this for now. Note that if we + * are transitioning from the initrd there might still + * be journal fd open, and we shouldn't attempt + * opening that before we parsed /proc/cmdline which + * might redirect output elsewhere. */ + log_set_target(LOG_TARGET_JOURNAL_OR_KMSG); + + } else if (getpid() == 1) { + /* Running inside a container, as PID 1 */ + arg_running_as = SYSTEMD_SYSTEM; + log_set_target(LOG_TARGET_CONSOLE); + log_open(); + + /* For the later on, see above... */ + log_set_target(LOG_TARGET_JOURNAL); + + /* clear the kernel timestamp, + * because we are in a container */ + kernel_timestamp.monotonic = 0ULL; + kernel_timestamp.realtime = 0ULL; + + } else { + /* Running as user instance */ + arg_running_as = SYSTEMD_USER; + log_set_target(LOG_TARGET_AUTO); + log_open(); + + /* clear the kernel timestamp, + * because we are not PID 1 */ + kernel_timestamp.monotonic = 0ULL; + kernel_timestamp.realtime = 0ULL; + } + + /* Initialize default unit */ + r = set_default_unit(SPECIAL_DEFAULT_TARGET); + if (r < 0) { + log_error("Failed to set default unit %s: %s", SPECIAL_DEFAULT_TARGET, strerror(-r)); + goto finish; + } + + r = initialize_join_controllers(); + if (r < 0) + goto finish; + + /* Mount /proc, /sys and friends, so that /proc/cmdline and + * /proc/$PID/fd is available. */ + if (getpid() == 1) { + r = mount_setup(loaded_policy); + if (r < 0) + goto finish; + } + + /* Reset all signal handlers. */ + assert_se(reset_all_signal_handlers() == 0); + + /* If we are init, we can block sigkill. Yay. */ + ignore_signals(SIGNALS_IGNORE, -1); + + if (parse_config_file() < 0) + goto finish; + + if (arg_running_as == SYSTEMD_SYSTEM) + if (parse_proc_cmdline() < 0) + goto finish; + + log_parse_environment(); + + if (parse_argv(argc, argv) < 0) + goto finish; + + if (arg_action == ACTION_TEST && + geteuid() == 0) { + log_error("Don't run test mode as root."); + goto finish; + } + + if (arg_running_as == SYSTEMD_USER && + arg_action == ACTION_RUN && + sd_booted() <= 0) { + log_error("Trying to run as user instance, but the system has not been booted with systemd."); + goto finish; + } + + if (arg_running_as == SYSTEMD_SYSTEM && + arg_action == ACTION_RUN && + running_in_chroot() > 0) { + log_error("Cannot be run in a chroot() environment."); + goto finish; + } + + if (arg_action == ACTION_HELP) { + retval = help(); + goto finish; + } else if (arg_action == ACTION_VERSION) { + retval = version(); + goto finish; + } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) { + unit_dump_config_items(stdout); + retval = EXIT_SUCCESS; + goto finish; + } else if (arg_action == ACTION_DONE) { + retval = EXIT_SUCCESS; + goto finish; + } + + assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST); + + /* Close logging fds, in order not to confuse fdset below */ + log_close(); + + /* Remember open file descriptors for later deserialization */ + r = fdset_new_fill(&fds); + if (r < 0) { + log_error("Failed to allocate fd set: %s", strerror(-r)); + goto finish; + } else + fdset_cloexec(fds, true); + + if (serialization) + assert_se(fdset_remove(fds, fileno(serialization)) >= 0); + + /* Set up PATH unless it is already set */ + setenv("PATH", +#ifdef HAVE_SPLIT_USR + "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", +#else + "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin", +#endif + arg_running_as == SYSTEMD_SYSTEM); + + if (arg_running_as == SYSTEMD_SYSTEM) { + /* Parse the data passed to us. We leave this + * variables set, but the manager later on will not + * pass them on to our children. */ + if (!in_initrd()) + parse_initrd_timestamp(&initrd_timestamp); + + /* Unset some environment variables passed in from the + * kernel that don't really make sense for us. */ + unsetenv("HOME"); + unsetenv("TERM"); + + /* When we are invoked by a shell, these might be set, + * but make little sense to pass on */ + unsetenv("PWD"); + unsetenv("SHLVL"); + unsetenv("_"); + + /* When we are invoked by a chroot-like tool such as + * nspawn, these might be set, but make little sense + * to pass on */ + unsetenv("USER"); + unsetenv("LOGNAME"); + + /* We suppress the socket activation env vars, as + * we'll try to match *any* open fd to units if + * possible. */ + unsetenv("LISTEN_FDS"); + unsetenv("LISTEN_PID"); + + /* All other variables are left as is, so that clients + * can still read them via /proc/1/environ */ + } + + /* Move out of the way, so that we won't block unmounts */ + assert_se(chdir("/") == 0); + + if (arg_running_as == SYSTEMD_SYSTEM) { + /* Become a session leader if we aren't one yet. */ + setsid(); + + /* Disable the umask logic */ + umask(0); + } + + /* Make sure D-Bus doesn't fiddle with the SIGPIPE handlers */ + dbus_connection_set_change_sigpipe(FALSE); + + /* Reset the console, but only if this is really init and we + * are freshly booted */ + if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN) + console_setup(getpid() == 1 && !skip_setup); + + /* Open the logging devices, if possible and necessary */ + log_open(); + + /* Make sure we leave a core dump without panicing the + * kernel. */ + if (getpid() == 1) { + install_crash_handler(); + + r = mount_cgroup_controllers(arg_join_controllers); + if (r < 0) + goto finish; + } + + if (arg_running_as == SYSTEMD_SYSTEM) { + const char *virtualization = NULL; + + log_info(PACKAGE_STRING " running in system mode. (" SYSTEMD_FEATURES ")"); + + detect_virtualization(&virtualization); + if (virtualization) + log_info("Detected virtualization '%s'.", virtualization); + + if (in_initrd()) + log_info("Running in initial RAM disk."); + + } else + log_debug(PACKAGE_STRING " running in user mode. (" SYSTEMD_FEATURES ")"); + + if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) { + locale_setup(); + + if (arg_show_status || plymouth_running()) + status_welcome(); + +#ifdef HAVE_KMOD + kmod_setup(); +#endif + hostname_setup(); + machine_id_setup(); + loopback_setup(); + + test_mtab(); + test_usr(); + test_cgroups(); + } + + if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0) + watchdog_set_timeout(&arg_runtime_watchdog); + + if (arg_timer_slack_nsec != (nsec_t) -1) + if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0) + log_error("Failed to adjust timer slack: %m"); + + if (arg_capability_bounding_set_drop) { + r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true); + if (r < 0) { + log_error("Failed to drop capability bounding set: %s", strerror(-r)); + goto finish; + } + r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop); + if (r < 0) { + log_error("Failed to drop capability bounding set of usermode helpers: %s", strerror(-r)); + goto finish; + } + } + + if (arg_running_as == SYSTEMD_USER) { + /* Become reaper of our children */ + if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) { + log_warning("Failed to make us a subreaper: %m"); + if (errno == EINVAL) + log_info("Perhaps the kernel version is too old (< 3.4?)"); + } + } + + if (arg_running_as == SYSTEMD_SYSTEM) + bump_rlimit_nofile(&saved_rlimit_nofile); + + r = manager_new(arg_running_as, &m); + if (r < 0) { + log_error("Failed to allocate manager object: %s", strerror(-r)); + goto finish; + } + + m->confirm_spawn = arg_confirm_spawn; + m->default_std_output = arg_default_std_output; + m->default_std_error = arg_default_std_error; + m->runtime_watchdog = arg_runtime_watchdog; + m->shutdown_watchdog = arg_shutdown_watchdog; + m->userspace_timestamp = userspace_timestamp; + m->kernel_timestamp = kernel_timestamp; + m->firmware_timestamp = firmware_timestamp; + m->loader_timestamp = loader_timestamp; + m->initrd_timestamp = initrd_timestamp; + + manager_set_default_rlimits(m, arg_default_rlimit); + + if (arg_default_controllers) + manager_set_default_controllers(m, arg_default_controllers); + + manager_set_show_status(m, arg_show_status); + + /* Remember whether we should queue the default job */ + queue_default_job = !serialization || arg_switched_root; + + before_startup = now(CLOCK_MONOTONIC); + + r = manager_startup(m, serialization, fds); + if (r < 0) + log_error("Failed to fully start up daemon: %s", strerror(-r)); + + /* This will close all file descriptors that were opened, but + * not claimed by any unit. */ + fdset_free(fds); + + if (serialization) { + fclose(serialization); + serialization = NULL; + } + + if (queue_default_job) { + DBusError error; + Unit *target = NULL; + Job *default_unit_job; + + dbus_error_init(&error); + + log_debug("Activating default unit: %s", arg_default_unit); + + r = manager_load_unit(m, arg_default_unit, NULL, &error, &target); + if (r < 0) { + log_error("Failed to load default target: %s", bus_error(&error, r)); + dbus_error_free(&error); + } else if (target->load_state == UNIT_ERROR) + log_error("Failed to load default target: %s", strerror(-target->load_error)); + else if (target->load_state == UNIT_MASKED) + log_error("Default target masked."); + + if (!target || target->load_state != UNIT_LOADED) { + log_info("Trying to load rescue target..."); + + r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target); + if (r < 0) { + log_error("Failed to load rescue target: %s", bus_error(&error, r)); + dbus_error_free(&error); + goto finish; + } else if (target->load_state == UNIT_ERROR) { + log_error("Failed to load rescue target: %s", strerror(-target->load_error)); + goto finish; + } else if (target->load_state == UNIT_MASKED) { + log_error("Rescue target masked."); + goto finish; + } + } + + assert(target->load_state == UNIT_LOADED); + + if (arg_action == ACTION_TEST) { + printf("-> By units:\n"); + manager_dump_units(m, stdout, "\t"); + } + + r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job); + if (r == -EPERM) { + log_debug("Default target could not be isolated, starting instead: %s", bus_error(&error, r)); + dbus_error_free(&error); + + r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job); + if (r < 0) { + log_error("Failed to start default target: %s", bus_error(&error, r)); + dbus_error_free(&error); + goto finish; + } + } else if (r < 0) { + log_error("Failed to isolate default target: %s", bus_error(&error, r)); + dbus_error_free(&error); + goto finish; + } + + m->default_unit_job_id = default_unit_job->id; + + after_startup = now(CLOCK_MONOTONIC); + log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG, + "Loaded units and determined initial transaction in %s.", + format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 0)); + + if (arg_action == ACTION_TEST) { + printf("-> By jobs:\n"); + manager_dump_jobs(m, stdout, "\t"); + retval = EXIT_SUCCESS; + goto finish; + } + } + + for (;;) { + r = manager_loop(m); + if (r < 0) { + log_error("Failed to run mainloop: %s", strerror(-r)); + goto finish; + } + + switch (m->exit_code) { + + case MANAGER_EXIT: + retval = EXIT_SUCCESS; + log_debug("Exit."); + goto finish; + + case MANAGER_RELOAD: + log_info("Reloading."); + r = manager_reload(m); + if (r < 0) + log_error("Failed to reload: %s", strerror(-r)); + break; + + case MANAGER_REEXECUTE: + + if (prepare_reexecute(m, &serialization, &fds, false) < 0) + goto finish; + + reexecute = true; + log_notice("Reexecuting."); + goto finish; + + case MANAGER_SWITCH_ROOT: + /* Steal the switch root parameters */ + switch_root_dir = m->switch_root; + switch_root_init = m->switch_root_init; + m->switch_root = m->switch_root_init = NULL; + + if (!switch_root_init) + if (prepare_reexecute(m, &serialization, &fds, true) < 0) + goto finish; + + reexecute = true; + log_notice("Switching root."); + goto finish; + + case MANAGER_REBOOT: + case MANAGER_POWEROFF: + case MANAGER_HALT: + case MANAGER_KEXEC: { + static const char * const table[_MANAGER_EXIT_CODE_MAX] = { + [MANAGER_REBOOT] = "reboot", + [MANAGER_POWEROFF] = "poweroff", + [MANAGER_HALT] = "halt", + [MANAGER_KEXEC] = "kexec" + }; + + assert_se(shutdown_verb = table[m->exit_code]); + arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT; + + log_notice("Shutting down."); + goto finish; + } + + default: + assert_not_reached("Unknown exit code."); + } + } + +finish: + if (m) + manager_free(m); + + for (j = 0; j < RLIMIT_NLIMITS; j++) + free(arg_default_rlimit[j]); + + free(arg_default_unit); + strv_free(arg_default_controllers); + free_join_controllers(); + + dbus_shutdown(); + label_finish(); + + if (reexecute) { + const char **args; + unsigned i, args_size; + + /* Close and disarm the watchdog, so that the new + * instance can reinitialize it, but doesn't get + * rebooted while we do that */ + watchdog_close(true); + + /* Reset the RLIMIT_NOFILE to the kernel default, so + * that the new systemd can pass the kernel default to + * its child processes */ + if (saved_rlimit_nofile.rlim_cur > 0) + setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile); + + if (switch_root_dir) { + /* Kill all remaining processes from the + * initrd, but don't wait for them, so that we + * can handle the SIGCHLD for them after + * deserializing. */ + broadcast_signal(SIGTERM, false); + + /* And switch root */ + r = switch_root(switch_root_dir); + if (r < 0) + log_error("Failed to switch root, ignoring: %s", strerror(-r)); + } + + args_size = MAX(6, argc+1); + args = newa(const char*, args_size); + + if (!switch_root_init) { + char sfd[16]; + + /* First try to spawn ourselves with the right + * path, and with full serialization. We do + * this only if the user didn't specify an + * explicit init to spawn. */ + + assert(serialization); + assert(fds); + + snprintf(sfd, sizeof(sfd), "%i", fileno(serialization)); + char_array_0(sfd); + + i = 0; + args[i++] = SYSTEMD_BINARY_PATH; + if (switch_root_dir) + args[i++] = "--switched-root"; + args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user"; + args[i++] = "--deserialize"; + args[i++] = sfd; + args[i++] = NULL; + + assert(i <= args_size); + execv(args[0], (char* const*) args); + } + + /* Try the fallback, if there is any, without any + * serialization. We pass the original argv[] and + * envp[]. (Well, modulo the ordering changes due to + * getopt() in argv[], and some cleanups in envp[], + * but let's hope that doesn't matter.) */ + + if (serialization) { + fclose(serialization); + serialization = NULL; + } + + if (fds) { + fdset_free(fds); + fds = NULL; + } + + /* Reopen the console */ + make_console_stdio(); + + for (j = 1, i = 1; j < argc; j++) + args[i++] = argv[j]; + args[i++] = NULL; + assert(i <= args_size); + + if (switch_root_init) { + args[0] = switch_root_init; + execv(args[0], (char* const*) args); + log_warning("Failed to execute configured init, trying fallback: %m"); + } + + args[0] = "/sbin/init"; + execv(args[0], (char* const*) args); + + if (errno == ENOENT) { + log_warning("No /sbin/init, trying fallback"); + + args[0] = "/bin/sh"; + args[1] = NULL; + execv(args[0], (char* const*) args); + log_error("Failed to execute /bin/sh, giving up: %m"); + } else + log_warning("Failed to execute /sbin/init, giving up: %m"); + } + + if (serialization) + fclose(serialization); + + if (fds) + fdset_free(fds); + + if (shutdown_verb) { + const char * command_line[] = { + SYSTEMD_SHUTDOWN_BINARY_PATH, + shutdown_verb, + NULL + }; + char **env_block; + + if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) { + char e[32]; + + /* If we reboot let's set the shutdown + * watchdog and tell the shutdown binary to + * repeatedly ping it */ + watchdog_set_timeout(&arg_shutdown_watchdog); + watchdog_close(false); + + /* Tell the binary how often to ping */ + snprintf(e, sizeof(e), "WATCHDOG_USEC=%llu", (unsigned long long) arg_shutdown_watchdog); + char_array_0(e); + + env_block = strv_append(environ, e); + } else { + env_block = strv_copy(environ); + watchdog_close(true); + } + + execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block); + free(env_block); + log_error("Failed to execute shutdown binary, freezing: %m"); + } + + if (getpid() == 1) + freeze(); + + return retval; +} diff --git a/src/core/manager.c b/src/core/manager.c new file mode 100644 index 0000000000..c7f8f20a70 --- /dev/null +++ b/src/core/manager.c @@ -0,0 +1,2674 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <assert.h> +#include <errno.h> +#include <string.h> +#include <sys/epoll.h> +#include <signal.h> +#include <sys/signalfd.h> +#include <sys/wait.h> +#include <unistd.h> +#include <sys/poll.h> +#include <sys/reboot.h> +#include <sys/ioctl.h> +#include <linux/kd.h> +#include <termios.h> +#include <fcntl.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <dirent.h> +#include <sys/timerfd.h> + +#ifdef HAVE_AUDIT +#include <libaudit.h> +#endif + +#include "systemd/sd-daemon.h" +#include "systemd/sd-id128.h" +#include "systemd/sd-messages.h" + +#include "manager.h" +#include "transaction.h" +#include "hashmap.h" +#include "macro.h" +#include "strv.h" +#include "log.h" +#include "util.h" +#include "mkdir.h" +#include "ratelimit.h" +#include "cgroup.h" +#include "mount-setup.h" +#include "unit-name.h" +#include "dbus-unit.h" +#include "dbus-job.h" +#include "missing.h" +#include "path-lookup.h" +#include "special.h" +#include "bus-errors.h" +#include "exit-status.h" +#include "virt.h" +#include "watchdog.h" +#include "cgroup-util.h" +#include "path-util.h" +#include "audit-fd.h" +#include "env-util.h" + +/* As soon as 16 units are in our GC queue, make sure to run a gc sweep */ +#define GC_QUEUE_ENTRIES_MAX 16 + +/* As soon as 5s passed since a unit was added to our GC queue, make sure to run a gc sweep */ +#define GC_QUEUE_USEC_MAX (10*USEC_PER_SEC) + +/* Initial delay and the interval for printing status messages about running jobs */ +#define JOBS_IN_PROGRESS_WAIT_SEC 5 +#define JOBS_IN_PROGRESS_PERIOD_SEC 1 +#define JOBS_IN_PROGRESS_PERIOD_DIVISOR 3 + +/* Where clients shall send notification messages to */ +#define NOTIFY_SOCKET "@/org/freedesktop/systemd1/notify" + +#define TIME_T_MAX (time_t)((1UL << ((sizeof(time_t) << 3) - 1)) - 1) + +static int manager_setup_notify(Manager *m) { + union { + struct sockaddr sa; + struct sockaddr_un un; + } sa = { + .sa.sa_family = AF_UNIX, + }; + struct epoll_event ev = { + .events = EPOLLIN, + .data.ptr = &m->notify_watch, + }; + int one = 1, r; + + m->notify_watch.type = WATCH_NOTIFY; + m->notify_watch.fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); + if (m->notify_watch.fd < 0) { + log_error("Failed to allocate notification socket: %m"); + return -errno; + } + + if (getpid() != 1 || detect_container(NULL) > 0) + snprintf(sa.un.sun_path, sizeof(sa.un.sun_path), NOTIFY_SOCKET "/%llu", random_ull()); + else + strncpy(sa.un.sun_path, NOTIFY_SOCKET, sizeof(sa.un.sun_path)); + + sa.un.sun_path[0] = 0; + + r = bind(m->notify_watch.fd, &sa.sa, + offsetof(struct sockaddr_un, sun_path) + 1 + strlen(sa.un.sun_path+1)); + if (r < 0) { + log_error("bind() failed: %m"); + return -errno; + } + + r = setsockopt(m->notify_watch.fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one)); + if (r < 0) { + log_error("SO_PASSCRED failed: %m"); + return -errno; + } + + r = epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, m->notify_watch.fd, &ev); + if (r < 0) { + log_error("Failed to add notification socket fd to epoll: %m"); + return -errno; + } + + sa.un.sun_path[0] = '@'; + m->notify_socket = strdup(sa.un.sun_path); + if (!m->notify_socket) + return log_oom(); + + log_debug("Using notification socket %s", m->notify_socket); + + return 0; +} + +static int manager_jobs_in_progress_mod_timer(Manager *m) { + struct itimerspec its = { + .it_value.tv_sec = JOBS_IN_PROGRESS_WAIT_SEC, + .it_interval.tv_sec = JOBS_IN_PROGRESS_PERIOD_SEC, + }; + + if (m->jobs_in_progress_watch.type != WATCH_JOBS_IN_PROGRESS) + return 0; + + if (timerfd_settime(m->jobs_in_progress_watch.fd, 0, &its, NULL) < 0) + return -errno; + + return 0; +} + +static int manager_watch_jobs_in_progress(Manager *m) { + struct epoll_event ev = { + .events = EPOLLIN, + .data.ptr = &m->jobs_in_progress_watch, + }; + int r; + + if (m->jobs_in_progress_watch.type != WATCH_INVALID) + return 0; + + m->jobs_in_progress_watch.type = WATCH_JOBS_IN_PROGRESS; + m->jobs_in_progress_watch.fd = timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK|TFD_CLOEXEC); + if (m->jobs_in_progress_watch.fd < 0) { + log_error("Failed to create timerfd: %m"); + r = -errno; + goto err; + } + + r = manager_jobs_in_progress_mod_timer(m); + if (r < 0) { + log_error("Failed to set up timer for jobs progress watch: %s", strerror(-r)); + goto err; + } + + if (epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, m->jobs_in_progress_watch.fd, &ev) < 0) { + log_error("Failed to add jobs progress timer fd to epoll: %m"); + r = -errno; + goto err; + } + + log_debug("Set up jobs progress timerfd."); + + return 0; + +err: + if (m->jobs_in_progress_watch.fd >= 0) + close_nointr_nofail(m->jobs_in_progress_watch.fd); + watch_init(&m->jobs_in_progress_watch); + return r; +} + +static void manager_unwatch_jobs_in_progress(Manager *m) { + if (m->jobs_in_progress_watch.type != WATCH_JOBS_IN_PROGRESS) + return; + + assert_se(epoll_ctl(m->epoll_fd, EPOLL_CTL_DEL, m->jobs_in_progress_watch.fd, NULL) >= 0); + close_nointr_nofail(m->jobs_in_progress_watch.fd); + watch_init(&m->jobs_in_progress_watch); + m->jobs_in_progress_iteration = 0; + + log_debug("Closed jobs progress timerfd."); +} + +#define CYLON_BUFFER_EXTRA (2*strlen(ANSI_RED_ON) + strlen(ANSI_HIGHLIGHT_RED_ON) + 2*strlen(ANSI_HIGHLIGHT_OFF)) +static void draw_cylon(char buffer[], size_t buflen, unsigned width, unsigned pos) { + char *p = buffer; + + assert(buflen >= CYLON_BUFFER_EXTRA + width + 1); + assert(pos <= width+1); /* 0 or width+1 mean that the center light is behind the corner */ + + if (pos > 1) { + if (pos > 2) + p = mempset(p, ' ', pos-2); + p = stpcpy(p, ANSI_RED_ON); + *p++ = '*'; + } + + if (pos > 0 && pos <= width) { + p = stpcpy(p, ANSI_HIGHLIGHT_RED_ON); + *p++ = '*'; + } + + p = stpcpy(p, ANSI_HIGHLIGHT_OFF); + + if (pos < width) { + p = stpcpy(p, ANSI_RED_ON); + *p++ = '*'; + if (pos < width-1) + p = mempset(p, ' ', width-1-pos); + p = stpcpy(p, ANSI_HIGHLIGHT_OFF); + } +} + +static void manager_print_jobs_in_progress(Manager *m) { + Iterator i; + Job *j; + char *job_of_n = NULL; + unsigned counter = 0, print_nr; + char cylon[6 + CYLON_BUFFER_EXTRA + 1]; + unsigned cylon_pos; + + print_nr = (m->jobs_in_progress_iteration / JOBS_IN_PROGRESS_PERIOD_DIVISOR) % m->n_running_jobs; + + HASHMAP_FOREACH(j, m->jobs, i) + if (j->state == JOB_RUNNING && counter++ == print_nr) + break; + + /* m->n_running_jobs must be consistent with the contents of m->jobs, + * so the above loop must have succeeded in finding j. */ + assert(counter == print_nr + 1); + + cylon_pos = m->jobs_in_progress_iteration % 14; + if (cylon_pos >= 8) + cylon_pos = 14 - cylon_pos; + draw_cylon(cylon, sizeof(cylon), 6, cylon_pos); + + if (m->n_running_jobs > 1) + if (asprintf(&job_of_n, "(%u of %u) ", counter, m->n_running_jobs) < 0) + job_of_n = NULL; + + manager_status_printf(m, true, cylon, "%sA %s job is running for %s", + strempty(job_of_n), job_type_to_string(j->type), unit_description(j->unit)); + free(job_of_n); + + m->jobs_in_progress_iteration++; +} + +static int manager_setup_time_change(Manager *m) { + struct epoll_event ev = { + .events = EPOLLIN, + .data.ptr = &m->time_change_watch, + }; + + /* We only care for the cancellation event, hence we set the + * timeout to the latest possible value. */ + struct itimerspec its = { + .it_value.tv_sec = TIME_T_MAX, + }; + assert_cc(sizeof(time_t) == sizeof(TIME_T_MAX)); + + assert(m->time_change_watch.type == WATCH_INVALID); + + /* Uses TFD_TIMER_CANCEL_ON_SET to get notifications whenever + * CLOCK_REALTIME makes a jump relative to CLOCK_MONOTONIC */ + + m->time_change_watch.type = WATCH_TIME_CHANGE; + m->time_change_watch.fd = timerfd_create(CLOCK_REALTIME, TFD_NONBLOCK|TFD_CLOEXEC); + if (m->time_change_watch.fd < 0) { + log_error("Failed to create timerfd: %m"); + return -errno; + } + + if (timerfd_settime(m->time_change_watch.fd, TFD_TIMER_ABSTIME|TFD_TIMER_CANCEL_ON_SET, &its, NULL) < 0) { + log_debug("Failed to set up TFD_TIMER_CANCEL_ON_SET, ignoring: %m"); + close_nointr_nofail(m->time_change_watch.fd); + watch_init(&m->time_change_watch); + return 0; + } + + if (epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, m->time_change_watch.fd, &ev) < 0) { + log_error("Failed to add timer change fd to epoll: %m"); + return -errno; + } + + log_debug("Set up TFD_TIMER_CANCEL_ON_SET timerfd."); + + return 0; +} + +static int enable_special_signals(Manager *m) { + int fd; + + assert(m); + + /* Enable that we get SIGINT on control-alt-del. In containers + * this will fail with EPERM (older) or EINVAL (newer), so + * ignore that. */ + if (reboot(RB_DISABLE_CAD) < 0 && errno != EPERM && errno != EINVAL) + log_warning("Failed to enable ctrl-alt-del handling: %m"); + + fd = open_terminal("/dev/tty0", O_RDWR|O_NOCTTY|O_CLOEXEC); + if (fd < 0) { + /* Support systems without virtual console */ + if (fd != -ENOENT) + log_warning("Failed to open /dev/tty0: %m"); + } else { + /* Enable that we get SIGWINCH on kbrequest */ + if (ioctl(fd, KDSIGACCEPT, SIGWINCH) < 0) + log_warning("Failed to enable kbrequest handling: %s", strerror(errno)); + + close_nointr_nofail(fd); + } + + return 0; +} + +static int manager_setup_signals(Manager *m) { + sigset_t mask; + struct epoll_event ev = { + .events = EPOLLIN, + .data.ptr = &m->signal_watch, + }; + struct sigaction sa = { + .sa_handler = SIG_DFL, + .sa_flags = SA_NOCLDSTOP|SA_RESTART, + }; + + assert(m); + + /* We are not interested in SIGSTOP and friends. */ + assert_se(sigaction(SIGCHLD, &sa, NULL) == 0); + + assert_se(sigemptyset(&mask) == 0); + + sigset_add_many(&mask, + SIGCHLD, /* Child died */ + SIGTERM, /* Reexecute daemon */ + SIGHUP, /* Reload configuration */ + SIGUSR1, /* systemd/upstart: reconnect to D-Bus */ + SIGUSR2, /* systemd: dump status */ + SIGINT, /* Kernel sends us this on control-alt-del */ + SIGWINCH, /* Kernel sends us this on kbrequest (alt-arrowup) */ + SIGPWR, /* Some kernel drivers and upsd send us this on power failure */ + SIGRTMIN+0, /* systemd: start default.target */ + SIGRTMIN+1, /* systemd: isolate rescue.target */ + SIGRTMIN+2, /* systemd: isolate emergency.target */ + SIGRTMIN+3, /* systemd: start halt.target */ + SIGRTMIN+4, /* systemd: start poweroff.target */ + SIGRTMIN+5, /* systemd: start reboot.target */ + SIGRTMIN+6, /* systemd: start kexec.target */ + SIGRTMIN+13, /* systemd: Immediate halt */ + SIGRTMIN+14, /* systemd: Immediate poweroff */ + SIGRTMIN+15, /* systemd: Immediate reboot */ + SIGRTMIN+16, /* systemd: Immediate kexec */ + SIGRTMIN+20, /* systemd: enable status messages */ + SIGRTMIN+21, /* systemd: disable status messages */ + SIGRTMIN+22, /* systemd: set log level to LOG_DEBUG */ + SIGRTMIN+23, /* systemd: set log level to LOG_INFO */ + SIGRTMIN+24, /* systemd: Immediate exit (--user only) */ + SIGRTMIN+26, /* systemd: set log target to journal-or-kmsg */ + SIGRTMIN+27, /* systemd: set log target to console */ + SIGRTMIN+28, /* systemd: set log target to kmsg */ + SIGRTMIN+29, /* systemd: set log target to syslog-or-kmsg */ + -1); + assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); + + m->signal_watch.type = WATCH_SIGNAL; + m->signal_watch.fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC); + if (m->signal_watch.fd < 0) + return -errno; + + if (epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, m->signal_watch.fd, &ev) < 0) + return -errno; + + if (m->running_as == SYSTEMD_SYSTEM) + return enable_special_signals(m); + + return 0; +} + +static void manager_strip_environment(Manager *m) { + assert(m); + + /* Remove variables from the inherited set that are part of + * the container interface: + * http://www.freedesktop.org/wiki/Software/systemd/ContainerInterface */ + strv_remove_prefix(m->environment, "container="); + strv_remove_prefix(m->environment, "container_"); + + /* Remove variables from the inherited set that are part of + * the initrd interface: + * http://www.freedesktop.org/wiki/Software/systemd/InitrdInterface */ + strv_remove_prefix(m->environment, "RD_"); + + /* Drop invalid entries */ + strv_env_clean(m->environment); +} + +int manager_new(SystemdRunningAs running_as, Manager **_m) { + Manager *m; + int r = -ENOMEM; + + assert(_m); + assert(running_as >= 0); + assert(running_as < _SYSTEMD_RUNNING_AS_MAX); + + m = new0(Manager, 1); + if (!m) + return -ENOMEM; + + m->running_as = running_as; + m->name_data_slot = m->conn_data_slot = m->subscribed_data_slot = -1; + m->exit_code = _MANAGER_EXIT_CODE_INVALID; + m->pin_cgroupfs_fd = -1; + m->idle_pipe[0] = m->idle_pipe[1] = -1; + + watch_init(&m->signal_watch); + watch_init(&m->mount_watch); + watch_init(&m->swap_watch); + watch_init(&m->udev_watch); + watch_init(&m->time_change_watch); + watch_init(&m->jobs_in_progress_watch); + + m->epoll_fd = m->dev_autofs_fd = -1; + m->current_job_id = 1; /* start as id #1, so that we can leave #0 around as "null-like" value */ + + m->environment = strv_copy(environ); + if (!m->environment) + goto fail; + + manager_strip_environment(m); + + if (running_as == SYSTEMD_SYSTEM) { + m->default_controllers = strv_new("cpu", NULL); + if (!m->default_controllers) + goto fail; + } + + if (!(m->units = hashmap_new(string_hash_func, string_compare_func))) + goto fail; + + if (!(m->jobs = hashmap_new(trivial_hash_func, trivial_compare_func))) + goto fail; + + if (!(m->watch_pids = hashmap_new(trivial_hash_func, trivial_compare_func))) + goto fail; + + if (!(m->cgroup_bondings = hashmap_new(string_hash_func, string_compare_func))) + goto fail; + + if (!(m->watch_bus = hashmap_new(string_hash_func, string_compare_func))) + goto fail; + + m->epoll_fd = epoll_create1(EPOLL_CLOEXEC); + if (m->epoll_fd < 0) + goto fail; + + r = manager_setup_signals(m); + if (r < 0) + goto fail; + + r = manager_setup_cgroup(m); + if (r < 0) + goto fail; + + r = manager_setup_notify(m); + if (r < 0) + goto fail; + + r = manager_setup_time_change(m); + if (r < 0) + goto fail; + + /* Try to connect to the busses, if possible. */ + r = bus_init(m, running_as != SYSTEMD_SYSTEM); + if (r < 0) + goto fail; + + m->taint_usr = dir_is_empty("/usr") > 0; + + *_m = m; + return 0; + +fail: + manager_free(m); + return r; +} + +static unsigned manager_dispatch_cleanup_queue(Manager *m) { + Unit *u; + unsigned n = 0; + + assert(m); + + while ((u = m->cleanup_queue)) { + assert(u->in_cleanup_queue); + + unit_free(u); + n++; + } + + return n; +} + +enum { + GC_OFFSET_IN_PATH, /* This one is on the path we were traveling */ + GC_OFFSET_UNSURE, /* No clue */ + GC_OFFSET_GOOD, /* We still need this unit */ + GC_OFFSET_BAD, /* We don't need this unit anymore */ + _GC_OFFSET_MAX +}; + +static void unit_gc_sweep(Unit *u, unsigned gc_marker) { + Iterator i; + Unit *other; + bool is_bad; + + assert(u); + + if (u->gc_marker == gc_marker + GC_OFFSET_GOOD || + u->gc_marker == gc_marker + GC_OFFSET_BAD || + u->gc_marker == gc_marker + GC_OFFSET_IN_PATH) + return; + + if (u->in_cleanup_queue) + goto bad; + + if (unit_check_gc(u)) + goto good; + + u->gc_marker = gc_marker + GC_OFFSET_IN_PATH; + + is_bad = true; + + SET_FOREACH(other, u->dependencies[UNIT_REFERENCED_BY], i) { + unit_gc_sweep(other, gc_marker); + + if (other->gc_marker == gc_marker + GC_OFFSET_GOOD) + goto good; + + if (other->gc_marker != gc_marker + GC_OFFSET_BAD) + is_bad = false; + } + + if (is_bad) + goto bad; + + /* We were unable to find anything out about this entry, so + * let's investigate it later */ + u->gc_marker = gc_marker + GC_OFFSET_UNSURE; + unit_add_to_gc_queue(u); + return; + +bad: + /* We definitely know that this one is not useful anymore, so + * let's mark it for deletion */ + u->gc_marker = gc_marker + GC_OFFSET_BAD; + unit_add_to_cleanup_queue(u); + return; + +good: + u->gc_marker = gc_marker + GC_OFFSET_GOOD; +} + +static unsigned manager_dispatch_gc_queue(Manager *m) { + Unit *u; + unsigned n = 0; + unsigned gc_marker; + + assert(m); + + if ((m->n_in_gc_queue < GC_QUEUE_ENTRIES_MAX) && + (m->gc_queue_timestamp <= 0 || + (m->gc_queue_timestamp + GC_QUEUE_USEC_MAX) > now(CLOCK_MONOTONIC))) + return 0; + + log_debug("Running GC..."); + + m->gc_marker += _GC_OFFSET_MAX; + if (m->gc_marker + _GC_OFFSET_MAX <= _GC_OFFSET_MAX) + m->gc_marker = 1; + + gc_marker = m->gc_marker; + + while ((u = m->gc_queue)) { + assert(u->in_gc_queue); + + unit_gc_sweep(u, gc_marker); + + LIST_REMOVE(Unit, gc_queue, m->gc_queue, u); + u->in_gc_queue = false; + + n++; + + if (u->gc_marker == gc_marker + GC_OFFSET_BAD || + u->gc_marker == gc_marker + GC_OFFSET_UNSURE) { + log_debug_unit(u->id, "Collecting %s", u->id); + u->gc_marker = gc_marker + GC_OFFSET_BAD; + unit_add_to_cleanup_queue(u); + } + } + + m->n_in_gc_queue = 0; + m->gc_queue_timestamp = 0; + + return n; +} + +static void manager_clear_jobs_and_units(Manager *m) { + Unit *u; + + assert(m); + + while ((u = hashmap_first(m->units))) + unit_free(u); + + manager_dispatch_cleanup_queue(m); + + assert(!m->load_queue); + assert(!m->run_queue); + assert(!m->dbus_unit_queue); + assert(!m->dbus_job_queue); + assert(!m->cleanup_queue); + assert(!m->gc_queue); + + assert(hashmap_isempty(m->jobs)); + assert(hashmap_isempty(m->units)); + + m->n_on_console = 0; + m->n_running_jobs = 0; +} + +void manager_free(Manager *m) { + UnitType c; + int i; + + assert(m); + + manager_clear_jobs_and_units(m); + + for (c = 0; c < _UNIT_TYPE_MAX; c++) + if (unit_vtable[c]->shutdown) + unit_vtable[c]->shutdown(m); + + /* If we reexecute ourselves, we keep the root cgroup + * around */ + manager_shutdown_cgroup(m, m->exit_code != MANAGER_REEXECUTE); + + manager_undo_generators(m); + + bus_done(m); + + hashmap_free(m->units); + hashmap_free(m->jobs); + hashmap_free(m->watch_pids); + hashmap_free(m->watch_bus); + + if (m->epoll_fd >= 0) + close_nointr_nofail(m->epoll_fd); + if (m->signal_watch.fd >= 0) + close_nointr_nofail(m->signal_watch.fd); + if (m->notify_watch.fd >= 0) + close_nointr_nofail(m->notify_watch.fd); + if (m->time_change_watch.fd >= 0) + close_nointr_nofail(m->time_change_watch.fd); + if (m->jobs_in_progress_watch.fd >= 0) + close_nointr_nofail(m->jobs_in_progress_watch.fd); + + free(m->notify_socket); + + lookup_paths_free(&m->lookup_paths); + strv_free(m->environment); + + strv_free(m->default_controllers); + + hashmap_free(m->cgroup_bondings); + set_free_free(m->unit_path_cache); + + close_pipe(m->idle_pipe); + + free(m->switch_root); + free(m->switch_root_init); + + for (i = 0; i < RLIMIT_NLIMITS; i++) + free(m->rlimit[i]); + + free(m); +} + +int manager_enumerate(Manager *m) { + int r = 0, q; + UnitType c; + + assert(m); + + /* Let's ask every type to load all units from disk/kernel + * that it might know */ + for (c = 0; c < _UNIT_TYPE_MAX; c++) + if (unit_vtable[c]->enumerate) + if ((q = unit_vtable[c]->enumerate(m)) < 0) + r = q; + + manager_dispatch_load_queue(m); + return r; +} + +int manager_coldplug(Manager *m) { + int r = 0, q; + Iterator i; + Unit *u; + char *k; + + assert(m); + + /* Then, let's set up their initial state. */ + HASHMAP_FOREACH_KEY(u, k, m->units, i) { + + /* ignore aliases */ + if (u->id != k) + continue; + + if ((q = unit_coldplug(u)) < 0) + r = q; + } + + return r; +} + +static void manager_build_unit_path_cache(Manager *m) { + char **i; + _cleanup_free_ DIR *d = NULL; + int r; + + assert(m); + + set_free_free(m->unit_path_cache); + + m->unit_path_cache = set_new(string_hash_func, string_compare_func); + if (!m->unit_path_cache) { + log_error("Failed to allocate unit path cache."); + return; + } + + /* This simply builds a list of files we know exist, so that + * we don't always have to go to disk */ + + STRV_FOREACH(i, m->lookup_paths.unit_path) { + struct dirent *de; + + d = opendir(*i); + if (!d) { + if (errno != ENOENT) + log_error("Failed to open directory %s: %m", *i); + continue; + } + + while ((de = readdir(d))) { + char *p; + + if (ignore_file(de->d_name)) + continue; + + p = strjoin(streq(*i, "/") ? "" : *i, "/", de->d_name, NULL); + if (!p) { + r = -ENOMEM; + goto fail; + } + + r = set_consume(m->unit_path_cache, p); + if (r < 0) + goto fail; + } + + closedir(d); + d = NULL; + } + + return; + +fail: + log_error("Failed to build unit path cache: %s", strerror(-r)); + + set_free_free(m->unit_path_cache); + m->unit_path_cache = NULL; +} + +int manager_startup(Manager *m, FILE *serialization, FDSet *fds) { + int r, q; + + assert(m); + + manager_run_generators(m); + + r = lookup_paths_init( + &m->lookup_paths, m->running_as, true, + m->generator_unit_path, + m->generator_unit_path_early, + m->generator_unit_path_late); + if (r < 0) + return r; + + manager_build_unit_path_cache(m); + + /* If we will deserialize make sure that during enumeration + * this is already known, so we increase the counter here + * already */ + if (serialization) + m->n_reloading ++; + + /* First, enumerate what we can from all config files */ + r = manager_enumerate(m); + + /* Second, deserialize if there is something to deserialize */ + if (serialization) { + q = manager_deserialize(m, serialization, fds); + if (q < 0) + r = q; + } + + /* Any fds left? Find some unit which wants them. This is + * useful to allow container managers to pass some file + * descriptors to us pre-initialized. This enables + * socket-based activation of entire containers. */ + if (fdset_size(fds) > 0) { + q = manager_distribute_fds(m, fds); + if (q < 0) + r = q; + } + + /* Third, fire things up! */ + q = manager_coldplug(m); + if (q < 0) + r = q; + + if (serialization) { + assert(m->n_reloading > 0); + m->n_reloading --; + } + + return r; +} + +int manager_add_job(Manager *m, JobType type, Unit *unit, JobMode mode, bool override, DBusError *e, Job **_ret) { + int r; + Transaction *tr; + + assert(m); + assert(type < _JOB_TYPE_MAX); + assert(unit); + assert(mode < _JOB_MODE_MAX); + + if (mode == JOB_ISOLATE && type != JOB_START) { + dbus_set_error(e, BUS_ERROR_INVALID_JOB_MODE, "Isolate is only valid for start."); + return -EINVAL; + } + + if (mode == JOB_ISOLATE && !unit->allow_isolate) { + dbus_set_error(e, BUS_ERROR_NO_ISOLATION, "Operation refused, unit may not be isolated."); + return -EPERM; + } + + log_debug_unit(unit->id, + "Trying to enqueue job %s/%s/%s", unit->id, + job_type_to_string(type), job_mode_to_string(mode)); + + job_type_collapse(&type, unit); + + tr = transaction_new(mode == JOB_REPLACE_IRREVERSIBLY); + if (!tr) + return -ENOMEM; + + r = transaction_add_job_and_dependencies(tr, type, unit, NULL, true, override, false, + mode == JOB_IGNORE_DEPENDENCIES || mode == JOB_IGNORE_REQUIREMENTS, + mode == JOB_IGNORE_DEPENDENCIES, e); + if (r < 0) + goto tr_abort; + + if (mode == JOB_ISOLATE) { + r = transaction_add_isolate_jobs(tr, m); + if (r < 0) + goto tr_abort; + } + + r = transaction_activate(tr, m, mode, e); + if (r < 0) + goto tr_abort; + + log_debug_unit(unit->id, + "Enqueued job %s/%s as %u", unit->id, + job_type_to_string(type), (unsigned) tr->anchor_job->id); + + if (_ret) + *_ret = tr->anchor_job; + + transaction_free(tr); + return 0; + +tr_abort: + transaction_abort(tr); + transaction_free(tr); + return r; +} + +int manager_add_job_by_name(Manager *m, JobType type, const char *name, JobMode mode, bool override, DBusError *e, Job **_ret) { + Unit *unit; + int r; + + assert(m); + assert(type < _JOB_TYPE_MAX); + assert(name); + assert(mode < _JOB_MODE_MAX); + + r = manager_load_unit(m, name, NULL, NULL, &unit); + if (r < 0) + return r; + + return manager_add_job(m, type, unit, mode, override, e, _ret); +} + +Job *manager_get_job(Manager *m, uint32_t id) { + assert(m); + + return hashmap_get(m->jobs, UINT32_TO_PTR(id)); +} + +Unit *manager_get_unit(Manager *m, const char *name) { + assert(m); + assert(name); + + return hashmap_get(m->units, name); +} + +unsigned manager_dispatch_load_queue(Manager *m) { + Unit *u; + unsigned n = 0; + + assert(m); + + /* Make sure we are not run recursively */ + if (m->dispatching_load_queue) + return 0; + + m->dispatching_load_queue = true; + + /* Dispatches the load queue. Takes a unit from the queue and + * tries to load its data until the queue is empty */ + + while ((u = m->load_queue)) { + assert(u->in_load_queue); + + unit_load(u); + n++; + } + + m->dispatching_load_queue = false; + return n; +} + +int manager_load_unit_prepare(Manager *m, const char *name, const char *path, DBusError *e, Unit **_ret) { + Unit *ret; + UnitType t; + int r; + + assert(m); + assert(name || path); + + /* This will prepare the unit for loading, but not actually + * load anything from disk. */ + + if (path && !is_path(path)) { + dbus_set_error(e, BUS_ERROR_INVALID_PATH, "Path %s is not absolute.", path); + return -EINVAL; + } + + if (!name) + name = path_get_file_name(path); + + t = unit_name_to_type(name); + + if (t == _UNIT_TYPE_INVALID || !unit_name_is_valid(name, false)) { + dbus_set_error(e, BUS_ERROR_INVALID_NAME, "Unit name %s is not valid.", name); + return -EINVAL; + } + + ret = manager_get_unit(m, name); + if (ret) { + *_ret = ret; + return 1; + } + + ret = unit_new(m, unit_vtable[t]->object_size); + if (!ret) + return -ENOMEM; + + if (path) { + ret->fragment_path = strdup(path); + if (!ret->fragment_path) { + unit_free(ret); + return -ENOMEM; + } + } + + if ((r = unit_add_name(ret, name)) < 0) { + unit_free(ret); + return r; + } + + unit_add_to_load_queue(ret); + unit_add_to_dbus_queue(ret); + unit_add_to_gc_queue(ret); + + if (_ret) + *_ret = ret; + + return 0; +} + +int manager_load_unit(Manager *m, const char *name, const char *path, DBusError *e, Unit **_ret) { + int r; + + assert(m); + + /* This will load the service information files, but not actually + * start any services or anything. */ + + r = manager_load_unit_prepare(m, name, path, e, _ret); + if (r != 0) + return r; + + manager_dispatch_load_queue(m); + + if (_ret) + *_ret = unit_follow_merge(*_ret); + + return 0; +} + +void manager_dump_jobs(Manager *s, FILE *f, const char *prefix) { + Iterator i; + Job *j; + + assert(s); + assert(f); + + HASHMAP_FOREACH(j, s->jobs, i) + job_dump(j, f, prefix); +} + +void manager_dump_units(Manager *s, FILE *f, const char *prefix) { + Iterator i; + Unit *u; + const char *t; + + assert(s); + assert(f); + + HASHMAP_FOREACH_KEY(u, t, s->units, i) + if (u->id == t) + unit_dump(u, f, prefix); +} + +void manager_clear_jobs(Manager *m) { + Job *j; + + assert(m); + + while ((j = hashmap_first(m->jobs))) + /* No need to recurse. We're cancelling all jobs. */ + job_finish_and_invalidate(j, JOB_CANCELED, false); +} + +unsigned manager_dispatch_run_queue(Manager *m) { + Job *j; + unsigned n = 0; + + if (m->dispatching_run_queue) + return 0; + + m->dispatching_run_queue = true; + + while ((j = m->run_queue)) { + assert(j->installed); + assert(j->in_run_queue); + + job_run_and_invalidate(j); + n++; + } + + m->dispatching_run_queue = false; + + if (m->n_running_jobs > 0) + manager_watch_jobs_in_progress(m); + + return n; +} + +unsigned manager_dispatch_dbus_queue(Manager *m) { + Job *j; + Unit *u; + unsigned n = 0; + + assert(m); + + if (m->dispatching_dbus_queue) + return 0; + + m->dispatching_dbus_queue = true; + + while ((u = m->dbus_unit_queue)) { + assert(u->in_dbus_queue); + + bus_unit_send_change_signal(u); + n++; + } + + while ((j = m->dbus_job_queue)) { + assert(j->in_dbus_queue); + + bus_job_send_change_signal(j); + n++; + } + + m->dispatching_dbus_queue = false; + return n; +} + +static int manager_process_notify_fd(Manager *m) { + ssize_t n; + + assert(m); + + for (;;) { + char buf[4096]; + struct iovec iovec = { + .iov_base = buf, + .iov_len = sizeof(buf)-1, + }; + + union { + struct cmsghdr cmsghdr; + uint8_t buf[CMSG_SPACE(sizeof(struct ucred))]; + } control = {}; + + struct msghdr msghdr = { + .msg_iov = &iovec, + .msg_iovlen = 1, + .msg_control = &control, + .msg_controllen = sizeof(control), + }; + struct ucred *ucred; + Unit *u; + _cleanup_strv_free_ char **tags = NULL; + + n = recvmsg(m->notify_watch.fd, &msghdr, MSG_DONTWAIT); + if (n <= 0) { + if (n == 0) + return -EIO; + + if (errno == EAGAIN || errno == EINTR) + break; + + return -errno; + } + + if (msghdr.msg_controllen < CMSG_LEN(sizeof(struct ucred)) || + control.cmsghdr.cmsg_level != SOL_SOCKET || + control.cmsghdr.cmsg_type != SCM_CREDENTIALS || + control.cmsghdr.cmsg_len != CMSG_LEN(sizeof(struct ucred))) { + log_warning("Received notify message without credentials. Ignoring."); + continue; + } + + ucred = (struct ucred*) CMSG_DATA(&control.cmsghdr); + + u = hashmap_get(m->watch_pids, LONG_TO_PTR(ucred->pid)); + if (!u) { + u = cgroup_unit_by_pid(m, ucred->pid); + if (!u) { + log_warning("Cannot find unit for notify message of PID %lu.", (unsigned long) ucred->pid); + continue; + } + } + + assert((size_t) n < sizeof(buf)); + buf[n] = 0; + tags = strv_split(buf, "\n\r"); + if (!tags) + return log_oom(); + + log_debug_unit(u->id, "Got notification message for unit %s", u->id); + + if (UNIT_VTABLE(u)->notify_message) + UNIT_VTABLE(u)->notify_message(u, ucred->pid, tags); + } + + return 0; +} + +static int manager_dispatch_sigchld(Manager *m) { + assert(m); + + for (;;) { + siginfo_t si = {}; + Unit *u; + int r; + + /* First we call waitd() for a PID and do not reap the + * zombie. That way we can still access /proc/$PID for + * it while it is a zombie. */ + if (waitid(P_ALL, 0, &si, WEXITED|WNOHANG|WNOWAIT) < 0) { + + if (errno == ECHILD) + break; + + if (errno == EINTR) + continue; + + return -errno; + } + + if (si.si_pid <= 0) + break; + + if (si.si_code == CLD_EXITED || si.si_code == CLD_KILLED || si.si_code == CLD_DUMPED) { + _cleanup_free_ char *name = NULL; + + get_process_comm(si.si_pid, &name); + log_debug("Got SIGCHLD for process %lu (%s)", (unsigned long) si.si_pid, strna(name)); + } + + /* Let's flush any message the dying child might still + * have queued for us. This ensures that the process + * still exists in /proc so that we can figure out + * which cgroup and hence unit it belongs to. */ + r = manager_process_notify_fd(m); + if (r < 0) + return r; + + /* And now figure out the unit this belongs to */ + u = hashmap_get(m->watch_pids, LONG_TO_PTR(si.si_pid)); + if (!u) + u = cgroup_unit_by_pid(m, si.si_pid); + + /* And now, we actually reap the zombie. */ + if (waitid(P_PID, si.si_pid, &si, WEXITED) < 0) { + if (errno == EINTR) + continue; + + return -errno; + } + + if (si.si_code != CLD_EXITED && si.si_code != CLD_KILLED && si.si_code != CLD_DUMPED) + continue; + + log_debug("Child %lu died (code=%s, status=%i/%s)", + (long unsigned) si.si_pid, + sigchld_code_to_string(si.si_code), + si.si_status, + strna(si.si_code == CLD_EXITED + ? exit_status_to_string(si.si_status, EXIT_STATUS_FULL) + : signal_to_string(si.si_status))); + + if (!u) + continue; + + log_debug_unit(u->id, + "Child %lu belongs to %s", (long unsigned) si.si_pid, u->id); + + hashmap_remove(m->watch_pids, LONG_TO_PTR(si.si_pid)); + UNIT_VTABLE(u)->sigchld_event(u, si.si_pid, si.si_code, si.si_status); + } + + return 0; +} + +static int manager_start_target(Manager *m, const char *name, JobMode mode) { + int r; + DBusError error; + + dbus_error_init(&error); + + log_debug_unit(name, "Activating special unit %s", name); + + r = manager_add_job_by_name(m, JOB_START, name, mode, true, &error, NULL); + if (r < 0) + log_error_unit(name, + "Failed to enqueue %s job: %s", name, bus_error(&error, r)); + + dbus_error_free(&error); + + return r; +} + +static int manager_process_signal_fd(Manager *m) { + ssize_t n; + struct signalfd_siginfo sfsi; + bool sigchld = false; + + assert(m); + + for (;;) { + n = read(m->signal_watch.fd, &sfsi, sizeof(sfsi)); + if (n != sizeof(sfsi)) { + + if (n >= 0) + return -EIO; + + if (errno == EINTR || errno == EAGAIN) + break; + + return -errno; + } + + if (sfsi.ssi_pid > 0) { + char *p = NULL; + + get_process_comm(sfsi.ssi_pid, &p); + + log_debug("Received SIG%s from PID %lu (%s).", + signal_to_string(sfsi.ssi_signo), + (unsigned long) sfsi.ssi_pid, strna(p)); + free(p); + } else + log_debug("Received SIG%s.", signal_to_string(sfsi.ssi_signo)); + + switch (sfsi.ssi_signo) { + + case SIGCHLD: + sigchld = true; + break; + + case SIGTERM: + if (m->running_as == SYSTEMD_SYSTEM) { + /* This is for compatibility with the + * original sysvinit */ + m->exit_code = MANAGER_REEXECUTE; + break; + } + + /* Fall through */ + + case SIGINT: + if (m->running_as == SYSTEMD_SYSTEM) { + manager_start_target(m, SPECIAL_CTRL_ALT_DEL_TARGET, JOB_REPLACE); + break; + } + + /* Run the exit target if there is one, if not, just exit. */ + if (manager_start_target(m, SPECIAL_EXIT_TARGET, JOB_REPLACE) < 0) { + m->exit_code = MANAGER_EXIT; + return 0; + } + + break; + + case SIGWINCH: + if (m->running_as == SYSTEMD_SYSTEM) + manager_start_target(m, SPECIAL_KBREQUEST_TARGET, JOB_REPLACE); + + /* This is a nop on non-init */ + break; + + case SIGPWR: + if (m->running_as == SYSTEMD_SYSTEM) + manager_start_target(m, SPECIAL_SIGPWR_TARGET, JOB_REPLACE); + + /* This is a nop on non-init */ + break; + + case SIGUSR1: { + Unit *u; + + u = manager_get_unit(m, SPECIAL_DBUS_SERVICE); + + if (!u || UNIT_IS_ACTIVE_OR_RELOADING(unit_active_state(u))) { + log_info("Trying to reconnect to bus..."); + bus_init(m, true); + } + + if (!u || !UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u))) { + log_info("Loading D-Bus service..."); + manager_start_target(m, SPECIAL_DBUS_SERVICE, JOB_REPLACE); + } + + break; + } + + case SIGUSR2: { + FILE *f; + char *dump = NULL; + size_t size; + + if (!(f = open_memstream(&dump, &size))) { + log_warning("Failed to allocate memory stream."); + break; + } + + manager_dump_units(m, f, "\t"); + manager_dump_jobs(m, f, "\t"); + + if (ferror(f)) { + fclose(f); + free(dump); + log_warning("Failed to write status stream"); + break; + } + + fclose(f); + log_dump(LOG_INFO, dump); + free(dump); + + break; + } + + case SIGHUP: + m->exit_code = MANAGER_RELOAD; + break; + + default: { + + /* Starting SIGRTMIN+0 */ + static const char * const target_table[] = { + [0] = SPECIAL_DEFAULT_TARGET, + [1] = SPECIAL_RESCUE_TARGET, + [2] = SPECIAL_EMERGENCY_TARGET, + [3] = SPECIAL_HALT_TARGET, + [4] = SPECIAL_POWEROFF_TARGET, + [5] = SPECIAL_REBOOT_TARGET, + [6] = SPECIAL_KEXEC_TARGET + }; + + /* Starting SIGRTMIN+13, so that target halt and system halt are 10 apart */ + static const ManagerExitCode code_table[] = { + [0] = MANAGER_HALT, + [1] = MANAGER_POWEROFF, + [2] = MANAGER_REBOOT, + [3] = MANAGER_KEXEC + }; + + if ((int) sfsi.ssi_signo >= SIGRTMIN+0 && + (int) sfsi.ssi_signo < SIGRTMIN+(int) ELEMENTSOF(target_table)) { + int idx = (int) sfsi.ssi_signo - SIGRTMIN; + manager_start_target(m, target_table[idx], + (idx == 1 || idx == 2) ? JOB_ISOLATE : JOB_REPLACE); + break; + } + + if ((int) sfsi.ssi_signo >= SIGRTMIN+13 && + (int) sfsi.ssi_signo < SIGRTMIN+13+(int) ELEMENTSOF(code_table)) { + m->exit_code = code_table[sfsi.ssi_signo - SIGRTMIN - 13]; + break; + } + + switch (sfsi.ssi_signo - SIGRTMIN) { + + case 20: + log_debug("Enabling showing of status."); + manager_set_show_status(m, true); + break; + + case 21: + log_debug("Disabling showing of status."); + manager_set_show_status(m, false); + break; + + case 22: + log_set_max_level(LOG_DEBUG); + log_notice("Setting log level to debug."); + break; + + case 23: + log_set_max_level(LOG_INFO); + log_notice("Setting log level to info."); + break; + + case 24: + if (m->running_as == SYSTEMD_USER) { + m->exit_code = MANAGER_EXIT; + return 0; + } + + /* This is a nop on init */ + break; + + case 26: + log_set_target(LOG_TARGET_JOURNAL_OR_KMSG); + log_notice("Setting log target to journal-or-kmsg."); + break; + + case 27: + log_set_target(LOG_TARGET_CONSOLE); + log_notice("Setting log target to console."); + break; + + case 28: + log_set_target(LOG_TARGET_KMSG); + log_notice("Setting log target to kmsg."); + break; + + case 29: + log_set_target(LOG_TARGET_SYSLOG_OR_KMSG); + log_notice("Setting log target to syslog-or-kmsg."); + break; + + default: + log_warning("Got unhandled signal <%s>.", signal_to_string(sfsi.ssi_signo)); + } + } + } + } + + if (sigchld) + return manager_dispatch_sigchld(m); + + return 0; +} + +static int process_event(Manager *m, struct epoll_event *ev) { + int r; + Watch *w; + + assert(m); + assert(ev); + + assert_se(w = ev->data.ptr); + + if (w->type == WATCH_INVALID) + return 0; + + switch (w->type) { + + case WATCH_SIGNAL: + + /* An incoming signal? */ + if (ev->events != EPOLLIN) + return -EINVAL; + + if ((r = manager_process_signal_fd(m)) < 0) + return r; + + break; + + case WATCH_NOTIFY: + + /* An incoming daemon notification event? */ + if (ev->events != EPOLLIN) + return -EINVAL; + + if ((r = manager_process_notify_fd(m)) < 0) + return r; + + break; + + case WATCH_FD: + + /* Some fd event, to be dispatched to the units */ + UNIT_VTABLE(w->data.unit)->fd_event(w->data.unit, w->fd, ev->events, w); + break; + + case WATCH_UNIT_TIMER: + case WATCH_JOB_TIMER: { + uint64_t v; + ssize_t k; + + /* Some timer event, to be dispatched to the units */ + k = read(w->fd, &v, sizeof(v)); + if (k != sizeof(v)) { + + if (k < 0 && (errno == EINTR || errno == EAGAIN)) + break; + + log_error("Failed to read timer event counter: %s", k < 0 ? strerror(-k) : "Short read"); + return k < 0 ? -errno : -EIO; + } + + if (w->type == WATCH_UNIT_TIMER) + UNIT_VTABLE(w->data.unit)->timer_event(w->data.unit, v, w); + else + job_timer_event(w->data.job, v, w); + break; + } + + case WATCH_MOUNT: + /* Some mount table change, intended for the mount subsystem */ + mount_fd_event(m, ev->events); + break; + + case WATCH_SWAP: + /* Some swap table change, intended for the swap subsystem */ + swap_fd_event(m, ev->events); + break; + + case WATCH_UDEV: + /* Some notification from udev, intended for the device subsystem */ + device_fd_event(m, ev->events); + break; + + case WATCH_DBUS_WATCH: + bus_watch_event(m, w, ev->events); + break; + + case WATCH_DBUS_TIMEOUT: + bus_timeout_event(m, w, ev->events); + break; + + case WATCH_TIME_CHANGE: { + Unit *u; + Iterator i; + + log_struct(LOG_INFO, + MESSAGE_ID(SD_MESSAGE_TIME_CHANGE), + "MESSAGE=Time has been changed", + NULL); + + /* Restart the watch */ + epoll_ctl(m->epoll_fd, EPOLL_CTL_DEL, m->time_change_watch.fd, + NULL); + close_nointr_nofail(m->time_change_watch.fd); + watch_init(&m->time_change_watch); + manager_setup_time_change(m); + + HASHMAP_FOREACH(u, m->units, i) { + if (UNIT_VTABLE(u)->time_change) + UNIT_VTABLE(u)->time_change(u); + } + + break; + } + + case WATCH_JOBS_IN_PROGRESS: { + uint64_t v; + + /* not interested in the data */ + read(w->fd, &v, sizeof(v)); + + manager_print_jobs_in_progress(m); + break; + } + + default: + log_error("event type=%i", w->type); + assert_not_reached("Unknown epoll event type."); + } + + return 0; +} + +int manager_loop(Manager *m) { + int r; + + RATELIMIT_DEFINE(rl, 1*USEC_PER_SEC, 50000); + + assert(m); + m->exit_code = MANAGER_RUNNING; + + /* Release the path cache */ + set_free_free(m->unit_path_cache); + m->unit_path_cache = NULL; + + manager_check_finished(m); + + /* There might still be some zombies hanging around from + * before we were exec()'ed. Leat's reap them */ + r = manager_dispatch_sigchld(m); + if (r < 0) + return r; + + while (m->exit_code == MANAGER_RUNNING) { + struct epoll_event event; + int n; + int wait_msec = -1; + + if (m->runtime_watchdog > 0 && m->running_as == SYSTEMD_SYSTEM) + watchdog_ping(); + + if (!ratelimit_test(&rl)) { + /* Yay, something is going seriously wrong, pause a little */ + log_warning("Looping too fast. Throttling execution a little."); + sleep(1); + continue; + } + + if (manager_dispatch_load_queue(m) > 0) + continue; + + if (manager_dispatch_run_queue(m) > 0) + continue; + + if (bus_dispatch(m) > 0) + continue; + + if (manager_dispatch_cleanup_queue(m) > 0) + continue; + + if (manager_dispatch_gc_queue(m) > 0) + continue; + + if (manager_dispatch_dbus_queue(m) > 0) + continue; + + if (swap_dispatch_reload(m) > 0) + continue; + + /* Sleep for half the watchdog time */ + if (m->runtime_watchdog > 0 && m->running_as == SYSTEMD_SYSTEM) { + wait_msec = (int) (m->runtime_watchdog / 2 / USEC_PER_MSEC); + if (wait_msec <= 0) + wait_msec = 1; + } else + wait_msec = -1; + + n = epoll_wait(m->epoll_fd, &event, 1, wait_msec); + if (n < 0) { + + if (errno == EINTR) + continue; + + return -errno; + } else if (n == 0) + continue; + + assert(n == 1); + + r = process_event(m, &event); + if (r < 0) + return r; + } + + return m->exit_code; +} + +int manager_load_unit_from_dbus_path(Manager *m, const char *s, DBusError *e, Unit **_u) { + char *n; + Unit *u; + int r; + + assert(m); + assert(s); + assert(_u); + + if (!startswith(s, "/org/freedesktop/systemd1/unit/")) + return -EINVAL; + + n = bus_path_unescape(s+31); + if (!n) + return -ENOMEM; + + r = manager_load_unit(m, n, NULL, e, &u); + free(n); + + if (r < 0) + return r; + + *_u = u; + + return 0; +} + +int manager_get_job_from_dbus_path(Manager *m, const char *s, Job **_j) { + Job *j; + unsigned id; + int r; + + assert(m); + assert(s); + assert(_j); + + if (!startswith(s, "/org/freedesktop/systemd1/job/")) + return -EINVAL; + + r = safe_atou(s + 30, &id); + if (r < 0) + return r; + + j = manager_get_job(m, id); + if (!j) + return -ENOENT; + + *_j = j; + + return 0; +} + +void manager_send_unit_audit(Manager *m, Unit *u, int type, bool success) { + +#ifdef HAVE_AUDIT + char *p; + int audit_fd; + + audit_fd = get_audit_fd(); + if (audit_fd < 0) + return; + + /* Don't generate audit events if the service was already + * started and we're just deserializing */ + if (m->n_reloading > 0) + return; + + if (m->running_as != SYSTEMD_SYSTEM) + return; + + if (u->type != UNIT_SERVICE) + return; + + p = unit_name_to_prefix_and_instance(u->id); + if (!p) { + log_error_unit(u->id, + "Failed to allocate unit name for audit message: %s", strerror(ENOMEM)); + return; + } + + if (audit_log_user_comm_message(audit_fd, type, "", p, NULL, NULL, NULL, success) < 0) { + if (errno == EPERM) { + /* We aren't allowed to send audit messages? + * Then let's not retry again. */ + close_audit_fd(); + } else + log_warning("Failed to send audit message: %m"); + } + + free(p); +#endif + +} + +void manager_send_unit_plymouth(Manager *m, Unit *u) { + int fd = -1; + union sockaddr_union sa; + int n = 0; + char *message = NULL; + + /* Don't generate plymouth events if the service was already + * started and we're just deserializing */ + if (m->n_reloading > 0) + return; + + if (m->running_as != SYSTEMD_SYSTEM) + return; + + if (u->type != UNIT_SERVICE && + u->type != UNIT_MOUNT && + u->type != UNIT_SWAP) + return; + + /* We set SOCK_NONBLOCK here so that we rather drop the + * message then wait for plymouth */ + fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); + if (fd < 0) { + log_error("socket() failed: %m"); + return; + } + + zero(sa); + sa.sa.sa_family = AF_UNIX; + strncpy(sa.un.sun_path+1, "/org/freedesktop/plymouthd", sizeof(sa.un.sun_path)-1); + if (connect(fd, &sa.sa, offsetof(struct sockaddr_un, sun_path) + 1 + strlen(sa.un.sun_path+1)) < 0) { + + if (errno != EPIPE && + errno != EAGAIN && + errno != ENOENT && + errno != ECONNREFUSED && + errno != ECONNRESET && + errno != ECONNABORTED) + log_error("connect() failed: %m"); + + goto finish; + } + + if (asprintf(&message, "U\002%c%s%n", (int) (strlen(u->id) + 1), u->id, &n) < 0) { + log_oom(); + goto finish; + } + + errno = 0; + if (write(fd, message, n + 1) != n + 1) { + + if (errno != EPIPE && + errno != EAGAIN && + errno != ENOENT && + errno != ECONNREFUSED && + errno != ECONNRESET && + errno != ECONNABORTED) + log_error("Failed to write Plymouth message: %m"); + + goto finish; + } + +finish: + if (fd >= 0) + close_nointr_nofail(fd); + + free(message); +} + +void manager_dispatch_bus_name_owner_changed( + Manager *m, + const char *name, + const char* old_owner, + const char *new_owner) { + + Unit *u; + + assert(m); + assert(name); + + if (!(u = hashmap_get(m->watch_bus, name))) + return; + + UNIT_VTABLE(u)->bus_name_owner_change(u, name, old_owner, new_owner); +} + +void manager_dispatch_bus_query_pid_done( + Manager *m, + const char *name, + pid_t pid) { + + Unit *u; + + assert(m); + assert(name); + assert(pid >= 1); + + if (!(u = hashmap_get(m->watch_bus, name))) + return; + + UNIT_VTABLE(u)->bus_query_pid_done(u, name, pid); +} + +int manager_open_serialization(Manager *m, FILE **_f) { + char *path = NULL; + int fd; + FILE *f; + + assert(_f); + + if (m->running_as == SYSTEMD_SYSTEM) + asprintf(&path, "/run/systemd/dump-%lu-XXXXXX", (unsigned long) getpid()); + else + asprintf(&path, "/tmp/systemd-dump-%lu-XXXXXX", (unsigned long) getpid()); + + if (!path) + return -ENOMEM; + + RUN_WITH_UMASK(0077) { + fd = mkostemp(path, O_RDWR|O_CLOEXEC); + } + + if (fd < 0) { + free(path); + return -errno; + } + + unlink(path); + + log_debug("Serializing state to %s", path); + free(path); + + f = fdopen(fd, "w+"); + if (!f) + return -errno; + + *_f = f; + + return 0; +} + +int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root) { + Iterator i; + Unit *u; + const char *t; + char **e; + int r; + + assert(m); + assert(f); + assert(fds); + + m->n_reloading ++; + + fprintf(f, "current-job-id=%i\n", m->current_job_id); + fprintf(f, "taint-usr=%s\n", yes_no(m->taint_usr)); + fprintf(f, "n-installed-jobs=%u\n", m->n_installed_jobs); + fprintf(f, "n-failed-jobs=%u\n", m->n_failed_jobs); + + dual_timestamp_serialize(f, "firmware-timestamp", &m->firmware_timestamp); + dual_timestamp_serialize(f, "kernel-timestamp", &m->kernel_timestamp); + dual_timestamp_serialize(f, "loader-timestamp", &m->loader_timestamp); + dual_timestamp_serialize(f, "initrd-timestamp", &m->initrd_timestamp); + + if (!in_initrd()) { + dual_timestamp_serialize(f, "userspace-timestamp", &m->userspace_timestamp); + dual_timestamp_serialize(f, "finish-timestamp", &m->finish_timestamp); + } + + if (!switching_root) { + STRV_FOREACH(e, m->environment) { + _cleanup_free_ char *ce; + + ce = cescape(*e); + if (ce) + fprintf(f, "env=%s\n", *e); + } + } + + fputc('\n', f); + + HASHMAP_FOREACH_KEY(u, t, m->units, i) { + if (u->id != t) + continue; + + if (!unit_can_serialize(u)) + continue; + + /* Start marker */ + fputs(u->id, f); + fputc('\n', f); + + if ((r = unit_serialize(u, f, fds, !switching_root)) < 0) { + m->n_reloading --; + return r; + } + } + + assert(m->n_reloading > 0); + m->n_reloading --; + + if (ferror(f)) + return -EIO; + + r = bus_fdset_add_all(m, fds); + if (r < 0) + return r; + + return 0; +} + +int manager_deserialize(Manager *m, FILE *f, FDSet *fds) { + int r = 0; + + assert(m); + assert(f); + + log_debug("Deserializing state..."); + + m->n_reloading ++; + + for (;;) { + char line[LINE_MAX], *l; + + if (!fgets(line, sizeof(line), f)) { + if (feof(f)) + r = 0; + else + r = -errno; + + goto finish; + } + + char_array_0(line); + l = strstrip(line); + + if (l[0] == 0) + break; + + if (startswith(l, "current-job-id=")) { + uint32_t id; + + if (safe_atou32(l+15, &id) < 0) + log_debug("Failed to parse current job id value %s", l+15); + else + m->current_job_id = MAX(m->current_job_id, id); + } else if (startswith(l, "n-installed-jobs=")) { + uint32_t n; + + if (safe_atou32(l+17, &n) < 0) + log_debug("Failed to parse installed jobs counter %s", l+17); + else + m->n_installed_jobs += n; + } else if (startswith(l, "n-failed-jobs=")) { + uint32_t n; + + if (safe_atou32(l+14, &n) < 0) + log_debug("Failed to parse failed jobs counter %s", l+14); + else + m->n_failed_jobs += n; + } else if (startswith(l, "taint-usr=")) { + int b; + + if ((b = parse_boolean(l+10)) < 0) + log_debug("Failed to parse taint /usr flag %s", l+10); + else + m->taint_usr = m->taint_usr || b; + } else if (startswith(l, "firmware-timestamp=")) + dual_timestamp_deserialize(l+19, &m->firmware_timestamp); + else if (startswith(l, "loader-timestamp=")) + dual_timestamp_deserialize(l+17, &m->loader_timestamp); + else if (startswith(l, "kernel-timestamp=")) + dual_timestamp_deserialize(l+17, &m->kernel_timestamp); + else if (startswith(l, "initrd-timestamp=")) + dual_timestamp_deserialize(l+17, &m->initrd_timestamp); + else if (startswith(l, "userspace-timestamp=")) + dual_timestamp_deserialize(l+20, &m->userspace_timestamp); + else if (startswith(l, "finish-timestamp=")) + dual_timestamp_deserialize(l+17, &m->finish_timestamp); + else if (startswith(l, "env=")) { + _cleanup_free_ char *uce = NULL; + char **e; + + uce = cunescape(l+4); + if (!uce) { + r = -ENOMEM; + goto finish; + } + + e = strv_env_set(m->environment, uce); + if (!e) { + r = -ENOMEM; + goto finish; + } + + strv_free(m->environment); + m->environment = e; + } else + log_debug("Unknown serialization item '%s'", l); + } + + for (;;) { + Unit *u; + char name[UNIT_NAME_MAX+2]; + + /* Start marker */ + if (!fgets(name, sizeof(name), f)) { + if (feof(f)) + r = 0; + else + r = -errno; + + goto finish; + } + + char_array_0(name); + + r = manager_load_unit(m, strstrip(name), NULL, NULL, &u); + if (r < 0) + goto finish; + + r = unit_deserialize(u, f, fds); + if (r < 0) + goto finish; + } + +finish: + if (ferror(f)) { + r = -EIO; + goto finish; + } + + assert(m->n_reloading > 0); + m->n_reloading --; + + return r; +} + +int manager_distribute_fds(Manager *m, FDSet *fds) { + Unit *u; + Iterator i; + int r; + + assert(m); + + HASHMAP_FOREACH(u, m->units, i) { + + if (fdset_size(fds) <= 0) + break; + + if (UNIT_VTABLE(u)->distribute_fds) { + r = UNIT_VTABLE(u)->distribute_fds(u, fds); + if (r < 0) + return r; + } + } + + return 0; +} + +int manager_reload(Manager *m) { + int r, q; + FILE *f; + FDSet *fds; + + assert(m); + + r = manager_open_serialization(m, &f); + if (r < 0) + return r; + + m->n_reloading ++; + + fds = fdset_new(); + if (!fds) { + m->n_reloading --; + r = -ENOMEM; + goto finish; + } + + r = manager_serialize(m, f, fds, false); + if (r < 0) { + m->n_reloading --; + goto finish; + } + + if (fseeko(f, 0, SEEK_SET) < 0) { + m->n_reloading --; + r = -errno; + goto finish; + } + + /* From here on there is no way back. */ + manager_clear_jobs_and_units(m); + manager_undo_generators(m); + lookup_paths_free(&m->lookup_paths); + + /* Find new unit paths */ + manager_run_generators(m); + + q = lookup_paths_init( + &m->lookup_paths, m->running_as, true, + m->generator_unit_path, + m->generator_unit_path_early, + m->generator_unit_path_late); + if (q < 0) + r = q; + + manager_build_unit_path_cache(m); + + /* First, enumerate what we can from all config files */ + q = manager_enumerate(m); + if (q < 0) + r = q; + + /* Second, deserialize our stored data */ + q = manager_deserialize(m, f, fds); + if (q < 0) + r = q; + + fclose(f); + f = NULL; + + /* Third, fire things up! */ + q = manager_coldplug(m); + if (q < 0) + r = q; + + assert(m->n_reloading > 0); + m->n_reloading--; + +finish: + if (f) + fclose(f); + + if (fds) + fdset_free(fds); + + return r; +} + +static bool manager_is_booting_or_shutting_down(Manager *m) { + Unit *u; + + assert(m); + + /* Is the initial job still around? */ + if (manager_get_job(m, m->default_unit_job_id)) + return true; + + /* Is there a job for the shutdown target? */ + u = manager_get_unit(m, SPECIAL_SHUTDOWN_TARGET); + if (u) + return !!u->job; + + return false; +} + +bool manager_is_reloading_or_reexecuting(Manager *m) { + assert(m); + + return m->n_reloading != 0; +} + +void manager_reset_failed(Manager *m) { + Unit *u; + Iterator i; + + assert(m); + + HASHMAP_FOREACH(u, m->units, i) + unit_reset_failed(u); +} + +bool manager_unit_inactive_or_pending(Manager *m, const char *name) { + Unit *u; + + assert(m); + assert(name); + + /* Returns true if the unit is inactive or going down */ + u = manager_get_unit(m, name); + if (!u) + return true; + + return unit_inactive_or_pending(u); +} + +void manager_check_finished(Manager *m) { + char userspace[FORMAT_TIMESPAN_MAX], initrd[FORMAT_TIMESPAN_MAX], kernel[FORMAT_TIMESPAN_MAX], sum[FORMAT_TIMESPAN_MAX]; + usec_t firmware_usec, loader_usec, kernel_usec, initrd_usec, userspace_usec, total_usec; + + assert(m); + + if (m->n_running_jobs == 0) + manager_unwatch_jobs_in_progress(m); + + if (hashmap_size(m->jobs) > 0) { + manager_jobs_in_progress_mod_timer(m); + return; + } + + /* Notify Type=idle units that we are done now */ + close_pipe(m->idle_pipe); + + /* Turn off confirm spawn now */ + m->confirm_spawn = false; + + if (dual_timestamp_is_set(&m->finish_timestamp)) + return; + + dual_timestamp_get(&m->finish_timestamp); + + if (m->running_as == SYSTEMD_SYSTEM && detect_container(NULL) <= 0) { + + /* Note that m->kernel_usec.monotonic is always at 0, + * and m->firmware_usec.monotonic and + * m->loader_usec.monotonic should be considered + * negative values. */ + + firmware_usec = m->firmware_timestamp.monotonic - m->loader_timestamp.monotonic; + loader_usec = m->loader_timestamp.monotonic - m->kernel_timestamp.monotonic; + userspace_usec = m->finish_timestamp.monotonic - m->userspace_timestamp.monotonic; + total_usec = m->firmware_timestamp.monotonic + m->finish_timestamp.monotonic; + + if (dual_timestamp_is_set(&m->initrd_timestamp)) { + + kernel_usec = m->initrd_timestamp.monotonic - m->kernel_timestamp.monotonic; + initrd_usec = m->userspace_timestamp.monotonic - m->initrd_timestamp.monotonic; + + if (!log_on_console()) + log_struct(LOG_INFO, + MESSAGE_ID(SD_MESSAGE_STARTUP_FINISHED), + "KERNEL_USEC=%llu", (unsigned long long) kernel_usec, + "INITRD_USEC=%llu", (unsigned long long) initrd_usec, + "USERSPACE_USEC=%llu", (unsigned long long) userspace_usec, + "MESSAGE=Startup finished in %s (kernel) + %s (initrd) + %s (userspace) = %s.", + format_timespan(kernel, sizeof(kernel), kernel_usec, USEC_PER_MSEC), + format_timespan(initrd, sizeof(initrd), initrd_usec, USEC_PER_MSEC), + format_timespan(userspace, sizeof(userspace), userspace_usec, USEC_PER_MSEC), + format_timespan(sum, sizeof(sum), total_usec, USEC_PER_MSEC), + NULL); + } else { + kernel_usec = m->userspace_timestamp.monotonic - m->kernel_timestamp.monotonic; + initrd_usec = 0; + + if (!log_on_console()) + log_struct(LOG_INFO, + MESSAGE_ID(SD_MESSAGE_STARTUP_FINISHED), + "KERNEL_USEC=%llu", (unsigned long long) kernel_usec, + "USERSPACE_USEC=%llu", (unsigned long long) userspace_usec, + "MESSAGE=Startup finished in %s (kernel) + %s (userspace) = %s.", + format_timespan(kernel, sizeof(kernel), kernel_usec, USEC_PER_MSEC), + format_timespan(userspace, sizeof(userspace), userspace_usec, USEC_PER_MSEC), + format_timespan(sum, sizeof(sum), total_usec, USEC_PER_MSEC), + NULL); + } + } else { + firmware_usec = loader_usec = initrd_usec = kernel_usec = 0; + total_usec = userspace_usec = m->finish_timestamp.monotonic - m->userspace_timestamp.monotonic; + + if (!log_on_console()) + log_struct(LOG_INFO, + MESSAGE_ID(SD_MESSAGE_STARTUP_FINISHED), + "USERSPACE_USEC=%llu", (unsigned long long) userspace_usec, + "MESSAGE=Startup finished in %s.", + format_timespan(sum, sizeof(sum), total_usec, USEC_PER_MSEC), + NULL); + } + + bus_broadcast_finished(m, firmware_usec, loader_usec, kernel_usec, initrd_usec, userspace_usec, total_usec); + + sd_notifyf(false, + "READY=1\nSTATUS=Startup finished in %s.", + format_timespan(sum, sizeof(sum), total_usec, USEC_PER_MSEC)); +} + +static int create_generator_dir(Manager *m, char **generator, const char *name) { + char *p; + int r; + + assert(m); + assert(generator); + assert(name); + + if (*generator) + return 0; + + if (m->running_as == SYSTEMD_SYSTEM && getpid() == 1) { + + p = strappend("/run/systemd/", name); + if (!p) + return log_oom(); + + r = mkdir_p_label(p, 0755); + if (r < 0) { + log_error("Failed to create generator directory %s: %s", + p, strerror(-r)); + free(p); + return r; + } + } else { + p = strjoin("/tmp/systemd-", name, ".XXXXXX", NULL); + if (!p) + return log_oom(); + + if (!mkdtemp(p)) { + log_error("Failed to create generator directory %s: %m", + p); + free(p); + return -errno; + } + } + + *generator = p; + return 0; +} + +static void trim_generator_dir(Manager *m, char **generator) { + assert(m); + assert(generator); + + if (!*generator) + return; + + if (rmdir(*generator) >= 0) { + free(*generator); + *generator = NULL; + } + + return; +} + +void manager_run_generators(Manager *m) { + DIR *d = NULL; + const char *generator_path; + const char *argv[5]; + int r; + + assert(m); + + generator_path = m->running_as == SYSTEMD_SYSTEM ? SYSTEM_GENERATOR_PATH : USER_GENERATOR_PATH; + d = opendir(generator_path); + if (!d) { + if (errno == ENOENT) + return; + + log_error("Failed to enumerate generator directory %s: %m", + generator_path); + return; + } + + r = create_generator_dir(m, &m->generator_unit_path, "generator"); + if (r < 0) + goto finish; + + r = create_generator_dir(m, &m->generator_unit_path_early, "generator.early"); + if (r < 0) + goto finish; + + r = create_generator_dir(m, &m->generator_unit_path_late, "generator.late"); + if (r < 0) + goto finish; + + argv[0] = NULL; /* Leave this empty, execute_directory() will fill something in */ + argv[1] = m->generator_unit_path; + argv[2] = m->generator_unit_path_early; + argv[3] = m->generator_unit_path_late; + argv[4] = NULL; + + RUN_WITH_UMASK(0022) { + execute_directory(generator_path, d, (char**) argv); + } + + trim_generator_dir(m, &m->generator_unit_path); + trim_generator_dir(m, &m->generator_unit_path_early); + trim_generator_dir(m, &m->generator_unit_path_late); + +finish: + if (d) + closedir(d); +} + +static void remove_generator_dir(Manager *m, char **generator) { + assert(m); + assert(generator); + + if (!*generator) + return; + + strv_remove(m->lookup_paths.unit_path, *generator); + rm_rf(*generator, false, true, false); + + free(*generator); + *generator = NULL; +} + +void manager_undo_generators(Manager *m) { + assert(m); + + remove_generator_dir(m, &m->generator_unit_path); + remove_generator_dir(m, &m->generator_unit_path_early); + remove_generator_dir(m, &m->generator_unit_path_late); +} + +int manager_set_default_controllers(Manager *m, char **controllers) { + char **l; + + assert(m); + + l = strv_copy(controllers); + if (!l) + return -ENOMEM; + + strv_free(m->default_controllers); + m->default_controllers = l; + + cg_shorten_controllers(m->default_controllers); + + return 0; +} + +int manager_set_default_rlimits(Manager *m, struct rlimit **default_rlimit) { + int i; + + assert(m); + + for (i = 0; i < RLIMIT_NLIMITS; i++) { + if (!default_rlimit[i]) + continue; + + m->rlimit[i] = newdup(struct rlimit, default_rlimit[i], 1); + if (!m->rlimit[i]) + return -ENOMEM; + } + + return 0; +} + +void manager_recheck_journal(Manager *m) { + Unit *u; + + assert(m); + + if (m->running_as != SYSTEMD_SYSTEM) + return; + + u = manager_get_unit(m, SPECIAL_JOURNALD_SOCKET); + if (u && SOCKET(u)->state != SOCKET_RUNNING) { + log_close_journal(); + return; + } + + u = manager_get_unit(m, SPECIAL_JOURNALD_SERVICE); + if (u && SERVICE(u)->state != SERVICE_RUNNING) { + log_close_journal(); + return; + } + + /* Hmm, OK, so the socket is fully up and the service is up + * too, then let's make use of the thing. */ + log_open(); +} + +void manager_set_show_status(Manager *m, bool b) { + assert(m); + + if (m->running_as != SYSTEMD_SYSTEM) + return; + + m->show_status = b; + + if (b) + touch("/run/systemd/show-status"); + else + unlink("/run/systemd/show-status"); +} + +static bool manager_get_show_status(Manager *m) { + assert(m); + + if (m->running_as != SYSTEMD_SYSTEM) + return false; + + if (m->show_status) + return true; + + /* If Plymouth is running make sure we show the status, so + * that there's something nice to see when people press Esc */ + + return plymouth_running(); +} + +void manager_status_printf(Manager *m, bool ephemeral, const char *status, const char *format, ...) { + va_list ap; + + if (!manager_get_show_status(m)) + return; + + /* XXX We should totally drop the check for ephemeral here + * and thus effectively make 'Type=idle' pointless. */ + if (ephemeral && m->n_on_console > 0) + return; + + if (!manager_is_booting_or_shutting_down(m)) + return; + + va_start(ap, format); + status_vprintf(status, true, ephemeral, format, ap); + va_end(ap); +} + +void watch_init(Watch *w) { + assert(w); + + w->type = WATCH_INVALID; + w->fd = -1; +} diff --git a/src/core/manager.h b/src/core/manager.h new file mode 100644 index 0000000000..bf833540ae --- /dev/null +++ b/src/core/manager.h @@ -0,0 +1,306 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> +#include <inttypes.h> +#include <stdio.h> +#include <dbus/dbus.h> + +#include "fdset.h" + +/* Enforce upper limit how many names we allow */ +#define MANAGER_MAX_NAMES 131072 /* 128K */ + +typedef struct Manager Manager; +typedef enum WatchType WatchType; +typedef struct Watch Watch; + +typedef enum ManagerExitCode { + MANAGER_RUNNING, + MANAGER_EXIT, + MANAGER_RELOAD, + MANAGER_REEXECUTE, + MANAGER_REBOOT, + MANAGER_POWEROFF, + MANAGER_HALT, + MANAGER_KEXEC, + MANAGER_SWITCH_ROOT, + _MANAGER_EXIT_CODE_MAX, + _MANAGER_EXIT_CODE_INVALID = -1 +} ManagerExitCode; + +enum WatchType { + WATCH_INVALID, + WATCH_SIGNAL, + WATCH_NOTIFY, + WATCH_FD, + WATCH_UNIT_TIMER, + WATCH_JOB_TIMER, + WATCH_MOUNT, + WATCH_SWAP, + WATCH_UDEV, + WATCH_DBUS_WATCH, + WATCH_DBUS_TIMEOUT, + WATCH_TIME_CHANGE, + WATCH_JOBS_IN_PROGRESS +}; + +struct Watch { + int fd; + WatchType type; + union { + struct Unit *unit; + struct Job *job; + DBusWatch *bus_watch; + DBusTimeout *bus_timeout; + } data; + bool fd_is_dupped:1; + bool socket_accept:1; +}; + +#include "unit.h" +#include "job.h" +#include "hashmap.h" +#include "list.h" +#include "set.h" +#include "dbus.h" +#include "path-lookup.h" +#include "execute.h" + +struct Manager { + /* Note that the set of units we know of is allowed to be + * inconsistent. However the subset of it that is loaded may + * not, and the list of jobs may neither. */ + + /* Active jobs and units */ + Hashmap *units; /* name string => Unit object n:1 */ + Hashmap *jobs; /* job id => Job object 1:1 */ + + /* To make it easy to iterate through the units of a specific + * type we maintain a per type linked list */ + LIST_HEAD(Unit, units_by_type[_UNIT_TYPE_MAX]); + + /* To optimize iteration of units that have requires_mounts_for set */ + LIST_HEAD(Unit, has_requires_mounts_for); + + /* Units that need to be loaded */ + LIST_HEAD(Unit, load_queue); /* this is actually more a stack than a queue, but uh. */ + + /* Jobs that need to be run */ + LIST_HEAD(Job, run_queue); /* more a stack than a queue, too */ + + /* Units and jobs that have not yet been announced via + * D-Bus. When something about a job changes it is added here + * if it is not in there yet. This allows easy coalescing of + * D-Bus change signals. */ + LIST_HEAD(Unit, dbus_unit_queue); + LIST_HEAD(Job, dbus_job_queue); + + /* Units to remove */ + LIST_HEAD(Unit, cleanup_queue); + + /* Units to check when doing GC */ + LIST_HEAD(Unit, gc_queue); + + Hashmap *watch_pids; /* pid => Unit object n:1 */ + + char *notify_socket; + + Watch notify_watch; + Watch signal_watch; + Watch time_change_watch; + Watch jobs_in_progress_watch; + + int epoll_fd; + + unsigned n_snapshots; + + LookupPaths lookup_paths; + Set *unit_path_cache; + + char **environment; + char **default_controllers; + + usec_t runtime_watchdog; + usec_t shutdown_watchdog; + + dual_timestamp firmware_timestamp; + dual_timestamp loader_timestamp; + dual_timestamp kernel_timestamp; + dual_timestamp initrd_timestamp; + dual_timestamp userspace_timestamp; + dual_timestamp finish_timestamp; + + char *generator_unit_path; + char *generator_unit_path_early; + char *generator_unit_path_late; + + /* Data specific to the device subsystem */ + struct udev* udev; + struct udev_monitor* udev_monitor; + Watch udev_watch; + Hashmap *devices_by_sysfs; + + /* Data specific to the mount subsystem */ + FILE *proc_self_mountinfo; + Watch mount_watch; + + /* Data specific to the swap filesystem */ + FILE *proc_swaps; + Hashmap *swaps_by_proc_swaps; + bool request_reload; + Watch swap_watch; + + /* Data specific to the D-Bus subsystem */ + DBusConnection *api_bus, *system_bus; + DBusServer *private_bus; + Set *bus_connections, *bus_connections_for_dispatch; + + DBusMessage *queued_message; /* This is used during reloading: + * before the reload we queue the + * reply message here, and + * afterwards we send it */ + DBusConnection *queued_message_connection; /* The connection to send the queued message on */ + + Hashmap *watch_bus; /* D-Bus names => Unit object n:1 */ + int32_t name_data_slot; + int32_t conn_data_slot; + int32_t subscribed_data_slot; + + uint32_t current_job_id; + uint32_t default_unit_job_id; + + /* Data specific to the Automount subsystem */ + int dev_autofs_fd; + + /* Data specific to the cgroup subsystem */ + Hashmap *cgroup_bondings; /* path string => CGroupBonding object 1:n */ + char *cgroup_hierarchy; + + usec_t gc_queue_timestamp; + int gc_marker; + unsigned n_in_gc_queue; + + /* Make sure the user cannot accidentally unmount our cgroup + * file system */ + int pin_cgroupfs_fd; + + /* Flags */ + SystemdRunningAs running_as; + ManagerExitCode exit_code:5; + + bool dispatching_load_queue:1; + bool dispatching_run_queue:1; + bool dispatching_dbus_queue:1; + + bool taint_usr:1; + + bool show_status; + bool confirm_spawn; + + ExecOutput default_std_output, default_std_error; + + struct rlimit *rlimit[RLIMIT_NLIMITS]; + + /* non-zero if we are reloading or reexecuting, */ + int n_reloading; + + unsigned n_installed_jobs; + unsigned n_failed_jobs; + + /* Jobs in progress watching */ + unsigned n_running_jobs; + unsigned n_on_console; + unsigned jobs_in_progress_iteration; + + /* Type=idle pipes */ + int idle_pipe[2]; + + char *switch_root; + char *switch_root_init; +}; + +int manager_new(SystemdRunningAs running_as, Manager **m); +void manager_free(Manager *m); + +int manager_enumerate(Manager *m); +int manager_coldplug(Manager *m); +int manager_startup(Manager *m, FILE *serialization, FDSet *fds); + +Job *manager_get_job(Manager *m, uint32_t id); +Unit *manager_get_unit(Manager *m, const char *name); + +int manager_get_job_from_dbus_path(Manager *m, const char *s, Job **_j); + +int manager_load_unit_prepare(Manager *m, const char *name, const char *path, DBusError *e, Unit **_ret); +int manager_load_unit(Manager *m, const char *name, const char *path, DBusError *e, Unit **_ret); +int manager_load_unit_from_dbus_path(Manager *m, const char *s, DBusError *e, Unit **_u); + +int manager_add_job(Manager *m, JobType type, Unit *unit, JobMode mode, bool force, DBusError *e, Job **_ret); +int manager_add_job_by_name(Manager *m, JobType type, const char *name, JobMode mode, bool force, DBusError *e, Job **_ret); + +void manager_dump_units(Manager *s, FILE *f, const char *prefix); +void manager_dump_jobs(Manager *s, FILE *f, const char *prefix); + +void manager_clear_jobs(Manager *m); + +unsigned manager_dispatch_load_queue(Manager *m); +unsigned manager_dispatch_run_queue(Manager *m); +unsigned manager_dispatch_dbus_queue(Manager *m); + +int manager_set_default_controllers(Manager *m, char **controllers); +int manager_set_default_rlimits(Manager *m, struct rlimit **default_rlimit); + +int manager_loop(Manager *m); + +void manager_dispatch_bus_name_owner_changed(Manager *m, const char *name, const char* old_owner, const char *new_owner); +void manager_dispatch_bus_query_pid_done(Manager *m, const char *name, pid_t pid); + +int manager_open_serialization(Manager *m, FILE **_f); + +int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root); +int manager_deserialize(Manager *m, FILE *f, FDSet *fds); +int manager_distribute_fds(Manager *m, FDSet *fds); + +int manager_reload(Manager *m); + +bool manager_is_reloading_or_reexecuting(Manager *m) _pure_; + +void manager_reset_failed(Manager *m); + +void manager_send_unit_audit(Manager *m, Unit *u, int type, bool success); +void manager_send_unit_plymouth(Manager *m, Unit *u); + +bool manager_unit_inactive_or_pending(Manager *m, const char *name); + +void manager_check_finished(Manager *m); + +void manager_run_generators(Manager *m); +void manager_undo_generators(Manager *m); + +void manager_recheck_journal(Manager *m); + +void manager_set_show_status(Manager *m, bool b); +void manager_status_printf(Manager *m, bool ephemeral, const char *status, const char *format, ...) _printf_attr_(4,5); + +void watch_init(Watch *w); diff --git a/src/core/mount-setup.c b/src/core/mount-setup.c new file mode 100644 index 0000000000..4629808a7a --- /dev/null +++ b/src/core/mount-setup.c @@ -0,0 +1,417 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/mount.h> +#include <errno.h> +#include <sys/stat.h> +#include <stdlib.h> +#include <string.h> +#include <libgen.h> +#include <assert.h> +#include <unistd.h> +#include <ftw.h> + +#include "mount-setup.h" +#include "dev-setup.h" +#include "log.h" +#include "macro.h" +#include "util.h" +#include "label.h" +#include "set.h" +#include "strv.h" +#include "mkdir.h" +#include "path-util.h" +#include "missing.h" +#include "virt.h" +#include "efivars.h" + +#ifndef TTY_GID +#define TTY_GID 5 +#endif + +typedef enum MountMode { + MNT_NONE = 0, + MNT_FATAL = 1 << 0, + MNT_IN_CONTAINER = 1 << 1, +} MountMode; + +typedef struct MountPoint { + const char *what; + const char *where; + const char *type; + const char *options; + unsigned long flags; + bool (*condition_fn)(void); + MountMode mode; +} MountPoint; + +/* The first three entries we might need before SELinux is up. The + * fourth (securityfs) is needed by IMA to load a custom policy. The + * other ones we can delay until SELinux and IMA are loaded. */ +#define N_EARLY_MOUNT 5 + +static const MountPoint mount_table[] = { + { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_NONE }, + { "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, + NULL, MNT_NONE }, + { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, + NULL, MNT_IN_CONTAINER }, + { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, + NULL, MNT_IN_CONTAINER }, +#ifdef HAVE_XATTR + { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_IN_CONTAINER }, +#endif + { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_IN_CONTAINER }, + { "pstore", "/sys/fs/pstore", "pstore", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_NONE }, +#ifdef ENABLE_EFI + { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + is_efi_boot, MNT_NONE }, +#endif +}; + +/* These are API file systems that might be mounted by other software, + * we just list them here so that we know that we should ignore them */ + +static const char ignore_paths[] = + /* SELinux file systems */ + "/sys/fs/selinux\0" + "/selinux\0" + /* Legacy cgroup mount points */ + "/dev/cgroup\0" + "/cgroup\0" + /* Legacy kernel file system */ + "/proc/bus/usb\0" + /* Container bind mounts */ + "/proc/sys\0" + "/dev/console\0" + "/proc/kmsg\0"; + +bool mount_point_is_api(const char *path) { + unsigned i; + + /* Checks if this mount point is considered "API", and hence + * should be ignored */ + + for (i = 0; i < ELEMENTSOF(mount_table); i ++) + if (path_equal(path, mount_table[i].where)) + return true; + + return path_startswith(path, "/sys/fs/cgroup/"); +} + +bool mount_point_ignore(const char *path) { + const char *i; + + NULSTR_FOREACH(i, ignore_paths) + if (path_equal(path, i)) + return true; + + return false; +} + +static int mount_one(const MountPoint *p, bool relabel) { + int r; + + assert(p); + + if (p->condition_fn && !p->condition_fn()) + return 0; + + /* Relabel first, just in case */ + if (relabel) + label_fix(p->where, true, true); + + r = path_is_mount_point(p->where, true); + if (r < 0) + return r; + + if (r > 0) + return 0; + + /* Skip securityfs in a container */ + if (!(p->mode & MNT_IN_CONTAINER) && detect_container(NULL) > 0) + return 0; + + /* The access mode here doesn't really matter too much, since + * the mounted file system will take precedence anyway. */ + mkdir_p_label(p->where, 0755); + + log_debug("Mounting %s to %s of type %s with options %s.", + p->what, + p->where, + p->type, + strna(p->options)); + + if (mount(p->what, + p->where, + p->type, + p->flags, + p->options) < 0) { + log_full((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, "Failed to mount %s: %s", p->where, strerror(errno)); + return (p->mode & MNT_FATAL) ? -errno : 0; + } + + /* Relabel again, since we now mounted something fresh here */ + if (relabel) + label_fix(p->where, false, false); + + return 1; +} + +int mount_setup_early(void) { + unsigned i; + int r = 0; + + assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table)); + + /* Do a minimal mount of /proc and friends to enable the most + * basic stuff, such as SELinux */ + for (i = 0; i < N_EARLY_MOUNT; i ++) { + int j; + + j = mount_one(mount_table + i, false); + if (r == 0) + r = j; + } + + return r; +} + +int mount_cgroup_controllers(char ***join_controllers) { + int r; + char buf[LINE_MAX]; + _cleanup_set_free_free_ Set *controllers = NULL; + _cleanup_fclose_ FILE *f; + + /* Mount all available cgroup controllers that are built into the kernel. */ + + f = fopen("/proc/cgroups", "re"); + if (!f) { + log_error("Failed to enumerate cgroup controllers: %m"); + return 0; + } + + controllers = set_new(string_hash_func, string_compare_func); + if (!controllers) + return log_oom(); + + /* Ignore the header line */ + (void) fgets(buf, sizeof(buf), f); + + for (;;) { + char *controller; + int enabled = 0; + + if (fscanf(f, "%ms %*i %*i %i", &controller, &enabled) != 2) { + + if (feof(f)) + break; + + log_error("Failed to parse /proc/cgroups."); + return -EIO; + } + + if (!enabled) { + free(controller); + continue; + } + + r = set_consume(controllers, controller); + if (r < 0) { + log_error("Failed to add controller to set."); + return r; + } + } + + for (;;) { + MountPoint p = { + .what = "cgroup", + .type = "cgroup", + .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV, + .mode = MNT_IN_CONTAINER, + }; + char ***k = NULL; + _cleanup_free_ char *options = NULL, *controller; + + controller = set_steal_first(controllers); + if (!controller) + break; + + if (join_controllers) + for (k = join_controllers; *k; k++) + if (strv_find(*k, controller)) + break; + + if (k && *k) { + char **i, **j; + + for (i = *k, j = *k; *i; i++) { + + if (!streq(*i, controller)) { + char _cleanup_free_ *t; + + t = set_remove(controllers, *i); + if (!t) { + free(*i); + continue; + } + } + + *(j++) = *i; + } + + *j = NULL; + + options = strv_join(*k, ","); + if (!options) + return log_oom(); + } else { + options = controller; + controller = NULL; + } + + p.where = strappenda("/sys/fs/cgroup/", options); + p.options = options; + + r = mount_one(&p, true); + if (r < 0) + return r; + + if (r > 0 && k && *k) { + char **i; + + for (i = *k; *i; i++) { + char *t = strappenda("/sys/fs/cgroup/", *i); + + r = symlink(options, t); + if (r < 0 && errno != EEXIST) { + log_error("Failed to create symlink %s: %m", t); + return -errno; + } + } + } + } + + return 0; +} + +static int nftw_cb( + const char *fpath, + const struct stat *sb, + int tflag, + struct FTW *ftwbuf) { + + /* No need to label /dev twice in a row... */ + if (_unlikely_(ftwbuf->level == 0)) + return FTW_CONTINUE; + + label_fix(fpath, false, false); + + /* /run/initramfs is static data and big, no need to + * dynamically relabel its contents at boot... */ + if (_unlikely_(ftwbuf->level == 1 && + tflag == FTW_D && + streq(fpath, "/run/initramfs"))) + return FTW_SKIP_SUBTREE; + + return FTW_CONTINUE; +}; + +int mount_setup(bool loaded_policy) { + + static const char relabel[] = + "/run/initramfs/root-fsck\0" + "/run/initramfs/shutdown\0"; + + int r; + unsigned i; + const char *j; + + for (i = 0; i < ELEMENTSOF(mount_table); i ++) { + r = mount_one(mount_table + i, true); + + if (r < 0) + return r; + } + + /* Nodes in devtmpfs and /run need to be manually updated for + * the appropriate labels, after mounting. The other virtual + * API file systems like /sys and /proc do not need that, they + * use the same label for all their files. */ + if (loaded_policy) { + usec_t before_relabel, after_relabel; + char timespan[FORMAT_TIMESPAN_MAX]; + + before_relabel = now(CLOCK_MONOTONIC); + + nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL); + nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL); + + /* Explicitly relabel these */ + NULSTR_FOREACH(j, relabel) + label_fix(j, true, false); + + after_relabel = now(CLOCK_MONOTONIC); + + log_info("Relabelled /dev and /run in %s.", + format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0)); + } + + /* Create a few default symlinks, which are normally created + * by udevd, but some scripts might need them before we start + * udevd. */ + dev_setup(NULL); + + /* Mark the root directory as shared in regards to mount + * propagation. The kernel defaults to "private", but we think + * it makes more sense to have a default of "shared" so that + * nspawn and the container tools work out of the box. If + * specific setups need other settings they can reset the + * propagation mode to private if needed. */ + if (detect_container(NULL) <= 0) + if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0) + log_warning("Failed to set up the root directory for shared mount propagation: %m"); + + /* Create a few directories we always want around, Note that + * sd_booted() checks for /run/systemd/system, so this mkdir + * really needs to stay for good, otherwise software that + * copied sd-daemon.c into their sources will misdetect + * systemd. */ + mkdir_label("/run/systemd", 0755); + mkdir_label("/run/systemd/system", 0755); + mkdir_label("/run/systemd/inaccessible", 0000); + + return 0; +} diff --git a/src/core/mount-setup.h b/src/core/mount-setup.h new file mode 100644 index 0000000000..4b521ad0e1 --- /dev/null +++ b/src/core/mount-setup.h @@ -0,0 +1,33 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +int mount_setup_early(void); + +int mount_setup(bool loaded_policy); + +int mount_cgroup_controllers(char ***join_controllers); + +bool mount_point_is_api(const char *path); +bool mount_point_ignore(const char *path); diff --git a/src/core/mount.c b/src/core/mount.c new file mode 100644 index 0000000000..10073b50be --- /dev/null +++ b/src/core/mount.c @@ -0,0 +1,1932 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <stdio.h> +#include <mntent.h> +#include <sys/epoll.h> +#include <signal.h> + +#include "manager.h" +#include "unit.h" +#include "mount.h" +#include "load-fragment.h" +#include "load-dropin.h" +#include "log.h" +#include "sd-messages.h" +#include "strv.h" +#include "mkdir.h" +#include "path-util.h" +#include "mount-setup.h" +#include "unit-name.h" +#include "dbus-mount.h" +#include "special.h" +#include "bus-errors.h" +#include "exit-status.h" +#include "def.h" + +static const UnitActiveState state_translation_table[_MOUNT_STATE_MAX] = { + [MOUNT_DEAD] = UNIT_INACTIVE, + [MOUNT_MOUNTING] = UNIT_ACTIVATING, + [MOUNT_MOUNTING_DONE] = UNIT_ACTIVE, + [MOUNT_MOUNTED] = UNIT_ACTIVE, + [MOUNT_REMOUNTING] = UNIT_RELOADING, + [MOUNT_UNMOUNTING] = UNIT_DEACTIVATING, + [MOUNT_MOUNTING_SIGTERM] = UNIT_DEACTIVATING, + [MOUNT_MOUNTING_SIGKILL] = UNIT_DEACTIVATING, + [MOUNT_REMOUNTING_SIGTERM] = UNIT_RELOADING, + [MOUNT_REMOUNTING_SIGKILL] = UNIT_RELOADING, + [MOUNT_UNMOUNTING_SIGTERM] = UNIT_DEACTIVATING, + [MOUNT_UNMOUNTING_SIGKILL] = UNIT_DEACTIVATING, + [MOUNT_FAILED] = UNIT_FAILED +}; + +static void mount_init(Unit *u) { + Mount *m = MOUNT(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + m->timeout_usec = DEFAULT_TIMEOUT_USEC; + m->directory_mode = 0755; + + exec_context_init(&m->exec_context); + + if (unit_has_name(u, "-.mount")) { + /* Don't allow start/stop for root directory */ + UNIT(m)->refuse_manual_start = true; + UNIT(m)->refuse_manual_stop = true; + } else { + /* The stdio/kmsg bridge socket is on /, in order to avoid a + * dep loop, don't use kmsg logging for -.mount */ + m->exec_context.std_output = u->manager->default_std_output; + m->exec_context.std_error = u->manager->default_std_error; + } + + kill_context_init(&m->kill_context); + + /* We need to make sure that /bin/mount is always called in + * the same process group as us, so that the autofs kernel + * side doesn't send us another mount request while we are + * already trying to comply its last one. */ + m->exec_context.same_pgrp = true; + + m->timer_watch.type = WATCH_INVALID; + + m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID; + + UNIT(m)->ignore_on_isolate = true; +} + +static void mount_unwatch_control_pid(Mount *m) { + assert(m); + + if (m->control_pid <= 0) + return; + + unit_unwatch_pid(UNIT(m), m->control_pid); + m->control_pid = 0; +} + +static void mount_parameters_done(MountParameters *p) { + assert(p); + + free(p->what); + free(p->options); + free(p->fstype); + + p->what = p->options = p->fstype = NULL; +} + +static void mount_done(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + free(m->where); + m->where = NULL; + + mount_parameters_done(&m->parameters_proc_self_mountinfo); + mount_parameters_done(&m->parameters_fragment); + + exec_context_done(&m->exec_context, manager_is_reloading_or_reexecuting(u->manager)); + exec_command_done_array(m->exec_command, _MOUNT_EXEC_COMMAND_MAX); + m->control_command = NULL; + + mount_unwatch_control_pid(m); + + unit_unwatch_timer(u, &m->timer_watch); +} + +_pure_ static MountParameters* get_mount_parameters_fragment(Mount *m) { + assert(m); + + if (m->from_fragment) + return &m->parameters_fragment; + + return NULL; +} + +_pure_ static MountParameters* get_mount_parameters(Mount *m) { + assert(m); + + if (m->from_proc_self_mountinfo) + return &m->parameters_proc_self_mountinfo; + + return get_mount_parameters_fragment(m); +} + +static int mount_add_mount_links(Mount *m) { + Unit *other; + int r; + MountParameters *pm; + + assert(m); + + pm = get_mount_parameters_fragment(m); + + /* Adds in links to other mount points that might lie below or + * above us in the hierarchy */ + + LIST_FOREACH(units_by_type, other, UNIT(m)->manager->units_by_type[UNIT_MOUNT]) { + Mount *n = MOUNT(other); + MountParameters *pn; + + if (n == m) + continue; + + if (UNIT(n)->load_state != UNIT_LOADED) + continue; + + pn = get_mount_parameters_fragment(n); + + if (path_startswith(m->where, n->where)) { + + if ((r = unit_add_dependency(UNIT(m), UNIT_AFTER, UNIT(n), true)) < 0) + return r; + + if (pn) + if ((r = unit_add_dependency(UNIT(m), UNIT_REQUIRES, UNIT(n), true)) < 0) + return r; + + } else if (path_startswith(n->where, m->where)) { + + if ((r = unit_add_dependency(UNIT(n), UNIT_AFTER, UNIT(m), true)) < 0) + return r; + + if (pm) + if ((r = unit_add_dependency(UNIT(n), UNIT_REQUIRES, UNIT(m), true)) < 0) + return r; + + } else if (pm && pm->what && path_startswith(pm->what, n->where)) { + + if ((r = unit_add_dependency(UNIT(m), UNIT_AFTER, UNIT(n), true)) < 0) + return r; + + if ((r = unit_add_dependency(UNIT(m), UNIT_REQUIRES, UNIT(n), true)) < 0) + return r; + + } else if (pn && pn->what && path_startswith(pn->what, m->where)) { + + if ((r = unit_add_dependency(UNIT(n), UNIT_AFTER, UNIT(m), true)) < 0) + return r; + + if ((r = unit_add_dependency(UNIT(n), UNIT_REQUIRES, UNIT(m), true)) < 0) + return r; + } + } + + return 0; +} + +static int mount_add_swap_links(Mount *m) { + Unit *other; + int r; + + assert(m); + + LIST_FOREACH(units_by_type, other, UNIT(m)->manager->units_by_type[UNIT_SWAP]) { + r = swap_add_one_mount_link(SWAP(other), m); + if (r < 0) + return r; + } + + return 0; +} + +static int mount_add_path_links(Mount *m) { + Unit *other; + int r; + + assert(m); + + LIST_FOREACH(units_by_type, other, UNIT(m)->manager->units_by_type[UNIT_PATH]) { + r = path_add_one_mount_link(PATH(other), m); + if (r < 0) + return r; + } + + return 0; +} + +static int mount_add_automount_links(Mount *m) { + Unit *other; + int r; + + assert(m); + + LIST_FOREACH(units_by_type, other, UNIT(m)->manager->units_by_type[UNIT_AUTOMOUNT]) { + r = automount_add_one_mount_link(AUTOMOUNT(other), m); + if (r < 0) + return r; + } + + return 0; +} + +static int mount_add_socket_links(Mount *m) { + Unit *other; + int r; + + assert(m); + + LIST_FOREACH(units_by_type, other, UNIT(m)->manager->units_by_type[UNIT_SOCKET]) { + r = socket_add_one_mount_link(SOCKET(other), m); + if (r < 0) + return r; + } + + return 0; +} + +static int mount_add_requires_mounts_links(Mount *m) { + Unit *other; + int r; + + assert(m); + + LIST_FOREACH(has_requires_mounts_for, other, UNIT(m)->manager->has_requires_mounts_for) { + r = unit_add_one_mount_link(other, m); + if (r < 0) + return r; + } + + return 0; +} + +static char* mount_test_option(const char *haystack, const char *needle) { + struct mntent me = { .mnt_opts = (char*) haystack }; + + assert(needle); + + /* Like glibc's hasmntopt(), but works on a string, not a + * struct mntent */ + + if (!haystack) + return NULL; + + return hasmntopt(&me, needle); +} + +static bool mount_is_network(MountParameters *p) { + assert(p); + + if (mount_test_option(p->options, "_netdev")) + return true; + + if (p->fstype && fstype_is_network(p->fstype)) + return true; + + return false; +} + +static bool mount_is_bind(MountParameters *p) { + assert(p); + + if (mount_test_option(p->options, "bind")) + return true; + + if (p->fstype && streq(p->fstype, "bind")) + return true; + + if (mount_test_option(p->options, "rbind")) + return true; + + if (p->fstype && streq(p->fstype, "rbind")) + return true; + + return false; +} + +static bool needs_quota(MountParameters *p) { + assert(p); + + if (mount_is_network(p)) + return false; + + if (mount_is_bind(p)) + return false; + + return mount_test_option(p->options, "usrquota") || + mount_test_option(p->options, "grpquota") || + mount_test_option(p->options, "quota") || + mount_test_option(p->options, "usrjquota") || + mount_test_option(p->options, "grpjquota"); +} + +static int mount_add_device_links(Mount *m) { + MountParameters *p; + int r; + + assert(m); + + p = get_mount_parameters_fragment(m); + if (!p) + return 0; + + if (!p->what) + return 0; + + if (mount_is_bind(p)) + return 0; + + if (!is_device_path(p->what)) + return 0; + + if (path_equal(m->where, "/")) + return 0; + + r = unit_add_node_link(UNIT(m), p->what, false); + if (r < 0) + return r; + + if (p->passno > 0 && + UNIT(m)->manager->running_as == SYSTEMD_SYSTEM) { + char *name; + Unit *fsck; + /* Let's add in the fsck service */ + + /* aka SPECIAL_FSCK_SERVICE */ + name = unit_name_from_path_instance("systemd-fsck", p->what, ".service"); + if (!name) + return -ENOMEM; + + r = manager_load_unit_prepare(UNIT(m)->manager, name, NULL, NULL, &fsck); + if (r < 0) { + log_warning_unit(name, + "Failed to prepare unit %s: %s", name, strerror(-r)); + free(name); + return r; + } + free(name); + + SERVICE(fsck)->fsck_passno = p->passno; + + r = unit_add_two_dependencies(UNIT(m), UNIT_AFTER, UNIT_REQUIRES, fsck, true); + if (r < 0) + return r; + } + + return 0; +} + +static int mount_add_quota_links(Mount *m) { + int r; + MountParameters *p; + + assert(m); + + if (UNIT(m)->manager->running_as != SYSTEMD_SYSTEM) + return 0; + + p = get_mount_parameters_fragment(m); + if (!p) + return 0; + + if (!needs_quota(p)) + return 0; + + r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_BEFORE, UNIT_WANTS, SPECIAL_QUOTACHECK_SERVICE, NULL, true); + if (r < 0) + return r; + + r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_BEFORE, UNIT_WANTS, SPECIAL_QUOTAON_SERVICE, NULL, true); + if (r < 0) + return r; + + return 0; +} + +static int mount_add_default_dependencies(Mount *m) { + const char *after, *after2, *online; + MountParameters *p; + int r; + + assert(m); + + if (UNIT(m)->manager->running_as != SYSTEMD_SYSTEM) + return 0; + + p = get_mount_parameters(m); + + if (!p) + return 0; + + if (path_equal(m->where, "/")) + return 0; + + if (mount_is_network(p)) { + after = SPECIAL_REMOTE_FS_PRE_TARGET; + after2 = SPECIAL_NETWORK_TARGET; + online = SPECIAL_NETWORK_ONLINE_TARGET; + } else { + after = SPECIAL_LOCAL_FS_PRE_TARGET; + after2 = NULL; + online = NULL; + } + + r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, after, NULL, true); + if (r < 0) + return r; + + if (after2) { + r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, after2, NULL, true); + if (r < 0) + return r; + } + + if (online) { + r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_WANTS, UNIT_AFTER, online, NULL, true); + if (r < 0) + return r; + } + + r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_UMOUNT_TARGET, NULL, true); + if (r < 0) + return r; + + return 0; +} + +static int mount_fix_timeouts(Mount *m) { + MountParameters *p; + const char *timeout = NULL; + Unit *other; + Iterator i; + usec_t u; + char *t; + int r; + + assert(m); + + p = get_mount_parameters_fragment(m); + if (!p) + return 0; + + /* Allow configuration how long we wait for a device that + * backs a mount point to show up. This is useful to support + * endless device timeouts for devices that show up only after + * user input, like crypto devices. */ + + if ((timeout = mount_test_option(p->options, "comment=systemd.device-timeout"))) + timeout += 31; + else if ((timeout = mount_test_option(p->options, "x-systemd.device-timeout"))) + timeout += 25; + else + return 0; + + t = strndup(timeout, strcspn(timeout, ",;" WHITESPACE)); + if (!t) + return -ENOMEM; + + r = parse_sec(t, &u); + free(t); + + if (r < 0) { + log_warning_unit(UNIT(m)->id, + "Failed to parse timeout for %s, ignoring: %s", + m->where, timeout); + return r; + } + + SET_FOREACH(other, UNIT(m)->dependencies[UNIT_AFTER], i) { + if (other->type != UNIT_DEVICE) + continue; + + other->job_timeout = u; + } + + return 0; +} + +static int mount_verify(Mount *m) { + bool b; + char *e; + assert(m); + + if (UNIT(m)->load_state != UNIT_LOADED) + return 0; + + if (!m->from_fragment && !m->from_proc_self_mountinfo) + return -ENOENT; + + if (!(e = unit_name_from_path(m->where, ".mount"))) + return -ENOMEM; + + b = unit_has_name(UNIT(m), e); + free(e); + + if (!b) { + log_error_unit(UNIT(m)->id, + "%s's Where setting doesn't match unit name. Refusing.", + UNIT(m)->id); + return -EINVAL; + } + + if (mount_point_is_api(m->where) || mount_point_ignore(m->where)) { + log_error_unit(UNIT(m)->id, + "Cannot create mount unit for API file system %s. Refusing.", + m->where); + return -EINVAL; + } + + if (UNIT(m)->fragment_path && !m->parameters_fragment.what) { + log_error_unit(UNIT(m)->id, + "%s's What setting is missing. Refusing.", UNIT(m)->id); + return -EBADMSG; + } + + if (m->exec_context.pam_name && m->kill_context.kill_mode != KILL_CONTROL_GROUP) { + log_error_unit(UNIT(m)->id, + "%s has PAM enabled. Kill mode must be set to control-group'. Refusing.", + UNIT(m)->id); + return -EINVAL; + } + + return 0; +} + +static int mount_add_extras(Mount *m) { + Unit *u = UNIT(m); + int r; + + if (UNIT(m)->fragment_path) + m->from_fragment = true; + + if (!m->where) { + m->where = unit_name_to_path(u->id); + if (!m->where) + return -ENOMEM; + } + + path_kill_slashes(m->where); + + r = unit_add_exec_dependencies(u, &m->exec_context); + if (r < 0) + return r; + + if (!UNIT(m)->description) { + r = unit_set_description(u, m->where); + if (r < 0) + return r; + } + + r = mount_add_device_links(m); + if (r < 0) + return r; + + r = mount_add_mount_links(m); + if (r < 0) + return r; + + r = mount_add_socket_links(m); + if (r < 0) + return r; + + r = mount_add_swap_links(m); + if (r < 0) + return r; + + r = mount_add_path_links(m); + if (r < 0) + return r; + + r = mount_add_requires_mounts_links(m); + if (r < 0) + return r; + + r = mount_add_automount_links(m); + if (r < 0) + return r; + + r = mount_add_quota_links(m); + if (r < 0) + return r; + + if (UNIT(m)->default_dependencies) { + r = mount_add_default_dependencies(m); + if (r < 0) + return r; + } + + r = unit_add_default_cgroups(u); + if (r < 0) + return r; + + r = mount_fix_timeouts(m); + if (r < 0) + return r; + + return 0; +} + +static int mount_load(Unit *u) { + Mount *m = MOUNT(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + if (m->from_proc_self_mountinfo) + r = unit_load_fragment_and_dropin_optional(u); + else + r = unit_load_fragment_and_dropin(u); + + if (r < 0) + return r; + + /* This is a new unit? Then let's add in some extras */ + if (u->load_state == UNIT_LOADED) { + r = mount_add_extras(m); + if (r < 0) + return r; + + r = unit_exec_context_defaults(u, &m->exec_context); + if (r < 0) + return r; + } + + return mount_verify(m); +} + +static int mount_notify_automount(Mount *m, int status) { + Unit *p; + int r; + Iterator i; + + assert(m); + + SET_FOREACH(p, UNIT(m)->dependencies[UNIT_TRIGGERED_BY], i) + if (p->type == UNIT_AUTOMOUNT) { + r = automount_send_ready(AUTOMOUNT(p), status); + if (r < 0) + return r; + } + + return 0; +} + +static void mount_set_state(Mount *m, MountState state) { + MountState old_state; + assert(m); + + old_state = m->state; + m->state = state; + + if (state != MOUNT_MOUNTING && + state != MOUNT_MOUNTING_DONE && + state != MOUNT_REMOUNTING && + state != MOUNT_UNMOUNTING && + state != MOUNT_MOUNTING_SIGTERM && + state != MOUNT_MOUNTING_SIGKILL && + state != MOUNT_UNMOUNTING_SIGTERM && + state != MOUNT_UNMOUNTING_SIGKILL && + state != MOUNT_REMOUNTING_SIGTERM && + state != MOUNT_REMOUNTING_SIGKILL) { + unit_unwatch_timer(UNIT(m), &m->timer_watch); + mount_unwatch_control_pid(m); + m->control_command = NULL; + m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID; + } + + if (state == MOUNT_MOUNTED || + state == MOUNT_REMOUNTING) + mount_notify_automount(m, 0); + else if (state == MOUNT_DEAD || + state == MOUNT_UNMOUNTING || + state == MOUNT_MOUNTING_SIGTERM || + state == MOUNT_MOUNTING_SIGKILL || + state == MOUNT_REMOUNTING_SIGTERM || + state == MOUNT_REMOUNTING_SIGKILL || + state == MOUNT_UNMOUNTING_SIGTERM || + state == MOUNT_UNMOUNTING_SIGKILL || + state == MOUNT_FAILED) { + if (state != old_state) + mount_notify_automount(m, -ENODEV); + } + + if (state != old_state) + log_debug_unit(UNIT(m)->id, + "%s changed %s -> %s", + UNIT(m)->id, + mount_state_to_string(old_state), + mount_state_to_string(state)); + + unit_notify(UNIT(m), state_translation_table[old_state], state_translation_table[state], m->reload_result == MOUNT_SUCCESS); + m->reload_result = MOUNT_SUCCESS; +} + +static int mount_coldplug(Unit *u) { + Mount *m = MOUNT(u); + MountState new_state = MOUNT_DEAD; + int r; + + assert(m); + assert(m->state == MOUNT_DEAD); + + if (m->deserialized_state != m->state) + new_state = m->deserialized_state; + else if (m->from_proc_self_mountinfo) + new_state = MOUNT_MOUNTED; + + if (new_state != m->state) { + + if (new_state == MOUNT_MOUNTING || + new_state == MOUNT_MOUNTING_DONE || + new_state == MOUNT_REMOUNTING || + new_state == MOUNT_UNMOUNTING || + new_state == MOUNT_MOUNTING_SIGTERM || + new_state == MOUNT_MOUNTING_SIGKILL || + new_state == MOUNT_UNMOUNTING_SIGTERM || + new_state == MOUNT_UNMOUNTING_SIGKILL || + new_state == MOUNT_REMOUNTING_SIGTERM || + new_state == MOUNT_REMOUNTING_SIGKILL) { + + if (m->control_pid <= 0) + return -EBADMSG; + + r = unit_watch_pid(UNIT(m), m->control_pid); + if (r < 0) + return r; + + r = unit_watch_timer(UNIT(m), CLOCK_MONOTONIC, true, m->timeout_usec, &m->timer_watch); + if (r < 0) + return r; + } + + mount_set_state(m, new_state); + } + + return 0; +} + +static void mount_dump(Unit *u, FILE *f, const char *prefix) { + Mount *m = MOUNT(u); + MountParameters *p; + + assert(m); + assert(f); + + p = get_mount_parameters(m); + + fprintf(f, + "%sMount State: %s\n" + "%sResult: %s\n" + "%sWhere: %s\n" + "%sWhat: %s\n" + "%sFile System Type: %s\n" + "%sOptions: %s\n" + "%sFrom /proc/self/mountinfo: %s\n" + "%sFrom fragment: %s\n" + "%sDirectoryMode: %04o\n", + prefix, mount_state_to_string(m->state), + prefix, mount_result_to_string(m->result), + prefix, m->where, + prefix, strna(p->what), + prefix, strna(p->fstype), + prefix, strna(p->options), + prefix, yes_no(m->from_proc_self_mountinfo), + prefix, yes_no(m->from_fragment), + prefix, m->directory_mode); + + if (m->control_pid > 0) + fprintf(f, + "%sControl PID: %lu\n", + prefix, (unsigned long) m->control_pid); + + exec_context_dump(&m->exec_context, f, prefix); + kill_context_dump(&m->kill_context, f, prefix); +} + +static int mount_spawn(Mount *m, ExecCommand *c, pid_t *_pid) { + pid_t pid; + int r; + + assert(m); + assert(c); + assert(_pid); + + r = unit_watch_timer(UNIT(m), CLOCK_MONOTONIC, true, m->timeout_usec, &m->timer_watch); + if (r < 0) + goto fail; + + if ((r = exec_spawn(c, + NULL, + &m->exec_context, + NULL, 0, + UNIT(m)->manager->environment, + true, + true, + true, + UNIT(m)->manager->confirm_spawn, + UNIT(m)->cgroup_bondings, + UNIT(m)->cgroup_attributes, + NULL, + UNIT(m)->id, + NULL, + &pid)) < 0) + goto fail; + + if ((r = unit_watch_pid(UNIT(m), pid)) < 0) + /* FIXME: we need to do something here */ + goto fail; + + *_pid = pid; + + return 0; + +fail: + unit_unwatch_timer(UNIT(m), &m->timer_watch); + + return r; +} + +static void mount_enter_dead(Mount *m, MountResult f) { + assert(m); + + if (f != MOUNT_SUCCESS) + m->result = f; + + exec_context_tmp_dirs_done(&m->exec_context); + mount_set_state(m, m->result != MOUNT_SUCCESS ? MOUNT_FAILED : MOUNT_DEAD); +} + +static void mount_enter_mounted(Mount *m, MountResult f) { + assert(m); + + if (f != MOUNT_SUCCESS) + m->result = f; + + mount_set_state(m, MOUNT_MOUNTED); +} + +static void mount_enter_signal(Mount *m, MountState state, MountResult f) { + int r; + + assert(m); + + if (f != MOUNT_SUCCESS) + m->result = f; + + r = unit_kill_context( + UNIT(m), + &m->kill_context, + state != MOUNT_MOUNTING_SIGTERM && state != MOUNT_UNMOUNTING_SIGTERM && state != MOUNT_REMOUNTING_SIGTERM, + -1, + m->control_pid, + false); + if (r < 0) + goto fail; + + if (r > 0) { + r = unit_watch_timer(UNIT(m), CLOCK_MONOTONIC, true, m->timeout_usec, &m->timer_watch); + if (r < 0) + goto fail; + + mount_set_state(m, state); + } else if (state == MOUNT_REMOUNTING_SIGTERM || state == MOUNT_REMOUNTING_SIGKILL) + mount_enter_mounted(m, MOUNT_SUCCESS); + else + mount_enter_dead(m, MOUNT_SUCCESS); + + return; + +fail: + log_warning_unit(UNIT(m)->id, + "%s failed to kill processes: %s", UNIT(m)->id, strerror(-r)); + + if (state == MOUNT_REMOUNTING_SIGTERM || state == MOUNT_REMOUNTING_SIGKILL) + mount_enter_mounted(m, MOUNT_FAILURE_RESOURCES); + else + mount_enter_dead(m, MOUNT_FAILURE_RESOURCES); +} + +void warn_if_dir_nonempty(const char *unit, const char* where) { + assert(unit); + assert(where); + + if (dir_is_empty(where) > 0) + return; + + log_struct_unit(LOG_NOTICE, + unit, + "MESSAGE=%s: Directory %s to mount over is not empty, mounting anyway.", + unit, where, + "WHERE=%s", where, + MESSAGE_ID(SD_MESSAGE_OVERMOUNTING), + NULL); +} + +static void mount_enter_unmounting(Mount *m) { + int r; + + assert(m); + + m->control_command_id = MOUNT_EXEC_UNMOUNT; + m->control_command = m->exec_command + MOUNT_EXEC_UNMOUNT; + + if ((r = exec_command_set( + m->control_command, + "/bin/umount", + m->where, + NULL)) < 0) + goto fail; + + mount_unwatch_control_pid(m); + + if ((r = mount_spawn(m, m->control_command, &m->control_pid)) < 0) + goto fail; + + mount_set_state(m, MOUNT_UNMOUNTING); + + return; + +fail: + log_warning_unit(UNIT(m)->id, + "%s failed to run 'umount' task: %s", + UNIT(m)->id, strerror(-r)); + mount_enter_mounted(m, MOUNT_FAILURE_RESOURCES); +} + +static void mount_enter_mounting(Mount *m) { + int r; + MountParameters *p; + + assert(m); + + m->control_command_id = MOUNT_EXEC_MOUNT; + m->control_command = m->exec_command + MOUNT_EXEC_MOUNT; + + mkdir_p_label(m->where, m->directory_mode); + + warn_if_dir_nonempty(m->meta.id, m->where); + + /* Create the source directory for bind-mounts if needed */ + p = get_mount_parameters_fragment(m); + if (p && mount_is_bind(p)) + mkdir_p_label(p->what, m->directory_mode); + + if (m->from_fragment) + r = exec_command_set( + m->control_command, + "/bin/mount", + m->parameters_fragment.what, + m->where, + "-t", m->parameters_fragment.fstype ? m->parameters_fragment.fstype : "auto", + m->parameters_fragment.options ? "-o" : NULL, m->parameters_fragment.options, + NULL); + else + r = -ENOENT; + + if (r < 0) + goto fail; + + mount_unwatch_control_pid(m); + + r = mount_spawn(m, m->control_command, &m->control_pid); + if (r < 0) + goto fail; + + mount_set_state(m, MOUNT_MOUNTING); + + return; + +fail: + log_warning_unit(UNIT(m)->id, + "%s failed to run 'mount' task: %s", + UNIT(m)->id, strerror(-r)); + mount_enter_dead(m, MOUNT_FAILURE_RESOURCES); +} + +static void mount_enter_mounting_done(Mount *m) { + assert(m); + + mount_set_state(m, MOUNT_MOUNTING_DONE); +} + +static void mount_enter_remounting(Mount *m) { + int r; + + assert(m); + + m->control_command_id = MOUNT_EXEC_REMOUNT; + m->control_command = m->exec_command + MOUNT_EXEC_REMOUNT; + + if (m->from_fragment) { + char *buf = NULL; + const char *o; + + if (m->parameters_fragment.options) { + if (!(buf = strappend("remount,", m->parameters_fragment.options))) { + r = -ENOMEM; + goto fail; + } + + o = buf; + } else + o = "remount"; + + r = exec_command_set( + m->control_command, + "/bin/mount", + m->parameters_fragment.what, + m->where, + "-t", m->parameters_fragment.fstype ? m->parameters_fragment.fstype : "auto", + "-o", o, + NULL); + + free(buf); + } else + r = -ENOENT; + + if (r < 0) + goto fail; + + mount_unwatch_control_pid(m); + + if ((r = mount_spawn(m, m->control_command, &m->control_pid)) < 0) + goto fail; + + mount_set_state(m, MOUNT_REMOUNTING); + + return; + +fail: + log_warning_unit(UNIT(m)->id, + "%s failed to run 'remount' task: %s", + UNIT(m)->id, strerror(-r)); + m->reload_result = MOUNT_FAILURE_RESOURCES; + mount_enter_mounted(m, MOUNT_SUCCESS); +} + +static int mount_start(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + /* We cannot fulfill this request right now, try again later + * please! */ + if (m->state == MOUNT_UNMOUNTING || + m->state == MOUNT_UNMOUNTING_SIGTERM || + m->state == MOUNT_UNMOUNTING_SIGKILL || + m->state == MOUNT_MOUNTING_SIGTERM || + m->state == MOUNT_MOUNTING_SIGKILL) + return -EAGAIN; + + /* Already on it! */ + if (m->state == MOUNT_MOUNTING) + return 0; + + assert(m->state == MOUNT_DEAD || m->state == MOUNT_FAILED); + + m->result = MOUNT_SUCCESS; + m->reload_result = MOUNT_SUCCESS; + + mount_enter_mounting(m); + return 0; +} + +static int mount_stop(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + /* Already on it */ + if (m->state == MOUNT_UNMOUNTING || + m->state == MOUNT_UNMOUNTING_SIGKILL || + m->state == MOUNT_UNMOUNTING_SIGTERM || + m->state == MOUNT_MOUNTING_SIGTERM || + m->state == MOUNT_MOUNTING_SIGKILL) + return 0; + + assert(m->state == MOUNT_MOUNTING || + m->state == MOUNT_MOUNTING_DONE || + m->state == MOUNT_MOUNTED || + m->state == MOUNT_REMOUNTING || + m->state == MOUNT_REMOUNTING_SIGTERM || + m->state == MOUNT_REMOUNTING_SIGKILL); + + mount_enter_unmounting(m); + return 0; +} + +static int mount_reload(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + if (m->state == MOUNT_MOUNTING_DONE) + return -EAGAIN; + + assert(m->state == MOUNT_MOUNTED); + + mount_enter_remounting(m); + return 0; +} + +static int mount_serialize(Unit *u, FILE *f, FDSet *fds) { + Mount *m = MOUNT(u); + + assert(m); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", mount_state_to_string(m->state)); + unit_serialize_item(u, f, "result", mount_result_to_string(m->result)); + unit_serialize_item(u, f, "reload-result", mount_result_to_string(m->reload_result)); + + if (m->control_pid > 0) + unit_serialize_item_format(u, f, "control-pid", "%lu", (unsigned long) m->control_pid); + + if (m->control_command_id >= 0) + unit_serialize_item(u, f, "control-command", mount_exec_command_to_string(m->control_command_id)); + + exec_context_serialize(&m->exec_context, UNIT(m), f); + + return 0; +} + +static int mount_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Mount *m = MOUNT(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + MountState state; + + if ((state = mount_state_from_string(value)) < 0) + log_debug_unit(u->id, "Failed to parse state value %s", value); + else + m->deserialized_state = state; + } else if (streq(key, "result")) { + MountResult f; + + f = mount_result_from_string(value); + if (f < 0) + log_debug_unit(UNIT(m)->id, + "Failed to parse result value %s", value); + else if (f != MOUNT_SUCCESS) + m->result = f; + + } else if (streq(key, "reload-result")) { + MountResult f; + + f = mount_result_from_string(value); + if (f < 0) + log_debug_unit(UNIT(m)->id, + "Failed to parse reload result value %s", value); + else if (f != MOUNT_SUCCESS) + m->reload_result = f; + + } else if (streq(key, "control-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_debug_unit(UNIT(m)->id, + "Failed to parse control-pid value %s", value); + else + m->control_pid = pid; + } else if (streq(key, "control-command")) { + MountExecCommand id; + + if ((id = mount_exec_command_from_string(value)) < 0) + log_debug_unit(UNIT(m)->id, + "Failed to parse exec-command value %s", value); + else { + m->control_command_id = id; + m->control_command = m->exec_command + id; + } + } else if (streq(key, "tmp-dir")) { + char *t; + + t = strdup(value); + if (!t) + return log_oom(); + + m->exec_context.tmp_dir = t; + } else if (streq(key, "var-tmp-dir")) { + char *t; + + t = strdup(value); + if (!t) + return log_oom(); + + m->exec_context.var_tmp_dir = t; + } else + log_debug_unit(UNIT(m)->id, + "Unknown serialization key '%s'", key); + + return 0; +} + +_pure_ static UnitActiveState mount_active_state(Unit *u) { + assert(u); + + return state_translation_table[MOUNT(u)->state]; +} + +_pure_ static const char *mount_sub_state_to_string(Unit *u) { + assert(u); + + return mount_state_to_string(MOUNT(u)->state); +} + +_pure_ static bool mount_check_gc(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + return m->from_proc_self_mountinfo; +} + +static void mount_sigchld_event(Unit *u, pid_t pid, int code, int status) { + Mount *m = MOUNT(u); + MountResult f; + + assert(m); + assert(pid >= 0); + + if (pid != m->control_pid) + return; + + m->control_pid = 0; + + if (is_clean_exit(code, status, NULL)) + f = MOUNT_SUCCESS; + else if (code == CLD_EXITED) + f = MOUNT_FAILURE_EXIT_CODE; + else if (code == CLD_KILLED) + f = MOUNT_FAILURE_SIGNAL; + else if (code == CLD_DUMPED) + f = MOUNT_FAILURE_CORE_DUMP; + else + assert_not_reached("Unknown code"); + + if (f != MOUNT_SUCCESS) + m->result = f; + + if (m->control_command) { + exec_status_exit(&m->control_command->exec_status, &m->exec_context, pid, code, status); + + m->control_command = NULL; + m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID; + } + + log_full_unit(f == MOUNT_SUCCESS ? LOG_DEBUG : LOG_NOTICE, u->id, + "%s mount process exited, code=%s status=%i", + u->id, sigchld_code_to_string(code), status); + + /* Note that mount(8) returning and the kernel sending us a + * mount table change event might happen out-of-order. If an + * operation succeed we assume the kernel will follow soon too + * and already change into the resulting state. If it fails + * we check if the kernel still knows about the mount. and + * change state accordingly. */ + + switch (m->state) { + + case MOUNT_MOUNTING: + case MOUNT_MOUNTING_DONE: + case MOUNT_MOUNTING_SIGKILL: + case MOUNT_MOUNTING_SIGTERM: + + if (f == MOUNT_SUCCESS) + mount_enter_mounted(m, f); + else if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, f); + else + mount_enter_dead(m, f); + break; + + case MOUNT_REMOUNTING: + case MOUNT_REMOUNTING_SIGKILL: + case MOUNT_REMOUNTING_SIGTERM: + + m->reload_result = f; + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_SUCCESS); + else + mount_enter_dead(m, MOUNT_SUCCESS); + + break; + + case MOUNT_UNMOUNTING: + case MOUNT_UNMOUNTING_SIGKILL: + case MOUNT_UNMOUNTING_SIGTERM: + + if (f == MOUNT_SUCCESS) + mount_enter_dead(m, f); + else if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, f); + else + mount_enter_dead(m, f); + break; + + default: + assert_not_reached("Uh, control process died at wrong time."); + } + + /* Notify clients about changed exit status */ + unit_add_to_dbus_queue(u); +} + +static void mount_timer_event(Unit *u, uint64_t elapsed, Watch *w) { + Mount *m = MOUNT(u); + + assert(m); + assert(elapsed == 1); + assert(w == &m->timer_watch); + + switch (m->state) { + + case MOUNT_MOUNTING: + case MOUNT_MOUNTING_DONE: + log_warning_unit(u->id, + "%s mounting timed out. Stopping.", u->id); + mount_enter_signal(m, MOUNT_MOUNTING_SIGTERM, MOUNT_FAILURE_TIMEOUT); + break; + + case MOUNT_REMOUNTING: + log_warning_unit(u->id, + "%s remounting timed out. Stopping.", u->id); + m->reload_result = MOUNT_FAILURE_TIMEOUT; + mount_enter_mounted(m, MOUNT_SUCCESS); + break; + + case MOUNT_UNMOUNTING: + log_warning_unit(u->id, + "%s unmounting timed out. Stopping.", u->id); + mount_enter_signal(m, MOUNT_UNMOUNTING_SIGTERM, MOUNT_FAILURE_TIMEOUT); + break; + + case MOUNT_MOUNTING_SIGTERM: + if (m->kill_context.send_sigkill) { + log_warning_unit(u->id, + "%s mounting timed out. Killing.", u->id); + mount_enter_signal(m, MOUNT_MOUNTING_SIGKILL, MOUNT_FAILURE_TIMEOUT); + } else { + log_warning_unit(u->id, + "%s mounting timed out. Skipping SIGKILL. Ignoring.", + u->id); + + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_FAILURE_TIMEOUT); + else + mount_enter_dead(m, MOUNT_FAILURE_TIMEOUT); + } + break; + + case MOUNT_REMOUNTING_SIGTERM: + if (m->kill_context.send_sigkill) { + log_warning_unit(u->id, + "%s remounting timed out. Killing.", u->id); + mount_enter_signal(m, MOUNT_REMOUNTING_SIGKILL, MOUNT_FAILURE_TIMEOUT); + } else { + log_warning_unit(u->id, + "%s remounting timed out. Skipping SIGKILL. Ignoring.", + u->id); + + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_FAILURE_TIMEOUT); + else + mount_enter_dead(m, MOUNT_FAILURE_TIMEOUT); + } + break; + + case MOUNT_UNMOUNTING_SIGTERM: + if (m->kill_context.send_sigkill) { + log_warning_unit(u->id, + "%s unmounting timed out. Killing.", u->id); + mount_enter_signal(m, MOUNT_UNMOUNTING_SIGKILL, MOUNT_FAILURE_TIMEOUT); + } else { + log_warning_unit(u->id, + "%s unmounting timed out. Skipping SIGKILL. Ignoring.", + u->id); + + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_FAILURE_TIMEOUT); + else + mount_enter_dead(m, MOUNT_FAILURE_TIMEOUT); + } + break; + + case MOUNT_MOUNTING_SIGKILL: + case MOUNT_REMOUNTING_SIGKILL: + case MOUNT_UNMOUNTING_SIGKILL: + log_warning_unit(u->id, + "%s mount process still around after SIGKILL. Ignoring.", + u->id); + + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_FAILURE_TIMEOUT); + else + mount_enter_dead(m, MOUNT_FAILURE_TIMEOUT); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } +} + +static int mount_add_one( + Manager *m, + const char *what, + const char *where, + const char *options, + const char *fstype, + int passno, + bool set_flags) { + int r; + Unit *u; + bool delete; + char *e, *w = NULL, *o = NULL, *f = NULL; + MountParameters *p; + bool load_extras = false; + + assert(m); + assert(what); + assert(where); + assert(options); + assert(fstype); + + /* Ignore API mount points. They should never be referenced in + * dependencies ever. */ + if (mount_point_is_api(where) || mount_point_ignore(where)) + return 0; + + if (streq(fstype, "autofs")) + return 0; + + /* probably some kind of swap, ignore */ + if (!is_path(where)) + return 0; + + e = unit_name_from_path(where, ".mount"); + if (!e) + return -ENOMEM; + + u = manager_get_unit(m, e); + if (!u) { + delete = true; + + u = unit_new(m, sizeof(Mount)); + if (!u) { + free(e); + return -ENOMEM; + } + + r = unit_add_name(u, e); + free(e); + + if (r < 0) + goto fail; + + MOUNT(u)->where = strdup(where); + if (!MOUNT(u)->where) { + r = -ENOMEM; + goto fail; + } + + u->source_path = strdup("/proc/self/mountinfo"); + if (!u->source_path) { + r = -ENOMEM; + goto fail; + } + + r = unit_add_dependency_by_name(u, UNIT_BEFORE, SPECIAL_LOCAL_FS_TARGET, NULL, true); + if (r < 0) + goto fail; + + r = unit_add_dependency_by_name(u, UNIT_CONFLICTS, SPECIAL_UMOUNT_TARGET, NULL, true); + if (r < 0) + goto fail; + + unit_add_to_load_queue(u); + } else { + delete = false; + free(e); + + if (!MOUNT(u)->where) { + MOUNT(u)->where = strdup(where); + if (!MOUNT(u)->where) { + r = -ENOMEM; + goto fail; + } + } + + if (u->load_state == UNIT_ERROR) { + u->load_state = UNIT_LOADED; + u->load_error = 0; + + /* Load in the extras later on, after we + * finished initialization of the unit */ + load_extras = true; + } + } + + if (!(w = strdup(what)) || + !(o = strdup(options)) || + !(f = strdup(fstype))) { + r = -ENOMEM; + goto fail; + } + + p = &MOUNT(u)->parameters_proc_self_mountinfo; + if (set_flags) { + MOUNT(u)->is_mounted = true; + MOUNT(u)->just_mounted = !MOUNT(u)->from_proc_self_mountinfo; + MOUNT(u)->just_changed = !streq_ptr(p->options, o); + } + + MOUNT(u)->from_proc_self_mountinfo = true; + + free(p->what); + p->what = w; + + free(p->options); + p->options = o; + + free(p->fstype); + p->fstype = f; + + p->passno = passno; + + if (load_extras) { + r = mount_add_extras(MOUNT(u)); + if (r < 0) + goto fail; + } + + unit_add_to_dbus_queue(u); + + return 0; + +fail: + free(w); + free(o); + free(f); + + if (delete && u) + unit_free(u); + + return r; +} + +static int mount_load_proc_self_mountinfo(Manager *m, bool set_flags) { + int r = 0; + unsigned i; + char *device, *path, *options, *options2, *fstype, *d, *p, *o; + + assert(m); + + rewind(m->proc_self_mountinfo); + + for (i = 1;; i++) { + int k; + + device = path = options = options2 = fstype = d = p = o = NULL; + + if ((k = fscanf(m->proc_self_mountinfo, + "%*s " /* (1) mount id */ + "%*s " /* (2) parent id */ + "%*s " /* (3) major:minor */ + "%*s " /* (4) root */ + "%ms " /* (5) mount point */ + "%ms" /* (6) mount options */ + "%*[^-]" /* (7) optional fields */ + "- " /* (8) separator */ + "%ms " /* (9) file system type */ + "%ms" /* (10) mount source */ + "%ms" /* (11) mount options 2 */ + "%*[^\n]", /* some rubbish at the end */ + &path, + &options, + &fstype, + &device, + &options2)) != 5) { + + if (k == EOF) + break; + + log_warning("Failed to parse /proc/self/mountinfo:%u.", i); + goto clean_up; + } + + o = strjoin(options, ",", options2, NULL); + if (!o) { + r = -ENOMEM; + goto finish; + } + + if (!(d = cunescape(device)) || + !(p = cunescape(path))) { + r = -ENOMEM; + goto finish; + } + + if ((k = mount_add_one(m, d, p, o, fstype, 0, set_flags)) < 0) + r = k; + +clean_up: + free(device); + free(path); + free(options); + free(options2); + free(fstype); + free(d); + free(p); + free(o); + } + +finish: + free(device); + free(path); + free(options); + free(options2); + free(fstype); + free(d); + free(p); + free(o); + + return r; +} + +static void mount_shutdown(Manager *m) { + assert(m); + + if (m->proc_self_mountinfo) { + fclose(m->proc_self_mountinfo); + m->proc_self_mountinfo = NULL; + } +} + +static int mount_enumerate(Manager *m) { + int r; + assert(m); + + if (!m->proc_self_mountinfo) { + struct epoll_event ev = { + .events = EPOLLPRI, + .data.ptr = &m->mount_watch, + }; + + m->proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); + if (!m->proc_self_mountinfo) + return -errno; + + m->mount_watch.type = WATCH_MOUNT; + m->mount_watch.fd = fileno(m->proc_self_mountinfo); + + if (epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, m->mount_watch.fd, &ev) < 0) + return -errno; + } + + r = mount_load_proc_self_mountinfo(m, false); + if (r < 0) + goto fail; + + return 0; + +fail: + mount_shutdown(m); + return r; +} + +void mount_fd_event(Manager *m, int events) { + Unit *u; + int r; + + assert(m); + assert(events & EPOLLPRI); + + /* The manager calls this for every fd event happening on the + * /proc/self/mountinfo file, which informs us about mounting + * table changes */ + + r = mount_load_proc_self_mountinfo(m, true); + if (r < 0) { + log_error("Failed to reread /proc/self/mountinfo: %s", strerror(-r)); + + /* Reset flags, just in case, for later calls */ + LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_MOUNT]) { + Mount *mount = MOUNT(u); + + mount->is_mounted = mount->just_mounted = mount->just_changed = false; + } + + return; + } + + manager_dispatch_load_queue(m); + + LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_MOUNT]) { + Mount *mount = MOUNT(u); + + if (!mount->is_mounted) { + /* This has just been unmounted. */ + + mount->from_proc_self_mountinfo = false; + + switch (mount->state) { + + case MOUNT_MOUNTED: + mount_enter_dead(mount, MOUNT_SUCCESS); + break; + + default: + mount_set_state(mount, mount->state); + break; + + } + + } else if (mount->just_mounted || mount->just_changed) { + + /* New or changed mount entry */ + + switch (mount->state) { + + case MOUNT_DEAD: + case MOUNT_FAILED: + mount_enter_mounted(mount, MOUNT_SUCCESS); + break; + + case MOUNT_MOUNTING: + mount_enter_mounting_done(mount); + break; + + default: + /* Nothing really changed, but let's + * issue an notification call + * nonetheless, in case somebody is + * waiting for this. (e.g. file system + * ro/rw remounts.) */ + mount_set_state(mount, mount->state); + break; + } + } + + /* Reset the flags for later calls */ + mount->is_mounted = mount->just_mounted = mount->just_changed = false; + } +} + +static void mount_reset_failed(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + if (m->state == MOUNT_FAILED) + mount_set_state(m, MOUNT_DEAD); + + m->result = MOUNT_SUCCESS; + m->reload_result = MOUNT_SUCCESS; +} + +static int mount_kill(Unit *u, KillWho who, int signo, DBusError *error) { + return unit_kill_common(u, who, signo, -1, MOUNT(u)->control_pid, error); +} + +static const char* const mount_state_table[_MOUNT_STATE_MAX] = { + [MOUNT_DEAD] = "dead", + [MOUNT_MOUNTING] = "mounting", + [MOUNT_MOUNTING_DONE] = "mounting-done", + [MOUNT_MOUNTED] = "mounted", + [MOUNT_REMOUNTING] = "remounting", + [MOUNT_UNMOUNTING] = "unmounting", + [MOUNT_MOUNTING_SIGTERM] = "mounting-sigterm", + [MOUNT_MOUNTING_SIGKILL] = "mounting-sigkill", + [MOUNT_REMOUNTING_SIGTERM] = "remounting-sigterm", + [MOUNT_REMOUNTING_SIGKILL] = "remounting-sigkill", + [MOUNT_UNMOUNTING_SIGTERM] = "unmounting-sigterm", + [MOUNT_UNMOUNTING_SIGKILL] = "unmounting-sigkill", + [MOUNT_FAILED] = "failed" +}; + +DEFINE_STRING_TABLE_LOOKUP(mount_state, MountState); + +static const char* const mount_exec_command_table[_MOUNT_EXEC_COMMAND_MAX] = { + [MOUNT_EXEC_MOUNT] = "ExecMount", + [MOUNT_EXEC_UNMOUNT] = "ExecUnmount", + [MOUNT_EXEC_REMOUNT] = "ExecRemount", +}; + +DEFINE_STRING_TABLE_LOOKUP(mount_exec_command, MountExecCommand); + +static const char* const mount_result_table[_MOUNT_RESULT_MAX] = { + [MOUNT_SUCCESS] = "success", + [MOUNT_FAILURE_RESOURCES] = "resources", + [MOUNT_FAILURE_TIMEOUT] = "timeout", + [MOUNT_FAILURE_EXIT_CODE] = "exit-code", + [MOUNT_FAILURE_SIGNAL] = "signal", + [MOUNT_FAILURE_CORE_DUMP] = "core-dump" +}; + +DEFINE_STRING_TABLE_LOOKUP(mount_result, MountResult); + +const UnitVTable mount_vtable = { + .object_size = sizeof(Mount), + + .sections = + "Unit\0" + "Mount\0" + "Install\0", + + .exec_context_offset = offsetof(Mount, exec_context), + .exec_section = "Mount", + + .no_alias = true, + .no_instances = true, + + .init = mount_init, + .load = mount_load, + .done = mount_done, + + .coldplug = mount_coldplug, + + .dump = mount_dump, + + .start = mount_start, + .stop = mount_stop, + .reload = mount_reload, + + .kill = mount_kill, + + .serialize = mount_serialize, + .deserialize_item = mount_deserialize_item, + + .active_state = mount_active_state, + .sub_state_to_string = mount_sub_state_to_string, + + .check_gc = mount_check_gc, + + .sigchld_event = mount_sigchld_event, + .timer_event = mount_timer_event, + + .reset_failed = mount_reset_failed, + + .bus_interface = "org.freedesktop.systemd1.Mount", + .bus_message_handler = bus_mount_message_handler, + .bus_invalidating_properties = bus_mount_invalidating_properties, + + .enumerate = mount_enumerate, + .shutdown = mount_shutdown, + + .status_message_formats = { + .starting_stopping = { + [0] = "Mounting %s...", + [1] = "Unmounting %s...", + }, + .finished_start_job = { + [JOB_DONE] = "Mounted %s.", + [JOB_FAILED] = "Failed to mount %s.", + [JOB_DEPENDENCY] = "Dependency failed for %s.", + [JOB_TIMEOUT] = "Timed out mounting %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Unmounted %s.", + [JOB_FAILED] = "Failed unmounting %s.", + [JOB_TIMEOUT] = "Timed out unmounting %s.", + }, + }, +}; diff --git a/src/core/mount.h b/src/core/mount.h new file mode 100644 index 0000000000..bcc10ee0d4 --- /dev/null +++ b/src/core/mount.h @@ -0,0 +1,123 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Mount Mount; + +#include "unit.h" +#include "kill.h" + +typedef enum MountState { + MOUNT_DEAD, + MOUNT_MOUNTING, /* /bin/mount is running, but the mount is not done yet. */ + MOUNT_MOUNTING_DONE, /* /bin/mount is running, and the mount is done. */ + MOUNT_MOUNTED, + MOUNT_REMOUNTING, + MOUNT_UNMOUNTING, + MOUNT_MOUNTING_SIGTERM, + MOUNT_MOUNTING_SIGKILL, + MOUNT_REMOUNTING_SIGTERM, + MOUNT_REMOUNTING_SIGKILL, + MOUNT_UNMOUNTING_SIGTERM, + MOUNT_UNMOUNTING_SIGKILL, + MOUNT_FAILED, + _MOUNT_STATE_MAX, + _MOUNT_STATE_INVALID = -1 +} MountState; + +typedef enum MountExecCommand { + MOUNT_EXEC_MOUNT, + MOUNT_EXEC_UNMOUNT, + MOUNT_EXEC_REMOUNT, + _MOUNT_EXEC_COMMAND_MAX, + _MOUNT_EXEC_COMMAND_INVALID = -1 +} MountExecCommand; + +typedef struct MountParameters { + char *what; + char *options; + char *fstype; + int passno; +} MountParameters; + +typedef enum MountResult { + MOUNT_SUCCESS, + MOUNT_FAILURE_RESOURCES, + MOUNT_FAILURE_TIMEOUT, + MOUNT_FAILURE_EXIT_CODE, + MOUNT_FAILURE_SIGNAL, + MOUNT_FAILURE_CORE_DUMP, + _MOUNT_RESULT_MAX, + _MOUNT_RESULT_INVALID = -1 +} MountResult; + +struct Mount { + Unit meta; + + char *where; + + MountParameters parameters_proc_self_mountinfo; + MountParameters parameters_fragment; + + bool from_proc_self_mountinfo:1; + bool from_fragment:1; + + /* Used while looking for mount points that vanished or got + * added from/to /proc/self/mountinfo */ + bool is_mounted:1; + bool just_mounted:1; + bool just_changed:1; + + MountResult result; + MountResult reload_result; + + mode_t directory_mode; + + usec_t timeout_usec; + + ExecCommand exec_command[_MOUNT_EXEC_COMMAND_MAX]; + ExecContext exec_context; + KillContext kill_context; + + MountState state, deserialized_state; + + ExecCommand* control_command; + MountExecCommand control_command_id; + pid_t control_pid; + + Watch timer_watch; +}; + +extern const UnitVTable mount_vtable; + +void mount_fd_event(Manager *m, int events); + +const char* mount_state_to_string(MountState i) _const_; +MountState mount_state_from_string(const char *s) _pure_; + +const char* mount_exec_command_to_string(MountExecCommand i) _const_; +MountExecCommand mount_exec_command_from_string(const char *s) _pure_; + +const char* mount_result_to_string(MountResult i) _const_; +MountResult mount_result_from_string(const char *s) _pure_; + +void warn_if_dir_nonempty(const char *unit, const char* where); diff --git a/src/core/namespace.c b/src/core/namespace.c new file mode 100644 index 0000000000..7e33d84156 --- /dev/null +++ b/src/core/namespace.c @@ -0,0 +1,277 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <sys/mount.h> +#include <string.h> +#include <stdio.h> +#include <unistd.h> +#include <sys/stat.h> +#include <sys/types.h> +#include <sched.h> +#include <sys/syscall.h> +#include <limits.h> +#include <linux/fs.h> + +#include "strv.h" +#include "util.h" +#include "path-util.h" +#include "namespace.h" +#include "missing.h" +#include "execute.h" + +typedef enum MountMode { + /* This is ordered by priority! */ + INACCESSIBLE, + READONLY, + PRIVATE_TMP, + PRIVATE_VAR_TMP, + READWRITE +} MountMode; + +typedef struct BindMount { + const char *path; + MountMode mode; + bool done; +} BindMount; + +static int append_mounts(BindMount **p, char **strv, MountMode mode) { + char **i; + + STRV_FOREACH(i, strv) { + + if (!path_is_absolute(*i)) + return -EINVAL; + + (*p)->path = *i; + (*p)->mode = mode; + (*p)++; + } + + return 0; +} + +static int mount_path_compare(const void *a, const void *b) { + const BindMount *p = a, *q = b; + + if (path_equal(p->path, q->path)) { + + /* If the paths are equal, check the mode */ + if (p->mode < q->mode) + return -1; + + if (p->mode > q->mode) + return 1; + + return 0; + } + + /* If the paths are not equal, then order prefixes first */ + if (path_startswith(p->path, q->path)) + return 1; + + if (path_startswith(q->path, p->path)) + return -1; + + return 0; +} + +static void drop_duplicates(BindMount *m, unsigned *n) { + BindMount *f, *t, *previous; + + assert(m); + assert(n); + + for (f = m, t = m, previous = NULL; f < m+*n; f++) { + + /* The first one wins */ + if (previous && path_equal(f->path, previous->path)) + continue; + + t->path = f->path; + t->mode = f->mode; + + previous = t; + + t++; + } + + *n = t - m; +} + +static int apply_mount( + BindMount *m, + const char *tmp_dir, + const char *var_tmp_dir) { + + const char *what; + int r; + + assert(m); + + switch (m->mode) { + + case INACCESSIBLE: + what = "/run/systemd/inaccessible"; + break; + + case READONLY: + case READWRITE: + what = m->path; + break; + + case PRIVATE_TMP: + what = tmp_dir; + break; + + case PRIVATE_VAR_TMP: + what = var_tmp_dir; + break; + + default: + assert_not_reached("Unknown mode"); + } + + assert(what); + + r = mount(what, m->path, NULL, MS_BIND|MS_REC, NULL); + if (r >= 0) + log_debug("Successfully mounted %s to %s", what, m->path); + + return r; +} + +static int make_read_only(BindMount *m) { + int r; + + assert(m); + + if (m->mode != INACCESSIBLE && m->mode != READONLY) + return 0; + + r = mount(NULL, m->path, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL); + if (r < 0) + return -errno; + + return 0; +} + +int setup_tmpdirs(char **tmp_dir, + char **var_tmp_dir) { + int r = 0; + char tmp_dir_template[] = "/tmp/systemd-private-XXXXXX", + var_tmp_dir_template[] = "/var/tmp/systemd-private-XXXXXX"; + + assert(tmp_dir); + assert(var_tmp_dir); + + r = create_tmp_dir(tmp_dir_template, tmp_dir); + if (r < 0) + return r; + + r = create_tmp_dir(var_tmp_dir_template, var_tmp_dir); + if (r == 0) + return 0; + + /* failure */ + rmdir(*tmp_dir); + rmdir(tmp_dir_template); + free(*tmp_dir); + *tmp_dir = NULL; + + return r; +} + +int setup_namespace(char** read_write_dirs, + char** read_only_dirs, + char** inaccessible_dirs, + char* tmp_dir, + char* var_tmp_dir, + bool private_tmp, + unsigned mount_flags) { + + unsigned n = strv_length(read_write_dirs) + + strv_length(read_only_dirs) + + strv_length(inaccessible_dirs) + + (private_tmp ? 2 : 0); + BindMount *m, *mounts; + int r = 0; + + if (!mount_flags) + mount_flags = MS_SHARED; + + if (unshare(CLONE_NEWNS) < 0) + return -errno; + + m = mounts = (BindMount *) alloca(n * sizeof(BindMount)); + if ((r = append_mounts(&m, read_write_dirs, READWRITE)) < 0 || + (r = append_mounts(&m, read_only_dirs, READONLY)) < 0 || + (r = append_mounts(&m, inaccessible_dirs, INACCESSIBLE)) < 0) + return r; + + if (private_tmp) { + m->path = "/tmp"; + m->mode = PRIVATE_TMP; + m++; + + m->path = "/var/tmp"; + m->mode = PRIVATE_VAR_TMP; + m++; + } + + assert(mounts + n == m); + + qsort(mounts, n, sizeof(BindMount), mount_path_compare); + drop_duplicates(mounts, &n); + + /* Remount / as SLAVE so that nothing now mounted in the namespace + shows up in the parent */ + if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) + return -errno; + + for (m = mounts; m < mounts + n; ++m) { + r = apply_mount(m, tmp_dir, var_tmp_dir); + if (r < 0) + goto undo_mounts; + } + + for (m = mounts; m < mounts + n; ++m) { + r = make_read_only(m); + if (r < 0) + goto undo_mounts; + } + + /* Remount / as the desired mode */ + if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { + r = -errno; + goto undo_mounts; + } + + return 0; + +undo_mounts: + for (m = mounts; m < mounts + n; ++m) { + if (m->done) + umount2(m->path, MNT_DETACH); + } + + return r; +} diff --git a/src/core/namespace.h b/src/core/namespace.h new file mode 100644 index 0000000000..ddb579468d --- /dev/null +++ b/src/core/namespace.h @@ -0,0 +1,33 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +int setup_tmpdirs(char **tmp_dir, char **var_tmp_dir); +int setup_namespace(char **read_write_dirs, + char **read_only_dirs, + char **inaccessible_dirs, + char *tmp_dir, + char *var_tmp_dir, + bool private_tmp, + unsigned mount_flags); diff --git a/src/core/org.freedesktop.systemd1.conf b/src/core/org.freedesktop.systemd1.conf new file mode 100644 index 0000000000..a07a8e1ce3 --- /dev/null +++ b/src/core/org.freedesktop.systemd1.conf @@ -0,0 +1,92 @@ +<?xml version="1.0"?> <!--*-nxml-*--> +<!DOCTYPE busconfig PUBLIC "-//freedesktop//DTD D-BUS Bus Configuration 1.0//EN" + "http://www.freedesktop.org/standards/dbus/1.0/busconfig.dtd"> + +<!-- + This file is part of systemd. + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. +--> + +<busconfig> + + <policy user="root"> + <allow own="org.freedesktop.systemd1"/> + + <!-- Root clients can do everything --> + <allow send_destination="org.freedesktop.systemd1"/> + <allow receive_sender="org.freedesktop.systemd1"/> + + <!-- systemd may receive activator requests --> + <allow receive_interface="org.freedesktop.systemd1.Activator" + receive_member="ActivationRequest"/> + </policy> + + <policy context="default"> + <deny send_destination="org.freedesktop.systemd1"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.DBus.Introspectable"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.DBus.Peer"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.DBus.Properties" + send_member="Get"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.DBus.Properties" + send_member="GetAll"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="GetUnit"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="GetUnitByPID"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="LoadUnit"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="GetJob"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="ListUnits"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="ListUnitFiles"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="GetUnitFileState"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="ListJobs"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="Subscribe"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="Unsubscribe"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="Dump"/> + + <allow receive_sender="org.freedesktop.systemd1"/> + </policy> + +</busconfig> diff --git a/src/core/org.freedesktop.systemd1.policy.in.in b/src/core/org.freedesktop.systemd1.policy.in.in new file mode 100644 index 0000000000..51bdafac45 --- /dev/null +++ b/src/core/org.freedesktop.systemd1.policy.in.in @@ -0,0 +1,41 @@ +<?xml version="1.0" encoding="UTF-8"?> <!--*-nxml-*--> +<!DOCTYPE policyconfig PUBLIC "-//freedesktop//DTD PolicyKit Policy Configuration 1.0//EN" + "http://www.freedesktop.org/standards/PolicyKit/1/policyconfig.dtd"> + +<!-- + This file is part of systemd. + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. +--> + +<policyconfig> + + <vendor>The systemd Project</vendor> + <vendor_url>http://www.freedesktop.org/wiki/Software/systemd</vendor_url> + + <action id="org.freedesktop.systemd1.reply-password"> + <_description>Send passphrase back to system</_description> + <_message>Authentication is required to send the entered passphrase back to the system.</_message> + <defaults> + <allow_any>no</allow_any> + <allow_inactive>no</allow_inactive> + <allow_active>auth_admin_keep</allow_active> + </defaults> + <annotate key="org.freedesktop.policykit.exec.path">@rootlibexecdir@/systemd-reply-password</annotate> + </action> + + <action id="org.freedesktop.systemd1.bus-access"> + <_description>Privileged system and service manager access</_description> + <_message>Authentication is required to access the system and service manager.</_message> + <defaults> + <allow_any>no</allow_any> + <allow_inactive>no</allow_inactive> + <allow_active>auth_admin_keep</allow_active> + </defaults> + <annotate key="org.freedesktop.policykit.exec.path">@bindir@/systemd-stdio-bridge</annotate> + </action> + +</policyconfig> diff --git a/src/core/org.freedesktop.systemd1.service b/src/core/org.freedesktop.systemd1.service new file mode 100644 index 0000000000..d4df3e93a2 --- /dev/null +++ b/src/core/org.freedesktop.systemd1.service @@ -0,0 +1,11 @@ +# This file is part of systemd. +# +# systemd is free software; you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published by +# the Free Software Foundation; either version 2.1 of the License, or +# (at your option) any later version. + +[D-BUS Service] +Name=org.freedesktop.systemd1 +Exec=/bin/false +User=root diff --git a/src/core/path.c b/src/core/path.c new file mode 100644 index 0000000000..8a09deb4ff --- /dev/null +++ b/src/core/path.c @@ -0,0 +1,833 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/inotify.h> +#include <sys/epoll.h> +#include <sys/ioctl.h> +#include <errno.h> +#include <unistd.h> + +#include "unit.h" +#include "unit-name.h" +#include "path.h" +#include "mkdir.h" +#include "dbus-path.h" +#include "special.h" +#include "dbus-common.h" +#include "path-util.h" +#include "macro.h" + +static const UnitActiveState state_translation_table[_PATH_STATE_MAX] = { + [PATH_DEAD] = UNIT_INACTIVE, + [PATH_WAITING] = UNIT_ACTIVE, + [PATH_RUNNING] = UNIT_ACTIVE, + [PATH_FAILED] = UNIT_FAILED +}; + +int path_spec_watch(PathSpec *s, Unit *u) { + + static const int flags_table[_PATH_TYPE_MAX] = { + [PATH_EXISTS] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB, + [PATH_EXISTS_GLOB] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB, + [PATH_CHANGED] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CLOSE_WRITE|IN_CREATE|IN_DELETE|IN_MOVED_FROM|IN_MOVED_TO, + [PATH_MODIFIED] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CLOSE_WRITE|IN_CREATE|IN_DELETE|IN_MOVED_FROM|IN_MOVED_TO|IN_MODIFY, + [PATH_DIRECTORY_NOT_EMPTY] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CREATE|IN_MOVED_TO + }; + + bool exists = false; + char *slash, *oldslash = NULL; + int r; + + assert(u); + assert(s); + + path_spec_unwatch(s, u); + + s->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC); + if (s->inotify_fd < 0) { + r = -errno; + goto fail; + } + + r = unit_watch_fd(u, s->inotify_fd, EPOLLIN, &s->watch); + if (r < 0) + goto fail; + + /* This assumes the path was passed through path_kill_slashes()! */ + + for (slash = strchr(s->path, '/'); ; slash = strchr(slash+1, '/')) { + char *cut = NULL; + int flags; + char tmp; + + if (slash) { + cut = slash + (slash == s->path); + tmp = *cut; + *cut = '\0'; + + flags = IN_MOVE_SELF | IN_DELETE_SELF | IN_ATTRIB | IN_CREATE | IN_MOVED_TO; + } else + flags = flags_table[s->type]; + + r = inotify_add_watch(s->inotify_fd, s->path, flags); + if (r < 0) { + if (errno == EACCES || errno == ENOENT) { + if (cut) + *cut = tmp; + break; + } + + log_warning("Failed to add watch on %s: %m", s->path); + r = -errno; + if (cut) + *cut = tmp; + goto fail; + } else { + exists = true; + + /* Path exists, we don't need to watch parent + too closely. */ + if (oldslash) { + char *cut2 = oldslash + (oldslash == s->path); + char tmp2 = *cut2; + *cut2 = '\0'; + + inotify_add_watch(s->inotify_fd, s->path, IN_MOVE_SELF); + /* Error is ignored, the worst can happen is + we get spurious events. */ + + *cut2 = tmp2; + } + } + + if (cut) + *cut = tmp; + + if (slash) + oldslash = slash; + else { + /* whole path has been iterated over */ + s->primary_wd = r; + break; + } + } + + if (!exists) { + log_error("Failed to add watch on any of the components of %s: %m", + s->path); + r = -errno; /* either EACCESS or ENOENT */ + goto fail; + } + + return 0; + +fail: + path_spec_unwatch(s, u); + return r; +} + +void path_spec_unwatch(PathSpec *s, Unit *u) { + + if (s->inotify_fd < 0) + return; + + unit_unwatch_fd(u, &s->watch); + + close_nointr_nofail(s->inotify_fd); + s->inotify_fd = -1; +} + +int path_spec_fd_event(PathSpec *s, uint32_t events) { + _cleanup_free_ uint8_t *buf = NULL; + struct inotify_event *e; + ssize_t k; + int l; + int r = 0; + + if (events != EPOLLIN) { + log_error("Got invalid poll event on inotify."); + return -EINVAL; + } + + if (ioctl(s->inotify_fd, FIONREAD, &l) < 0) { + log_error("FIONREAD failed: %m"); + return -errno; + } + + assert(l > 0); + + buf = malloc(l); + if (!buf) + return log_oom(); + + k = read(s->inotify_fd, buf, l); + if (k < 0) { + log_error("Failed to read inotify event: %m"); + return -errno; + } + + e = (struct inotify_event*) buf; + + while (k > 0) { + size_t step; + + if ((s->type == PATH_CHANGED || s->type == PATH_MODIFIED) && + s->primary_wd == e->wd) + r = 1; + + step = sizeof(struct inotify_event) + e->len; + assert(step <= (size_t) k); + + e = (struct inotify_event*) ((uint8_t*) e + step); + k -= step; + } + + return r; +} + +static bool path_spec_check_good(PathSpec *s, bool initial) { + bool good = false; + + switch (s->type) { + + case PATH_EXISTS: + good = access(s->path, F_OK) >= 0; + break; + + case PATH_EXISTS_GLOB: + good = glob_exists(s->path) > 0; + break; + + case PATH_DIRECTORY_NOT_EMPTY: { + int k; + + k = dir_is_empty(s->path); + good = !(k == -ENOENT || k > 0); + break; + } + + case PATH_CHANGED: + case PATH_MODIFIED: { + bool b; + + b = access(s->path, F_OK) >= 0; + good = !initial && b != s->previous_exists; + s->previous_exists = b; + break; + } + + default: + ; + } + + return good; +} + +static bool path_spec_startswith(PathSpec *s, const char *what) { + return path_startswith(s->path, what); +} + +static void path_spec_mkdir(PathSpec *s, mode_t mode) { + int r; + + if (s->type == PATH_EXISTS || s->type == PATH_EXISTS_GLOB) + return; + + r = mkdir_p_label(s->path, mode); + if (r < 0) + log_warning("mkdir(%s) failed: %s", s->path, strerror(-r)); +} + +static void path_spec_dump(PathSpec *s, FILE *f, const char *prefix) { + fprintf(f, + "%s%s: %s\n", + prefix, + path_type_to_string(s->type), + s->path); +} + +void path_spec_done(PathSpec *s) { + assert(s); + assert(s->inotify_fd == -1); + + free(s->path); +} + +static void path_init(Unit *u) { + Path *p = PATH(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + p->directory_mode = 0755; +} + +void path_free_specs(Path *p) { + PathSpec *s; + + assert(p); + + while ((s = p->specs)) { + path_spec_unwatch(s, UNIT(p)); + LIST_REMOVE(PathSpec, spec, p->specs, s); + path_spec_done(s); + free(s); + } +} + +static void path_done(Unit *u) { + Path *p = PATH(u); + + assert(p); + + path_free_specs(p); +} + +int path_add_one_mount_link(Path *p, Mount *m) { + PathSpec *s; + int r; + + assert(p); + assert(m); + + if (UNIT(p)->load_state != UNIT_LOADED || + UNIT(m)->load_state != UNIT_LOADED) + return 0; + + LIST_FOREACH(spec, s, p->specs) { + if (!path_spec_startswith(s, m->where)) + continue; + + r = unit_add_two_dependencies(UNIT(p), UNIT_AFTER, UNIT_REQUIRES, + UNIT(m), true); + if (r < 0) + return r; + } + + return 0; +} + +static int path_add_mount_links(Path *p) { + Unit *other; + int r; + + assert(p); + + LIST_FOREACH(units_by_type, other, UNIT(p)->manager->units_by_type[UNIT_MOUNT]) { + r = path_add_one_mount_link(p, MOUNT(other)); + if (r < 0) + return r; + } + + return 0; +} + +static int path_verify(Path *p) { + assert(p); + + if (UNIT(p)->load_state != UNIT_LOADED) + return 0; + + if (!p->specs) { + log_error_unit(UNIT(p)->id, + "%s lacks path setting. Refusing.", UNIT(p)->id); + return -EINVAL; + } + + return 0; +} + +static int path_add_default_dependencies(Path *p) { + int r; + + assert(p); + + r = unit_add_dependency_by_name(UNIT(p), UNIT_BEFORE, + SPECIAL_PATHS_TARGET, NULL, true); + if (r < 0) + return r; + + if (UNIT(p)->manager->running_as == SYSTEMD_SYSTEM) { + r = unit_add_two_dependencies_by_name(UNIT(p), UNIT_AFTER, UNIT_REQUIRES, + SPECIAL_SYSINIT_TARGET, NULL, true); + if (r < 0) + return r; + } + + return unit_add_two_dependencies_by_name(UNIT(p), UNIT_BEFORE, UNIT_CONFLICTS, + SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +static int path_load(Unit *u) { + Path *p = PATH(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + r = unit_load_fragment_and_dropin(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + + if (set_isempty(u->dependencies[UNIT_TRIGGERS])) { + Unit *x; + + r = unit_load_related_unit(u, ".service", &x); + if (r < 0) + return r; + + r = unit_add_two_dependencies(u, UNIT_BEFORE, UNIT_TRIGGERS, x, true); + if (r < 0) + return r; + } + + r = path_add_mount_links(p); + if (r < 0) + return r; + + if (UNIT(p)->default_dependencies) { + r = path_add_default_dependencies(p); + if (r < 0) + return r; + } + } + + return path_verify(p); +} + +static void path_dump(Unit *u, FILE *f, const char *prefix) { + Path *p = PATH(u); + Unit *trigger; + PathSpec *s; + + assert(p); + assert(f); + + trigger = UNIT_TRIGGER(u); + + fprintf(f, + "%sPath State: %s\n" + "%sResult: %s\n" + "%sUnit: %s\n" + "%sMakeDirectory: %s\n" + "%sDirectoryMode: %04o\n", + prefix, path_state_to_string(p->state), + prefix, path_result_to_string(p->result), + prefix, trigger ? trigger->id : "n/a", + prefix, yes_no(p->make_directory), + prefix, p->directory_mode); + + LIST_FOREACH(spec, s, p->specs) + path_spec_dump(s, f, prefix); +} + +static void path_unwatch(Path *p) { + PathSpec *s; + + assert(p); + + LIST_FOREACH(spec, s, p->specs) + path_spec_unwatch(s, UNIT(p)); +} + +static int path_watch(Path *p) { + int r; + PathSpec *s; + + assert(p); + + LIST_FOREACH(spec, s, p->specs) { + r = path_spec_watch(s, UNIT(p)); + if (r < 0) + return r; + } + + return 0; +} + +static void path_set_state(Path *p, PathState state) { + PathState old_state; + assert(p); + + old_state = p->state; + p->state = state; + + if (state != PATH_WAITING && + (state != PATH_RUNNING || p->inotify_triggered)) + path_unwatch(p); + + if (state != old_state) + log_debug("%s changed %s -> %s", + UNIT(p)->id, + path_state_to_string(old_state), + path_state_to_string(state)); + + unit_notify(UNIT(p), state_translation_table[old_state], state_translation_table[state], true); +} + +static void path_enter_waiting(Path *p, bool initial, bool recheck); + +static int path_coldplug(Unit *u) { + Path *p = PATH(u); + + assert(p); + assert(p->state == PATH_DEAD); + + if (p->deserialized_state != p->state) { + + if (p->deserialized_state == PATH_WAITING || + p->deserialized_state == PATH_RUNNING) + path_enter_waiting(p, true, true); + else + path_set_state(p, p->deserialized_state); + } + + return 0; +} + +static void path_enter_dead(Path *p, PathResult f) { + assert(p); + + if (f != PATH_SUCCESS) + p->result = f; + + path_set_state(p, p->result != PATH_SUCCESS ? PATH_FAILED : PATH_DEAD); +} + +static void path_enter_running(Path *p) { + _cleanup_dbus_error_free_ DBusError error; + int r; + + assert(p); + + dbus_error_init(&error); + + /* Don't start job if we are supposed to go down */ + if (unit_stop_pending(UNIT(p))) + return; + + r = manager_add_job(UNIT(p)->manager, JOB_START, UNIT_TRIGGER(UNIT(p)), + JOB_REPLACE, true, &error, NULL); + if (r < 0) + goto fail; + + p->inotify_triggered = false; + + r = path_watch(p); + if (r < 0) + goto fail; + + path_set_state(p, PATH_RUNNING); + return; + +fail: + log_warning("%s failed to queue unit startup job: %s", + UNIT(p)->id, bus_error(&error, r)); + path_enter_dead(p, PATH_FAILURE_RESOURCES); +} + +static bool path_check_good(Path *p, bool initial) { + PathSpec *s; + bool good = false; + + assert(p); + + LIST_FOREACH(spec, s, p->specs) { + good = path_spec_check_good(s, initial); + + if (good) + break; + } + + return good; +} + +static void path_enter_waiting(Path *p, bool initial, bool recheck) { + int r; + + if (recheck) + if (path_check_good(p, initial)) { + log_debug("%s got triggered.", UNIT(p)->id); + path_enter_running(p); + return; + } + + r = path_watch(p); + if (r < 0) + goto fail; + + /* Hmm, so now we have created inotify watches, but the file + * might have appeared/been removed by now, so we must + * recheck */ + + if (recheck) + if (path_check_good(p, false)) { + log_debug("%s got triggered.", UNIT(p)->id); + path_enter_running(p); + return; + } + + path_set_state(p, PATH_WAITING); + return; + +fail: + log_warning("%s failed to enter waiting state: %s", + UNIT(p)->id, strerror(-r)); + path_enter_dead(p, PATH_FAILURE_RESOURCES); +} + +static void path_mkdir(Path *p) { + PathSpec *s; + + assert(p); + + if (!p->make_directory) + return; + + LIST_FOREACH(spec, s, p->specs) + path_spec_mkdir(s, p->directory_mode); +} + +static int path_start(Unit *u) { + Path *p = PATH(u); + + assert(p); + assert(p->state == PATH_DEAD || p->state == PATH_FAILED); + + if (UNIT_TRIGGER(u)->load_state != UNIT_LOADED) + return -ENOENT; + + path_mkdir(p); + + p->result = PATH_SUCCESS; + path_enter_waiting(p, true, true); + + return 0; +} + +static int path_stop(Unit *u) { + Path *p = PATH(u); + + assert(p); + assert(p->state == PATH_WAITING || p->state == PATH_RUNNING); + + path_enter_dead(p, PATH_SUCCESS); + return 0; +} + +static int path_serialize(Unit *u, FILE *f, FDSet *fds) { + Path *p = PATH(u); + + assert(u); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", path_state_to_string(p->state)); + unit_serialize_item(u, f, "result", path_result_to_string(p->result)); + + return 0; +} + +static int path_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Path *p = PATH(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + PathState state; + + state = path_state_from_string(value); + if (state < 0) + log_debug("Failed to parse state value %s", value); + else + p->deserialized_state = state; + + } else if (streq(key, "result")) { + PathResult f; + + f = path_result_from_string(value); + if (f < 0) + log_debug("Failed to parse result value %s", value); + else if (f != PATH_SUCCESS) + p->result = f; + + } else + log_debug("Unknown serialization key '%s'", key); + + return 0; +} + +_pure_ static UnitActiveState path_active_state(Unit *u) { + assert(u); + + return state_translation_table[PATH(u)->state]; +} + +_pure_ static const char *path_sub_state_to_string(Unit *u) { + assert(u); + + return path_state_to_string(PATH(u)->state); +} + +static void path_fd_event(Unit *u, int fd, uint32_t events, Watch *w) { + Path *p = PATH(u); + PathSpec *s; + int changed; + + assert(p); + assert(fd >= 0); + + if (p->state != PATH_WAITING && + p->state != PATH_RUNNING) + return; + + /* log_debug("inotify wakeup on %s.", u->id); */ + + LIST_FOREACH(spec, s, p->specs) + if (path_spec_owns_inotify_fd(s, fd)) + break; + + if (!s) { + log_error("Got event on unknown fd."); + goto fail; + } + + changed = path_spec_fd_event(s, events); + if (changed < 0) + goto fail; + + /* If we are already running, then remember that one event was + * dispatched so that we restart the service only if something + * actually changed on disk */ + p->inotify_triggered = true; + + if (changed) + path_enter_running(p); + else + path_enter_waiting(p, false, true); + + return; + +fail: + path_enter_dead(p, PATH_FAILURE_RESOURCES); +} + +static void path_trigger_notify(Unit *u, Unit *other) { + Path *p = PATH(u); + + assert(u); + assert(other); + + /* Invoked whenever the unit we trigger changes state or gains + * or loses a job */ + + if (other->load_state != UNIT_LOADED) + return; + + if (p->state == PATH_RUNNING && + UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(other))) { + log_debug_unit(UNIT(p)->id, + "%s got notified about unit deactivation.", + UNIT(p)->id); + + /* Hmm, so inotify was triggered since the + * last activation, so I guess we need to + * recheck what is going on. */ + path_enter_waiting(p, false, p->inotify_triggered); + } +} + +static void path_reset_failed(Unit *u) { + Path *p = PATH(u); + + assert(p); + + if (p->state == PATH_FAILED) + path_set_state(p, PATH_DEAD); + + p->result = PATH_SUCCESS; +} + +static const char* const path_state_table[_PATH_STATE_MAX] = { + [PATH_DEAD] = "dead", + [PATH_WAITING] = "waiting", + [PATH_RUNNING] = "running", + [PATH_FAILED] = "failed" +}; + +DEFINE_STRING_TABLE_LOOKUP(path_state, PathState); + +static const char* const path_type_table[_PATH_TYPE_MAX] = { + [PATH_EXISTS] = "PathExists", + [PATH_EXISTS_GLOB] = "PathExistsGlob", + [PATH_CHANGED] = "PathChanged", + [PATH_MODIFIED] = "PathModified", + [PATH_DIRECTORY_NOT_EMPTY] = "DirectoryNotEmpty" +}; + +DEFINE_STRING_TABLE_LOOKUP(path_type, PathType); + +static const char* const path_result_table[_PATH_RESULT_MAX] = { + [PATH_SUCCESS] = "success", + [PATH_FAILURE_RESOURCES] = "resources" +}; + +DEFINE_STRING_TABLE_LOOKUP(path_result, PathResult); + +const UnitVTable path_vtable = { + .object_size = sizeof(Path), + .sections = + "Unit\0" + "Path\0" + "Install\0", + + .init = path_init, + .done = path_done, + .load = path_load, + + .coldplug = path_coldplug, + + .dump = path_dump, + + .start = path_start, + .stop = path_stop, + + .serialize = path_serialize, + .deserialize_item = path_deserialize_item, + + .active_state = path_active_state, + .sub_state_to_string = path_sub_state_to_string, + + .fd_event = path_fd_event, + + .trigger_notify = path_trigger_notify, + + .reset_failed = path_reset_failed, + + .bus_interface = "org.freedesktop.systemd1.Path", + .bus_message_handler = bus_path_message_handler, + .bus_invalidating_properties = bus_path_invalidating_properties +}; diff --git a/src/core/path.h b/src/core/path.h new file mode 100644 index 0000000000..6adab5897d --- /dev/null +++ b/src/core/path.h @@ -0,0 +1,108 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Path Path; + +#include "unit.h" +#include "mount.h" + +typedef enum PathState { + PATH_DEAD, + PATH_WAITING, + PATH_RUNNING, + PATH_FAILED, + _PATH_STATE_MAX, + _PATH_STATE_INVALID = -1 +} PathState; + +typedef enum PathType { + PATH_EXISTS, + PATH_EXISTS_GLOB, + PATH_DIRECTORY_NOT_EMPTY, + PATH_CHANGED, + PATH_MODIFIED, + _PATH_TYPE_MAX, + _PATH_TYPE_INVALID = -1 +} PathType; + +typedef struct PathSpec { + char *path; + + Watch watch; + + LIST_FIELDS(struct PathSpec, spec); + + PathType type; + int inotify_fd; + int primary_wd; + + bool previous_exists; +} PathSpec; + +int path_spec_watch(PathSpec *s, Unit *u); +void path_spec_unwatch(PathSpec *s, Unit *u); +int path_spec_fd_event(PathSpec *s, uint32_t events); +void path_spec_done(PathSpec *s); + +static inline bool path_spec_owns_inotify_fd(PathSpec *s, int fd) { + return s->inotify_fd == fd; +} + +typedef enum PathResult { + PATH_SUCCESS, + PATH_FAILURE_RESOURCES, + _PATH_RESULT_MAX, + _PATH_RESULT_INVALID = -1 +} PathResult; + +struct Path { + Unit meta; + + LIST_HEAD(PathSpec, specs); + + PathState state, deserialized_state; + + bool inotify_triggered; + + bool make_directory; + mode_t directory_mode; + + PathResult result; +}; + +/* Called from the mount code figure out if a mount is a dependency of + * any of the paths of this path object */ +int path_add_one_mount_link(Path *p, Mount *m); + +void path_free_specs(Path *p); + +extern const UnitVTable path_vtable; + +const char* path_state_to_string(PathState i) _const_; +PathState path_state_from_string(const char *s) _pure_; + +const char* path_type_to_string(PathType i) _const_; +PathType path_type_from_string(const char *s) _pure_; + +const char* path_result_to_string(PathResult i) _const_; +PathResult path_result_from_string(const char *s) _pure_; diff --git a/src/core/securebits.h b/src/core/securebits.h new file mode 100644 index 0000000000..ba0bba5353 --- /dev/null +++ b/src/core/securebits.h @@ -0,0 +1,45 @@ +#ifndef _LINUX_SECUREBITS_H +#define _LINUX_SECUREBITS_H 1 + +/* This is minimal version of Linux' linux/securebits.h header file, + * which is licensed GPL2 */ + +#define SECUREBITS_DEFAULT 0x00000000 + +/* When set UID 0 has no special privileges. When unset, we support + inheritance of root-permissions and suid-root executable under + compatibility mode. We raise the effective and inheritable bitmasks + *of the executable file* if the effective uid of the new process is + 0. If the real uid is 0, we raise the effective (legacy) bit of the + executable file. */ +#define SECURE_NOROOT 0 +#define SECURE_NOROOT_LOCKED 1 /* make bit-0 immutable */ + +/* When set, setuid to/from uid 0 does not trigger capability-"fixup". + When unset, to provide compatibility with old programs relying on + set*uid to gain/lose privilege, transitions to/from uid 0 cause + capabilities to be gained/lost. */ +#define SECURE_NO_SETUID_FIXUP 2 +#define SECURE_NO_SETUID_FIXUP_LOCKED 3 /* make bit-2 immutable */ + +/* When set, a process can retain its capabilities even after + transitioning to a non-root user (the set-uid fixup suppressed by + bit 2). Bit-4 is cleared when a process calls exec(); setting both + bit 4 and 5 will create a barrier through exec that no exec()'d + child can use this feature again. */ +#define SECURE_KEEP_CAPS 4 +#define SECURE_KEEP_CAPS_LOCKED 5 /* make bit-4 immutable */ + +/* Each securesetting is implemented using two bits. One bit specifies + whether the setting is on or off. The other bit specify whether the + setting is locked or not. A setting which is locked cannot be + changed from user-level. */ +#define issecure_mask(X) (1 << (X)) +#define issecure(X) (issecure_mask(X) & current_cred_xxx(securebits)) + +#define SECURE_ALL_BITS (issecure_mask(SECURE_NOROOT) | \ + issecure_mask(SECURE_NO_SETUID_FIXUP) | \ + issecure_mask(SECURE_KEEP_CAPS)) +#define SECURE_ALL_LOCKS (SECURE_ALL_BITS << 1) + +#endif /* !_LINUX_SECUREBITS_H */ diff --git a/src/core/selinux-access.c b/src/core/selinux-access.c new file mode 100644 index 0000000000..426aed07d2 --- /dev/null +++ b/src/core/selinux-access.c @@ -0,0 +1,439 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2012 Dan Walsh + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "selinux-access.h" + +#ifdef HAVE_SELINUX + +#include <stdio.h> +#include <string.h> +#include <errno.h> +#include <limits.h> +#include <selinux/selinux.h> +#include <selinux/avc.h> +#ifdef HAVE_AUDIT +#include <libaudit.h> +#endif +#include <dbus.h> + +#include "util.h" +#include "log.h" +#include "bus-errors.h" +#include "dbus-common.h" +#include "audit.h" +#include "selinux-util.h" +#include "audit-fd.h" + +static bool initialized = false; + +struct auditstruct { + const char *path; + char *cmdline; + uid_t loginuid; + uid_t uid; + gid_t gid; +}; + +static int bus_get_selinux_security_context( + DBusConnection *connection, + const char *name, + char **scon, + DBusError *error) { + + _cleanup_dbus_message_unref_ DBusMessage *m = NULL, *reply = NULL; + DBusMessageIter iter, sub; + const char *bytes; + char *b; + int nbytes; + + m = dbus_message_new_method_call( + DBUS_SERVICE_DBUS, + DBUS_PATH_DBUS, + DBUS_INTERFACE_DBUS, + "GetConnectionSELinuxSecurityContext"); + if (!m) { + dbus_set_error_const(error, DBUS_ERROR_NO_MEMORY, NULL); + return -ENOMEM; + } + + if (!dbus_message_append_args( + m, + DBUS_TYPE_STRING, &name, + DBUS_TYPE_INVALID)) { + dbus_set_error_const(error, DBUS_ERROR_NO_MEMORY, NULL); + return -ENOMEM; + } + + reply = dbus_connection_send_with_reply_and_block(connection, m, -1, error); + if (!reply) + return -EIO; + + if (dbus_set_error_from_message(error, reply)) + return -EIO; + + if (!dbus_message_iter_init(reply, &iter)) + return -EIO; + + if (dbus_message_iter_get_arg_type(&iter) != DBUS_TYPE_ARRAY) + return -EIO; + + dbus_message_iter_recurse(&iter, &sub); + dbus_message_iter_get_fixed_array(&sub, &bytes, &nbytes); + + b = strndup(bytes, nbytes); + if (!b) + return -ENOMEM; + + *scon = b; + + return 0; +} + +static int bus_get_audit_data( + DBusConnection *connection, + const char *name, + struct auditstruct *audit, + DBusError *error) { + + pid_t pid; + int r; + + pid = bus_get_unix_process_id(connection, name, error); + if (pid <= 0) + return -EIO; + + r = audit_loginuid_from_pid(pid, &audit->loginuid); + if (r < 0) + return r; + + r = get_process_uid(pid, &audit->uid); + if (r < 0) + return r; + + r = get_process_gid(pid, &audit->gid); + if (r < 0) + return r; + + r = get_process_cmdline(pid, 0, true, &audit->cmdline); + if (r < 0) + return r; + + return 0; +} + +/* + Any time an access gets denied this callback will be called + with the aduit data. We then need to just copy the audit data into the msgbuf. +*/ +static int audit_callback( + void *auditdata, + security_class_t cls, + char *msgbuf, + size_t msgbufsize) { + + struct auditstruct *audit = (struct auditstruct *) auditdata; + + snprintf(msgbuf, msgbufsize, + "auid=%d uid=%d gid=%d%s%s%s%s%s%s", + audit->loginuid, + audit->uid, + audit->gid, + (audit->path ? " path=\"" : ""), + strempty(audit->path), + (audit->path ? "\"" : ""), + (audit->cmdline ? " cmdline=\"" : ""), + strempty(audit->cmdline), + (audit->cmdline ? "\"" : "")); + + msgbuf[msgbufsize-1] = 0; + + return 0; +} + +/* + Any time an access gets denied this callback will be called + code copied from dbus. If audit is turned on the messages will go as + user_avc's into the /var/log/audit/audit.log, otherwise they will be + sent to syslog. +*/ +_printf_attr_(2, 3) static int log_callback(int type, const char *fmt, ...) { + va_list ap; + + va_start(ap, fmt); + +#ifdef HAVE_AUDIT + if (get_audit_fd() >= 0) { + _cleanup_free_ char *buf = NULL; + int r; + + r = vasprintf(&buf, fmt, ap); + va_end(ap); + + if (r >= 0) { + audit_log_user_avc_message(get_audit_fd(), AUDIT_USER_AVC, buf, NULL, NULL, NULL, 0); + return 0; + } + + va_start(ap, fmt); + } +#endif + log_metav(LOG_USER | LOG_INFO, __FILE__, __LINE__, __FUNCTION__, fmt, ap); + va_end(ap); + + return 0; +} + +/* + Function must be called once to initialize the SELinux AVC environment. + Sets up callbacks. + If you want to cleanup memory you should need to call selinux_access_finish. +*/ +static int access_init(void) { + int r; + + if (avc_open(NULL, 0)) { + log_error("avc_open() failed: %m"); + return -errno; + } + + selinux_set_callback(SELINUX_CB_AUDIT, (union selinux_callback) audit_callback); + selinux_set_callback(SELINUX_CB_LOG, (union selinux_callback) log_callback); + + if (security_getenforce() >= 0) + return 0; + + r = -errno; + avc_destroy(); + + return r; +} + +static int selinux_access_init(DBusError *error) { + int r; + + if (initialized) + return 0; + + if (use_selinux()) { + r = access_init(); + if (r < 0) { + dbus_set_error(error, DBUS_ERROR_ACCESS_DENIED, "Failed to initialize SELinux."); + return r; + } + } + + initialized = true; + return 0; +} + +void selinux_access_free(void) { + if (!initialized) + return; + + avc_destroy(); + initialized = false; +} + +static int get_audit_data( + DBusConnection *connection, + DBusMessage *message, + struct auditstruct *audit, + DBusError *error) { + + const char *sender; + int r, fd; + struct ucred ucred; + socklen_t len = sizeof(ucred); + + sender = dbus_message_get_sender(message); + if (sender) + return bus_get_audit_data(connection, sender, audit, error); + + if (!dbus_connection_get_unix_fd(connection, &fd)) + return -EINVAL; + + r = getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &ucred, &len); + if (r < 0) { + log_error("Failed to determine peer credentials: %m"); + return -errno; + } + + audit->uid = ucred.uid; + audit->gid = ucred.gid; + + r = audit_loginuid_from_pid(ucred.pid, &audit->loginuid); + if (r < 0) + return r; + + r = get_process_cmdline(ucred.pid, 0, true, &audit->cmdline); + if (r < 0) + return r; + + return 0; +} + +/* + This function returns the security context of the remote end of the dbus + connections. Whether it is on the bus or a local connection. +*/ +static int get_calling_context( + DBusConnection *connection, + DBusMessage *message, + security_context_t *scon, + DBusError *error) { + + const char *sender; + int r; + int fd; + + /* + If sender exists then + if sender is NULL this indicates a local connection. Grab the fd + from dbus and do an getpeercon to peers process context + */ + sender = dbus_message_get_sender(message); + if (sender) { + r = bus_get_selinux_security_context(connection, sender, scon, error); + if (r >= 0) + return r; + + log_error("bus_get_selinux_security_context failed: %m"); + return r; + } + + if (!dbus_connection_get_unix_fd(connection, &fd)) { + log_error("bus_connection_get_unix_fd failed %m"); + return -EINVAL; + } + + r = getpeercon(fd, scon); + if (r < 0) { + log_error("getpeercon failed %m"); + return -errno; + } + + return 0; +} + +/* + This function communicates with the kernel to check whether or not it should + allow the access. + If the machine is in permissive mode it will return ok. Audit messages will + still be generated if the access would be denied in enforcing mode. +*/ +int selinux_access_check( + DBusConnection *connection, + DBusMessage *message, + const char *path, + const char *permission, + DBusError *error) { + + security_context_t scon = NULL, fcon = NULL; + int r = 0; + const char *tclass = NULL; + struct auditstruct audit; + + assert(connection); + assert(message); + assert(permission); + assert(error); + + if (!use_selinux()) + return 0; + + r = selinux_access_init(error); + if (r < 0) + return r; + + audit.uid = audit.loginuid = (uid_t) -1; + audit.gid = (gid_t) -1; + audit.cmdline = NULL; + audit.path = path; + + r = get_calling_context(connection, message, &scon, error); + if (r < 0) { + log_error("Failed to get caller's security context on: %m"); + goto finish; + } + + if (path) { + tclass = "service"; + /* get the file context of the unit file */ + r = getfilecon(path, &fcon); + if (r < 0) { + dbus_set_error(error, DBUS_ERROR_ACCESS_DENIED, "Failed to get file context on %s.", path); + r = -errno; + log_error("Failed to get security context on %s: %m",path); + goto finish; + } + + } else { + tclass = "system"; + r = getcon(&fcon); + if (r < 0) { + dbus_set_error(error, DBUS_ERROR_ACCESS_DENIED, "Failed to get current context."); + r = -errno; + log_error("Failed to get current process context on: %m"); + goto finish; + } + } + + (void) get_audit_data(connection, message, &audit, error); + + errno = 0; + r = selinux_check_access(scon, fcon, tclass, permission, &audit); + if (r < 0) { + dbus_set_error(error, DBUS_ERROR_ACCESS_DENIED, "SELinux policy denies access."); + r = -errno; + log_error("SELinux policy denies access."); + } + + log_debug("SELinux access check scon=%s tcon=%s tclass=%s perm=%s path=%s cmdline=%s: %i", scon, fcon, tclass, permission, path, audit.cmdline, r); + +finish: + free(audit.cmdline); + freecon(scon); + freecon(fcon); + + if (r && security_getenforce() != 1) { + dbus_error_init(error); + r = 0; + } + + return r; +} + +#else + +int selinux_access_check( + DBusConnection *connection, + DBusMessage *message, + const char *path, + const char *permission, + DBusError *error) { + + return 0; +} + +void selinux_access_free(void) { +} + +#endif diff --git a/src/core/selinux-access.h b/src/core/selinux-access.h new file mode 100644 index 0000000000..9183cbc9a6 --- /dev/null +++ b/src/core/selinux-access.h @@ -0,0 +1,62 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Dan Walsh + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dbus.h> + +void selinux_access_free(void); + +int selinux_access_check(DBusConnection *connection, DBusMessage *message, const char *path, const char *permission, DBusError *error); + +#ifdef HAVE_SELINUX + +#define SELINUX_ACCESS_CHECK(connection, message, permission) \ + do { \ + DBusError _error; \ + int _r; \ + DBusConnection *_c = (connection); \ + DBusMessage *_m = (message); \ + dbus_error_init(&_error); \ + _r = selinux_access_check(_c, _m, NULL, (permission), &_error); \ + if (_r < 0) \ + return bus_send_error_reply(_c, _m, &_error, _r); \ + } while (false) + +#define SELINUX_UNIT_ACCESS_CHECK(unit, connection, message, permission) \ + do { \ + DBusError _error; \ + int _r; \ + DBusConnection *_c = (connection); \ + DBusMessage *_m = (message); \ + Unit *_u = (unit); \ + dbus_error_init(&_error); \ + _r = selinux_access_check(_c, _m, _u->source_path ?: _u->fragment_path, (permission), &_error); \ + if (_r < 0) \ + return bus_send_error_reply(_c, _m, &_error, _r); \ + } while (false) + +#else + +#define SELINUX_ACCESS_CHECK(connection, message, permission) do { } while (false) +#define SELINUX_UNIT_ACCESS_CHECK(unit, connection, message, permission) do { } while (false) + +#endif diff --git a/src/core/selinux-setup.c b/src/core/selinux-setup.c new file mode 100644 index 0000000000..7a32ed59a0 --- /dev/null +++ b/src/core/selinux-setup.c @@ -0,0 +1,120 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <unistd.h> +#include <stdio.h> +#include <errno.h> +#include <string.h> +#include <stdlib.h> + +#ifdef HAVE_SELINUX +#include <selinux/selinux.h> +#endif + +#include "selinux-setup.h" +#include "selinux-util.h" +#include "label.h" +#include "mount-setup.h" +#include "macro.h" +#include "util.h" +#include "log.h" + +#ifdef HAVE_SELINUX +static int null_log(int type, const char *fmt, ...) { + return 0; +} +#endif + +int selinux_setup(bool *loaded_policy) { + +#ifdef HAVE_SELINUX + int enforce = 0; + usec_t before_load, after_load; + security_context_t con; + int r; + union selinux_callback cb; + + assert(loaded_policy); + + /* Turn off all of SELinux' own logging, we want to do that */ + cb.func_log = null_log; + selinux_set_callback(SELINUX_CB_LOG, cb); + + /* Already initialized by somebody else? */ + r = getcon_raw(&con); + if (r == 0) { + bool initialized; + + initialized = !streq(con, "kernel"); + freecon(con); + + if (initialized) + return 0; + } + + /* Make sure we have no fds open while loading the policy and + * transitioning */ + log_close(); + + /* Now load the policy */ + before_load = now(CLOCK_MONOTONIC); + r = selinux_init_load_policy(&enforce); + if (r == 0) { + char timespan[FORMAT_TIMESPAN_MAX]; + char *label; + + retest_selinux(); + + /* Transition to the new context */ + r = label_get_create_label_from_exe(SYSTEMD_BINARY_PATH, &label); + if (r < 0 || label == NULL) { + log_open(); + log_error("Failed to compute init label, ignoring."); + } else { + r = setcon(label); + + log_open(); + if (r < 0) + log_error("Failed to transition into init label '%s', ignoring.", label); + + label_free(label); + } + + after_load = now(CLOCK_MONOTONIC); + + log_info("Successfully loaded SELinux policy in %s.", + format_timespan(timespan, sizeof(timespan), after_load - before_load, 0)); + + *loaded_policy = true; + + } else { + log_open(); + + if (enforce > 0) { + log_error("Failed to load SELinux policy. Freezing."); + return -EIO; + } else + log_debug("Unable to load SELinux policy. Ignoring."); + } +#endif + + return 0; +} diff --git a/src/core/selinux-setup.h b/src/core/selinux-setup.h new file mode 100644 index 0000000000..39e2bc25bb --- /dev/null +++ b/src/core/selinux-setup.h @@ -0,0 +1,26 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +int selinux_setup(bool *loaded_policy); diff --git a/src/core/service.c b/src/core/service.c new file mode 100644 index 0000000000..3617c24711 --- /dev/null +++ b/src/core/service.c @@ -0,0 +1,3888 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <signal.h> +#include <dirent.h> +#include <unistd.h> +#include <sys/reboot.h> + +#include "manager.h" +#include "unit.h" +#include "service.h" +#include "load-fragment.h" +#include "load-dropin.h" +#include "log.h" +#include "strv.h" +#include "unit-name.h" +#include "unit-printf.h" +#include "dbus-service.h" +#include "special.h" +#include "dbus-common.h" +#include "exit-status.h" +#include "def.h" +#include "path-util.h" +#include "util.h" +#include "utf8.h" +#include "env-util.h" +#include "fileio.h" + +#ifdef HAVE_SYSV_COMPAT + +#define DEFAULT_SYSV_TIMEOUT_USEC (5*USEC_PER_MINUTE) + +typedef enum RunlevelType { + RUNLEVEL_UP, + RUNLEVEL_DOWN +} RunlevelType; + +static const struct { + const char *path; + const char *target; + const RunlevelType type; +} rcnd_table[] = { + /* Standard SysV runlevels for start-up */ + { "rc1.d", SPECIAL_RESCUE_TARGET, RUNLEVEL_UP }, + { "rc2.d", SPECIAL_RUNLEVEL2_TARGET, RUNLEVEL_UP }, + { "rc3.d", SPECIAL_RUNLEVEL3_TARGET, RUNLEVEL_UP }, + { "rc4.d", SPECIAL_RUNLEVEL4_TARGET, RUNLEVEL_UP }, + { "rc5.d", SPECIAL_RUNLEVEL5_TARGET, RUNLEVEL_UP }, + + /* Standard SysV runlevels for shutdown */ + { "rc0.d", SPECIAL_POWEROFF_TARGET, RUNLEVEL_DOWN }, + { "rc6.d", SPECIAL_REBOOT_TARGET, RUNLEVEL_DOWN } + + /* Note that the order here matters, as we read the + directories in this order, and we want to make sure that + sysv_start_priority is known when we first load the + unit. And that value we only know from S links. Hence + UP must be read before DOWN */ +}; + +#define RUNLEVELS_UP "12345" +#endif + +static const UnitActiveState state_translation_table[_SERVICE_STATE_MAX] = { + [SERVICE_DEAD] = UNIT_INACTIVE, + [SERVICE_START_PRE] = UNIT_ACTIVATING, + [SERVICE_START] = UNIT_ACTIVATING, + [SERVICE_START_POST] = UNIT_ACTIVATING, + [SERVICE_RUNNING] = UNIT_ACTIVE, + [SERVICE_EXITED] = UNIT_ACTIVE, + [SERVICE_RELOAD] = UNIT_RELOADING, + [SERVICE_STOP] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGTERM] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGKILL] = UNIT_DEACTIVATING, + [SERVICE_STOP_POST] = UNIT_DEACTIVATING, + [SERVICE_FINAL_SIGTERM] = UNIT_DEACTIVATING, + [SERVICE_FINAL_SIGKILL] = UNIT_DEACTIVATING, + [SERVICE_FAILED] = UNIT_FAILED, + [SERVICE_AUTO_RESTART] = UNIT_ACTIVATING +}; + +/* For Type=idle we never want to delay any other jobs, hence we + * consider idle jobs active as soon as we start working on them */ +static const UnitActiveState state_translation_table_idle[_SERVICE_STATE_MAX] = { + [SERVICE_DEAD] = UNIT_INACTIVE, + [SERVICE_START_PRE] = UNIT_ACTIVE, + [SERVICE_START] = UNIT_ACTIVE, + [SERVICE_START_POST] = UNIT_ACTIVE, + [SERVICE_RUNNING] = UNIT_ACTIVE, + [SERVICE_EXITED] = UNIT_ACTIVE, + [SERVICE_RELOAD] = UNIT_RELOADING, + [SERVICE_STOP] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGTERM] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGKILL] = UNIT_DEACTIVATING, + [SERVICE_STOP_POST] = UNIT_DEACTIVATING, + [SERVICE_FINAL_SIGTERM] = UNIT_DEACTIVATING, + [SERVICE_FINAL_SIGKILL] = UNIT_DEACTIVATING, + [SERVICE_FAILED] = UNIT_FAILED, + [SERVICE_AUTO_RESTART] = UNIT_ACTIVATING +}; + +static void service_init(Unit *u) { + Service *s = SERVICE(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + s->timeout_start_usec = DEFAULT_TIMEOUT_USEC; + s->timeout_stop_usec = DEFAULT_TIMEOUT_USEC; + s->restart_usec = DEFAULT_RESTART_USEC; + s->type = _SERVICE_TYPE_INVALID; + + watch_init(&s->watchdog_watch); + watch_init(&s->timer_watch); + +#ifdef HAVE_SYSV_COMPAT + s->sysv_start_priority = -1; + s->sysv_start_priority_from_rcnd = -1; +#endif + s->socket_fd = -1; + s->guess_main_pid = true; + + exec_context_init(&s->exec_context); + kill_context_init(&s->kill_context); + + RATELIMIT_INIT(s->start_limit, 10*USEC_PER_SEC, 5); + + s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; +} + +static void service_unwatch_control_pid(Service *s) { + assert(s); + + if (s->control_pid <= 0) + return; + + unit_unwatch_pid(UNIT(s), s->control_pid); + s->control_pid = 0; +} + +static void service_unwatch_main_pid(Service *s) { + assert(s); + + if (s->main_pid <= 0) + return; + + unit_unwatch_pid(UNIT(s), s->main_pid); + s->main_pid = 0; +} + +static void service_unwatch_pid_file(Service *s) { + if (!s->pid_file_pathspec) + return; + + log_debug_unit(UNIT(s)->id, "Stopping watch for %s's PID file %s", + UNIT(s)->id, s->pid_file_pathspec->path); + path_spec_unwatch(s->pid_file_pathspec, UNIT(s)); + path_spec_done(s->pid_file_pathspec); + free(s->pid_file_pathspec); + s->pid_file_pathspec = NULL; +} + +static int service_set_main_pid(Service *s, pid_t pid) { + pid_t ppid; + + assert(s); + + if (pid <= 1) + return -EINVAL; + + if (pid == getpid()) + return -EINVAL; + + s->main_pid = pid; + s->main_pid_known = true; + + if (get_parent_of_pid(pid, &ppid) >= 0 && ppid != getpid()) { + log_warning_unit(UNIT(s)->id, + "%s: Supervising process %lu which is not our child. We'll most likely not notice when it exits.", + UNIT(s)->id, (unsigned long) pid); + + s->main_pid_alien = true; + } else + s->main_pid_alien = false; + + exec_status_start(&s->main_exec_status, pid); + + return 0; +} + +static void service_close_socket_fd(Service *s) { + assert(s); + + if (s->socket_fd < 0) + return; + + close_nointr_nofail(s->socket_fd); + s->socket_fd = -1; +} + +static void service_connection_unref(Service *s) { + assert(s); + + if (!UNIT_DEREF(s->accept_socket)) + return; + + socket_connection_unref(SOCKET(UNIT_DEREF(s->accept_socket))); + unit_ref_unset(&s->accept_socket); +} + +static void service_stop_watchdog(Service *s) { + assert(s); + + unit_unwatch_timer(UNIT(s), &s->watchdog_watch); + s->watchdog_timestamp.realtime = 0; + s->watchdog_timestamp.monotonic = 0; +} + +static void service_enter_dead(Service *s, ServiceResult f, bool allow_restart); + +static void service_handle_watchdog(Service *s) { + usec_t offset; + int r; + + assert(s); + + if (s->watchdog_usec == 0) + return; + + offset = now(CLOCK_MONOTONIC) - s->watchdog_timestamp.monotonic; + if (offset >= s->watchdog_usec) { + log_error_unit(UNIT(s)->id, "%s watchdog timeout!", UNIT(s)->id); + service_enter_dead(s, SERVICE_FAILURE_WATCHDOG, true); + return; + } + + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, s->watchdog_usec - offset, &s->watchdog_watch); + if (r < 0) + log_warning_unit(UNIT(s)->id, + "%s failed to install watchdog timer: %s", + UNIT(s)->id, strerror(-r)); +} + +static void service_reset_watchdog(Service *s) { + assert(s); + + dual_timestamp_get(&s->watchdog_timestamp); + service_handle_watchdog(s); +} + +static void service_done(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + free(s->pid_file); + s->pid_file = NULL; + +#ifdef HAVE_SYSV_COMPAT + free(s->sysv_runlevels); + s->sysv_runlevels = NULL; +#endif + + free(s->status_text); + s->status_text = NULL; + + exec_context_done(&s->exec_context, manager_is_reloading_or_reexecuting(u->manager)); + exec_command_free_array(s->exec_command, _SERVICE_EXEC_COMMAND_MAX); + s->control_command = NULL; + s->main_command = NULL; + + set_free(s->restart_ignore_status.code); + s->restart_ignore_status.code = NULL; + set_free(s->restart_ignore_status.signal); + s->restart_ignore_status.signal = NULL; + + set_free(s->success_status.code); + s->success_status.code = NULL; + set_free(s->success_status.signal); + s->success_status.signal = NULL; + + /* This will leak a process, but at least no memory or any of + * our resources */ + service_unwatch_main_pid(s); + service_unwatch_control_pid(s); + service_unwatch_pid_file(s); + + if (s->bus_name) { + unit_unwatch_bus_name(u, s->bus_name); + free(s->bus_name); + s->bus_name = NULL; + } + + service_close_socket_fd(s); + service_connection_unref(s); + + unit_ref_unset(&s->accept_socket); + + service_stop_watchdog(s); + + unit_unwatch_timer(u, &s->timer_watch); +} + +#ifdef HAVE_SYSV_COMPAT +static char *sysv_translate_name(const char *name) { + char *r; + + r = new(char, strlen(name) + sizeof(".service")); + if (!r) + return NULL; + + if (endswith(name, ".sh")) + /* Drop .sh suffix */ + strcpy(stpcpy(r, name) - 3, ".service"); + else + /* Normal init script name */ + strcpy(stpcpy(r, name), ".service"); + + return r; +} + +static int sysv_translate_facility(const char *name, const char *filename, char **_r) { + + /* We silently ignore the $ prefix here. According to the LSB + * spec it simply indicates whether something is a + * standardized name or a distribution-specific one. Since we + * just follow what already exists and do not introduce new + * uses or names we don't care who introduced a new name. */ + + static const char * const table[] = { + /* LSB defined facilities */ + "local_fs", NULL, + "network", SPECIAL_NETWORK_TARGET, + "named", SPECIAL_NSS_LOOKUP_TARGET, + "portmap", SPECIAL_RPCBIND_TARGET, + "remote_fs", SPECIAL_REMOTE_FS_TARGET, + "syslog", NULL, + "time", SPECIAL_TIME_SYNC_TARGET, + }; + + unsigned i; + char *r; + const char *n; + + assert(name); + assert(_r); + + n = *name == '$' ? name + 1 : name; + + for (i = 0; i < ELEMENTSOF(table); i += 2) { + + if (!streq(table[i], n)) + continue; + + if (!table[i+1]) + return 0; + + r = strdup(table[i+1]); + if (!r) + return log_oom(); + + goto finish; + } + + /* If we don't know this name, fallback heuristics to figure + * out whether something is a target or a service alias. */ + + if (*name == '$') { + if (!unit_prefix_is_valid(n)) + return -EINVAL; + + /* Facilities starting with $ are most likely targets */ + r = unit_name_build(n, NULL, ".target"); + } else if (filename && streq(name, filename)) + /* Names equaling the file name of the services are redundant */ + return 0; + else + /* Everything else we assume to be normal service names */ + r = sysv_translate_name(n); + + if (!r) + return -ENOMEM; + +finish: + *_r = r; + + return 1; +} + +static int sysv_fix_order(Service *s) { + Unit *other; + int r; + + assert(s); + + if (s->sysv_start_priority < 0) + return 0; + + /* For each pair of services where at least one lacks a LSB + * header, we use the start priority value to order things. */ + + LIST_FOREACH(units_by_type, other, UNIT(s)->manager->units_by_type[UNIT_SERVICE]) { + Service *t; + UnitDependency d; + bool special_s, special_t; + + t = SERVICE(other); + + if (s == t) + continue; + + if (UNIT(t)->load_state != UNIT_LOADED) + continue; + + if (t->sysv_start_priority < 0) + continue; + + /* If both units have modern headers we don't care + * about the priorities */ + if ((UNIT(s)->fragment_path || s->sysv_has_lsb) && + (UNIT(t)->fragment_path || t->sysv_has_lsb)) + continue; + + special_s = s->sysv_runlevels && !chars_intersect(RUNLEVELS_UP, s->sysv_runlevels); + special_t = t->sysv_runlevels && !chars_intersect(RUNLEVELS_UP, t->sysv_runlevels); + + if (special_t && !special_s) + d = UNIT_AFTER; + else if (special_s && !special_t) + d = UNIT_BEFORE; + else if (t->sysv_start_priority < s->sysv_start_priority) + d = UNIT_AFTER; + else if (t->sysv_start_priority > s->sysv_start_priority) + d = UNIT_BEFORE; + else + continue; + + /* FIXME: Maybe we should compare the name here lexicographically? */ + + if ((r = unit_add_dependency(UNIT(s), d, UNIT(t), true)) < 0) + return r; + } + + return 0; +} + +static ExecCommand *exec_command_new(const char *path, const char *arg1) { + ExecCommand *c; + + if (!(c = new0(ExecCommand, 1))) + return NULL; + + if (!(c->path = strdup(path))) { + free(c); + return NULL; + } + + if (!(c->argv = strv_new(path, arg1, NULL))) { + free(c->path); + free(c); + return NULL; + } + + return c; +} + +static int sysv_exec_commands(Service *s, const bool supports_reload) { + ExecCommand *c; + + assert(s); + assert(s->is_sysv); + assert(UNIT(s)->source_path); + + c = exec_command_new(UNIT(s)->source_path, "start"); + if (!c) + return -ENOMEM; + exec_command_append_list(s->exec_command+SERVICE_EXEC_START, c); + + c = exec_command_new(UNIT(s)->source_path, "stop"); + if (!c) + return -ENOMEM; + exec_command_append_list(s->exec_command+SERVICE_EXEC_STOP, c); + + if (supports_reload) { + c = exec_command_new(UNIT(s)->source_path, "reload"); + if (!c) + return -ENOMEM; + exec_command_append_list(s->exec_command+SERVICE_EXEC_RELOAD, c); + } + + return 0; +} + +static bool usage_contains_reload(const char *line) { + return (strcasestr(line, "{reload|") || + strcasestr(line, "{reload}") || + strcasestr(line, "{reload\"") || + strcasestr(line, "|reload|") || + strcasestr(line, "|reload}") || + strcasestr(line, "|reload\"")); +} + +static int service_load_sysv_path(Service *s, const char *path) { + FILE *f; + Unit *u; + unsigned line = 0; + int r; + enum { + NORMAL, + DESCRIPTION, + LSB, + LSB_DESCRIPTION, + USAGE_CONTINUATION + } state = NORMAL; + char *short_description = NULL, *long_description = NULL, *chkconfig_description = NULL, *description; + struct stat st; + bool supports_reload = false; + + assert(s); + assert(path); + + u = UNIT(s); + + f = fopen(path, "re"); + if (!f) { + r = errno == ENOENT ? 0 : -errno; + goto finish; + } + + if (fstat(fileno(f), &st) < 0) { + r = -errno; + goto finish; + } + + free(u->source_path); + u->source_path = strdup(path); + if (!u->source_path) { + r = -ENOMEM; + goto finish; + } + u->source_mtime = timespec_load(&st.st_mtim); + + if (null_or_empty(&st)) { + u->load_state = UNIT_MASKED; + r = 0; + goto finish; + } + + s->is_sysv = true; + + while (!feof(f)) { + char l[LINE_MAX], *t; + + if (!fgets(l, sizeof(l), f)) { + if (feof(f)) + break; + + r = -errno; + log_error_unit(u->id, + "Failed to read configuration file '%s': %s", + path, strerror(-r)); + goto finish; + } + + line++; + + t = strstrip(l); + if (*t != '#') { + /* Try to figure out whether this init script supports + * the reload operation. This heuristic looks for + * "Usage" lines which include the reload option. */ + if ( state == USAGE_CONTINUATION || + (state == NORMAL && strcasestr(t, "usage"))) { + if (usage_contains_reload(t)) { + supports_reload = true; + state = NORMAL; + } else if (t[strlen(t)-1] == '\\') + state = USAGE_CONTINUATION; + else + state = NORMAL; + } + + continue; + } + + if (state == NORMAL && streq(t, "### BEGIN INIT INFO")) { + state = LSB; + s->sysv_has_lsb = true; + continue; + } + + if ((state == LSB_DESCRIPTION || state == LSB) && streq(t, "### END INIT INFO")) { + state = NORMAL; + continue; + } + + t++; + t += strspn(t, WHITESPACE); + + if (state == NORMAL) { + + /* Try to parse Red Hat style chkconfig headers */ + + if (startswith_no_case(t, "chkconfig:")) { + int start_priority; + char runlevels[16], *k; + + state = NORMAL; + + if (sscanf(t+10, "%15s %i %*i", + runlevels, + &start_priority) != 2) { + + log_warning_unit(u->id, + "[%s:%u] Failed to parse chkconfig line. Ignoring.", + path, line); + continue; + } + + /* A start priority gathered from the + * symlink farms is preferred over the + * data from the LSB header. */ + if (start_priority < 0 || start_priority > 99) + log_warning_unit(u->id, + "[%s:%u] Start priority out of range. Ignoring.", + path, line); + else + s->sysv_start_priority = start_priority; + + char_array_0(runlevels); + k = delete_chars(runlevels, WHITESPACE "-"); + + if (k[0]) { + char *d; + + if (!(d = strdup(k))) { + r = -ENOMEM; + goto finish; + } + + free(s->sysv_runlevels); + s->sysv_runlevels = d; + } + + } else if (startswith_no_case(t, "description:")) { + + size_t k = strlen(t); + char *d; + const char *j; + + if (t[k-1] == '\\') { + state = DESCRIPTION; + t[k-1] = 0; + } + + if ((j = strstrip(t+12)) && *j) { + if (!(d = strdup(j))) { + r = -ENOMEM; + goto finish; + } + } else + d = NULL; + + free(chkconfig_description); + chkconfig_description = d; + + } else if (startswith_no_case(t, "pidfile:")) { + + char *fn; + + state = NORMAL; + + fn = strstrip(t+8); + if (!path_is_absolute(fn)) { + log_warning_unit(u->id, + "[%s:%u] PID file not absolute. Ignoring.", + path, line); + continue; + } + + if (!(fn = strdup(fn))) { + r = -ENOMEM; + goto finish; + } + + free(s->pid_file); + s->pid_file = fn; + } + + } else if (state == DESCRIPTION) { + + /* Try to parse Red Hat style description + * continuation */ + + size_t k = strlen(t); + char *j; + + if (t[k-1] == '\\') + t[k-1] = 0; + else + state = NORMAL; + + if ((j = strstrip(t)) && *j) { + char *d = NULL; + + if (chkconfig_description) + d = strjoin(chkconfig_description, " ", j, NULL); + else + d = strdup(j); + + if (!d) { + r = -ENOMEM; + goto finish; + } + + free(chkconfig_description); + chkconfig_description = d; + } + + } else if (state == LSB || state == LSB_DESCRIPTION) { + + if (startswith_no_case(t, "Provides:")) { + char *i, *w; + size_t z; + + state = LSB; + + FOREACH_WORD_QUOTED(w, z, t+9, i) { + char *n, *m; + + if (!(n = strndup(w, z))) { + r = -ENOMEM; + goto finish; + } + + r = sysv_translate_facility(n, path_get_file_name(path), &m); + free(n); + + if (r < 0) + goto finish; + + if (r == 0) + continue; + + if (unit_name_to_type(m) == UNIT_SERVICE) + r = unit_merge_by_name(u, m); + else + /* NB: SysV targets + * which are provided + * by a service are + * pulled in by the + * services, as an + * indication that the + * generic service is + * now available. This + * is strictly + * one-way. The + * targets do NOT pull + * in the SysV + * services! */ + r = unit_add_two_dependencies_by_name(u, UNIT_BEFORE, UNIT_WANTS, m, NULL, true); + + if (r < 0) + log_error_unit(u->id, + "[%s:%u] Failed to add LSB Provides name %s, ignoring: %s", + path, line, m, strerror(-r)); + + free(m); + } + + } else if (startswith_no_case(t, "Required-Start:") || + startswith_no_case(t, "Should-Start:") || + startswith_no_case(t, "X-Start-Before:") || + startswith_no_case(t, "X-Start-After:")) { + char *i, *w; + size_t z; + + state = LSB; + + FOREACH_WORD_QUOTED(w, z, strchr(t, ':')+1, i) { + char *n, *m; + + if (!(n = strndup(w, z))) { + r = -ENOMEM; + goto finish; + } + + r = sysv_translate_facility(n, path_get_file_name(path), &m); + if (r < 0) { + log_error_unit(u->id, + "[%s:%u] Failed to translate LSB dependency %s, ignoring: %s", + path, line, n, strerror(-r)); + free(n); + continue; + } + + free(n); + + if (r == 0) + continue; + + r = unit_add_dependency_by_name(u, startswith_no_case(t, "X-Start-Before:") ? UNIT_BEFORE : UNIT_AFTER, m, NULL, true); + + if (r < 0) + log_error_unit(u->id, "[%s:%u] Failed to add dependency on %s, ignoring: %s", + path, line, m, strerror(-r)); + + free(m); + } + } else if (startswith_no_case(t, "Default-Start:")) { + char *k, *d; + + state = LSB; + + k = delete_chars(t+14, WHITESPACE "-"); + + if (k[0] != 0) { + if (!(d = strdup(k))) { + r = -ENOMEM; + goto finish; + } + + free(s->sysv_runlevels); + s->sysv_runlevels = d; + } + + } else if (startswith_no_case(t, "Description:")) { + char *d, *j; + + state = LSB_DESCRIPTION; + + if ((j = strstrip(t+12)) && *j) { + if (!(d = strdup(j))) { + r = -ENOMEM; + goto finish; + } + } else + d = NULL; + + free(long_description); + long_description = d; + + } else if (startswith_no_case(t, "Short-Description:")) { + char *d, *j; + + state = LSB; + + if ((j = strstrip(t+18)) && *j) { + if (!(d = strdup(j))) { + r = -ENOMEM; + goto finish; + } + } else + d = NULL; + + free(short_description); + short_description = d; + + } else if (state == LSB_DESCRIPTION) { + + if (startswith(l, "#\t") || startswith(l, "# ")) { + char *j; + + if ((j = strstrip(t)) && *j) { + char *d = NULL; + + if (long_description) + d = strjoin(long_description, " ", t, NULL); + else + d = strdup(j); + + if (!d) { + r = -ENOMEM; + goto finish; + } + + free(long_description); + long_description = d; + } + + } else + state = LSB; + } + } + } + + if ((r = sysv_exec_commands(s, supports_reload)) < 0) + goto finish; + + if (s->sysv_runlevels && !chars_intersect(RUNLEVELS_UP, s->sysv_runlevels)) { + /* If there a runlevels configured for this service + * but none of the standard ones, then we assume this + * is some special kind of service (which might be + * needed for early boot) and don't create any links + * to it. */ + + UNIT(s)->default_dependencies = false; + + /* Don't timeout special services during boot (like fsck) */ + s->timeout_start_usec = 0; + s->timeout_stop_usec = 0; + } else { + s->timeout_start_usec = DEFAULT_SYSV_TIMEOUT_USEC; + s->timeout_stop_usec = DEFAULT_SYSV_TIMEOUT_USEC; + } + + /* Special setting for all SysV services */ + s->type = SERVICE_FORKING; + s->remain_after_exit = !s->pid_file; + s->guess_main_pid = false; + s->restart = SERVICE_RESTART_NO; + s->exec_context.ignore_sigpipe = false; + s->kill_context.kill_mode = KILL_PROCESS; + + /* We use the long description only if + * no short description is set. */ + + if (short_description) + description = short_description; + else if (chkconfig_description) + description = chkconfig_description; + else if (long_description) + description = long_description; + else + description = NULL; + + if (description) { + char *d; + + if (!(d = strappend(s->sysv_has_lsb ? "LSB: " : "SYSV: ", description))) { + r = -ENOMEM; + goto finish; + } + + u->description = d; + } + + /* The priority that has been set in /etc/rcN.d/ hierarchies + * takes precedence over what is stored as default in the LSB + * header */ + if (s->sysv_start_priority_from_rcnd >= 0) + s->sysv_start_priority = s->sysv_start_priority_from_rcnd; + + u->load_state = UNIT_LOADED; + r = 0; + +finish: + + if (f) + fclose(f); + + free(short_description); + free(long_description); + free(chkconfig_description); + + return r; +} + +static int service_load_sysv_name(Service *s, const char *name) { + char **p; + + assert(s); + assert(name); + + /* For SysV services we strip the *.sh suffixes. */ + if (endswith(name, ".sh.service")) + return -ENOENT; + + STRV_FOREACH(p, UNIT(s)->manager->lookup_paths.sysvinit_path) { + char *path; + int r; + + path = strjoin(*p, "/", name, NULL); + if (!path) + return -ENOMEM; + + assert(endswith(path, ".service")); + path[strlen(path)-8] = 0; + + r = service_load_sysv_path(s, path); + + if (r >= 0 && UNIT(s)->load_state == UNIT_STUB) { + /* Try *.sh source'able init scripts */ + strcat(path, ".sh"); + r = service_load_sysv_path(s, path); + } + free(path); + + if (r < 0) + return r; + + if (UNIT(s)->load_state != UNIT_STUB) + break; + } + + return 0; +} + +static int service_load_sysv(Service *s) { + const char *t; + Iterator i; + int r; + + assert(s); + + /* Load service data from SysV init scripts, preferably with + * LSB headers ... */ + + if (strv_isempty(UNIT(s)->manager->lookup_paths.sysvinit_path)) + return 0; + + if ((t = UNIT(s)->id)) + if ((r = service_load_sysv_name(s, t)) < 0) + return r; + + if (UNIT(s)->load_state == UNIT_STUB) + SET_FOREACH(t, UNIT(s)->names, i) { + if (t == UNIT(s)->id) + continue; + + if ((r = service_load_sysv_name(s, t)) < 0) + return r; + + if (UNIT(s)->load_state != UNIT_STUB) + break; + } + + return 0; +} +#endif + +static int fsck_fix_order(Service *s) { + Unit *other; + int r; + + assert(s); + + if (s->fsck_passno <= 0) + return 0; + + /* For each pair of services where both have an fsck priority + * we order things based on it. */ + + LIST_FOREACH(units_by_type, other, UNIT(s)->manager->units_by_type[UNIT_SERVICE]) { + Service *t; + UnitDependency d; + + t = SERVICE(other); + + if (s == t) + continue; + + if (UNIT(t)->load_state != UNIT_LOADED) + continue; + + if (t->fsck_passno <= 0) + continue; + + if (t->fsck_passno < s->fsck_passno) + d = UNIT_AFTER; + else if (t->fsck_passno > s->fsck_passno) + d = UNIT_BEFORE; + else + continue; + + r = unit_add_dependency(UNIT(s), d, UNIT(t), true); + if (r < 0) + return r; + } + + return 0; +} + +static int service_verify(Service *s) { + assert(s); + + if (UNIT(s)->load_state != UNIT_LOADED) + return 0; + + if (!s->exec_command[SERVICE_EXEC_START]) { + log_error_unit(UNIT(s)->id, + "%s lacks ExecStart setting. Refusing.", UNIT(s)->id); + return -EINVAL; + } + + if (s->type != SERVICE_ONESHOT && + s->exec_command[SERVICE_EXEC_START]->command_next) { + log_error_unit(UNIT(s)->id, + "%s has more than one ExecStart setting, which is only allowed for Type=oneshot services. Refusing.", UNIT(s)->id); + return -EINVAL; + } + + if (s->type == SERVICE_DBUS && !s->bus_name) { + log_error_unit(UNIT(s)->id, + "%s is of type D-Bus but no D-Bus service name has been specified. Refusing.", UNIT(s)->id); + return -EINVAL; + } + + if (s->bus_name && s->type != SERVICE_DBUS) + log_warning_unit(UNIT(s)->id, + "%s has a D-Bus service name specified, but is not of type dbus. Ignoring.", UNIT(s)->id); + + if (s->exec_context.pam_name && s->kill_context.kill_mode != KILL_CONTROL_GROUP) { + log_error_unit(UNIT(s)->id, + "%s has PAM enabled. Kill mode must be set to 'control-group'. Refusing.", UNIT(s)->id); + return -EINVAL; + } + + return 0; +} + +static int service_add_default_dependencies(Service *s) { + int r; + + assert(s); + + /* Add a number of automatic dependencies useful for the + * majority of services. */ + + /* First, pull in base system */ + if (UNIT(s)->manager->running_as == SYSTEMD_SYSTEM) { + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, + SPECIAL_BASIC_TARGET, NULL, true); + if (r < 0) + return r; + + } else if (UNIT(s)->manager->running_as == SYSTEMD_USER) { + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, + SPECIAL_SOCKETS_TARGET, NULL, true); + if (r < 0) + return r; + + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, + SPECIAL_TIMERS_TARGET, NULL, true); + if (r < 0) + return r; + + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, + SPECIAL_PATHS_TARGET, NULL, true); + if (r < 0) + return r; + } + + /* Second, activate normal shutdown */ + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_BEFORE, UNIT_CONFLICTS, + SPECIAL_SHUTDOWN_TARGET, NULL, true); + return r; +} + +static void service_fix_output(Service *s) { + assert(s); + + /* If nothing has been explicitly configured, patch default + * output in. If input is socket/tty we avoid this however, + * since in that case we want output to default to the same + * place as we read input from. */ + + if (s->exec_context.std_error == EXEC_OUTPUT_INHERIT && + s->exec_context.std_output == EXEC_OUTPUT_INHERIT && + s->exec_context.std_input == EXEC_INPUT_NULL) + s->exec_context.std_error = UNIT(s)->manager->default_std_error; + + if (s->exec_context.std_output == EXEC_OUTPUT_INHERIT && + s->exec_context.std_input == EXEC_INPUT_NULL) + s->exec_context.std_output = UNIT(s)->manager->default_std_output; +} + +static int service_load(Unit *u) { + int r; + Service *s = SERVICE(u); + + assert(s); + + /* Load a .service file */ + if ((r = unit_load_fragment(u)) < 0) + return r; + +#ifdef HAVE_SYSV_COMPAT + /* Load a classic init script as a fallback, if we couldn't find anything */ + if (u->load_state == UNIT_STUB) + if ((r = service_load_sysv(s)) < 0) + return r; +#endif + + /* Still nothing found? Then let's give up */ + if (u->load_state == UNIT_STUB) + return -ENOENT; + + /* We were able to load something, then let's add in the + * dropin directories. */ + if ((r = unit_load_dropin(unit_follow_merge(u))) < 0) + return r; + + /* This is a new unit? Then let's add in some extras */ + if (u->load_state == UNIT_LOADED) { + if (s->type == _SERVICE_TYPE_INVALID) + s->type = s->bus_name ? SERVICE_DBUS : SERVICE_SIMPLE; + + /* Oneshot services have disabled start timeout by default */ + if (s->type == SERVICE_ONESHOT && !s->start_timeout_defined) + s->timeout_start_usec = 0; + + service_fix_output(s); + + r = unit_add_exec_dependencies(u, &s->exec_context); + if (r < 0) + return r; + + r = unit_add_default_cgroups(u); + if (r < 0) + return r; + +#ifdef HAVE_SYSV_COMPAT + r = sysv_fix_order(s); + if (r < 0) + return r; +#endif + + r = fsck_fix_order(s); + if (r < 0) + return r; + + if (s->bus_name) + if ((r = unit_watch_bus_name(u, s->bus_name)) < 0) + return r; + + if (s->type == SERVICE_NOTIFY && s->notify_access == NOTIFY_NONE) + s->notify_access = NOTIFY_MAIN; + + if (s->watchdog_usec > 0 && s->notify_access == NOTIFY_NONE) + s->notify_access = NOTIFY_MAIN; + + if (s->type == SERVICE_DBUS || s->bus_name) { + r = unit_add_two_dependencies_by_name(u, UNIT_AFTER, UNIT_REQUIRES, + SPECIAL_DBUS_SOCKET, NULL, true); + if (r < 0) + return r; + } + + if (UNIT(s)->default_dependencies) { + r = service_add_default_dependencies(s); + if (r < 0) + return r; + } + + r = unit_exec_context_defaults(u, &s->exec_context); + if (r < 0) + return r; + } + + return service_verify(s); +} + +static void service_dump(Unit *u, FILE *f, const char *prefix) { + + ServiceExecCommand c; + Service *s = SERVICE(u); + const char *prefix2; + _cleanup_free_ char *p2 = NULL; + + assert(s); + + p2 = strappend(prefix, "\t"); + prefix2 = p2 ? p2 : prefix; + + fprintf(f, + "%sService State: %s\n" + "%sResult: %s\n" + "%sReload Result: %s\n" + "%sPermissionsStartOnly: %s\n" + "%sRootDirectoryStartOnly: %s\n" + "%sRemainAfterExit: %s\n" + "%sGuessMainPID: %s\n" + "%sType: %s\n" + "%sRestart: %s\n" + "%sNotifyAccess: %s\n", + prefix, service_state_to_string(s->state), + prefix, service_result_to_string(s->result), + prefix, service_result_to_string(s->reload_result), + prefix, yes_no(s->permissions_start_only), + prefix, yes_no(s->root_directory_start_only), + prefix, yes_no(s->remain_after_exit), + prefix, yes_no(s->guess_main_pid), + prefix, service_type_to_string(s->type), + prefix, service_restart_to_string(s->restart), + prefix, notify_access_to_string(s->notify_access)); + + if (s->control_pid > 0) + fprintf(f, + "%sControl PID: %lu\n", + prefix, (unsigned long) s->control_pid); + + if (s->main_pid > 0) + fprintf(f, + "%sMain PID: %lu\n" + "%sMain PID Known: %s\n" + "%sMain PID Alien: %s\n", + prefix, (unsigned long) s->main_pid, + prefix, yes_no(s->main_pid_known), + prefix, yes_no(s->main_pid_alien)); + + if (s->pid_file) + fprintf(f, + "%sPIDFile: %s\n", + prefix, s->pid_file); + + if (s->bus_name) + fprintf(f, + "%sBusName: %s\n" + "%sBus Name Good: %s\n", + prefix, s->bus_name, + prefix, yes_no(s->bus_name_good)); + + kill_context_dump(&s->kill_context, f, prefix); + exec_context_dump(&s->exec_context, f, prefix); + + for (c = 0; c < _SERVICE_EXEC_COMMAND_MAX; c++) { + + if (!s->exec_command[c]) + continue; + + fprintf(f, "%s-> %s:\n", + prefix, service_exec_command_to_string(c)); + + exec_command_dump_list(s->exec_command[c], f, prefix2); + } + +#ifdef HAVE_SYSV_COMPAT + if (s->is_sysv) + fprintf(f, + "%sSysV Init Script has LSB Header: %s\n" + "%sSysVEnabled: %s\n", + prefix, yes_no(s->sysv_has_lsb), + prefix, yes_no(s->sysv_enabled)); + + if (s->sysv_start_priority >= 0) + fprintf(f, + "%sSysVStartPriority: %i\n", + prefix, s->sysv_start_priority); + + if (s->sysv_runlevels) + fprintf(f, "%sSysVRunLevels: %s\n", + prefix, s->sysv_runlevels); +#endif + + if (s->fsck_passno > 0) + fprintf(f, + "%sFsckPassNo: %i\n", + prefix, s->fsck_passno); + + if (s->status_text) + fprintf(f, "%sStatus Text: %s\n", + prefix, s->status_text); +} + +static int service_load_pid_file(Service *s, bool may_warn) { + _cleanup_free_ char *k = NULL; + int r; + pid_t pid; + + assert(s); + + if (!s->pid_file) + return -ENOENT; + + r = read_one_line_file(s->pid_file, &k); + if (r < 0) { + if (may_warn) + log_info_unit(UNIT(s)->id, + "PID file %s not readable (yet?) after %s.", + s->pid_file, service_state_to_string(s->state)); + return r; + } + + r = parse_pid(k, &pid); + if (r < 0) { + if (may_warn) + log_info_unit(UNIT(s)->id, + "Failed to read PID from file %s: %s", + s->pid_file, strerror(-r)); + return r; + } + + if (kill(pid, 0) < 0 && errno != EPERM) { + if (may_warn) + log_info_unit(UNIT(s)->id, + "PID %lu read from file %s does not exist.", + (unsigned long) pid, s->pid_file); + return -ESRCH; + } + + if (s->main_pid_known) { + if (pid == s->main_pid) + return 0; + + log_debug_unit(UNIT(s)->id, + "Main PID changing: %lu -> %lu", + (unsigned long) s->main_pid, (unsigned long) pid); + service_unwatch_main_pid(s); + s->main_pid_known = false; + } else + log_debug_unit(UNIT(s)->id, + "Main PID loaded: %lu", (unsigned long) pid); + + r = service_set_main_pid(s, pid); + if (r < 0) + return r; + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) { + /* FIXME: we need to do something here */ + log_warning_unit(UNIT(s)->id, + "Failed to watch PID %lu from service %s", + (unsigned long) pid, UNIT(s)->id); + return r; + } + + return 0; +} + +static int service_search_main_pid(Service *s) { + pid_t pid; + int r; + + assert(s); + + /* If we know it anyway, don't ever fallback to unreliable + * heuristics */ + if (s->main_pid_known) + return 0; + + if (!s->guess_main_pid) + return 0; + + assert(s->main_pid <= 0); + + pid = cgroup_bonding_search_main_pid_list(UNIT(s)->cgroup_bondings); + if (pid <= 0) + return -ENOENT; + + log_debug_unit(UNIT(s)->id, + "Main PID guessed: %lu", (unsigned long) pid); + r = service_set_main_pid(s, pid); + if (r < 0) + return r; + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) + /* FIXME: we need to do something here */ + log_warning_unit(UNIT(s)->id, + "Failed to watch PID %lu from service %s", + (unsigned long) pid, UNIT(s)->id); + return r; + + return 0; +} + +static void service_notify_sockets_dead(Service *s, bool failed_permanent) { + Iterator i; + Unit *u; + + assert(s); + + /* Notifies all our sockets when we die */ + + if (s->socket_fd >= 0) + return; + + SET_FOREACH(u, UNIT(s)->dependencies[UNIT_TRIGGERED_BY], i) + if (u->type == UNIT_SOCKET) + socket_notify_service_dead(SOCKET(u), failed_permanent); + + return; +} + +static void service_set_state(Service *s, ServiceState state) { + ServiceState old_state; + const UnitActiveState *table; + assert(s); + + table = s->type == SERVICE_IDLE ? state_translation_table_idle : state_translation_table; + + old_state = s->state; + s->state = state; + + service_unwatch_pid_file(s); + + if (state != SERVICE_START_PRE && + state != SERVICE_START && + state != SERVICE_START_POST && + state != SERVICE_RELOAD && + state != SERVICE_STOP && + state != SERVICE_STOP_SIGTERM && + state != SERVICE_STOP_SIGKILL && + state != SERVICE_STOP_POST && + state != SERVICE_FINAL_SIGTERM && + state != SERVICE_FINAL_SIGKILL && + state != SERVICE_AUTO_RESTART) + unit_unwatch_timer(UNIT(s), &s->timer_watch); + + if (state != SERVICE_START && + state != SERVICE_START_POST && + state != SERVICE_RUNNING && + state != SERVICE_RELOAD && + state != SERVICE_STOP && + state != SERVICE_STOP_SIGTERM && + state != SERVICE_STOP_SIGKILL) { + service_unwatch_main_pid(s); + s->main_command = NULL; + } + + if (state != SERVICE_START_PRE && + state != SERVICE_START && + state != SERVICE_START_POST && + state != SERVICE_RELOAD && + state != SERVICE_STOP && + state != SERVICE_STOP_SIGTERM && + state != SERVICE_STOP_SIGKILL && + state != SERVICE_STOP_POST && + state != SERVICE_FINAL_SIGTERM && + state != SERVICE_FINAL_SIGKILL) { + service_unwatch_control_pid(s); + s->control_command = NULL; + s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; + } + + if (state == SERVICE_FAILED) + service_notify_sockets_dead(s, s->result == SERVICE_FAILURE_START_LIMIT); + + if (state == SERVICE_DEAD || + state == SERVICE_STOP || + state == SERVICE_STOP_SIGTERM || + state == SERVICE_STOP_SIGKILL || + state == SERVICE_STOP_POST || + state == SERVICE_FINAL_SIGTERM || + state == SERVICE_FINAL_SIGKILL || + state == SERVICE_AUTO_RESTART) + service_notify_sockets_dead(s, false); + + if (state != SERVICE_START_PRE && + state != SERVICE_START && + state != SERVICE_START_POST && + state != SERVICE_RUNNING && + state != SERVICE_RELOAD && + state != SERVICE_STOP && + state != SERVICE_STOP_SIGTERM && + state != SERVICE_STOP_SIGKILL && + state != SERVICE_STOP_POST && + state != SERVICE_FINAL_SIGTERM && + state != SERVICE_FINAL_SIGKILL && + !(state == SERVICE_DEAD && UNIT(s)->job)) { + service_close_socket_fd(s); + service_connection_unref(s); + } + + if (state == SERVICE_STOP || state == SERVICE_STOP_SIGTERM) + service_stop_watchdog(s); + + /* For the inactive states unit_notify() will trim the cgroup, + * but for exit we have to do that ourselves... */ + if (state == SERVICE_EXITED && UNIT(s)->manager->n_reloading <= 0) + cgroup_bonding_trim_list(UNIT(s)->cgroup_bondings, true); + + if (old_state != state) + log_debug_unit(UNIT(s)->id, + "%s changed %s -> %s", UNIT(s)->id, + service_state_to_string(old_state), + service_state_to_string(state)); + + unit_notify(UNIT(s), table[old_state], table[state], s->reload_result == SERVICE_SUCCESS); + s->reload_result = SERVICE_SUCCESS; +} + +static int service_coldplug(Unit *u) { + Service *s = SERVICE(u); + int r; + + assert(s); + assert(s->state == SERVICE_DEAD); + + if (s->deserialized_state != s->state) { + + if (s->deserialized_state == SERVICE_START_PRE || + s->deserialized_state == SERVICE_START || + s->deserialized_state == SERVICE_START_POST || + s->deserialized_state == SERVICE_RELOAD || + s->deserialized_state == SERVICE_STOP || + s->deserialized_state == SERVICE_STOP_SIGTERM || + s->deserialized_state == SERVICE_STOP_SIGKILL || + s->deserialized_state == SERVICE_STOP_POST || + s->deserialized_state == SERVICE_FINAL_SIGTERM || + s->deserialized_state == SERVICE_FINAL_SIGKILL || + s->deserialized_state == SERVICE_AUTO_RESTART) { + if (s->deserialized_state == SERVICE_AUTO_RESTART || s->timeout_start_usec > 0) { + usec_t k; + + k = s->deserialized_state == SERVICE_AUTO_RESTART ? s->restart_usec : s->timeout_start_usec; + + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, k, &s->timer_watch); + if (r < 0) + return r; + } + } + + if ((s->deserialized_state == SERVICE_START && + (s->type == SERVICE_FORKING || + s->type == SERVICE_DBUS || + s->type == SERVICE_ONESHOT || + s->type == SERVICE_NOTIFY)) || + s->deserialized_state == SERVICE_START_POST || + s->deserialized_state == SERVICE_RUNNING || + s->deserialized_state == SERVICE_RELOAD || + s->deserialized_state == SERVICE_STOP || + s->deserialized_state == SERVICE_STOP_SIGTERM || + s->deserialized_state == SERVICE_STOP_SIGKILL) + if (s->main_pid > 0) { + r = unit_watch_pid(UNIT(s), s->main_pid); + if (r < 0) + return r; + } + + if (s->deserialized_state == SERVICE_START_PRE || + s->deserialized_state == SERVICE_START || + s->deserialized_state == SERVICE_START_POST || + s->deserialized_state == SERVICE_RELOAD || + s->deserialized_state == SERVICE_STOP || + s->deserialized_state == SERVICE_STOP_SIGTERM || + s->deserialized_state == SERVICE_STOP_SIGKILL || + s->deserialized_state == SERVICE_STOP_POST || + s->deserialized_state == SERVICE_FINAL_SIGTERM || + s->deserialized_state == SERVICE_FINAL_SIGKILL) + if (s->control_pid > 0) { + r = unit_watch_pid(UNIT(s), s->control_pid); + if (r < 0) + return r; + } + + if (s->deserialized_state == SERVICE_START_POST || + s->deserialized_state == SERVICE_RUNNING) + service_handle_watchdog(s); + + service_set_state(s, s->deserialized_state); + } + return 0; +} + +static int service_collect_fds(Service *s, int **fds, unsigned *n_fds) { + Iterator i; + int r; + int *rfds = NULL; + unsigned rn_fds = 0; + Unit *u; + + assert(s); + assert(fds); + assert(n_fds); + + if (s->socket_fd >= 0) + return 0; + + SET_FOREACH(u, UNIT(s)->dependencies[UNIT_TRIGGERED_BY], i) { + int *cfds; + unsigned cn_fds; + Socket *sock; + + if (u->type != UNIT_SOCKET) + continue; + + sock = SOCKET(u); + + r = socket_collect_fds(sock, &cfds, &cn_fds); + if (r < 0) + goto fail; + + if (!cfds) + continue; + + if (!rfds) { + rfds = cfds; + rn_fds = cn_fds; + } else { + int *t; + + t = new(int, rn_fds+cn_fds); + if (!t) { + free(cfds); + r = -ENOMEM; + goto fail; + } + + memcpy(t, rfds, rn_fds * sizeof(int)); + memcpy(t+rn_fds, cfds, cn_fds * sizeof(int)); + free(rfds); + free(cfds); + + rfds = t; + rn_fds = rn_fds+cn_fds; + } + } + + *fds = rfds; + *n_fds = rn_fds; + + return 0; + +fail: + free(rfds); + + return r; +} + +static int service_spawn( + Service *s, + ExecCommand *c, + bool timeout, + bool pass_fds, + bool apply_permissions, + bool apply_chroot, + bool apply_tty_stdin, + bool set_notify_socket, + bool is_control, + pid_t *_pid) { + + pid_t pid; + int r; + int *fds = NULL; + _cleanup_free_ int *fdsbuf = NULL; + unsigned n_fds = 0, n_env = 0; + _cleanup_strv_free_ char + **argv = NULL, **final_env = NULL, **our_env = NULL; + + assert(s); + assert(c); + assert(_pid); + + if (pass_fds || + s->exec_context.std_input == EXEC_INPUT_SOCKET || + s->exec_context.std_output == EXEC_OUTPUT_SOCKET || + s->exec_context.std_error == EXEC_OUTPUT_SOCKET) { + + if (s->socket_fd >= 0) { + fds = &s->socket_fd; + n_fds = 1; + } else { + r = service_collect_fds(s, &fdsbuf, &n_fds); + if (r < 0) + goto fail; + + fds = fdsbuf; + } + } + + if (timeout && s->timeout_start_usec) { + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, + s->timeout_start_usec, &s->timer_watch); + if (r < 0) + goto fail; + } else + unit_unwatch_timer(UNIT(s), &s->timer_watch); + + argv = unit_full_printf_strv(UNIT(s), c->argv); + if (!argv) { + r = -ENOMEM; + goto fail; + } + + our_env = new0(char*, 5); + if (!our_env) { + r = -ENOMEM; + goto fail; + } + + if (set_notify_socket) + if (asprintf(our_env + n_env++, "NOTIFY_SOCKET=%s", UNIT(s)->manager->notify_socket) < 0) { + r = -ENOMEM; + goto fail; + } + + if (s->main_pid > 0) + if (asprintf(our_env + n_env++, "MAINPID=%lu", (unsigned long) s->main_pid) < 0) { + r = -ENOMEM; + goto fail; + } + + if (s->watchdog_usec > 0) + if (asprintf(our_env + n_env++, "WATCHDOG_USEC=%llu", (unsigned long long) s->watchdog_usec) < 0) { + r = -ENOMEM; + goto fail; + } + + if (s->meta.manager->running_as != SYSTEMD_SYSTEM) + if (asprintf(our_env + n_env++, "MANAGERPID=%lu", (unsigned long) getpid()) < 0) { + r = -ENOMEM; + goto fail; + } + + final_env = strv_env_merge(2, UNIT(s)->manager->environment, our_env, NULL); + if (!final_env) { + r = -ENOMEM; + goto fail; + } + + r = exec_spawn(c, + argv, + &s->exec_context, + fds, n_fds, + final_env, + apply_permissions, + apply_chroot, + apply_tty_stdin, + UNIT(s)->manager->confirm_spawn, + UNIT(s)->cgroup_bondings, + UNIT(s)->cgroup_attributes, + is_control ? "control" : NULL, + UNIT(s)->id, + s->type == SERVICE_IDLE ? UNIT(s)->manager->idle_pipe : NULL, + &pid); + if (r < 0) + goto fail; + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) + /* FIXME: we need to do something here */ + goto fail; + + *_pid = pid; + + return 0; + +fail: + if (timeout) + unit_unwatch_timer(UNIT(s), &s->timer_watch); + + return r; +} + +static int main_pid_good(Service *s) { + assert(s); + + /* Returns 0 if the pid is dead, 1 if it is good, -1 if we + * don't know */ + + /* If we know the pid file, then lets just check if it is + * still valid */ + if (s->main_pid_known) { + + /* If it's an alien child let's check if it is still + * alive ... */ + if (s->main_pid_alien) + return kill(s->main_pid, 0) >= 0 || errno != ESRCH; + + /* .. otherwise assume we'll get a SIGCHLD for it, + * which we really should wait for to collect exit + * status and code */ + return s->main_pid > 0; + } + + /* We don't know the pid */ + return -EAGAIN; +} + +_pure_ static int control_pid_good(Service *s) { + assert(s); + + return s->control_pid > 0; +} + +static int cgroup_good(Service *s) { + int r; + + assert(s); + + r = cgroup_bonding_is_empty_list(UNIT(s)->cgroup_bondings); + if (r < 0) + return r; + + return !r; +} + +static void service_enter_dead(Service *s, ServiceResult f, bool allow_restart) { + int r; + assert(s); + + if (f != SERVICE_SUCCESS) + s->result = f; + + service_set_state(s, s->result != SERVICE_SUCCESS ? SERVICE_FAILED : SERVICE_DEAD); + + if (allow_restart && + !s->forbid_restart && + (s->restart == SERVICE_RESTART_ALWAYS || + (s->restart == SERVICE_RESTART_ON_SUCCESS && s->result == SERVICE_SUCCESS) || + (s->restart == SERVICE_RESTART_ON_FAILURE && s->result != SERVICE_SUCCESS) || + (s->restart == SERVICE_RESTART_ON_ABORT && (s->result == SERVICE_FAILURE_SIGNAL || + s->result == SERVICE_FAILURE_CORE_DUMP))) && + (s->result != SERVICE_FAILURE_EXIT_CODE || + !set_contains(s->restart_ignore_status.code, INT_TO_PTR(s->main_exec_status.status))) && + (s->result != SERVICE_FAILURE_SIGNAL || + !set_contains(s->restart_ignore_status.signal, INT_TO_PTR(s->main_exec_status.status))) + ) { + + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, s->restart_usec, &s->timer_watch); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_AUTO_RESTART); + } + + s->forbid_restart = false; + + /* we want fresh tmpdirs in case service is started again immediately */ + exec_context_tmp_dirs_done(&s->exec_context); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run install restart timer: %s", + UNIT(s)->id, strerror(-r)); + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, false); +} + +static void service_enter_signal(Service *s, ServiceState state, ServiceResult f); + +static void service_enter_stop_post(Service *s, ServiceResult f) { + int r; + assert(s); + + if (f != SERVICE_SUCCESS) + s->result = f; + + service_unwatch_control_pid(s); + + s->control_command = s->exec_command[SERVICE_EXEC_STOP_POST]; + if (s->control_command) { + s->control_command_id = SERVICE_EXEC_STOP_POST; + + r = service_spawn(s, + s->control_command, + true, + false, + !s->permissions_start_only, + !s->root_directory_start_only, + true, + false, + true, + &s->control_pid); + if (r < 0) + goto fail; + + + service_set_state(s, SERVICE_STOP_POST); + } else + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_SUCCESS); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'stop-post' task: %s", + UNIT(s)->id, strerror(-r)); + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); +} + +static void service_enter_signal(Service *s, ServiceState state, ServiceResult f) { + int r; + + assert(s); + + if (f != SERVICE_SUCCESS) + s->result = f; + + r = unit_kill_context( + UNIT(s), + &s->kill_context, + state != SERVICE_STOP_SIGTERM && state != SERVICE_FINAL_SIGTERM, + s->main_pid, + s->control_pid, + s->main_pid_alien); + if (r < 0) + goto fail; + + if (r > 0) { + if (s->timeout_stop_usec > 0) { + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, + s->timeout_stop_usec, &s->timer_watch); + if (r < 0) + goto fail; + } + + service_set_state(s, state); + } else if (state == SERVICE_STOP_SIGTERM || state == SERVICE_STOP_SIGKILL) + service_enter_stop_post(s, SERVICE_SUCCESS); + else + service_enter_dead(s, SERVICE_SUCCESS, true); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to kill processes: %s", UNIT(s)->id, strerror(-r)); + + if (state == SERVICE_STOP_SIGTERM || state == SERVICE_STOP_SIGKILL) + service_enter_stop_post(s, SERVICE_FAILURE_RESOURCES); + else + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, true); +} + +static void service_enter_stop(Service *s, ServiceResult f) { + int r; + + assert(s); + + if (f != SERVICE_SUCCESS) + s->result = f; + + service_unwatch_control_pid(s); + + s->control_command = s->exec_command[SERVICE_EXEC_STOP]; + if (s->control_command) { + s->control_command_id = SERVICE_EXEC_STOP; + + r = service_spawn(s, + s->control_command, + true, + false, + !s->permissions_start_only, + !s->root_directory_start_only, + false, + false, + true, + &s->control_pid); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_STOP); + } else + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_SUCCESS); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'stop' task: %s", UNIT(s)->id, strerror(-r)); + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_RESOURCES); +} + +static void service_enter_running(Service *s, ServiceResult f) { + int main_pid_ok, cgroup_ok; + assert(s); + + if (f != SERVICE_SUCCESS) + s->result = f; + + main_pid_ok = main_pid_good(s); + cgroup_ok = cgroup_good(s); + + if ((main_pid_ok > 0 || (main_pid_ok < 0 && cgroup_ok != 0)) && + (s->bus_name_good || s->type != SERVICE_DBUS)) + service_set_state(s, SERVICE_RUNNING); + else if (s->remain_after_exit) + service_set_state(s, SERVICE_EXITED); + else + service_enter_stop(s, SERVICE_SUCCESS); +} + +static void service_enter_start_post(Service *s) { + int r; + assert(s); + + service_unwatch_control_pid(s); + + if (s->watchdog_usec > 0) + service_reset_watchdog(s); + + s->control_command = s->exec_command[SERVICE_EXEC_START_POST]; + if (s->control_command) { + s->control_command_id = SERVICE_EXEC_START_POST; + + r = service_spawn(s, + s->control_command, + true, + false, + !s->permissions_start_only, + !s->root_directory_start_only, + false, + false, + true, + &s->control_pid); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_START_POST); + } else + service_enter_running(s, SERVICE_SUCCESS); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'start-post' task: %s", UNIT(s)->id, strerror(-r)); + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); +} + +static void service_enter_start(Service *s) { + pid_t pid; + int r; + ExecCommand *c; + + assert(s); + + assert(s->exec_command[SERVICE_EXEC_START]); + assert(!s->exec_command[SERVICE_EXEC_START]->command_next || s->type == SERVICE_ONESHOT); + + if (s->type == SERVICE_FORKING) + service_unwatch_control_pid(s); + else + service_unwatch_main_pid(s); + + /* We want to ensure that nobody leaks processes from + * START_PRE here, so let's go on a killing spree, People + * should not spawn long running processes from START_PRE. */ + cgroup_bonding_kill_list(UNIT(s)->cgroup_bondings, SIGKILL, true, true, NULL, "control"); + + if (s->type == SERVICE_FORKING) { + s->control_command_id = SERVICE_EXEC_START; + c = s->control_command = s->exec_command[SERVICE_EXEC_START]; + + s->main_command = NULL; + } else { + s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; + s->control_command = NULL; + + c = s->main_command = s->exec_command[SERVICE_EXEC_START]; + } + + r = service_spawn(s, + c, + s->type == SERVICE_FORKING || s->type == SERVICE_DBUS || + s->type == SERVICE_NOTIFY || s->type == SERVICE_ONESHOT, + true, + true, + true, + true, + s->notify_access != NOTIFY_NONE, + false, + &pid); + if (r < 0) + goto fail; + + if (s->type == SERVICE_SIMPLE || s->type == SERVICE_IDLE) { + /* For simple services we immediately start + * the START_POST binaries. */ + + service_set_main_pid(s, pid); + service_enter_start_post(s); + + } else if (s->type == SERVICE_FORKING) { + + /* For forking services we wait until the start + * process exited. */ + + s->control_pid = pid; + service_set_state(s, SERVICE_START); + + } else if (s->type == SERVICE_ONESHOT || + s->type == SERVICE_DBUS || + s->type == SERVICE_NOTIFY) { + + /* For oneshot services we wait until the start + * process exited, too, but it is our main process. */ + + /* For D-Bus services we know the main pid right away, + * but wait for the bus name to appear on the + * bus. Notify services are similar. */ + + service_set_main_pid(s, pid); + service_set_state(s, SERVICE_START); + } else + assert_not_reached("Unknown service type"); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'start' task: %s", UNIT(s)->id, strerror(-r)); + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); +} + +static void service_enter_start_pre(Service *s) { + int r; + + assert(s); + + service_unwatch_control_pid(s); + + s->control_command = s->exec_command[SERVICE_EXEC_START_PRE]; + if (s->control_command) { + + /* Before we start anything, let's clear up what might + * be left from previous runs. */ + cgroup_bonding_kill_list(UNIT(s)->cgroup_bondings, SIGKILL, + true,true, NULL, "control"); + + s->control_command_id = SERVICE_EXEC_START_PRE; + + r = service_spawn(s, + s->control_command, + true, + false, + !s->permissions_start_only, + !s->root_directory_start_only, + true, + false, + true, + &s->control_pid); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_START_PRE); + } else + service_enter_start(s); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'start-pre' task: %s", UNIT(s)->id, strerror(-r)); + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, true); +} + +static void service_enter_restart(Service *s) { + int r; + DBusError error; + + assert(s); + dbus_error_init(&error); + + if (UNIT(s)->job && UNIT(s)->job->type == JOB_STOP) { + /* Don't restart things if we are going down anyway */ + log_info_unit(UNIT(s)->id, + "Stop job pending for unit, delaying automatic restart."); + + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, s->restart_usec, &s->timer_watch); + if (r < 0) + goto fail; + + return; + } + + /* Any units that are bound to this service must also be + * restarted. We use JOB_RESTART (instead of the more obvious + * JOB_START) here so that those dependency jobs will be added + * as well. */ + r = manager_add_job(UNIT(s)->manager, JOB_RESTART, UNIT(s), JOB_FAIL, false, &error, NULL); + if (r < 0) + goto fail; + + /* Note that we stay in the SERVICE_AUTO_RESTART state here, + * it will be canceled as part of the service_stop() call that + * is executed as part of JOB_RESTART. */ + + log_debug_unit(UNIT(s)->id, + "%s scheduled restart job.", UNIT(s)->id); + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to schedule restart job: %s", + UNIT(s)->id, bus_error(&error, -r)); + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, false); + + dbus_error_free(&error); +} + +static void service_enter_reload(Service *s) { + int r; + + assert(s); + + service_unwatch_control_pid(s); + + s->control_command = s->exec_command[SERVICE_EXEC_RELOAD]; + if (s->control_command) { + s->control_command_id = SERVICE_EXEC_RELOAD; + + r = service_spawn(s, + s->control_command, + true, + false, + !s->permissions_start_only, + !s->root_directory_start_only, + false, + false, + true, + &s->control_pid); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_RELOAD); + } else + service_enter_running(s, SERVICE_SUCCESS); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'reload' task: %s", + UNIT(s)->id, strerror(-r)); + s->reload_result = SERVICE_FAILURE_RESOURCES; + service_enter_running(s, SERVICE_SUCCESS); +} + +static void service_run_next_control(Service *s) { + int r; + + assert(s); + assert(s->control_command); + assert(s->control_command->command_next); + + assert(s->control_command_id != SERVICE_EXEC_START); + + s->control_command = s->control_command->command_next; + service_unwatch_control_pid(s); + + r = service_spawn(s, + s->control_command, + true, + false, + !s->permissions_start_only, + !s->root_directory_start_only, + s->control_command_id == SERVICE_EXEC_START_PRE || + s->control_command_id == SERVICE_EXEC_STOP_POST, + false, + true, + &s->control_pid); + if (r < 0) + goto fail; + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run next control task: %s", + UNIT(s)->id, strerror(-r)); + + if (s->state == SERVICE_START_PRE) + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); + else if (s->state == SERVICE_STOP) + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_RESOURCES); + else if (s->state == SERVICE_STOP_POST) + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, true); + else if (s->state == SERVICE_RELOAD) { + s->reload_result = SERVICE_FAILURE_RESOURCES; + service_enter_running(s, SERVICE_SUCCESS); + } else + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); +} + +static void service_run_next_main(Service *s) { + pid_t pid; + int r; + + assert(s); + assert(s->main_command); + assert(s->main_command->command_next); + assert(s->type == SERVICE_ONESHOT); + + s->main_command = s->main_command->command_next; + service_unwatch_main_pid(s); + + r = service_spawn(s, + s->main_command, + true, + true, + true, + true, + true, + s->notify_access != NOTIFY_NONE, + false, + &pid); + if (r < 0) + goto fail; + + service_set_main_pid(s, pid); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run next main task: %s", UNIT(s)->id, strerror(-r)); + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); +} + +static int service_start_limit_test(Service *s) { + assert(s); + + if (ratelimit_test(&s->start_limit)) + return 0; + + switch (s->start_limit_action) { + + case SERVICE_START_LIMIT_NONE: + log_warning_unit(UNIT(s)->id, + "%s start request repeated too quickly, refusing to start.", + UNIT(s)->id); + break; + + case SERVICE_START_LIMIT_REBOOT: { + DBusError error; + int r; + + dbus_error_init(&error); + + log_warning_unit(UNIT(s)->id, + "%s start request repeated too quickly, rebooting.", UNIT(s)->id); + + r = manager_add_job_by_name(UNIT(s)->manager, JOB_START, + SPECIAL_REBOOT_TARGET, JOB_REPLACE, + true, &error, NULL); + if (r < 0) { + log_error_unit(UNIT(s)->id, + "Failed to reboot: %s.", bus_error(&error, r)); + dbus_error_free(&error); + } + + break; + } + + case SERVICE_START_LIMIT_REBOOT_FORCE: + log_warning_unit(UNIT(s)->id, + "%s start request repeated too quickly, forcibly rebooting.", UNIT(s)->id); + UNIT(s)->manager->exit_code = MANAGER_REBOOT; + break; + + case SERVICE_START_LIMIT_REBOOT_IMMEDIATE: + log_warning_unit(UNIT(s)->id, + "%s start request repeated too quickly, rebooting immediately.", UNIT(s)->id); + sync(); + reboot(RB_AUTOBOOT); + break; + + default: + log_error_unit(UNIT(s)->id, + "start limit action=%i", s->start_limit_action); + assert_not_reached("Unknown StartLimitAction."); + } + + return -ECANCELED; +} + +static int service_start(Unit *u) { + Service *s = SERVICE(u); + int r; + + assert(s); + + /* We cannot fulfill this request right now, try again later + * please! */ + if (s->state == SERVICE_STOP || + s->state == SERVICE_STOP_SIGTERM || + s->state == SERVICE_STOP_SIGKILL || + s->state == SERVICE_STOP_POST || + s->state == SERVICE_FINAL_SIGTERM || + s->state == SERVICE_FINAL_SIGKILL) + return -EAGAIN; + + /* Already on it! */ + if (s->state == SERVICE_START_PRE || + s->state == SERVICE_START || + s->state == SERVICE_START_POST) + return 0; + + /* A service that will be restarted must be stopped first to + * trigger BindsTo and/or OnFailure dependencies. If a user + * does not want to wait for the holdoff time to elapse, the + * service should be manually restarted, not started. We + * simply return EAGAIN here, so that any start jobs stay + * queued, and assume that the auto restart timer will + * eventually trigger the restart. */ + if (s->state == SERVICE_AUTO_RESTART) + return -EAGAIN; + + assert(s->state == SERVICE_DEAD || s->state == SERVICE_FAILED); + + /* Make sure we don't enter a busy loop of some kind. */ + r = service_start_limit_test(s); + if (r < 0) { + service_enter_dead(s, SERVICE_FAILURE_START_LIMIT, false); + return r; + } + + s->result = SERVICE_SUCCESS; + s->reload_result = SERVICE_SUCCESS; + s->main_pid_known = false; + s->main_pid_alien = false; + s->forbid_restart = false; + + service_enter_start_pre(s); + return 0; +} + +static int service_stop(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + /* Don't create restart jobs from here. */ + s->forbid_restart = true; + + /* Already on it */ + if (s->state == SERVICE_STOP || + s->state == SERVICE_STOP_SIGTERM || + s->state == SERVICE_STOP_SIGKILL || + s->state == SERVICE_STOP_POST || + s->state == SERVICE_FINAL_SIGTERM || + s->state == SERVICE_FINAL_SIGKILL) + return 0; + + /* A restart will be scheduled or is in progress. */ + if (s->state == SERVICE_AUTO_RESTART) { + service_set_state(s, SERVICE_DEAD); + return 0; + } + + /* If there's already something running we go directly into + * kill mode. */ + if (s->state == SERVICE_START_PRE || + s->state == SERVICE_START || + s->state == SERVICE_START_POST || + s->state == SERVICE_RELOAD) { + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_SUCCESS); + return 0; + } + + assert(s->state == SERVICE_RUNNING || + s->state == SERVICE_EXITED); + + service_enter_stop(s, SERVICE_SUCCESS); + return 0; +} + +static int service_reload(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + assert(s->state == SERVICE_RUNNING || s->state == SERVICE_EXITED); + + service_enter_reload(s); + return 0; +} + +_pure_ static bool service_can_reload(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + return !!s->exec_command[SERVICE_EXEC_RELOAD]; +} + +static int service_serialize(Unit *u, FILE *f, FDSet *fds) { + Service *s = SERVICE(u); + + assert(u); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", service_state_to_string(s->state)); + unit_serialize_item(u, f, "result", service_result_to_string(s->result)); + unit_serialize_item(u, f, "reload-result", service_result_to_string(s->reload_result)); + + if (s->control_pid > 0) + unit_serialize_item_format(u, f, "control-pid", "%lu", + (unsigned long) s->control_pid); + + if (s->main_pid_known && s->main_pid > 0) + unit_serialize_item_format(u, f, "main-pid", "%lu", (unsigned long) s->main_pid); + + unit_serialize_item(u, f, "main-pid-known", yes_no(s->main_pid_known)); + + if (s->status_text) + unit_serialize_item(u, f, "status-text", s->status_text); + + /* FIXME: There's a minor uncleanliness here: if there are + * multiple commands attached here, we will start from the + * first one again */ + if (s->control_command_id >= 0) + unit_serialize_item(u, f, "control-command", + service_exec_command_to_string(s->control_command_id)); + + if (s->socket_fd >= 0) { + int copy; + + if ((copy = fdset_put_dup(fds, s->socket_fd)) < 0) + return copy; + + unit_serialize_item_format(u, f, "socket-fd", "%i", copy); + } + + if (s->main_exec_status.pid > 0) { + unit_serialize_item_format(u, f, "main-exec-status-pid", "%lu", + (unsigned long) s->main_exec_status.pid); + dual_timestamp_serialize(f, "main-exec-status-start", + &s->main_exec_status.start_timestamp); + dual_timestamp_serialize(f, "main-exec-status-exit", + &s->main_exec_status.exit_timestamp); + + if (dual_timestamp_is_set(&s->main_exec_status.exit_timestamp)) { + unit_serialize_item_format(u, f, "main-exec-status-code", "%i", + s->main_exec_status.code); + unit_serialize_item_format(u, f, "main-exec-status-status", "%i", + s->main_exec_status.status); + } + } + if (dual_timestamp_is_set(&s->watchdog_timestamp)) + dual_timestamp_serialize(f, "watchdog-timestamp", + &s->watchdog_timestamp); + + if (s->exec_context.tmp_dir) + unit_serialize_item(u, f, "tmp-dir", s->exec_context.tmp_dir); + + if (s->exec_context.var_tmp_dir) + unit_serialize_item(u, f, "var-tmp-dir", s->exec_context.var_tmp_dir); + + return 0; +} + +static int service_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Service *s = SERVICE(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + ServiceState state; + + state = service_state_from_string(value); + if (state < 0) + log_debug_unit(u->id, "Failed to parse state value %s", value); + else + s->deserialized_state = state; + } else if (streq(key, "result")) { + ServiceResult f; + + f = service_result_from_string(value); + if (f < 0) + log_debug_unit(u->id, "Failed to parse result value %s", value); + else if (f != SERVICE_SUCCESS) + s->result = f; + + } else if (streq(key, "reload-result")) { + ServiceResult f; + + f = service_result_from_string(value); + if (f < 0) + log_debug_unit(u->id, "Failed to parse reload result value %s", value); + else if (f != SERVICE_SUCCESS) + s->reload_result = f; + + } else if (streq(key, "control-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_debug_unit(u->id, "Failed to parse control-pid value %s", value); + else + s->control_pid = pid; + } else if (streq(key, "main-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_debug_unit(u->id, "Failed to parse main-pid value %s", value); + else + service_set_main_pid(s, (pid_t) pid); + } else if (streq(key, "main-pid-known")) { + int b; + + b = parse_boolean(value); + if (b < 0) + log_debug_unit(u->id, "Failed to parse main-pid-known value %s", value); + else + s->main_pid_known = b; + } else if (streq(key, "status-text")) { + char *t; + + t = strdup(value); + if (!t) + log_oom(); + else { + free(s->status_text); + s->status_text = t; + } + + } else if (streq(key, "control-command")) { + ServiceExecCommand id; + + id = service_exec_command_from_string(value); + if (id < 0) + log_debug_unit(u->id, "Failed to parse exec-command value %s", value); + else { + s->control_command_id = id; + s->control_command = s->exec_command[id]; + } + } else if (streq(key, "socket-fd")) { + int fd; + + if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_debug_unit(u->id, "Failed to parse socket-fd value %s", value); + else { + + if (s->socket_fd >= 0) + close_nointr_nofail(s->socket_fd); + s->socket_fd = fdset_remove(fds, fd); + } + } else if (streq(key, "main-exec-status-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_debug_unit(u->id, "Failed to parse main-exec-status-pid value %s", value); + else + s->main_exec_status.pid = pid; + } else if (streq(key, "main-exec-status-code")) { + int i; + + if (safe_atoi(value, &i) < 0) + log_debug_unit(u->id, "Failed to parse main-exec-status-code value %s", value); + else + s->main_exec_status.code = i; + } else if (streq(key, "main-exec-status-status")) { + int i; + + if (safe_atoi(value, &i) < 0) + log_debug_unit(u->id, "Failed to parse main-exec-status-status value %s", value); + else + s->main_exec_status.status = i; + } else if (streq(key, "main-exec-status-start")) + dual_timestamp_deserialize(value, &s->main_exec_status.start_timestamp); + else if (streq(key, "main-exec-status-exit")) + dual_timestamp_deserialize(value, &s->main_exec_status.exit_timestamp); + else if (streq(key, "watchdog-timestamp")) + dual_timestamp_deserialize(value, &s->watchdog_timestamp); + else if (streq(key, "tmp-dir")) { + char *t; + + t = strdup(value); + if (!t) + return log_oom(); + + s->exec_context.tmp_dir = t; + } else if (streq(key, "var-tmp-dir")) { + char *t; + + t = strdup(value); + if (!t) + return log_oom(); + + s->exec_context.var_tmp_dir = t; + } else + log_debug_unit(u->id, "Unknown serialization key '%s'", key); + + return 0; +} + +_pure_ static UnitActiveState service_active_state(Unit *u) { + const UnitActiveState *table; + + assert(u); + + table = SERVICE(u)->type == SERVICE_IDLE ? state_translation_table_idle : state_translation_table; + + return table[SERVICE(u)->state]; +} + +static const char *service_sub_state_to_string(Unit *u) { + assert(u); + + return service_state_to_string(SERVICE(u)->state); +} + +static bool service_check_gc(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + /* Never clean up services that still have a process around, + * even if the service is formally dead. */ + if (cgroup_good(s) > 0 || + main_pid_good(s) > 0 || + control_pid_good(s) > 0) + return true; + +#ifdef HAVE_SYSV_COMPAT + if (s->is_sysv) + return true; +#endif + + return false; +} + +_pure_ static bool service_check_snapshot(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + return !s->got_socket_fd; +} + +static int service_retry_pid_file(Service *s) { + int r; + + assert(s->pid_file); + assert(s->state == SERVICE_START || s->state == SERVICE_START_POST); + + r = service_load_pid_file(s, false); + if (r < 0) + return r; + + service_unwatch_pid_file(s); + + service_enter_running(s, SERVICE_SUCCESS); + return 0; +} + +static int service_watch_pid_file(Service *s) { + int r; + + log_debug_unit(UNIT(s)->id, + "Setting watch for %s's PID file %s", + UNIT(s)->id, s->pid_file_pathspec->path); + r = path_spec_watch(s->pid_file_pathspec, UNIT(s)); + if (r < 0) + goto fail; + + /* the pidfile might have appeared just before we set the watch */ + log_debug_unit(UNIT(s)->id, + "Trying to read %s's PID file %s in case it changed", + UNIT(s)->id, s->pid_file_pathspec->path); + service_retry_pid_file(s); + + return 0; +fail: + log_error_unit(UNIT(s)->id, + "Failed to set a watch for %s's PID file %s: %s", + UNIT(s)->id, s->pid_file_pathspec->path, strerror(-r)); + service_unwatch_pid_file(s); + return r; +} + +static int service_demand_pid_file(Service *s) { + PathSpec *ps; + + assert(s->pid_file); + assert(!s->pid_file_pathspec); + + ps = new0(PathSpec, 1); + if (!ps) + return -ENOMEM; + + ps->path = strdup(s->pid_file); + if (!ps->path) { + free(ps); + return -ENOMEM; + } + + path_kill_slashes(ps->path); + + /* PATH_CHANGED would not be enough. There are daemons (sendmail) that + * keep their PID file open all the time. */ + ps->type = PATH_MODIFIED; + ps->inotify_fd = -1; + + s->pid_file_pathspec = ps; + + return service_watch_pid_file(s); +} + +static void service_fd_event(Unit *u, int fd, uint32_t events, Watch *w) { + Service *s = SERVICE(u); + + assert(s); + assert(fd >= 0); + assert(s->state == SERVICE_START || s->state == SERVICE_START_POST); + assert(s->pid_file_pathspec); + assert(path_spec_owns_inotify_fd(s->pid_file_pathspec, fd)); + + log_debug_unit(u->id, "inotify event for %s", u->id); + + if (path_spec_fd_event(s->pid_file_pathspec, events) < 0) + goto fail; + + if (service_retry_pid_file(s) == 0) + return; + + if (service_watch_pid_file(s) < 0) + goto fail; + + return; +fail: + service_unwatch_pid_file(s); + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_RESOURCES); +} + +static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { + Service *s = SERVICE(u); + ServiceResult f; + + assert(s); + assert(pid >= 0); + + if (UNIT(s)->fragment_path ? is_clean_exit(code, status, &s->success_status) : + is_clean_exit_lsb(code, status, &s->success_status)) + f = SERVICE_SUCCESS; + else if (code == CLD_EXITED) + f = SERVICE_FAILURE_EXIT_CODE; + else if (code == CLD_KILLED) + f = SERVICE_FAILURE_SIGNAL; + else if (code == CLD_DUMPED) + f = SERVICE_FAILURE_CORE_DUMP; + else + assert_not_reached("Unknown code"); + + if (s->main_pid == pid) { + /* Forking services may occasionally move to a new PID. + * As long as they update the PID file before exiting the old + * PID, they're fine. */ + if (service_load_pid_file(s, false) == 0) + return; + + s->main_pid = 0; + exec_status_exit(&s->main_exec_status, &s->exec_context, pid, code, status); + + if (s->main_command) { + /* If this is not a forking service than the + * main process got started and hence we copy + * the exit status so that it is recorded both + * as main and as control process exit + * status */ + + s->main_command->exec_status = s->main_exec_status; + + if (s->main_command->ignore) + f = SERVICE_SUCCESS; + } else if (s->exec_command[SERVICE_EXEC_START]) { + + /* If this is a forked process, then we should + * ignore the return value if this was + * configured for the starter process */ + + if (s->exec_command[SERVICE_EXEC_START]->ignore) + f = SERVICE_SUCCESS; + } + + log_struct_unit(f == SERVICE_SUCCESS ? LOG_DEBUG : LOG_NOTICE, + u->id, + "MESSAGE=%s: main process exited, code=%s, status=%i/%s", + u->id, sigchld_code_to_string(code), status, + strna(code == CLD_EXITED + ? exit_status_to_string(status, EXIT_STATUS_FULL) + : signal_to_string(status)), + "EXIT_CODE=%s", sigchld_code_to_string(code), + "EXIT_STATUS=%i", status, + NULL); + + if (f != SERVICE_SUCCESS) + s->result = f; + + if (s->main_command && + s->main_command->command_next && + f == SERVICE_SUCCESS) { + + /* There is another command to * + * execute, so let's do that. */ + + log_debug_unit(u->id, + "%s running next main command for state %s", + u->id, service_state_to_string(s->state)); + service_run_next_main(s); + + } else { + + /* The service exited, so the service is officially + * gone. */ + s->main_command = NULL; + + switch (s->state) { + + case SERVICE_START_POST: + case SERVICE_RELOAD: + case SERVICE_STOP: + /* Need to wait until the operation is + * done */ + break; + + case SERVICE_START: + if (s->type == SERVICE_ONESHOT) { + /* This was our main goal, so let's go on */ + if (f == SERVICE_SUCCESS) + service_enter_start_post(s); + else + service_enter_signal(s, SERVICE_FINAL_SIGTERM, f); + break; + } + + /* Fall through */ + + case SERVICE_RUNNING: + service_enter_running(s, f); + break; + + case SERVICE_STOP_SIGTERM: + case SERVICE_STOP_SIGKILL: + + if (!control_pid_good(s)) + service_enter_stop_post(s, f); + + /* If there is still a control process, wait for that first */ + break; + + default: + assert_not_reached("Uh, main process died at wrong time."); + } + } + + } else if (s->control_pid == pid) { + + s->control_pid = 0; + + if (s->control_command) { + exec_status_exit(&s->control_command->exec_status, + &s->exec_context, pid, code, status); + + if (s->control_command->ignore) + f = SERVICE_SUCCESS; + } + + log_full_unit(f == SERVICE_SUCCESS ? LOG_DEBUG : LOG_NOTICE, u->id, + "%s: control process exited, code=%s status=%i", + u->id, sigchld_code_to_string(code), status); + + if (f != SERVICE_SUCCESS) + s->result = f; + + /* Immediately get rid of the cgroup, so that the + * kernel doesn't delay the cgroup empty messages for + * the service cgroup any longer than necessary */ + cgroup_bonding_kill_list(UNIT(s)->cgroup_bondings, SIGKILL, + true, true, NULL, "control"); + + if (s->control_command && + s->control_command->command_next && + f == SERVICE_SUCCESS) { + + /* There is another command to * + * execute, so let's do that. */ + + log_debug_unit(u->id, + "%s running next control command for state %s", + u->id, service_state_to_string(s->state)); + service_run_next_control(s); + + } else { + /* No further commands for this step, so let's + * figure out what to do next */ + + s->control_command = NULL; + s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; + + log_debug_unit(u->id, + "%s got final SIGCHLD for state %s", + u->id, service_state_to_string(s->state)); + + switch (s->state) { + + case SERVICE_START_PRE: + if (f == SERVICE_SUCCESS) + service_enter_start(s); + else + service_enter_signal(s, SERVICE_FINAL_SIGTERM, f); + break; + + case SERVICE_START: + if (s->type != SERVICE_FORKING) + /* Maybe spurious event due to a reload that changed the type? */ + break; + + if (f != SERVICE_SUCCESS) { + service_enter_signal(s, SERVICE_FINAL_SIGTERM, f); + break; + } + + if (s->pid_file) { + bool has_start_post; + int r; + + /* Let's try to load the pid file here if we can. + * The PID file might actually be created by a START_POST + * script. In that case don't worry if the loading fails. */ + + has_start_post = !!s->exec_command[SERVICE_EXEC_START_POST]; + r = service_load_pid_file(s, !has_start_post); + if (!has_start_post && r < 0) { + r = service_demand_pid_file(s); + if (r < 0 || !cgroup_good(s)) + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); + break; + } + } else + service_search_main_pid(s); + + service_enter_start_post(s); + break; + + case SERVICE_START_POST: + if (f != SERVICE_SUCCESS) { + service_enter_stop(s, f); + break; + } + + if (s->pid_file) { + int r; + + r = service_load_pid_file(s, true); + if (r < 0) { + r = service_demand_pid_file(s); + if (r < 0 || !cgroup_good(s)) + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); + break; + } + } else + service_search_main_pid(s); + + service_enter_running(s, SERVICE_SUCCESS); + break; + + case SERVICE_RELOAD: + if (f == SERVICE_SUCCESS) { + service_load_pid_file(s, true); + service_search_main_pid(s); + } + + s->reload_result = f; + service_enter_running(s, SERVICE_SUCCESS); + break; + + case SERVICE_STOP: + service_enter_signal(s, SERVICE_STOP_SIGTERM, f); + break; + + case SERVICE_STOP_SIGTERM: + case SERVICE_STOP_SIGKILL: + if (main_pid_good(s) <= 0) + service_enter_stop_post(s, f); + + /* If there is still a service + * process around, wait until + * that one quit, too */ + break; + + case SERVICE_STOP_POST: + case SERVICE_FINAL_SIGTERM: + case SERVICE_FINAL_SIGKILL: + service_enter_dead(s, f, true); + break; + + default: + assert_not_reached("Uh, control process died at wrong time."); + } + } + } + + /* Notify clients about changed exit status */ + unit_add_to_dbus_queue(u); +} + +static void service_timer_event(Unit *u, uint64_t elapsed, Watch* w) { + Service *s = SERVICE(u); + + assert(s); + assert(elapsed == 1); + + if (w == &s->watchdog_watch) { + service_handle_watchdog(s); + return; + } + + assert(w == &s->timer_watch); + + switch (s->state) { + + case SERVICE_START_PRE: + case SERVICE_START: + log_warning_unit(u->id, + "%s operation timed out. Terminating.", u->id); + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_START_POST: + log_warning_unit(u->id, + "%s operation timed out. Stopping.", u->id); + service_enter_stop(s, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_RELOAD: + log_warning_unit(u->id, + "%s operation timed out. Stopping.", u->id); + s->reload_result = SERVICE_FAILURE_TIMEOUT; + service_enter_running(s, SERVICE_SUCCESS); + break; + + case SERVICE_STOP: + log_warning_unit(u->id, + "%s stopping timed out. Terminating.", u->id); + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_STOP_SIGTERM: + if (s->kill_context.send_sigkill) { + log_warning_unit(u->id, + "%s stopping timed out. Killing.", u->id); + service_enter_signal(s, SERVICE_STOP_SIGKILL, SERVICE_FAILURE_TIMEOUT); + } else { + log_warning_unit(u->id, + "%s stopping timed out. Skipping SIGKILL.", u->id); + service_enter_stop_post(s, SERVICE_FAILURE_TIMEOUT); + } + + break; + + case SERVICE_STOP_SIGKILL: + /* Uh, we sent a SIGKILL and it is still not gone? + * Must be something we cannot kill, so let's just be + * weirded out and continue */ + + log_warning_unit(u->id, + "%s still around after SIGKILL. Ignoring.", u->id); + service_enter_stop_post(s, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_STOP_POST: + log_warning_unit(u->id, + "%s stopping timed out (2). Terminating.", u->id); + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_FINAL_SIGTERM: + if (s->kill_context.send_sigkill) { + log_warning_unit(u->id, + "%s stopping timed out (2). Killing.", u->id); + service_enter_signal(s, SERVICE_FINAL_SIGKILL, SERVICE_FAILURE_TIMEOUT); + } else { + log_warning_unit(u->id, + "%s stopping timed out (2). Skipping SIGKILL. Entering failed mode.", + u->id); + service_enter_dead(s, SERVICE_FAILURE_TIMEOUT, false); + } + + break; + + case SERVICE_FINAL_SIGKILL: + log_warning_unit(u->id, + "%s still around after SIGKILL (2). Entering failed mode.", u->id); + service_enter_dead(s, SERVICE_FAILURE_TIMEOUT, true); + break; + + case SERVICE_AUTO_RESTART: + log_info_unit(u->id, + "%s holdoff time over, scheduling restart.", u->id); + service_enter_restart(s); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } +} + +static void service_cgroup_notify_event(Unit *u) { + Service *s = SERVICE(u); + + assert(u); + + log_debug_unit(u->id, + "%s: cgroup is empty", u->id); + + switch (s->state) { + + /* Waiting for SIGCHLD is usually more interesting, + * because it includes return codes/signals. Which is + * why we ignore the cgroup events for most cases, + * except when we don't know pid which to expect the + * SIGCHLD for. */ + + case SERVICE_START: + case SERVICE_START_POST: + /* If we were hoping for the daemon to write its PID file, + * we can give up now. */ + if (s->pid_file_pathspec) { + log_warning_unit(u->id, + "%s never wrote its PID file. Failing.", UNIT(s)->id); + service_unwatch_pid_file(s); + if (s->state == SERVICE_START) + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); + else + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); + } + break; + + case SERVICE_RUNNING: + /* service_enter_running() will figure out what to do */ + service_enter_running(s, SERVICE_SUCCESS); + break; + + case SERVICE_STOP_SIGTERM: + case SERVICE_STOP_SIGKILL: + + if (main_pid_good(s) <= 0 && !control_pid_good(s)) + service_enter_stop_post(s, SERVICE_SUCCESS); + + break; + + case SERVICE_FINAL_SIGTERM: + case SERVICE_FINAL_SIGKILL: + if (main_pid_good(s) <= 0 && !control_pid_good(s)) + service_enter_dead(s, SERVICE_SUCCESS, true); + + break; + + default: + ; + } +} + +static void service_notify_message(Unit *u, pid_t pid, char **tags) { + Service *s = SERVICE(u); + const char *e; + + assert(u); + + if (s->notify_access == NOTIFY_NONE) { + log_warning_unit(u->id, + "%s: Got notification message from PID %lu, but reception is disabled.", + u->id, (unsigned long) pid); + return; + } + + if (s->notify_access == NOTIFY_MAIN && pid != s->main_pid) { + log_warning_unit(u->id, + "%s: Got notification message from PID %lu, but reception only permitted for PID %lu", + u->id, (unsigned long) pid, (unsigned long) s->main_pid); + return; + } + + log_debug_unit(u->id, + "%s: Got message", u->id); + + /* Interpret MAINPID= */ + if ((e = strv_find_prefix(tags, "MAINPID=")) && + (s->state == SERVICE_START || + s->state == SERVICE_START_POST || + s->state == SERVICE_RUNNING || + s->state == SERVICE_RELOAD)) { + + if (parse_pid(e + 8, &pid) < 0) + log_warning_unit(u->id, + "Failed to parse notification message %s", e); + else { + log_debug_unit(u->id, + "%s: got %s", u->id, e); + service_set_main_pid(s, pid); + } + } + + /* Interpret READY= */ + if (s->type == SERVICE_NOTIFY && + s->state == SERVICE_START && + strv_find(tags, "READY=1")) { + log_debug_unit(u->id, + "%s: got READY=1", u->id); + + service_enter_start_post(s); + } + + /* Interpret STATUS= */ + e = strv_find_prefix(tags, "STATUS="); + if (e) { + char *t; + + if (e[7]) { + + if (!utf8_is_valid(e+7)) { + log_warning_unit(u->id, + "Status message in notification is not UTF-8 clean."); + return; + } + + t = strdup(e+7); + if (!t) { + log_error_unit(u->id, + "Failed to allocate string."); + return; + } + + log_debug_unit(u->id, + "%s: got %s", u->id, e); + + free(s->status_text); + s->status_text = t; + } else { + free(s->status_text); + s->status_text = NULL; + } + + } + if (strv_find(tags, "WATCHDOG=1")) { + log_debug_unit(u->id, + "%s: got WATCHDOG=1", u->id); + if (dual_timestamp_is_set(&s->watchdog_timestamp)) + service_reset_watchdog(s); + } + + /* Notify clients about changed status or main pid */ + unit_add_to_dbus_queue(u); +} + +#ifdef HAVE_SYSV_COMPAT + +static int service_enumerate(Manager *m) { + char **p; + unsigned i; + _cleanup_closedir_ DIR *d = NULL; + _cleanup_free_ char *path = NULL, *fpath = NULL, *name = NULL; + Set *runlevel_services[ELEMENTSOF(rcnd_table)] = {}; + _cleanup_set_free_ Set *shutdown_services = NULL; + Unit *service; + Iterator j; + int r; + + assert(m); + + if (m->running_as != SYSTEMD_SYSTEM) + return 0; + + STRV_FOREACH(p, m->lookup_paths.sysvrcnd_path) + for (i = 0; i < ELEMENTSOF(rcnd_table); i ++) { + struct dirent *de; + + free(path); + path = strjoin(*p, "/", rcnd_table[i].path, NULL); + if (!path) { + r = -ENOMEM; + goto finish; + } + + if (d) + closedir(d); + + d = opendir(path); + if (!d) { + if (errno != ENOENT) + log_warning("opendir(%s) failed: %s", path, strerror(errno)); + + continue; + } + + while ((de = readdir(d))) { + int a, b; + + if (ignore_file(de->d_name)) + continue; + + if (de->d_name[0] != 'S' && de->d_name[0] != 'K') + continue; + + if (strlen(de->d_name) < 4) + continue; + + a = undecchar(de->d_name[1]); + b = undecchar(de->d_name[2]); + + if (a < 0 || b < 0) + continue; + + free(fpath); + fpath = strjoin(path, "/", de->d_name, NULL); + if (!fpath) { + r = -ENOMEM; + goto finish; + } + + if (access(fpath, X_OK) < 0) { + + if (errno != ENOENT) + log_warning("access() failed on %s: %s", fpath, strerror(errno)); + + continue; + } + + free(name); + name = sysv_translate_name(de->d_name + 3); + if (!name) { + r = log_oom(); + goto finish; + } + + r = manager_load_unit_prepare(m, name, NULL, NULL, &service); + if (r < 0) { + log_warning("Failed to prepare unit %s: %s", name, strerror(-r)); + continue; + } + + if (de->d_name[0] == 'S') { + + if (rcnd_table[i].type == RUNLEVEL_UP) { + SERVICE(service)->sysv_start_priority_from_rcnd = + MAX(a*10 + b, SERVICE(service)->sysv_start_priority_from_rcnd); + + SERVICE(service)->sysv_enabled = true; + } + + r = set_ensure_allocated(&runlevel_services[i], + trivial_hash_func, trivial_compare_func); + if (r < 0) + goto finish; + + r = set_put(runlevel_services[i], service); + if (r < 0) + goto finish; + + } else if (de->d_name[0] == 'K' && + (rcnd_table[i].type == RUNLEVEL_DOWN)) { + + r = set_ensure_allocated(&shutdown_services, + trivial_hash_func, trivial_compare_func); + if (r < 0) + goto finish; + + r = set_put(shutdown_services, service); + if (r < 0) + goto finish; + } + } + } + + /* Now we loaded all stubs and are aware of the lowest + start-up priority for all services, not let's actually load + the services, this will also tell us which services are + actually native now */ + manager_dispatch_load_queue(m); + + /* If this is a native service, rely on native ways to pull in + * a service, don't pull it in via sysv rcN.d links. */ + for (i = 0; i < ELEMENTSOF(rcnd_table); i ++) + SET_FOREACH(service, runlevel_services[i], j) { + service = unit_follow_merge(service); + + if (service->fragment_path) + continue; + + r = unit_add_two_dependencies_by_name_inverse( + service, UNIT_AFTER, UNIT_WANTS, + rcnd_table[i].target, NULL, true); + if (r < 0) + goto finish; + } + + /* We honour K links only for halt/reboot. For the normal + * runlevels we assume the stop jobs will be implicitly added + * by the core logic. Also, we don't really distinguish here + * between the runlevels 0 and 6 and just add them to the + * special shutdown target. */ + SET_FOREACH(service, shutdown_services, j) { + service = unit_follow_merge(service); + + if (service->fragment_path) + continue; + + r = unit_add_two_dependencies_by_name( + service, UNIT_BEFORE, UNIT_CONFLICTS, + SPECIAL_SHUTDOWN_TARGET, NULL, true); + if (r < 0) + goto finish; + } + + r = 0; + +finish: + + for (i = 0; i < ELEMENTSOF(rcnd_table); i++) + set_free(runlevel_services[i]); + + return r; +} +#endif + +static void service_bus_name_owner_change( + Unit *u, + const char *name, + const char *old_owner, + const char *new_owner) { + + Service *s = SERVICE(u); + + assert(s); + assert(name); + + assert(streq(s->bus_name, name)); + assert(old_owner || new_owner); + + if (old_owner && new_owner) + log_debug_unit(u->id, + "%s's D-Bus name %s changed owner from %s to %s", + u->id, name, old_owner, new_owner); + else if (old_owner) + log_debug_unit(u->id, + "%s's D-Bus name %s no longer registered by %s", + u->id, name, old_owner); + else + log_debug_unit(u->id, + "%s's D-Bus name %s now registered by %s", + u->id, name, new_owner); + + s->bus_name_good = !!new_owner; + + if (s->type == SERVICE_DBUS) { + + /* service_enter_running() will figure out what to + * do */ + if (s->state == SERVICE_RUNNING) + service_enter_running(s, SERVICE_SUCCESS); + else if (s->state == SERVICE_START && new_owner) + service_enter_start_post(s); + + } else if (new_owner && + s->main_pid <= 0 && + (s->state == SERVICE_START || + s->state == SERVICE_START_POST || + s->state == SERVICE_RUNNING || + s->state == SERVICE_RELOAD)) { + + /* Try to acquire PID from bus service */ + log_debug_unit(u->id, + "Trying to acquire PID from D-Bus name..."); + + bus_query_pid(u->manager, name); + } +} + +static void service_bus_query_pid_done( + Unit *u, + const char *name, + pid_t pid) { + + Service *s = SERVICE(u); + + assert(s); + assert(name); + + log_debug_unit(u->id, + "%s's D-Bus name %s is now owned by process %u", + u->id, name, (unsigned) pid); + + if (s->main_pid <= 0 && + (s->state == SERVICE_START || + s->state == SERVICE_START_POST || + s->state == SERVICE_RUNNING || + s->state == SERVICE_RELOAD)) + service_set_main_pid(s, pid); +} + +int service_set_socket_fd(Service *s, int fd, Socket *sock) { + + assert(s); + assert(fd >= 0); + + /* This is called by the socket code when instantiating a new + * service for a stream socket and the socket needs to be + * configured. */ + + if (UNIT(s)->load_state != UNIT_LOADED) + return -EINVAL; + + if (s->socket_fd >= 0) + return -EBUSY; + + if (s->state != SERVICE_DEAD) + return -EAGAIN; + + s->socket_fd = fd; + s->got_socket_fd = true; + + unit_ref_set(&s->accept_socket, UNIT(sock)); + + return unit_add_two_dependencies(UNIT(sock), UNIT_BEFORE, UNIT_TRIGGERS, UNIT(s), false); +} + +static void service_reset_failed(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + if (s->state == SERVICE_FAILED) + service_set_state(s, SERVICE_DEAD); + + s->result = SERVICE_SUCCESS; + s->reload_result = SERVICE_SUCCESS; + + RATELIMIT_RESET(s->start_limit); +} + +static int service_kill(Unit *u, KillWho who, int signo, DBusError *error) { + Service *s = SERVICE(u); + return unit_kill_common(u, who, signo, s->main_pid, s->control_pid, error); +} + +static const char* const service_state_table[_SERVICE_STATE_MAX] = { + [SERVICE_DEAD] = "dead", + [SERVICE_START_PRE] = "start-pre", + [SERVICE_START] = "start", + [SERVICE_START_POST] = "start-post", + [SERVICE_RUNNING] = "running", + [SERVICE_EXITED] = "exited", + [SERVICE_RELOAD] = "reload", + [SERVICE_STOP] = "stop", + [SERVICE_STOP_SIGTERM] = "stop-sigterm", + [SERVICE_STOP_SIGKILL] = "stop-sigkill", + [SERVICE_STOP_POST] = "stop-post", + [SERVICE_FINAL_SIGTERM] = "final-sigterm", + [SERVICE_FINAL_SIGKILL] = "final-sigkill", + [SERVICE_FAILED] = "failed", + [SERVICE_AUTO_RESTART] = "auto-restart", +}; + +DEFINE_STRING_TABLE_LOOKUP(service_state, ServiceState); + +static const char* const service_restart_table[_SERVICE_RESTART_MAX] = { + [SERVICE_RESTART_NO] = "no", + [SERVICE_RESTART_ON_SUCCESS] = "on-success", + [SERVICE_RESTART_ON_FAILURE] = "on-failure", + [SERVICE_RESTART_ON_ABORT] = "on-abort", + [SERVICE_RESTART_ALWAYS] = "always" +}; + +DEFINE_STRING_TABLE_LOOKUP(service_restart, ServiceRestart); + +static const char* const service_type_table[_SERVICE_TYPE_MAX] = { + [SERVICE_SIMPLE] = "simple", + [SERVICE_FORKING] = "forking", + [SERVICE_ONESHOT] = "oneshot", + [SERVICE_DBUS] = "dbus", + [SERVICE_NOTIFY] = "notify", + [SERVICE_IDLE] = "idle" +}; + +DEFINE_STRING_TABLE_LOOKUP(service_type, ServiceType); + +static const char* const service_exec_command_table[_SERVICE_EXEC_COMMAND_MAX] = { + [SERVICE_EXEC_START_PRE] = "ExecStartPre", + [SERVICE_EXEC_START] = "ExecStart", + [SERVICE_EXEC_START_POST] = "ExecStartPost", + [SERVICE_EXEC_RELOAD] = "ExecReload", + [SERVICE_EXEC_STOP] = "ExecStop", + [SERVICE_EXEC_STOP_POST] = "ExecStopPost", +}; + +DEFINE_STRING_TABLE_LOOKUP(service_exec_command, ServiceExecCommand); + +static const char* const notify_access_table[_NOTIFY_ACCESS_MAX] = { + [NOTIFY_NONE] = "none", + [NOTIFY_MAIN] = "main", + [NOTIFY_ALL] = "all" +}; + +DEFINE_STRING_TABLE_LOOKUP(notify_access, NotifyAccess); + +static const char* const service_result_table[_SERVICE_RESULT_MAX] = { + [SERVICE_SUCCESS] = "success", + [SERVICE_FAILURE_RESOURCES] = "resources", + [SERVICE_FAILURE_TIMEOUT] = "timeout", + [SERVICE_FAILURE_EXIT_CODE] = "exit-code", + [SERVICE_FAILURE_SIGNAL] = "signal", + [SERVICE_FAILURE_CORE_DUMP] = "core-dump", + [SERVICE_FAILURE_WATCHDOG] = "watchdog", + [SERVICE_FAILURE_START_LIMIT] = "start-limit" +}; + +DEFINE_STRING_TABLE_LOOKUP(service_result, ServiceResult); + +static const char* const start_limit_action_table[_SERVICE_START_LIMIT_MAX] = { + [SERVICE_START_LIMIT_NONE] = "none", + [SERVICE_START_LIMIT_REBOOT] = "reboot", + [SERVICE_START_LIMIT_REBOOT_FORCE] = "reboot-force", + [SERVICE_START_LIMIT_REBOOT_IMMEDIATE] = "reboot-immediate" +}; +DEFINE_STRING_TABLE_LOOKUP(start_limit_action, StartLimitAction); + +const UnitVTable service_vtable = { + .object_size = sizeof(Service), + + .sections = + "Unit\0" + "Service\0" + "Install\0", + + .exec_context_offset = offsetof(Service, exec_context), + .exec_section = "Service", + + .init = service_init, + .done = service_done, + .load = service_load, + + .coldplug = service_coldplug, + + .dump = service_dump, + + .start = service_start, + .stop = service_stop, + .reload = service_reload, + + .can_reload = service_can_reload, + + .kill = service_kill, + + .serialize = service_serialize, + .deserialize_item = service_deserialize_item, + + .active_state = service_active_state, + .sub_state_to_string = service_sub_state_to_string, + + .check_gc = service_check_gc, + .check_snapshot = service_check_snapshot, + + .sigchld_event = service_sigchld_event, + .timer_event = service_timer_event, + .fd_event = service_fd_event, + + .reset_failed = service_reset_failed, + + .cgroup_notify_empty = service_cgroup_notify_event, + .notify_message = service_notify_message, + + .bus_name_owner_change = service_bus_name_owner_change, + .bus_query_pid_done = service_bus_query_pid_done, + + .bus_interface = "org.freedesktop.systemd1.Service", + .bus_message_handler = bus_service_message_handler, + .bus_invalidating_properties = bus_service_invalidating_properties, + +#ifdef HAVE_SYSV_COMPAT + .enumerate = service_enumerate, +#endif + .status_message_formats = { + .starting_stopping = { + [0] = "Starting %s...", + [1] = "Stopping %s...", + }, + .finished_start_job = { + [JOB_DONE] = "Started %s.", + [JOB_FAILED] = "Failed to start %s.", + [JOB_DEPENDENCY] = "Dependency failed for %s.", + [JOB_TIMEOUT] = "Timed out starting %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Stopped %s.", + [JOB_FAILED] = "Stopped (with error) %s.", + [JOB_TIMEOUT] = "Timed out stopping %s.", + }, + }, +}; diff --git a/src/core/service.h b/src/core/service.h new file mode 100644 index 0000000000..703d3faa45 --- /dev/null +++ b/src/core/service.h @@ -0,0 +1,225 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Service Service; + +#include "unit.h" +#include "path.h" +#include "ratelimit.h" +#include "service.h" +#include "kill.h" +#include "exit-status.h" + +typedef enum ServiceState { + SERVICE_DEAD, + SERVICE_START_PRE, + SERVICE_START, + SERVICE_START_POST, + SERVICE_RUNNING, + SERVICE_EXITED, /* Nothing is running anymore, but RemainAfterExit is true hence this is OK */ + SERVICE_RELOAD, + SERVICE_STOP, /* No STOP_PRE state, instead just register multiple STOP executables */ + SERVICE_STOP_SIGTERM, + SERVICE_STOP_SIGKILL, + SERVICE_STOP_POST, + SERVICE_FINAL_SIGTERM, /* In case the STOP_POST executable hangs, we shoot that down, too */ + SERVICE_FINAL_SIGKILL, + SERVICE_FAILED, + SERVICE_AUTO_RESTART, + _SERVICE_STATE_MAX, + _SERVICE_STATE_INVALID = -1 +} ServiceState; + +typedef enum ServiceRestart { + SERVICE_RESTART_NO, + SERVICE_RESTART_ON_SUCCESS, + SERVICE_RESTART_ON_FAILURE, + SERVICE_RESTART_ON_ABORT, + SERVICE_RESTART_ALWAYS, + _SERVICE_RESTART_MAX, + _SERVICE_RESTART_INVALID = -1 +} ServiceRestart; + +typedef enum ServiceType { + SERVICE_SIMPLE, /* we fork and go on right-away (i.e. modern socket activated daemons) */ + SERVICE_FORKING, /* forks by itself (i.e. traditional daemons) */ + SERVICE_ONESHOT, /* we fork and wait until the program finishes (i.e. programs like fsck which run and need to finish before we continue) */ + SERVICE_DBUS, /* we fork and wait until a specific D-Bus name appears on the bus */ + SERVICE_NOTIFY, /* we fork and wait until a daemon sends us a ready message with sd_notify() */ + SERVICE_IDLE, /* much like simple, but delay exec() until all jobs are dispatched. */ + _SERVICE_TYPE_MAX, + _SERVICE_TYPE_INVALID = -1 +} ServiceType; + +typedef enum ServiceExecCommand { + SERVICE_EXEC_START_PRE, + SERVICE_EXEC_START, + SERVICE_EXEC_START_POST, + SERVICE_EXEC_RELOAD, + SERVICE_EXEC_STOP, + SERVICE_EXEC_STOP_POST, + _SERVICE_EXEC_COMMAND_MAX, + _SERVICE_EXEC_COMMAND_INVALID = -1 +} ServiceExecCommand; + +typedef enum NotifyAccess { + NOTIFY_NONE, + NOTIFY_ALL, + NOTIFY_MAIN, + _NOTIFY_ACCESS_MAX, + _NOTIFY_ACCESS_INVALID = -1 +} NotifyAccess; + +typedef enum ServiceResult { + SERVICE_SUCCESS, + SERVICE_FAILURE_RESOURCES, + SERVICE_FAILURE_TIMEOUT, + SERVICE_FAILURE_EXIT_CODE, + SERVICE_FAILURE_SIGNAL, + SERVICE_FAILURE_CORE_DUMP, + SERVICE_FAILURE_WATCHDOG, + SERVICE_FAILURE_START_LIMIT, + _SERVICE_RESULT_MAX, + _SERVICE_RESULT_INVALID = -1 +} ServiceResult; + +typedef enum StartLimitAction { + SERVICE_START_LIMIT_NONE, + SERVICE_START_LIMIT_REBOOT, + SERVICE_START_LIMIT_REBOOT_FORCE, + SERVICE_START_LIMIT_REBOOT_IMMEDIATE, + _SERVICE_START_LIMIT_MAX, + _SERVICE_START_LIMIT_INVALID = -1 +} StartLimitAction; + +struct Service { + Unit meta; + + ServiceType type; + ServiceRestart restart; + ExitStatusSet restart_ignore_status; + ExitStatusSet success_status; + + /* If set we'll read the main daemon PID from this file */ + char *pid_file; + + usec_t restart_usec; + usec_t timeout_start_usec; + usec_t timeout_stop_usec; + + dual_timestamp watchdog_timestamp; + usec_t watchdog_usec; + Watch watchdog_watch; + + ExecCommand* exec_command[_SERVICE_EXEC_COMMAND_MAX]; + + ExecContext exec_context; + KillContext kill_context; + + ServiceState state, deserialized_state; + + /* The exit status of the real main process */ + ExecStatus main_exec_status; + + /* The currently executed control process */ + ExecCommand *control_command; + + /* The currently executed main process, which may be NULL if + * the main process got started via forking mode and not by + * us */ + ExecCommand *main_command; + + /* The ID of the control command currently being executed */ + ServiceExecCommand control_command_id; + + pid_t main_pid, control_pid; + int socket_fd; + + int fsck_passno; + + bool permissions_start_only; + bool root_directory_start_only; + bool remain_after_exit; + bool guess_main_pid; + + /* If we shut down, remember why */ + ServiceResult result; + ServiceResult reload_result; + + bool main_pid_known:1; + bool main_pid_alien:1; + bool bus_name_good:1; + bool forbid_restart:1; + bool got_socket_fd:1; + bool start_timeout_defined:1; +#ifdef HAVE_SYSV_COMPAT + bool is_sysv:1; + bool sysv_has_lsb:1; + bool sysv_enabled:1; + int sysv_start_priority_from_rcnd; + int sysv_start_priority; + + char *sysv_runlevels; +#endif + + char *bus_name; + + char *status_text; + + RateLimit start_limit; + StartLimitAction start_limit_action; + + UnitRef accept_socket; + + Watch timer_watch; + PathSpec *pid_file_pathspec; + + NotifyAccess notify_access; +}; + +extern const UnitVTable service_vtable; + +struct Socket; + +int service_set_socket_fd(Service *s, int fd, struct Socket *socket); + +const char* service_state_to_string(ServiceState i) _const_; +ServiceState service_state_from_string(const char *s) _pure_; + +const char* service_restart_to_string(ServiceRestart i) _const_; +ServiceRestart service_restart_from_string(const char *s) _pure_; + +const char* service_type_to_string(ServiceType i) _const_; +ServiceType service_type_from_string(const char *s) _pure_; + +const char* service_exec_command_to_string(ServiceExecCommand i) _const_; +ServiceExecCommand service_exec_command_from_string(const char *s) _pure_; + +const char* notify_access_to_string(NotifyAccess i) _const_; +NotifyAccess notify_access_from_string(const char *s) _pure_; + +const char* service_result_to_string(ServiceResult i) _const_; +ServiceResult service_result_from_string(const char *s) _pure_; + +const char* start_limit_action_to_string(StartLimitAction i) _const_; +StartLimitAction start_limit_action_from_string(const char *s) _pure_; diff --git a/src/core/shutdown.c b/src/core/shutdown.c new file mode 100644 index 0000000000..2db761de36 --- /dev/null +++ b/src/core/shutdown.c @@ -0,0 +1,332 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 ProFUSION embedded systems + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/mman.h> +#include <sys/types.h> +#include <sys/reboot.h> +#include <linux/reboot.h> +#include <sys/wait.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <sys/mount.h> +#include <sys/syscall.h> +#include <fcntl.h> +#include <dirent.h> +#include <errno.h> +#include <unistd.h> +#include <signal.h> +#include <stdbool.h> +#include <stdlib.h> +#include <string.h> + +#include "missing.h" +#include "log.h" +#include "umount.h" +#include "util.h" +#include "mkdir.h" +#include "virt.h" +#include "watchdog.h" +#include "killall.h" + +#define FINALIZE_ATTEMPTS 50 + +static int prepare_new_root(void) { + static const char dirs[] = + "/run/initramfs/oldroot\0" + "/run/initramfs/proc\0" + "/run/initramfs/sys\0" + "/run/initramfs/dev\0" + "/run/initramfs/run\0"; + + const char *dir; + + if (mount("/run/initramfs", "/run/initramfs", NULL, MS_BIND, NULL) < 0) { + log_error("Failed to mount bind /run/initramfs on /run/initramfs: %m"); + return -errno; + } + + if (mount(NULL, "/run/initramfs", NULL, MS_PRIVATE, NULL) < 0) { + log_error("Failed to make /run/initramfs private mount: %m"); + return -errno; + } + + NULSTR_FOREACH(dir, dirs) + if (mkdir_p_label(dir, 0755) < 0 && errno != EEXIST) { + log_error("Failed to mkdir %s: %m", dir); + return -errno; + } + + if (mount("/sys", "/run/initramfs/sys", NULL, MS_BIND, NULL) < 0) { + log_error("Failed to mount bind /sys on /run/initramfs/sys: %m"); + return -errno; + } + + if (mount("/proc", "/run/initramfs/proc", NULL, MS_BIND, NULL) < 0) { + log_error("Failed to mount bind /proc on /run/initramfs/proc: %m"); + return -errno; + } + + if (mount("/dev", "/run/initramfs/dev", NULL, MS_BIND, NULL) < 0) { + log_error("Failed to mount bind /dev on /run/initramfs/dev: %m"); + return -errno; + } + + if (mount("/run", "/run/initramfs/run", NULL, MS_BIND, NULL) < 0) { + log_error("Failed to mount bind /run on /run/initramfs/run: %m"); + return -errno; + } + + return 0; +} + +static int pivot_to_new_root(void) { + + if (chdir("/run/initramfs") < 0) { + log_error("Failed to change directory to /run/initramfs: %m"); + return -errno; + } + + /* Work-around for a kernel bug: for some reason the kernel + * refuses switching root if any file systems are mounted + * MS_SHARED. Hence remount them MS_PRIVATE here as a + * work-around. + * + * https://bugzilla.redhat.com/show_bug.cgi?id=847418 */ + if (mount(NULL, "/", NULL, MS_REC|MS_PRIVATE, NULL) < 0) + log_warning("Failed to make \"/\" private mount: %m"); + + if (pivot_root(".", "oldroot") < 0) { + log_error("pivot failed: %m"); + /* only chroot if pivot root succeeded */ + return -errno; + } + + chroot("."); + + setsid(); + make_console_stdio(); + + log_info("Successfully changed into root pivot."); + + return 0; +} + +int main(int argc, char *argv[]) { + int cmd, r; + unsigned retries; + bool need_umount = true, need_swapoff = true, need_loop_detach = true, need_dm_detach = true; + bool in_container, use_watchdog = false; + char *arguments[3]; + + log_parse_environment(); + log_set_target(LOG_TARGET_CONSOLE); /* syslog will die if not gone yet */ + log_open(); + + umask(0022); + + if (getpid() != 1) { + log_error("Not executed by init (pid 1)."); + r = -EPERM; + goto error; + } + + if (argc != 2) { + log_error("Invalid number of arguments."); + r = -EINVAL; + goto error; + } + + in_container = detect_container(NULL) > 0; + + if (streq(argv[1], "reboot")) + cmd = RB_AUTOBOOT; + else if (streq(argv[1], "poweroff")) + cmd = RB_POWER_OFF; + else if (streq(argv[1], "halt")) + cmd = RB_HALT_SYSTEM; + else if (streq(argv[1], "kexec")) + cmd = LINUX_REBOOT_CMD_KEXEC; + else { + log_error("Unknown action '%s'.", argv[1]); + r = -EINVAL; + goto error; + } + + use_watchdog = !!getenv("WATCHDOG_USEC"); + + /* lock us into memory */ + mlockall(MCL_CURRENT|MCL_FUTURE); + + log_info("Sending SIGTERM to remaining processes..."); + broadcast_signal(SIGTERM, true); + + log_info("Sending SIGKILL to remaining processes..."); + broadcast_signal(SIGKILL, true); + + if (in_container) { + need_swapoff = false; + need_dm_detach = false; + need_loop_detach = false; + } + + /* Unmount all mountpoints, swaps, and loopback devices */ + for (retries = 0; retries < FINALIZE_ATTEMPTS; retries++) { + bool changed = false; + + if (use_watchdog) + watchdog_ping(); + + if (need_umount) { + log_info("Unmounting file systems."); + r = umount_all(&changed); + if (r == 0) { + need_umount = false; + log_info("All filesystems unmounted."); + } else if (r > 0) + log_info("Not all file systems unmounted, %d left.", r); + else + log_error("Failed to unmount file systems: %s", strerror(-r)); + } + + if (need_swapoff) { + log_info("Deactivating swaps."); + r = swapoff_all(&changed); + if (r == 0) { + need_swapoff = false; + log_info("All swaps deactivated."); + } else if (r > 0) + log_info("Not all swaps deactivated, %d left.", r); + else + log_error("Failed to deactivate swaps: %s", strerror(-r)); + } + + if (need_loop_detach) { + log_info("Detaching loop devices."); + r = loopback_detach_all(&changed); + if (r == 0) { + need_loop_detach = false; + log_info("All loop devices detached."); + } else if (r > 0) + log_info("Not all loop devices detached, %d left.", r); + else + log_error("Failed to detach loop devices: %s", strerror(-r)); + } + + if (need_dm_detach) { + log_info("Detaching DM devices."); + r = dm_detach_all(&changed); + if (r == 0) { + need_dm_detach = false; + log_info("All DM devices detached."); + } else if (r > 0) + log_info("Not all DM devices detached, %d left.", r); + else + log_error("Failed to detach DM devices: %s", strerror(-r)); + } + + if (!need_umount && !need_swapoff && !need_loop_detach && !need_dm_detach) { + if (retries > 0) + log_info("All filesystems, swaps, loop devices, DM devices detached."); + /* Yay, done */ + break; + } + + /* If in this iteration we didn't manage to + * unmount/deactivate anything, we simply give up */ + if (!changed) { + log_error("Cannot finalize remaining file systems and devices, giving up."); + break; + } + + log_debug("Couldn't finalize remaining file systems and devices after %u retries, trying again.", retries+1); + } + + if (retries >= FINALIZE_ATTEMPTS) + log_error("Too many iterations, giving up."); + else + log_info("Storage is finalized."); + + arguments[0] = NULL; + arguments[1] = argv[1]; + arguments[2] = NULL; + execute_directory(SYSTEM_SHUTDOWN_PATH, NULL, arguments); + + if (!in_container && !in_initrd() && + access("/run/initramfs/shutdown", X_OK) == 0) { + + if (prepare_new_root() >= 0 && + pivot_to_new_root() >= 0) { + + log_info("Returning to initrd..."); + + execv("/shutdown", argv); + log_error("Failed to execute shutdown binary: %m"); + } + } + + /* The kernel will automaticall flush ATA disks and suchlike + * on reboot(), but the file systems need to be synce'd + * explicitly in advance. So let's do this here, but not + * needlessly slow down containers. */ + if (!in_container) + sync(); + + if (cmd == LINUX_REBOOT_CMD_KEXEC) { + + if (!in_container) { + /* We cheat and exec kexec to avoid doing all its work */ + pid_t pid = fork(); + + if (pid < 0) + log_error("Could not fork: %m. Falling back to normal reboot."); + else if (pid > 0) { + wait_for_terminate_and_warn("kexec", pid); + log_warning("kexec failed. Falling back to normal reboot."); + } else { + /* Child */ + const char *args[3] = { "/sbin/kexec", "-e", NULL }; + execv(args[0], (char * const *) args); + return EXIT_FAILURE; + } + } + + cmd = RB_AUTOBOOT; + } + + reboot(cmd); + + if (errno == EPERM && in_container) { + /* If we are in a container, and we lacked + * CAP_SYS_BOOT just exit, this will kill our + * container for good. */ + log_error("Exiting container."); + exit(0); + } + + log_error("Failed to invoke reboot(): %m"); + r = -errno; + + error: + log_error("Critical error while doing system shutdown: %s", strerror(-r)); + + freeze(); + return EXIT_FAILURE; +} diff --git a/src/core/smack-setup.c b/src/core/smack-setup.c new file mode 100644 index 0000000000..73eeb04190 --- /dev/null +++ b/src/core/smack-setup.c @@ -0,0 +1,151 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright (C) 2013 Intel Corporation + Authors: + Nathaniel Chen <nathaniel.chen@intel.com> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published + by the Free Software Foundation; either version 2.1 of the License, + or (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdio.h> +#include <errno.h> +#include <string.h> +#include <unistd.h> +#include <stdlib.h> +#include <sys/vfs.h> +#include <fcntl.h> +#include <sys/types.h> +#include <dirent.h> +#include <sys/mount.h> +#include <stdint.h> + +#include "macro.h" +#include "smack-setup.h" +#include "util.h" +#include "log.h" +#include "label.h" + +#define SMACK_CONFIG "/etc/smack/accesses.d/" +#define CIPSO_CONFIG "/etc/smack/cipso/" + +static int write_rules(const char* dstpath, const char* srcdir) { + _cleanup_fclose_ FILE *dst = NULL; + _cleanup_closedir_ DIR *dir = NULL; + struct dirent *entry; + char buf[NAME_MAX]; + int dfd = -1; + int r = 0; + + dst = fopen(dstpath, "we"); + if (!dst) { + if (errno != ENOENT) + log_warning("Failed to open %s: %m", dstpath); + return -errno; /* negative error */ + } + + /* write rules to dst from every file in the directory */ + dir = opendir(srcdir); + if (!dir) { + if (errno != ENOENT) + log_warning("Failed to opendir %s: %m", srcdir); + return errno; /* positive on purpose */ + } + + dfd = dirfd(dir); + assert(dfd >= 0); + + FOREACH_DIRENT(entry, dir, return 0) { + int fd; + _cleanup_fclose_ FILE *policy = NULL; + + fd = openat(dfd, entry->d_name, O_RDONLY|O_CLOEXEC); + if (fd < 0) { + if (r == 0) + r = -errno; + log_warning("Failed to open %s: %m", entry->d_name); + continue; + } + + policy = fdopen(fd, "re"); + if (!policy) { + if (r == 0) + r = -errno; + close_nointr_nofail(fd); + log_error("Failed to open %s: %m", entry->d_name); + continue; + } + + /* load2 write rules in the kernel require a line buffered stream */ + FOREACH_LINE(buf, policy, + log_error("Failed to read line from %s: %m", + entry->d_name)) { + if (!fputs(buf, dst)) { + if (r == 0) + r = -EINVAL; + log_error("Failed to write line to %s", dstpath); + break; + } + if (fflush(dst)) { + if (r == 0) + r = -errno; + log_error("Failed to flush writes to %s: %m", dstpath); + break; + } + } + } + + return r; +} + + +int smack_setup(void) { + int r; + + r = write_rules("/sys/fs/smackfs/load2", SMACK_CONFIG); + switch(r) { + case -ENOENT: + log_debug("Smack is not enabled in the kernel."); + return 0; + case ENOENT: + log_debug("Smack access rules directory " SMACK_CONFIG " not found"); + return 0; + case 0: + log_info("Successfully loaded Smack policies."); + break; + default: + log_warning("Failed to load Smack access rules: %s, ignoring.", + strerror(abs(r))); + return 0; + } + + r = write_rules("/sys/fs/smackfs/cipso2", CIPSO_CONFIG); + switch(r) { + case -ENOENT: + log_debug("Smack/CIPSO is not enabled in the kernel."); + return 0; + case ENOENT: + log_debug("Smack/CIPSO access rules directory " CIPSO_CONFIG " not found"); + return 0; + case 0: + log_info("Successfully loaded Smack/CIPSO policies."); + return 0; + default: + log_warning("Failed to load Smack/CIPSO access rules: %s, ignoring.", + strerror(abs(r))); + return 0; + } +} diff --git a/src/core/smack-setup.h b/src/core/smack-setup.h new file mode 100644 index 0000000000..ffe91843c3 --- /dev/null +++ b/src/core/smack-setup.h @@ -0,0 +1,26 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright (C) 2013 Intel Corporation + Authors: + Nathaniel Chen <nathaniel.chen@intel.com> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published + by the Free Software Foundation; either version 2.1 of the License, + or (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int smack_setup(void); diff --git a/src/core/snapshot.c b/src/core/snapshot.c new file mode 100644 index 0000000000..a63eccd8de --- /dev/null +++ b/src/core/snapshot.c @@ -0,0 +1,309 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "unit.h" +#include "snapshot.h" +#include "unit-name.h" +#include "dbus-snapshot.h" +#include "bus-errors.h" + +static const UnitActiveState state_translation_table[_SNAPSHOT_STATE_MAX] = { + [SNAPSHOT_DEAD] = UNIT_INACTIVE, + [SNAPSHOT_ACTIVE] = UNIT_ACTIVE +}; + +static void snapshot_init(Unit *u) { + Snapshot *s = SNAPSHOT(u); + + assert(s); + assert(UNIT(s)->load_state == UNIT_STUB); + + UNIT(s)->ignore_on_isolate = true; + UNIT(s)->ignore_on_snapshot = true; +} + +static void snapshot_set_state(Snapshot *s, SnapshotState state) { + SnapshotState old_state; + assert(s); + + old_state = s->state; + s->state = state; + + if (state != old_state) + log_debug("%s changed %s -> %s", + UNIT(s)->id, + snapshot_state_to_string(old_state), + snapshot_state_to_string(state)); + + unit_notify(UNIT(s), state_translation_table[old_state], state_translation_table[state], true); +} + +static int snapshot_load(Unit *u) { + Snapshot *s = SNAPSHOT(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + /* Make sure that only snapshots created via snapshot_create() + * can be loaded */ + if (!s->by_snapshot_create && UNIT(s)->manager->n_reloading <= 0) + return -ENOENT; + + u->load_state = UNIT_LOADED; + return 0; +} + +static int snapshot_coldplug(Unit *u) { + Snapshot *s = SNAPSHOT(u); + + assert(s); + assert(s->state == SNAPSHOT_DEAD); + + if (s->deserialized_state != s->state) + snapshot_set_state(s, s->deserialized_state); + + return 0; +} + +static void snapshot_dump(Unit *u, FILE *f, const char *prefix) { + Snapshot *s = SNAPSHOT(u); + + assert(s); + assert(f); + + fprintf(f, + "%sSnapshot State: %s\n" + "%sClean Up: %s\n", + prefix, snapshot_state_to_string(s->state), + prefix, yes_no(s->cleanup)); +} + +static int snapshot_start(Unit *u) { + Snapshot *s = SNAPSHOT(u); + + assert(s); + assert(s->state == SNAPSHOT_DEAD); + + snapshot_set_state(s, SNAPSHOT_ACTIVE); + + if (s->cleanup) + unit_add_to_cleanup_queue(u); + + return 0; +} + +static int snapshot_stop(Unit *u) { + Snapshot *s = SNAPSHOT(u); + + assert(s); + assert(s->state == SNAPSHOT_ACTIVE); + + snapshot_set_state(s, SNAPSHOT_DEAD); + return 0; +} + +static int snapshot_serialize(Unit *u, FILE *f, FDSet *fds) { + Snapshot *s = SNAPSHOT(u); + Unit *other; + Iterator i; + + assert(s); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", snapshot_state_to_string(s->state)); + unit_serialize_item(u, f, "cleanup", yes_no(s->cleanup)); + SET_FOREACH(other, u->dependencies[UNIT_WANTS], i) + unit_serialize_item(u, f, "wants", other->id); + + return 0; +} + +static int snapshot_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Snapshot *s = SNAPSHOT(u); + int r; + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + SnapshotState state; + + if ((state = snapshot_state_from_string(value)) < 0) + log_debug("Failed to parse state value %s", value); + else + s->deserialized_state = state; + + } else if (streq(key, "cleanup")) { + + if ((r = parse_boolean(value)) < 0) + log_debug("Failed to parse cleanup value %s", value); + else + s->cleanup = r; + + } else if (streq(key, "wants")) { + + if ((r = unit_add_two_dependencies_by_name(u, UNIT_AFTER, UNIT_WANTS, value, NULL, true)) < 0) + return r; + } else + log_debug("Unknown serialization key '%s'", key); + + return 0; +} + +_pure_ static UnitActiveState snapshot_active_state(Unit *u) { + assert(u); + + return state_translation_table[SNAPSHOT(u)->state]; +} + +_pure_ static const char *snapshot_sub_state_to_string(Unit *u) { + assert(u); + + return snapshot_state_to_string(SNAPSHOT(u)->state); +} + +int snapshot_create(Manager *m, const char *name, bool cleanup, DBusError *e, Snapshot **_s) { + Iterator i; + Unit *other, *u = NULL; + char *n = NULL; + int r; + const char *k; + + assert(m); + assert(_s); + + if (name) { + if (!unit_name_is_valid(name, false)) { + dbus_set_error(e, BUS_ERROR_INVALID_NAME, "Unit name %s is not valid.", name); + return -EINVAL; + } + + if (unit_name_to_type(name) != UNIT_SNAPSHOT) { + dbus_set_error(e, BUS_ERROR_UNIT_TYPE_MISMATCH, "Unit name %s lacks snapshot suffix.", name); + return -EINVAL; + } + + if (manager_get_unit(m, name)) { + dbus_set_error(e, BUS_ERROR_UNIT_EXISTS, "Snapshot %s exists already.", name); + return -EEXIST; + } + + } else { + + for (;;) { + if (asprintf(&n, "snapshot-%u.snapshot", ++ m->n_snapshots) < 0) + return -ENOMEM; + + if (!manager_get_unit(m, n)) + break; + + free(n); + } + + name = n; + } + + r = manager_load_unit_prepare(m, name, NULL, e, &u); + free(n); + + if (r < 0) + goto fail; + + SNAPSHOT(u)->by_snapshot_create = true; + manager_dispatch_load_queue(m); + assert(u->load_state == UNIT_LOADED); + + HASHMAP_FOREACH_KEY(other, k, m->units, i) { + + if (other->ignore_on_snapshot) + continue; + + if (k != other->id) + continue; + + if (UNIT_VTABLE(other)->check_snapshot) + if (!UNIT_VTABLE(other)->check_snapshot(other)) + continue; + + if (!UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(other))) + continue; + + if ((r = unit_add_two_dependencies(u, UNIT_AFTER, UNIT_WANTS, other, true)) < 0) + goto fail; + } + + SNAPSHOT(u)->cleanup = cleanup; + u->allow_isolate = true; + *_s = SNAPSHOT(u); + + return 0; + +fail: + if (u) + unit_add_to_cleanup_queue(u); + + return r; +} + +void snapshot_remove(Snapshot *s) { + assert(s); + + unit_add_to_cleanup_queue(UNIT(s)); +} + +static const char* const snapshot_state_table[_SNAPSHOT_STATE_MAX] = { + [SNAPSHOT_DEAD] = "dead", + [SNAPSHOT_ACTIVE] = "active" +}; + +DEFINE_STRING_TABLE_LOOKUP(snapshot_state, SnapshotState); + +const UnitVTable snapshot_vtable = { + .object_size = sizeof(Snapshot), + + .no_alias = true, + .no_instances = true, + .no_gc = true, + + .init = snapshot_init, + + .load = snapshot_load, + .coldplug = snapshot_coldplug, + + .dump = snapshot_dump, + + .start = snapshot_start, + .stop = snapshot_stop, + + .serialize = snapshot_serialize, + .deserialize_item = snapshot_deserialize_item, + + .active_state = snapshot_active_state, + .sub_state_to_string = snapshot_sub_state_to_string, + + .bus_interface = "org.freedesktop.systemd1.Snapshot", + .bus_message_handler = bus_snapshot_message_handler +}; diff --git a/src/core/snapshot.h b/src/core/snapshot.h new file mode 100644 index 0000000000..56f87cff4d --- /dev/null +++ b/src/core/snapshot.h @@ -0,0 +1,50 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Snapshot Snapshot; + +#include "unit.h" + +typedef enum SnapshotState { + SNAPSHOT_DEAD, + SNAPSHOT_ACTIVE, + _SNAPSHOT_STATE_MAX, + _SNAPSHOT_STATE_INVALID = -1 +} SnapshotState; + +struct Snapshot { + Unit meta; + + SnapshotState state, deserialized_state; + + bool cleanup; + bool by_snapshot_create:1; +}; + +extern const UnitVTable snapshot_vtable; + +int snapshot_create(Manager *m, const char *name, bool cleanup, DBusError *e, Snapshot **s); +void snapshot_remove(Snapshot *s); + +const char* snapshot_state_to_string(SnapshotState i) _const_; +SnapshotState snapshot_state_from_string(const char *s) _pure_; diff --git a/src/core/socket.c b/src/core/socket.c new file mode 100644 index 0000000000..1b08f0a5fd --- /dev/null +++ b/src/core/socket.c @@ -0,0 +1,2411 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/types.h> +#include <sys/stat.h> +#include <unistd.h> +#include <errno.h> +#include <fcntl.h> +#include <sys/epoll.h> +#include <signal.h> +#include <arpa/inet.h> +#include <mqueue.h> +#ifdef HAVE_XATTR +#include <attr/xattr.h> +#endif + +#include "unit.h" +#include "socket.h" +#include "netinet/tcp.h" +#include "log.h" +#include "load-dropin.h" +#include "load-fragment.h" +#include "strv.h" +#include "mkdir.h" +#include "path-util.h" +#include "unit-name.h" +#include "unit-printf.h" +#include "dbus-socket.h" +#include "missing.h" +#include "special.h" +#include "dbus-common.h" +#include "label.h" +#include "exit-status.h" +#include "def.h" + +static const UnitActiveState state_translation_table[_SOCKET_STATE_MAX] = { + [SOCKET_DEAD] = UNIT_INACTIVE, + [SOCKET_START_PRE] = UNIT_ACTIVATING, + [SOCKET_START_POST] = UNIT_ACTIVATING, + [SOCKET_LISTENING] = UNIT_ACTIVE, + [SOCKET_RUNNING] = UNIT_ACTIVE, + [SOCKET_STOP_PRE] = UNIT_DEACTIVATING, + [SOCKET_STOP_PRE_SIGTERM] = UNIT_DEACTIVATING, + [SOCKET_STOP_PRE_SIGKILL] = UNIT_DEACTIVATING, + [SOCKET_STOP_POST] = UNIT_DEACTIVATING, + [SOCKET_FINAL_SIGTERM] = UNIT_DEACTIVATING, + [SOCKET_FINAL_SIGKILL] = UNIT_DEACTIVATING, + [SOCKET_FAILED] = UNIT_FAILED +}; + +static void socket_init(Unit *u) { + Socket *s = SOCKET(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + s->backlog = SOMAXCONN; + s->timeout_usec = DEFAULT_TIMEOUT_USEC; + s->directory_mode = 0755; + s->socket_mode = 0666; + + s->max_connections = 64; + + s->priority = -1; + s->ip_tos = -1; + s->ip_ttl = -1; + s->mark = -1; + + exec_context_init(&s->exec_context); + s->exec_context.std_output = u->manager->default_std_output; + s->exec_context.std_error = u->manager->default_std_error; + kill_context_init(&s->kill_context); + + s->control_command_id = _SOCKET_EXEC_COMMAND_INVALID; +} + +static void socket_unwatch_control_pid(Socket *s) { + assert(s); + + if (s->control_pid <= 0) + return; + + unit_unwatch_pid(UNIT(s), s->control_pid); + s->control_pid = 0; +} + +void socket_free_ports(Socket *s) { + SocketPort *p; + + assert(s); + + while ((p = s->ports)) { + LIST_REMOVE(SocketPort, port, s->ports, p); + + if (p->fd >= 0) { + unit_unwatch_fd(UNIT(s), &p->fd_watch); + close_nointr_nofail(p->fd); + } + + free(p->path); + free(p); + } +} + +static void socket_done(Unit *u) { + Socket *s = SOCKET(u); + + assert(s); + + socket_free_ports(s); + + exec_context_done(&s->exec_context, manager_is_reloading_or_reexecuting(u->manager)); + exec_command_free_array(s->exec_command, _SOCKET_EXEC_COMMAND_MAX); + s->control_command = NULL; + + socket_unwatch_control_pid(s); + + unit_ref_unset(&s->service); + + free(s->tcp_congestion); + s->tcp_congestion = NULL; + + free(s->bind_to_device); + s->bind_to_device = NULL; + + free(s->smack); + free(s->smack_ip_in); + free(s->smack_ip_out); + + unit_unwatch_timer(u, &s->timer_watch); +} + +static int socket_instantiate_service(Socket *s) { + char *prefix, *name; + int r; + Unit *u; + + assert(s); + + /* This fills in s->service if it isn't filled in yet. For + * Accept=yes sockets we create the next connection service + * here. For Accept=no this is mostly a NOP since the service + * is figured out at load time anyway. */ + + if (UNIT_DEREF(s->service)) + return 0; + + assert(s->accept); + + if (!(prefix = unit_name_to_prefix(UNIT(s)->id))) + return -ENOMEM; + + r = asprintf(&name, "%s@%u.service", prefix, s->n_accepted); + free(prefix); + + if (r < 0) + return -ENOMEM; + + r = manager_load_unit(UNIT(s)->manager, name, NULL, NULL, &u); + free(name); + + if (r < 0) + return r; + +#ifdef HAVE_SYSV_COMPAT + if (SERVICE(u)->is_sysv) { + log_error("Using SysV services for socket activation is not supported. Refusing."); + return -ENOENT; + } +#endif + + u->no_gc = true; + unit_ref_set(&s->service, u); + + return unit_add_two_dependencies(UNIT(s), UNIT_BEFORE, UNIT_TRIGGERS, u, false); +} + +static bool have_non_accept_socket(Socket *s) { + SocketPort *p; + + assert(s); + + if (!s->accept) + return true; + + LIST_FOREACH(port, p, s->ports) { + + if (p->type != SOCKET_SOCKET) + return true; + + if (!socket_address_can_accept(&p->address)) + return true; + } + + return false; +} + +static int socket_verify(Socket *s) { + assert(s); + + if (UNIT(s)->load_state != UNIT_LOADED) + return 0; + + if (!s->ports) { + log_error_unit(UNIT(s)->id, + "%s lacks Listen setting. Refusing.", UNIT(s)->id); + return -EINVAL; + } + + if (s->accept && have_non_accept_socket(s)) { + log_error_unit(UNIT(s)->id, + "%s configured for accepting sockets, but sockets are non-accepting. Refusing.", + UNIT(s)->id); + return -EINVAL; + } + + if (s->accept && s->max_connections <= 0) { + log_error_unit(UNIT(s)->id, + "%s's MaxConnection setting too small. Refusing.", UNIT(s)->id); + return -EINVAL; + } + + if (s->accept && UNIT_DEREF(s->service)) { + log_error_unit(UNIT(s)->id, + "Explicit service configuration for accepting sockets not supported on %s. Refusing.", + UNIT(s)->id); + return -EINVAL; + } + + if (s->exec_context.pam_name && s->kill_context.kill_mode != KILL_CONTROL_GROUP) { + log_error_unit(UNIT(s)->id, + "%s has PAM enabled. Kill mode must be set to 'control-group'. Refusing.", + UNIT(s)->id); + return -EINVAL; + } + + return 0; +} + +static bool socket_needs_mount(Socket *s, const char *prefix) { + SocketPort *p; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + + if (p->type == SOCKET_SOCKET) { + if (socket_address_needs_mount(&p->address, prefix)) + return true; + } else if (p->type == SOCKET_FIFO || p->type == SOCKET_SPECIAL) { + if (path_startswith(p->path, prefix)) + return true; + } + } + + return false; +} + +int socket_add_one_mount_link(Socket *s, Mount *m) { + int r; + + assert(s); + assert(m); + + if (UNIT(s)->load_state != UNIT_LOADED || + UNIT(m)->load_state != UNIT_LOADED) + return 0; + + if (!socket_needs_mount(s, m->where)) + return 0; + + r = unit_add_two_dependencies(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, UNIT(m), true); + if (r < 0) + return r; + + return 0; +} + +static int socket_add_mount_links(Socket *s) { + Unit *other; + int r; + + assert(s); + + LIST_FOREACH(units_by_type, other, UNIT(s)->manager->units_by_type[UNIT_MOUNT]) { + r = socket_add_one_mount_link(s, MOUNT(other)); + if (r < 0) + return r; + } + + return 0; +} + +static int socket_add_device_link(Socket *s) { + char *t; + int r; + + assert(s); + + if (!s->bind_to_device || streq(s->bind_to_device, "lo")) + return 0; + + if (asprintf(&t, "/sys/subsystem/net/devices/%s", s->bind_to_device) < 0) + return -ENOMEM; + + r = unit_add_node_link(UNIT(s), t, false); + free(t); + + return r; +} + +static int socket_add_default_dependencies(Socket *s) { + int r; + assert(s); + + r = unit_add_dependency_by_name(UNIT(s), UNIT_BEFORE, SPECIAL_SOCKETS_TARGET, NULL, true); + if (r < 0) + return r; + + if (UNIT(s)->manager->running_as == SYSTEMD_SYSTEM) { + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, SPECIAL_SYSINIT_TARGET, NULL, true); + if (r < 0) + return r; + } + + return unit_add_two_dependencies_by_name(UNIT(s), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +_pure_ static bool socket_has_exec(Socket *s) { + unsigned i; + assert(s); + + for (i = 0; i < _SOCKET_EXEC_COMMAND_MAX; i++) + if (s->exec_command[i]) + return true; + + return false; +} + +static int socket_load(Unit *u) { + Socket *s = SOCKET(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + if ((r = unit_load_fragment_and_dropin(u)) < 0) + return r; + + /* This is a new unit? Then let's add in some extras */ + if (u->load_state == UNIT_LOADED) { + + if (have_non_accept_socket(s)) { + + if (!UNIT_DEREF(s->service)) { + Unit *x; + + r = unit_load_related_unit(u, ".service", &x); + if (r < 0) + return r; + + unit_ref_set(&s->service, x); + } + + r = unit_add_two_dependencies(u, UNIT_BEFORE, UNIT_TRIGGERS, UNIT_DEREF(s->service), true); + if (r < 0) + return r; + } + + if ((r = socket_add_mount_links(s)) < 0) + return r; + + if ((r = socket_add_device_link(s)) < 0) + return r; + + if (socket_has_exec(s)) + if ((r = unit_add_exec_dependencies(u, &s->exec_context)) < 0) + return r; + + if ((r = unit_add_default_cgroups(u)) < 0) + return r; + + if (UNIT(s)->default_dependencies) + if ((r = socket_add_default_dependencies(s)) < 0) + return r; + + r = unit_exec_context_defaults(u, &s->exec_context); + if (r < 0) + return r; + } + + return socket_verify(s); +} + +_const_ static const char* listen_lookup(int family, int type) { + + if (family == AF_NETLINK) + return "ListenNetlink"; + + if (type == SOCK_STREAM) + return "ListenStream"; + else if (type == SOCK_DGRAM) + return "ListenDatagram"; + else if (type == SOCK_SEQPACKET) + return "ListenSequentialPacket"; + + assert_not_reached("Unknown socket type"); + return NULL; +} + +static void socket_dump(Unit *u, FILE *f, const char *prefix) { + + SocketExecCommand c; + Socket *s = SOCKET(u); + SocketPort *p; + const char *prefix2; + char *p2; + + assert(s); + assert(f); + + p2 = strappend(prefix, "\t"); + prefix2 = p2 ? p2 : prefix; + + fprintf(f, + "%sSocket State: %s\n" + "%sResult: %s\n" + "%sBindIPv6Only: %s\n" + "%sBacklog: %u\n" + "%sSocketMode: %04o\n" + "%sDirectoryMode: %04o\n" + "%sKeepAlive: %s\n" + "%sFreeBind: %s\n" + "%sTransparent: %s\n" + "%sBroadcast: %s\n" + "%sPassCredentials: %s\n" + "%sPassSecurity: %s\n" + "%sTCPCongestion: %s\n", + prefix, socket_state_to_string(s->state), + prefix, socket_result_to_string(s->result), + prefix, socket_address_bind_ipv6_only_to_string(s->bind_ipv6_only), + prefix, s->backlog, + prefix, s->socket_mode, + prefix, s->directory_mode, + prefix, yes_no(s->keep_alive), + prefix, yes_no(s->free_bind), + prefix, yes_no(s->transparent), + prefix, yes_no(s->broadcast), + prefix, yes_no(s->pass_cred), + prefix, yes_no(s->pass_sec), + prefix, strna(s->tcp_congestion)); + + if (s->control_pid > 0) + fprintf(f, + "%sControl PID: %lu\n", + prefix, (unsigned long) s->control_pid); + + if (s->bind_to_device) + fprintf(f, + "%sBindToDevice: %s\n", + prefix, s->bind_to_device); + + if (s->accept) + fprintf(f, + "%sAccepted: %u\n" + "%sNConnections: %u\n" + "%sMaxConnections: %u\n", + prefix, s->n_accepted, + prefix, s->n_connections, + prefix, s->max_connections); + + if (s->priority >= 0) + fprintf(f, + "%sPriority: %i\n", + prefix, s->priority); + + if (s->receive_buffer > 0) + fprintf(f, + "%sReceiveBuffer: %zu\n", + prefix, s->receive_buffer); + + if (s->send_buffer > 0) + fprintf(f, + "%sSendBuffer: %zu\n", + prefix, s->send_buffer); + + if (s->ip_tos >= 0) + fprintf(f, + "%sIPTOS: %i\n", + prefix, s->ip_tos); + + if (s->ip_ttl >= 0) + fprintf(f, + "%sIPTTL: %i\n", + prefix, s->ip_ttl); + + if (s->pipe_size > 0) + fprintf(f, + "%sPipeSize: %zu\n", + prefix, s->pipe_size); + + if (s->mark >= 0) + fprintf(f, + "%sMark: %i\n", + prefix, s->mark); + + if (s->mq_maxmsg > 0) + fprintf(f, + "%sMessageQueueMaxMessages: %li\n", + prefix, s->mq_maxmsg); + + if (s->mq_msgsize > 0) + fprintf(f, + "%sMessageQueueMessageSize: %li\n", + prefix, s->mq_msgsize); + + if (s->smack) + fprintf(f, + "%sSmackLabel: %s\n", + prefix, s->smack); + + if (s->smack_ip_in) + fprintf(f, + "%sSmackLabelIPIn: %s\n", + prefix, s->smack_ip_in); + + if (s->smack_ip_out) + fprintf(f, + "%sSmackLabelIPOut: %s\n", + prefix, s->smack_ip_out); + + LIST_FOREACH(port, p, s->ports) { + + if (p->type == SOCKET_SOCKET) { + const char *t; + int r; + char *k = NULL; + + if ((r = socket_address_print(&p->address, &k)) < 0) + t = strerror(-r); + else + t = k; + + fprintf(f, "%s%s: %s\n", prefix, listen_lookup(socket_address_family(&p->address), p->address.type), t); + free(k); + } else if (p->type == SOCKET_SPECIAL) + fprintf(f, "%sListenSpecial: %s\n", prefix, p->path); + else if (p->type == SOCKET_MQUEUE) + fprintf(f, "%sListenMessageQueue: %s\n", prefix, p->path); + else + fprintf(f, "%sListenFIFO: %s\n", prefix, p->path); + } + + exec_context_dump(&s->exec_context, f, prefix); + kill_context_dump(&s->kill_context, f, prefix); + + for (c = 0; c < _SOCKET_EXEC_COMMAND_MAX; c++) { + if (!s->exec_command[c]) + continue; + + fprintf(f, "%s-> %s:\n", + prefix, socket_exec_command_to_string(c)); + + exec_command_dump_list(s->exec_command[c], f, prefix2); + } + + free(p2); +} + +static int instance_from_socket(int fd, unsigned nr, char **instance) { + socklen_t l; + char *r; + union { + struct sockaddr sa; + struct sockaddr_un un; + struct sockaddr_in in; + struct sockaddr_in6 in6; + struct sockaddr_storage storage; + } local, remote; + + assert(fd >= 0); + assert(instance); + + l = sizeof(local); + if (getsockname(fd, &local.sa, &l) < 0) + return -errno; + + l = sizeof(remote); + if (getpeername(fd, &remote.sa, &l) < 0) + return -errno; + + switch (local.sa.sa_family) { + + case AF_INET: { + uint32_t + a = ntohl(local.in.sin_addr.s_addr), + b = ntohl(remote.in.sin_addr.s_addr); + + if (asprintf(&r, + "%u-%u.%u.%u.%u:%u-%u.%u.%u.%u:%u", + nr, + a >> 24, (a >> 16) & 0xFF, (a >> 8) & 0xFF, a & 0xFF, + ntohs(local.in.sin_port), + b >> 24, (b >> 16) & 0xFF, (b >> 8) & 0xFF, b & 0xFF, + ntohs(remote.in.sin_port)) < 0) + return -ENOMEM; + + break; + } + + case AF_INET6: { + static const unsigned char ipv4_prefix[] = { + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0xFF, 0xFF + }; + + if (memcmp(&local.in6.sin6_addr, ipv4_prefix, sizeof(ipv4_prefix)) == 0 && + memcmp(&remote.in6.sin6_addr, ipv4_prefix, sizeof(ipv4_prefix)) == 0) { + const uint8_t + *a = local.in6.sin6_addr.s6_addr+12, + *b = remote.in6.sin6_addr.s6_addr+12; + + if (asprintf(&r, + "%u-%u.%u.%u.%u:%u-%u.%u.%u.%u:%u", + nr, + a[0], a[1], a[2], a[3], + ntohs(local.in6.sin6_port), + b[0], b[1], b[2], b[3], + ntohs(remote.in6.sin6_port)) < 0) + return -ENOMEM; + } else { + char a[INET6_ADDRSTRLEN], b[INET6_ADDRSTRLEN]; + + if (asprintf(&r, + "%u-%s:%u-%s:%u", + nr, + inet_ntop(AF_INET6, &local.in6.sin6_addr, a, sizeof(a)), + ntohs(local.in6.sin6_port), + inet_ntop(AF_INET6, &remote.in6.sin6_addr, b, sizeof(b)), + ntohs(remote.in6.sin6_port)) < 0) + return -ENOMEM; + } + + break; + } + + case AF_UNIX: { + struct ucred ucred; + + l = sizeof(ucred); + if (getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &ucred, &l) < 0) + return -errno; + + if (asprintf(&r, + "%u-%lu-%lu", + nr, + (unsigned long) ucred.pid, + (unsigned long) ucred.uid) < 0) + return -ENOMEM; + + break; + } + + default: + assert_not_reached("Unhandled socket type."); + } + + *instance = r; + return 0; +} + +static void socket_close_fds(Socket *s) { + SocketPort *p; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + if (p->fd < 0) + continue; + + unit_unwatch_fd(UNIT(s), &p->fd_watch); + close_nointr_nofail(p->fd); + + /* One little note: we should never delete any sockets + * in the file system here! After all some other + * process we spawned might still have a reference of + * this fd and wants to continue to use it. Therefore + * we delete sockets in the file system before we + * create a new one, not after we stopped using + * one! */ + + p->fd = -1; + } +} + +static void socket_apply_socket_options(Socket *s, int fd) { + assert(s); + assert(fd >= 0); + + if (s->keep_alive) { + int b = s->keep_alive; + if (setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, &b, sizeof(b)) < 0) + log_warning_unit(UNIT(s)->id, "SO_KEEPALIVE failed: %m"); + } + + if (s->broadcast) { + int one = 1; + if (setsockopt(fd, SOL_SOCKET, SO_BROADCAST, &one, sizeof(one)) < 0) + log_warning_unit(UNIT(s)->id, "SO_BROADCAST failed: %m"); + } + + if (s->pass_cred) { + int one = 1; + if (setsockopt(fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one)) < 0) + log_warning_unit(UNIT(s)->id, "SO_PASSCRED failed: %m"); + } + + if (s->pass_sec) { + int one = 1; + if (setsockopt(fd, SOL_SOCKET, SO_PASSSEC, &one, sizeof(one)) < 0) + log_warning_unit(UNIT(s)->id, "SO_PASSSEC failed: %m"); + } + + if (s->priority >= 0) + if (setsockopt(fd, SOL_SOCKET, SO_PRIORITY, &s->priority, sizeof(s->priority)) < 0) + log_warning_unit(UNIT(s)->id, "SO_PRIORITY failed: %m"); + + if (s->receive_buffer > 0) { + int value = (int) s->receive_buffer; + + /* We first try with SO_RCVBUFFORCE, in case we have the perms for that */ + + if (setsockopt(fd, SOL_SOCKET, SO_RCVBUFFORCE, &value, sizeof(value)) < 0) + if (setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &value, sizeof(value)) < 0) + log_warning_unit(UNIT(s)->id, "SO_RCVBUF failed: %m"); + } + + if (s->send_buffer > 0) { + int value = (int) s->send_buffer; + if (setsockopt(fd, SOL_SOCKET, SO_SNDBUFFORCE, &value, sizeof(value)) < 0) + if (setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &value, sizeof(value)) < 0) + log_warning_unit(UNIT(s)->id, "SO_SNDBUF failed: %m"); + } + + if (s->mark >= 0) + if (setsockopt(fd, SOL_SOCKET, SO_MARK, &s->mark, sizeof(s->mark)) < 0) + log_warning_unit(UNIT(s)->id, "SO_MARK failed: %m"); + + if (s->ip_tos >= 0) + if (setsockopt(fd, IPPROTO_IP, IP_TOS, &s->ip_tos, sizeof(s->ip_tos)) < 0) + log_warning_unit(UNIT(s)->id, "IP_TOS failed: %m"); + + if (s->ip_ttl >= 0) { + int r, x; + + r = setsockopt(fd, IPPROTO_IP, IP_TTL, &s->ip_ttl, sizeof(s->ip_ttl)); + + if (socket_ipv6_is_supported()) + x = setsockopt(fd, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &s->ip_ttl, sizeof(s->ip_ttl)); + else { + x = -1; + errno = EAFNOSUPPORT; + } + + if (r < 0 && x < 0) + log_warning_unit(UNIT(s)->id, + "IP_TTL/IPV6_UNICAST_HOPS failed: %m"); + } + + if (s->tcp_congestion) + if (setsockopt(fd, SOL_TCP, TCP_CONGESTION, s->tcp_congestion, strlen(s->tcp_congestion)+1) < 0) + log_warning_unit(UNIT(s)->id, "TCP_CONGESTION failed: %m"); + +#ifdef HAVE_XATTR + if (s->smack_ip_in) + if (fsetxattr(fd, "security.SMACK64IPIN", s->smack_ip_in, strlen(s->smack_ip_in), 0) < 0) + log_error_unit(UNIT(s)->id, + "fsetxattr(\"security.SMACK64IPIN\"): %m"); + + if (s->smack_ip_out) + if (fsetxattr(fd, "security.SMACK64IPOUT", s->smack_ip_out, strlen(s->smack_ip_out), 0) < 0) + log_error_unit(UNIT(s)->id, + "fsetxattr(\"security.SMACK64IPOUT\"): %m"); +#endif +} + +static void socket_apply_fifo_options(Socket *s, int fd) { + assert(s); + assert(fd >= 0); + + if (s->pipe_size > 0) + if (fcntl(fd, F_SETPIPE_SZ, s->pipe_size) < 0) + log_warning_unit(UNIT(s)->id, + "F_SETPIPE_SZ: %m"); + +#ifdef HAVE_XATTR + if (s->smack) + if (fsetxattr(fd, "security.SMACK64", s->smack, strlen(s->smack), 0) < 0) + log_error_unit(UNIT(s)->id, + "fsetxattr(\"security.SMACK64\"): %m"); +#endif +} + +static int fifo_address_create( + const char *path, + mode_t directory_mode, + mode_t socket_mode, + int *_fd) { + + int fd = -1, r = 0; + struct stat st; + mode_t old_mask; + + assert(path); + assert(_fd); + + mkdir_parents_label(path, directory_mode); + + r = label_context_set(path, S_IFIFO); + if (r < 0) + goto fail; + + /* Enforce the right access mode for the fifo */ + old_mask = umask(~ socket_mode); + + /* Include the original umask in our mask */ + umask(~socket_mode | old_mask); + + r = mkfifo(path, socket_mode); + umask(old_mask); + + if (r < 0 && errno != EEXIST) { + r = -errno; + goto fail; + } + + if ((fd = open(path, O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK|O_NOFOLLOW)) < 0) { + r = -errno; + goto fail; + } + + label_context_clear(); + + if (fstat(fd, &st) < 0) { + r = -errno; + goto fail; + } + + if (!S_ISFIFO(st.st_mode) || + (st.st_mode & 0777) != (socket_mode & ~old_mask) || + st.st_uid != getuid() || + st.st_gid != getgid()) { + + r = -EEXIST; + goto fail; + } + + *_fd = fd; + return 0; + +fail: + label_context_clear(); + + if (fd >= 0) + close_nointr_nofail(fd); + + return r; +} + +static int special_address_create( + const char *path, + int *_fd) { + + int fd = -1, r = 0; + struct stat st; + + assert(path); + assert(_fd); + + if ((fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NONBLOCK|O_NOFOLLOW)) < 0) { + r = -errno; + goto fail; + } + + if (fstat(fd, &st) < 0) { + r = -errno; + goto fail; + } + + /* Check whether this is a /proc, /sys or /dev file or char device */ + if (!S_ISREG(st.st_mode) && !S_ISCHR(st.st_mode)) { + r = -EEXIST; + goto fail; + } + + *_fd = fd; + return 0; + +fail: + if (fd >= 0) + close_nointr_nofail(fd); + + return r; +} + +static int mq_address_create( + const char *path, + mode_t mq_mode, + long maxmsg, + long msgsize, + int *_fd) { + + int fd = -1, r = 0; + struct stat st; + mode_t old_mask; + struct mq_attr _attr, *attr = NULL; + + assert(path); + assert(_fd); + + if (maxmsg > 0 && msgsize > 0) { + zero(_attr); + _attr.mq_flags = O_NONBLOCK; + _attr.mq_maxmsg = maxmsg; + _attr.mq_msgsize = msgsize; + attr = &_attr; + } + + /* Enforce the right access mode for the mq */ + old_mask = umask(~ mq_mode); + + /* Include the original umask in our mask */ + umask(~mq_mode | old_mask); + + fd = mq_open(path, O_RDONLY|O_CLOEXEC|O_NONBLOCK|O_CREAT, mq_mode, attr); + umask(old_mask); + + if (fd < 0) { + r = -errno; + goto fail; + } + + if (fstat(fd, &st) < 0) { + r = -errno; + goto fail; + } + + if ((st.st_mode & 0777) != (mq_mode & ~old_mask) || + st.st_uid != getuid() || + st.st_gid != getgid()) { + + r = -EEXIST; + goto fail; + } + + *_fd = fd; + return 0; + +fail: + if (fd >= 0) + close_nointr_nofail(fd); + + return r; +} + +static int socket_open_fds(Socket *s) { + SocketPort *p; + int r; + char *label = NULL; + bool know_label = false; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + + if (p->fd >= 0) + continue; + + if (p->type == SOCKET_SOCKET) { + + if (!know_label) { + + if ((r = socket_instantiate_service(s)) < 0) + return r; + + if (UNIT_DEREF(s->service) && + SERVICE(UNIT_DEREF(s->service))->exec_command[SERVICE_EXEC_START]) { + r = label_get_create_label_from_exe(SERVICE(UNIT_DEREF(s->service))->exec_command[SERVICE_EXEC_START]->path, &label); + + if (r < 0) { + if (r != -EPERM) + return r; + } + } + + know_label = true; + } + + if ((r = socket_address_listen( + &p->address, + s->backlog, + s->bind_ipv6_only, + s->bind_to_device, + s->free_bind, + s->transparent, + s->directory_mode, + s->socket_mode, + label, + &p->fd)) < 0) + goto rollback; + + socket_apply_socket_options(s, p->fd); + + } else if (p->type == SOCKET_SPECIAL) { + + if ((r = special_address_create( + p->path, + &p->fd)) < 0) + goto rollback; + + } else if (p->type == SOCKET_FIFO) { + + if ((r = fifo_address_create( + p->path, + s->directory_mode, + s->socket_mode, + &p->fd)) < 0) + goto rollback; + + socket_apply_fifo_options(s, p->fd); + } else if (p->type == SOCKET_MQUEUE) { + + if ((r = mq_address_create( + p->path, + s->socket_mode, + s->mq_maxmsg, + s->mq_msgsize, + &p->fd)) < 0) + goto rollback; + } else + assert_not_reached("Unknown port type"); + } + + label_free(label); + return 0; + +rollback: + socket_close_fds(s); + label_free(label); + return r; +} + +static void socket_unwatch_fds(Socket *s) { + SocketPort *p; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + if (p->fd < 0) + continue; + + unit_unwatch_fd(UNIT(s), &p->fd_watch); + } +} + +static int socket_watch_fds(Socket *s) { + SocketPort *p; + int r; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + if (p->fd < 0) + continue; + + p->fd_watch.socket_accept = + s->accept && + p->type == SOCKET_SOCKET && + socket_address_can_accept(&p->address); + + if ((r = unit_watch_fd(UNIT(s), p->fd, EPOLLIN, &p->fd_watch)) < 0) + goto fail; + } + + return 0; + +fail: + socket_unwatch_fds(s); + return r; +} + +static void socket_set_state(Socket *s, SocketState state) { + SocketState old_state; + assert(s); + + old_state = s->state; + s->state = state; + + if (state != SOCKET_START_PRE && + state != SOCKET_START_POST && + state != SOCKET_STOP_PRE && + state != SOCKET_STOP_PRE_SIGTERM && + state != SOCKET_STOP_PRE_SIGKILL && + state != SOCKET_STOP_POST && + state != SOCKET_FINAL_SIGTERM && + state != SOCKET_FINAL_SIGKILL) { + unit_unwatch_timer(UNIT(s), &s->timer_watch); + socket_unwatch_control_pid(s); + s->control_command = NULL; + s->control_command_id = _SOCKET_EXEC_COMMAND_INVALID; + } + + if (state != SOCKET_LISTENING) + socket_unwatch_fds(s); + + if (state != SOCKET_START_POST && + state != SOCKET_LISTENING && + state != SOCKET_RUNNING && + state != SOCKET_STOP_PRE && + state != SOCKET_STOP_PRE_SIGTERM && + state != SOCKET_STOP_PRE_SIGKILL) + socket_close_fds(s); + + if (state != old_state) + log_debug_unit(UNIT(s)->id, + "%s changed %s -> %s", UNIT(s)->id, + socket_state_to_string(old_state), + socket_state_to_string(state)); + + unit_notify(UNIT(s), state_translation_table[old_state], state_translation_table[state], true); +} + +static int socket_coldplug(Unit *u) { + Socket *s = SOCKET(u); + int r; + + assert(s); + assert(s->state == SOCKET_DEAD); + + if (s->deserialized_state != s->state) { + + if (s->deserialized_state == SOCKET_START_PRE || + s->deserialized_state == SOCKET_START_POST || + s->deserialized_state == SOCKET_STOP_PRE || + s->deserialized_state == SOCKET_STOP_PRE_SIGTERM || + s->deserialized_state == SOCKET_STOP_PRE_SIGKILL || + s->deserialized_state == SOCKET_STOP_POST || + s->deserialized_state == SOCKET_FINAL_SIGTERM || + s->deserialized_state == SOCKET_FINAL_SIGKILL) { + + if (s->control_pid <= 0) + return -EBADMSG; + + r = unit_watch_pid(UNIT(s), s->control_pid); + if (r < 0) + return r; + + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, s->timeout_usec, &s->timer_watch); + if (r < 0) + return r; + } + + if (s->deserialized_state == SOCKET_START_POST || + s->deserialized_state == SOCKET_LISTENING || + s->deserialized_state == SOCKET_RUNNING || + s->deserialized_state == SOCKET_STOP_PRE || + s->deserialized_state == SOCKET_STOP_PRE_SIGTERM || + s->deserialized_state == SOCKET_STOP_PRE_SIGKILL) + if ((r = socket_open_fds(s)) < 0) + return r; + + if (s->deserialized_state == SOCKET_LISTENING) + if ((r = socket_watch_fds(s)) < 0) + return r; + + socket_set_state(s, s->deserialized_state); + } + + return 0; +} + +static int socket_spawn(Socket *s, ExecCommand *c, pid_t *_pid) { + pid_t pid; + int r; + char **argv; + + assert(s); + assert(c); + assert(_pid); + + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, s->timeout_usec, &s->timer_watch); + if (r < 0) + goto fail; + + argv = unit_full_printf_strv(UNIT(s), c->argv); + if (!argv) { + r = -ENOMEM; + goto fail; + } + + r = exec_spawn(c, + argv, + &s->exec_context, + NULL, 0, + UNIT(s)->manager->environment, + true, + true, + true, + UNIT(s)->manager->confirm_spawn, + UNIT(s)->cgroup_bondings, + UNIT(s)->cgroup_attributes, + NULL, + UNIT(s)->id, + NULL, + &pid); + + strv_free(argv); + if (r < 0) + goto fail; + + if ((r = unit_watch_pid(UNIT(s), pid)) < 0) + /* FIXME: we need to do something here */ + goto fail; + + *_pid = pid; + + return 0; + +fail: + unit_unwatch_timer(UNIT(s), &s->timer_watch); + + return r; +} + +static void socket_enter_dead(Socket *s, SocketResult f) { + assert(s); + + if (f != SOCKET_SUCCESS) + s->result = f; + + exec_context_tmp_dirs_done(&s->exec_context); + socket_set_state(s, s->result != SOCKET_SUCCESS ? SOCKET_FAILED : SOCKET_DEAD); +} + +static void socket_enter_signal(Socket *s, SocketState state, SocketResult f); + +static void socket_enter_stop_post(Socket *s, SocketResult f) { + int r; + assert(s); + + if (f != SOCKET_SUCCESS) + s->result = f; + + socket_unwatch_control_pid(s); + + s->control_command_id = SOCKET_EXEC_STOP_POST; + + if ((s->control_command = s->exec_command[SOCKET_EXEC_STOP_POST])) { + if ((r = socket_spawn(s, s->control_command, &s->control_pid)) < 0) + goto fail; + + socket_set_state(s, SOCKET_STOP_POST); + } else + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_SUCCESS); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'stop-post' task: %s", + UNIT(s)->id, strerror(-r)); + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_signal(Socket *s, SocketState state, SocketResult f) { + int r; + + assert(s); + + if (f != SOCKET_SUCCESS) + s->result = f; + + r = unit_kill_context( + UNIT(s), + &s->kill_context, + state != SOCKET_STOP_PRE_SIGTERM && state != SOCKET_FINAL_SIGTERM, + -1, + s->control_pid, + false); + if (r < 0) + goto fail; + + if (r > 0) { + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, s->timeout_usec, &s->timer_watch); + if (r < 0) + goto fail; + + socket_set_state(s, state); + } else if (state == SOCKET_STOP_PRE_SIGTERM || state == SOCKET_STOP_PRE_SIGKILL) + socket_enter_stop_post(s, SOCKET_SUCCESS); + else + socket_enter_dead(s, SOCKET_SUCCESS); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to kill processes: %s", + UNIT(s)->id, strerror(-r)); + + if (state == SOCKET_STOP_PRE_SIGTERM || state == SOCKET_STOP_PRE_SIGKILL) + socket_enter_stop_post(s, SOCKET_FAILURE_RESOURCES); + else + socket_enter_dead(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_stop_pre(Socket *s, SocketResult f) { + int r; + assert(s); + + if (f != SOCKET_SUCCESS) + s->result = f; + + socket_unwatch_control_pid(s); + + s->control_command_id = SOCKET_EXEC_STOP_PRE; + + if ((s->control_command = s->exec_command[SOCKET_EXEC_STOP_PRE])) { + if ((r = socket_spawn(s, s->control_command, &s->control_pid)) < 0) + goto fail; + + socket_set_state(s, SOCKET_STOP_PRE); + } else + socket_enter_stop_post(s, SOCKET_SUCCESS); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'stop-pre' task: %s", + UNIT(s)->id, strerror(-r)); + socket_enter_stop_post(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_listening(Socket *s) { + int r; + assert(s); + + r = socket_watch_fds(s); + if (r < 0) { + log_warning_unit(UNIT(s)->id, + "%s failed to watch sockets: %s", + UNIT(s)->id, strerror(-r)); + goto fail; + } + + socket_set_state(s, SOCKET_LISTENING); + return; + +fail: + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_start_post(Socket *s) { + int r; + assert(s); + + r = socket_open_fds(s); + if (r < 0) { + log_warning_unit(UNIT(s)->id, + "%s failed to listen on sockets: %s", + UNIT(s)->id, strerror(-r)); + goto fail; + } + + socket_unwatch_control_pid(s); + + s->control_command_id = SOCKET_EXEC_START_POST; + + if ((s->control_command = s->exec_command[SOCKET_EXEC_START_POST])) { + r = socket_spawn(s, s->control_command, &s->control_pid); + if (r < 0) { + log_warning_unit(UNIT(s)->id, + "%s failed to run 'start-post' task: %s", + UNIT(s)->id, strerror(-r)); + goto fail; + } + + socket_set_state(s, SOCKET_START_POST); + } else + socket_enter_listening(s); + + return; + +fail: + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_start_pre(Socket *s) { + int r; + assert(s); + + socket_unwatch_control_pid(s); + + s->control_command_id = SOCKET_EXEC_START_PRE; + + if ((s->control_command = s->exec_command[SOCKET_EXEC_START_PRE])) { + if ((r = socket_spawn(s, s->control_command, &s->control_pid)) < 0) + goto fail; + + socket_set_state(s, SOCKET_START_PRE); + } else + socket_enter_start_post(s); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'start-pre' task: %s", + UNIT(s)->id, strerror(-r)); + socket_enter_dead(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_running(Socket *s, int cfd) { + int r; + DBusError error; + + assert(s); + dbus_error_init(&error); + + /* We don't take connections anymore if we are supposed to + * shut down anyway */ + if (unit_stop_pending(UNIT(s))) { + log_debug_unit(UNIT(s)->id, + "Suppressing connection request on %s since unit stop is scheduled.", + UNIT(s)->id); + + if (cfd >= 0) + close_nointr_nofail(cfd); + else { + /* Flush all sockets by closing and reopening them */ + socket_close_fds(s); + + r = socket_watch_fds(s); + if (r < 0) { + log_warning_unit(UNIT(s)->id, + "%s failed to watch sockets: %s", + UNIT(s)->id, strerror(-r)); + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); + } + } + + return; + } + + if (cfd < 0) { + Iterator i; + Unit *u; + bool pending = false; + + /* If there's already a start pending don't bother to + * do anything */ + SET_FOREACH(u, UNIT(s)->dependencies[UNIT_TRIGGERS], i) + if (unit_active_or_pending(u)) { + pending = true; + break; + } + + if (!pending) { + r = manager_add_job(UNIT(s)->manager, JOB_START, UNIT_DEREF(s->service), JOB_REPLACE, true, &error, NULL); + if (r < 0) + goto fail; + } + + socket_set_state(s, SOCKET_RUNNING); + } else { + char *prefix, *instance = NULL, *name; + Service *service; + + if (s->n_connections >= s->max_connections) { + log_warning_unit(UNIT(s)->id, + "%s: Too many incoming connections (%u)", + UNIT(s)->id, s->n_connections); + close_nointr_nofail(cfd); + return; + } + + r = socket_instantiate_service(s); + if (r < 0) + goto fail; + + r = instance_from_socket(cfd, s->n_accepted, &instance); + if (r < 0) { + if (r != -ENOTCONN) + goto fail; + + /* ENOTCONN is legitimate if TCP RST was received. + * This connection is over, but the socket unit lives on. */ + close_nointr_nofail(cfd); + return; + } + + prefix = unit_name_to_prefix(UNIT(s)->id); + if (!prefix) { + free(instance); + r = -ENOMEM; + goto fail; + } + + name = unit_name_build(prefix, instance, ".service"); + free(prefix); + free(instance); + + if (!name) { + r = -ENOMEM; + goto fail; + } + + r = unit_add_name(UNIT_DEREF(s->service), name); + if (r < 0) { + free(name); + goto fail; + } + + service = SERVICE(UNIT_DEREF(s->service)); + unit_ref_unset(&s->service); + s->n_accepted ++; + + UNIT(service)->no_gc = false; + + unit_choose_id(UNIT(service), name); + free(name); + + r = service_set_socket_fd(service, cfd, s); + if (r < 0) + goto fail; + + cfd = -1; + s->n_connections ++; + + r = manager_add_job(UNIT(s)->manager, JOB_START, UNIT(service), JOB_REPLACE, true, &error, NULL); + if (r < 0) + goto fail; + + /* Notify clients about changed counters */ + unit_add_to_dbus_queue(UNIT(s)); + } + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to queue service startup job (Maybe the service file is missing or not a %s unit?): %s", + UNIT(s)->id, + cfd >= 0 ? "template" : "non-template", + bus_error(&error, r)); + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); + + if (cfd >= 0) + close_nointr_nofail(cfd); + + dbus_error_free(&error); +} + +static void socket_run_next(Socket *s) { + int r; + + assert(s); + assert(s->control_command); + assert(s->control_command->command_next); + + socket_unwatch_control_pid(s); + + s->control_command = s->control_command->command_next; + + if ((r = socket_spawn(s, s->control_command, &s->control_pid)) < 0) + goto fail; + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run next task: %s", + UNIT(s)->id, strerror(-r)); + + if (s->state == SOCKET_START_POST) + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); + else if (s->state == SOCKET_STOP_POST) + socket_enter_dead(s, SOCKET_FAILURE_RESOURCES); + else + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_FAILURE_RESOURCES); +} + +static int socket_start(Unit *u) { + Socket *s = SOCKET(u); + + assert(s); + + /* We cannot fulfill this request right now, try again later + * please! */ + if (s->state == SOCKET_STOP_PRE || + s->state == SOCKET_STOP_PRE_SIGKILL || + s->state == SOCKET_STOP_PRE_SIGTERM || + s->state == SOCKET_STOP_POST || + s->state == SOCKET_FINAL_SIGTERM || + s->state == SOCKET_FINAL_SIGKILL) + return -EAGAIN; + + if (s->state == SOCKET_START_PRE || + s->state == SOCKET_START_POST) + return 0; + + /* Cannot run this without the service being around */ + if (UNIT_DEREF(s->service)) { + Service *service; + + service = SERVICE(UNIT_DEREF(s->service)); + + if (UNIT(service)->load_state != UNIT_LOADED) { + log_error_unit(u->id, + "Socket service %s not loaded, refusing.", + UNIT(service)->id); + return -ENOENT; + } + + /* If the service is already active we cannot start the + * socket */ + if (service->state != SERVICE_DEAD && + service->state != SERVICE_FAILED && + service->state != SERVICE_AUTO_RESTART) { + log_error_unit(u->id, + "Socket service %s already active, refusing.", + UNIT(service)->id); + return -EBUSY; + } + +#ifdef HAVE_SYSV_COMPAT + if (service->is_sysv) { + log_error_unit(u->id, + "Using SysV services for socket activation is not supported. Refusing."); + return -ENOENT; + } +#endif + } + + assert(s->state == SOCKET_DEAD || s->state == SOCKET_FAILED); + + s->result = SOCKET_SUCCESS; + socket_enter_start_pre(s); + return 0; +} + +static int socket_stop(Unit *u) { + Socket *s = SOCKET(u); + + assert(s); + + /* Already on it */ + if (s->state == SOCKET_STOP_PRE || + s->state == SOCKET_STOP_PRE_SIGTERM || + s->state == SOCKET_STOP_PRE_SIGKILL || + s->state == SOCKET_STOP_POST || + s->state == SOCKET_FINAL_SIGTERM || + s->state == SOCKET_FINAL_SIGKILL) + return 0; + + /* If there's already something running we go directly into + * kill mode. */ + if (s->state == SOCKET_START_PRE || + s->state == SOCKET_START_POST) { + socket_enter_signal(s, SOCKET_STOP_PRE_SIGTERM, SOCKET_SUCCESS); + return -EAGAIN; + } + + assert(s->state == SOCKET_LISTENING || s->state == SOCKET_RUNNING); + + socket_enter_stop_pre(s, SOCKET_SUCCESS); + return 0; +} + +static int socket_serialize(Unit *u, FILE *f, FDSet *fds) { + Socket *s = SOCKET(u); + SocketPort *p; + int r; + + assert(u); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", socket_state_to_string(s->state)); + unit_serialize_item(u, f, "result", socket_result_to_string(s->result)); + unit_serialize_item_format(u, f, "n-accepted", "%u", s->n_accepted); + + if (s->control_pid > 0) + unit_serialize_item_format(u, f, "control-pid", "%lu", (unsigned long) s->control_pid); + + if (s->control_command_id >= 0) + unit_serialize_item(u, f, "control-command", socket_exec_command_to_string(s->control_command_id)); + + LIST_FOREACH(port, p, s->ports) { + int copy; + + if (p->fd < 0) + continue; + + if ((copy = fdset_put_dup(fds, p->fd)) < 0) + return copy; + + if (p->type == SOCKET_SOCKET) { + char *t; + + r = socket_address_print(&p->address, &t); + if (r < 0) + return r; + + if (socket_address_family(&p->address) == AF_NETLINK) + unit_serialize_item_format(u, f, "netlink", "%i %s", copy, t); + else + unit_serialize_item_format(u, f, "socket", "%i %i %s", copy, p->address.type, t); + free(t); + } else if (p->type == SOCKET_SPECIAL) + unit_serialize_item_format(u, f, "special", "%i %s", copy, p->path); + else if (p->type == SOCKET_MQUEUE) + unit_serialize_item_format(u, f, "mqueue", "%i %s", copy, p->path); + else { + assert(p->type == SOCKET_FIFO); + unit_serialize_item_format(u, f, "fifo", "%i %s", copy, p->path); + } + } + + exec_context_serialize(&s->exec_context, UNIT(s), f); + + return 0; +} + +static int socket_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Socket *s = SOCKET(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + SocketState state; + + state = socket_state_from_string(value); + if (state < 0) + log_debug_unit(u->id, + "Failed to parse state value %s", value); + else + s->deserialized_state = state; + } else if (streq(key, "result")) { + SocketResult f; + + f = socket_result_from_string(value); + if (f < 0) + log_debug_unit(u->id, + "Failed to parse result value %s", value); + else if (f != SOCKET_SUCCESS) + s->result = f; + + } else if (streq(key, "n-accepted")) { + unsigned k; + + if (safe_atou(value, &k) < 0) + log_debug_unit(u->id, + "Failed to parse n-accepted value %s", value); + else + s->n_accepted += k; + } else if (streq(key, "control-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_debug_unit(u->id, + "Failed to parse control-pid value %s", value); + else + s->control_pid = pid; + } else if (streq(key, "control-command")) { + SocketExecCommand id; + + id = socket_exec_command_from_string(value); + if (id < 0) + log_debug_unit(u->id, + "Failed to parse exec-command value %s", value); + else { + s->control_command_id = id; + s->control_command = s->exec_command[id]; + } + } else if (streq(key, "fifo")) { + int fd, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %n", &fd, &skip) < 1 || fd < 0 || !fdset_contains(fds, fd)) + log_debug_unit(u->id, + "Failed to parse fifo value %s", value); + else { + + LIST_FOREACH(port, p, s->ports) + if (p->type == SOCKET_FIFO && + streq_ptr(p->path, value+skip)) + break; + + if (p) { + if (p->fd >= 0) + close_nointr_nofail(p->fd); + p->fd = fdset_remove(fds, fd); + } + } + + } else if (streq(key, "special")) { + int fd, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %n", &fd, &skip) < 1 || fd < 0 || !fdset_contains(fds, fd)) + log_debug_unit(u->id, + "Failed to parse special value %s", value); + else { + + LIST_FOREACH(port, p, s->ports) + if (p->type == SOCKET_SPECIAL && + streq_ptr(p->path, value+skip)) + break; + + if (p) { + if (p->fd >= 0) + close_nointr_nofail(p->fd); + p->fd = fdset_remove(fds, fd); + } + } + + } else if (streq(key, "mqueue")) { + int fd, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %n", &fd, &skip) < 1 || fd < 0 || !fdset_contains(fds, fd)) + log_debug_unit(u->id, + "Failed to parse mqueue value %s", value); + else { + + LIST_FOREACH(port, p, s->ports) + if (p->type == SOCKET_MQUEUE && + streq_ptr(p->path, value+skip)) + break; + + if (p) { + if (p->fd >= 0) + close_nointr_nofail(p->fd); + p->fd = fdset_remove(fds, fd); + } + } + + } else if (streq(key, "socket")) { + int fd, type, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %i %n", &fd, &type, &skip) < 2 || fd < 0 || type < 0 || !fdset_contains(fds, fd)) + log_debug_unit(u->id, + "Failed to parse socket value %s", value); + else { + + LIST_FOREACH(port, p, s->ports) + if (socket_address_is(&p->address, value+skip, type)) + break; + + if (p) { + if (p->fd >= 0) + close_nointr_nofail(p->fd); + p->fd = fdset_remove(fds, fd); + } + } + + } else if (streq(key, "netlink")) { + int fd, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %n", &fd, &skip) < 1 || fd < 0 || !fdset_contains(fds, fd)) + log_debug_unit(u->id, + "Failed to parse socket value %s", value); + else { + + LIST_FOREACH(port, p, s->ports) + if (socket_address_is_netlink(&p->address, value+skip)) + break; + + if (p) { + if (p->fd >= 0) + close_nointr_nofail(p->fd); + p->fd = fdset_remove(fds, fd); + } + } + } else if (streq(key, "tmp-dir")) { + char *t; + + t = strdup(value); + if (!t) + return log_oom(); + + s->exec_context.tmp_dir = t; + } else if (streq(key, "var-tmp-dir")) { + char *t; + + t = strdup(value); + if (!t) + return log_oom(); + + s->exec_context.var_tmp_dir = t; + } else + log_debug_unit(UNIT(s)->id, + "Unknown serialization key '%s'", key); + + return 0; +} + +static int socket_distribute_fds(Unit *u, FDSet *fds) { + Socket *s = SOCKET(u); + SocketPort *p; + + assert(u); + + LIST_FOREACH(port, p, s->ports) { + Iterator i; + int fd; + + if (p->type != SOCKET_SOCKET) + continue; + + if (p->fd >= 0) + continue; + + FDSET_FOREACH(fd, fds, i) { + if (socket_address_matches_fd(&p->address, fd)) { + p->fd = fdset_remove(fds, fd); + s->deserialized_state = SOCKET_LISTENING; + break; + } + } + } + + return 0; +} + +_pure_ static UnitActiveState socket_active_state(Unit *u) { + assert(u); + + return state_translation_table[SOCKET(u)->state]; +} + +_pure_ static const char *socket_sub_state_to_string(Unit *u) { + assert(u); + + return socket_state_to_string(SOCKET(u)->state); +} + +const char* socket_port_type_to_string(SocketPort *p) { + + assert(p); + + switch (p->type) { + case SOCKET_SOCKET: + switch (p->address.type) { + case SOCK_STREAM: return "Stream"; + case SOCK_DGRAM: return "Datagram"; + case SOCK_SEQPACKET: return "SequentialPacket"; + case SOCK_RAW: + if (socket_address_family(&p->address) == AF_NETLINK) + return "Netlink"; + default: return "Invalid"; + } + case SOCKET_SPECIAL: return "Special"; + case SOCKET_MQUEUE: return "MessageQueue"; + case SOCKET_FIFO: return "FIFO"; + default: return NULL; + } +} + +_pure_ static bool socket_check_gc(Unit *u) { + Socket *s = SOCKET(u); + + assert(u); + + return s->n_connections > 0; +} + +static void socket_fd_event(Unit *u, int fd, uint32_t events, Watch *w) { + Socket *s = SOCKET(u); + int cfd = -1; + + assert(s); + assert(fd >= 0); + + if (s->state != SOCKET_LISTENING) + return; + + log_debug_unit(u->id, "Incoming traffic on %s", u->id); + + if (events != EPOLLIN) { + + if (events & EPOLLHUP) + log_error_unit(u->id, + "%s: Got POLLHUP on a listening socket. The service probably invoked shutdown() on it, and should better not do that.", + u->id); + else + log_error_unit(u->id, + "%s: Got unexpected poll event (0x%x) on socket.", + u->id, events); + + goto fail; + } + + if (w->socket_accept) { + for (;;) { + + cfd = accept4(fd, NULL, NULL, SOCK_NONBLOCK); + if (cfd < 0) { + + if (errno == EINTR) + continue; + + log_error_unit(u->id, + "Failed to accept socket: %m"); + goto fail; + } + + break; + } + + socket_apply_socket_options(s, cfd); + } + + socket_enter_running(s, cfd); + return; + +fail: + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_sigchld_event(Unit *u, pid_t pid, int code, int status) { + Socket *s = SOCKET(u); + SocketResult f; + + assert(s); + assert(pid >= 0); + + if (pid != s->control_pid) + return; + + s->control_pid = 0; + + if (is_clean_exit(code, status, NULL)) + f = SOCKET_SUCCESS; + else if (code == CLD_EXITED) + f = SOCKET_FAILURE_EXIT_CODE; + else if (code == CLD_KILLED) + f = SOCKET_FAILURE_SIGNAL; + else if (code == CLD_DUMPED) + f = SOCKET_FAILURE_CORE_DUMP; + else + assert_not_reached("Unknown code"); + + if (s->control_command) { + exec_status_exit(&s->control_command->exec_status, &s->exec_context, pid, code, status); + + if (s->control_command->ignore) + f = SOCKET_SUCCESS; + } + + log_full_unit(f == SOCKET_SUCCESS ? LOG_DEBUG : LOG_NOTICE, + u->id, + "%s control process exited, code=%s status=%i", + u->id, sigchld_code_to_string(code), status); + + if (f != SOCKET_SUCCESS) + s->result = f; + + if (s->control_command && + s->control_command->command_next && + f == SOCKET_SUCCESS) { + + log_debug_unit(u->id, + "%s running next command for state %s", + u->id, socket_state_to_string(s->state)); + socket_run_next(s); + } else { + s->control_command = NULL; + s->control_command_id = _SOCKET_EXEC_COMMAND_INVALID; + + /* No further commands for this step, so let's figure + * out what to do next */ + + log_debug_unit(u->id, + "%s got final SIGCHLD for state %s", + u->id, socket_state_to_string(s->state)); + + switch (s->state) { + + case SOCKET_START_PRE: + if (f == SOCKET_SUCCESS) + socket_enter_start_post(s); + else + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, f); + break; + + case SOCKET_START_POST: + if (f == SOCKET_SUCCESS) + socket_enter_listening(s); + else + socket_enter_stop_pre(s, f); + break; + + case SOCKET_STOP_PRE: + case SOCKET_STOP_PRE_SIGTERM: + case SOCKET_STOP_PRE_SIGKILL: + socket_enter_stop_post(s, f); + break; + + case SOCKET_STOP_POST: + case SOCKET_FINAL_SIGTERM: + case SOCKET_FINAL_SIGKILL: + socket_enter_dead(s, f); + break; + + default: + assert_not_reached("Uh, control process died at wrong time."); + } + } + + /* Notify clients about changed exit status */ + unit_add_to_dbus_queue(u); +} + +static void socket_timer_event(Unit *u, uint64_t elapsed, Watch *w) { + Socket *s = SOCKET(u); + + assert(s); + assert(elapsed == 1); + assert(w == &s->timer_watch); + + switch (s->state) { + + case SOCKET_START_PRE: + log_warning_unit(u->id, + "%s starting timed out. Terminating.", u->id); + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_START_POST: + log_warning_unit(u->id, + "%s starting timed out. Stopping.", u->id); + socket_enter_stop_pre(s, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_STOP_PRE: + log_warning_unit(u->id, + "%s stopping timed out. Terminating.", u->id); + socket_enter_signal(s, SOCKET_STOP_PRE_SIGTERM, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_STOP_PRE_SIGTERM: + if (s->kill_context.send_sigkill) { + log_warning_unit(u->id, + "%s stopping timed out. Killing.", u->id); + socket_enter_signal(s, SOCKET_STOP_PRE_SIGKILL, SOCKET_FAILURE_TIMEOUT); + } else { + log_warning_unit(u->id, + "%s stopping timed out. Skipping SIGKILL. Ignoring.", + u->id); + socket_enter_stop_post(s, SOCKET_FAILURE_TIMEOUT); + } + break; + + case SOCKET_STOP_PRE_SIGKILL: + log_warning_unit(u->id, + "%s still around after SIGKILL. Ignoring.", u->id); + socket_enter_stop_post(s, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_STOP_POST: + log_warning_unit(u->id, + "%s stopping timed out (2). Terminating.", u->id); + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_FINAL_SIGTERM: + if (s->kill_context.send_sigkill) { + log_warning_unit(u->id, + "%s stopping timed out (2). Killing.", u->id); + socket_enter_signal(s, SOCKET_FINAL_SIGKILL, SOCKET_FAILURE_TIMEOUT); + } else { + log_warning_unit(u->id, + "%s stopping timed out (2). Skipping SIGKILL. Ignoring.", + u->id); + socket_enter_dead(s, SOCKET_FAILURE_TIMEOUT); + } + break; + + case SOCKET_FINAL_SIGKILL: + log_warning_unit(u->id, + "%s still around after SIGKILL (2). Entering failed mode.", + u->id); + socket_enter_dead(s, SOCKET_FAILURE_TIMEOUT); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } +} + +int socket_collect_fds(Socket *s, int **fds, unsigned *n_fds) { + int *rfds; + unsigned rn_fds, k; + SocketPort *p; + + assert(s); + assert(fds); + assert(n_fds); + + /* Called from the service code for requesting our fds */ + + rn_fds = 0; + LIST_FOREACH(port, p, s->ports) + if (p->fd >= 0) + rn_fds++; + + if (rn_fds <= 0) { + *fds = NULL; + *n_fds = 0; + return 0; + } + + if (!(rfds = new(int, rn_fds))) + return -ENOMEM; + + k = 0; + LIST_FOREACH(port, p, s->ports) + if (p->fd >= 0) + rfds[k++] = p->fd; + + assert(k == rn_fds); + + *fds = rfds; + *n_fds = rn_fds; + + return 0; +} + +void socket_notify_service_dead(Socket *s, bool failed_permanent) { + assert(s); + + /* The service is dead. Dang! + * + * This is strictly for one-instance-for-all-connections + * services. */ + + if (s->state == SOCKET_RUNNING) { + log_debug_unit(UNIT(s)->id, + "%s got notified about service death (failed permanently: %s)", + UNIT(s)->id, yes_no(failed_permanent)); + if (failed_permanent) + socket_enter_stop_pre(s, SOCKET_FAILURE_SERVICE_FAILED_PERMANENT); + else + socket_enter_listening(s); + } +} + +void socket_connection_unref(Socket *s) { + assert(s); + + /* The service is dead. Yay! + * + * This is strictly for one-instance-per-connection + * services. */ + + assert(s->n_connections > 0); + s->n_connections--; + + log_debug_unit(UNIT(s)->id, + "%s: One connection closed, %u left.", UNIT(s)->id, s->n_connections); +} + +static void socket_reset_failed(Unit *u) { + Socket *s = SOCKET(u); + + assert(s); + + if (s->state == SOCKET_FAILED) + socket_set_state(s, SOCKET_DEAD); + + s->result = SOCKET_SUCCESS; +} + +static int socket_kill(Unit *u, KillWho who, int signo, DBusError *error) { + return unit_kill_common(u, who, signo, -1, SOCKET(u)->control_pid, error); +} + +static const char* const socket_state_table[_SOCKET_STATE_MAX] = { + [SOCKET_DEAD] = "dead", + [SOCKET_START_PRE] = "start-pre", + [SOCKET_START_POST] = "start-post", + [SOCKET_LISTENING] = "listening", + [SOCKET_RUNNING] = "running", + [SOCKET_STOP_PRE] = "stop-pre", + [SOCKET_STOP_PRE_SIGTERM] = "stop-pre-sigterm", + [SOCKET_STOP_PRE_SIGKILL] = "stop-pre-sigkill", + [SOCKET_STOP_POST] = "stop-post", + [SOCKET_FINAL_SIGTERM] = "final-sigterm", + [SOCKET_FINAL_SIGKILL] = "final-sigkill", + [SOCKET_FAILED] = "failed" +}; + +DEFINE_STRING_TABLE_LOOKUP(socket_state, SocketState); + +static const char* const socket_exec_command_table[_SOCKET_EXEC_COMMAND_MAX] = { + [SOCKET_EXEC_START_PRE] = "StartPre", + [SOCKET_EXEC_START_POST] = "StartPost", + [SOCKET_EXEC_STOP_PRE] = "StopPre", + [SOCKET_EXEC_STOP_POST] = "StopPost" +}; + +DEFINE_STRING_TABLE_LOOKUP(socket_exec_command, SocketExecCommand); + +static const char* const socket_result_table[_SOCKET_RESULT_MAX] = { + [SOCKET_SUCCESS] = "success", + [SOCKET_FAILURE_RESOURCES] = "resources", + [SOCKET_FAILURE_TIMEOUT] = "timeout", + [SOCKET_FAILURE_EXIT_CODE] = "exit-code", + [SOCKET_FAILURE_SIGNAL] = "signal", + [SOCKET_FAILURE_CORE_DUMP] = "core-dump", + [SOCKET_FAILURE_SERVICE_FAILED_PERMANENT] = "service-failed-permanent" +}; + +DEFINE_STRING_TABLE_LOOKUP(socket_result, SocketResult); + +const UnitVTable socket_vtable = { + .object_size = sizeof(Socket), + + .sections = + "Unit\0" + "Socket\0" + "Install\0", + + .exec_context_offset = offsetof(Socket, exec_context), + .exec_section = "Socket", + + .init = socket_init, + .done = socket_done, + .load = socket_load, + + .kill = socket_kill, + + .coldplug = socket_coldplug, + + .dump = socket_dump, + + .start = socket_start, + .stop = socket_stop, + + .serialize = socket_serialize, + .deserialize_item = socket_deserialize_item, + .distribute_fds = socket_distribute_fds, + + .active_state = socket_active_state, + .sub_state_to_string = socket_sub_state_to_string, + + .check_gc = socket_check_gc, + + .fd_event = socket_fd_event, + .sigchld_event = socket_sigchld_event, + .timer_event = socket_timer_event, + + .reset_failed = socket_reset_failed, + + .bus_interface = "org.freedesktop.systemd1.Socket", + .bus_message_handler = bus_socket_message_handler, + .bus_invalidating_properties = bus_socket_invalidating_properties, + + .status_message_formats = { + /*.starting_stopping = { + [0] = "Starting socket %s...", + [1] = "Stopping socket %s...", + },*/ + .finished_start_job = { + [JOB_DONE] = "Listening on %s.", + [JOB_FAILED] = "Failed to listen on %s.", + [JOB_DEPENDENCY] = "Dependency failed for %s.", + [JOB_TIMEOUT] = "Timed out starting %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Closed %s.", + [JOB_FAILED] = "Failed stopping %s.", + [JOB_TIMEOUT] = "Timed out stopping %s.", + }, + }, +}; diff --git a/src/core/socket.h b/src/core/socket.h new file mode 100644 index 0000000000..9d48cde0a6 --- /dev/null +++ b/src/core/socket.h @@ -0,0 +1,180 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Socket Socket; + +#include "manager.h" +#include "unit.h" +#include "socket-util.h" +#include "mount.h" +#include "service.h" + +typedef enum SocketState { + SOCKET_DEAD, + SOCKET_START_PRE, + SOCKET_START_POST, + SOCKET_LISTENING, + SOCKET_RUNNING, + SOCKET_STOP_PRE, + SOCKET_STOP_PRE_SIGTERM, + SOCKET_STOP_PRE_SIGKILL, + SOCKET_STOP_POST, + SOCKET_FINAL_SIGTERM, + SOCKET_FINAL_SIGKILL, + SOCKET_FAILED, + _SOCKET_STATE_MAX, + _SOCKET_STATE_INVALID = -1 +} SocketState; + +typedef enum SocketExecCommand { + SOCKET_EXEC_START_PRE, + SOCKET_EXEC_START_POST, + SOCKET_EXEC_STOP_PRE, + SOCKET_EXEC_STOP_POST, + _SOCKET_EXEC_COMMAND_MAX, + _SOCKET_EXEC_COMMAND_INVALID = -1 +} SocketExecCommand; + +typedef enum SocketType { + SOCKET_SOCKET, + SOCKET_FIFO, + SOCKET_SPECIAL, + SOCKET_MQUEUE, + _SOCKET_FIFO_MAX, + _SOCKET_FIFO_INVALID = -1 +} SocketType; + +typedef enum SocketResult { + SOCKET_SUCCESS, + SOCKET_FAILURE_RESOURCES, + SOCKET_FAILURE_TIMEOUT, + SOCKET_FAILURE_EXIT_CODE, + SOCKET_FAILURE_SIGNAL, + SOCKET_FAILURE_CORE_DUMP, + SOCKET_FAILURE_SERVICE_FAILED_PERMANENT, + _SOCKET_RESULT_MAX, + _SOCKET_RESULT_INVALID = -1 +} SocketResult; + +typedef struct SocketPort { + SocketType type; + int fd; + + SocketAddress address; + char *path; + Watch fd_watch; + + LIST_FIELDS(struct SocketPort, port); +} SocketPort; + +struct Socket { + Unit meta; + + LIST_HEAD(SocketPort, ports); + + unsigned n_accepted; + unsigned n_connections; + unsigned max_connections; + + unsigned backlog; + usec_t timeout_usec; + + ExecCommand* exec_command[_SOCKET_EXEC_COMMAND_MAX]; + ExecContext exec_context; + KillContext kill_context; + + /* For Accept=no sockets refers to the one service we'll + activate. For Accept=yes sockets is either NULL, or filled + when the next service we spawn. */ + UnitRef service; + + SocketState state, deserialized_state; + + Watch timer_watch; + + ExecCommand* control_command; + SocketExecCommand control_command_id; + pid_t control_pid; + + mode_t directory_mode; + mode_t socket_mode; + + SocketResult result; + + bool accept; + + /* Socket options */ + bool keep_alive; + bool free_bind; + bool transparent; + bool broadcast; + bool pass_cred; + bool pass_sec; + + /* Only for INET6 sockets: issue IPV6_V6ONLY sockopt */ + SocketAddressBindIPv6Only bind_ipv6_only; + + int priority; + int mark; + size_t receive_buffer; + size_t send_buffer; + int ip_tos; + int ip_ttl; + size_t pipe_size; + char *bind_to_device; + char *tcp_congestion; + long mq_maxmsg; + long mq_msgsize; + + char *smack; + char *smack_ip_in; + char *smack_ip_out; +}; + +/* Called from the service code when collecting fds */ +int socket_collect_fds(Socket *s, int **fds, unsigned *n_fds); + +/* Called from the service when it shut down */ +void socket_notify_service_dead(Socket *s, bool failed_permanent); + +/* Called from the mount code figure out if a mount is a dependency of + * any of the sockets of this socket */ +int socket_add_one_mount_link(Socket *s, Mount *m); + +/* Called from the service code when a per-connection service ended */ +void socket_connection_unref(Socket *s); + +void socket_free_ports(Socket *s); + +extern const UnitVTable socket_vtable; + +const char* socket_state_to_string(SocketState i) _const_; +SocketState socket_state_from_string(const char *s) _pure_; + +const char* socket_exec_command_to_string(SocketExecCommand i) _const_; +SocketExecCommand socket_exec_command_from_string(const char *s) _pure_; + +const char* socket_result_to_string(SocketResult i) _const_; +SocketResult socket_result_from_string(const char *s) _pure_; + +const char* socket_port_type_to_string(SocketPort *p) _pure_; diff --git a/src/core/special.h b/src/core/special.h new file mode 100644 index 0000000000..a9b50bce05 --- /dev/null +++ b/src/core/special.h @@ -0,0 +1,115 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#define SPECIAL_DEFAULT_TARGET "default.target" + +/* Shutdown targets */ +#define SPECIAL_UMOUNT_TARGET "umount.target" +/* This is not really intended to be started by directly. This is + * mostly so that other targets (reboot/halt/poweroff) can depend on + * it to bring all services down that want to be brought down on + * system shutdown. */ +#define SPECIAL_SHUTDOWN_TARGET "shutdown.target" +#define SPECIAL_HALT_TARGET "halt.target" +#define SPECIAL_POWEROFF_TARGET "poweroff.target" +#define SPECIAL_REBOOT_TARGET "reboot.target" +#define SPECIAL_KEXEC_TARGET "kexec.target" +#define SPECIAL_EXIT_TARGET "exit.target" +#define SPECIAL_SUSPEND_TARGET "suspend.target" +#define SPECIAL_HIBERNATE_TARGET "hibernate.target" +#define SPECIAL_HYBRID_SLEEP_TARGET "hybrid-sleep.target" + +/* Special boot targets */ +#define SPECIAL_RESCUE_TARGET "rescue.target" +#define SPECIAL_EMERGENCY_TARGET "emergency.target" + +/* Early boot targets */ +#define SPECIAL_SYSINIT_TARGET "sysinit.target" +#define SPECIAL_SOCKETS_TARGET "sockets.target" +#define SPECIAL_TIMERS_TARGET "timers.target" +#define SPECIAL_PATHS_TARGET "paths.target" +#define SPECIAL_LOCAL_FS_TARGET "local-fs.target" +#define SPECIAL_LOCAL_FS_PRE_TARGET "local-fs-pre.target" +#define SPECIAL_INITRD_FS_TARGET "initrd-fs.target" +#define SPECIAL_INITRD_ROOT_FS_TARGET "initrd-root-fs.target" +#define SPECIAL_REMOTE_FS_TARGET "remote-fs.target" /* LSB's $remote_fs */ +#define SPECIAL_REMOTE_FS_PRE_TARGET "remote-fs-pre.target" +#define SPECIAL_SWAP_TARGET "swap.target" +#define SPECIAL_NETWORK_ONLINE_TARGET "network-online.target" +#define SPECIAL_BASIC_TARGET "basic.target" + +/* LSB compatibility */ +#define SPECIAL_NETWORK_TARGET "network.target" /* LSB's $network */ +#define SPECIAL_NSS_LOOKUP_TARGET "nss-lookup.target" /* LSB's $named */ +#define SPECIAL_RPCBIND_TARGET "rpcbind.target" /* LSB's $portmap */ +#define SPECIAL_TIME_SYNC_TARGET "time-sync.target" /* LSB's $time */ + +/* + * Rules regarding adding further high level targets like the above: + * + * - Be conservative, only add more of these when we really need + * them. We need strong usecases for further additions. + * + * - When there can be multiple implementations running side-by-side, + * it needs to be a .target unit which can pull in all + * implementations. + * + * - If something can be implemented with socket activation, and + * without, it needs to be a .target unit, so that it can pull in + * the appropriate unit. + * + * - Otherwise, it should be a .service unit. + * + * - In some cases it is OK to have both a .service and a .target + * unit, i.e. if there can be multiple parallel implementations, but + * only one is the "system" one. Example: syslog. + * + * Or to put this in other words: .service symlinks can be used to + * arbitrate between multiple implementations if there can be only one + * of a kind. .target units can be used to support multiple + * implementations that can run side-by-side. + */ + +/* Magic early boot services */ +#define SPECIAL_FSCK_SERVICE "systemd-fsck@.service" +#define SPECIAL_QUOTACHECK_SERVICE "systemd-quotacheck.service" +#define SPECIAL_QUOTAON_SERVICE "quotaon.service" +#define SPECIAL_REMOUNT_FS_SERVICE "systemd-remount-fs.service" + +/* Services systemd relies on */ +#define SPECIAL_DBUS_SERVICE "dbus.service" +#define SPECIAL_DBUS_SOCKET "dbus.socket" +#define SPECIAL_JOURNALD_SOCKET "systemd-journald.socket" +#define SPECIAL_JOURNALD_SERVICE "systemd-journald.service" + +/* Magic init signals */ +#define SPECIAL_KBREQUEST_TARGET "kbrequest.target" +#define SPECIAL_SIGPWR_TARGET "sigpwr.target" +#define SPECIAL_CTRL_ALT_DEL_TARGET "ctrl-alt-del.target" + +/* For SysV compatibility. Usually an alias for a saner target. On + * SysV-free systems this doesn't exist. */ +#define SPECIAL_RUNLEVEL2_TARGET "runlevel2.target" +#define SPECIAL_RUNLEVEL3_TARGET "runlevel3.target" +#define SPECIAL_RUNLEVEL4_TARGET "runlevel4.target" +#define SPECIAL_RUNLEVEL5_TARGET "runlevel5.target" diff --git a/src/core/swap.c b/src/core/swap.c new file mode 100644 index 0000000000..d503fe20df --- /dev/null +++ b/src/core/swap.c @@ -0,0 +1,1385 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <limits.h> +#include <unistd.h> +#include <fcntl.h> +#include <sys/epoll.h> +#include <sys/stat.h> +#include <sys/swap.h> +#include <libudev.h> + +#include "unit.h" +#include "swap.h" +#include "load-fragment.h" +#include "load-dropin.h" +#include "unit-name.h" +#include "dbus-swap.h" +#include "special.h" +#include "bus-errors.h" +#include "exit-status.h" +#include "def.h" +#include "path-util.h" +#include "virt.h" + +static const UnitActiveState state_translation_table[_SWAP_STATE_MAX] = { + [SWAP_DEAD] = UNIT_INACTIVE, + [SWAP_ACTIVATING] = UNIT_ACTIVATING, + [SWAP_ACTIVE] = UNIT_ACTIVE, + [SWAP_DEACTIVATING] = UNIT_DEACTIVATING, + [SWAP_ACTIVATING_SIGTERM] = UNIT_DEACTIVATING, + [SWAP_ACTIVATING_SIGKILL] = UNIT_DEACTIVATING, + [SWAP_DEACTIVATING_SIGTERM] = UNIT_DEACTIVATING, + [SWAP_DEACTIVATING_SIGKILL] = UNIT_DEACTIVATING, + [SWAP_FAILED] = UNIT_FAILED +}; + +static void swap_unset_proc_swaps(Swap *s) { + Swap *first; + Hashmap *swaps; + + assert(s); + + if (!s->parameters_proc_swaps.what) + return; + + /* Remove this unit from the chain of swaps which share the + * same kernel swap device. */ + swaps = UNIT(s)->manager->swaps_by_proc_swaps; + first = hashmap_get(swaps, s->parameters_proc_swaps.what); + LIST_REMOVE(Swap, same_proc_swaps, first, s); + + if (first) + hashmap_remove_and_replace(swaps, + s->parameters_proc_swaps.what, + first->parameters_proc_swaps.what, + first); + else + hashmap_remove(swaps, s->parameters_proc_swaps.what); + + free(s->parameters_proc_swaps.what); + s->parameters_proc_swaps.what = NULL; +} + +static void swap_init(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + assert(UNIT(s)->load_state == UNIT_STUB); + + s->timeout_usec = DEFAULT_TIMEOUT_USEC; + + exec_context_init(&s->exec_context); + s->exec_context.std_output = u->manager->default_std_output; + s->exec_context.std_error = u->manager->default_std_error; + kill_context_init(&s->kill_context); + + s->parameters_proc_swaps.priority = s->parameters_fragment.priority = -1; + + s->timer_watch.type = WATCH_INVALID; + + s->control_command_id = _SWAP_EXEC_COMMAND_INVALID; + + UNIT(s)->ignore_on_isolate = true; +} + +static void swap_unwatch_control_pid(Swap *s) { + assert(s); + + if (s->control_pid <= 0) + return; + + unit_unwatch_pid(UNIT(s), s->control_pid); + s->control_pid = 0; +} + +static void swap_done(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + swap_unset_proc_swaps(s); + + free(s->what); + s->what = NULL; + + free(s->parameters_fragment.what); + s->parameters_fragment.what = NULL; + + exec_context_done(&s->exec_context, manager_is_reloading_or_reexecuting(u->manager)); + exec_command_done_array(s->exec_command, _SWAP_EXEC_COMMAND_MAX); + s->control_command = NULL; + + swap_unwatch_control_pid(s); + + unit_unwatch_timer(u, &s->timer_watch); +} + +int swap_add_one_mount_link(Swap *s, Mount *m) { + int r; + + assert(s); + assert(m); + + if (UNIT(s)->load_state != UNIT_LOADED || + UNIT(m)->load_state != UNIT_LOADED) + return 0; + + if (is_device_path(s->what)) + return 0; + + if (!path_startswith(s->what, m->where)) + return 0; + + r = unit_add_two_dependencies(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, UNIT(m), true); + if (r < 0) + return r; + + return 0; +} + +static int swap_add_mount_links(Swap *s) { + Unit *other; + int r; + + assert(s); + + LIST_FOREACH(units_by_type, other, UNIT(s)->manager->units_by_type[UNIT_MOUNT]) + if ((r = swap_add_one_mount_link(s, MOUNT(other))) < 0) + return r; + + return 0; +} + +static int swap_add_device_links(Swap *s) { + SwapParameters *p; + + assert(s); + + if (!s->what) + return 0; + + if (s->from_fragment) + p = &s->parameters_fragment; + else + return 0; + + if (is_device_path(s->what)) + return unit_add_node_link(UNIT(s), s->what, + !p->noauto && p->nofail && + UNIT(s)->manager->running_as == SYSTEMD_SYSTEM); + else + /* File based swap devices need to be ordered after + * systemd-remount-fs.service, since they might need a + * writable file system. */ + return unit_add_dependency_by_name(UNIT(s), UNIT_AFTER, SPECIAL_REMOUNT_FS_SERVICE, NULL, true); +} + +static int swap_add_default_dependencies(Swap *s) { + int r; + + assert(s); + + if (UNIT(s)->manager->running_as != SYSTEMD_SYSTEM) + return 0; + + if (detect_container(NULL) > 0) + return 0; + + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_UMOUNT_TARGET, NULL, true); + if (r < 0) + return r; + + return 0; +} + +static int swap_verify(Swap *s) { + bool b; + _cleanup_free_ char *e = NULL; + + if (UNIT(s)->load_state != UNIT_LOADED) + return 0; + + e = unit_name_from_path(s->what, ".swap"); + if (e == NULL) + return log_oom(); + + b = unit_has_name(UNIT(s), e); + if (!b) { + log_error_unit(UNIT(s)->id, + "%s: Value of \"What\" and unit name do not match, not loading.", + UNIT(s)->id); + return -EINVAL; + } + + if (s->exec_context.pam_name && s->kill_context.kill_mode != KILL_CONTROL_GROUP) { + log_error_unit(UNIT(s)->id, + "%s has PAM enabled. Kill mode must be set to 'control-group'. Refusing to load.", + UNIT(s)->id); + return -EINVAL; + } + + return 0; +} + +static int swap_load(Unit *u) { + int r; + Swap *s = SWAP(u); + + assert(s); + assert(u->load_state == UNIT_STUB); + + /* Load a .swap file */ + r = unit_load_fragment_and_dropin_optional(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + r = unit_add_exec_dependencies(u, &s->exec_context); + if (r < 0) + return r; + + if (UNIT(s)->fragment_path) + s->from_fragment = true; + + if (!s->what) { + if (s->parameters_fragment.what) + s->what = strdup(s->parameters_fragment.what); + else if (s->parameters_proc_swaps.what) + s->what = strdup(s->parameters_proc_swaps.what); + else + s->what = unit_name_to_path(u->id); + + if (!s->what) + return -ENOMEM; + } + + path_kill_slashes(s->what); + + if (!UNIT(s)->description) + if ((r = unit_set_description(u, s->what)) < 0) + return r; + + r = swap_add_device_links(s); + if (r < 0) + return r; + + r = swap_add_mount_links(s); + if (r < 0) + return r; + + r = unit_add_default_cgroups(u); + if (r < 0) + return r; + + if (UNIT(s)->default_dependencies) { + r = swap_add_default_dependencies(s); + if (r < 0) + return r; + } + + r = unit_exec_context_defaults(u, &s->exec_context); + if (r < 0) + return r; + } + + return swap_verify(s); +} + +static int swap_add_one( + Manager *m, + const char *what, + const char *what_proc_swaps, + int priority, + bool noauto, + bool nofail, + bool set_flags) { + + Unit *u = NULL; + _cleanup_free_ char *e = NULL; + char *wp = NULL; + bool delete = false; + int r; + SwapParameters *p; + Swap *first; + + assert(m); + assert(what); + assert(what_proc_swaps); + + e = unit_name_from_path(what, ".swap"); + if (!e) + return log_oom(); + + u = manager_get_unit(m, e); + + if (u && + SWAP(u)->from_proc_swaps && + !path_equal(SWAP(u)->parameters_proc_swaps.what, what_proc_swaps)) + return -EEXIST; + + if (!u) { + delete = true; + + u = unit_new(m, sizeof(Swap)); + if (!u) + return log_oom(); + + r = unit_add_name(u, e); + if (r < 0) + goto fail; + + SWAP(u)->what = strdup(what); + if (!SWAP(u)->what) { + r = log_oom(); + goto fail; + } + + unit_add_to_load_queue(u); + } else + delete = false; + + p = &SWAP(u)->parameters_proc_swaps; + + if (!p->what) { + wp = strdup(what_proc_swaps); + if (!wp) { + r = log_oom(); + goto fail; + } + + if (!m->swaps_by_proc_swaps) { + m->swaps_by_proc_swaps = hashmap_new(string_hash_func, string_compare_func); + if (!m->swaps_by_proc_swaps) { + r = log_oom(); + goto fail; + } + } + + free(p->what); + p->what = wp; + + first = hashmap_get(m->swaps_by_proc_swaps, wp); + LIST_PREPEND(Swap, same_proc_swaps, first, SWAP(u)); + + r = hashmap_replace(m->swaps_by_proc_swaps, wp, first); + if (r < 0) + goto fail; + } + + if (set_flags) { + SWAP(u)->is_active = true; + SWAP(u)->just_activated = !SWAP(u)->from_proc_swaps; + } + + SWAP(u)->from_proc_swaps = true; + + p->priority = priority; + p->noauto = noauto; + p->nofail = nofail; + + unit_add_to_dbus_queue(u); + + return 0; + +fail: + log_warning_unit(e, "Failed to load swap unit: %s", strerror(-r)); + + free(wp); + + if (delete && u) + unit_free(u); + + return r; +} + +static int swap_process_new_swap(Manager *m, const char *device, int prio, bool set_flags) { + struct stat st; + int r = 0, k; + + assert(m); + + if (stat(device, &st) >= 0 && S_ISBLK(st.st_mode)) { + struct udev_device *d; + const char *dn; + struct udev_list_entry *item = NULL, *first = NULL; + + /* So this is a proper swap device. Create swap units + * for all names this swap device is known under */ + + d = udev_device_new_from_devnum(m->udev, 'b', st.st_rdev); + if (!d) + return log_oom(); + + dn = udev_device_get_devnode(d); + /* Skip dn==device, since that case will be handled below */ + if (dn && !streq(dn, device)) + r = swap_add_one(m, dn, device, prio, false, false, set_flags); + + /* Add additional units for all symlinks */ + first = udev_device_get_devlinks_list_entry(d); + udev_list_entry_foreach(item, first) { + const char *p; + + /* Don't bother with the /dev/block links */ + p = udev_list_entry_get_name(item); + + if (path_startswith(p, "/dev/block/")) + continue; + + if (stat(p, &st) >= 0) + if ((!S_ISBLK(st.st_mode)) || + st.st_rdev != udev_device_get_devnum(d)) + continue; + + k = swap_add_one(m, p, device, prio, false, false, set_flags); + if (k < 0) + r = k; + } + + udev_device_unref(d); + } + + k = swap_add_one(m, device, device, prio, false, false, set_flags); + if (k < 0) + r = k; + + return r; +} + +static void swap_set_state(Swap *s, SwapState state) { + SwapState old_state; + + assert(s); + + old_state = s->state; + s->state = state; + + if (state != SWAP_ACTIVATING && + state != SWAP_ACTIVATING_SIGTERM && + state != SWAP_ACTIVATING_SIGKILL && + state != SWAP_DEACTIVATING && + state != SWAP_DEACTIVATING_SIGTERM && + state != SWAP_DEACTIVATING_SIGKILL) { + unit_unwatch_timer(UNIT(s), &s->timer_watch); + swap_unwatch_control_pid(s); + s->control_command = NULL; + s->control_command_id = _SWAP_EXEC_COMMAND_INVALID; + } + + if (state != old_state) + log_debug_unit(UNIT(s)->id, + "%s changed %s -> %s", + UNIT(s)->id, + swap_state_to_string(old_state), + swap_state_to_string(state)); + + unit_notify(UNIT(s), state_translation_table[old_state], + state_translation_table[state], true); +} + +static int swap_coldplug(Unit *u) { + Swap *s = SWAP(u); + SwapState new_state = SWAP_DEAD; + int r; + + assert(s); + assert(s->state == SWAP_DEAD); + + if (s->deserialized_state != s->state) + new_state = s->deserialized_state; + else if (s->from_proc_swaps) + new_state = SWAP_ACTIVE; + + if (new_state != s->state) { + + if (new_state == SWAP_ACTIVATING || + new_state == SWAP_ACTIVATING_SIGTERM || + new_state == SWAP_ACTIVATING_SIGKILL || + new_state == SWAP_DEACTIVATING || + new_state == SWAP_DEACTIVATING_SIGTERM || + new_state == SWAP_DEACTIVATING_SIGKILL) { + + if (s->control_pid <= 0) + return -EBADMSG; + + r = unit_watch_pid(UNIT(s), s->control_pid); + if (r < 0) + return r; + + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, s->timeout_usec, &s->timer_watch); + if (r < 0) + return r; + } + + swap_set_state(s, new_state); + } + + return 0; +} + +static void swap_dump(Unit *u, FILE *f, const char *prefix) { + Swap *s = SWAP(u); + SwapParameters *p; + + assert(s); + assert(f); + + if (s->from_proc_swaps) + p = &s->parameters_proc_swaps; + else if (s->from_fragment) + p = &s->parameters_fragment; + else + p = NULL; + + fprintf(f, + "%sSwap State: %s\n" + "%sResult: %s\n" + "%sWhat: %s\n" + "%sFrom /proc/swaps: %s\n" + "%sFrom fragment: %s\n", + prefix, swap_state_to_string(s->state), + prefix, swap_result_to_string(s->result), + prefix, s->what, + prefix, yes_no(s->from_proc_swaps), + prefix, yes_no(s->from_fragment)); + + if (p) + fprintf(f, + "%sPriority: %i\n" + "%sNoAuto: %s\n" + "%sNoFail: %s\n", + prefix, p->priority, + prefix, yes_no(p->noauto), + prefix, yes_no(p->nofail)); + + if (s->control_pid > 0) + fprintf(f, + "%sControl PID: %lu\n", + prefix, (unsigned long) s->control_pid); + + exec_context_dump(&s->exec_context, f, prefix); + kill_context_dump(&s->kill_context, f, prefix); +} + +static int swap_spawn(Swap *s, ExecCommand *c, pid_t *_pid) { + pid_t pid; + int r; + + assert(s); + assert(c); + assert(_pid); + + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, s->timeout_usec, &s->timer_watch); + if (r < 0) + goto fail; + + r = exec_spawn(c, + NULL, + &s->exec_context, + NULL, 0, + UNIT(s)->manager->environment, + true, + true, + true, + UNIT(s)->manager->confirm_spawn, + UNIT(s)->cgroup_bondings, + UNIT(s)->cgroup_attributes, + NULL, + UNIT(s)->id, + NULL, + &pid); + if (r < 0) + goto fail; + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) + /* FIXME: we need to do something here */ + goto fail; + + *_pid = pid; + + return 0; + +fail: + unit_unwatch_timer(UNIT(s), &s->timer_watch); + + return r; +} + +static void swap_enter_dead(Swap *s, SwapResult f) { + assert(s); + + if (f != SWAP_SUCCESS) + s->result = f; + + exec_context_tmp_dirs_done(&s->exec_context); + swap_set_state(s, s->result != SWAP_SUCCESS ? SWAP_FAILED : SWAP_DEAD); +} + +static void swap_enter_active(Swap *s, SwapResult f) { + assert(s); + + if (f != SWAP_SUCCESS) + s->result = f; + + swap_set_state(s, SWAP_ACTIVE); +} + +static void swap_enter_signal(Swap *s, SwapState state, SwapResult f) { + int r; + + assert(s); + + if (f != SWAP_SUCCESS) + s->result = f; + + r = unit_kill_context( + UNIT(s), + &s->kill_context, + state != SWAP_ACTIVATING_SIGTERM && state != SWAP_DEACTIVATING_SIGTERM, + -1, + s->control_pid, + false); + if (r < 0) + goto fail; + + if (r > 0) { + r = unit_watch_timer(UNIT(s), CLOCK_MONOTONIC, true, s->timeout_usec, &s->timer_watch); + if (r < 0) + goto fail; + + swap_set_state(s, state); + } else + swap_enter_dead(s, SWAP_SUCCESS); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to kill processes: %s", UNIT(s)->id, strerror(-r)); + + swap_enter_dead(s, SWAP_FAILURE_RESOURCES); +} + +static void swap_enter_activating(Swap *s) { + int r, priority; + + assert(s); + + s->control_command_id = SWAP_EXEC_ACTIVATE; + s->control_command = s->exec_command + SWAP_EXEC_ACTIVATE; + + if (s->from_fragment) + priority = s->parameters_fragment.priority; + else + priority = -1; + + if (priority >= 0) { + char p[LINE_MAX]; + + snprintf(p, sizeof(p), "%i", priority); + char_array_0(p); + + r = exec_command_set( + s->control_command, + "/sbin/swapon", + "-p", + p, + s->what, + NULL); + } else + r = exec_command_set( + s->control_command, + "/sbin/swapon", + s->what, + NULL); + + if (r < 0) + goto fail; + + swap_unwatch_control_pid(s); + + r = swap_spawn(s, s->control_command, &s->control_pid); + if (r < 0) + goto fail; + + swap_set_state(s, SWAP_ACTIVATING); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'swapon' task: %s", + UNIT(s)->id, strerror(-r)); + swap_enter_dead(s, SWAP_FAILURE_RESOURCES); +} + +static void swap_enter_deactivating(Swap *s) { + int r; + + assert(s); + + s->control_command_id = SWAP_EXEC_DEACTIVATE; + s->control_command = s->exec_command + SWAP_EXEC_DEACTIVATE; + + r = exec_command_set(s->control_command, + "/sbin/swapoff", + s->what, + NULL); + if (r < 0) + goto fail; + + swap_unwatch_control_pid(s); + + r = swap_spawn(s, s->control_command, &s->control_pid); + if (r < 0) + goto fail; + + swap_set_state(s, SWAP_DEACTIVATING); + + return; + +fail: + log_warning_unit(UNIT(s)->id, + "%s failed to run 'swapoff' task: %s", + UNIT(s)->id, strerror(-r)); + swap_enter_active(s, SWAP_FAILURE_RESOURCES); +} + +static int swap_start(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + /* We cannot fulfill this request right now, try again later + * please! */ + + if (s->state == SWAP_DEACTIVATING || + s->state == SWAP_DEACTIVATING_SIGTERM || + s->state == SWAP_DEACTIVATING_SIGKILL || + s->state == SWAP_ACTIVATING_SIGTERM || + s->state == SWAP_ACTIVATING_SIGKILL) + return -EAGAIN; + + if (s->state == SWAP_ACTIVATING) + return 0; + + assert(s->state == SWAP_DEAD || s->state == SWAP_FAILED); + + if (detect_container(NULL) > 0) + return -EPERM; + + s->result = SWAP_SUCCESS; + swap_enter_activating(s); + return 0; +} + +static int swap_stop(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + if (s->state == SWAP_DEACTIVATING || + s->state == SWAP_DEACTIVATING_SIGTERM || + s->state == SWAP_DEACTIVATING_SIGKILL || + s->state == SWAP_ACTIVATING_SIGTERM || + s->state == SWAP_ACTIVATING_SIGKILL) + return 0; + + assert(s->state == SWAP_ACTIVATING || + s->state == SWAP_ACTIVE); + + if (detect_container(NULL) > 0) + return -EPERM; + + swap_enter_deactivating(s); + return 0; +} + +static int swap_serialize(Unit *u, FILE *f, FDSet *fds) { + Swap *s = SWAP(u); + + assert(s); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", swap_state_to_string(s->state)); + unit_serialize_item(u, f, "result", swap_result_to_string(s->result)); + + if (s->control_pid > 0) + unit_serialize_item_format(u, f, "control-pid", "%lu", (unsigned long) s->control_pid); + + if (s->control_command_id >= 0) + unit_serialize_item(u, f, "control-command", swap_exec_command_to_string(s->control_command_id)); + + exec_context_serialize(&s->exec_context, UNIT(s), f); + + return 0; +} + +static int swap_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Swap *s = SWAP(u); + + assert(s); + assert(fds); + + if (streq(key, "state")) { + SwapState state; + + state = swap_state_from_string(value); + if (state < 0) + log_debug_unit(u->id, "Failed to parse state value %s", value); + else + s->deserialized_state = state; + } else if (streq(key, "result")) { + SwapResult f; + + f = swap_result_from_string(value); + if (f < 0) + log_debug_unit(u->id, "Failed to parse result value %s", value); + else if (f != SWAP_SUCCESS) + s->result = f; + } else if (streq(key, "control-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_debug_unit(u->id, "Failed to parse control-pid value %s", value); + else + s->control_pid = pid; + + } else if (streq(key, "control-command")) { + SwapExecCommand id; + + id = swap_exec_command_from_string(value); + if (id < 0) + log_debug_unit(u->id, "Failed to parse exec-command value %s", value); + else { + s->control_command_id = id; + s->control_command = s->exec_command + id; + } + } else if (streq(key, "tmp-dir")) { + char *t; + + t = strdup(value); + if (!t) + return log_oom(); + + s->exec_context.tmp_dir = t; + } else if (streq(key, "var-tmp-dir")) { + char *t; + + t = strdup(value); + if (!t) + return log_oom(); + + s->exec_context.var_tmp_dir = t; + } else + log_debug_unit(u->id, "Unknown serialization key '%s'", key); + + return 0; +} + +_pure_ static UnitActiveState swap_active_state(Unit *u) { + assert(u); + + return state_translation_table[SWAP(u)->state]; +} + +_pure_ static const char *swap_sub_state_to_string(Unit *u) { + assert(u); + + return swap_state_to_string(SWAP(u)->state); +} + +_pure_ static bool swap_check_gc(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + return s->from_proc_swaps; +} + +static void swap_sigchld_event(Unit *u, pid_t pid, int code, int status) { + Swap *s = SWAP(u); + SwapResult f; + + assert(s); + assert(pid >= 0); + + if (pid != s->control_pid) + return; + + s->control_pid = 0; + + if (is_clean_exit(code, status, NULL)) + f = SWAP_SUCCESS; + else if (code == CLD_EXITED) + f = SWAP_FAILURE_EXIT_CODE; + else if (code == CLD_KILLED) + f = SWAP_FAILURE_SIGNAL; + else if (code == CLD_DUMPED) + f = SWAP_FAILURE_CORE_DUMP; + else + assert_not_reached("Unknown code"); + + if (f != SWAP_SUCCESS) + s->result = f; + + if (s->control_command) { + exec_status_exit(&s->control_command->exec_status, &s->exec_context, pid, code, status); + + s->control_command = NULL; + s->control_command_id = _SWAP_EXEC_COMMAND_INVALID; + } + + log_full_unit(f == SWAP_SUCCESS ? LOG_DEBUG : LOG_NOTICE, + u->id, + "%s swap process exited, code=%s status=%i", + u->id, sigchld_code_to_string(code), status); + + switch (s->state) { + + case SWAP_ACTIVATING: + case SWAP_ACTIVATING_SIGTERM: + case SWAP_ACTIVATING_SIGKILL: + + if (f == SWAP_SUCCESS) + swap_enter_active(s, f); + else + swap_enter_dead(s, f); + break; + + case SWAP_DEACTIVATING: + case SWAP_DEACTIVATING_SIGKILL: + case SWAP_DEACTIVATING_SIGTERM: + + if (f == SWAP_SUCCESS) + swap_enter_dead(s, f); + else + swap_enter_dead(s, f); + break; + + default: + assert_not_reached("Uh, control process died at wrong time."); + } + + /* Notify clients about changed exit status */ + unit_add_to_dbus_queue(u); + + /* Request a reload of /proc/swaps, so that following units + * can follow our state change */ + u->manager->request_reload = true; +} + +static void swap_timer_event(Unit *u, uint64_t elapsed, Watch *w) { + Swap *s = SWAP(u); + + assert(s); + assert(elapsed == 1); + assert(w == &s->timer_watch); + + switch (s->state) { + + case SWAP_ACTIVATING: + log_warning_unit(u->id, "%s activation timed out. Stopping.", u->id); + swap_enter_signal(s, SWAP_ACTIVATING_SIGTERM, SWAP_FAILURE_TIMEOUT); + break; + + case SWAP_DEACTIVATING: + log_warning_unit(u->id, "%s deactivation timed out. Stopping.", u->id); + swap_enter_signal(s, SWAP_DEACTIVATING_SIGTERM, SWAP_FAILURE_TIMEOUT); + break; + + case SWAP_ACTIVATING_SIGTERM: + if (s->kill_context.send_sigkill) { + log_warning_unit(u->id, "%s activation timed out. Killing.", u->id); + swap_enter_signal(s, SWAP_ACTIVATING_SIGKILL, SWAP_FAILURE_TIMEOUT); + } else { + log_warning_unit(u->id, "%s activation timed out. Skipping SIGKILL. Ignoring.", u->id); + swap_enter_dead(s, SWAP_FAILURE_TIMEOUT); + } + break; + + case SWAP_DEACTIVATING_SIGTERM: + if (s->kill_context.send_sigkill) { + log_warning_unit(u->id, "%s deactivation timed out. Killing.", u->id); + swap_enter_signal(s, SWAP_DEACTIVATING_SIGKILL, SWAP_FAILURE_TIMEOUT); + } else { + log_warning_unit(u->id, "%s deactivation timed out. Skipping SIGKILL. Ignoring.", u->id); + swap_enter_dead(s, SWAP_FAILURE_TIMEOUT); + } + break; + + case SWAP_ACTIVATING_SIGKILL: + case SWAP_DEACTIVATING_SIGKILL: + log_warning_unit(u->id, "%s swap process still around after SIGKILL. Ignoring.", u->id); + swap_enter_dead(s, SWAP_FAILURE_TIMEOUT); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } +} + +static int swap_load_proc_swaps(Manager *m, bool set_flags) { + unsigned i; + int r = 0; + + assert(m); + + rewind(m->proc_swaps); + + (void) fscanf(m->proc_swaps, "%*s %*s %*s %*s %*s\n"); + + for (i = 1;; i++) { + char *dev = NULL, *d; + int prio = 0, k; + + k = fscanf(m->proc_swaps, + "%ms " /* device/file */ + "%*s " /* type of swap */ + "%*s " /* swap size */ + "%*s " /* used */ + "%i\n", /* priority */ + &dev, &prio); + if (k != 2) { + if (k == EOF) + break; + + log_warning("Failed to parse /proc/swaps:%u", i); + free(dev); + continue; + } + + d = cunescape(dev); + free(dev); + + if (!d) + return -ENOMEM; + + k = swap_process_new_swap(m, d, prio, set_flags); + free(d); + + if (k < 0) + r = k; + } + + return r; +} + +int swap_dispatch_reload(Manager *m) { + /* This function should go as soon as the kernel properly notifies us */ + + if (_likely_(!m->request_reload)) + return 0; + + m->request_reload = false; + + return swap_fd_event(m, EPOLLPRI); +} + +int swap_fd_event(Manager *m, int events) { + Unit *u; + int r; + + assert(m); + assert(events & EPOLLPRI); + + r = swap_load_proc_swaps(m, true); + if (r < 0) { + log_error("Failed to reread /proc/swaps: %s", strerror(-r)); + + /* Reset flags, just in case, for late calls */ + LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_SWAP]) { + Swap *swap = SWAP(u); + + swap->is_active = swap->just_activated = false; + } + + return 0; + } + + manager_dispatch_load_queue(m); + + LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_SWAP]) { + Swap *swap = SWAP(u); + + if (!swap->is_active) { + /* This has just been deactivated */ + + swap->from_proc_swaps = false; + swap_unset_proc_swaps(swap); + + switch (swap->state) { + + case SWAP_ACTIVE: + swap_enter_dead(swap, SWAP_SUCCESS); + break; + + default: + swap_set_state(swap, swap->state); + break; + } + + } else if (swap->just_activated) { + + /* New swap entry */ + + switch (swap->state) { + + case SWAP_DEAD: + case SWAP_FAILED: + swap_enter_active(swap, SWAP_SUCCESS); + break; + + default: + /* Nothing really changed, but let's + * issue an notification call + * nonetheless, in case somebody is + * waiting for this. */ + swap_set_state(swap, swap->state); + break; + } + } + + /* Reset the flags for later calls */ + swap->is_active = swap->just_activated = false; + } + + return 1; +} + +static Unit *swap_following(Unit *u) { + Swap *s = SWAP(u); + Swap *other, *first = NULL; + + assert(s); + + if (streq_ptr(s->what, s->parameters_proc_swaps.what)) + return NULL; + + /* Make everybody follow the unit that's named after the swap + * device in the kernel */ + + LIST_FOREACH_AFTER(same_proc_swaps, other, s) + if (streq_ptr(other->what, other->parameters_proc_swaps.what)) + return UNIT(other); + + LIST_FOREACH_BEFORE(same_proc_swaps, other, s) { + if (streq_ptr(other->what, other->parameters_proc_swaps.what)) + return UNIT(other); + + first = other; + } + + return UNIT(first); +} + +static int swap_following_set(Unit *u, Set **_set) { + Swap *s = SWAP(u); + Swap *other; + Set *set; + int r; + + assert(s); + assert(_set); + + if (LIST_JUST_US(same_proc_swaps, s)) { + *_set = NULL; + return 0; + } + + if (!(set = set_new(NULL, NULL))) + return -ENOMEM; + + LIST_FOREACH_AFTER(same_proc_swaps, other, s) + if ((r = set_put(set, other)) < 0) + goto fail; + + LIST_FOREACH_BEFORE(same_proc_swaps, other, s) + if ((r = set_put(set, other)) < 0) + goto fail; + + *_set = set; + return 1; + +fail: + set_free(set); + return r; +} + +static void swap_shutdown(Manager *m) { + assert(m); + + if (m->proc_swaps) { + fclose(m->proc_swaps); + m->proc_swaps = NULL; + } + + hashmap_free(m->swaps_by_proc_swaps); + m->swaps_by_proc_swaps = NULL; +} + +static int swap_enumerate(Manager *m) { + int r; + assert(m); + + if (!m->proc_swaps) { + struct epoll_event ev = { + .events = EPOLLPRI, + .data.ptr = &m->swap_watch, + }; + + m->proc_swaps = fopen("/proc/swaps", "re"); + if (!m->proc_swaps) + return (errno == ENOENT) ? 0 : -errno; + + m->swap_watch.type = WATCH_SWAP; + m->swap_watch.fd = fileno(m->proc_swaps); + + if (epoll_ctl(m->epoll_fd, EPOLL_CTL_ADD, m->swap_watch.fd, &ev) < 0) + return -errno; + } + + r = swap_load_proc_swaps(m, false); + if (r < 0) + swap_shutdown(m); + + return r; +} + +static void swap_reset_failed(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + if (s->state == SWAP_FAILED) + swap_set_state(s, SWAP_DEAD); + + s->result = SWAP_SUCCESS; +} + +static int swap_kill(Unit *u, KillWho who, int signo, DBusError *error) { + return unit_kill_common(u, who, signo, -1, SWAP(u)->control_pid, error); +} + +static const char* const swap_state_table[_SWAP_STATE_MAX] = { + [SWAP_DEAD] = "dead", + [SWAP_ACTIVATING] = "activating", + [SWAP_ACTIVE] = "active", + [SWAP_DEACTIVATING] = "deactivating", + [SWAP_ACTIVATING_SIGTERM] = "activating-sigterm", + [SWAP_ACTIVATING_SIGKILL] = "activating-sigkill", + [SWAP_DEACTIVATING_SIGTERM] = "deactivating-sigterm", + [SWAP_DEACTIVATING_SIGKILL] = "deactivating-sigkill", + [SWAP_FAILED] = "failed" +}; + +DEFINE_STRING_TABLE_LOOKUP(swap_state, SwapState); + +static const char* const swap_exec_command_table[_SWAP_EXEC_COMMAND_MAX] = { + [SWAP_EXEC_ACTIVATE] = "ExecActivate", + [SWAP_EXEC_DEACTIVATE] = "ExecDeactivate", +}; + +DEFINE_STRING_TABLE_LOOKUP(swap_exec_command, SwapExecCommand); + +static const char* const swap_result_table[_SWAP_RESULT_MAX] = { + [SWAP_SUCCESS] = "success", + [SWAP_FAILURE_RESOURCES] = "resources", + [SWAP_FAILURE_TIMEOUT] = "timeout", + [SWAP_FAILURE_EXIT_CODE] = "exit-code", + [SWAP_FAILURE_SIGNAL] = "signal", + [SWAP_FAILURE_CORE_DUMP] = "core-dump" +}; + +DEFINE_STRING_TABLE_LOOKUP(swap_result, SwapResult); + +const UnitVTable swap_vtable = { + .object_size = sizeof(Swap), + + .sections = + "Unit\0" + "Swap\0" + "Install\0", + + .exec_context_offset = offsetof(Swap, exec_context), + .exec_section = "Swap", + + .no_alias = true, + .no_instances = true, + + .init = swap_init, + .load = swap_load, + .done = swap_done, + + .coldplug = swap_coldplug, + + .dump = swap_dump, + + .start = swap_start, + .stop = swap_stop, + + .kill = swap_kill, + + .serialize = swap_serialize, + .deserialize_item = swap_deserialize_item, + + .active_state = swap_active_state, + .sub_state_to_string = swap_sub_state_to_string, + + .check_gc = swap_check_gc, + + .sigchld_event = swap_sigchld_event, + .timer_event = swap_timer_event, + + .reset_failed = swap_reset_failed, + + .bus_interface = "org.freedesktop.systemd1.Swap", + .bus_message_handler = bus_swap_message_handler, + .bus_invalidating_properties = bus_swap_invalidating_properties, + + .following = swap_following, + .following_set = swap_following_set, + + .enumerate = swap_enumerate, + .shutdown = swap_shutdown, + + .status_message_formats = { + .starting_stopping = { + [0] = "Activating swap %s...", + [1] = "Deactivating swap %s...", + }, + .finished_start_job = { + [JOB_DONE] = "Activated swap %s.", + [JOB_FAILED] = "Failed to activate swap %s.", + [JOB_DEPENDENCY] = "Dependency failed for %s.", + [JOB_TIMEOUT] = "Timed out activating swap %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Deactivated swap %s.", + [JOB_FAILED] = "Failed deactivating swap %s.", + [JOB_TIMEOUT] = "Timed out deactivating swap %s.", + }, + }, +}; diff --git a/src/core/swap.h b/src/core/swap.h new file mode 100644 index 0000000000..121889d1d5 --- /dev/null +++ b/src/core/swap.h @@ -0,0 +1,121 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2010 Maarten Lankhorst + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Swap Swap; + +#include "unit.h" + +typedef enum SwapState { + SWAP_DEAD, + SWAP_ACTIVATING, + SWAP_ACTIVE, + SWAP_DEACTIVATING, + SWAP_ACTIVATING_SIGTERM, + SWAP_ACTIVATING_SIGKILL, + SWAP_DEACTIVATING_SIGTERM, + SWAP_DEACTIVATING_SIGKILL, + SWAP_FAILED, + _SWAP_STATE_MAX, + _SWAP_STATE_INVALID = -1 +} SwapState; + +typedef enum SwapExecCommand { + SWAP_EXEC_ACTIVATE, + SWAP_EXEC_DEACTIVATE, + _SWAP_EXEC_COMMAND_MAX, + _SWAP_EXEC_COMMAND_INVALID = -1 +} SwapExecCommand; + +typedef struct SwapParameters { + char *what; + int priority; + bool noauto:1; + bool nofail:1; +} SwapParameters; + +typedef enum SwapResult { + SWAP_SUCCESS, + SWAP_FAILURE_RESOURCES, + SWAP_FAILURE_TIMEOUT, + SWAP_FAILURE_EXIT_CODE, + SWAP_FAILURE_SIGNAL, + SWAP_FAILURE_CORE_DUMP, + _SWAP_RESULT_MAX, + _SWAP_RESULT_INVALID = -1 +} SwapResult; + +struct Swap { + Unit meta; + + char *what; + + SwapParameters parameters_proc_swaps; + SwapParameters parameters_fragment; + + bool from_proc_swaps:1; + bool from_fragment:1; + + /* Used while looking for swaps that vanished or got added + * from/to /proc/swaps */ + bool is_active:1; + bool just_activated:1; + + SwapResult result; + + usec_t timeout_usec; + + ExecCommand exec_command[_SWAP_EXEC_COMMAND_MAX]; + ExecContext exec_context; + KillContext kill_context; + + SwapState state, deserialized_state; + + ExecCommand* control_command; + SwapExecCommand control_command_id; + pid_t control_pid; + + Watch timer_watch; + + /* In order to be able to distinguish dependencies on + different device nodes we might end up creating multiple + devices for the same swap. We chain them up here. */ + + LIST_FIELDS(struct Swap, same_proc_swaps); +}; + +extern const UnitVTable swap_vtable; + +int swap_add_one_mount_link(Swap *s, Mount *m); + +int swap_dispatch_reload(Manager *m); +int swap_fd_event(Manager *m, int events); + +const char* swap_state_to_string(SwapState i) _const_; +SwapState swap_state_from_string(const char *s) _pure_; + +const char* swap_exec_command_to_string(SwapExecCommand i) _const_; +SwapExecCommand swap_exec_command_from_string(const char *s) _pure_; + +const char* swap_result_to_string(SwapResult i) _const_; +SwapResult swap_result_from_string(const char *s) _pure_; diff --git a/src/core/switch-root.c b/src/core/switch-root.c new file mode 100644 index 0000000000..ce0e41d510 --- /dev/null +++ b/src/core/switch-root.c @@ -0,0 +1,167 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2012 Harald Hoyer, Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/stat.h> +#include <stdbool.h> +#include <errno.h> +#include <string.h> +#include <sys/mount.h> +#include <unistd.h> +#include <fcntl.h> + +#include "util.h" +#include "path-util.h" +#include "switch-root.h" +#include "missing.h" + +int switch_root(const char *new_root) { + + /* Don't try to unmount/move the old "/", there's no way to do it. */ + static const char move_mounts[] = + "/dev\0" + "/proc\0" + "/sys\0" + "/run\0"; + + int r, old_root_fd = -1; + struct stat new_root_stat; + bool old_root_remove; + const char *i; + _cleanup_free_ char *temporary_old_root = NULL; + + if (path_equal(new_root, "/")) + return 0; + + /* When using pivot_root() we assume that /mnt exists as place + * we can temporarily move the old root to. As we immediately + * unmount it from there it doesn't matter much which + * directory we choose for this, but it should be more likely + * than not that /mnt exists and is suitable as mount point + * and is on the same fs as the old root dir */ + temporary_old_root = strappend(new_root, "/mnt"); + if (!temporary_old_root) + return -ENOMEM; + + old_root_remove = in_initrd(); + + if (stat(new_root, &new_root_stat) < 0) { + r = -errno; + log_error("Failed to stat directory %s: %m", new_root); + goto fail; + } + + /* Work-around for a kernel bug: for some reason the kernel + * refuses switching root if any file systems are mounted + * MS_SHARED. Hence remount them MS_PRIVATE here as a + * work-around. + * + * https://bugzilla.redhat.com/show_bug.cgi?id=847418 */ + if (mount(NULL, "/", NULL, MS_REC|MS_PRIVATE, NULL) < 0) + log_warning("Failed to make \"/\" private mount: %m"); + + NULSTR_FOREACH(i, move_mounts) { + char new_mount[PATH_MAX]; + struct stat sb; + + snprintf(new_mount, sizeof(new_mount), "%s%s", new_root, i); + char_array_0(new_mount); + + if ((stat(new_mount, &sb) < 0) || + sb.st_dev != new_root_stat.st_dev) { + + /* Mount point seems to be mounted already or + * stat failed. Unmount the old mount + * point. */ + if (umount2(i, MNT_DETACH) < 0) + log_warning("Failed to unmount %s: %m", i); + continue; + } + + if (mount(i, new_mount, NULL, MS_MOVE, NULL) < 0) { + log_error("Failed to move mount %s to %s, forcing unmount: %m", i, new_mount); + + if (umount2(i, MNT_FORCE) < 0) + log_warning("Failed to unmount %s: %m", i); + } + } + + if (chdir(new_root) < 0) { + r = -errno; + log_error("Failed to change directory to %s: %m", new_root); + goto fail; + } + + if (old_root_remove) { + old_root_fd = open("/", O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_NOCTTY|O_DIRECTORY); + if (old_root_fd < 0) + log_warning("Failed to open root directory: %m"); + } + + /* We first try a pivot_root() so that we can umount the old + * root dir. In many cases (i.e. where rootfs is /), that's + * not possible however, and hence we simply overmount root */ + if (pivot_root(new_root, temporary_old_root) >= 0) { + + /* Immediately get rid of the old root. Since we are + * running off it we need to do this lazily. */ + if (umount2(temporary_old_root, MNT_DETACH) < 0) { + r = -errno; + log_error("Failed to umount old root dir %s: %m", temporary_old_root); + goto fail; + } + + } else if (mount(new_root, "/", NULL, MS_MOVE, NULL) < 0) { + r = -errno; + log_error("Failed to mount moving %s to /: %m", new_root); + goto fail; + } + + if (chroot(".") < 0) { + r = -errno; + log_error("Failed to change root: %m"); + goto fail; + } + + if (chdir("/") < 0) { + r = -errno; + log_error("Failed to change directory: %m"); + goto fail; + } + + if (old_root_fd >= 0) { + struct stat rb; + + if (fstat(old_root_fd, &rb) < 0) + log_warning("Failed to stat old root directory, leaving: %m"); + else { + rm_rf_children(old_root_fd, false, false, &rb); + old_root_fd = -1; + } + } + + r = 0; + +fail: + if (old_root_fd >= 0) + close_nointr_nofail(old_root_fd); + + return r; +} diff --git a/src/core/switch-root.h b/src/core/switch-root.h new file mode 100644 index 0000000000..0c4cd1e403 --- /dev/null +++ b/src/core/switch-root.h @@ -0,0 +1,27 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#ifndef fooswitchroothfoo +#define fooswitchroothfoo + +/*** + This file is part of systemd. + + Copyright 2012 Harald Hoyer, Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int switch_root(const char *switch_root); + +#endif diff --git a/src/core/sync.c b/src/core/sync.c new file mode 100644 index 0000000000..7e74b63071 --- /dev/null +++ b/src/core/sync.c @@ -0,0 +1,65 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <pthread.h> +#include <unistd.h> + +#include "sync.h" + +static void *sync_thread(void *p) { + sync(); + return NULL; +} + +int asynchronous_sync(void) { + pthread_attr_t a; + pthread_t t; + int r; + + /* It kinda sucks that we have to resort to threads to + * implement an asynchronous sync(), but well, such is + * life. + * + * Note that issuing this command right before exiting a + * process will cause the process to wait for the sync() to + * complete. This function hence is nicely asynchronous really + * only in long running processes. */ + + r = pthread_attr_init(&a); + if (r != 0) + return -r; + + r = pthread_attr_setdetachstate(&a, PTHREAD_CREATE_DETACHED); + if (r != 0) { + r = -r; + goto finish; + } + + r = pthread_create(&t, &a, sync_thread, NULL); + if (r != 0) { + r = -r; + goto finish; + } + +finish: + pthread_attr_destroy(&a); + return r; +} diff --git a/src/core/sync.h b/src/core/sync.h new file mode 100644 index 0000000000..eb26c88deb --- /dev/null +++ b/src/core/sync.h @@ -0,0 +1,24 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int asynchronous_sync(void); diff --git a/src/core/syscall-list.c b/src/core/syscall-list.c new file mode 100644 index 0000000000..35216b2a88 --- /dev/null +++ b/src/core/syscall-list.c @@ -0,0 +1,57 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/syscall.h> +#include <string.h> + +#include "util.h" + +#include "syscall-list.h" + +static const struct syscall_name* lookup_syscall(register const char *str, + register unsigned int len); + +#include "syscall-to-name.h" +#include "syscall-from-name.h" + +const char *syscall_to_name(int id) { + id = SYSCALL_TO_INDEX(id); + if (id < 0 || id >= (int) ELEMENTSOF(syscall_names)) + return NULL; + + return syscall_names[id]; +} + +int syscall_from_name(const char *name) { + const struct syscall_name *sc; + + assert(name); + + sc = lookup_syscall(name, strlen(name)); + if (!sc) + return -1; + + return sc->id; +} + +int syscall_max(void) { + return ELEMENTSOF(syscall_names); +} diff --git a/src/core/syscall-list.h b/src/core/syscall-list.h new file mode 100644 index 0000000000..6bb4d91f91 --- /dev/null +++ b/src/core/syscall-list.h @@ -0,0 +1,44 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#ifndef foosyscalllisthfoo +#define foosyscalllisthfoo + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#if defined __x86_64__ && defined __ILP32__ +/* The x32 ABI defines all of its syscalls with bit 30 set, which causes + issues when attempting to use syscalls as simple indices into an array. + Instead, use the syscall id & ~SYSCALL_MASK as the index, and | the + internal id with the syscall mask as needed. +*/ +#include <asm/unistd.h> +#define SYSCALL_TO_INDEX(x) ((x) & ~__X32_SYSCALL_BIT) +#define INDEX_TO_SYSCALL(x) ((x) | __X32_SYSCALL_BIT) +#else +#define SYSCALL_TO_INDEX(x) (x) +#define INDEX_TO_SYSCALL(x) (x) +#endif + +const char *syscall_to_name(int id); +int syscall_from_name(const char *name); + +int syscall_max(void); + +#endif diff --git a/src/core/sysfs-show.h b/src/core/sysfs-show.h new file mode 100644 index 0000000000..9ffd129c48 --- /dev/null +++ b/src/core/sysfs-show.h @@ -0,0 +1,24 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int show_sysfs(const char *seat, const char *prefix, unsigned columns); diff --git a/src/core/system.conf b/src/core/system.conf new file mode 100644 index 0000000000..508e0f5fa4 --- /dev/null +++ b/src/core/system.conf @@ -0,0 +1,43 @@ +# This file is part of systemd. +# +# systemd is free software; you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published by +# the Free Software Foundation; either version 2.1 of the License, or +# (at your option) any later version. +# +# See systemd-system.conf(5) for details + +[Manager] +#LogLevel=info +#LogTarget=journal-or-kmsg +#LogColor=yes +#LogLocation=no +#DumpCore=yes +#CrashShell=no +#ShowStatus=yes +#CrashChVT=1 +#CPUAffinity=1 2 +#DefaultControllers=cpu +#DefaultStandardOutput=journal +#DefaultStandardError=inherit +#JoinControllers=cpu,cpuacct,cpuset net_cls,net_prio +#RuntimeWatchdogSec=0 +#ShutdownWatchdogSec=10min +#CapabilityBoundingSet= +#TimerSlackNSec= +#DefaultLimitCPU= +#DefaultLimitFSIZE= +#DefaultLimitDATA= +#DefaultLimitSTACK= +#DefaultLimitCORE= +#DefaultLimitRSS= +#DefaultLimitNOFILE= +#DefaultLimitAS= +#DefaultLimitNPROC= +#DefaultLimitMEMLOCK= +#DefaultLimitLOCKS= +#DefaultLimitSIGPENDING= +#DefaultLimitMSGQUEUE= +#DefaultLimitNICE= +#DefaultLimitRTPRIO= +#DefaultLimitRTTIME= diff --git a/src/core/systemd.pc.in b/src/core/systemd.pc.in new file mode 100644 index 0000000000..2f49d5df52 --- /dev/null +++ b/src/core/systemd.pc.in @@ -0,0 +1,23 @@ +# This file is part of systemd. +# +# systemd is free software; you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published by +# the Free Software Foundation; either version 2.1 of the License, or +# (at your option) any later version. + +prefix=@prefix@ +exec_prefix=@exec_prefix@ +systemdutildir=@rootlibexecdir@ +systemdsystemunitdir=@systemunitdir@ +systemdsystempresetdir=@systempresetdir@ +systemduserunitdir=@userunitdir@ +systemduserpresetdir=@userpresetdir@ +systemdsystemconfdir=@pkgsysconfdir@/system +systemduserconfdir=@pkgsysconfdir@/user +systemdsystemunitpath=${systemdsystemconfdir}:/etc/systemd/system:/run/systemd/system:/usr/local/lib/systemd/system:${systemdsystemunitdir}:/usr/lib/systemd/system:/lib/systemd/system +systemduserunitpath=${systemduserconfdir}:/etc/systemd/user:/run/systemd/user:/usr/local/lib/systemd/user:/usr/local/share/systemd/user:${systemduserunitdir}:/usr/lib/systemd/user:/usr/share/systemd/user + +Name: systemd +Description: systemd System and Service Manager +URL: @PACKAGE_URL@ +Version: @PACKAGE_VERSION@ diff --git a/src/core/target.c b/src/core/target.c new file mode 100644 index 0000000000..3fffa0d2f5 --- /dev/null +++ b/src/core/target.c @@ -0,0 +1,239 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <signal.h> +#include <unistd.h> + +#include "unit.h" +#include "target.h" +#include "load-fragment.h" +#include "log.h" +#include "dbus-target.h" +#include "special.h" +#include "unit-name.h" + +static const UnitActiveState state_translation_table[_TARGET_STATE_MAX] = { + [TARGET_DEAD] = UNIT_INACTIVE, + [TARGET_ACTIVE] = UNIT_ACTIVE +}; + +static void target_set_state(Target *t, TargetState state) { + TargetState old_state; + assert(t); + + old_state = t->state; + t->state = state; + + if (state != old_state) + log_debug("%s changed %s -> %s", + UNIT(t)->id, + target_state_to_string(old_state), + target_state_to_string(state)); + + unit_notify(UNIT(t), state_translation_table[old_state], state_translation_table[state], true); +} + +static int target_add_default_dependencies(Target *t) { + + static const UnitDependency deps[] = { + UNIT_REQUIRES, + UNIT_REQUIRES_OVERRIDABLE, + UNIT_REQUISITE, + UNIT_REQUISITE_OVERRIDABLE, + UNIT_WANTS, + UNIT_BINDS_TO, + UNIT_PART_OF + }; + + Iterator i; + Unit *other; + int r; + unsigned k; + + assert(t); + + /* Imply ordering for requirement dependencies on target + * units. Note that when the user created a contradicting + * ordering manually we won't add anything in here to make + * sure we don't create a loop. */ + + for (k = 0; k < ELEMENTSOF(deps); k++) + SET_FOREACH(other, UNIT(t)->dependencies[deps[k]], i) { + r = unit_add_default_target_dependency(other, UNIT(t)); + if (r < 0) + return r; + } + + /* Make sure targets are unloaded on shutdown */ + return unit_add_dependency_by_name(UNIT(t), UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +static int target_load(Unit *u) { + Target *t = TARGET(u); + int r; + + assert(t); + + r = unit_load_fragment_and_dropin(u); + if (r < 0) + return r; + + /* This is a new unit? Then let's add in some extras */ + if (u->load_state == UNIT_LOADED && u->default_dependencies) { + r = target_add_default_dependencies(t); + if (r < 0) + return r; + } + + return 0; +} + +static int target_coldplug(Unit *u) { + Target *t = TARGET(u); + + assert(t); + assert(t->state == TARGET_DEAD); + + if (t->deserialized_state != t->state) + target_set_state(t, t->deserialized_state); + + return 0; +} + +static void target_dump(Unit *u, FILE *f, const char *prefix) { + Target *t = TARGET(u); + + assert(t); + assert(f); + + fprintf(f, + "%sTarget State: %s\n", + prefix, target_state_to_string(t->state)); +} + +static int target_start(Unit *u) { + Target *t = TARGET(u); + + assert(t); + assert(t->state == TARGET_DEAD); + + target_set_state(t, TARGET_ACTIVE); + return 0; +} + +static int target_stop(Unit *u) { + Target *t = TARGET(u); + + assert(t); + assert(t->state == TARGET_ACTIVE); + + target_set_state(t, TARGET_DEAD); + return 0; +} + +static int target_serialize(Unit *u, FILE *f, FDSet *fds) { + Target *s = TARGET(u); + + assert(s); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", target_state_to_string(s->state)); + return 0; +} + +static int target_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Target *s = TARGET(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + TargetState state; + + state = target_state_from_string(value); + if (state < 0) + log_debug("Failed to parse state value %s", value); + else + s->deserialized_state = state; + + } else + log_debug("Unknown serialization key '%s'", key); + + return 0; +} + +_pure_ static UnitActiveState target_active_state(Unit *u) { + assert(u); + + return state_translation_table[TARGET(u)->state]; +} + +_pure_ static const char *target_sub_state_to_string(Unit *u) { + assert(u); + + return target_state_to_string(TARGET(u)->state); +} + +static const char* const target_state_table[_TARGET_STATE_MAX] = { + [TARGET_DEAD] = "dead", + [TARGET_ACTIVE] = "active" +}; + +DEFINE_STRING_TABLE_LOOKUP(target_state, TargetState); + +const UnitVTable target_vtable = { + .object_size = sizeof(Target), + .sections = + "Unit\0" + "Target\0" + "Install\0", + + .load = target_load, + .coldplug = target_coldplug, + + .dump = target_dump, + + .start = target_start, + .stop = target_stop, + + .serialize = target_serialize, + .deserialize_item = target_deserialize_item, + + .active_state = target_active_state, + .sub_state_to_string = target_sub_state_to_string, + + .bus_interface = "org.freedesktop.systemd1.Target", + .bus_message_handler = bus_target_message_handler, + + .status_message_formats = { + .finished_start_job = { + [JOB_DONE] = "Reached target %s.", + [JOB_DEPENDENCY] = "Dependency failed for %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Stopped target %s.", + }, + }, +}; diff --git a/src/core/target.h b/src/core/target.h new file mode 100644 index 0000000000..a5398d9e98 --- /dev/null +++ b/src/core/target.h @@ -0,0 +1,44 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Target Target; + +#include "unit.h" + +typedef enum TargetState { + TARGET_DEAD, + TARGET_ACTIVE, + _TARGET_STATE_MAX, + _TARGET_STATE_INVALID = -1 +} TargetState; + +struct Target { + Unit meta; + + TargetState state, deserialized_state; +}; + +extern const UnitVTable target_vtable; + +const char* target_state_to_string(TargetState i) _const_; +TargetState target_state_from_string(const char *s) _pure_; diff --git a/src/core/tcpwrap.c b/src/core/tcpwrap.c new file mode 100644 index 0000000000..6c630fac60 --- /dev/null +++ b/src/core/tcpwrap.c @@ -0,0 +1,68 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/socket.h> +#include <sys/un.h> +#include <arpa/inet.h> +#include <netinet/in.h> + +#ifdef HAVE_LIBWRAP +#include <tcpd.h> +#endif + +#include "tcpwrap.h" +#include "log.h" + +bool socket_tcpwrap(int fd, const char *name) { +#ifdef HAVE_LIBWRAP + struct request_info req; + union { + struct sockaddr sa; + struct sockaddr_in in; + struct sockaddr_in6 in6; + struct sockaddr_un un; + struct sockaddr_storage storage; + } sa_union; + socklen_t l = sizeof(sa_union); + + if (getsockname(fd, &sa_union.sa, &l) < 0) + return true; + + if (sa_union.sa.sa_family != AF_INET && + sa_union.sa.sa_family != AF_INET6) + return true; + + request_init(&req, + RQ_DAEMON, name, + RQ_FILE, fd, + NULL); + + fromhost(&req); + + if (!hosts_access(&req)) { + log_warning("Connection refused by tcpwrap."); + return false; + } + + log_debug("Connection accepted by tcpwrap."); +#endif + return true; +} diff --git a/src/core/tcpwrap.h b/src/core/tcpwrap.h new file mode 100644 index 0000000000..3353b6596e --- /dev/null +++ b/src/core/tcpwrap.h @@ -0,0 +1,26 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +bool socket_tcpwrap(int fd, const char *name); diff --git a/src/core/timer.c b/src/core/timer.c new file mode 100644 index 0000000000..9166c1e2fc --- /dev/null +++ b/src/core/timer.c @@ -0,0 +1,622 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "unit.h" +#include "unit-name.h" +#include "timer.h" +#include "dbus-timer.h" +#include "special.h" +#include "dbus-common.h" + +static const UnitActiveState state_translation_table[_TIMER_STATE_MAX] = { + [TIMER_DEAD] = UNIT_INACTIVE, + [TIMER_WAITING] = UNIT_ACTIVE, + [TIMER_RUNNING] = UNIT_ACTIVE, + [TIMER_ELAPSED] = UNIT_ACTIVE, + [TIMER_FAILED] = UNIT_FAILED +}; + +static void timer_init(Unit *u) { + Timer *t = TIMER(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + t->next_elapse_monotonic = (usec_t) -1; + t->next_elapse_realtime = (usec_t) -1; + watch_init(&t->monotonic_watch); + watch_init(&t->realtime_watch); +} + +void timer_free_values(Timer *t) { + TimerValue *v; + + assert(t); + + while ((v = t->values)) { + LIST_REMOVE(TimerValue, value, t->values, v); + + if (v->calendar_spec) + calendar_spec_free(v->calendar_spec); + + free(v); + } +} + +static void timer_done(Unit *u) { + Timer *t = TIMER(u); + + assert(t); + + timer_free_values(t); + + unit_unwatch_timer(u, &t->monotonic_watch); + unit_unwatch_timer(u, &t->realtime_watch); +} + +static int timer_verify(Timer *t) { + assert(t); + + if (UNIT(t)->load_state != UNIT_LOADED) + return 0; + + if (!t->values) { + log_error_unit(UNIT(t)->id, + "%s lacks value setting. Refusing.", UNIT(t)->id); + return -EINVAL; + } + + return 0; +} + +static int timer_add_default_dependencies(Timer *t) { + int r; + + assert(t); + + r = unit_add_dependency_by_name(UNIT(t), UNIT_BEFORE, SPECIAL_TIMERS_TARGET, NULL, true); + if (r < 0) + return r; + + if (UNIT(t)->manager->running_as == SYSTEMD_SYSTEM) { + r = unit_add_two_dependencies_by_name(UNIT(t), UNIT_AFTER, UNIT_REQUIRES, SPECIAL_SYSINIT_TARGET, NULL, true); + if (r < 0) + return r; + } + + return unit_add_two_dependencies_by_name(UNIT(t), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +static int timer_load(Unit *u) { + Timer *t = TIMER(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + r = unit_load_fragment_and_dropin(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + + if (set_isempty(u->dependencies[UNIT_TRIGGERS])) { + Unit *x; + + r = unit_load_related_unit(u, ".service", &x); + if (r < 0) + return r; + + r = unit_add_two_dependencies(u, UNIT_BEFORE, UNIT_TRIGGERS, x, true); + if (r < 0) + return r; + } + + if (UNIT(t)->default_dependencies) { + r = timer_add_default_dependencies(t); + if (r < 0) + return r; + } + } + + return timer_verify(t); +} + +static void timer_dump(Unit *u, FILE *f, const char *prefix) { + Timer *t = TIMER(u); + Unit *trigger; + TimerValue *v; + + trigger = UNIT_TRIGGER(u); + + fprintf(f, + "%sTimer State: %s\n" + "%sResult: %s\n" + "%sUnit: %s\n", + prefix, timer_state_to_string(t->state), + prefix, timer_result_to_string(t->result), + prefix, trigger ? trigger->id : "n/a"); + + LIST_FOREACH(value, v, t->values) { + + if (v->base == TIMER_CALENDAR) { + _cleanup_free_ char *p = NULL; + + calendar_spec_to_string(v->calendar_spec, &p); + + fprintf(f, + "%s%s: %s\n", + prefix, + timer_base_to_string(v->base), + strna(p)); + } else { + char timespan1[FORMAT_TIMESPAN_MAX]; + + fprintf(f, + "%s%s: %s\n", + prefix, + timer_base_to_string(v->base), + strna(format_timespan(timespan1, sizeof(timespan1), v->value, 0))); + } + } +} + +static void timer_set_state(Timer *t, TimerState state) { + TimerState old_state; + assert(t); + + old_state = t->state; + t->state = state; + + if (state != TIMER_WAITING) { + unit_unwatch_timer(UNIT(t), &t->monotonic_watch); + unit_unwatch_timer(UNIT(t), &t->realtime_watch); + } + + if (state != old_state) + log_debug_unit(UNIT(t)->id, + "%s changed %s -> %s", UNIT(t)->id, + timer_state_to_string(old_state), + timer_state_to_string(state)); + + unit_notify(UNIT(t), state_translation_table[old_state], state_translation_table[state], true); +} + +static void timer_enter_waiting(Timer *t, bool initial); + +static int timer_coldplug(Unit *u) { + Timer *t = TIMER(u); + + assert(t); + assert(t->state == TIMER_DEAD); + + if (t->deserialized_state != t->state) { + + if (t->deserialized_state == TIMER_WAITING) + timer_enter_waiting(t, false); + else + timer_set_state(t, t->deserialized_state); + } + + return 0; +} + +static void timer_enter_dead(Timer *t, TimerResult f) { + assert(t); + + if (f != TIMER_SUCCESS) + t->result = f; + + timer_set_state(t, t->result != TIMER_SUCCESS ? TIMER_FAILED : TIMER_DEAD); +} + +static void timer_enter_waiting(Timer *t, bool initial) { + TimerValue *v; + usec_t base = 0; + dual_timestamp ts; + bool found_monotonic = false, found_realtime = false; + int r; + + dual_timestamp_get(&ts); + t->next_elapse_monotonic = t->next_elapse_realtime = 0; + + LIST_FOREACH(value, v, t->values) { + + if (v->disabled) + continue; + + if (v->base == TIMER_CALENDAR) { + + r = calendar_spec_next_usec(v->calendar_spec, ts.realtime, &v->next_elapse); + if (r < 0) + continue; + + if (!found_realtime) + t->next_elapse_realtime = v->next_elapse; + else + t->next_elapse_realtime = MIN(t->next_elapse_realtime, v->next_elapse); + + found_realtime = true; + + } else { + switch (v->base) { + + case TIMER_ACTIVE: + if (state_translation_table[t->state] == UNIT_ACTIVE) + base = UNIT(t)->inactive_exit_timestamp.monotonic; + else + base = ts.monotonic; + break; + + case TIMER_BOOT: + /* CLOCK_MONOTONIC equals the uptime on Linux */ + base = 0; + break; + + case TIMER_STARTUP: + base = UNIT(t)->manager->userspace_timestamp.monotonic; + break; + + case TIMER_UNIT_ACTIVE: + + base = UNIT_TRIGGER(UNIT(t))->inactive_exit_timestamp.monotonic; + + if (base <= 0) + base = t->last_trigger_monotonic; + + if (base <= 0) + continue; + + break; + + case TIMER_UNIT_INACTIVE: + + base = UNIT_TRIGGER(UNIT(t))->inactive_enter_timestamp.monotonic; + + if (base <= 0) + base = t->last_trigger_monotonic; + + if (base <= 0) + continue; + + break; + + default: + assert_not_reached("Unknown timer base"); + } + + v->next_elapse = base + v->value; + + if (!initial && + v->next_elapse < ts.monotonic && + (v->base == TIMER_ACTIVE || v->base == TIMER_BOOT || v->base == TIMER_STARTUP)) { + /* This is a one time trigger, disable it now */ + v->disabled = true; + continue; + } + + if (!found_monotonic) + t->next_elapse_monotonic = v->next_elapse; + else + t->next_elapse_monotonic = MIN(t->next_elapse_monotonic, v->next_elapse); + + found_monotonic = true; + } + } + + if (!found_monotonic && !found_realtime) { + log_debug_unit(UNIT(t)->id, "%s: Timer is elapsed.", UNIT(t)->id); + timer_set_state(t, TIMER_ELAPSED); + return; + } + + if (found_monotonic) { + char buf[FORMAT_TIMESPAN_MAX]; + log_debug_unit(UNIT(t)->id, + "%s: Monotonic timer elapses in %s.", + UNIT(t)->id, + format_timespan(buf, sizeof(buf), t->next_elapse_monotonic > ts.monotonic ? t->next_elapse_monotonic - ts.monotonic : 0, 0)); + + r = unit_watch_timer(UNIT(t), CLOCK_MONOTONIC, false, t->next_elapse_monotonic, &t->monotonic_watch); + if (r < 0) + goto fail; + } else + unit_unwatch_timer(UNIT(t), &t->monotonic_watch); + + if (found_realtime) { + char buf[FORMAT_TIMESTAMP_MAX]; + log_debug_unit(UNIT(t)->id, + "%s: Realtime timer elapses at %s.", + UNIT(t)->id, + format_timestamp(buf, sizeof(buf), t->next_elapse_realtime)); + + r = unit_watch_timer(UNIT(t), CLOCK_REALTIME, false, t->next_elapse_realtime, &t->realtime_watch); + if (r < 0) + goto fail; + } else + unit_unwatch_timer(UNIT(t), &t->realtime_watch); + + timer_set_state(t, TIMER_WAITING); + return; + +fail: + log_warning_unit(UNIT(t)->id, + "%s failed to enter waiting state: %s", + UNIT(t)->id, strerror(-r)); + timer_enter_dead(t, TIMER_FAILURE_RESOURCES); +} + +static void timer_enter_running(Timer *t) { + DBusError error; + int r; + + assert(t); + dbus_error_init(&error); + + /* Don't start job if we are supposed to go down */ + if (unit_stop_pending(UNIT(t))) + return; + + r = manager_add_job(UNIT(t)->manager, JOB_START, UNIT_TRIGGER(UNIT(t)), + JOB_REPLACE, true, &error, NULL); + if (r < 0) + goto fail; + + t->last_trigger_monotonic = now(CLOCK_MONOTONIC); + + timer_set_state(t, TIMER_RUNNING); + return; + +fail: + log_warning_unit(UNIT(t)->id, + "%s failed to queue unit startup job: %s", + UNIT(t)->id, bus_error(&error, r)); + timer_enter_dead(t, TIMER_FAILURE_RESOURCES); + + dbus_error_free(&error); +} + +static int timer_start(Unit *u) { + Timer *t = TIMER(u); + + assert(t); + assert(t->state == TIMER_DEAD || t->state == TIMER_FAILED); + + if (UNIT_TRIGGER(u)->load_state != UNIT_LOADED) + return -ENOENT; + + t->result = TIMER_SUCCESS; + timer_enter_waiting(t, true); + return 0; +} + +static int timer_stop(Unit *u) { + Timer *t = TIMER(u); + + assert(t); + assert(t->state == TIMER_WAITING || t->state == TIMER_RUNNING || t->state == TIMER_ELAPSED); + + timer_enter_dead(t, TIMER_SUCCESS); + return 0; +} + +static int timer_serialize(Unit *u, FILE *f, FDSet *fds) { + Timer *t = TIMER(u); + + assert(u); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", timer_state_to_string(t->state)); + unit_serialize_item(u, f, "result", timer_result_to_string(t->result)); + + return 0; +} + +static int timer_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Timer *t = TIMER(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + TimerState state; + + state = timer_state_from_string(value); + if (state < 0) + log_debug_unit(u->id, "Failed to parse state value %s", value); + else + t->deserialized_state = state; + } else if (streq(key, "result")) { + TimerResult f; + + f = timer_result_from_string(value); + if (f < 0) + log_debug_unit(u->id, "Failed to parse result value %s", value); + else if (f != TIMER_SUCCESS) + t->result = f; + + } else + log_debug_unit(u->id, "Unknown serialization key '%s'", key); + + return 0; +} + +_pure_ static UnitActiveState timer_active_state(Unit *u) { + assert(u); + + return state_translation_table[TIMER(u)->state]; +} + +_pure_ static const char *timer_sub_state_to_string(Unit *u) { + assert(u); + + return timer_state_to_string(TIMER(u)->state); +} + +static void timer_timer_event(Unit *u, uint64_t elapsed, Watch *w) { + Timer *t = TIMER(u); + + assert(t); + assert(elapsed == 1); + + if (t->state != TIMER_WAITING) + return; + + log_debug_unit(u->id, "Timer elapsed on %s", u->id); + timer_enter_running(t); +} + +static void timer_trigger_notify(Unit *u, Unit *other) { + Timer *t = TIMER(u); + TimerValue *v; + + assert(u); + assert(other); + + if (other->load_state != UNIT_LOADED) + return; + + /* Reenable all timers that depend on unit state */ + LIST_FOREACH(value, v, t->values) + if (v->base == TIMER_UNIT_ACTIVE || + v->base == TIMER_UNIT_INACTIVE) + v->disabled = false; + + switch (t->state) { + + case TIMER_WAITING: + case TIMER_ELAPSED: + + /* Recalculate sleep time */ + timer_enter_waiting(t, false); + break; + + case TIMER_RUNNING: + + if (UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(other))) { + log_debug_unit(UNIT(t)->id, + "%s got notified about unit deactivation.", + UNIT(t)->id); + timer_enter_waiting(t, false); + } + break; + + case TIMER_DEAD: + case TIMER_FAILED: + break; + + default: + assert_not_reached("Unknown timer state"); + } +} + +static void timer_reset_failed(Unit *u) { + Timer *t = TIMER(u); + + assert(t); + + if (t->state == TIMER_FAILED) + timer_set_state(t, TIMER_DEAD); + + t->result = TIMER_SUCCESS; +} + +static void timer_time_change(Unit *u) { + Timer *t = TIMER(u); + + assert(u); + + if (t->state != TIMER_WAITING) + return; + + log_debug_unit(u->id, + "%s: time change, recalculating next elapse.", u->id); + timer_enter_waiting(t, false); +} + +static const char* const timer_state_table[_TIMER_STATE_MAX] = { + [TIMER_DEAD] = "dead", + [TIMER_WAITING] = "waiting", + [TIMER_RUNNING] = "running", + [TIMER_ELAPSED] = "elapsed", + [TIMER_FAILED] = "failed" +}; + +DEFINE_STRING_TABLE_LOOKUP(timer_state, TimerState); + +static const char* const timer_base_table[_TIMER_BASE_MAX] = { + [TIMER_ACTIVE] = "OnActiveSec", + [TIMER_BOOT] = "OnBootSec", + [TIMER_STARTUP] = "OnStartupSec", + [TIMER_UNIT_ACTIVE] = "OnUnitActiveSec", + [TIMER_UNIT_INACTIVE] = "OnUnitInactiveSec", + [TIMER_CALENDAR] = "OnCalendar" +}; + +DEFINE_STRING_TABLE_LOOKUP(timer_base, TimerBase); + +static const char* const timer_result_table[_TIMER_RESULT_MAX] = { + [TIMER_SUCCESS] = "success", + [TIMER_FAILURE_RESOURCES] = "resources" +}; + +DEFINE_STRING_TABLE_LOOKUP(timer_result, TimerResult); + +const UnitVTable timer_vtable = { + .object_size = sizeof(Timer), + .sections = + "Unit\0" + "Timer\0" + "Install\0", + + .init = timer_init, + .done = timer_done, + .load = timer_load, + + .coldplug = timer_coldplug, + + .dump = timer_dump, + + .start = timer_start, + .stop = timer_stop, + + .serialize = timer_serialize, + .deserialize_item = timer_deserialize_item, + + .active_state = timer_active_state, + .sub_state_to_string = timer_sub_state_to_string, + + .timer_event = timer_timer_event, + + .trigger_notify = timer_trigger_notify, + + .reset_failed = timer_reset_failed, + .time_change = timer_time_change, + + .bus_interface = "org.freedesktop.systemd1.Timer", + .bus_message_handler = bus_timer_message_handler, + .bus_invalidating_properties = bus_timer_invalidating_properties +}; diff --git a/src/core/timer.h b/src/core/timer.h new file mode 100644 index 0000000000..4168553e9d --- /dev/null +++ b/src/core/timer.h @@ -0,0 +1,97 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Timer Timer; + +#include "unit.h" +#include "calendarspec.h" + +typedef enum TimerState { + TIMER_DEAD, + TIMER_WAITING, + TIMER_RUNNING, + TIMER_ELAPSED, + TIMER_FAILED, + _TIMER_STATE_MAX, + _TIMER_STATE_INVALID = -1 +} TimerState; + +typedef enum TimerBase { + TIMER_ACTIVE, + TIMER_BOOT, + TIMER_STARTUP, + TIMER_UNIT_ACTIVE, + TIMER_UNIT_INACTIVE, + TIMER_CALENDAR, + _TIMER_BASE_MAX, + _TIMER_BASE_INVALID = -1 +} TimerBase; + +typedef struct TimerValue { + TimerBase base; + bool disabled; + clockid_t clock_id; + + usec_t value; /* only for monotonic events */ + CalendarSpec *calendar_spec; /* only for calendar events */ + usec_t next_elapse; + + LIST_FIELDS(struct TimerValue, value); +} TimerValue; + +typedef enum TimerResult { + TIMER_SUCCESS, + TIMER_FAILURE_RESOURCES, + _TIMER_RESULT_MAX, + _TIMER_RESULT_INVALID = -1 +} TimerResult; + +struct Timer { + Unit meta; + + LIST_HEAD(TimerValue, values); + usec_t next_elapse_monotonic; + usec_t next_elapse_realtime; + + TimerState state, deserialized_state; + + Watch monotonic_watch; + Watch realtime_watch; + + TimerResult result; + + usec_t last_trigger_monotonic; +}; + +void timer_free_values(Timer *t); + +extern const UnitVTable timer_vtable; + +const char *timer_state_to_string(TimerState i) _const_; +TimerState timer_state_from_string(const char *s) _pure_; + +const char *timer_base_to_string(TimerBase i) _const_; +TimerBase timer_base_from_string(const char *s) _pure_; + +const char* timer_result_to_string(TimerResult i) _const_; +TimerResult timer_result_from_string(const char *s) _pure_; diff --git a/src/core/transaction.c b/src/core/transaction.c new file mode 100644 index 0000000000..fa97b69755 --- /dev/null +++ b/src/core/transaction.c @@ -0,0 +1,1132 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <unistd.h> +#include <fcntl.h> + +#include "transaction.h" +#include "bus-errors.h" +#include "dbus-common.h" + +static void transaction_unlink_job(Transaction *tr, Job *j, bool delete_dependencies); + +static void transaction_delete_job(Transaction *tr, Job *j, bool delete_dependencies) { + assert(tr); + assert(j); + + /* Deletes one job from the transaction */ + + transaction_unlink_job(tr, j, delete_dependencies); + + job_free(j); +} + +static void transaction_delete_unit(Transaction *tr, Unit *u) { + Job *j; + + /* Deletes all jobs associated with a certain unit from the + * transaction */ + + while ((j = hashmap_get(tr->jobs, u))) + transaction_delete_job(tr, j, true); +} + +void transaction_abort(Transaction *tr) { + Job *j; + + assert(tr); + + while ((j = hashmap_first(tr->jobs))) + transaction_delete_job(tr, j, false); + + assert(hashmap_isempty(tr->jobs)); +} + +static void transaction_find_jobs_that_matter_to_anchor(Job *j, unsigned generation) { + JobDependency *l; + + /* A recursive sweep through the graph that marks all units + * that matter to the anchor job, i.e. are directly or + * indirectly a dependency of the anchor job via paths that + * are fully marked as mattering. */ + + j->matters_to_anchor = true; + j->generation = generation; + + LIST_FOREACH(subject, l, j->subject_list) { + + /* This link does not matter */ + if (!l->matters) + continue; + + /* This unit has already been marked */ + if (l->object->generation == generation) + continue; + + transaction_find_jobs_that_matter_to_anchor(l->object, generation); + } +} + +static void transaction_merge_and_delete_job(Transaction *tr, Job *j, Job *other, JobType t) { + JobDependency *l, *last; + + assert(j); + assert(other); + assert(j->unit == other->unit); + assert(!j->installed); + + /* Merges 'other' into 'j' and then deletes 'other'. */ + + j->type = t; + j->state = JOB_WAITING; + j->override = j->override || other->override; + j->irreversible = j->irreversible || other->irreversible; + + j->matters_to_anchor = j->matters_to_anchor || other->matters_to_anchor; + + /* Patch us in as new owner of the JobDependency objects */ + last = NULL; + LIST_FOREACH(subject, l, other->subject_list) { + assert(l->subject == other); + l->subject = j; + last = l; + } + + /* Merge both lists */ + if (last) { + last->subject_next = j->subject_list; + if (j->subject_list) + j->subject_list->subject_prev = last; + j->subject_list = other->subject_list; + } + + /* Patch us in as new owner of the JobDependency objects */ + last = NULL; + LIST_FOREACH(object, l, other->object_list) { + assert(l->object == other); + l->object = j; + last = l; + } + + /* Merge both lists */ + if (last) { + last->object_next = j->object_list; + if (j->object_list) + j->object_list->object_prev = last; + j->object_list = other->object_list; + } + + /* Kill the other job */ + other->subject_list = NULL; + other->object_list = NULL; + transaction_delete_job(tr, other, true); +} + +_pure_ static bool job_is_conflicted_by(Job *j) { + JobDependency *l; + + assert(j); + + /* Returns true if this job is pulled in by a least one + * ConflictedBy dependency. */ + + LIST_FOREACH(object, l, j->object_list) + if (l->conflicts) + return true; + + return false; +} + +static int delete_one_unmergeable_job(Transaction *tr, Job *j) { + Job *k; + + assert(j); + + /* Tries to delete one item in the linked list + * j->transaction_next->transaction_next->... that conflicts + * with another one, in an attempt to make an inconsistent + * transaction work. */ + + /* We rely here on the fact that if a merged with b does not + * merge with c, either a or b merge with c neither */ + LIST_FOREACH(transaction, j, j) + LIST_FOREACH(transaction, k, j->transaction_next) { + Job *d; + + /* Is this one mergeable? Then skip it */ + if (job_type_is_mergeable(j->type, k->type)) + continue; + + /* Ok, we found two that conflict, let's see if we can + * drop one of them */ + if (!j->matters_to_anchor && !k->matters_to_anchor) { + + /* Both jobs don't matter, so let's + * find the one that is smarter to + * remove. Let's think positive and + * rather remove stops then starts -- + * except if something is being + * stopped because it is conflicted by + * another unit in which case we + * rather remove the start. */ + + log_debug_unit(j->unit->id, + "Looking at job %s/%s conflicted_by=%s", + j->unit->id, job_type_to_string(j->type), + yes_no(j->type == JOB_STOP && job_is_conflicted_by(j))); + log_debug_unit(k->unit->id, + "Looking at job %s/%s conflicted_by=%s", + k->unit->id, job_type_to_string(k->type), + yes_no(k->type == JOB_STOP && job_is_conflicted_by(k))); + + if (j->type == JOB_STOP) { + + if (job_is_conflicted_by(j)) + d = k; + else + d = j; + + } else if (k->type == JOB_STOP) { + + if (job_is_conflicted_by(k)) + d = j; + else + d = k; + } else + d = j; + + } else if (!j->matters_to_anchor) + d = j; + else if (!k->matters_to_anchor) + d = k; + else + return -ENOEXEC; + + /* Ok, we can drop one, so let's do so. */ + log_debug_unit(d->unit->id, + "Fixing conflicting jobs by deleting job %s/%s", + d->unit->id, job_type_to_string(d->type)); + transaction_delete_job(tr, d, true); + return 0; + } + + return -EINVAL; +} + +static int transaction_merge_jobs(Transaction *tr, DBusError *e) { + Job *j; + Iterator i; + int r; + + assert(tr); + + /* First step, check whether any of the jobs for one specific + * task conflict. If so, try to drop one of them. */ + HASHMAP_FOREACH(j, tr->jobs, i) { + JobType t; + Job *k; + + t = j->type; + LIST_FOREACH(transaction, k, j->transaction_next) { + if (job_type_merge_and_collapse(&t, k->type, j->unit) >= 0) + continue; + + /* OK, we could not merge all jobs for this + * action. Let's see if we can get rid of one + * of them */ + + r = delete_one_unmergeable_job(tr, j); + if (r >= 0) + /* Ok, we managed to drop one, now + * let's ask our callers to call us + * again after garbage collecting */ + return -EAGAIN; + + /* We couldn't merge anything. Failure */ + dbus_set_error(e, BUS_ERROR_TRANSACTION_JOBS_CONFLICTING, "Transaction contains conflicting jobs '%s' and '%s' for %s. Probably contradicting requirement dependencies configured.", + job_type_to_string(t), job_type_to_string(k->type), k->unit->id); + return r; + } + } + + /* Second step, merge the jobs. */ + HASHMAP_FOREACH(j, tr->jobs, i) { + JobType t = j->type; + Job *k; + + /* Merge all transaction jobs for j->unit */ + LIST_FOREACH(transaction, k, j->transaction_next) + assert_se(job_type_merge_and_collapse(&t, k->type, j->unit) == 0); + + while ((k = j->transaction_next)) { + if (tr->anchor_job == k) { + transaction_merge_and_delete_job(tr, k, j, t); + j = k; + } else + transaction_merge_and_delete_job(tr, j, k, t); + } + + assert(!j->transaction_next); + assert(!j->transaction_prev); + } + + return 0; +} + +static void transaction_drop_redundant(Transaction *tr) { + Job *j; + Iterator i; + + /* Goes through the transaction and removes all jobs of the units + * whose jobs are all noops. If not all of a unit's jobs are + * redundant, they are kept. */ + + assert(tr); + +rescan: + HASHMAP_FOREACH(j, tr->jobs, i) { + Job *k; + + LIST_FOREACH(transaction, k, j) { + + if (tr->anchor_job == k || + !job_type_is_redundant(k->type, unit_active_state(k->unit)) || + (k->unit->job && job_type_is_conflicting(k->type, k->unit->job->type))) + goto next_unit; + } + + /* log_debug("Found redundant job %s/%s, dropping.", j->unit->id, job_type_to_string(j->type)); */ + transaction_delete_job(tr, j, false); + goto rescan; + next_unit:; + } +} + +_pure_ static bool unit_matters_to_anchor(Unit *u, Job *j) { + assert(u); + assert(!j->transaction_prev); + + /* Checks whether at least one of the jobs for this unit + * matters to the anchor. */ + + LIST_FOREACH(transaction, j, j) + if (j->matters_to_anchor) + return true; + + return false; +} + +static int transaction_verify_order_one(Transaction *tr, Job *j, Job *from, unsigned generation, DBusError *e) { + Iterator i; + Unit *u; + int r; + + assert(tr); + assert(j); + assert(!j->transaction_prev); + + /* Does a recursive sweep through the ordering graph, looking + * for a cycle. If we find cycle we try to break it. */ + + /* Have we seen this before? */ + if (j->generation == generation) { + Job *k, *delete; + + /* If the marker is NULL we have been here already and + * decided the job was loop-free from here. Hence + * shortcut things and return right-away. */ + if (!j->marker) + return 0; + + /* So, the marker is not NULL and we already have been + * here. We have a cycle. Let's try to break it. We go + * backwards in our path and try to find a suitable + * job to remove. We use the marker to find our way + * back, since smart how we are we stored our way back + * in there. */ + log_warning_unit(j->unit->id, + "Found ordering cycle on %s/%s", + j->unit->id, job_type_to_string(j->type)); + + delete = NULL; + for (k = from; k; k = ((k->generation == generation && k->marker != k) ? k->marker : NULL)) { + + /* logging for j not k here here to provide consistent narrative */ + log_info_unit(j->unit->id, + "Walked on cycle path to %s/%s", + k->unit->id, job_type_to_string(k->type)); + + if (!delete && + !unit_matters_to_anchor(k->unit, k)) { + /* Ok, we can drop this one, so let's + * do so. */ + delete = k; + } + + /* Check if this in fact was the beginning of + * the cycle */ + if (k == j) + break; + } + + + if (delete) { + /* logging for j not k here here to provide consistent narrative */ + log_warning_unit(j->unit->id, + "Breaking ordering cycle by deleting job %s/%s", + delete->unit->id, job_type_to_string(delete->type)); + log_error_unit(delete->unit->id, + "Job %s/%s deleted to break ordering cycle starting with %s/%s", + delete->unit->id, job_type_to_string(delete->type), + j->unit->id, job_type_to_string(j->type)); + unit_status_printf(delete->unit, ANSI_HIGHLIGHT_RED_ON " SKIP " ANSI_HIGHLIGHT_OFF, + "Ordering cycle found, skipping %s"); + transaction_delete_unit(tr, delete->unit); + return -EAGAIN; + } + + log_error("Unable to break cycle"); + + dbus_set_error(e, BUS_ERROR_TRANSACTION_ORDER_IS_CYCLIC, + "Transaction order is cyclic. See system logs for details."); + return -ENOEXEC; + } + + /* Make the marker point to where we come from, so that we can + * find our way backwards if we want to break a cycle. We use + * a special marker for the beginning: we point to + * ourselves. */ + j->marker = from ? from : j; + j->generation = generation; + + /* We assume that the dependencies are bidirectional, and + * hence can ignore UNIT_AFTER */ + SET_FOREACH(u, j->unit->dependencies[UNIT_BEFORE], i) { + Job *o; + + /* Is there a job for this unit? */ + o = hashmap_get(tr->jobs, u); + if (!o) { + /* Ok, there is no job for this in the + * transaction, but maybe there is already one + * running? */ + o = u->job; + if (!o) + continue; + } + + r = transaction_verify_order_one(tr, o, j, generation, e); + if (r < 0) + return r; + } + + /* Ok, let's backtrack, and remember that this entry is not on + * our path anymore. */ + j->marker = NULL; + + return 0; +} + +static int transaction_verify_order(Transaction *tr, unsigned *generation, DBusError *e) { + Job *j; + int r; + Iterator i; + unsigned g; + + assert(tr); + assert(generation); + + /* Check if the ordering graph is cyclic. If it is, try to fix + * that up by dropping one of the jobs. */ + + g = (*generation)++; + + HASHMAP_FOREACH(j, tr->jobs, i) + if ((r = transaction_verify_order_one(tr, j, NULL, g, e)) < 0) + return r; + + return 0; +} + +static void transaction_collect_garbage(Transaction *tr) { + Iterator i; + Job *j; + + assert(tr); + + /* Drop jobs that are not required by any other job */ + +rescan: + HASHMAP_FOREACH(j, tr->jobs, i) { + if (tr->anchor_job == j || j->object_list) { + /* log_debug("Keeping job %s/%s because of %s/%s", */ + /* j->unit->id, job_type_to_string(j->type), */ + /* j->object_list->subject ? j->object_list->subject->unit->id : "root", */ + /* j->object_list->subject ? job_type_to_string(j->object_list->subject->type) : "root"); */ + continue; + } + + /* log_debug("Garbage collecting job %s/%s", j->unit->id, job_type_to_string(j->type)); */ + transaction_delete_job(tr, j, true); + goto rescan; + } +} + +static int transaction_is_destructive(Transaction *tr, JobMode mode, DBusError *e) { + Iterator i; + Job *j; + + assert(tr); + + /* Checks whether applying this transaction means that + * existing jobs would be replaced */ + + HASHMAP_FOREACH(j, tr->jobs, i) { + + /* Assume merged */ + assert(!j->transaction_prev); + assert(!j->transaction_next); + + if (j->unit->job && (mode == JOB_FAIL || j->unit->job->irreversible) && + !job_type_is_superset(j->type, j->unit->job->type)) { + + dbus_set_error(e, BUS_ERROR_TRANSACTION_IS_DESTRUCTIVE, "Transaction is destructive."); + return -EEXIST; + } + } + + return 0; +} + +static void transaction_minimize_impact(Transaction *tr) { + Job *j; + Iterator i; + + assert(tr); + + /* Drops all unnecessary jobs that reverse already active jobs + * or that stop a running service. */ + +rescan: + HASHMAP_FOREACH(j, tr->jobs, i) { + LIST_FOREACH(transaction, j, j) { + bool stops_running_service, changes_existing_job; + + /* If it matters, we shouldn't drop it */ + if (j->matters_to_anchor) + continue; + + /* Would this stop a running service? + * Would this change an existing job? + * If so, let's drop this entry */ + + stops_running_service = + j->type == JOB_STOP && UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(j->unit)); + + changes_existing_job = + j->unit->job && + job_type_is_conflicting(j->type, j->unit->job->type); + + if (!stops_running_service && !changes_existing_job) + continue; + + if (stops_running_service) + log_debug_unit(j->unit->id, + "%s/%s would stop a running service.", + j->unit->id, job_type_to_string(j->type)); + + if (changes_existing_job) + log_debug_unit(j->unit->id, + "%s/%s would change existing job.", + j->unit->id, job_type_to_string(j->type)); + + /* Ok, let's get rid of this */ + log_debug_unit(j->unit->id, + "Deleting %s/%s to minimize impact.", + j->unit->id, job_type_to_string(j->type)); + + transaction_delete_job(tr, j, true); + goto rescan; + } + } +} + +static int transaction_apply(Transaction *tr, Manager *m, JobMode mode) { + Iterator i; + Job *j; + int r; + + /* Moves the transaction jobs to the set of active jobs */ + + if (mode == JOB_ISOLATE) { + + /* When isolating first kill all installed jobs which + * aren't part of the new transaction */ + HASHMAP_FOREACH(j, m->jobs, i) { + assert(j->installed); + + if (hashmap_get(tr->jobs, j->unit)) + continue; + + /* Not invalidating recursively. Avoids triggering + * OnFailure= actions of dependent jobs. Also avoids + * invalidating our iterator. */ + job_finish_and_invalidate(j, JOB_CANCELED, false); + } + } + + HASHMAP_FOREACH(j, tr->jobs, i) { + /* Assume merged */ + assert(!j->transaction_prev); + assert(!j->transaction_next); + + r = hashmap_put(m->jobs, UINT32_TO_PTR(j->id), j); + if (r < 0) + goto rollback; + } + + while ((j = hashmap_steal_first(tr->jobs))) { + Job *installed_job; + + /* Clean the job dependencies */ + transaction_unlink_job(tr, j, false); + + installed_job = job_install(j); + if (installed_job != j) { + /* j has been merged into a previously installed job */ + if (tr->anchor_job == j) + tr->anchor_job = installed_job; + hashmap_remove(m->jobs, UINT32_TO_PTR(j->id)); + job_free(j); + j = installed_job; + } + + job_add_to_run_queue(j); + job_add_to_dbus_queue(j); + job_start_timer(j); + job_shutdown_magic(j); + } + + return 0; + +rollback: + + HASHMAP_FOREACH(j, tr->jobs, i) + hashmap_remove(m->jobs, UINT32_TO_PTR(j->id)); + + return r; +} + +int transaction_activate(Transaction *tr, Manager *m, JobMode mode, DBusError *e) { + Iterator i; + Job *j; + int r; + unsigned generation = 1; + + assert(tr); + + /* This applies the changes recorded in tr->jobs to + * the actual list of jobs, if possible. */ + + /* Reset the generation counter of all installed jobs. The detection of cycles + * looks at installed jobs. If they had a non-zero generation from some previous + * walk of the graph, the algorithm would break. */ + HASHMAP_FOREACH(j, m->jobs, i) + j->generation = 0; + + /* First step: figure out which jobs matter */ + transaction_find_jobs_that_matter_to_anchor(tr->anchor_job, generation++); + + /* Second step: Try not to stop any running services if + * we don't have to. Don't try to reverse running + * jobs if we don't have to. */ + if (mode == JOB_FAIL) + transaction_minimize_impact(tr); + + /* Third step: Drop redundant jobs */ + transaction_drop_redundant(tr); + + for (;;) { + /* Fourth step: Let's remove unneeded jobs that might + * be lurking. */ + if (mode != JOB_ISOLATE) + transaction_collect_garbage(tr); + + /* Fifth step: verify order makes sense and correct + * cycles if necessary and possible */ + r = transaction_verify_order(tr, &generation, e); + if (r >= 0) + break; + + if (r != -EAGAIN) { + log_warning("Requested transaction contains an unfixable cyclic ordering dependency: %s", bus_error(e, r)); + return r; + } + + /* Let's see if the resulting transaction ordering + * graph is still cyclic... */ + } + + for (;;) { + /* Sixth step: let's drop unmergeable entries if + * necessary and possible, merge entries we can + * merge */ + r = transaction_merge_jobs(tr, e); + if (r >= 0) + break; + + if (r != -EAGAIN) { + log_warning("Requested transaction contains unmergeable jobs: %s", bus_error(e, r)); + return r; + } + + /* Seventh step: an entry got dropped, let's garbage + * collect its dependencies. */ + if (mode != JOB_ISOLATE) + transaction_collect_garbage(tr); + + /* Let's see if the resulting transaction still has + * unmergeable entries ... */ + } + + /* Eights step: Drop redundant jobs again, if the merging now allows us to drop more. */ + transaction_drop_redundant(tr); + + /* Ninth step: check whether we can actually apply this */ + r = transaction_is_destructive(tr, mode, e); + if (r < 0) { + log_notice("Requested transaction contradicts existing jobs: %s", bus_error(e, r)); + return r; + } + + /* Tenth step: apply changes */ + r = transaction_apply(tr, m, mode); + if (r < 0) { + log_warning("Failed to apply transaction: %s", strerror(-r)); + return r; + } + + assert(hashmap_isempty(tr->jobs)); + + if (!hashmap_isempty(m->jobs)) { + /* Are there any jobs now? Then make sure we have the + * idle pipe around. We don't really care too much + * whether this works or not, as the idle pipe is a + * feature for cosmetics, not actually useful for + * anything beyond that. */ + + if (m->idle_pipe[0] < 0 && m->idle_pipe[1] < 0) + pipe2(m->idle_pipe, O_NONBLOCK|O_CLOEXEC); + } + + return 0; +} + +static Job* transaction_add_one_job(Transaction *tr, JobType type, Unit *unit, bool override, bool *is_new) { + Job *j, *f; + + assert(tr); + assert(unit); + + /* Looks for an existing prospective job and returns that. If + * it doesn't exist it is created and added to the prospective + * jobs list. */ + + f = hashmap_get(tr->jobs, unit); + + LIST_FOREACH(transaction, j, f) { + assert(j->unit == unit); + + if (j->type == type) { + if (is_new) + *is_new = false; + return j; + } + } + + j = job_new(unit, type); + if (!j) + return NULL; + + j->generation = 0; + j->marker = NULL; + j->matters_to_anchor = false; + j->override = override; + j->irreversible = tr->irreversible; + + LIST_PREPEND(Job, transaction, f, j); + + if (hashmap_replace(tr->jobs, unit, f) < 0) { + LIST_REMOVE(Job, transaction, f, j); + job_free(j); + return NULL; + } + + if (is_new) + *is_new = true; + + /* log_debug("Added job %s/%s to transaction.", unit->id, job_type_to_string(type)); */ + + return j; +} + +static void transaction_unlink_job(Transaction *tr, Job *j, bool delete_dependencies) { + assert(tr); + assert(j); + + if (j->transaction_prev) + j->transaction_prev->transaction_next = j->transaction_next; + else if (j->transaction_next) + hashmap_replace(tr->jobs, j->unit, j->transaction_next); + else + hashmap_remove_value(tr->jobs, j->unit, j); + + if (j->transaction_next) + j->transaction_next->transaction_prev = j->transaction_prev; + + j->transaction_prev = j->transaction_next = NULL; + + while (j->subject_list) + job_dependency_free(j->subject_list); + + while (j->object_list) { + Job *other = j->object_list->matters ? j->object_list->subject : NULL; + + job_dependency_free(j->object_list); + + if (other && delete_dependencies) { + log_debug_unit(other->unit->id, + "Deleting job %s/%s as dependency of job %s/%s", + other->unit->id, job_type_to_string(other->type), + j->unit->id, job_type_to_string(j->type)); + transaction_delete_job(tr, other, delete_dependencies); + } + } +} + +int transaction_add_job_and_dependencies( + Transaction *tr, + JobType type, + Unit *unit, + Job *by, + bool matters, + bool override, + bool conflicts, + bool ignore_requirements, + bool ignore_order, + DBusError *e) { + Job *ret; + Iterator i; + Unit *dep; + int r; + bool is_new; + + assert(tr); + assert(type < _JOB_TYPE_MAX); + assert(type < _JOB_TYPE_MAX_IN_TRANSACTION); + assert(unit); + + /* log_debug("Pulling in %s/%s from %s/%s", */ + /* unit->id, job_type_to_string(type), */ + /* by ? by->unit->id : "NA", */ + /* by ? job_type_to_string(by->type) : "NA"); */ + + if (unit->load_state != UNIT_LOADED && + unit->load_state != UNIT_ERROR && + unit->load_state != UNIT_MASKED) { + dbus_set_error(e, BUS_ERROR_LOAD_FAILED, "Unit %s is not loaded properly.", unit->id); + return -EINVAL; + } + + if (type != JOB_STOP && unit->load_state == UNIT_ERROR) { + dbus_set_error(e, BUS_ERROR_LOAD_FAILED, + "Unit %s failed to load: %s. " + "See system logs and 'systemctl status %s' for details.", + unit->id, + strerror(-unit->load_error), + unit->id); + return -EINVAL; + } + + if (type != JOB_STOP && unit->load_state == UNIT_MASKED) { + dbus_set_error(e, BUS_ERROR_MASKED, "Unit %s is masked.", unit->id); + return -EADDRNOTAVAIL; + } + + if (!unit_job_is_applicable(unit, type)) { + dbus_set_error(e, BUS_ERROR_JOB_TYPE_NOT_APPLICABLE, "Job type %s is not applicable for unit %s.", job_type_to_string(type), unit->id); + return -EBADR; + } + + /* First add the job. */ + ret = transaction_add_one_job(tr, type, unit, override, &is_new); + if (!ret) + return -ENOMEM; + + ret->ignore_order = ret->ignore_order || ignore_order; + + /* Then, add a link to the job. */ + if (by) { + if (!job_dependency_new(by, ret, matters, conflicts)) + return -ENOMEM; + } else { + /* If the job has no parent job, it is the anchor job. */ + assert(!tr->anchor_job); + tr->anchor_job = ret; + } + + if (is_new && !ignore_requirements && type != JOB_NOP) { + Set *following; + + /* If we are following some other unit, make sure we + * add all dependencies of everybody following. */ + if (unit_following_set(ret->unit, &following) > 0) { + SET_FOREACH(dep, following, i) { + r = transaction_add_job_and_dependencies(tr, type, dep, ret, false, override, false, false, ignore_order, e); + if (r < 0) { + log_warning_unit(dep->id, + "Cannot add dependency job for unit %s, ignoring: %s", + dep->id, bus_error(e, r)); + + if (e) + dbus_error_free(e); + } + } + + set_free(following); + } + + /* Finally, recursively add in all dependencies. */ + if (type == JOB_START || type == JOB_RESTART) { + SET_FOREACH(dep, ret->unit->dependencies[UNIT_REQUIRES], i) { + r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, true, override, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) + goto fail; + + if (e) + dbus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_BINDS_TO], i) { + r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, true, override, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) + goto fail; + + if (e) + dbus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_REQUIRES_OVERRIDABLE], i) { + r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, !override, override, false, false, ignore_order, e); + if (r < 0) { + log_full_unit(r == -EADDRNOTAVAIL ? LOG_DEBUG : LOG_WARNING, dep->id, + "Cannot add dependency job for unit %s, ignoring: %s", + dep->id, bus_error(e, r)); + + if (e) + dbus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_WANTS], i) { + r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, false, false, false, false, ignore_order, e); + if (r < 0) { + log_full_unit(r == -EADDRNOTAVAIL ? LOG_DEBUG : LOG_WARNING, dep->id, + "Cannot add dependency job for unit %s, ignoring: %s", + dep->id, bus_error(e, r)); + + if (e) + dbus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_REQUISITE], i) { + r = transaction_add_job_and_dependencies(tr, JOB_VERIFY_ACTIVE, dep, ret, true, override, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) + goto fail; + + if (e) + dbus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_REQUISITE_OVERRIDABLE], i) { + r = transaction_add_job_and_dependencies(tr, JOB_VERIFY_ACTIVE, dep, ret, !override, override, false, false, ignore_order, e); + if (r < 0) { + log_full_unit(r == -EADDRNOTAVAIL ? LOG_DEBUG : LOG_WARNING, dep->id, + "Cannot add dependency job for unit %s, ignoring: %s", + dep->id, bus_error(e, r)); + + if (e) + dbus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_CONFLICTS], i) { + r = transaction_add_job_and_dependencies(tr, JOB_STOP, dep, ret, true, override, true, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) + goto fail; + + if (e) + dbus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_CONFLICTED_BY], i) { + r = transaction_add_job_and_dependencies(tr, JOB_STOP, dep, ret, false, override, false, false, ignore_order, e); + if (r < 0) { + log_warning_unit(dep->id, + "Cannot add dependency job for unit %s, ignoring: %s", + dep->id, bus_error(e, r)); + + if (e) + dbus_error_free(e); + } + } + + } + + if (type == JOB_STOP || type == JOB_RESTART) { + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_REQUIRED_BY], i) { + r = transaction_add_job_and_dependencies(tr, type, dep, ret, true, override, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) + goto fail; + + if (e) + dbus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_BOUND_BY], i) { + r = transaction_add_job_and_dependencies(tr, type, dep, ret, true, override, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) + goto fail; + + if (e) + dbus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_CONSISTS_OF], i) { + r = transaction_add_job_and_dependencies(tr, type, dep, ret, true, override, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) + goto fail; + + if (e) + dbus_error_free(e); + } + } + + } + + if (type == JOB_RELOAD) { + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_PROPAGATES_RELOAD_TO], i) { + r = transaction_add_job_and_dependencies(tr, JOB_RELOAD, dep, ret, false, override, false, false, ignore_order, e); + if (r < 0) { + log_warning_unit(dep->id, + "Cannot add dependency reload job for unit %s, ignoring: %s", + dep->id, bus_error(e, r)); + + if (e) + dbus_error_free(e); + } + } + } + + /* JOB_VERIFY_STARTED require no dependency handling */ + } + + return 0; + +fail: + return r; +} + +int transaction_add_isolate_jobs(Transaction *tr, Manager *m) { + Iterator i; + Unit *u; + char *k; + int r; + + assert(tr); + assert(m); + + HASHMAP_FOREACH_KEY(u, k, m->units, i) { + + /* ignore aliases */ + if (u->id != k) + continue; + + if (u->ignore_on_isolate) + continue; + + /* No need to stop inactive jobs */ + if (UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(u)) && !u->job) + continue; + + /* Is there already something listed for this? */ + if (hashmap_get(tr->jobs, u)) + continue; + + r = transaction_add_job_and_dependencies(tr, JOB_STOP, u, tr->anchor_job, true, false, false, false, false, NULL); + if (r < 0) + log_warning_unit(u->id, + "Cannot add isolate job for unit %s, ignoring: %s", + u->id, strerror(-r)); + } + + return 0; +} + +Transaction *transaction_new(bool irreversible) { + Transaction *tr; + + tr = new0(Transaction, 1); + if (!tr) + return NULL; + + tr->jobs = hashmap_new(trivial_hash_func, trivial_compare_func); + if (!tr->jobs) { + free(tr); + return NULL; + } + + tr->irreversible = irreversible; + + return tr; +} + +void transaction_free(Transaction *tr) { + assert(hashmap_isempty(tr->jobs)); + hashmap_free(tr->jobs); + free(tr); +} diff --git a/src/core/transaction.h b/src/core/transaction.h new file mode 100644 index 0000000000..12f9194927 --- /dev/null +++ b/src/core/transaction.h @@ -0,0 +1,57 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#ifndef footransactionhfoo +#define footransactionhfoo + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Transaction Transaction; + +#include "unit.h" +#include "manager.h" +#include "job.h" +#include "hashmap.h" + +struct Transaction { + /* Jobs to be added */ + Hashmap *jobs; /* Unit object => Job object list 1:1 */ + Job *anchor_job; /* the job the user asked for */ + bool irreversible; +}; + +Transaction *transaction_new(bool irreversible); +void transaction_free(Transaction *tr); + +int transaction_add_job_and_dependencies( + Transaction *tr, + JobType type, + Unit *unit, + Job *by, + bool matters, + bool override, + bool conflicts, + bool ignore_requirements, + bool ignore_order, + DBusError *e); +int transaction_activate(Transaction *tr, Manager *m, JobMode mode, DBusError *e); +int transaction_add_isolate_jobs(Transaction *tr, Manager *m); +void transaction_abort(Transaction *tr); + +#endif diff --git a/src/core/umount.c b/src/core/umount.c new file mode 100644 index 0000000000..1e95ad70dd --- /dev/null +++ b/src/core/umount.c @@ -0,0 +1,639 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 ProFUSION embedded systems + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <fcntl.h> +#include <string.h> +#include <sys/mount.h> +#include <sys/swap.h> +#include <unistd.h> +#include <linux/loop.h> +#include <linux/dm-ioctl.h> +#include <libudev.h> + +#include "list.h" +#include "mount-setup.h" +#include "umount.h" +#include "path-util.h" +#include "util.h" +#include "virt.h" + +typedef struct MountPoint { + char *path; + dev_t devnum; + LIST_FIELDS (struct MountPoint, mount_point); +} MountPoint; + +static void mount_point_free(MountPoint **head, MountPoint *m) { + assert(head); + assert(m); + + LIST_REMOVE(MountPoint, mount_point, *head, m); + + free(m->path); + free(m); +} + +static void mount_points_list_free(MountPoint **head) { + assert(head); + + while (*head) + mount_point_free(head, *head); +} + +static int mount_points_list_get(MountPoint **head) { + FILE *proc_self_mountinfo; + char *path, *p; + unsigned int i; + int r; + + assert(head); + + if (!(proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"))) + return -errno; + + for (i = 1;; i++) { + int k; + MountPoint *m; + + path = p = NULL; + + if ((k = fscanf(proc_self_mountinfo, + "%*s " /* (1) mount id */ + "%*s " /* (2) parent id */ + "%*s " /* (3) major:minor */ + "%*s " /* (4) root */ + "%ms " /* (5) mount point */ + "%*s" /* (6) mount options */ + "%*[^-]" /* (7) optional fields */ + "- " /* (8) separator */ + "%*s " /* (9) file system type */ + "%*s" /* (10) mount source */ + "%*s" /* (11) mount options 2 */ + "%*[^\n]", /* some rubbish at the end */ + &path)) != 1) { + if (k == EOF) + break; + + log_warning("Failed to parse /proc/self/mountinfo:%u.", i); + + free(path); + continue; + } + + p = cunescape(path); + free(path); + + if (!p) { + r = -ENOMEM; + goto finish; + } + + /* Ignore mount points we can't unmount because they + * are API or because we are keeping them open (like + * /dev/console) */ + if (mount_point_is_api(p) || + mount_point_ignore(p) || + path_equal(p, "/dev/console")) { + free(p); + continue; + } + + if (!(m = new0(MountPoint, 1))) { + free(p); + r = -ENOMEM; + goto finish; + } + + m->path = p; + LIST_PREPEND(MountPoint, mount_point, *head, m); + } + + r = 0; + +finish: + fclose(proc_self_mountinfo); + + return r; +} + +static int swap_list_get(MountPoint **head) { + FILE *proc_swaps; + unsigned int i; + int r; + + assert(head); + + if (!(proc_swaps = fopen("/proc/swaps", "re"))) + return (errno == ENOENT) ? 0 : -errno; + + (void) fscanf(proc_swaps, "%*s %*s %*s %*s %*s\n"); + + for (i = 2;; i++) { + MountPoint *swap; + char *dev = NULL, *d; + int k; + + if ((k = fscanf(proc_swaps, + "%ms " /* device/file */ + "%*s " /* type of swap */ + "%*s " /* swap size */ + "%*s " /* used */ + "%*s\n", /* priority */ + &dev)) != 1) { + + if (k == EOF) + break; + + log_warning("Failed to parse /proc/swaps:%u.", i); + + free(dev); + continue; + } + + if (endswith(dev, "(deleted)")) { + free(dev); + continue; + } + + d = cunescape(dev); + free(dev); + + if (!d) { + r = -ENOMEM; + goto finish; + } + + if (!(swap = new0(MountPoint, 1))) { + free(d); + r = -ENOMEM; + goto finish; + } + + swap->path = d; + LIST_PREPEND(MountPoint, mount_point, *head, swap); + } + + r = 0; + +finish: + fclose(proc_swaps); + + return r; +} + +static int loopback_list_get(MountPoint **head) { + int r; + struct udev *udev; + struct udev_enumerate *e = NULL; + struct udev_list_entry *item = NULL, *first = NULL; + + assert(head); + + if (!(udev = udev_new())) { + r = -ENOMEM; + goto finish; + } + + if (!(e = udev_enumerate_new(udev))) { + r = -ENOMEM; + goto finish; + } + + if (udev_enumerate_add_match_subsystem(e, "block") < 0 || + udev_enumerate_add_match_sysname(e, "loop*") < 0 || + udev_enumerate_add_match_sysattr(e, "loop/backing_file", NULL) < 0) { + r = -EIO; + goto finish; + } + + if (udev_enumerate_scan_devices(e) < 0) { + r = -EIO; + goto finish; + } + + first = udev_enumerate_get_list_entry(e); + udev_list_entry_foreach(item, first) { + MountPoint *lb; + struct udev_device *d; + char *loop; + const char *dn; + + if (!(d = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item)))) { + r = -ENOMEM; + goto finish; + } + + if (!(dn = udev_device_get_devnode(d))) { + udev_device_unref(d); + continue; + } + + loop = strdup(dn); + udev_device_unref(d); + + if (!loop) { + r = -ENOMEM; + goto finish; + } + + if (!(lb = new0(MountPoint, 1))) { + free(loop); + r = -ENOMEM; + goto finish; + } + + lb->path = loop; + LIST_PREPEND(MountPoint, mount_point, *head, lb); + } + + r = 0; + +finish: + if (e) + udev_enumerate_unref(e); + + if (udev) + udev_unref(udev); + + return r; +} + +static int dm_list_get(MountPoint **head) { + int r; + struct udev *udev; + struct udev_enumerate *e = NULL; + struct udev_list_entry *item = NULL, *first = NULL; + + assert(head); + + if (!(udev = udev_new())) { + r = -ENOMEM; + goto finish; + } + + if (!(e = udev_enumerate_new(udev))) { + r = -ENOMEM; + goto finish; + } + + if (udev_enumerate_add_match_subsystem(e, "block") < 0 || + udev_enumerate_add_match_sysname(e, "dm-*") < 0) { + r = -EIO; + goto finish; + } + + if (udev_enumerate_scan_devices(e) < 0) { + r = -EIO; + goto finish; + } + + first = udev_enumerate_get_list_entry(e); + + udev_list_entry_foreach(item, first) { + MountPoint *m; + struct udev_device *d; + dev_t devnum; + char *node; + const char *dn; + + if (!(d = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item)))) { + r = -ENOMEM; + goto finish; + } + + devnum = udev_device_get_devnum(d); + dn = udev_device_get_devnode(d); + + if (major(devnum) == 0 || !dn) { + udev_device_unref(d); + continue; + } + + node = strdup(dn); + udev_device_unref(d); + + if (!node) { + r = -ENOMEM; + goto finish; + } + + if (!(m = new(MountPoint, 1))) { + free(node); + r = -ENOMEM; + goto finish; + } + + m->path = node; + m->devnum = devnum; + LIST_PREPEND(MountPoint, mount_point, *head, m); + } + + r = 0; + +finish: + if (e) + udev_enumerate_unref(e); + + if (udev) + udev_unref(udev); + + return r; +} + +static int delete_loopback(const char *device) { + int fd, r; + + if ((fd = open(device, O_RDONLY|O_CLOEXEC)) < 0) + return errno == ENOENT ? 0 : -errno; + + r = ioctl(fd, LOOP_CLR_FD, 0); + close_nointr_nofail(fd); + + if (r >= 0) + return 1; + + /* ENXIO: not bound, so no error */ + if (errno == ENXIO) + return 0; + + return -errno; +} + +static int delete_dm(dev_t devnum) { + _cleanup_close_ int fd = -1; + int r; + struct dm_ioctl dm = { + .version = {DM_VERSION_MAJOR, + DM_VERSION_MINOR, + DM_VERSION_PATCHLEVEL}, + .data_size = sizeof(dm), + .dev = devnum, + }; + + assert(major(devnum) != 0); + + fd = open("/dev/mapper/control", O_RDWR|O_CLOEXEC); + if (fd < 0) + return -errno; + + r = ioctl(fd, DM_DEV_REMOVE, &dm); + return r >= 0 ? 0 : -errno; +} + +static int mount_points_list_umount(MountPoint **head, bool *changed, bool log_error) { + MountPoint *m, *n; + int n_failed = 0; + + assert(head); + + LIST_FOREACH_SAFE(mount_point, m, n, *head) { + + /* If we are in a container, don't attempt to + read-only mount anything as that brings no real + benefits, but might confuse the host, as we remount + the superblock here, not the bind mound. */ + if (detect_container(NULL) <= 0) { + /* We always try to remount directories + * read-only first, before we go on and umount + * them. + * + * Mount points can be stacked. If a mount + * point is stacked below / or /usr, we + * cannot umount or remount it directly, + * since there is no way to refer to the + * underlying mount. There's nothing we can do + * about it for the general case, but we can + * do something about it if it is aliased + * somehwere else via a bind mount. If we + * explicitly remount the super block of that + * alias read-only we hence should be + * relatively safe regarding keeping the fs we + * can otherwise not see dirty. */ + mount(NULL, m->path, NULL, MS_REMOUNT|MS_RDONLY, NULL); + } + + /* Skip / and /usr since we cannot unmount that + * anyway, since we are running from it. They have + * already been remounted ro. */ + if (path_equal(m->path, "/") +#ifndef HAVE_SPLIT_USR + || path_equal(m->path, "/usr") +#endif + ) + continue; + + /* Trying to umount. We don't force here since we rely + * on busy NFS and FUSE file systems to return EBUSY + * until we closed everything on top of them. */ + log_info("Unmounting %s.", m->path); + if (umount2(m->path, 0) == 0) { + if (changed) + *changed = true; + + mount_point_free(head, m); + } else if (log_error) { + log_warning("Could not unmount %s: %m", m->path); + n_failed++; + } + } + + return n_failed; +} + +static int swap_points_list_off(MountPoint **head, bool *changed) { + MountPoint *m, *n; + int n_failed = 0; + + assert(head); + + LIST_FOREACH_SAFE(mount_point, m, n, *head) { + log_info("Deactivating swap %s.", m->path); + if (swapoff(m->path) == 0) { + if (changed) + *changed = true; + + mount_point_free(head, m); + } else { + log_warning("Could not deactivate swap %s: %m", m->path); + n_failed++; + } + } + + return n_failed; +} + +static int loopback_points_list_detach(MountPoint **head, bool *changed) { + MountPoint *m, *n; + int n_failed = 0, k; + struct stat root_st; + + assert(head); + + k = lstat("/", &root_st); + + LIST_FOREACH_SAFE(mount_point, m, n, *head) { + int r; + struct stat loopback_st; + + if (k >= 0 && + major(root_st.st_dev) != 0 && + lstat(m->path, &loopback_st) >= 0 && + root_st.st_dev == loopback_st.st_rdev) { + n_failed ++; + continue; + } + + log_info("Detaching loopback %s.", m->path); + r = delete_loopback(m->path); + if (r >= 0) { + if (r > 0 && changed) + *changed = true; + + mount_point_free(head, m); + } else { + log_warning("Could not detach loopback %s: %m", m->path); + n_failed++; + } + } + + return n_failed; +} + +static int dm_points_list_detach(MountPoint **head, bool *changed) { + MountPoint *m, *n; + int n_failed = 0, k; + struct stat root_st; + + assert(head); + + k = lstat("/", &root_st); + + LIST_FOREACH_SAFE(mount_point, m, n, *head) { + int r; + + if (k >= 0 && + major(root_st.st_dev) != 0 && + root_st.st_dev == m->devnum) { + n_failed ++; + continue; + } + + log_info("Detaching DM %u:%u.", major(m->devnum), minor(m->devnum)); + r = delete_dm(m->devnum); + if (r >= 0) { + if (changed) + *changed = true; + + mount_point_free(head, m); + } else { + log_warning("Could not detach DM %s: %m", m->path); + n_failed++; + } + } + + return n_failed; +} + +int umount_all(bool *changed) { + int r; + bool umount_changed; + LIST_HEAD(MountPoint, mp_list_head); + + LIST_HEAD_INIT(MountPoint, mp_list_head); + r = mount_points_list_get(&mp_list_head); + if (r < 0) + goto end; + + /* retry umount, until nothing can be umounted anymore */ + do { + umount_changed = false; + + mount_points_list_umount(&mp_list_head, &umount_changed, false); + if (umount_changed) + *changed = true; + + } while (umount_changed); + + /* umount one more time with logging enabled */ + r = mount_points_list_umount(&mp_list_head, &umount_changed, true); + if (r <= 0) + goto end; + + end: + mount_points_list_free(&mp_list_head); + + return r; +} + +int swapoff_all(bool *changed) { + int r; + LIST_HEAD(MountPoint, swap_list_head); + + LIST_HEAD_INIT(MountPoint, swap_list_head); + + r = swap_list_get(&swap_list_head); + if (r < 0) + goto end; + + r = swap_points_list_off(&swap_list_head, changed); + + end: + mount_points_list_free(&swap_list_head); + + return r; +} + +int loopback_detach_all(bool *changed) { + int r; + LIST_HEAD(MountPoint, loopback_list_head); + + LIST_HEAD_INIT(MountPoint, loopback_list_head); + + r = loopback_list_get(&loopback_list_head); + if (r < 0) + goto end; + + r = loopback_points_list_detach(&loopback_list_head, changed); + + end: + mount_points_list_free(&loopback_list_head); + + return r; +} + +int dm_detach_all(bool *changed) { + int r; + LIST_HEAD(MountPoint, dm_list_head); + + LIST_HEAD_INIT(MountPoint, dm_list_head); + + r = dm_list_get(&dm_list_head); + if (r < 0) + goto end; + + r = dm_points_list_detach(&dm_list_head, changed); + + end: + mount_points_list_free(&dm_list_head); + + return r; +} diff --git a/src/core/umount.h b/src/core/umount.h new file mode 100644 index 0000000000..8439ffe58f --- /dev/null +++ b/src/core/umount.h @@ -0,0 +1,30 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 ProFUSION embedded systems + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int umount_all(bool *changed); + +int swapoff_all(bool *changed); + +int loopback_detach_all(bool *changed); + +int dm_detach_all(bool *changed); diff --git a/src/core/unit-printf.c b/src/core/unit-printf.c new file mode 100644 index 0000000000..85a05b872a --- /dev/null +++ b/src/core/unit-printf.c @@ -0,0 +1,329 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "systemd/sd-id128.h" +#include "unit.h" +#include "specifier.h" +#include "path-util.h" +#include "strv.h" +#include "unit-name.h" +#include "unit-printf.h" +#include "macro.h" + +static char *specifier_prefix_and_instance(char specifier, void *data, void *userdata) { + Unit *u = userdata; + assert(u); + + return unit_name_to_prefix_and_instance(u->id); +} + +static char *specifier_prefix(char specifier, void *data, void *userdata) { + Unit *u = userdata; + assert(u); + + return unit_name_to_prefix(u->id); +} + +static char *specifier_prefix_unescaped(char specifier, void *data, void *userdata) { + Unit *u = userdata; + char *p, *r; + + assert(u); + + p = unit_name_to_prefix(u->id); + if (!p) + return NULL; + + r = unit_name_unescape(p); + free(p); + + return r; +} + +static char *specifier_instance_unescaped(char specifier, void *data, void *userdata) { + Unit *u = userdata; + assert(u); + + if (u->instance) + return unit_name_unescape(u->instance); + + return strdup(""); +} + +static char *specifier_filename(char specifier, void *data, void *userdata) { + Unit *u = userdata; + assert(u); + + if (u->instance) + return unit_name_path_unescape(u->instance); + + return unit_name_to_path(u->id); +} + +static char *specifier_cgroup(char specifier, void *data, void *userdata) { + Unit *u = userdata; + assert(u); + + return unit_default_cgroup_path(u); +} + +static char *specifier_cgroup_root(char specifier, void *data, void *userdata) { + Unit *u = userdata; + char *p; + assert(u); + + if (specifier == 'r') + return strdup(u->manager->cgroup_hierarchy); + + if (path_get_parent(u->manager->cgroup_hierarchy, &p) < 0) + return strdup(""); + + if (streq(p, "/")) { + free(p); + return strdup(""); + } + + return p; +} + +static char *specifier_runtime(char specifier, void *data, void *userdata) { + Unit *u = userdata; + assert(u); + + if (u->manager->running_as == SYSTEMD_USER) { + const char *e; + + e = getenv("XDG_RUNTIME_DIR"); + if (e) + return strdup(e); + } + + return strdup("/run"); +} + +static char *specifier_user_name(char specifier, void *data, void *userdata) { + Unit *u = userdata; + ExecContext *c; + int r; + const char *username; + _cleanup_free_ char *tmp = NULL; + uid_t uid; + char *printed = NULL; + + assert(u); + + c = unit_get_exec_context(u); + + if (c && c->user) + username = c->user; + else + /* get USER env from env or our own uid */ + username = tmp = getusername_malloc(); + + /* fish username from passwd */ + r = get_user_creds(&username, &uid, NULL, NULL, NULL); + if (r < 0) + return NULL; + + switch (specifier) { + case 'U': + if (asprintf(&printed, "%d", uid) < 0) + return NULL; + break; + case 'u': + printed = strdup(username); + break; + } + + return printed; +} + +static char *specifier_user_home(char specifier, void *data, void *userdata) { + Unit *u = userdata; + ExecContext *c; + int r; + const char *username, *home; + + assert(u); + + c = unit_get_exec_context(u); + + /* return HOME if set, otherwise from passwd */ + if (!c || !c->user) { + char *h; + + r = get_home_dir(&h); + if (r < 0) + return NULL; + + return h; + } + + username = c->user; + r = get_user_creds(&username, NULL, NULL, &home, NULL); + if (r < 0) + return NULL; + + return strdup(home); +} + +static char *specifier_user_shell(char specifier, void *data, void *userdata) { + Unit *u = userdata; + ExecContext *c; + int r; + const char *username, *shell; + char *ret; + + assert(u); + + c = unit_get_exec_context(u); + + if (c && c->user) + username = c->user; + else + username = "root"; + + /* return /bin/sh for root, otherwise the value from passwd */ + r = get_user_creds(&username, NULL, NULL, NULL, &shell); + if (r < 0) { + log_warning_unit(u->id, + "Failed to determine shell: %s", + strerror(-r)); + return NULL; + } + + if (!path_is_absolute(shell)) { + log_warning_unit(u->id, + "Shell %s is not absolute, ignoring.", + shell); + } + + ret = strdup(shell); + if (!ret) + log_oom(); + + return ret; +} + +char *unit_name_printf(Unit *u, const char* format) { + + /* + * This will use the passed string as format string and + * replace the following specifiers: + * + * %n: the full id of the unit (foo@bar.waldo) + * %N: the id of the unit without the suffix (foo@bar) + * %p: the prefix (foo) + * %i: the instance (bar) + */ + + const Specifier table[] = { + { 'n', specifier_string, u->id }, + { 'N', specifier_prefix_and_instance, NULL }, + { 'p', specifier_prefix, NULL }, + { 'i', specifier_string, u->instance }, + { 0, NULL, NULL } + }; + + assert(u); + assert(format); + + return specifier_printf(format, table, u); +} + +char *unit_full_printf(Unit *u, const char *format) { + + /* This is similar to unit_name_printf() but also supports + * unescaping. Also, adds a couple of additional codes: + * + * %f the the instance if set, otherwise the id + * %c cgroup path of unit + * %r root cgroup path of this systemd instance (e.g. "/user/lennart/shared/systemd-4711") + * %R parent of root cgroup path (e.g. "/usr/lennart/shared") + * %t the runtime directory to place sockets in (e.g. "/run" or $XDG_RUNTIME_DIR) + * %U the UID of the configured user or running user + * %u the username of the configured user or running user + * %h the homedir of the configured user or running user + * %s the shell of the configured user or running user + * %m the machine ID of the running system + * %H the host name of the running system + * %b the boot ID of the running system + */ + + const Specifier table[] = { + { 'n', specifier_string, u->id }, + { 'N', specifier_prefix_and_instance, NULL }, + { 'p', specifier_prefix, NULL }, + { 'P', specifier_prefix_unescaped, NULL }, + { 'i', specifier_string, u->instance }, + { 'I', specifier_instance_unescaped, NULL }, + + { 'f', specifier_filename, NULL }, + { 'c', specifier_cgroup, NULL }, + { 'r', specifier_cgroup_root, NULL }, + { 'R', specifier_cgroup_root, NULL }, + { 't', specifier_runtime, NULL }, + { 'U', specifier_user_name, NULL }, + { 'u', specifier_user_name, NULL }, + { 'h', specifier_user_home, NULL }, + { 's', specifier_user_shell, NULL }, + + { 'm', specifier_machine_id, NULL }, + { 'H', specifier_host_name, NULL }, + { 'b', specifier_boot_id, NULL }, + { 0, NULL, NULL } + }; + + assert(format); + + return specifier_printf(format, table, u); +} + +char **unit_full_printf_strv(Unit *u, char **l) { + size_t n; + char **r, **i, **j; + + /* Applies unit_full_printf to every entry in l */ + + assert(u); + + n = strv_length(l); + r = new(char*, n+1); + if (!r) + return NULL; + + for (i = l, j = r; *i; i++, j++) { + *j = unit_full_printf(u, *i); + if (!*j) + goto fail; + } + + *j = NULL; + return r; + +fail: + for (j--; j >= r; j--) + free(*j); + + free(r); + + return NULL; +} diff --git a/src/core/unit-printf.h b/src/core/unit-printf.h new file mode 100644 index 0000000000..d2f4ccd178 --- /dev/null +++ b/src/core/unit-printf.h @@ -0,0 +1,28 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "unit.h" + +char *unit_name_printf(Unit *u, const char* text); +char *unit_full_printf(Unit *u, const char *text); +char **unit_full_printf_strv(Unit *u, char **l); diff --git a/src/core/unit.c b/src/core/unit.c new file mode 100644 index 0000000000..9b36b225fa --- /dev/null +++ b/src/core/unit.c @@ -0,0 +1,3089 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <assert.h> +#include <errno.h> +#include <string.h> +#include <sys/epoll.h> +#include <sys/timerfd.h> +#include <sys/poll.h> +#include <stdlib.h> +#include <unistd.h> +#include <sys/stat.h> + +#include "systemd/sd-id128.h" +#include "systemd/sd-messages.h" +#include "set.h" +#include "unit.h" +#include "macro.h" +#include "strv.h" +#include "path-util.h" +#include "load-fragment.h" +#include "load-dropin.h" +#include "log.h" +#include "unit-name.h" +#include "dbus-unit.h" +#include "special.h" +#include "cgroup-util.h" +#include "missing.h" +#include "cgroup-attr.h" +#include "mkdir.h" +#include "label.h" +#include "fileio-label.h" +#include "bus-errors.h" + +const UnitVTable * const unit_vtable[_UNIT_TYPE_MAX] = { + [UNIT_SERVICE] = &service_vtable, + [UNIT_TIMER] = &timer_vtable, + [UNIT_SOCKET] = &socket_vtable, + [UNIT_TARGET] = &target_vtable, + [UNIT_DEVICE] = &device_vtable, + [UNIT_MOUNT] = &mount_vtable, + [UNIT_AUTOMOUNT] = &automount_vtable, + [UNIT_SNAPSHOT] = &snapshot_vtable, + [UNIT_SWAP] = &swap_vtable, + [UNIT_PATH] = &path_vtable +}; + +Unit *unit_new(Manager *m, size_t size) { + Unit *u; + + assert(m); + assert(size >= sizeof(Unit)); + + u = malloc0(size); + if (!u) + return NULL; + + u->names = set_new(string_hash_func, string_compare_func); + if (!u->names) { + free(u); + return NULL; + } + + u->manager = m; + u->type = _UNIT_TYPE_INVALID; + u->deserialized_job = _JOB_TYPE_INVALID; + u->default_dependencies = true; + u->unit_file_state = _UNIT_FILE_STATE_INVALID; + + return u; +} + +bool unit_has_name(Unit *u, const char *name) { + assert(u); + assert(name); + + return !!set_get(u->names, (char*) name); +} + +int unit_add_name(Unit *u, const char *text) { + UnitType t; + char *s, *i = NULL; + int r; + + assert(u); + assert(text); + + if (unit_name_is_template(text)) { + if (!u->instance) + return -EINVAL; + + s = unit_name_replace_instance(text, u->instance); + } else + s = strdup(text); + + if (!s) + return -ENOMEM; + + if (!unit_name_is_valid(s, false)) { + r = -EINVAL; + goto fail; + } + + assert_se((t = unit_name_to_type(s)) >= 0); + + if (u->type != _UNIT_TYPE_INVALID && t != u->type) { + r = -EINVAL; + goto fail; + } + + if ((r = unit_name_to_instance(s, &i)) < 0) + goto fail; + + if (i && unit_vtable[t]->no_instances) { + r = -EINVAL; + goto fail; + } + + /* Ensure that this unit is either instanced or not instanced, + * but not both. */ + if (u->type != _UNIT_TYPE_INVALID && !u->instance != !i) { + r = -EINVAL; + goto fail; + } + + if (unit_vtable[t]->no_alias && + !set_isempty(u->names) && + !set_get(u->names, s)) { + r = -EEXIST; + goto fail; + } + + if (hashmap_size(u->manager->units) >= MANAGER_MAX_NAMES) { + r = -E2BIG; + goto fail; + } + + if ((r = set_put(u->names, s)) < 0) { + if (r == -EEXIST) + r = 0; + goto fail; + } + + if ((r = hashmap_put(u->manager->units, s, u)) < 0) { + set_remove(u->names, s); + goto fail; + } + + if (u->type == _UNIT_TYPE_INVALID) { + + u->type = t; + u->id = s; + u->instance = i; + + LIST_PREPEND(Unit, units_by_type, u->manager->units_by_type[t], u); + + if (UNIT_VTABLE(u)->init) + UNIT_VTABLE(u)->init(u); + } else + free(i); + + unit_add_to_dbus_queue(u); + return 0; + +fail: + free(s); + free(i); + + return r; +} + +int unit_choose_id(Unit *u, const char *name) { + char *s, *t = NULL, *i; + int r; + + assert(u); + assert(name); + + if (unit_name_is_template(name)) { + + if (!u->instance) + return -EINVAL; + + if (!(t = unit_name_replace_instance(name, u->instance))) + return -ENOMEM; + + name = t; + } + + /* Selects one of the names of this unit as the id */ + s = set_get(u->names, (char*) name); + free(t); + + if (!s) + return -ENOENT; + + if ((r = unit_name_to_instance(s, &i)) < 0) + return r; + + u->id = s; + + free(u->instance); + u->instance = i; + + unit_add_to_dbus_queue(u); + + return 0; +} + +int unit_set_description(Unit *u, const char *description) { + char *s; + + assert(u); + + if (!(s = strdup(description))) + return -ENOMEM; + + free(u->description); + u->description = s; + + unit_add_to_dbus_queue(u); + return 0; +} + +bool unit_check_gc(Unit *u) { + assert(u); + + if (u->load_state == UNIT_STUB) + return true; + + if (UNIT_VTABLE(u)->no_gc) + return true; + + if (u->no_gc) + return true; + + if (u->job) + return true; + + if (u->nop_job) + return true; + + if (unit_active_state(u) != UNIT_INACTIVE) + return true; + + if (u->refs) + return true; + + if (UNIT_VTABLE(u)->check_gc) + if (UNIT_VTABLE(u)->check_gc(u)) + return true; + + return false; +} + +void unit_add_to_load_queue(Unit *u) { + assert(u); + assert(u->type != _UNIT_TYPE_INVALID); + + if (u->load_state != UNIT_STUB || u->in_load_queue) + return; + + LIST_PREPEND(Unit, load_queue, u->manager->load_queue, u); + u->in_load_queue = true; +} + +void unit_add_to_cleanup_queue(Unit *u) { + assert(u); + + if (u->in_cleanup_queue) + return; + + LIST_PREPEND(Unit, cleanup_queue, u->manager->cleanup_queue, u); + u->in_cleanup_queue = true; +} + +void unit_add_to_gc_queue(Unit *u) { + assert(u); + + if (u->in_gc_queue || u->in_cleanup_queue) + return; + + if (unit_check_gc(u)) + return; + + LIST_PREPEND(Unit, gc_queue, u->manager->gc_queue, u); + u->in_gc_queue = true; + + u->manager->n_in_gc_queue ++; + + if (u->manager->gc_queue_timestamp <= 0) + u->manager->gc_queue_timestamp = now(CLOCK_MONOTONIC); +} + +void unit_add_to_dbus_queue(Unit *u) { + assert(u); + assert(u->type != _UNIT_TYPE_INVALID); + + if (u->load_state == UNIT_STUB || u->in_dbus_queue) + return; + + /* Shortcut things if nobody cares */ + if (!bus_has_subscriber(u->manager)) { + u->sent_dbus_new_signal = true; + return; + } + + LIST_PREPEND(Unit, dbus_queue, u->manager->dbus_unit_queue, u); + u->in_dbus_queue = true; +} + +static void bidi_set_free(Unit *u, Set *s) { + Iterator i; + Unit *other; + + assert(u); + + /* Frees the set and makes sure we are dropped from the + * inverse pointers */ + + SET_FOREACH(other, s, i) { + UnitDependency d; + + for (d = 0; d < _UNIT_DEPENDENCY_MAX; d++) + set_remove(other->dependencies[d], u); + + unit_add_to_gc_queue(other); + } + + set_free(s); +} + +void unit_free(Unit *u) { + UnitDependency d; + Iterator i; + char *t; + + assert(u); + + bus_unit_send_removed_signal(u); + + if (u->load_state != UNIT_STUB) + if (UNIT_VTABLE(u)->done) + UNIT_VTABLE(u)->done(u); + + SET_FOREACH(t, u->names, i) + hashmap_remove_value(u->manager->units, t, u); + + if (u->job) { + Job *j = u->job; + job_uninstall(j); + job_free(j); + } + + if (u->nop_job) { + Job *j = u->nop_job; + job_uninstall(j); + job_free(j); + } + + for (d = 0; d < _UNIT_DEPENDENCY_MAX; d++) + bidi_set_free(u, u->dependencies[d]); + + if (u->requires_mounts_for) { + LIST_REMOVE(Unit, has_requires_mounts_for, u->manager->has_requires_mounts_for, u); + strv_free(u->requires_mounts_for); + } + + if (u->type != _UNIT_TYPE_INVALID) + LIST_REMOVE(Unit, units_by_type, u->manager->units_by_type[u->type], u); + + if (u->in_load_queue) + LIST_REMOVE(Unit, load_queue, u->manager->load_queue, u); + + if (u->in_dbus_queue) + LIST_REMOVE(Unit, dbus_queue, u->manager->dbus_unit_queue, u); + + if (u->in_cleanup_queue) + LIST_REMOVE(Unit, cleanup_queue, u->manager->cleanup_queue, u); + + if (u->in_gc_queue) { + LIST_REMOVE(Unit, gc_queue, u->manager->gc_queue, u); + u->manager->n_in_gc_queue--; + } + + cgroup_bonding_free_list(u->cgroup_bondings, u->manager->n_reloading <= 0); + cgroup_attribute_free_list(u->cgroup_attributes); + + free(u->description); + strv_free(u->documentation); + free(u->fragment_path); + free(u->source_path); + strv_free(u->dropin_paths); + free(u->instance); + + set_free_free(u->names); + + condition_free_list(u->conditions); + + while (u->refs) + unit_ref_unset(u->refs); + + free(u); +} + +UnitActiveState unit_active_state(Unit *u) { + assert(u); + + if (u->load_state == UNIT_MERGED) + return unit_active_state(unit_follow_merge(u)); + + /* After a reload it might happen that a unit is not correctly + * loaded but still has a process around. That's why we won't + * shortcut failed loading to UNIT_INACTIVE_FAILED. */ + + return UNIT_VTABLE(u)->active_state(u); +} + +const char* unit_sub_state_to_string(Unit *u) { + assert(u); + + return UNIT_VTABLE(u)->sub_state_to_string(u); +} + +static void complete_move(Set **s, Set **other) { + assert(s); + assert(other); + + if (!*other) + return; + + if (*s) + set_move(*s, *other); + else { + *s = *other; + *other = NULL; + } +} + +static void merge_names(Unit *u, Unit *other) { + char *t; + Iterator i; + + assert(u); + assert(other); + + complete_move(&u->names, &other->names); + + set_free_free(other->names); + other->names = NULL; + other->id = NULL; + + SET_FOREACH(t, u->names, i) + assert_se(hashmap_replace(u->manager->units, t, u) == 0); +} + +static void merge_dependencies(Unit *u, Unit *other, UnitDependency d) { + Iterator i; + Unit *back; + int r; + + assert(u); + assert(other); + assert(d < _UNIT_DEPENDENCY_MAX); + + /* Fix backwards pointers */ + SET_FOREACH(back, other->dependencies[d], i) { + UnitDependency k; + + for (k = 0; k < _UNIT_DEPENDENCY_MAX; k++) + if ((r = set_remove_and_put(back->dependencies[k], other, u)) < 0) { + + if (r == -EEXIST) + set_remove(back->dependencies[k], other); + else + assert(r == -ENOENT); + } + } + + complete_move(&u->dependencies[d], &other->dependencies[d]); + + set_free(other->dependencies[d]); + other->dependencies[d] = NULL; +} + +int unit_merge(Unit *u, Unit *other) { + UnitDependency d; + + assert(u); + assert(other); + assert(u->manager == other->manager); + assert(u->type != _UNIT_TYPE_INVALID); + + other = unit_follow_merge(other); + + if (other == u) + return 0; + + if (u->type != other->type) + return -EINVAL; + + if (!u->instance != !other->instance) + return -EINVAL; + + if (other->load_state != UNIT_STUB && + other->load_state != UNIT_ERROR) + return -EEXIST; + + if (other->job) + return -EEXIST; + + if (other->nop_job) + return -EEXIST; + + if (!UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(other))) + return -EEXIST; + + /* Merge names */ + merge_names(u, other); + + /* Redirect all references */ + while (other->refs) + unit_ref_set(other->refs, u); + + /* Merge dependencies */ + for (d = 0; d < _UNIT_DEPENDENCY_MAX; d++) + merge_dependencies(u, other, d); + + other->load_state = UNIT_MERGED; + other->merged_into = u; + + /* If there is still some data attached to the other node, we + * don't need it anymore, and can free it. */ + if (other->load_state != UNIT_STUB) + if (UNIT_VTABLE(other)->done) + UNIT_VTABLE(other)->done(other); + + unit_add_to_dbus_queue(u); + unit_add_to_cleanup_queue(other); + + return 0; +} + +int unit_merge_by_name(Unit *u, const char *name) { + Unit *other; + int r; + char *s = NULL; + + assert(u); + assert(name); + + if (unit_name_is_template(name)) { + if (!u->instance) + return -EINVAL; + + if (!(s = unit_name_replace_instance(name, u->instance))) + return -ENOMEM; + + name = s; + } + + if (!(other = manager_get_unit(u->manager, name))) + r = unit_add_name(u, name); + else + r = unit_merge(u, other); + + free(s); + return r; +} + +Unit* unit_follow_merge(Unit *u) { + assert(u); + + while (u->load_state == UNIT_MERGED) + assert_se(u = u->merged_into); + + return u; +} + +int unit_add_exec_dependencies(Unit *u, ExecContext *c) { + int r; + + assert(u); + assert(c); + + if (c->std_output != EXEC_OUTPUT_KMSG && + c->std_output != EXEC_OUTPUT_SYSLOG && + c->std_output != EXEC_OUTPUT_JOURNAL && + c->std_output != EXEC_OUTPUT_KMSG_AND_CONSOLE && + c->std_output != EXEC_OUTPUT_SYSLOG_AND_CONSOLE && + c->std_output != EXEC_OUTPUT_JOURNAL_AND_CONSOLE && + c->std_error != EXEC_OUTPUT_KMSG && + c->std_error != EXEC_OUTPUT_SYSLOG && + c->std_error != EXEC_OUTPUT_JOURNAL && + c->std_error != EXEC_OUTPUT_KMSG_AND_CONSOLE && + c->std_error != EXEC_OUTPUT_JOURNAL_AND_CONSOLE && + c->std_error != EXEC_OUTPUT_SYSLOG_AND_CONSOLE) + return 0; + + /* If syslog or kernel logging is requested, make sure our own + * logging daemon is run first. */ + + if (u->manager->running_as == SYSTEMD_SYSTEM) { + r = unit_add_dependency_by_name(u, UNIT_AFTER, SPECIAL_JOURNALD_SOCKET, NULL, true); + if (r < 0) + return r; + } + + return 0; +} + +const char *unit_description(Unit *u) { + assert(u); + + if (u->description) + return u->description; + + return strna(u->id); +} + +void unit_dump(Unit *u, FILE *f, const char *prefix) { + char *t, **j; + UnitDependency d; + Iterator i; + char *p2; + const char *prefix2; + char + timestamp1[FORMAT_TIMESTAMP_MAX], + timestamp2[FORMAT_TIMESTAMP_MAX], + timestamp3[FORMAT_TIMESTAMP_MAX], + timestamp4[FORMAT_TIMESTAMP_MAX], + timespan[FORMAT_TIMESPAN_MAX]; + Unit *following; + + assert(u); + assert(u->type >= 0); + + if (!prefix) + prefix = ""; + p2 = strappend(prefix, "\t"); + prefix2 = p2 ? p2 : prefix; + + fprintf(f, + "%s-> Unit %s:\n" + "%s\tDescription: %s\n" + "%s\tInstance: %s\n" + "%s\tUnit Load State: %s\n" + "%s\tUnit Active State: %s\n" + "%s\tInactive Exit Timestamp: %s\n" + "%s\tActive Enter Timestamp: %s\n" + "%s\tActive Exit Timestamp: %s\n" + "%s\tInactive Enter Timestamp: %s\n" + "%s\tGC Check Good: %s\n" + "%s\tNeed Daemon Reload: %s\n", + prefix, u->id, + prefix, unit_description(u), + prefix, strna(u->instance), + prefix, unit_load_state_to_string(u->load_state), + prefix, unit_active_state_to_string(unit_active_state(u)), + prefix, strna(format_timestamp(timestamp1, sizeof(timestamp1), u->inactive_exit_timestamp.realtime)), + prefix, strna(format_timestamp(timestamp2, sizeof(timestamp2), u->active_enter_timestamp.realtime)), + prefix, strna(format_timestamp(timestamp3, sizeof(timestamp3), u->active_exit_timestamp.realtime)), + prefix, strna(format_timestamp(timestamp4, sizeof(timestamp4), u->inactive_enter_timestamp.realtime)), + prefix, yes_no(unit_check_gc(u)), + prefix, yes_no(unit_need_daemon_reload(u))); + + SET_FOREACH(t, u->names, i) + fprintf(f, "%s\tName: %s\n", prefix, t); + + STRV_FOREACH(j, u->documentation) + fprintf(f, "%s\tDocumentation: %s\n", prefix, *j); + + if ((following = unit_following(u))) + fprintf(f, "%s\tFollowing: %s\n", prefix, following->id); + + if (u->fragment_path) + fprintf(f, "%s\tFragment Path: %s\n", prefix, u->fragment_path); + + if (u->source_path) + fprintf(f, "%s\tSource Path: %s\n", prefix, u->source_path); + + STRV_FOREACH(j, u->dropin_paths) + fprintf(f, "%s\tDropIn Path: %s\n", prefix, *j); + + if (u->job_timeout > 0) + fprintf(f, "%s\tJob Timeout: %s\n", prefix, format_timespan(timespan, sizeof(timespan), u->job_timeout, 0)); + + condition_dump_list(u->conditions, f, prefix); + + if (dual_timestamp_is_set(&u->condition_timestamp)) + fprintf(f, + "%s\tCondition Timestamp: %s\n" + "%s\tCondition Result: %s\n", + prefix, strna(format_timestamp(timestamp1, sizeof(timestamp1), u->condition_timestamp.realtime)), + prefix, yes_no(u->condition_result)); + + for (d = 0; d < _UNIT_DEPENDENCY_MAX; d++) { + Unit *other; + + SET_FOREACH(other, u->dependencies[d], i) + fprintf(f, "%s\t%s: %s\n", prefix, unit_dependency_to_string(d), other->id); + } + + if (!strv_isempty(u->requires_mounts_for)) { + fprintf(f, + "%s\tRequiresMountsFor:", prefix); + + STRV_FOREACH(j, u->requires_mounts_for) + fprintf(f, " %s", *j); + + fputs("\n", f); + } + + if (u->load_state == UNIT_LOADED) { + CGroupBonding *b; + CGroupAttribute *a; + + fprintf(f, + "%s\tStopWhenUnneeded: %s\n" + "%s\tRefuseManualStart: %s\n" + "%s\tRefuseManualStop: %s\n" + "%s\tDefaultDependencies: %s\n" + "%s\tOnFailureIsolate: %s\n" + "%s\tIgnoreOnIsolate: %s\n" + "%s\tIgnoreOnSnapshot: %s\n", + prefix, yes_no(u->stop_when_unneeded), + prefix, yes_no(u->refuse_manual_start), + prefix, yes_no(u->refuse_manual_stop), + prefix, yes_no(u->default_dependencies), + prefix, yes_no(u->on_failure_isolate), + prefix, yes_no(u->ignore_on_isolate), + prefix, yes_no(u->ignore_on_snapshot)); + + LIST_FOREACH(by_unit, b, u->cgroup_bondings) + fprintf(f, "%s\tControlGroup: %s:%s\n", + prefix, b->controller, b->path); + + LIST_FOREACH(by_unit, a, u->cgroup_attributes) { + _cleanup_free_ char *v = NULL; + + if (a->semantics && a->semantics->map_write) + a->semantics->map_write(a->semantics, a->value, &v); + + fprintf(f, "%s\tControlGroupAttribute: %s %s \"%s\"\n", + prefix, a->controller, a->name, v ? v : a->value); + } + + if (UNIT_VTABLE(u)->dump) + UNIT_VTABLE(u)->dump(u, f, prefix2); + + } else if (u->load_state == UNIT_MERGED) + fprintf(f, + "%s\tMerged into: %s\n", + prefix, u->merged_into->id); + else if (u->load_state == UNIT_ERROR) + fprintf(f, "%s\tLoad Error Code: %s\n", prefix, strerror(-u->load_error)); + + + if (u->job) + job_dump(u->job, f, prefix2); + + if (u->nop_job) + job_dump(u->nop_job, f, prefix2); + + free(p2); +} + +/* Common implementation for multiple backends */ +int unit_load_fragment_and_dropin(Unit *u) { + int r; + + assert(u); + + /* Load a .service file */ + if ((r = unit_load_fragment(u)) < 0) + return r; + + if (u->load_state == UNIT_STUB) + return -ENOENT; + + /* Load drop-in directory data */ + if ((r = unit_load_dropin(unit_follow_merge(u))) < 0) + return r; + + return 0; +} + +/* Common implementation for multiple backends */ +int unit_load_fragment_and_dropin_optional(Unit *u) { + int r; + + assert(u); + + /* Same as unit_load_fragment_and_dropin(), but whether + * something can be loaded or not doesn't matter. */ + + /* Load a .service file */ + if ((r = unit_load_fragment(u)) < 0) + return r; + + if (u->load_state == UNIT_STUB) + u->load_state = UNIT_LOADED; + + /* Load drop-in directory data */ + if ((r = unit_load_dropin(unit_follow_merge(u))) < 0) + return r; + + return 0; +} + +int unit_add_default_target_dependency(Unit *u, Unit *target) { + assert(u); + assert(target); + + if (target->type != UNIT_TARGET) + return 0; + + /* Only add the dependency if both units are loaded, so that + * that loop check below is reliable */ + if (u->load_state != UNIT_LOADED || + target->load_state != UNIT_LOADED) + return 0; + + /* If either side wants no automatic dependencies, then let's + * skip this */ + if (!u->default_dependencies || + !target->default_dependencies) + return 0; + + /* Don't create loops */ + if (set_get(target->dependencies[UNIT_BEFORE], u)) + return 0; + + return unit_add_dependency(target, UNIT_AFTER, u, true); +} + +static int unit_add_default_dependencies(Unit *u) { + static const UnitDependency deps[] = { + UNIT_REQUIRED_BY, + UNIT_REQUIRED_BY_OVERRIDABLE, + UNIT_WANTED_BY, + UNIT_BOUND_BY + }; + + Unit *target; + Iterator i; + int r; + unsigned k; + + assert(u); + + for (k = 0; k < ELEMENTSOF(deps); k++) + SET_FOREACH(target, u->dependencies[deps[k]], i) + if ((r = unit_add_default_target_dependency(u, target)) < 0) + return r; + + return 0; +} + +int unit_load(Unit *u) { + int r; + + assert(u); + + if (u->in_load_queue) { + LIST_REMOVE(Unit, load_queue, u->manager->load_queue, u); + u->in_load_queue = false; + } + + if (u->type == _UNIT_TYPE_INVALID) + return -EINVAL; + + if (u->load_state != UNIT_STUB) + return 0; + + if (UNIT_VTABLE(u)->load) + if ((r = UNIT_VTABLE(u)->load(u)) < 0) + goto fail; + + if (u->load_state == UNIT_STUB) { + r = -ENOENT; + goto fail; + } + + if (u->load_state == UNIT_LOADED && + u->default_dependencies) + if ((r = unit_add_default_dependencies(u)) < 0) + goto fail; + + if (u->load_state == UNIT_LOADED) { + r = unit_add_mount_links(u); + if (r < 0) + return r; + } + + if (u->on_failure_isolate && + set_size(u->dependencies[UNIT_ON_FAILURE]) > 1) { + + log_error_unit(u->id, + "More than one OnFailure= dependencies specified for %s but OnFailureIsolate= enabled. Refusing.", u->id); + + r = -EINVAL; + goto fail; + } + + assert((u->load_state != UNIT_MERGED) == !u->merged_into); + + unit_add_to_dbus_queue(unit_follow_merge(u)); + unit_add_to_gc_queue(u); + + return 0; + +fail: + u->load_state = UNIT_ERROR; + u->load_error = r; + unit_add_to_dbus_queue(u); + unit_add_to_gc_queue(u); + + log_debug_unit(u->id, "Failed to load configuration for %s: %s", + u->id, strerror(-r)); + + return r; +} + +bool unit_condition_test(Unit *u) { + assert(u); + + dual_timestamp_get(&u->condition_timestamp); + u->condition_result = condition_test_list(u->conditions); + + return u->condition_result; +} + +_pure_ static const char* unit_get_status_message_format(Unit *u, JobType t) { + const UnitStatusMessageFormats *format_table; + + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + if (t != JOB_START && t != JOB_STOP) + return NULL; + + format_table = &UNIT_VTABLE(u)->status_message_formats; + if (!format_table) + return NULL; + + return format_table->starting_stopping[t == JOB_STOP]; +} + +_pure_ static const char *unit_get_status_message_format_try_harder(Unit *u, JobType t) { + const char *format; + + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + format = unit_get_status_message_format(u, t); + if (format) + return format; + + /* Return generic strings */ + if (t == JOB_START) + return "Starting %s."; + else if (t == JOB_STOP) + return "Stopping %s."; + else if (t == JOB_RELOAD) + return "Reloading %s."; + + return NULL; +} + +static void unit_status_print_starting_stopping(Unit *u, JobType t) { + const char *format; + + assert(u); + + /* We only print status messages for selected units on + * selected operations. */ + + format = unit_get_status_message_format(u, t); + if (!format) + return; + + unit_status_printf(u, "", format); +} + +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wformat-nonliteral" +static void unit_status_log_starting_stopping_reloading(Unit *u, JobType t) { + const char *format; + char buf[LINE_MAX]; + sd_id128_t mid; + + assert(u); + + if (t != JOB_START && t != JOB_STOP && t != JOB_RELOAD) + return; + + if (log_on_console()) + return; + + /* We log status messages for all units and all operations. */ + + format = unit_get_status_message_format_try_harder(u, t); + if (!format) + return; + + snprintf(buf, sizeof(buf), format, unit_description(u)); + char_array_0(buf); + + mid = t == JOB_START ? SD_MESSAGE_UNIT_STARTING : + t == JOB_STOP ? SD_MESSAGE_UNIT_STOPPING : + SD_MESSAGE_UNIT_RELOADING; + + log_struct_unit(LOG_INFO, + u->id, + MESSAGE_ID(mid), + "MESSAGE=%s", buf, + NULL); +} +#pragma GCC diagnostic pop + +/* Errors: + * -EBADR: This unit type does not support starting. + * -EALREADY: Unit is already started. + * -EAGAIN: An operation is already in progress. Retry later. + * -ECANCELED: Too many requests for now. + */ +int unit_start(Unit *u) { + UnitActiveState state; + Unit *following; + + assert(u); + + if (u->load_state != UNIT_LOADED) + return -EINVAL; + + /* If this is already started, then this will succeed. Note + * that this will even succeed if this unit is not startable + * by the user. This is relied on to detect when we need to + * wait for units and when waiting is finished. */ + state = unit_active_state(u); + if (UNIT_IS_ACTIVE_OR_RELOADING(state)) + return -EALREADY; + + /* If the conditions failed, don't do anything at all. If we + * already are activating this call might still be useful to + * speed up activation in case there is some hold-off time, + * but we don't want to recheck the condition in that case. */ + if (state != UNIT_ACTIVATING && + !unit_condition_test(u)) { + log_debug_unit(u->id, "Starting of %s requested but condition failed. Ignoring.", u->id); + return -EALREADY; + } + + /* Forward to the main object, if we aren't it. */ + if ((following = unit_following(u))) { + log_debug_unit(u->id, "Redirecting start request from %s to %s.", + u->id, following->id); + return unit_start(following); + } + + unit_status_log_starting_stopping_reloading(u, JOB_START); + unit_status_print_starting_stopping(u, JOB_START); + + /* If it is stopped, but we cannot start it, then fail */ + if (!UNIT_VTABLE(u)->start) + return -EBADR; + + /* We don't suppress calls to ->start() here when we are + * already starting, to allow this request to be used as a + * "hurry up" call, for example when the unit is in some "auto + * restart" state where it waits for a holdoff timer to elapse + * before it will start again. */ + + unit_add_to_dbus_queue(u); + + return UNIT_VTABLE(u)->start(u); +} + +bool unit_can_start(Unit *u) { + assert(u); + + return !!UNIT_VTABLE(u)->start; +} + +bool unit_can_isolate(Unit *u) { + assert(u); + + return unit_can_start(u) && + u->allow_isolate; +} + +/* Errors: + * -EBADR: This unit type does not support stopping. + * -EALREADY: Unit is already stopped. + * -EAGAIN: An operation is already in progress. Retry later. + */ +int unit_stop(Unit *u) { + UnitActiveState state; + Unit *following; + + assert(u); + + state = unit_active_state(u); + if (UNIT_IS_INACTIVE_OR_FAILED(state)) + return -EALREADY; + + if ((following = unit_following(u))) { + log_debug_unit(u->id, "Redirecting stop request from %s to %s.", + u->id, following->id); + return unit_stop(following); + } + + unit_status_log_starting_stopping_reloading(u, JOB_STOP); + unit_status_print_starting_stopping(u, JOB_STOP); + + if (!UNIT_VTABLE(u)->stop) + return -EBADR; + + unit_add_to_dbus_queue(u); + + return UNIT_VTABLE(u)->stop(u); +} + +/* Errors: + * -EBADR: This unit type does not support reloading. + * -ENOEXEC: Unit is not started. + * -EAGAIN: An operation is already in progress. Retry later. + */ +int unit_reload(Unit *u) { + UnitActiveState state; + Unit *following; + + assert(u); + + if (u->load_state != UNIT_LOADED) + return -EINVAL; + + if (!unit_can_reload(u)) + return -EBADR; + + state = unit_active_state(u); + if (state == UNIT_RELOADING) + return -EALREADY; + + if (state != UNIT_ACTIVE) + return -ENOEXEC; + + if ((following = unit_following(u))) { + log_debug_unit(u->id, "Redirecting reload request from %s to %s.", + u->id, following->id); + return unit_reload(following); + } + + unit_status_log_starting_stopping_reloading(u, JOB_RELOAD); + + unit_add_to_dbus_queue(u); + return UNIT_VTABLE(u)->reload(u); +} + +bool unit_can_reload(Unit *u) { + assert(u); + + if (!UNIT_VTABLE(u)->reload) + return false; + + if (!UNIT_VTABLE(u)->can_reload) + return true; + + return UNIT_VTABLE(u)->can_reload(u); +} + +static void unit_check_unneeded(Unit *u) { + Iterator i; + Unit *other; + + assert(u); + + /* If this service shall be shut down when unneeded then do + * so. */ + + if (!u->stop_when_unneeded) + return; + + if (!UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u))) + return; + + SET_FOREACH(other, u->dependencies[UNIT_REQUIRED_BY], i) + if (unit_active_or_pending(other)) + return; + + SET_FOREACH(other, u->dependencies[UNIT_REQUIRED_BY_OVERRIDABLE], i) + if (unit_active_or_pending(other)) + return; + + SET_FOREACH(other, u->dependencies[UNIT_WANTED_BY], i) + if (unit_active_or_pending(other)) + return; + + SET_FOREACH(other, u->dependencies[UNIT_BOUND_BY], i) + if (unit_active_or_pending(other)) + return; + + log_info_unit(u->id, "Service %s is not needed anymore. Stopping.", u->id); + + /* Ok, nobody needs us anymore. Sniff. Then let's commit suicide */ + manager_add_job(u->manager, JOB_STOP, u, JOB_FAIL, true, NULL, NULL); +} + +static void retroactively_start_dependencies(Unit *u) { + Iterator i; + Unit *other; + + assert(u); + assert(UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u))); + + SET_FOREACH(other, u->dependencies[UNIT_REQUIRES], i) + if (!set_get(u->dependencies[UNIT_AFTER], other) && + !UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_START, other, JOB_REPLACE, true, NULL, NULL); + + SET_FOREACH(other, u->dependencies[UNIT_BINDS_TO], i) + if (!set_get(u->dependencies[UNIT_AFTER], other) && + !UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_START, other, JOB_REPLACE, true, NULL, NULL); + + SET_FOREACH(other, u->dependencies[UNIT_REQUIRES_OVERRIDABLE], i) + if (!set_get(u->dependencies[UNIT_AFTER], other) && + !UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_START, other, JOB_FAIL, false, NULL, NULL); + + SET_FOREACH(other, u->dependencies[UNIT_WANTS], i) + if (!set_get(u->dependencies[UNIT_AFTER], other) && + !UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_START, other, JOB_FAIL, false, NULL, NULL); + + SET_FOREACH(other, u->dependencies[UNIT_CONFLICTS], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_STOP, other, JOB_REPLACE, true, NULL, NULL); + + SET_FOREACH(other, u->dependencies[UNIT_CONFLICTED_BY], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_STOP, other, JOB_REPLACE, true, NULL, NULL); +} + +static void retroactively_stop_dependencies(Unit *u) { + Iterator i; + Unit *other; + + assert(u); + assert(UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(u))); + + /* Pull down units which are bound to us recursively if enabled */ + SET_FOREACH(other, u->dependencies[UNIT_BOUND_BY], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_STOP, other, JOB_REPLACE, true, NULL, NULL); +} + +static void check_unneeded_dependencies(Unit *u) { + Iterator i; + Unit *other; + + assert(u); + assert(UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(u))); + + /* Garbage collect services that might not be needed anymore, if enabled */ + SET_FOREACH(other, u->dependencies[UNIT_REQUIRES], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); + SET_FOREACH(other, u->dependencies[UNIT_REQUIRES_OVERRIDABLE], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); + SET_FOREACH(other, u->dependencies[UNIT_WANTS], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); + SET_FOREACH(other, u->dependencies[UNIT_REQUISITE], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); + SET_FOREACH(other, u->dependencies[UNIT_REQUISITE_OVERRIDABLE], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); + SET_FOREACH(other, u->dependencies[UNIT_BINDS_TO], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); +} + +void unit_start_on_failure(Unit *u) { + Unit *other; + Iterator i; + + assert(u); + + if (set_size(u->dependencies[UNIT_ON_FAILURE]) <= 0) + return; + + log_info_unit(u->id, "Triggering OnFailure= dependencies of %s.", u->id); + + SET_FOREACH(other, u->dependencies[UNIT_ON_FAILURE], i) { + int r; + + r = manager_add_job(u->manager, JOB_START, other, u->on_failure_isolate ? JOB_ISOLATE : JOB_REPLACE, true, NULL, NULL); + if (r < 0) + log_error_unit(u->id, "Failed to enqueue OnFailure= job: %s", strerror(-r)); + } +} + +void unit_trigger_notify(Unit *u) { + Unit *other; + Iterator i; + + assert(u); + + SET_FOREACH(other, u->dependencies[UNIT_TRIGGERED_BY], i) + if (UNIT_VTABLE(other)->trigger_notify) + UNIT_VTABLE(other)->trigger_notify(other, u); +} + +void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_success) { + Manager *m; + bool unexpected; + + assert(u); + assert(os < _UNIT_ACTIVE_STATE_MAX); + assert(ns < _UNIT_ACTIVE_STATE_MAX); + + /* Note that this is called for all low-level state changes, + * even if they might map to the same high-level + * UnitActiveState! That means that ns == os is OK an expected + * behavior here. For example: if a mount point is remounted + * this function will be called too! */ + + m = u->manager; + + if (m->n_reloading <= 0) { + dual_timestamp ts; + + dual_timestamp_get(&ts); + + if (UNIT_IS_INACTIVE_OR_FAILED(os) && !UNIT_IS_INACTIVE_OR_FAILED(ns)) + u->inactive_exit_timestamp = ts; + else if (!UNIT_IS_INACTIVE_OR_FAILED(os) && UNIT_IS_INACTIVE_OR_FAILED(ns)) + u->inactive_enter_timestamp = ts; + + if (!UNIT_IS_ACTIVE_OR_RELOADING(os) && UNIT_IS_ACTIVE_OR_RELOADING(ns)) + u->active_enter_timestamp = ts; + else if (UNIT_IS_ACTIVE_OR_RELOADING(os) && !UNIT_IS_ACTIVE_OR_RELOADING(ns)) + u->active_exit_timestamp = ts; + } + + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) + cgroup_bonding_trim_list(u->cgroup_bondings, true); + + if (UNIT_IS_INACTIVE_OR_FAILED(os) != UNIT_IS_INACTIVE_OR_FAILED(ns)) { + ExecContext *ec = unit_get_exec_context(u); + if (ec && exec_context_may_touch_console(ec)) { + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) + m->n_on_console--; + else + m->n_on_console++; + } + } + + if (u->job) { + unexpected = false; + + if (u->job->state == JOB_WAITING) + + /* So we reached a different state for this + * job. Let's see if we can run it now if it + * failed previously due to EAGAIN. */ + job_add_to_run_queue(u->job); + + /* Let's check whether this state change constitutes a + * finished job, or maybe contradicts a running job and + * hence needs to invalidate jobs. */ + + switch (u->job->type) { + + case JOB_START: + case JOB_VERIFY_ACTIVE: + + if (UNIT_IS_ACTIVE_OR_RELOADING(ns)) + job_finish_and_invalidate(u->job, JOB_DONE, true); + else if (u->job->state == JOB_RUNNING && ns != UNIT_ACTIVATING) { + unexpected = true; + + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) + job_finish_and_invalidate(u->job, ns == UNIT_FAILED ? JOB_FAILED : JOB_DONE, true); + } + + break; + + case JOB_RELOAD: + case JOB_RELOAD_OR_START: + + if (u->job->state == JOB_RUNNING) { + if (ns == UNIT_ACTIVE) + job_finish_and_invalidate(u->job, reload_success ? JOB_DONE : JOB_FAILED, true); + else if (ns != UNIT_ACTIVATING && ns != UNIT_RELOADING) { + unexpected = true; + + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) + job_finish_and_invalidate(u->job, ns == UNIT_FAILED ? JOB_FAILED : JOB_DONE, true); + } + } + + break; + + case JOB_STOP: + case JOB_RESTART: + case JOB_TRY_RESTART: + + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) + job_finish_and_invalidate(u->job, JOB_DONE, true); + else if (u->job->state == JOB_RUNNING && ns != UNIT_DEACTIVATING) { + unexpected = true; + job_finish_and_invalidate(u->job, JOB_FAILED, true); + } + + break; + + default: + assert_not_reached("Job type unknown"); + } + + } else + unexpected = true; + + if (m->n_reloading <= 0) { + + /* If this state change happened without being + * requested by a job, then let's retroactively start + * or stop dependencies. We skip that step when + * deserializing, since we don't want to create any + * additional jobs just because something is already + * activated. */ + + if (unexpected) { + if (UNIT_IS_INACTIVE_OR_FAILED(os) && UNIT_IS_ACTIVE_OR_ACTIVATING(ns)) + retroactively_start_dependencies(u); + else if (UNIT_IS_ACTIVE_OR_ACTIVATING(os) && UNIT_IS_INACTIVE_OR_DEACTIVATING(ns)) + retroactively_stop_dependencies(u); + } + + /* stop unneeded units regardless if going down was expected or not */ + if (UNIT_IS_ACTIVE_OR_ACTIVATING(os) && UNIT_IS_INACTIVE_OR_DEACTIVATING(ns)) + check_unneeded_dependencies(u); + + if (ns != os && ns == UNIT_FAILED) { + log_notice_unit(u->id, + "Unit %s entered failed state.", u->id); + unit_start_on_failure(u); + } + } + + /* Some names are special */ + if (UNIT_IS_ACTIVE_OR_RELOADING(ns)) { + + if (unit_has_name(u, SPECIAL_DBUS_SERVICE)) + /* The bus just might have become available, + * hence try to connect to it, if we aren't + * yet connected. */ + bus_init(m, true); + + if (u->type == UNIT_SERVICE && + !UNIT_IS_ACTIVE_OR_RELOADING(os) && + m->n_reloading <= 0) { + /* Write audit record if we have just finished starting up */ + manager_send_unit_audit(m, u, AUDIT_SERVICE_START, true); + u->in_audit = true; + } + + if (!UNIT_IS_ACTIVE_OR_RELOADING(os)) + manager_send_unit_plymouth(m, u); + + } else { + + /* We don't care about D-Bus here, since we'll get an + * asynchronous notification for it anyway. */ + + if (u->type == UNIT_SERVICE && + UNIT_IS_INACTIVE_OR_FAILED(ns) && + !UNIT_IS_INACTIVE_OR_FAILED(os) && + m->n_reloading <= 0) { + + /* Hmm, if there was no start record written + * write it now, so that we always have a nice + * pair */ + if (!u->in_audit) { + manager_send_unit_audit(m, u, AUDIT_SERVICE_START, ns == UNIT_INACTIVE); + + if (ns == UNIT_INACTIVE) + manager_send_unit_audit(m, u, AUDIT_SERVICE_STOP, true); + } else + /* Write audit record if we have just finished shutting down */ + manager_send_unit_audit(m, u, AUDIT_SERVICE_STOP, ns == UNIT_INACTIVE); + + u->in_audit = false; + } + } + + manager_recheck_journal(m); + unit_trigger_notify(u); + + /* Maybe we finished startup and are now ready for being + * stopped because unneeded? */ + if (u->manager->n_reloading <= 0) + unit_check_unneeded(u); + + unit_add_to_dbus_queue(u); + unit_add_to_gc_queue(u); +} + +int unit_watch_fd(Unit *u, int fd, uint32_t events, Watch *w) { + struct epoll_event ev = { + .data.ptr = w, + .events = events, + }; + + assert(u); + assert(fd >= 0); + assert(w); + assert(w->type == WATCH_INVALID || (w->type == WATCH_FD && w->fd == fd && w->data.unit == u)); + + if (epoll_ctl(u->manager->epoll_fd, + w->type == WATCH_INVALID ? EPOLL_CTL_ADD : EPOLL_CTL_MOD, + fd, + &ev) < 0) + return -errno; + + w->fd = fd; + w->type = WATCH_FD; + w->data.unit = u; + + return 0; +} + +void unit_unwatch_fd(Unit *u, Watch *w) { + assert(u); + assert(w); + + if (w->type == WATCH_INVALID) + return; + + assert(w->type == WATCH_FD); + assert(w->data.unit == u); + assert_se(epoll_ctl(u->manager->epoll_fd, EPOLL_CTL_DEL, w->fd, NULL) >= 0); + + w->fd = -1; + w->type = WATCH_INVALID; + w->data.unit = NULL; +} + +int unit_watch_pid(Unit *u, pid_t pid) { + assert(u); + assert(pid >= 1); + + /* Watch a specific PID. We only support one unit watching + * each PID for now. */ + + return hashmap_put(u->manager->watch_pids, LONG_TO_PTR(pid), u); +} + +void unit_unwatch_pid(Unit *u, pid_t pid) { + assert(u); + assert(pid >= 1); + + hashmap_remove_value(u->manager->watch_pids, LONG_TO_PTR(pid), u); +} + +int unit_watch_timer(Unit *u, clockid_t clock_id, bool relative, usec_t usec, Watch *w) { + struct itimerspec its = {}; + int flags, fd; + bool ours; + + assert(u); + assert(w); + assert(w->type == WATCH_INVALID || (w->type == WATCH_UNIT_TIMER && w->data.unit == u)); + + /* This will try to reuse the old timer if there is one */ + + if (w->type == WATCH_UNIT_TIMER) { + assert(w->data.unit == u); + assert(w->fd >= 0); + + ours = false; + fd = w->fd; + } else if (w->type == WATCH_INVALID) { + + ours = true; + fd = timerfd_create(clock_id, TFD_NONBLOCK|TFD_CLOEXEC); + if (fd < 0) + return -errno; + } else + assert_not_reached("Invalid watch type"); + + if (usec <= 0) { + /* Set absolute time in the past, but not 0, since we + * don't want to disarm the timer */ + its.it_value.tv_sec = 0; + its.it_value.tv_nsec = 1; + + flags = TFD_TIMER_ABSTIME; + } else { + timespec_store(&its.it_value, usec); + flags = relative ? 0 : TFD_TIMER_ABSTIME; + } + + /* This will also flush the elapse counter */ + if (timerfd_settime(fd, flags, &its, NULL) < 0) + goto fail; + + if (w->type == WATCH_INVALID) { + struct epoll_event ev = { + .data.ptr = w, + .events = EPOLLIN, + }; + + if (epoll_ctl(u->manager->epoll_fd, EPOLL_CTL_ADD, fd, &ev) < 0) + goto fail; + } + + w->type = WATCH_UNIT_TIMER; + w->fd = fd; + w->data.unit = u; + + return 0; + +fail: + if (ours) + close_nointr_nofail(fd); + + return -errno; +} + +void unit_unwatch_timer(Unit *u, Watch *w) { + assert(u); + assert(w); + + if (w->type == WATCH_INVALID) + return; + + assert(w->type == WATCH_UNIT_TIMER); + assert(w->data.unit == u); + assert(w->fd >= 0); + + assert_se(epoll_ctl(u->manager->epoll_fd, EPOLL_CTL_DEL, w->fd, NULL) >= 0); + close_nointr_nofail(w->fd); + + w->fd = -1; + w->type = WATCH_INVALID; + w->data.unit = NULL; +} + +bool unit_job_is_applicable(Unit *u, JobType j) { + assert(u); + assert(j >= 0 && j < _JOB_TYPE_MAX); + + switch (j) { + + case JOB_VERIFY_ACTIVE: + case JOB_START: + case JOB_STOP: + case JOB_NOP: + return true; + + case JOB_RESTART: + case JOB_TRY_RESTART: + return unit_can_start(u); + + case JOB_RELOAD: + return unit_can_reload(u); + + case JOB_RELOAD_OR_START: + return unit_can_reload(u) && unit_can_start(u); + + default: + assert_not_reached("Invalid job type"); + } +} + +int unit_add_dependency(Unit *u, UnitDependency d, Unit *other, bool add_reference) { + + static const UnitDependency inverse_table[_UNIT_DEPENDENCY_MAX] = { + [UNIT_REQUIRES] = UNIT_REQUIRED_BY, + [UNIT_REQUIRES_OVERRIDABLE] = UNIT_REQUIRED_BY_OVERRIDABLE, + [UNIT_WANTS] = UNIT_WANTED_BY, + [UNIT_REQUISITE] = UNIT_REQUIRED_BY, + [UNIT_REQUISITE_OVERRIDABLE] = UNIT_REQUIRED_BY_OVERRIDABLE, + [UNIT_BINDS_TO] = UNIT_BOUND_BY, + [UNIT_PART_OF] = UNIT_CONSISTS_OF, + [UNIT_REQUIRED_BY] = _UNIT_DEPENDENCY_INVALID, + [UNIT_REQUIRED_BY_OVERRIDABLE] = _UNIT_DEPENDENCY_INVALID, + [UNIT_WANTED_BY] = _UNIT_DEPENDENCY_INVALID, + [UNIT_BOUND_BY] = UNIT_BINDS_TO, + [UNIT_CONSISTS_OF] = UNIT_PART_OF, + [UNIT_CONFLICTS] = UNIT_CONFLICTED_BY, + [UNIT_CONFLICTED_BY] = UNIT_CONFLICTS, + [UNIT_BEFORE] = UNIT_AFTER, + [UNIT_AFTER] = UNIT_BEFORE, + [UNIT_ON_FAILURE] = _UNIT_DEPENDENCY_INVALID, + [UNIT_REFERENCES] = UNIT_REFERENCED_BY, + [UNIT_REFERENCED_BY] = UNIT_REFERENCES, + [UNIT_TRIGGERS] = UNIT_TRIGGERED_BY, + [UNIT_TRIGGERED_BY] = UNIT_TRIGGERS, + [UNIT_PROPAGATES_RELOAD_TO] = UNIT_RELOAD_PROPAGATED_FROM, + [UNIT_RELOAD_PROPAGATED_FROM] = UNIT_PROPAGATES_RELOAD_TO, + }; + int r, q = 0, v = 0, w = 0; + + assert(u); + assert(d >= 0 && d < _UNIT_DEPENDENCY_MAX); + assert(other); + + u = unit_follow_merge(u); + other = unit_follow_merge(other); + + /* We won't allow dependencies on ourselves. We will not + * consider them an error however. */ + if (u == other) + return 0; + + if ((r = set_ensure_allocated(&u->dependencies[d], trivial_hash_func, trivial_compare_func)) < 0) + return r; + + if (inverse_table[d] != _UNIT_DEPENDENCY_INVALID) + if ((r = set_ensure_allocated(&other->dependencies[inverse_table[d]], trivial_hash_func, trivial_compare_func)) < 0) + return r; + + if (add_reference) + if ((r = set_ensure_allocated(&u->dependencies[UNIT_REFERENCES], trivial_hash_func, trivial_compare_func)) < 0 || + (r = set_ensure_allocated(&other->dependencies[UNIT_REFERENCED_BY], trivial_hash_func, trivial_compare_func)) < 0) + return r; + + if ((q = set_put(u->dependencies[d], other)) < 0) + return q; + + if (inverse_table[d] != _UNIT_DEPENDENCY_INVALID) + if ((v = set_put(other->dependencies[inverse_table[d]], u)) < 0) { + r = v; + goto fail; + } + + if (add_reference) { + if ((w = set_put(u->dependencies[UNIT_REFERENCES], other)) < 0) { + r = w; + goto fail; + } + + if ((r = set_put(other->dependencies[UNIT_REFERENCED_BY], u)) < 0) + goto fail; + } + + unit_add_to_dbus_queue(u); + return 0; + +fail: + if (q > 0) + set_remove(u->dependencies[d], other); + + if (v > 0) + set_remove(other->dependencies[inverse_table[d]], u); + + if (w > 0) + set_remove(u->dependencies[UNIT_REFERENCES], other); + + return r; +} + +int unit_add_two_dependencies(Unit *u, UnitDependency d, UnitDependency e, Unit *other, bool add_reference) { + int r; + + assert(u); + + if ((r = unit_add_dependency(u, d, other, add_reference)) < 0) + return r; + + if ((r = unit_add_dependency(u, e, other, add_reference)) < 0) + return r; + + return 0; +} + +static const char *resolve_template(Unit *u, const char *name, const char*path, char **p) { + char *s; + + assert(u); + assert(name || path); + assert(p); + + if (!name) + name = path_get_file_name(path); + + if (!unit_name_is_template(name)) { + *p = NULL; + return name; + } + + if (u->instance) + s = unit_name_replace_instance(name, u->instance); + else { + _cleanup_free_ char *i = NULL; + + i = unit_name_to_prefix(u->id); + if (!i) + return NULL; + + s = unit_name_replace_instance(name, i); + } + + if (!s) + return NULL; + + *p = s; + return s; +} + +int unit_add_dependency_by_name(Unit *u, UnitDependency d, const char *name, const char *path, bool add_reference) { + Unit *other; + int r; + _cleanup_free_ char *s = NULL; + + assert(u); + assert(name || path); + + name = resolve_template(u, name, path, &s); + if (!name) + return -ENOMEM; + + r = manager_load_unit(u->manager, name, path, NULL, &other); + if (r < 0) + return r; + + return unit_add_dependency(u, d, other, add_reference); +} + +int unit_add_two_dependencies_by_name(Unit *u, UnitDependency d, UnitDependency e, const char *name, const char *path, bool add_reference) { + Unit *other; + int r; + char *s; + + assert(u); + assert(name || path); + + if (!(name = resolve_template(u, name, path, &s))) + return -ENOMEM; + + if ((r = manager_load_unit(u->manager, name, path, NULL, &other)) < 0) + goto finish; + + r = unit_add_two_dependencies(u, d, e, other, add_reference); + +finish: + free(s); + return r; +} + +int unit_add_dependency_by_name_inverse(Unit *u, UnitDependency d, const char *name, const char *path, bool add_reference) { + Unit *other; + int r; + char *s; + + assert(u); + assert(name || path); + + if (!(name = resolve_template(u, name, path, &s))) + return -ENOMEM; + + if ((r = manager_load_unit(u->manager, name, path, NULL, &other)) < 0) + goto finish; + + r = unit_add_dependency(other, d, u, add_reference); + +finish: + free(s); + return r; +} + +int unit_add_two_dependencies_by_name_inverse(Unit *u, UnitDependency d, UnitDependency e, const char *name, const char *path, bool add_reference) { + Unit *other; + int r; + char *s; + + assert(u); + assert(name || path); + + if (!(name = resolve_template(u, name, path, &s))) + return -ENOMEM; + + if ((r = manager_load_unit(u->manager, name, path, NULL, &other)) < 0) + goto finish; + + if ((r = unit_add_two_dependencies(other, d, e, u, add_reference)) < 0) + goto finish; + +finish: + free(s); + return r; +} + +int set_unit_path(const char *p) { + _cleanup_free_ char *c = NULL; + + /* This is mostly for debug purposes */ + c = path_make_absolute_cwd(p); + if (setenv("SYSTEMD_UNIT_PATH", c, 0) < 0) + return -errno; + + return 0; +} + +char *unit_dbus_path(Unit *u) { + assert(u); + + if (!u->id) + return NULL; + + return unit_dbus_path_from_name(u->id); +} + +static int unit_add_cgroup(Unit *u, CGroupBonding *b) { + int r; + + assert(u); + assert(b); + + assert(b->path); + + if (!b->controller) { + b->controller = strdup(SYSTEMD_CGROUP_CONTROLLER); + if (!b->controller) + return log_oom(); + + b->ours = true; + } + + /* Ensure this hasn't been added yet */ + assert(!b->unit); + + if (streq(b->controller, SYSTEMD_CGROUP_CONTROLLER)) { + CGroupBonding *l; + + l = hashmap_get(u->manager->cgroup_bondings, b->path); + LIST_PREPEND(CGroupBonding, by_path, l, b); + + r = hashmap_replace(u->manager->cgroup_bondings, b->path, l); + if (r < 0) { + LIST_REMOVE(CGroupBonding, by_path, l, b); + return r; + } + } + + LIST_PREPEND(CGroupBonding, by_unit, u->cgroup_bondings, b); + b->unit = u; + + return 0; +} + +char *unit_default_cgroup_path(Unit *u) { + _cleanup_free_ char *escaped_instance = NULL; + + assert(u); + + escaped_instance = cg_escape(u->id); + if (!escaped_instance) + return NULL; + + if (u->instance) { + _cleanup_free_ char *t = NULL, *escaped_template = NULL; + + t = unit_name_template(u->id); + if (!t) + return NULL; + + escaped_template = cg_escape(t); + if (!escaped_template) + return NULL; + + return strjoin(u->manager->cgroup_hierarchy, "/", escaped_template, "/", escaped_instance, NULL); + } else + return strjoin(u->manager->cgroup_hierarchy, "/", escaped_instance, NULL); +} + +int unit_add_cgroup_from_text(Unit *u, const char *name, bool overwrite, CGroupBonding **ret) { + char *controller = NULL, *path = NULL; + CGroupBonding *b = NULL; + bool ours = false; + int r; + + assert(u); + assert(name); + + r = cg_split_spec(name, &controller, &path); + if (r < 0) + return r; + + if (!path) { + path = unit_default_cgroup_path(u); + ours = true; + } + + if (!controller) { + controller = strdup("systemd"); + ours = true; + } + + if (!path || !controller) { + free(path); + free(controller); + return log_oom(); + } + + if (streq(controller, "systemd")) { + /* Within the systemd unit hierarchy we do not allow changes. */ + if (path_startswith(path, "/system")) { + log_warning_unit(u->id, "Manipulating the systemd:/system cgroup hierarchy is not permitted."); + free(path); + free(controller); + return -EPERM; + } + } + + b = cgroup_bonding_find_list(u->cgroup_bondings, controller); + if (b) { + if (streq(path, b->path)) { + free(path); + free(controller); + + if (ret) + *ret = b; + return 0; + } + + if (overwrite && !b->essential) { + free(controller); + + free(b->path); + b->path = path; + + b->ours = ours; + b->realized = false; + + if (ret) + *ret = b; + + return 1; + } + + r = -EEXIST; + b = NULL; + goto fail; + } + + b = new0(CGroupBonding, 1); + if (!b) { + r = -ENOMEM; + goto fail; + } + + b->controller = controller; + b->path = path; + b->ours = ours; + b->essential = streq(controller, SYSTEMD_CGROUP_CONTROLLER); + + r = unit_add_cgroup(u, b); + if (r < 0) + goto fail; + + if (ret) + *ret = b; + + return 1; + +fail: + free(path); + free(controller); + free(b); + + return r; +} + +static int unit_add_one_default_cgroup(Unit *u, const char *controller) { + CGroupBonding *b = NULL; + int r = -ENOMEM; + + assert(u); + + if (controller && !cg_controller_is_valid(controller, true)) + return -EINVAL; + + if (!controller) + controller = SYSTEMD_CGROUP_CONTROLLER; + + if (cgroup_bonding_find_list(u->cgroup_bondings, controller)) + return 0; + + b = new0(CGroupBonding, 1); + if (!b) + return -ENOMEM; + + b->controller = strdup(controller); + if (!b->controller) + goto fail; + + b->path = unit_default_cgroup_path(u); + if (!b->path) + goto fail; + + b->ours = true; + b->essential = streq(controller, SYSTEMD_CGROUP_CONTROLLER); + + r = unit_add_cgroup(u, b); + if (r < 0) + goto fail; + + return 1; + +fail: + free(b->path); + free(b->controller); + free(b); + + return r; +} + +int unit_add_default_cgroups(Unit *u) { + CGroupAttribute *a; + char **c; + int r; + + assert(u); + + /* Adds in the default cgroups, if they weren't specified + * otherwise. */ + + if (!u->manager->cgroup_hierarchy) + return 0; + + r = unit_add_one_default_cgroup(u, NULL); + if (r < 0) + return r; + + STRV_FOREACH(c, u->manager->default_controllers) + unit_add_one_default_cgroup(u, *c); + + LIST_FOREACH(by_unit, a, u->cgroup_attributes) + unit_add_one_default_cgroup(u, a->controller); + + return 0; +} + +CGroupBonding* unit_get_default_cgroup(Unit *u) { + assert(u); + + return cgroup_bonding_find_list(u->cgroup_bondings, NULL); +} + +int unit_add_cgroup_attribute( + Unit *u, + const CGroupSemantics *semantics, + const char *controller, + const char *name, + const char *value, + CGroupAttribute **ret) { + + _cleanup_free_ char *c = NULL; + CGroupAttribute *a; + int r; + + assert(u); + assert(value); + + if (semantics) { + /* Semantics always take precedence */ + if (semantics->name) + name = semantics->name; + + if (semantics->controller) + controller = semantics->controller; + } + + if (!name) + return -EINVAL; + + if (!controller) { + r = cg_controller_from_attr(name, &c); + if (r < 0) + return -EINVAL; + + controller = c; + } + + if (!controller || + streq(controller, SYSTEMD_CGROUP_CONTROLLER) || + streq(controller, "systemd")) + return -EINVAL; + + if (!filename_is_safe(name)) + return -EINVAL; + + if (!cg_controller_is_valid(controller, false)) + return -EINVAL; + + /* Check if this attribute already exists. Note that we will + * explicitly check for the value here too, as there are + * attributes which accept multiple values. */ + a = cgroup_attribute_find_list(u->cgroup_attributes, controller, name); + if (a) { + if (streq(value, a->value)) { + /* Exactly the same value is always OK, let's ignore this */ + if (ret) + *ret = a; + + return 0; + } + + if (semantics && !semantics->multiple) { + char *v; + + /* If this is a single-item entry, we can + * simply patch the existing attribute */ + + v = strdup(value); + if (!v) + return -ENOMEM; + + free(a->value); + a->value = v; + + if (ret) + *ret = a; + return 1; + } + } + + a = new0(CGroupAttribute, 1); + if (!a) + return -ENOMEM; + + if (c) { + a->controller = c; + c = NULL; + } else + a->controller = strdup(controller); + + a->name = strdup(name); + a->value = strdup(value); + + if (!a->controller || !a->name || !a->value) { + free(a->controller); + free(a->name); + free(a->value); + free(a); + return -ENOMEM; + } + + a->semantics = semantics; + a->unit = u; + + LIST_PREPEND(CGroupAttribute, by_unit, u->cgroup_attributes, a); + + if (ret) + *ret = a; + + return 1; +} + +int unit_load_related_unit(Unit *u, const char *type, Unit **_found) { + _cleanup_free_ char *t = NULL; + int r; + + assert(u); + assert(type); + assert(_found); + + t = unit_name_change_suffix(u->id, type); + if (!t) + return -ENOMEM; + + assert(!unit_has_name(u, t)); + + r = manager_load_unit(u->manager, t, NULL, NULL, _found); + assert(r < 0 || *_found != u); + return r; +} + +int unit_get_related_unit(Unit *u, const char *type, Unit **_found) { + _cleanup_free_ char *t = NULL; + Unit *found; + + assert(u); + assert(type); + assert(_found); + + t = unit_name_change_suffix(u->id, type); + if (!t) + return -ENOMEM; + + assert(!unit_has_name(u, t)); + + found = manager_get_unit(u->manager, t); + if (!found) + return -ENOENT; + + *_found = found; + return 0; +} + +int unit_watch_bus_name(Unit *u, const char *name) { + assert(u); + assert(name); + + /* Watch a specific name on the bus. We only support one unit + * watching each name for now. */ + + return hashmap_put(u->manager->watch_bus, name, u); +} + +void unit_unwatch_bus_name(Unit *u, const char *name) { + assert(u); + assert(name); + + hashmap_remove_value(u->manager->watch_bus, name, u); +} + +bool unit_can_serialize(Unit *u) { + assert(u); + + return UNIT_VTABLE(u)->serialize && UNIT_VTABLE(u)->deserialize_item; +} + +int unit_serialize(Unit *u, FILE *f, FDSet *fds, bool serialize_jobs) { + int r; + + assert(u); + assert(f); + assert(fds); + + if (!unit_can_serialize(u)) + return 0; + + if ((r = UNIT_VTABLE(u)->serialize(u, f, fds)) < 0) + return r; + + + if (serialize_jobs) { + if (u->job) { + fprintf(f, "job\n"); + job_serialize(u->job, f, fds); + } + + if (u->nop_job) { + fprintf(f, "job\n"); + job_serialize(u->nop_job, f, fds); + } + } + + dual_timestamp_serialize(f, "inactive-exit-timestamp", &u->inactive_exit_timestamp); + dual_timestamp_serialize(f, "active-enter-timestamp", &u->active_enter_timestamp); + dual_timestamp_serialize(f, "active-exit-timestamp", &u->active_exit_timestamp); + dual_timestamp_serialize(f, "inactive-enter-timestamp", &u->inactive_enter_timestamp); + dual_timestamp_serialize(f, "condition-timestamp", &u->condition_timestamp); + + if (dual_timestamp_is_set(&u->condition_timestamp)) + unit_serialize_item(u, f, "condition-result", yes_no(u->condition_result)); + + /* End marker */ + fputc('\n', f); + return 0; +} + +void unit_serialize_item_format(Unit *u, FILE *f, const char *key, const char *format, ...) { + va_list ap; + + assert(u); + assert(f); + assert(key); + assert(format); + + fputs(key, f); + fputc('=', f); + + va_start(ap, format); + vfprintf(f, format, ap); + va_end(ap); + + fputc('\n', f); +} + +void unit_serialize_item(Unit *u, FILE *f, const char *key, const char *value) { + assert(u); + assert(f); + assert(key); + assert(value); + + fprintf(f, "%s=%s\n", key, value); +} + +int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { + int r; + + assert(u); + assert(f); + assert(fds); + + if (!unit_can_serialize(u)) + return 0; + + for (;;) { + char line[LINE_MAX], *l, *v; + size_t k; + + if (!fgets(line, sizeof(line), f)) { + if (feof(f)) + return 0; + return -errno; + } + + char_array_0(line); + l = strstrip(line); + + /* End marker */ + if (l[0] == 0) + return 0; + + k = strcspn(l, "="); + + if (l[k] == '=') { + l[k] = 0; + v = l+k+1; + } else + v = l+k; + + if (streq(l, "job")) { + if (v[0] == '\0') { + /* new-style serialized job */ + Job *j = job_new_raw(u); + if (!j) + return -ENOMEM; + + r = job_deserialize(j, f, fds); + if (r < 0) { + job_free(j); + return r; + } + + r = hashmap_put(u->manager->jobs, UINT32_TO_PTR(j->id), j); + if (r < 0) { + job_free(j); + return r; + } + + r = job_install_deserialized(j); + if (r < 0) { + hashmap_remove(u->manager->jobs, UINT32_TO_PTR(j->id)); + job_free(j); + return r; + } + + if (j->state == JOB_RUNNING) + u->manager->n_running_jobs++; + } else { + /* legacy */ + JobType type = job_type_from_string(v); + if (type < 0) + log_debug("Failed to parse job type value %s", v); + else + u->deserialized_job = type; + } + continue; + } else if (streq(l, "inactive-exit-timestamp")) { + dual_timestamp_deserialize(v, &u->inactive_exit_timestamp); + continue; + } else if (streq(l, "active-enter-timestamp")) { + dual_timestamp_deserialize(v, &u->active_enter_timestamp); + continue; + } else if (streq(l, "active-exit-timestamp")) { + dual_timestamp_deserialize(v, &u->active_exit_timestamp); + continue; + } else if (streq(l, "inactive-enter-timestamp")) { + dual_timestamp_deserialize(v, &u->inactive_enter_timestamp); + continue; + } else if (streq(l, "condition-timestamp")) { + dual_timestamp_deserialize(v, &u->condition_timestamp); + continue; + } else if (streq(l, "condition-result")) { + int b; + + if ((b = parse_boolean(v)) < 0) + log_debug("Failed to parse condition result value %s", v); + else + u->condition_result = b; + + continue; + } + + if ((r = UNIT_VTABLE(u)->deserialize_item(u, l, v, fds)) < 0) + return r; + } +} + +int unit_add_node_link(Unit *u, const char *what, bool wants) { + Unit *device; + char *e; + int r; + + assert(u); + + if (!what) + return 0; + + /* Adds in links to the device node that this unit is based on */ + + if (!is_device_path(what)) + return 0; + + e = unit_name_from_path(what, ".device"); + if (!e) + return -ENOMEM; + + r = manager_load_unit(u->manager, e, NULL, NULL, &device); + free(e); + if (r < 0) + return r; + + r = unit_add_two_dependencies(u, UNIT_AFTER, UNIT_BINDS_TO, device, true); + if (r < 0) + return r; + + if (wants) { + r = unit_add_dependency(device, UNIT_WANTS, u, false); + if (r < 0) + return r; + } + + return 0; +} + +int unit_coldplug(Unit *u) { + int r; + + assert(u); + + if (UNIT_VTABLE(u)->coldplug) + if ((r = UNIT_VTABLE(u)->coldplug(u)) < 0) + return r; + + if (u->job) { + r = job_coldplug(u->job); + if (r < 0) + return r; + } else if (u->deserialized_job >= 0) { + /* legacy */ + r = manager_add_job(u->manager, u->deserialized_job, u, JOB_IGNORE_REQUIREMENTS, false, NULL, NULL); + if (r < 0) + return r; + + u->deserialized_job = _JOB_TYPE_INVALID; + } + + return 0; +} + +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wformat-nonliteral" +void unit_status_printf(Unit *u, const char *status, const char *unit_status_msg_format) { + manager_status_printf(u->manager, false, status, unit_status_msg_format, unit_description(u)); +} +#pragma GCC diagnostic pop + +bool unit_need_daemon_reload(Unit *u) { + _cleanup_strv_free_ char **t = NULL; + char **path; + struct stat st; + unsigned loaded_cnt, current_cnt; + + assert(u); + + if (u->fragment_path) { + zero(st); + if (stat(u->fragment_path, &st) < 0) + /* What, cannot access this anymore? */ + return true; + + if (u->fragment_mtime > 0 && + timespec_load(&st.st_mtim) != u->fragment_mtime) + return true; + } + + if (u->source_path) { + zero(st); + if (stat(u->source_path, &st) < 0) + return true; + + if (u->source_mtime > 0 && + timespec_load(&st.st_mtim) != u->source_mtime) + return true; + } + + t = unit_find_dropin_paths(u); + loaded_cnt = strv_length(t); + current_cnt = strv_length(u->dropin_paths); + + if (loaded_cnt == current_cnt) { + if (loaded_cnt == 0) + return false; + + if (strv_overlap(u->dropin_paths, t)) { + STRV_FOREACH(path, u->dropin_paths) { + zero(st); + if (stat(*path, &st) < 0) + return true; + + if (u->dropin_mtime > 0 && + timespec_load(&st.st_mtim) > u->dropin_mtime) + return true; + } + + return false; + } else + return true; + } else + return true; +} + +void unit_reset_failed(Unit *u) { + assert(u); + + if (UNIT_VTABLE(u)->reset_failed) + UNIT_VTABLE(u)->reset_failed(u); +} + +Unit *unit_following(Unit *u) { + assert(u); + + if (UNIT_VTABLE(u)->following) + return UNIT_VTABLE(u)->following(u); + + return NULL; +} + +bool unit_stop_pending(Unit *u) { + assert(u); + + /* This call does check the current state of the unit. It's + * hence useful to be called from state change calls of the + * unit itself, where the state isn't updated yet. This is + * different from unit_inactive_or_pending() which checks both + * the current state and for a queued job. */ + + return u->job && u->job->type == JOB_STOP; +} + +bool unit_inactive_or_pending(Unit *u) { + assert(u); + + /* Returns true if the unit is inactive or going down */ + + if (UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(u))) + return true; + + if (unit_stop_pending(u)) + return true; + + return false; +} + +bool unit_active_or_pending(Unit *u) { + assert(u); + + /* Returns true if the unit is active or going up */ + + if (UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u))) + return true; + + if (u->job && + (u->job->type == JOB_START || + u->job->type == JOB_RELOAD_OR_START || + u->job->type == JOB_RESTART)) + return true; + + return false; +} + +int unit_kill(Unit *u, KillWho w, int signo, DBusError *error) { + assert(u); + assert(w >= 0 && w < _KILL_WHO_MAX); + assert(signo > 0); + assert(signo < _NSIG); + + if (!UNIT_VTABLE(u)->kill) + return -ENOTSUP; + + return UNIT_VTABLE(u)->kill(u, w, signo, error); +} + +int unit_kill_common( + Unit *u, + KillWho who, + int signo, + pid_t main_pid, + pid_t control_pid, + DBusError *error) { + + int r = 0; + + if (who == KILL_MAIN && main_pid <= 0) { + if (main_pid < 0) + dbus_set_error(error, BUS_ERROR_NO_SUCH_PROCESS, "%s units have no main processes", unit_type_to_string(u->type)); + else + dbus_set_error(error, BUS_ERROR_NO_SUCH_PROCESS, "No main process to kill"); + return -ESRCH; + } + + if (who == KILL_CONTROL && control_pid <= 0) { + if (control_pid < 0) + dbus_set_error(error, BUS_ERROR_NO_SUCH_PROCESS, "%s units have no control processes", unit_type_to_string(u->type)); + else + dbus_set_error(error, BUS_ERROR_NO_SUCH_PROCESS, "No control process to kill"); + return -ESRCH; + } + + if (who == KILL_CONTROL || who == KILL_ALL) + if (control_pid > 0) + if (kill(control_pid, signo) < 0) + r = -errno; + + if (who == KILL_MAIN || who == KILL_ALL) + if (main_pid > 0) + if (kill(main_pid, signo) < 0) + r = -errno; + + if (who == KILL_ALL) { + _cleanup_set_free_ Set *pid_set = NULL; + int q; + + pid_set = set_new(trivial_hash_func, trivial_compare_func); + if (!pid_set) + return -ENOMEM; + + /* Exclude the control/main pid from being killed via the cgroup */ + if (control_pid > 0) { + q = set_put(pid_set, LONG_TO_PTR(control_pid)); + if (q < 0) + return q; + } + + if (main_pid > 0) { + q = set_put(pid_set, LONG_TO_PTR(main_pid)); + if (q < 0) + return q; + } + + q = cgroup_bonding_kill_list(u->cgroup_bondings, signo, false, false, pid_set, NULL); + if (q < 0 && q != -EAGAIN && q != -ESRCH && q != -ENOENT) + r = q; + } + + return r; +} + +int unit_following_set(Unit *u, Set **s) { + assert(u); + assert(s); + + if (UNIT_VTABLE(u)->following_set) + return UNIT_VTABLE(u)->following_set(u, s); + + *s = NULL; + return 0; +} + +UnitFileState unit_get_unit_file_state(Unit *u) { + assert(u); + + if (u->unit_file_state < 0 && u->fragment_path) + u->unit_file_state = unit_file_get_state( + u->manager->running_as == SYSTEMD_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, + NULL, path_get_file_name(u->fragment_path)); + + return u->unit_file_state; +} + +Unit* unit_ref_set(UnitRef *ref, Unit *u) { + assert(ref); + assert(u); + + if (ref->unit) + unit_ref_unset(ref); + + ref->unit = u; + LIST_PREPEND(UnitRef, refs, u->refs, ref); + return u; +} + +void unit_ref_unset(UnitRef *ref) { + assert(ref); + + if (!ref->unit) + return; + + LIST_REMOVE(UnitRef, refs, ref->unit->refs, ref); + ref->unit = NULL; +} + +int unit_add_one_mount_link(Unit *u, Mount *m) { + char **i; + + assert(u); + assert(m); + + if (u->load_state != UNIT_LOADED || + UNIT(m)->load_state != UNIT_LOADED) + return 0; + + STRV_FOREACH(i, u->requires_mounts_for) { + + if (UNIT(m) == u) + continue; + + if (!path_startswith(*i, m->where)) + continue; + + return unit_add_two_dependencies(u, UNIT_AFTER, UNIT_REQUIRES, UNIT(m), true); + } + + return 0; +} + +int unit_add_mount_links(Unit *u) { + Unit *other; + int r; + + assert(u); + + LIST_FOREACH(units_by_type, other, u->manager->units_by_type[UNIT_MOUNT]) { + r = unit_add_one_mount_link(u, MOUNT(other)); + if (r < 0) + return r; + } + + return 0; +} + +int unit_exec_context_defaults(Unit *u, ExecContext *c) { + unsigned i; + int r; + + assert(u); + assert(c); + + /* This only copies in the ones that need memory */ + + for (i = 0; i < RLIMIT_NLIMITS; i++) + if (u->manager->rlimit[i] && !c->rlimit[i]) { + c->rlimit[i] = newdup(struct rlimit, u->manager->rlimit[i], 1); + if (!c->rlimit[i]) + return -ENOMEM; + } + + if (u->manager->running_as == SYSTEMD_USER && + !c->working_directory) { + + r = get_home_dir(&c->working_directory); + if (r < 0) + return r; + } + + return 0; +} + +ExecContext *unit_get_exec_context(Unit *u) { + size_t offset; + assert(u); + + offset = UNIT_VTABLE(u)->exec_context_offset; + if (offset <= 0) + return NULL; + + return (ExecContext*) ((uint8_t*) u + offset); +} + +static int drop_in_file(Unit *u, bool runtime, const char *name, char **_p, char **_q) { + char *p, *q; + int r; + + assert(u); + assert(name); + assert(_p); + assert(_q); + + if (u->manager->running_as == SYSTEMD_USER && runtime) + return -ENOTSUP; + + if (!filename_is_safe(name)) + return -EINVAL; + + if (u->manager->running_as == SYSTEMD_USER) { + _cleanup_free_ char *c = NULL; + + r = user_config_home(&c); + if (r < 0) + return r; + if (r == 0) + return -ENOENT; + + p = strjoin(c, "/", u->id, ".d", NULL); + } else if (runtime) + p = strjoin("/run/systemd/system/", u->id, ".d", NULL); + else + p = strjoin("/etc/systemd/system/", u->id, ".d", NULL); + if (!p) + return -ENOMEM; + + q = strjoin(p, "/50-", name, ".conf", NULL); + if (!q) { + free(p); + return -ENOMEM; + } + + *_p = p; + *_q = q; + return 0; +} + +int unit_write_drop_in(Unit *u, bool runtime, const char *name, const char *data) { + _cleanup_free_ char *p = NULL, *q = NULL; + int r; + + assert(u); + + r = drop_in_file(u, runtime, name, &p, &q); + if (r < 0) + return r; + + mkdir_p(p, 0755); + return write_string_file_atomic_label(q, data); +} + +int unit_remove_drop_in(Unit *u, bool runtime, const char *name) { + _cleanup_free_ char *p = NULL, *q = NULL; + int r; + + assert(u); + + r = drop_in_file(u, runtime, name, &p, &q); + if (unlink(q) < 0) + r = -errno; + else + r = 0; + + rmdir(p); + return r; +} + +int unit_kill_context( + Unit *u, + KillContext *c, + bool sigkill, + pid_t main_pid, + pid_t control_pid, + bool main_pid_alien) { + + int sig, wait_for_exit = 0, r; + + assert(u); + assert(c); + + if (c->kill_mode == KILL_NONE) + return 0; + + sig = sigkill ? SIGKILL : c->kill_signal; + + if (main_pid > 0) { + r = kill_and_sigcont(main_pid, sig); + + if (r < 0 && r != -ESRCH) { + _cleanup_free_ char *comm = NULL; + get_process_comm(main_pid, &comm); + + log_warning_unit(u->id, "Failed to kill main process %li (%s): %s", + (long) main_pid, strna(comm), strerror(-r)); + } else + wait_for_exit = !main_pid_alien; + } + + if (control_pid > 0) { + r = kill_and_sigcont(control_pid, sig); + + if (r < 0 && r != -ESRCH) { + _cleanup_free_ char *comm = NULL; + get_process_comm(control_pid, &comm); + + log_warning_unit(u->id, + "Failed to kill control process %li (%s): %s", + (long) control_pid, strna(comm), strerror(-r)); + } else + wait_for_exit = true; + } + + if (c->kill_mode == KILL_CONTROL_GROUP) { + _cleanup_set_free_ Set *pid_set = NULL; + + pid_set = set_new(trivial_hash_func, trivial_compare_func); + if (!pid_set) + return -ENOMEM; + + /* Exclude the main/control pids from being killed via the cgroup */ + if (main_pid > 0) { + r = set_put(pid_set, LONG_TO_PTR(main_pid)); + if (r < 0) + return r; + } + + if (control_pid > 0) { + r = set_put(pid_set, LONG_TO_PTR(control_pid)); + if (r < 0) + return r; + } + + r = cgroup_bonding_kill_list(u->cgroup_bondings, sig, true, false, pid_set, NULL); + if (r < 0) { + if (r != -EAGAIN && r != -ESRCH && r != -ENOENT) + log_warning_unit(u->id, "Failed to kill control group: %s", strerror(-r)); + } else if (r > 0) + wait_for_exit = true; + } + + return wait_for_exit; +} + +static const char* const unit_active_state_table[_UNIT_ACTIVE_STATE_MAX] = { + [UNIT_ACTIVE] = "active", + [UNIT_RELOADING] = "reloading", + [UNIT_INACTIVE] = "inactive", + [UNIT_FAILED] = "failed", + [UNIT_ACTIVATING] = "activating", + [UNIT_DEACTIVATING] = "deactivating" +}; + +DEFINE_STRING_TABLE_LOOKUP(unit_active_state, UnitActiveState); + +static const char* const unit_dependency_table[_UNIT_DEPENDENCY_MAX] = { + [UNIT_REQUIRES] = "Requires", + [UNIT_REQUIRES_OVERRIDABLE] = "RequiresOverridable", + [UNIT_REQUISITE] = "Requisite", + [UNIT_REQUISITE_OVERRIDABLE] = "RequisiteOverridable", + [UNIT_WANTS] = "Wants", + [UNIT_BINDS_TO] = "BindsTo", + [UNIT_PART_OF] = "PartOf", + [UNIT_REQUIRED_BY] = "RequiredBy", + [UNIT_REQUIRED_BY_OVERRIDABLE] = "RequiredByOverridable", + [UNIT_WANTED_BY] = "WantedBy", + [UNIT_BOUND_BY] = "BoundBy", + [UNIT_CONSISTS_OF] = "ConsistsOf", + [UNIT_CONFLICTS] = "Conflicts", + [UNIT_CONFLICTED_BY] = "ConflictedBy", + [UNIT_BEFORE] = "Before", + [UNIT_AFTER] = "After", + [UNIT_ON_FAILURE] = "OnFailure", + [UNIT_TRIGGERS] = "Triggers", + [UNIT_TRIGGERED_BY] = "TriggeredBy", + [UNIT_PROPAGATES_RELOAD_TO] = "PropagatesReloadTo", + [UNIT_RELOAD_PROPAGATED_FROM] = "ReloadPropagatedFrom", + [UNIT_REFERENCES] = "References", + [UNIT_REFERENCED_BY] = "ReferencedBy", +}; + +DEFINE_STRING_TABLE_LOOKUP(unit_dependency, UnitDependency); diff --git a/src/core/unit.h b/src/core/unit.h new file mode 100644 index 0000000000..b04475e4fb --- /dev/null +++ b/src/core/unit.h @@ -0,0 +1,588 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> +#include <stdlib.h> +#include <unistd.h> + +typedef struct Unit Unit; +typedef struct UnitVTable UnitVTable; +typedef enum UnitActiveState UnitActiveState; +typedef enum UnitDependency UnitDependency; +typedef struct UnitRef UnitRef; +typedef struct UnitStatusMessageFormats UnitStatusMessageFormats; + +#include "set.h" +#include "util.h" +#include "list.h" +#include "socket-util.h" +#include "execute.h" +#include "condition.h" +#include "install.h" +#include "unit-name.h" +#include "cgroup-semantics.h" + +enum UnitActiveState { + UNIT_ACTIVE, + UNIT_RELOADING, + UNIT_INACTIVE, + UNIT_FAILED, + UNIT_ACTIVATING, + UNIT_DEACTIVATING, + _UNIT_ACTIVE_STATE_MAX, + _UNIT_ACTIVE_STATE_INVALID = -1 +}; + +static inline bool UNIT_IS_ACTIVE_OR_RELOADING(UnitActiveState t) { + return t == UNIT_ACTIVE || t == UNIT_RELOADING; +} + +static inline bool UNIT_IS_ACTIVE_OR_ACTIVATING(UnitActiveState t) { + return t == UNIT_ACTIVE || t == UNIT_ACTIVATING || t == UNIT_RELOADING; +} + +static inline bool UNIT_IS_INACTIVE_OR_DEACTIVATING(UnitActiveState t) { + return t == UNIT_INACTIVE || t == UNIT_FAILED || t == UNIT_DEACTIVATING; +} + +static inline bool UNIT_IS_INACTIVE_OR_FAILED(UnitActiveState t) { + return t == UNIT_INACTIVE || t == UNIT_FAILED; +} + +enum UnitDependency { + /* Positive dependencies */ + UNIT_REQUIRES, + UNIT_REQUIRES_OVERRIDABLE, + UNIT_REQUISITE, + UNIT_REQUISITE_OVERRIDABLE, + UNIT_WANTS, + UNIT_BINDS_TO, + UNIT_PART_OF, + + /* Inverse of the above */ + UNIT_REQUIRED_BY, /* inverse of 'requires' and 'requisite' is 'required_by' */ + UNIT_REQUIRED_BY_OVERRIDABLE, /* inverse of 'requires_overridable' and 'requisite_overridable' is 'soft_required_by' */ + UNIT_WANTED_BY, /* inverse of 'wants' */ + UNIT_BOUND_BY, /* inverse of 'binds_to' */ + UNIT_CONSISTS_OF, /* inverse of 'part_of' */ + + /* Negative dependencies */ + UNIT_CONFLICTS, /* inverse of 'conflicts' is 'conflicted_by' */ + UNIT_CONFLICTED_BY, + + /* Order */ + UNIT_BEFORE, /* inverse of 'before' is 'after' and vice versa */ + UNIT_AFTER, + + /* On Failure */ + UNIT_ON_FAILURE, + + /* Triggers (i.e. a socket triggers a service) */ + UNIT_TRIGGERS, + UNIT_TRIGGERED_BY, + + /* Propagate reloads */ + UNIT_PROPAGATES_RELOAD_TO, + UNIT_RELOAD_PROPAGATED_FROM, + + /* Reference information for GC logic */ + UNIT_REFERENCES, /* Inverse of 'references' is 'referenced_by' */ + UNIT_REFERENCED_BY, + + _UNIT_DEPENDENCY_MAX, + _UNIT_DEPENDENCY_INVALID = -1 +}; + +#include "manager.h" +#include "job.h" +#include "cgroup.h" +#include "cgroup-attr.h" + +struct Unit { + Manager *manager; + + UnitType type; + UnitLoadState load_state; + Unit *merged_into; + + char *id; /* One name is special because we use it for identification. Points to an entry in the names set */ + char *instance; + + Set *names; + Set *dependencies[_UNIT_DEPENDENCY_MAX]; + + char **requires_mounts_for; + + char *description; + char **documentation; + + char *fragment_path; /* if loaded from a config file this is the primary path to it */ + char *source_path; /* if converted, the source file */ + char **dropin_paths; + usec_t fragment_mtime; + usec_t source_mtime; + usec_t dropin_mtime; + + /* If there is something to do with this unit, then this is the installed job for it */ + Job *job; + + /* JOB_NOP jobs are special and can be installed without disturbing the real job. */ + Job *nop_job; + + usec_t job_timeout; + + /* References to this */ + LIST_HEAD(UnitRef, refs); + + /* Conditions to check */ + LIST_HEAD(Condition, conditions); + + dual_timestamp condition_timestamp; + + dual_timestamp inactive_exit_timestamp; + dual_timestamp active_enter_timestamp; + dual_timestamp active_exit_timestamp; + dual_timestamp inactive_enter_timestamp; + + /* Counterparts in the cgroup filesystem */ + CGroupBonding *cgroup_bondings; + CGroupAttribute *cgroup_attributes; + + /* Per type list */ + LIST_FIELDS(Unit, units_by_type); + + /* All units which have requires_mounts_for set */ + LIST_FIELDS(Unit, has_requires_mounts_for); + + /* Load queue */ + LIST_FIELDS(Unit, load_queue); + + /* D-Bus queue */ + LIST_FIELDS(Unit, dbus_queue); + + /* Cleanup queue */ + LIST_FIELDS(Unit, cleanup_queue); + + /* GC queue */ + LIST_FIELDS(Unit, gc_queue); + + /* Used during GC sweeps */ + unsigned gc_marker; + + /* When deserializing, temporarily store the job type for this + * unit here, if there was a job scheduled. + * Only for deserializing from a legacy version. New style uses full + * serialized jobs. */ + int deserialized_job; /* This is actually of type JobType */ + + /* Error code when we didn't manage to load the unit (negative) */ + int load_error; + + /* Cached unit file state */ + UnitFileState unit_file_state; + + /* Garbage collect us we nobody wants or requires us anymore */ + bool stop_when_unneeded; + + /* Create default dependencies */ + bool default_dependencies; + + /* Refuse manual starting, allow starting only indirectly via dependency. */ + bool refuse_manual_start; + + /* Don't allow the user to stop this unit manually, allow stopping only indirectly via dependency. */ + bool refuse_manual_stop; + + /* Allow isolation requests */ + bool allow_isolate; + + /* Isolate OnFailure unit */ + bool on_failure_isolate; + + /* Ignore this unit when isolating */ + bool ignore_on_isolate; + + /* Ignore this unit when snapshotting */ + bool ignore_on_snapshot; + + /* Did the last condition check succeed? */ + bool condition_result; + + bool in_load_queue:1; + bool in_dbus_queue:1; + bool in_cleanup_queue:1; + bool in_gc_queue:1; + + bool sent_dbus_new_signal:1; + + bool no_gc:1; + + bool in_audit:1; +}; + +struct UnitRef { + /* Keeps tracks of references to a unit. This is useful so + * that we can merge two units if necessary and correct all + * references to them */ + + Unit* unit; + LIST_FIELDS(UnitRef, refs); +}; + +struct UnitStatusMessageFormats { + const char *starting_stopping[2]; + const char *finished_start_job[_JOB_RESULT_MAX]; + const char *finished_stop_job[_JOB_RESULT_MAX]; +}; + +#include "service.h" +#include "timer.h" +#include "socket.h" +#include "target.h" +#include "device.h" +#include "mount.h" +#include "automount.h" +#include "snapshot.h" +#include "swap.h" +#include "path.h" + +struct UnitVTable { + /* How much memory does an object of this unit type need */ + size_t object_size; + + /* If greater than 0, the offset into the object where + * ExecContext is found, if the unit type has that */ + size_t exec_context_offset; + + /* The name of the section with the exec settings of ExecContext */ + const char *exec_section; + + /* Config file sections this unit type understands, separated + * by NUL chars */ + const char *sections; + + /* This should reset all type-specific variables. This should + * not allocate memory, and is called with zero-initialized + * data. It should hence only initialize variables that need + * to be set != 0. */ + void (*init)(Unit *u); + + /* This should free all type-specific variables. It should be + * idempotent. */ + void (*done)(Unit *u); + + /* Actually load data from disk. This may fail, and should set + * load_state to UNIT_LOADED, UNIT_MERGED or leave it at + * UNIT_STUB if no configuration could be found. */ + int (*load)(Unit *u); + + /* If a lot of units got created via enumerate(), this is + * where to actually set the state and call unit_notify(). */ + int (*coldplug)(Unit *u); + + void (*dump)(Unit *u, FILE *f, const char *prefix); + + int (*start)(Unit *u); + int (*stop)(Unit *u); + int (*reload)(Unit *u); + + int (*kill)(Unit *u, KillWho w, int signo, DBusError *error); + + bool (*can_reload)(Unit *u); + + /* Write all data that cannot be restored from other sources + * away using unit_serialize_item() */ + int (*serialize)(Unit *u, FILE *f, FDSet *fds); + + /* Restore one item from the serialization */ + int (*deserialize_item)(Unit *u, const char *key, const char *data, FDSet *fds); + + /* Try to match up fds with what we need for this unit */ + int (*distribute_fds)(Unit *u, FDSet *fds); + + /* Boils down the more complex internal state of this unit to + * a simpler one that the engine can understand */ + UnitActiveState (*active_state)(Unit *u); + + /* Returns the substate specific to this unit type as + * string. This is purely information so that we can give the + * user a more fine grained explanation in which actual state a + * unit is in. */ + const char* (*sub_state_to_string)(Unit *u); + + /* Return true when there is reason to keep this entry around + * even nothing references it and it isn't active in any + * way */ + bool (*check_gc)(Unit *u); + + /* Return true when this unit is suitable for snapshotting */ + bool (*check_snapshot)(Unit *u); + + void (*fd_event)(Unit *u, int fd, uint32_t events, Watch *w); + void (*sigchld_event)(Unit *u, pid_t pid, int code, int status); + void (*timer_event)(Unit *u, uint64_t n_elapsed, Watch *w); + + /* Reset failed state if we are in failed state */ + void (*reset_failed)(Unit *u); + + /* Called whenever any of the cgroups this unit watches for + * ran empty */ + void (*cgroup_notify_empty)(Unit *u); + + /* Called whenever a process of this unit sends us a message */ + void (*notify_message)(Unit *u, pid_t pid, char **tags); + + /* Called whenever a name this Unit registered for comes or + * goes away. */ + void (*bus_name_owner_change)(Unit *u, const char *name, const char *old_owner, const char *new_owner); + + /* Called whenever a bus PID lookup finishes */ + void (*bus_query_pid_done)(Unit *u, const char *name, pid_t pid); + + /* Called for each message received on the bus */ + DBusHandlerResult (*bus_message_handler)(Unit *u, DBusConnection *c, DBusMessage *message); + + /* Return the unit this unit is following */ + Unit *(*following)(Unit *u); + + /* Return the set of units that are following each other */ + int (*following_set)(Unit *u, Set **s); + + /* Invoked each time a unit this unit is triggering changes + * state or gains/loses a job */ + void (*trigger_notify)(Unit *u, Unit *trigger); + + /* Called whenever CLOCK_REALTIME made a jump */ + void (*time_change)(Unit *u); + + /* This is called for each unit type and should be used to + * enumerate existing devices and load them. However, + * everything that is loaded here should still stay in + * inactive state. It is the job of the coldplug() call above + * to put the units into the initial state. */ + int (*enumerate)(Manager *m); + + /* Type specific cleanups. */ + void (*shutdown)(Manager *m); + + /* When sending out PropertiesChanged signal, which properties + * shall be invalidated? This is a NUL separated list of + * strings, to minimize relocations a little. */ + const char *bus_invalidating_properties; + + /* The interface name */ + const char *bus_interface; + + UnitStatusMessageFormats status_message_formats; + + /* Can units of this type have multiple names? */ + bool no_alias:1; + + /* Instances make no sense for this type */ + bool no_instances:1; + + /* Exclude from automatic gc */ + bool no_gc:1; +}; + +extern const UnitVTable * const unit_vtable[_UNIT_TYPE_MAX]; + +#define UNIT_VTABLE(u) unit_vtable[(u)->type] + +/* For casting a unit into the various unit types */ +#define DEFINE_CAST(UPPERCASE, MixedCase) \ + static inline MixedCase* UPPERCASE(Unit *u) { \ + if (_unlikely_(!u || u->type != UNIT_##UPPERCASE)) \ + return NULL; \ + \ + return (MixedCase*) u; \ + } + +/* For casting the various unit types into a unit */ +#define UNIT(u) (&(u)->meta) + +#define UNIT_TRIGGER(u) ((Unit*) set_first((u)->dependencies[UNIT_TRIGGERS])) + +DEFINE_CAST(SOCKET, Socket); +DEFINE_CAST(TIMER, Timer); +DEFINE_CAST(SERVICE, Service); +DEFINE_CAST(TARGET, Target); +DEFINE_CAST(DEVICE, Device); +DEFINE_CAST(MOUNT, Mount); +DEFINE_CAST(AUTOMOUNT, Automount); +DEFINE_CAST(SNAPSHOT, Snapshot); +DEFINE_CAST(SWAP, Swap); +DEFINE_CAST(PATH, Path); + +Unit *unit_new(Manager *m, size_t size); +void unit_free(Unit *u); + +int unit_add_name(Unit *u, const char *name); + +int unit_add_dependency(Unit *u, UnitDependency d, Unit *other, bool add_reference); +int unit_add_two_dependencies(Unit *u, UnitDependency d, UnitDependency e, Unit *other, bool add_reference); + +int unit_add_dependency_by_name(Unit *u, UnitDependency d, const char *name, const char *filename, bool add_reference); +int unit_add_two_dependencies_by_name(Unit *u, UnitDependency d, UnitDependency e, const char *name, const char *path, bool add_reference); + +int unit_add_dependency_by_name_inverse(Unit *u, UnitDependency d, const char *name, const char *filename, bool add_reference); +int unit_add_two_dependencies_by_name_inverse(Unit *u, UnitDependency d, UnitDependency e, const char *name, const char *path, bool add_reference); + +int unit_add_exec_dependencies(Unit *u, ExecContext *c); + +int unit_add_cgroup_from_text(Unit *u, const char *name, bool overwrite, CGroupBonding **ret); +int unit_add_default_cgroups(Unit *u); +CGroupBonding* unit_get_default_cgroup(Unit *u); +int unit_add_cgroup_attribute(Unit *u, const CGroupSemantics *semantics, const char *controller, const char *name, const char *value, CGroupAttribute **ret); + +int unit_choose_id(Unit *u, const char *name); +int unit_set_description(Unit *u, const char *description); + +bool unit_check_gc(Unit *u); + +void unit_add_to_load_queue(Unit *u); +void unit_add_to_dbus_queue(Unit *u); +void unit_add_to_cleanup_queue(Unit *u); +void unit_add_to_gc_queue(Unit *u); + +int unit_merge(Unit *u, Unit *other); +int unit_merge_by_name(Unit *u, const char *other); + +Unit *unit_follow_merge(Unit *u) _pure_; + +int unit_load_fragment_and_dropin(Unit *u); +int unit_load_fragment_and_dropin_optional(Unit *u); +int unit_load(Unit *unit); + +const char *unit_description(Unit *u) _pure_; + +bool unit_has_name(Unit *u, const char *name); + +UnitActiveState unit_active_state(Unit *u); + +const char* unit_sub_state_to_string(Unit *u); + +void unit_dump(Unit *u, FILE *f, const char *prefix); + +bool unit_can_reload(Unit *u) _pure_; +bool unit_can_start(Unit *u) _pure_; +bool unit_can_isolate(Unit *u) _pure_; + +int unit_start(Unit *u); +int unit_stop(Unit *u); +int unit_reload(Unit *u); + +int unit_kill(Unit *u, KillWho w, int signo, DBusError *error); +int unit_kill_common(Unit *u, KillWho who, int signo, pid_t main_pid, pid_t control_pid, DBusError *error); + +void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_success); + +int unit_watch_fd(Unit *u, int fd, uint32_t events, Watch *w); +void unit_unwatch_fd(Unit *u, Watch *w); + +int unit_watch_pid(Unit *u, pid_t pid); +void unit_unwatch_pid(Unit *u, pid_t pid); + +int unit_watch_timer(Unit *u, clockid_t, bool relative, usec_t usec, Watch *w); +void unit_unwatch_timer(Unit *u, Watch *w); + +int unit_watch_bus_name(Unit *u, const char *name); +void unit_unwatch_bus_name(Unit *u, const char *name); + +bool unit_job_is_applicable(Unit *u, JobType j); + +int set_unit_path(const char *p); + +char *unit_dbus_path(Unit *u); + +int unit_load_related_unit(Unit *u, const char *type, Unit **_found); +int unit_get_related_unit(Unit *u, const char *type, Unit **_found); + +bool unit_can_serialize(Unit *u) _pure_; +int unit_serialize(Unit *u, FILE *f, FDSet *fds, bool serialize_jobs); +void unit_serialize_item_format(Unit *u, FILE *f, const char *key, const char *value, ...) _printf_attr_(4,5); +void unit_serialize_item(Unit *u, FILE *f, const char *key, const char *value); +int unit_deserialize(Unit *u, FILE *f, FDSet *fds); + +int unit_add_node_link(Unit *u, const char *what, bool wants); + +int unit_coldplug(Unit *u); + +void unit_status_printf(Unit *u, const char *status, const char *unit_status_msg_format) _printf_attr_(3, 0); + +bool unit_need_daemon_reload(Unit *u); + +void unit_reset_failed(Unit *u); + +Unit *unit_following(Unit *u); + +bool unit_stop_pending(Unit *u) _pure_; +bool unit_inactive_or_pending(Unit *u) _pure_; +bool unit_active_or_pending(Unit *u); + +int unit_add_default_target_dependency(Unit *u, Unit *target); + +char *unit_default_cgroup_path(Unit *u); + +int unit_following_set(Unit *u, Set **s); + +void unit_start_on_failure(Unit *u); +void unit_trigger_notify(Unit *u); + +bool unit_condition_test(Unit *u); + +UnitFileState unit_get_unit_file_state(Unit *u); + +Unit* unit_ref_set(UnitRef *ref, Unit *u); +void unit_ref_unset(UnitRef *ref); + +#define UNIT_DEREF(ref) ((ref).unit) + +int unit_add_one_mount_link(Unit *u, Mount *m); +int unit_add_mount_links(Unit *u); + +int unit_exec_context_defaults(Unit *u, ExecContext *c); + +ExecContext *unit_get_exec_context(Unit *u) _pure_; + +int unit_write_drop_in(Unit *u, bool runtime, const char *name, const char *data); +int unit_remove_drop_in(Unit *u, bool runtime, const char *name); + +int unit_kill_context(Unit *u, KillContext *c, bool sigkill, pid_t main_pid, pid_t control_pid, bool main_pid_alien); + +const char *unit_active_state_to_string(UnitActiveState i) _const_; +UnitActiveState unit_active_state_from_string(const char *s) _pure_; + +const char *unit_dependency_to_string(UnitDependency i) _const_; +UnitDependency unit_dependency_from_string(const char *s) _pure_; + +/* Macros which append UNIT= or USER_UNIT= to the message */ + +#define log_full_unit(level, unit, ...) log_meta_object(level, __FILE__, __LINE__, __func__, getpid() == 1 ? "UNIT=" : "USER_UNIT=", unit, __VA_ARGS__) +#define log_debug_unit(unit, ...) log_full_unit(LOG_DEBUG, unit, __VA_ARGS__) +#define log_info_unit(unit, ...) log_full_unit(LOG_INFO, unit, __VA_ARGS__) +#define log_notice_unit(unit, ...) log_full_unit(LOG_NOTICE, unit, __VA_ARGS__) +#define log_warning_unit(unit, ...) log_full_unit(LOG_WARNING, unit, __VA_ARGS__) +#define log_error_unit(unit, ...) log_full_unit(LOG_ERR, unit, __VA_ARGS__) + +#define log_struct_unit(level, unit, ...) log_struct(level, getpid() == 1 ? "UNIT=%s" : "USER_UNIT=%s", unit, __VA_ARGS__) diff --git a/src/core/user.conf b/src/core/user.conf new file mode 100644 index 0000000000..4252451eb7 --- /dev/null +++ b/src/core/user.conf @@ -0,0 +1,17 @@ +# This file is part of systemd. +# +# systemd is free software; you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published by +# the Free Software Foundation; either version 2.1 of the License, or +# (at your option) any later version. +# +# See systemd-user.conf(5) for details + +[Manager] +#LogLevel=info +#LogTarget=console +#LogColor=yes +#LogLocation=no +#DefaultControllers=cpu +#DefaultStandardOutput=inherit +#DefaultStandardError=inherit |