/*
Unix SMB/CIFS implementation.
Locking functions
Copyright (C) Andrew Tridgell 1992-2000
Copyright (C) Jeremy Allison 1992-2006
Copyright (C) Volker Lendecke 2005
This program 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 3 of the License, or
(at your option) any later version.
This program 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 this program. If not, see .
Revision History:
12 aug 96: Erik.Devriendt@te6.siemens.be
added support for shared memory implementation of share mode locking
May 1997. Jeremy Allison (jallison@whistle.com). Modified share mode
locking to deal with multiple share modes per open file.
September 1997. Jeremy Allison (jallison@whistle.com). Added oplock
support.
rewritten completely to use new tdb code. Tridge, Dec '99
Added POSIX locking support. Jeremy Allison (jeremy@valinux.com), Apr. 2000.
Added Unix Extensions POSIX locking support. Jeremy Allison Mar 2006.
*/
#include "includes.h"
#include "system/filesys.h"
#include "lib/util/server_id.h"
#include "locking/proto.h"
#include "smbd/globals.h"
#include "dbwrap/dbwrap.h"
#include "dbwrap/dbwrap_open.h"
#include "../libcli/security/security.h"
#include "serverid.h"
#include "messages.h"
#include "util_tdb.h"
#include "../librpc/gen_ndr/ndr_open_files.h"
#include "source3/lib/dbwrap/dbwrap_watch.h"
#include "locking/leases_db.h"
#include "../lib/util/memcache.h"
#include "lib/util/tevent_ntstatus.h"
#undef DBGC_CLASS
#define DBGC_CLASS DBGC_LOCKING
#define NO_LOCKING_COUNT (-1)
/* the locking database handle */
static struct db_context *lock_db;
static struct db_context *share_entries_db;
static bool locking_init_internal(bool read_only)
{
struct db_context *backend;
char *db_path;
brl_init(read_only);
if (lock_db)
return True;
db_path = lock_path(talloc_tos(), "locking.tdb");
if (db_path == NULL) {
return false;
}
backend = db_open(NULL, db_path,
SMB_OPEN_DATABASE_TDB_HASH_SIZE,
TDB_DEFAULT|
TDB_VOLATILE|
TDB_CLEAR_IF_FIRST|
TDB_INCOMPATIBLE_HASH|
TDB_SEQNUM,
read_only?O_RDONLY:O_RDWR|O_CREAT, 0644,
DBWRAP_LOCK_ORDER_1, DBWRAP_FLAG_NONE);
TALLOC_FREE(db_path);
if (!backend) {
DEBUG(0,("ERROR: Failed to initialise locking database\n"));
return False;
}
lock_db = db_open_watched(NULL, &backend, global_messaging_context());
if (lock_db == NULL) {
DBG_ERR("db_open_watched failed\n");
TALLOC_FREE(backend);
return false;
}
db_path = lock_path(talloc_tos(), "share_entries.tdb");
if (db_path == NULL) {
return false;
}
share_entries_db = db_open(
NULL, db_path,
SMB_OPEN_DATABASE_TDB_HASH_SIZE,
TDB_DEFAULT|
TDB_VOLATILE|
TDB_CLEAR_IF_FIRST|
TDB_INCOMPATIBLE_HASH,
read_only?O_RDONLY:O_RDWR|O_CREAT, 0644,
DBWRAP_LOCK_ORDER_3, DBWRAP_FLAG_NONE);
TALLOC_FREE(db_path);
if (share_entries_db == NULL) {
TALLOC_FREE(lock_db);
return false;
}
if (!posix_locking_init(read_only)) {
TALLOC_FREE(share_entries_db);
TALLOC_FREE(lock_db);
return False;
}
return True;
}
bool locking_init(void)
{
return locking_init_internal(false);
}
bool locking_init_readonly(void)
{
return locking_init_internal(true);
}
/*******************************************************************
Deinitialize the share_mode management.
******************************************************************/
bool locking_end(void)
{
brl_shutdown();
TALLOC_FREE(lock_db);
return true;
}
/*******************************************************************
Form a static locking key for a dev/inode pair.
******************************************************************/
static TDB_DATA locking_key(const struct file_id *id)
{
return make_tdb_data((const uint8_t *)id, sizeof(*id));
}
/*******************************************************************
Share mode cache utility functions that store/delete/retrieve
entries from memcache.
For now share the statcache (global cache) memory space. If
a lock record gets orphaned (which shouldn't happen as we're
using the same locking_key data as lookup) it will eventually
fall out of the cache via the normal LRU trim mechanism. If
necessary we can always make this a separate (smaller) cache.
******************************************************************/
static DATA_BLOB memcache_key(const struct file_id *id)
{
return data_blob_const((const void *)id, sizeof(*id));
}
static void share_mode_memcache_store(struct share_mode_data *d)
{
const DATA_BLOB key = memcache_key(&d->id);
struct file_id_buf idbuf;
DBG_DEBUG("stored entry for file %s seq %"PRIx64" key %s\n",
d->base_name,
d->sequence_number,
file_id_str_buf(d->id, &idbuf));
/* Ensure everything stored in the cache is pristine. */
d->modified = false;
d->fresh = false;
/*
* Ensure the memory going into the cache
* doesn't have a destructor so it can be
* cleanly evicted by the memcache LRU
* mechanism.
*/
talloc_set_destructor(d, NULL);
/* Cache will own d after this call. */
memcache_add_talloc(NULL,
SHARE_MODE_LOCK_CACHE,
key,
&d);
}
/*
* NB. We use ndr_pull_hyper on a stack-created
* struct ndr_pull with no talloc allowed, as we
* need this to be really fast as an ndr-peek into
* the first 9 bytes of the blob.
*/
static enum ndr_err_code get_share_mode_blob_header(
DATA_BLOB *blob, uint64_t *pseq, uint16_t *pflags)
{
struct ndr_pull ndr = {.data = blob->data, .data_size = blob->length};
NDR_CHECK(ndr_pull_hyper(&ndr, NDR_SCALARS, pseq));
NDR_CHECK(ndr_pull_uint16(&ndr, NDR_SCALARS, pflags));
return NDR_ERR_SUCCESS;
}
struct fsp_update_share_mode_flags_state {
enum ndr_err_code ndr_err;
uint16_t share_mode_flags;
};
static void fsp_update_share_mode_flags_fn(
TDB_DATA value, bool *modified_dependent, void *private_data)
{
struct fsp_update_share_mode_flags_state *state = private_data;
DATA_BLOB blob = { .data = value.dptr, .length = value.dsize };
uint64_t seq;
state->ndr_err = get_share_mode_blob_header(
&blob, &seq, &state->share_mode_flags);
}
static NTSTATUS fsp_update_share_mode_flags(struct files_struct *fsp)
{
struct fsp_update_share_mode_flags_state state = {0};
int seqnum = dbwrap_get_seqnum(lock_db);
NTSTATUS status;
if (seqnum == fsp->share_mode_flags_seqnum) {
return NT_STATUS_OK;
}
status = share_mode_do_locked(
fsp->file_id, fsp_update_share_mode_flags_fn, &state);
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("share_mode_do_locked returned %s\n",
nt_errstr(status));
return status;
}
if (!NDR_ERR_CODE_IS_SUCCESS(state.ndr_err)) {
DBG_DEBUG("get_share_mode_blob_header returned %s\n",
ndr_errstr(state.ndr_err));
return ndr_map_error2ntstatus(state.ndr_err);
}
fsp->share_mode_flags_seqnum = seqnum;
fsp->share_mode_flags = state.share_mode_flags;
return NT_STATUS_OK;
}
bool file_has_read_lease(struct files_struct *fsp)
{
NTSTATUS status;
status = fsp_update_share_mode_flags(fsp);
if (!NT_STATUS_IS_OK(status)) {
/* Safe default for leases */
return true;
}
return (fsp->share_mode_flags & SHARE_MODE_LEASE_READ) != 0;
}
static int share_mode_data_nofree_destructor(struct share_mode_data *d)
{
return -1;
}
static struct share_mode_data *share_mode_memcache_fetch(TALLOC_CTX *mem_ctx,
const TDB_DATA id_key,
DATA_BLOB *blob)
{
enum ndr_err_code ndr_err;
struct share_mode_data *d;
uint64_t sequence_number;
uint16_t flags;
void *ptr;
struct file_id id;
struct file_id_buf idbuf;
DATA_BLOB key;
/* Ensure this is a locking_key record. */
if (id_key.dsize != sizeof(id)) {
return NULL;
}
memcpy(&id, id_key.dptr, id_key.dsize);
key = memcache_key(&id);
ptr = memcache_lookup_talloc(NULL,
SHARE_MODE_LOCK_CACHE,
key);
if (ptr == NULL) {
DBG_DEBUG("failed to find entry for key %s\n",
file_id_str_buf(id, &idbuf));
return NULL;
}
/* sequence number key is at start of blob. */
ndr_err = get_share_mode_blob_header(blob, &sequence_number, &flags);
if (ndr_err != NDR_ERR_SUCCESS) {
/* Bad blob. Remove entry. */
DBG_DEBUG("bad blob %u key %s\n",
(unsigned int)ndr_err,
file_id_str_buf(id, &idbuf));
memcache_delete(NULL,
SHARE_MODE_LOCK_CACHE,
key);
return NULL;
}
d = (struct share_mode_data *)ptr;
if (d->sequence_number != sequence_number) {
DBG_DEBUG("seq changed (cached %"PRIx64") (new %"PRIx64") "
"for key %s\n",
d->sequence_number,
sequence_number,
file_id_str_buf(id, &idbuf));
/* Cache out of date. Remove entry. */
memcache_delete(NULL,
SHARE_MODE_LOCK_CACHE,
key);
return NULL;
}
/* Move onto mem_ctx. */
d = talloc_move(mem_ctx, &ptr);
/*
* Now we own d, prevent the cache from freeing it
* when we delete the entry.
*/
talloc_set_destructor(d, share_mode_data_nofree_destructor);
/* Remove from the cache. We own it now. */
memcache_delete(NULL,
SHARE_MODE_LOCK_CACHE,
key);
/* And reset the destructor to none. */
talloc_set_destructor(d, NULL);
DBG_DEBUG("fetched entry for file %s seq %"PRIx64" key %s\n",
d->base_name,
d->sequence_number,
file_id_str_buf(id, &idbuf));
return d;
}
/*******************************************************************
Get all share mode entries for a dev/inode pair.
********************************************************************/
static struct share_mode_data *parse_share_modes(TALLOC_CTX *mem_ctx,
const TDB_DATA key,
const TDB_DATA dbuf)
{
struct share_mode_data *d;
enum ndr_err_code ndr_err;
DATA_BLOB blob;
blob.data = dbuf.dptr;
blob.length = dbuf.dsize;
/* See if we already have a cached copy of this key. */
d = share_mode_memcache_fetch(mem_ctx, key, &blob);
if (d != NULL) {
return d;
}
d = talloc(mem_ctx, struct share_mode_data);
if (d == NULL) {
DEBUG(0, ("talloc failed\n"));
goto fail;
}
ndr_err = ndr_pull_struct_blob_all(
&blob, d, d, (ndr_pull_flags_fn_t)ndr_pull_share_mode_data);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
DBG_WARNING("ndr_pull_share_mode_data failed: %s\n",
ndr_errstr(ndr_err));
goto fail;
}
if (DEBUGLEVEL >= 10) {
DEBUG(10, ("parse_share_modes:\n"));
NDR_PRINT_DEBUG(share_mode_data, d);
}
/*
* We have a non-zero locking.tdb record that was correctly
* parsed. This means a share_entries.tdb entry exists,
* otherwise we'd have paniced before in
* share_mode_data_store()
*/
d->have_share_modes = true;
return d;
fail:
TALLOC_FREE(d);
return NULL;
}
/*******************************************************************
If modified, store the share_mode_data back into the database.
********************************************************************/
static NTSTATUS share_mode_data_store(
struct share_mode_data *d, struct db_record *rec)
{
DATA_BLOB blob;
enum ndr_err_code ndr_err;
NTSTATUS status;
if (!d->modified) {
DBG_DEBUG("not modified\n");
return NT_STATUS_OK;
}
if (DEBUGLEVEL >= 10) {
DBG_DEBUG("\n");
NDR_PRINT_DEBUG(share_mode_data, d);
}
d->sequence_number += 1;
if (!d->have_share_modes) {
TDB_DATA key = dbwrap_record_get_key(rec);
bool share_entries_exist;
share_entries_exist = dbwrap_exists(share_entries_db, key);
SMB_ASSERT(!share_entries_exist);
TALLOC_FREE(d->delete_tokens);
d->num_delete_tokens = 0;
if (d->fresh) {
DBG_DEBUG("Ignoring fresh empty record\n");
return NT_STATUS_OK;
}
status = dbwrap_record_delete(rec);
return status;
}
ndr_err = ndr_push_struct_blob(
&blob, d, d, (ndr_push_flags_fn_t)ndr_push_share_mode_data);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
DBG_DEBUG("ndr_push_share_mode_data failed: %s\n",
ndr_errstr(ndr_err));
return ndr_map_error2ntstatus(ndr_err);
}
status = dbwrap_record_store(
rec,
(TDB_DATA) { .dptr = blob.data, .dsize = blob.length },
TDB_REPLACE);
TALLOC_FREE(blob.data);
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("dbwrap_record_store failed: %s\n",
nt_errstr(status));
}
return status;
}
/*******************************************************************
Allocate a new share_mode_data struct, mark it unmodified.
fresh is set to note that currently there is no database entry.
********************************************************************/
static struct share_mode_data *fresh_share_mode_lock(
TALLOC_CTX *mem_ctx, const char *servicepath,
const struct smb_filename *smb_fname,
const struct timespec *old_write_time)
{
struct share_mode_data *d;
if ((servicepath == NULL) || (smb_fname == NULL) ||
(old_write_time == NULL)) {
return NULL;
}
d = talloc_zero(mem_ctx, struct share_mode_data);
if (d == NULL) {
goto fail;
}
/* New record - new sequence number. */
generate_random_buffer((uint8_t *)&d->sequence_number, 8);
d->base_name = talloc_strdup(d, smb_fname->base_name);
if (d->base_name == NULL) {
goto fail;
}
if (smb_fname->stream_name != NULL) {
d->stream_name = talloc_strdup(d, smb_fname->stream_name);
if (d->stream_name == NULL) {
goto fail;
}
}
d->servicepath = talloc_strdup(d, servicepath);
if (d->servicepath == NULL) {
goto fail;
}
d->old_write_time = full_timespec_to_nt_time(old_write_time);
d->flags = SHARE_MODE_SHARE_DELETE |
SHARE_MODE_SHARE_WRITE |
SHARE_MODE_SHARE_READ;
d->modified = false;
d->fresh = true;
return d;
fail:
DEBUG(0, ("talloc failed\n"));
TALLOC_FREE(d);
return NULL;
}
/*
* We can only ever have one share mode locked. Use a static
* share_mode_data pointer that is shared by multiple nested
* share_mode_lock structures, explicitly refcounted.
*/
static struct share_mode_data *static_share_mode_data = NULL;
static size_t static_share_mode_data_refcount = 0;
/*
* db_record for the above. With dbwrap_do_locked we can get a
* db_record on the stack, which we can't TALLOC_FREE but which we
* need to share with a nested get_share_mode_lock call.
*/
static struct db_record *static_share_mode_record = NULL;
static TDB_DATA static_share_mode_record_value = {0};
static bool static_share_mode_record_talloced = false;
/*******************************************************************
Either fetch a share mode from the database, or allocate a fresh
one if the record doesn't exist.
********************************************************************/
static NTSTATUS get_static_share_mode_data(
struct db_record *rec,
struct file_id id,
const char *servicepath,
const struct smb_filename *smb_fname,
const struct timespec *old_write_time)
{
struct share_mode_data *d;
TDB_DATA value = dbwrap_record_get_value(rec);
SMB_ASSERT(static_share_mode_data == NULL);
if (value.dptr == NULL) {
d = fresh_share_mode_lock(
lock_db, servicepath, smb_fname, old_write_time);
if (d == NULL) {
if (smb_fname == NULL) {
return NT_STATUS_NOT_FOUND;
}
return NT_STATUS_NO_MEMORY;
}
} else {
TDB_DATA key = locking_key(&id);
d = parse_share_modes(lock_db, key, value);
if (d == NULL) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
}
d->id = id;
static_share_mode_data = d;
return NT_STATUS_OK;
}
/*******************************************************************
Get a share_mode_lock, Reference counted to allow nested calls.
********************************************************************/
static int share_mode_lock_destructor(struct share_mode_lock *lck);
struct share_mode_lock *get_share_mode_lock(
TALLOC_CTX *mem_ctx,
struct file_id id,
const char *servicepath,
const struct smb_filename *smb_fname,
const struct timespec *old_write_time)
{
TDB_DATA key = locking_key(&id);
struct share_mode_lock *lck = NULL;
NTSTATUS status;
lck = talloc(mem_ctx, struct share_mode_lock);
if (lck == NULL) {
DEBUG(1, ("talloc failed\n"));
return NULL;
}
if (static_share_mode_data != NULL) {
if (!file_id_equal(&static_share_mode_data->id, &id)) {
DEBUG(1, ("Can not lock two share modes "
"simultaneously\n"));
goto fail;
}
goto done;
}
SMB_ASSERT(static_share_mode_data_refcount == 0);
if (static_share_mode_record == NULL) {
static_share_mode_record = dbwrap_fetch_locked(
lock_db, lock_db, key);
if (static_share_mode_record == NULL) {
DEBUG(3, ("Could not lock share entry\n"));
goto fail;
}
static_share_mode_record_talloced = true;
static_share_mode_record_value = dbwrap_record_get_value(
static_share_mode_record);
} else {
TDB_DATA static_key;
int cmp;
static_key = dbwrap_record_get_key(static_share_mode_record);
cmp = tdb_data_cmp(static_key, key);
if (cmp != 0) {
DBG_WARNING("Can not lock two share modes "
"simultaneously\n");
return NULL;
}
}
status = get_static_share_mode_data(static_share_mode_record,
id,
servicepath,
smb_fname,
old_write_time);
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("get_static_share_mode_data failed: %s\n",
nt_errstr(status));
TALLOC_FREE(static_share_mode_record);
goto fail;
}
done:
static_share_mode_data_refcount += 1;
lck->data = static_share_mode_data;
talloc_set_destructor(lck, share_mode_lock_destructor);
return lck;
fail:
TALLOC_FREE(lck);
return NULL;
}
static int share_mode_lock_destructor(struct share_mode_lock *lck)
{
NTSTATUS status;
SMB_ASSERT(static_share_mode_data_refcount > 0);
static_share_mode_data_refcount -= 1;
if (static_share_mode_data_refcount > 0) {
return 0;
}
status = share_mode_data_store(
static_share_mode_data, static_share_mode_record);
if (!NT_STATUS_IS_OK(status)) {
DBG_ERR("share_mode_data_store failed: %s\n",
nt_errstr(status));
smb_panic("Could not store share mode data\n");
}
if (static_share_mode_record_talloced) {
TALLOC_FREE(static_share_mode_record);
}
if (static_share_mode_data->have_share_modes) {
/*
* This is worth keeping. Without share modes,
* share_mode_data_store above has left nothing in the
* database.
*/
share_mode_memcache_store(static_share_mode_data);
static_share_mode_data = NULL;
} else {
/*
* The next opener of this file will find an empty
* locking.tdb record. Don't store the share_mode_data
* in the memcache, fresh_share_mode_lock() will
* generate a fresh seqnum anyway, obsoleting the
* cache entry.
*/
TALLOC_FREE(static_share_mode_data);
}
return 0;
}
struct share_mode_do_locked_state {
void (*fn)(TDB_DATA value,
bool *modified_dependent,
void *private_data);
void *private_data;
};
static void share_mode_do_locked_fn(
struct db_record *rec,
TDB_DATA value,
void *private_data)
{
struct share_mode_do_locked_state *state = private_data;
bool modified_dependent = false;
bool reset_static_share_mode_record = false;
if (static_share_mode_record == NULL) {
static_share_mode_record = rec;
static_share_mode_record_value = value;
static_share_mode_record_talloced = false;
reset_static_share_mode_record = true;
} else {
SMB_ASSERT(static_share_mode_record == rec);
}
state->fn(value, &modified_dependent, state->private_data);
if (modified_dependent) {
dbwrap_watched_wakeup(rec);
}
if (reset_static_share_mode_record) {
static_share_mode_record = NULL;
}
}
NTSTATUS share_mode_do_locked(
struct file_id id,
void (*fn)(TDB_DATA value,
bool *modified_dependent,
void *private_data),
void *private_data)
{
TDB_DATA key = locking_key(&id);
size_t refcount = static_share_mode_data_refcount;
if (static_share_mode_record != NULL) {
bool modified_dependent = false;
TDB_DATA static_key;
int cmp;
static_key = dbwrap_record_get_key(static_share_mode_record);
cmp = tdb_data_cmp(static_key, key);
if (cmp != 0) {
DBG_WARNING("Can not lock two share modes "
"simultaneously\n");
return NT_STATUS_INVALID_LOCK_SEQUENCE;
}
fn(static_share_mode_record_value,
&modified_dependent,
private_data);
if (modified_dependent) {
dbwrap_watched_wakeup(static_share_mode_record);
}
} else {
struct share_mode_do_locked_state state = {
.fn = fn, .private_data = private_data,
};
NTSTATUS status;
status = dbwrap_do_locked(
lock_db, key, share_mode_do_locked_fn, &state);
if (!NT_STATUS_IS_OK(status)) {
DBG_WARNING("dbwrap_do_locked failed: %s\n",
nt_errstr(status));
return status;
}
}
SMB_ASSERT(refcount == static_share_mode_data_refcount);
return NT_STATUS_OK;
}
static void share_mode_wakeup_waiters_fn(TDB_DATA value,
bool *modified_dependent,
void *private_data)
{
*modified_dependent = true;
}
NTSTATUS share_mode_wakeup_waiters(struct file_id id)
{
return share_mode_do_locked(id, share_mode_wakeup_waiters_fn, NULL);
}
bool share_mode_have_entries(struct share_mode_lock *lck)
{
return lck->data->have_share_modes;
}
struct share_mode_watch_state {
struct tevent_context *ev;
bool blockerdead;
struct server_id blocker;
struct tevent_req *subreq;
};
static void share_mode_watch_fn(
TDB_DATA value, bool *modified_dependent, void *private_data)
{
struct share_mode_watch_state *state = talloc_get_type_abort(
private_data, struct share_mode_watch_state);
state->subreq = dbwrap_watched_watch_send(
state, state->ev, static_share_mode_record, state->blocker);
}
static void share_mode_watch_done(struct tevent_req *subreq);
struct tevent_req *share_mode_watch_send(
TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct file_id id,
struct server_id blocker)
{
struct tevent_req *req = NULL;
struct share_mode_watch_state *state = NULL;
NTSTATUS status;
req = tevent_req_create(
mem_ctx, &state, struct share_mode_watch_state);
if (req == NULL) {
return NULL;
}
state->ev = ev;
state->blocker = blocker;
status = share_mode_do_locked(id, share_mode_watch_fn, state);
if (tevent_req_nterror(req, status)) {
return tevent_req_post(req, ev);
}
if (tevent_req_nomem(state->subreq, req)) {
return tevent_req_post(req, ev);
}
tevent_req_set_callback(state->subreq, share_mode_watch_done, req);
return req;
}
static void share_mode_watch_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct share_mode_watch_state *state = tevent_req_data(
req, struct share_mode_watch_state);
NTSTATUS status;
status = dbwrap_watched_watch_recv(
subreq, &state->blockerdead, &state->blocker);
if (tevent_req_nterror(req, status)) {
return;
}
tevent_req_done(req);
}
NTSTATUS share_mode_watch_recv(
struct tevent_req *req, bool *blockerdead, struct server_id *blocker)
{
struct share_mode_watch_state *state = tevent_req_data(
req, struct share_mode_watch_state);
NTSTATUS status;
if (tevent_req_is_nterror(req, &status)) {
return status;
}
if (blockerdead != NULL) {
*blockerdead = state->blockerdead;
}
if (blocker != NULL) {
*blocker = state->blocker;
}
return NT_STATUS_OK;
}
struct fetch_share_mode_unlocked_state {
TALLOC_CTX *mem_ctx;
struct share_mode_lock *lck;
};
static void fetch_share_mode_unlocked_parser(
TDB_DATA key, TDB_DATA data, void *private_data)
{
struct fetch_share_mode_unlocked_state *state = private_data;
if (data.dsize == 0) {
/* Likely a ctdb tombstone record, ignore it */
return;
}
state->lck = talloc(state->mem_ctx, struct share_mode_lock);
if (state->lck == NULL) {
DEBUG(0, ("talloc failed\n"));
return;
}
state->lck->data = parse_share_modes(state->lck, key, data);
}
/*******************************************************************
Get a share_mode_lock without locking the database or reference
counting. Used by smbstatus to display existing share modes.
********************************************************************/
struct share_mode_lock *fetch_share_mode_unlocked(TALLOC_CTX *mem_ctx,
struct file_id id)
{
struct fetch_share_mode_unlocked_state state = { .mem_ctx = mem_ctx };
TDB_DATA key = locking_key(&id);
NTSTATUS status;
status = dbwrap_parse_record(
lock_db, key, fetch_share_mode_unlocked_parser, &state);
if (!NT_STATUS_IS_OK(status)) {
return NULL;
}
return state.lck;
}
static void fetch_share_mode_done(struct tevent_req *subreq);
struct fetch_share_mode_state {
struct file_id id;
TDB_DATA key;
struct fetch_share_mode_unlocked_state parser_state;
enum dbwrap_req_state req_state;
};
/**
* @brief Get a share_mode_lock without locking or refcounting
*
* This can be used in a clustered Samba environment where the async dbwrap
* request is sent over a socket to the local ctdbd. If the send queue is full
* and the caller was issuing multiple async dbwrap requests in a loop, the
* caller knows it's probably time to stop sending requests for now and try
* again later.
*
* @param[in] mem_ctx The talloc memory context to use.
*
* @param[in] ev The event context to work on.
*
* @param[in] id The file id for the locking.tdb key
*
* @param[out] queued This boolean out parameter tells the caller whether the
* async request is blocked in a full send queue:
*
* false := request is dispatched
*
* true := send queue is full, request waiting to be
* dispatched
*
* @return The new async request, NULL on error.
**/
struct tevent_req *fetch_share_mode_send(TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct file_id id,
bool *queued)
{
struct tevent_req *req = NULL;
struct fetch_share_mode_state *state = NULL;
struct tevent_req *subreq = NULL;
*queued = false;
req = tevent_req_create(mem_ctx, &state,
struct fetch_share_mode_state);
if (req == NULL) {
return NULL;
}
state->id = id;
state->key = locking_key(&state->id);
state->parser_state.mem_ctx = state;
subreq = dbwrap_parse_record_send(state,
ev,
lock_db,
state->key,
fetch_share_mode_unlocked_parser,
&state->parser_state,
&state->req_state);
if (tevent_req_nomem(subreq, req)) {
return tevent_req_post(req, ev);
}
tevent_req_set_callback(subreq, fetch_share_mode_done, req);
if (state->req_state < DBWRAP_REQ_DISPATCHED) {
*queued = true;
}
return req;
}
static void fetch_share_mode_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
NTSTATUS status;
status = dbwrap_parse_record_recv(subreq);
TALLOC_FREE(subreq);
if (tevent_req_nterror(req, status)) {
return;
}
tevent_req_done(req);
return;
}
NTSTATUS fetch_share_mode_recv(struct tevent_req *req,
TALLOC_CTX *mem_ctx,
struct share_mode_lock **_lck)
{
struct fetch_share_mode_state *state = tevent_req_data(
req, struct fetch_share_mode_state);
struct share_mode_lock *lck = NULL;
NTSTATUS status;
if (tevent_req_is_nterror(req, &status)) {
tevent_req_received(req);
return status;
}
if (state->parser_state.lck->data == NULL) {
tevent_req_received(req);
return NT_STATUS_NOT_FOUND;
}
lck = talloc_move(mem_ctx, &state->parser_state.lck);
if (DEBUGLEVEL >= 10) {
DBG_DEBUG("share_mode_data:\n");
NDR_PRINT_DEBUG(share_mode_data, lck->data);
}
*_lck = lck;
tevent_req_received(req);
return NT_STATUS_OK;
}
struct share_mode_forall_state {
int (*fn)(struct file_id fid, const struct share_mode_data *data,
void *private_data);
void *private_data;
};
static int share_mode_traverse_fn(struct db_record *rec, void *_state)
{
struct share_mode_forall_state *state =
(struct share_mode_forall_state *)_state;
TDB_DATA key;
TDB_DATA value;
DATA_BLOB blob;
enum ndr_err_code ndr_err;
struct share_mode_data *d;
struct file_id fid;
int ret;
key = dbwrap_record_get_key(rec);
value = dbwrap_record_get_value(rec);
/* Ensure this is a locking_key record. */
if (key.dsize != sizeof(fid)) {
return 0;
}
memcpy(&fid, key.dptr, sizeof(fid));
d = talloc(talloc_tos(), struct share_mode_data);
if (d == NULL) {
return 0;
}
blob.data = value.dptr;
blob.length = value.dsize;
ndr_err = ndr_pull_struct_blob_all(
&blob, d, d, (ndr_pull_flags_fn_t)ndr_pull_share_mode_data);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
DEBUG(1, ("ndr_pull_share_mode_lock failed\n"));
return 0;
}
if (DEBUGLEVEL > 10) {
DEBUG(11, ("parse_share_modes:\n"));
NDR_PRINT_DEBUG(share_mode_data, d);
}
ret = state->fn(fid, d, state->private_data);
TALLOC_FREE(d);
return ret;
}
int share_mode_forall(int (*fn)(struct file_id fid,
const struct share_mode_data *data,
void *private_data),
void *private_data)
{
struct share_mode_forall_state state = {
.fn = fn,
.private_data = private_data
};
NTSTATUS status;
int count;
if (lock_db == NULL) {
return 0;
}
status = dbwrap_traverse_read(lock_db, share_mode_traverse_fn,
&state, &count);
if (!NT_STATUS_IS_OK(status)) {
return -1;
}
return count;
}
struct share_entry_forall_state {
struct file_id fid;
const struct share_mode_data *data;
int (*fn)(struct file_id fid,
const struct share_mode_data *data,
const struct share_mode_entry *entry,
void *private_data);
void *private_data;
int ret;
};
static bool share_entry_traverse_walker(
struct share_mode_entry *e,
bool *modified,
void *private_data)
{
struct share_entry_forall_state *state = private_data;
state->ret = state->fn(
state->fid, state->data, e, state->private_data);
return (state->ret != 0);
}
static int share_entry_traverse_fn(struct file_id fid,
const struct share_mode_data *data,
void *private_data)
{
struct share_entry_forall_state *state = private_data;
struct share_mode_lock lck = {
.data = discard_const_p(struct share_mode_data, data)
};
bool ok;
state->fid = fid;
state->data = data;
ok = share_mode_forall_entries(
&lck, share_entry_traverse_walker, state);
if (!ok) {
DBG_DEBUG("share_mode_forall_entries failed\n");
return false;
}
return state->ret;
}
/*******************************************************************
Call the specified function on each entry under management by the
share mode system.
********************************************************************/
int share_entry_forall(int (*fn)(struct file_id fid,
const struct share_mode_data *data,
const struct share_mode_entry *entry,
void *private_data),
void *private_data)
{
struct share_entry_forall_state state = {
.fn = fn, .private_data = private_data };
return share_mode_forall(share_entry_traverse_fn, &state);
}
struct cleanup_disconnected_state {
struct share_mode_lock *lck;
uint64_t open_persistent_id;
size_t num_disconnected;
bool found_connected;
};
static bool cleanup_disconnected_lease(struct share_mode_entry *e,
void *private_data)
{
struct cleanup_disconnected_state *state = private_data;
NTSTATUS status;
status = leases_db_del(
&e->client_guid, &e->lease_key, &state->lck->data->id);
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("leases_db_del failed: %s\n",
nt_errstr(status));
}
return false;
}
static bool share_mode_find_connected_fn(
struct share_mode_entry *e,
bool *modified,
void *private_data)
{
struct cleanup_disconnected_state *state = private_data;
struct share_mode_data *d = state->lck->data;
bool disconnected;
disconnected = server_id_is_disconnected(&e->pid);
if (!disconnected) {
struct file_id_buf tmp1;
struct server_id_buf tmp2;
DBG_INFO("file (file-id='%s', servicepath='%s', "
"base_name='%s%s%s') "
"is used by server %s ==> do not cleanup\n",
file_id_str_buf(d->id, &tmp1),
d->servicepath,
d->base_name,
(d->stream_name == NULL)
? "" : "', stream_name='",
(d->stream_name == NULL)
? "" : d->stream_name,
server_id_str_buf(e->pid, &tmp2));
state->found_connected = true;
return true;
}
if (state->open_persistent_id != e->share_file_id) {
struct file_id_buf tmp;
DBG_INFO("entry for file "
"(file-id='%s', servicepath='%s', "
"base_name='%s%s%s') "
"has share_file_id %"PRIu64" but expected "
"%"PRIu64"==> do not cleanup\n",
file_id_str_buf(d->id, &tmp),
d->servicepath,
d->base_name,
(d->stream_name == NULL)
? "" : "', stream_name='",
(d->stream_name == NULL)
? "" : d->stream_name,
e->share_file_id,
state->open_persistent_id);
state->found_connected = true;
return true;
}
state->num_disconnected += 1;
return false;
}
bool share_mode_cleanup_disconnected(struct file_id fid,
uint64_t open_persistent_id)
{
struct cleanup_disconnected_state state = {
.open_persistent_id = open_persistent_id
};
struct share_mode_data *data;
bool ret = false;
TALLOC_CTX *frame = talloc_stackframe();
struct file_id_buf idbuf;
NTSTATUS status;
bool ok;
state.lck = get_existing_share_mode_lock(frame, fid);
if (state.lck == NULL) {
DBG_INFO("Could not fetch share mode entry for %s\n",
file_id_str_buf(fid, &idbuf));
goto done;
}
data = state.lck->data;
ok = share_mode_forall_entries(
state.lck, share_mode_find_connected_fn, &state);
if (!ok) {
DBG_DEBUG("share_mode_forall_entries failed\n");
goto done;
}
if (state.found_connected) {
DBG_DEBUG("Found connected entry\n");
goto done;
}
ok = share_mode_forall_leases(
state.lck, cleanup_disconnected_lease, &state);
if (!ok) {
DBG_DEBUG("failed to clean up leases associated "
"with file (file-id='%s', servicepath='%s', "
"base_name='%s%s%s') and open_persistent_id %"PRIu64" "
"==> do not cleanup\n",
file_id_str_buf(fid, &idbuf),
data->servicepath,
data->base_name,
(data->stream_name == NULL)
? "" : "', stream_name='",
(data->stream_name == NULL)
? "" : data->stream_name,
open_persistent_id);
}
ok = brl_cleanup_disconnected(fid, open_persistent_id);
if (!ok) {
DBG_DEBUG("failed to clean up byte range locks associated "
"with file (file-id='%s', servicepath='%s', "
"base_name='%s%s%s') and open_persistent_id %"PRIu64" "
"==> do not cleanup\n",
file_id_str_buf(fid, &idbuf),
data->servicepath,
data->base_name,
(data->stream_name == NULL)
? "" : "', stream_name='",
(data->stream_name == NULL)
? "" : data->stream_name,
open_persistent_id);
goto done;
}
DBG_DEBUG("cleaning up %zu entries for file "
"(file-id='%s', servicepath='%s', "
"base_name='%s%s%s') "
"from open_persistent_id %"PRIu64"\n",
state.num_disconnected,
file_id_str_buf(fid, &idbuf),
data->servicepath,
data->base_name,
(data->stream_name == NULL)
? "" : "', stream_name='",
(data->stream_name == NULL)
? "" : data->stream_name,
open_persistent_id);
/*
* No connected share entries left, wipe them all
*/
status = dbwrap_delete(share_entries_db, locking_key(&fid));
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("dbwrap_delete failed: %s\n",
nt_errstr(status));
goto done;
}
data->have_share_modes = false;
data->modified = true;
ret = true;
done:
talloc_free(frame);
return ret;
}
static int share_mode_entry_cmp(
struct server_id pid1,
uint64_t share_file_id1,
struct server_id pid2,
uint64_t share_file_id2)
{
int cmp;
cmp = server_id_cmp(&pid1, &pid2);
if (cmp != 0) {
return cmp;
}
if (share_file_id1 != share_file_id2) {
return (share_file_id1 < share_file_id2) ? -1 : 1;
}
return 0;
}
/*
* 132 is the sizeof an ndr-encoded struct share_mode_entry_buf.
* Reading/writing entries will immediately error out if this
* size differs (push/pull is done without allocs).
*/
struct share_mode_entry_buf {
uint8_t buf[132];
};
#define SHARE_MODE_ENTRY_SIZE (sizeof(struct share_mode_entry_buf))
static bool share_mode_entry_put(
const struct share_mode_entry *e,
struct share_mode_entry_buf *dst)
{
DATA_BLOB blob = { .data = dst->buf, .length = sizeof(dst->buf) };
enum ndr_err_code ndr_err;
if (DEBUGLEVEL>=10) {
DBG_DEBUG("share_mode_entry:\n");
NDR_PRINT_DEBUG(share_mode_entry, discard_const_p(void, e));
}
ndr_err = ndr_push_struct_into_fixed_blob(
&blob,
e,
(ndr_push_flags_fn_t)ndr_push_share_mode_entry);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
DBG_WARNING("ndr_push_share_mode_entry failed: %s\n",
ndr_errstr(ndr_err));
return false;
}
return true;
}
static bool share_mode_entry_get(
DATA_BLOB blob, struct share_mode_entry *e)
{
enum ndr_err_code ndr_err = NDR_ERR_SUCCESS;
ndr_err = ndr_pull_struct_blob_all_noalloc(
&blob, e, (ndr_pull_flags_fn_t)ndr_pull_share_mode_entry);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
DBG_WARNING("ndr_pull_share_mode_entry failed\n");
return false;
}
return true;
}
static size_t share_mode_entry_find(
uint8_t *data,
size_t num_share_modes,
struct server_id pid,
uint64_t share_file_id,
struct share_mode_entry *e,
bool *match)
{
ssize_t left, right, middle;
if (num_share_modes == 0) {
*match = false;
return 0;
}
left = 0;
right = (num_share_modes-1);
while (left <= right) {
DATA_BLOB blob;
int cmp;
bool ok;
middle = left + ((right - left) / 2);
DBG_DEBUG("left=%zu, right=%zu, middle=%zu\n",
left,
right,
middle);
blob = (DATA_BLOB) {
.data = data + middle * SHARE_MODE_ENTRY_SIZE,
.length = SHARE_MODE_ENTRY_SIZE,
};
DBG_DEBUG("blob.data=%p, blob.length=%zu\n",
blob.data,
blob.length);
ok = share_mode_entry_get(blob, e);
if (!ok) {
DBG_DEBUG("share_mode_entry_get failed\n");
return false;
}
cmp = share_mode_entry_cmp(
e->pid, e->share_file_id, pid, share_file_id);
if (cmp == 0) {
*match = true;
return middle;
}
if (cmp < 0) {
right = middle-1;
} else {
left = middle+1;
}
}
*match = false;
return left;
}
struct set_share_mode_state {
struct share_mode_entry e;
bool created_share_mode_record;
NTSTATUS status;
};
static void set_share_mode_fn(
struct db_record *rec,
TDB_DATA data,
void *private_data)
{
struct set_share_mode_state *state = private_data;
size_t idx, num_share_modes;
struct share_mode_entry tmp;
struct share_mode_entry_buf buf;
bool ok, found;
TDB_DATA dbufs[3];
size_t num_dbufs = 0;
if ((data.dsize % SHARE_MODE_ENTRY_SIZE) != 0) {
DBG_WARNING("Got invalid record size %zu\n", data.dsize);
state->status = NT_STATUS_INTERNAL_DB_CORRUPTION;
return;
}
num_share_modes = data.dsize / SHARE_MODE_ENTRY_SIZE;
ok = share_mode_entry_put(&state->e, &buf);
if (!ok) {
DBG_DEBUG("share_mode_entry_put failed\n");
state->status = NT_STATUS_INTERNAL_ERROR;
return;
}
DBG_DEBUG("num_share_modes=%zu\n", num_share_modes);
idx = share_mode_entry_find(
data.dptr,
num_share_modes,
state->e.pid,
state->e.share_file_id,
&tmp,
&found);
if (found) {
DBG_WARNING("Found duplicate share mode\n");
state->status = NT_STATUS_INTERNAL_DB_CORRUPTION;
return;
}
DBG_DEBUG("idx=%zu, found=%d\n", idx, (int)found);
if (idx > 0) {
dbufs[num_dbufs] = (TDB_DATA) {
.dptr = data.dptr,
.dsize = idx * SHARE_MODE_ENTRY_SIZE,
};
num_dbufs += 1;
}
dbufs[num_dbufs] = (TDB_DATA) {
.dptr = buf.buf, .dsize = SHARE_MODE_ENTRY_SIZE,
};
num_dbufs += 1;
if (idx < num_share_modes) {
dbufs[num_dbufs] = (TDB_DATA) {
.dptr = data.dptr + idx * SHARE_MODE_ENTRY_SIZE,
.dsize = (num_share_modes-idx) * SHARE_MODE_ENTRY_SIZE,
};
num_dbufs += 1;
}
{
size_t i;
for (i=0; icreated_share_mode_record = (num_share_modes == 0);
state->status = dbwrap_record_storev(rec, dbufs, num_dbufs, 0);
}
bool set_share_mode(struct share_mode_lock *lck,
struct files_struct *fsp,
uid_t uid,
uint64_t mid,
uint16_t op_type,
uint32_t share_access,
uint32_t access_mask)
{
struct share_mode_data *d = lck->data;
struct set_share_mode_state state = {
.status = NT_STATUS_OK,
.e.pid = messaging_server_id(fsp->conn->sconn->msg_ctx),
.e.share_access = share_access,
.e.private_options = fsp->fh->private_options,
.e.access_mask = access_mask,
.e.op_mid = mid,
.e.op_type = op_type,
.e.time.tv_sec = fsp->open_time.tv_sec,
.e.time.tv_usec = fsp->open_time.tv_usec,
.e.share_file_id = fsp->fh->gen_id,
.e.uid = (uint32_t)uid,
.e.flags = (fsp->posix_flags & FSP_POSIX_FLAGS_OPEN) ?
SHARE_MODE_FLAG_POSIX_OPEN : 0,
.e.name_hash = fsp->name_hash,
};
NTSTATUS status;
if (op_type == LEASE_OPLOCK) {
const struct GUID *client_guid = fsp_client_guid(fsp);
state.e.client_guid = *client_guid;
state.e.lease_key = fsp->lease->lease.lease_key;
}
status = dbwrap_do_locked(
share_entries_db,
locking_key(&d->id),
set_share_mode_fn,
&state);
if (!NT_STATUS_IS_OK(status)) {
DBG_WARNING("dbwrap_do_locked failed: %s\n",
nt_errstr(status));
return false;
}
if (!NT_STATUS_IS_OK(state.status)) {
DBG_WARNING("set_share_mode_fn failed: %s\n",
nt_errstr(state.status));
return false;
}
if (state.created_share_mode_record) {
d->have_share_modes = true;
d->modified = true;
}
return true;
}
struct share_mode_forall_entries_state {
struct share_mode_lock *lck;
bool (*fn)(struct share_mode_entry *e,
bool *modified,
void *private_data);
void *private_data;
bool ok;
};
static bool share_mode_for_one_entry(
struct share_mode_forall_entries_state *state,
size_t *i,
size_t *num_share_modes,
TDB_DATA data,
bool *writeback)
{
DATA_BLOB blob = {
.data = data.dptr + (*i) * SHARE_MODE_ENTRY_SIZE,
.length = SHARE_MODE_ENTRY_SIZE,
};
struct share_mode_entry e = {.pid.pid=0};
enum ndr_err_code ndr_err = NDR_ERR_SUCCESS;
bool modified = false;
bool stop = false;
struct server_id e_pid;
uint64_t e_share_file_id;
ndr_err = ndr_pull_struct_blob_all_noalloc(
&blob,
&e,
(ndr_pull_flags_fn_t)ndr_pull_share_mode_entry);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
DBG_WARNING("ndr_pull_share_mode_entry failed\n");
*i += 1;
return false;
}
if (DEBUGLEVEL >= 10) {
DBG_DEBUG("entry[%zu]:\n", *i);
NDR_PRINT_DEBUG(share_mode_entry, &e);
}
e_pid = e.pid;
e_share_file_id = e.share_file_id;
stop = state->fn(&e, &modified, state->private_data);
DBG_DEBUG("entry[%zu]: modified=%d, e.stale=%d\n",
*i,
(int)modified,
(int)e.stale);
if (modified) {
if (DEBUGLEVEL>=10) {
DBG_DEBUG("share_mode_entry:\n");
NDR_PRINT_DEBUG(share_mode_entry, &e);
}
/*
* Make sure sorting order is kept intact
*/
SMB_ASSERT(server_id_equal(&e_pid, &e.pid));
SMB_ASSERT(e_share_file_id == e.share_file_id);
ndr_err = ndr_push_struct_into_fixed_blob(
&blob,
&e,
(ndr_push_flags_fn_t)
ndr_push_share_mode_entry);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
DBG_WARNING("ndr_push_share_mode_entry "
"failed: %s\n",
ndr_errstr(ndr_err));
/*
* Not much we can do, just ignore it
*/
}
*i += 1;
*writeback = true;
return stop;
}
if (e.stale) {
if (*i < *num_share_modes) {
memmove(blob.data,
blob.data + SHARE_MODE_ENTRY_SIZE,
(*num_share_modes - *i - 1) *
SHARE_MODE_ENTRY_SIZE);
}
*num_share_modes -= 1;
*writeback = true;
return stop;
}
if (stop) {
return true;
}
*i += 1;
return false;
}
static void share_mode_forall_entries_fn(
struct db_record *rec,
TDB_DATA data,
void *private_data)
{
struct share_mode_forall_entries_state *state = private_data;
struct share_mode_data *d = state->lck->data;
size_t num_share_modes;
bool writeback = false;
NTSTATUS status;
bool stop = false;
size_t i;
if ((data.dsize % SHARE_MODE_ENTRY_SIZE) != 0) {
DBG_WARNING("Invalid data size %zu\n", data.dsize);
return;
}
num_share_modes = data.dsize / SHARE_MODE_ENTRY_SIZE;
DBG_DEBUG("num_share_modes=%zu\n", num_share_modes);
i = 0;
while (iok = true;
return;
}
if (num_share_modes == 0) {
if (data.dsize != 0) {
d->have_share_modes = false;
d->modified = true;
}
status = dbwrap_record_delete(rec);
} else {
TDB_DATA value = {
.dptr = data.dptr,
.dsize = num_share_modes * SHARE_MODE_ENTRY_SIZE,
};
status = dbwrap_record_store(rec, value, 0);
}
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("Storing record with %zu entries failed: %s\n",
num_share_modes,
nt_errstr(status));
return;
}
state->ok = true;
}
bool share_mode_forall_entries(
struct share_mode_lock *lck,
bool (*fn)(struct share_mode_entry *e,
bool *modified,
void *private_data),
void *private_data)
{
struct share_mode_forall_entries_state state = {
.lck = lck,
.fn = fn,
.private_data = private_data,
};
NTSTATUS status;
status = dbwrap_do_locked(
share_entries_db,
locking_key(&lck->data->id),
share_mode_forall_entries_fn,
&state);
if (NT_STATUS_EQUAL(status, NT_STATUS_NOT_FOUND)) {
status = NT_STATUS_OK;
state.ok = true;
}
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("dbwrap_do_locked failed: %s\n",
nt_errstr(status));
return false;
}
return state.ok;
}
struct share_mode_count_entries_state {
size_t num_share_modes;
NTSTATUS status;
};
static void share_mode_count_entries_fn(
TDB_DATA key, TDB_DATA data, void *private_data)
{
struct share_mode_count_entries_state *state = private_data;
if ((data.dsize % SHARE_MODE_ENTRY_SIZE) != 0) {
DBG_WARNING("Invalid data size %zu\n", data.dsize);
state->status = NT_STATUS_INTERNAL_DB_CORRUPTION;
return;
}
state->num_share_modes = data.dsize / SHARE_MODE_ENTRY_SIZE;
state->status = NT_STATUS_OK;
}
NTSTATUS share_mode_count_entries(struct file_id fid, size_t *num_share_modes)
{
struct share_mode_count_entries_state state = {
.status = NT_STATUS_NOT_FOUND,
};
NTSTATUS status;
status = dbwrap_parse_record(
share_entries_db,
locking_key(&fid),
share_mode_count_entries_fn,
&state);
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("dbwrap_parse_record failed: %s\n",
nt_errstr(status));
return status;
}
if (!NT_STATUS_IS_OK(state.status)) {
DBG_DEBUG("share_mode_forall_entries_fn failed: %s\n",
nt_errstr(state.status));
return state.status;
}
*num_share_modes = state.num_share_modes;
return NT_STATUS_OK;
}
struct share_mode_entry_do_state {
struct server_id pid;
uint64_t share_file_id;
void (*fn)(struct share_mode_entry *e,
size_t num_share_modes,
bool *modified,
void *private_data);
void *private_data;
size_t num_share_modes;
NTSTATUS status;
};
static void share_mode_entry_do_fn(
struct db_record *rec,
TDB_DATA data,
void *private_data)
{
struct share_mode_entry_do_state *state = private_data;
size_t idx;
bool found = false;
bool modified;
struct share_mode_entry e;
struct share_mode_entry_buf buf;
TDB_DATA dbufs[3];
size_t num_dbufs = 0;
if ((data.dsize % SHARE_MODE_ENTRY_SIZE) != 0) {
DBG_WARNING("Invalid data size %zu\n", data.dsize);
state->status = NT_STATUS_INTERNAL_DB_CORRUPTION;
return;
}
state->num_share_modes = data.dsize / SHARE_MODE_ENTRY_SIZE;
DBG_DEBUG("state->num_share_modes=%zu\n", state->num_share_modes);
idx = share_mode_entry_find(
data.dptr,
state->num_share_modes,
state->pid,
state->share_file_id,
&e,
&found);
if (!found) {
DBG_WARNING("Did not find share mode entry for %"PRIu64"\n",
state->share_file_id);
state->status = NT_STATUS_NOT_FOUND;
return;
}
state->fn(&e, state->num_share_modes, &modified, state->private_data);
if (!e.stale && !modified) {
state->status = NT_STATUS_OK;
return;
}
if (idx > 0) {
dbufs[num_dbufs] = (TDB_DATA) {
.dptr = data.dptr,
.dsize = idx * SHARE_MODE_ENTRY_SIZE,
};
num_dbufs += 1;
}
if (!e.stale) {
bool ok = share_mode_entry_put(&e, &buf);
if (!ok) {
DBG_DEBUG("share_mode_entry_put failed\n");
state->status = NT_STATUS_INTERNAL_ERROR;
return;
}
dbufs[num_dbufs] = (TDB_DATA) {
.dptr = buf.buf, .dsize = SHARE_MODE_ENTRY_SIZE,
};
num_dbufs += 1;
}
idx += 1;
if (idx < state->num_share_modes) {
size_t behind = state->num_share_modes - idx;
dbufs[num_dbufs] = (TDB_DATA) {
.dptr = data.dptr + idx * SHARE_MODE_ENTRY_SIZE,
.dsize = behind * SHARE_MODE_ENTRY_SIZE,
};
num_dbufs += 1;
}
if (e.stale) {
state->num_share_modes -= 1;
}
if (state->num_share_modes == 0) {
state->status = dbwrap_record_delete(rec);
if (!NT_STATUS_IS_OK(state->status)) {
DBG_DEBUG("dbwrap_record_delete failed: %s\n",
nt_errstr(state->status));
}
return;
}
state->status = dbwrap_record_storev(rec, dbufs, num_dbufs, 0);
if (!NT_STATUS_IS_OK(state->status)) {
DBG_DEBUG("dbwrap_record_storev failed: %s\n",
nt_errstr(state->status));
return;
}
}
static bool share_mode_entry_do(
struct share_mode_lock *lck,
struct server_id pid,
uint64_t share_file_id,
void (*fn)(struct share_mode_entry *e,
size_t num_share_modes,
bool *modified,
void *private_data),
void *private_data)
{
struct share_mode_data *d = lck->data;
struct share_mode_entry_do_state state = {
.pid = pid,
.share_file_id = share_file_id,
.fn = fn,
.private_data = private_data,
};
NTSTATUS status;
bool have_share_modes;
status = dbwrap_do_locked(
share_entries_db,
locking_key(&d->id),
share_mode_entry_do_fn,
&state);
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("share_mode_forall_entries failed: %s\n",
nt_errstr(status));
return false;
}
if (!NT_STATUS_IS_OK(state.status)) {
DBG_DEBUG("share_mode_entry_do_fn failed: %s\n",
nt_errstr(status));
return false;
}
have_share_modes = (state.num_share_modes != 0);
if (d->have_share_modes != have_share_modes) {
d->have_share_modes = have_share_modes;
d->modified = true;
}
return true;
}
struct del_share_mode_state {
bool ok;
};
static void del_share_mode_fn(
struct share_mode_entry *e,
size_t num_share_modes,
bool *modified,
void *private_data)
{
struct del_share_mode_state *state = private_data;
e->stale = true;
state->ok = true;
}
bool del_share_mode(struct share_mode_lock *lck, files_struct *fsp)
{
struct del_share_mode_state state = { .ok = false };
bool ok;
ok = share_mode_entry_do(
lck,
messaging_server_id(fsp->conn->sconn->msg_ctx),
fsp->fh->gen_id,
del_share_mode_fn,
&state);
if (!ok) {
DBG_DEBUG("share_mode_entry_do failed\n");
return false;
}
if (!state.ok) {
DBG_DEBUG("del_share_mode_fn failed\n");
return false;
}
return true;
}
struct remove_share_oplock_state {
bool ok;
};
static void remove_share_oplock_fn(
struct share_mode_entry *e,
size_t num_share_modes,
bool *modified,
void *private_data)
{
struct remove_share_oplock_state *state = private_data;
e->op_type = NO_OPLOCK;
*modified = true;
state->ok = true;
}
bool remove_share_oplock(struct share_mode_lock *lck, files_struct *fsp)
{
struct remove_share_oplock_state state = { .ok = false };
bool ok;
ok = share_mode_entry_do(
lck,
messaging_server_id(fsp->conn->sconn->msg_ctx),
fsp->fh->gen_id,
remove_share_oplock_fn,
&state);
if (!ok) {
DBG_DEBUG("share_mode_entry_do failed\n");
return false;
}
if (!state.ok) {
DBG_DEBUG("remove_share_oplock_fn failed\n");
return false;
}
if (fsp->oplock_type == LEASE_OPLOCK) {
remove_lease_if_stale(
lck,
fsp_client_guid(fsp),
&fsp->lease->lease.lease_key);
}
share_mode_wakeup_waiters(fsp->file_id);
return true;
}
struct downgrade_share_oplock_state {
bool ok;
};
static void downgrade_share_oplock_fn(
struct share_mode_entry *e,
size_t num_share_modes,
bool *modified,
void *private_data)
{
struct downgrade_share_oplock_state *state = private_data;
e->op_type = LEVEL_II_OPLOCK;
*modified = true;
state->ok = true;
}
bool downgrade_share_oplock(struct share_mode_lock *lck, files_struct *fsp)
{
struct downgrade_share_oplock_state state = { .ok = false };
bool ok;
ok = share_mode_entry_do(
lck,
messaging_server_id(fsp->conn->sconn->msg_ctx),
fsp->fh->gen_id,
downgrade_share_oplock_fn,
&state);
if (!ok) {
DBG_DEBUG("share_mode_entry_do failed\n");
return false;
}
if (!state.ok) {
DBG_DEBUG("downgrade_share_oplock_fn failed\n");
return false;
}
lck->data->flags |= SHARE_MODE_LEASE_READ;
lck->data->modified = true;
return true;
}
struct mark_share_mode_disconnected_state {
uint64_t open_persistent_id;
bool ok;
};
static void mark_share_mode_disconnected_fn(
struct share_mode_entry *e,
size_t num_share_modes,
bool *modified,
void *private_data)
{
struct mark_share_mode_disconnected_state *state = private_data;
if (num_share_modes != 1) {
state->ok = false;
return;
}
server_id_set_disconnected(&e->pid);
e->share_file_id = state->open_persistent_id;
*modified = true;
state->ok = true;
}
bool mark_share_mode_disconnected(struct share_mode_lock *lck,
struct files_struct *fsp)
{
struct mark_share_mode_disconnected_state state;
bool ok;
if (fsp->op == NULL) {
return false;
}
if (!fsp->op->global->durable) {
return false;
}
state = (struct mark_share_mode_disconnected_state) {
.open_persistent_id = fsp->op->global->open_persistent_id,
};
ok = share_mode_entry_do(
lck,
messaging_server_id(fsp->conn->sconn->msg_ctx),
fsp->fh->gen_id,
mark_share_mode_disconnected_fn,
&state);
if (!ok) {
DBG_DEBUG("share_mode_entry_do failed\n");
return false;
}
if (!state.ok) {
DBG_DEBUG("mark_share_mode_disconnected_fn failed\n");
return false;
}
lck->data->modified = true;
return true;
}
static void reset_share_mode_entry_del_fn(
struct share_mode_entry *e,
size_t num_share_modes,
bool *modified,
void *private_data)
{
struct set_share_mode_state *state = private_data;
state->e = *e;
e->stale = true;
state->status = NT_STATUS_OK;
}
bool reset_share_mode_entry(
struct share_mode_lock *lck,
struct server_id old_pid,
uint64_t old_share_file_id,
struct server_id new_pid,
uint64_t new_mid,
uint64_t new_share_file_id)
{
struct share_mode_data *d = lck->data;
struct set_share_mode_state state = {
.status = NT_STATUS_INTERNAL_ERROR,
};
NTSTATUS status;
bool ok;
ok = share_mode_entry_do(
lck,
old_pid,
old_share_file_id,
reset_share_mode_entry_del_fn,
&state);
if (!ok) {
DBG_DEBUG("share_mode_entry_do failed\n");
return false;
}
if (!NT_STATUS_IS_OK(state.status)) {
DBG_DEBUG("reset_share_mode_entry_del_fn failed: %s\n",
nt_errstr(state.status));
return false;
}
state.status = NT_STATUS_INTERNAL_ERROR;
state.e.pid = new_pid;
state.e.op_mid = new_mid;
state.e.share_file_id = new_share_file_id;
status = dbwrap_do_locked(
share_entries_db,
locking_key(&d->id),
set_share_mode_fn,
&state);
if (!NT_STATUS_IS_OK(status)) {
DBG_WARNING("dbwrap_do_locked failed: %s\n",
nt_errstr(status));
return false;
}
if (!NT_STATUS_IS_OK(state.status)) {
DBG_WARNING("set_share_mode_fn failed: %s\n",
nt_errstr(state.status));
return false;
}
d->have_share_modes = true;
return true;
}