s3: messages: Implement cleanup of dead records.

When a smbd process dies, pending messages.tdb records for this process
might not get cleaned up. Implement a cleanup for dead records that is
triggered after a smbd dies uncleanly; the records for that PID are
deleted.

Based on a patchset from Christof Schmitt <cs@samba.org>.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>

(cherry picked from commit 837671f47670b16726aa96ba7a0902974a1037eb)

Bug: https://bugzilla.samba.org/show_bug.cgi?id=10534
Cleanup messages.tdb record after unclean smbd shutdown

Autobuild-User(v4-0-test): Karolin Seeger <kseeger@samba.org>
Autobuild-Date(v4-0-test): Mon Apr  7 11:55:50 CEST 2014 on sn-devel-104

4 files changed, 68 insertions, 0 deletions

diff --git a/source3/include/messages.h b/source3/include/messages.h
index 4b45901afda..a45f5a5d028 100644
--- a/source3/include/messages.h
+++ b/source3/include/messages.h
@@ -101,6 +101,9 @@ bool messaging_tdb_parent_init(TALLOC_CTX *mem_ctx);
 void *messaging_tdb_event(TALLOC_CTX *mem_ctx, struct messaging_context *msg,
 			  struct tevent_context *ev);
 
+NTSTATUS messaging_tdb_cleanup(struct messaging_context *msg_ctx,
+			struct server_id pid);
+
 NTSTATUS messaging_ctdbd_init(struct messaging_context *msg_ctx,
 			      TALLOC_CTX *mem_ctx,
 			      struct messaging_backend **presult);
@@ -140,6 +143,9 @@ NTSTATUS messaging_send_buf(struct messaging_context *msg_ctx,
 void messaging_dispatch_rec(struct messaging_context *msg_ctx,
 			    struct messaging_rec *rec);
 
+void messaging_cleanup_server(struct messaging_context *msg_ctx,
+				struct server_id pid);
+
 #include "librpc/gen_ndr/ndr_messaging.h"
 
 #endif
diff --git a/source3/lib/messages.c b/source3/lib/messages.c
index cd763e7265a..5e738c74619 100644
--- a/source3/lib/messages.c
+++ b/source3/lib/messages.c
@@ -397,4 +397,21 @@ void messaging_dispatch_rec(struct messaging_context *msg_ctx,
 	return;
 }
 
+/*
+  Call when a process has terminated abnormally.
+*/
+void messaging_cleanup_server(struct messaging_context *msg_ctx,
+				struct server_id server)
+{
+	if (server_id_is_disconnected(&server)) {
+		return;
+	}
+
+	if (!procid_is_local(&server)) {
+		return;
+	}
+
+	(void)messaging_tdb_cleanup(msg_ctx, server);
+
+}
 /** @} **/
diff --git a/source3/lib/messages_local.c b/source3/lib/messages_local.c
index 6b63d729254..859eeb858e3 100644
--- a/source3/lib/messages_local.c
+++ b/source3/lib/messages_local.c
@@ -45,6 +45,7 @@
 #include "includes.h"
 #include "system/filesys.h"
 #include "messages.h"
+#include "serverid.h"
 #include "lib/tdb_wrap/tdb_wrap.h"
 #include "lib/param/param.h"
 
@@ -200,6 +201,43 @@ static TDB_DATA message_key_pid(TALLOC_CTX *mem_ctx, struct server_id pid)
 	return kbuf;
 }
 
+/*******************************************************************
+ Called when a process has terminated abnormally. Remove all messages
+ pending for it.
+******************************************************************/
+
+NTSTATUS messaging_tdb_cleanup(struct messaging_context *msg_ctx,
+				struct server_id pid)
+{
+	struct messaging_tdb_context *ctx = talloc_get_type(
+					msg_ctx->local->private_data,
+					struct messaging_tdb_context);
+	struct tdb_wrap *tdb = ctx->tdb;
+	TDB_DATA key;
+	TALLOC_CTX *frame = talloc_stackframe();
+
+	key = message_key_pid(frame, pid);
+	/*
+	 * We have to lock the key to avoid
+	 * races in case the server_id was
+	 * re-used and is active (a remote
+	 * possibility, true). We only
+	 * clean up the database if we
+	 * know server_id doesn't exist
+	 * while checked under the chainlock.
+	 */
+	if (tdb_chainlock(tdb->tdb, key) != 0) {
+		TALLOC_FREE(frame);
+		return NT_STATUS_LOCK_NOT_GRANTED;
+	}
+	if (!serverid_exists(&pid)) {
+		(void)tdb_delete(tdb->tdb, key);
+	}
+	tdb_chainunlock(tdb->tdb, key);
+	TALLOC_FREE(frame);
+	return NT_STATUS_OK;
+}
+
 /*
   Fetch the messaging array for a process
  */
diff --git a/source3/smbd/server.c b/source3/smbd/server.c
index a86fa488c80..86c20baf435 100644
--- a/source3/smbd/server.c
+++ b/source3/smbd/server.c
@@ -455,6 +455,13 @@ static void remove_child_pid(struct smbd_parent_context *parent,
 						parent);
 			DEBUG(1,("Scheduled cleanup of brl and lock database after unclean shutdown\n"));
 		}
+
+		/*
+		 * Ensure we flush any stored messages
+		 * queued for the child process that
+		 * terminated uncleanly.
+		 */
+		messaging_cleanup_server(parent->msg_ctx, child_id);
 	}
 
 	if (!serverid_deregister(child_id)) {