Attachment 8791 Details for Bug 9806 – Patch against samba 4.0.5

[patch] Patch against samba 4.0.5

0001-standard-process-mode-restart-crashed-task-services.patch (text/plain), 2.45 KB, created by Samjam on 2013-04-19 10:36:16 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Samjam

Created: 2013-04-19 10:36:16 UTC

Size: 2.45 KB

patch

obsolete

>From 53ac493fa391a126ed737a8d6dfeae7acfe61bc6 Mon Sep 17 00:00:00 2001
>From: Sam Liddicott <sam@liddicott.com>
>Date: Fri, 19 Apr 2013 11:33:17 +0100
>Subject: [PATCH] standard process mode restart crashed task services
>
>Although it should not happen task services can crash and leave
>the whole server set unusable, see bug #9806
>
>This patch has the process_standard model restart crashed
>servcies.
>
>Signed-off-by: Sam Liddicott <sam@liddicott.com>
>---
> source4/smbd/process_standard.c |   45 ++++++++++++++++++++++++++++-----------
> 1 file changed, 33 insertions(+), 12 deletions(-)
>
>diff --git a/source4/smbd/process_standard.c b/source4/smbd/process_standard.c
>index c5377b3..dce9c3b 100644
>--- a/source4/smbd/process_standard.c
>+++ b/source4/smbd/process_standard.c
>@@ -172,18 +172,39 @@ static void standard_new_task(struct tevent_context *ev,
> 		      standard_pipe_handler, NULL);
> 	close(child_pipe[1]);
> 
>-	/* Ensure that the forked children do not expose identical random streams */
>-	set_need_random_reseed();
>-
>-	setproctitle("task %s server_id[%d]", service_name, (int)pid);
>-
>-	/* setup this new task.  Cluster ID is PID based for this process model */
>-	new_task(ev, lp_ctx, cluster_id(pid, 0), private_data);
>-
>-	/* we can't return to the top level here, as that event context is gone,
>-	   so we now process events in the new event context until there are no
>-	   more to process */	   
>-	tevent_loop_wait(ev);
>+	setproctitle("*%s*", service_name);
>+	signal(SIGCHLD, SIG_DFL);
>+
>+	while(1) {
>+		int status;
>+		pid_t cpid = fork();
>+
>+		if (cpid == 0) {
>+			/* Ensure that the forked children do not expose identical random streams */
>+			set_need_random_reseed();
>+
>+			setproctitle("task %s server_id[%d]", service_name, (int)pid);
>+
>+			/* setup this new task.  Cluster ID is PID based for this process model */
>+			new_task(ev, lp_ctx, cluster_id(pid, 0), private_data);
>+
>+			/* we can't return to the top level here, as that event context is gone,
>+			   so we now process events in the new event context until there are no
>+			   more to process */	   
>+			tevent_loop_wait(ev);
>+
>+			talloc_free(ev);
>+			exit(0);
>+		}
>+
>+		DEBUG(0, ("waiting for child %d %s\n", (int)cpid, service_name));
>+		waitpid(cpid, &status, 0);
>+		if (WIFEXITED(status)) {
>+			DEBUG(0, ("Exit after %d from child %d %s\n", status, (int)cpid, service_name));
>+			break;
>+		}
>+		DEBUG(0, ("Restart after %d from child %d %s\n", status, (int)cpid, service_name));
>+	}
> 
> 	talloc_free(ev);
> 	exit(0);
>-- 
>1.7.10.4
>

Actions: View

Attachments on bug 9806: 8782 | 8791 | 8946