[gdm-list] (PATCH) Re: gdm hung after killing child

[Simon Bowden <simonb cse unsw EDU AU>]

Hello,

In the spirit of solving my own problems...

I have found that this is because when a gdm slave is killed, it shuts 
down the session and exits with DISPLAY_ABORT, which means the master 
drops that display.

The attached patch makes it exit with DISPLAY_REMANAGE for displays of 
type TYPE_LOCAL. Incidentally, this replicates xdm's behaviour when its 
slave is killed.

If this is not acceptable (why?) then I have an alternative patch that 
handles the HUP signal and exits with REMANAGE when it receives one of 
them.

On a side note, the slave's global "d" variable is initialised after the 
signal handlers are installed, so there is a race condition if a signal 
arrives before gdm_slave_run (so if one arrives, then blech). It's very 
unlikely to happen, but why not just d = display or d = 0 earlier?

Cheers,

 - Simon

On Tue, 20 Sep 2005, Simon Bowden wrote:

> Hello,
>
> Running a Debian GDM: 2.6.0.8.
>
> When GDM starts, it launches a child process which in turn launches the X 
> server, resulting in a structure like this:
>
> root     18325     1  0 16:57 ?        00:00:00 gdm
> root     18326 18325  0 16:57 ?        00:00:00 gdm
> root     18586 18326  0 17:08 ?        00:00:03 /usr/X11R6/bin/X :0 -audit 0 
> -auth /var/lib/gdm/:0.Xauth
>
> Now, we have something that due to historical reasons sends a SIGTERM to the 
> parent of the X server (18326 above). That process dies, takes the X server 
> with it (which is intended), BUT, then nothing starts up again.
>
> The 18325 process just sits there indefinitely. No X server is running now. 
> If I SIGHUP the process then it wakes up, but that's manual intervention.
>
> # strace -f -p 18325
> Process 18325 attached - interrupt to quit
> poll(
> [never returns in many minutes waited]
>
> Obviously one could argue that we should just kill the parent of it, but I 
> really don't think gdm should react like that anyway.
>
> Some further debug info:
> It logs this:
> Sep 20 18:19:53 gdm[18325]: gdm_child_action: Aborting display :0
>
> The strace does this just before the poll:
> send(7, "<30>Sep 20 18:19:53 gdm[18325]: gdm_child_action: Aborting display 
> :0", 69, 0) = 69
> rt_sigaction(SIGPIPE, {SIG_IGN}, NULL, 8) = 0
> time(NULL)                              = 1127204393
> close(8)                                = 0
> gettimeofday({1127204393, 22553}, {4294966696, 0}) = 0
> waitpid(-1, 0xbfb601f8, WNOHANG)        = -1 ECHILD (No child processes)
> poll( <unfinished ...>
>
>
> Cheers,
>
> - Simon
--- daemon/slave.c.orig	2005-10-25 13:55:56.000000000 +1000
+++ daemon/slave.c	2005-10-25 13:49:55.000000000 +1000
@@ -4359,7 +4359,11 @@
 
 	gdm_debug ("gdm_slave_term_handler: %s got TERM/INT signal", d->name);
 
-	exit_code_to_use = DISPLAY_ABORT;
+        if (d && d->type == TYPE_LOCAL)
+            exit_code_to_use = DISPLAY_REMANAGE;
+        else
+            exit_code_to_use = DISPLAY_ABORT;
+
 	need_to_quit_after_session_stop = TRUE;
 
 	if (already_in_slave_start_jmp ||