kernel/exit.c: if init dies, log a signal which killed it, if any
authorDenys Vlasenko <vda.linux@googlemail.com>
Fri, 23 Mar 2012 22:01:54 +0000 (15:01 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 23 Mar 2012 23:58:32 +0000 (16:58 -0700)
I just received another user's pleas for help when their init
mysteriously died.  I again explained that they need to check whether it
died because of bad instruction, a segv, or something else.  Which was
an annoying detour into writing a trivial C program to spawn his init
and print its exit code:

  http://lists.busybox.net/pipermail/busybox/2012-January/077172.html

I hear you saying "just test it under /bin/sh".  Well, the crashing init
_was_ /bin/sh.

Which prompted me to make kernel do this first step automatically.  We can
print exit code, which makes it possible to see that death was from e.g.
SIGILL without writing test programs.

[akpm@linux-foundation.org: add 0x to hex number output]
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kernel/exit.c

index 456329fd4ea35cff20cd10b082497e7e4088b4ab..3db1909faed9331488cecad09e10359db175a746 100644 (file)
@@ -711,8 +711,11 @@ static struct task_struct *find_new_reaper(struct task_struct *father)
 
        if (unlikely(pid_ns->child_reaper == father)) {
                write_unlock_irq(&tasklist_lock);
-               if (unlikely(pid_ns == &init_pid_ns))
-                       panic("Attempted to kill init!");
+               if (unlikely(pid_ns == &init_pid_ns)) {
+                       panic("Attempted to kill init! exitcode=0x%08x\n",
+                               father->signal->group_exit_code ?:
+                                       father->exit_code);
+               }
 
                zap_pid_ns_processes(pid_ns);
                write_lock_irq(&tasklist_lock);