Comment 87 for bug 769927

Revision history for this message
In , Ian (ian-redhat-bugs) wrote :

(In reply to comment #34)
> I rolled up a small test program that can reproduce this now without a gui:
>
> -----------------[snip]------------------
> #include <stdio.h>
> #include <sys/types.h>
> #include <sys/stat.h>
> #include <unistd.h>
>
> int
> main(int argc, char **argv)
> {
> struct stat buf;
>
> fork();
>
> stat(argv[1], &buf);
>
> return 0;
> }
> -----------------[snip]------------------

OK, I'll try this in a VM.
Only have an F15, a 2.6.38 based kernel I think, but that shouldn't make a difference.

>
> ...when I run this with /misc/salusa/scratch/foo as the first argument, then
> unmounting /misc/salusa triggers the panic. Ian, were you trying to reproduce
> this on a single-cpu box before? If so, then that might explain why you
> couldn't get it to happen.

I use at least 2-cpus in vms, and I think I had 4 in the one I
used to try and reproduce this.

>
> With this, I think we can rule out inotify as the problem. The issue is
> something with simultaneous d_automounts running.

It does look that way, yes.
At least that will save me the time trying to write an inotify
based reproducer.

Ian