Re: 2.6.17-mm3

[Reuben Farrelly <reuben-lkml@reub.net>]

On 27/06/2006 8:52 p.m., Andrew Morton wrote:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.17/2.6.17-mm3/
>
>
> - Added the x86 Geode development tree to the -mm lineup, as git-geode.patch
>   (Jordan Crouse).
>
> - The locking validator patches are back, in a reworked series.

Still seeing the same breakage as I posted about here in -mm2:

http://www.ussg.iu.edu/hypermail/linux/kernel/0606.3/0195.html

I've since noticed that mail delivery is held up by the deliver process. 
'deliver' is a component of the dovecot mail system (http://www.dovecot.org/) 
which does delivery of mails to a user's Maildir and in the process recreates an 
index.

[root@tornado ~]# nice top
top - 23:05:13 up 47 min,  4 users,  load average: 6.48, 4.80, 4.10
Tasks: 170 total,   5 running, 165 sleeping,   0 stopped,   0 zombie
Cpu(s)%:  0.2 us, 98.9 sy,  0.0 ni,  0.6 id,  0.2 wa,  0.2 hi,  0.0 si,  0.0 st
Mem:   1016764k total,   847184k used,   169580k free,    56840k buffers
Swap:   497936k total,        0k used,   497936k free,   267064k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
 2231 reuben    25   0  7984 1392 1136 R   55  0.1  23:20.16 deliver
 2591 reuben    25   0  7988 1392 1136 R   53  0.1  22:41.52 deliver
 2232 reuben    25   0  7988 1392 1136 R   50  0.1  23:23.78 deliver
 2586 reuben    25   0  7984 1392 1136 R   49  0.1  22:28.72 deliver


[root@tornado ~]# strace -p 2232
Process 2232 attached - interrupt to quit

I can't kill these processes with a KILL signal, nor can I strace them.

Revert to -mm1 or any previous kernel release, and it all works fine.  That 
leads me to believe this is probably not a userspace issue.

What patches which could have possibly caused this?

Reuben
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/