lkml.org 
[lkml]   [2006]   [Mar]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectSystem lockup with processes in D state in 2.6.16.1
Date
Some time ago, we were seeing a problem in the kernel in the reiserfs code 
where a lock inversion issue could cause processes to get stuck in D state,
requiring a system reboot.

http://marc.theaimsgroup.com/?t=108932517300001&r=1&w=2

This link describes the actual call path that causes the problem.
http://marc.theaimsgroup.com/?l=linux-kernel&m=109035413201491&w=2

At the time, the solution we got was to add a patch the basically bypasses
reiser_file_write and just calls generic_file_write. This semed to fix the
problem and we've been running for over a year fine with that patch.

Recent we brought the issue up again with some reiser people, who mentioned
that:

> There was a patch for the problem referenced by this link. (By Cris Mason,
> I think). This patch is long included into vanilla kernel
> (2.6.15 certainly contains it). If you still see deadlocks, I guess you
> need
> to gather some more info again (sysrq-t and friends).

So we recently built 2.6.16.1, without the patch. However after just 1 hour
of stress testing with cyrus again, we were able to lock up the system with
lots of processes stuck in D state and a load running to 500+. The machine
had about 1500 processes running on it, and the dmesg buffer was only 1M, so
it seems we weren't able to capture all the traces with a sysrq-t, but
there's still a lot of info. I've put the sysrq-t and kernel config output
at the links below:

http://kernel.robm.fastmail.fm/sysrq-t-2006-03-30-1.txt
http://kernel.robm.fastmail.fm/kernel-config-2006-03-30-1.txt

Any idea if this is related to the previous problem or is something
different?

Rob


-
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/

\
 
 \ /
  Last update: 2006-03-31 00:44    [W:0.048 / U:0.228 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site