Re: Processes hang when using ReiserFS with ACL-Patch

Hello,

this sounds like an oops in the kernel with some locks held (which causes the 
locks to stay held when the offending process is killed). Later processes 
then hang in "D" waiting for the lock to become available, which never 
happens. Please proceed as follows:

	(1) run a file system check,
	(2) echo 1 > /proc/sys/kernel/sysrq
	(3) reproduce the problem, and watch for oopses in the system log,
	(4) when processes start to hang, press sysrq+t to get backtraces of all
	    running tasks; this shows in which locks they are hanging.

Please send the following information:

	- the list of patches you have applied (+versions/links),
	- the kernel configuration used,
	- the System.map for that kernel.

You may want to grab the most recent sources of the "kernel of the day" from 
ftp://ftp.suse.com/pub/people/mantel/kotd/i386/kernel-source.src.rpm which 
contains all the patches that SUSE is using right now, including reiserfs 
acls. There are also binary kernels at the same place that you may want to 
try. I am pretty sure that those kernels won't have the problem.

On Friday 09 April 2004 01:54, Peter Bücker wrote:
> Hi,
>
> I recently started using ReiserFS with the latest ReiserFS-ACL-Patch and
> kernel 2.6.5.
> Since then, the server has become very unstable:
>
> Usually, files are transferred to and from the server using Samba
> 3.0.2a, with ACL-support compiled in.
> After quite a short time (sometimes 30 seconds, sometimes a few
> minutes), Windows reports that the "network drive is no longer
> available" and the transfer aborts.
> At the same time the corresponding smbd-process and syslogd switch to
> state D and never come back until I reboot the machine.
>
> A few other processes also switch to state D after smbd has "crashed",
> most noticably cron and it's child processes.
> After about 10 minutes, even sshd will stop working. I can still
> connect, but it does not seem to be able to spawn a shell anymore.
>
> However, this does not only happen when accessing the partition using
> samba, but also using proftpd.
>
> My disk setup is as follows:
> --------------------------------------------------------------
> /dev/hda1 on / type reiserfs (rw)
> /dev/mapper/ftp-real on /var/ftp type reiserfs (rw,acl)
> --------------------------------------------------------------
>
> /var/ftp is on a logical volume using LVM2.
> When I mount /var/ftp with ACL-support disabled, things work fine.
>
> I will try to gather some detailed debug output from smbd tomorrow.
> So far I can only say that it complains about failing to break an oplock
> on the file being transferred.

Regards,
-- 
Andreas Gruenbacher <[email protected]>
SUSE Labs, SUSE LINUX AG
_______________________________________________
acl-devel mailing list
[email protected]
http://acl.bestbits.at/mailman/listinfo/acl-devel