Date: Tue, 25 May 2010 18:26:49 +0200 From: Jan Henrik Sylvester <me@janh.de> To: Benjamin Kaduk <kaduk@MIT.EDU> Cc: afs-list freebsd <freebsd-afs@freebsd.org> Subject: Re: AFS on FreeBSD 8? Message-ID: <4BFBFA49.8040900@janh.de> In-Reply-To: <alpine.GSO.1.10.1005201320510.29136@multics.mit.edu> References: <558186334.2619.1271175586732.JavaMail.root@thunderbeast.private.linuxbox.com> <4BC4A5F1.6050406@janh.de> <alpine.GSO.1.10.1004131314070.29136@multics.mit.edu> <4BC5EEB5.5050109@janh.de> <alpine.GSO.1.10.1004141242470.29136@multics.mit.edu> <4BE01AF0.3080309@janh.de> <alpine.GSO.1.10.1005041225120.29136@multics.mit.edu> <4BF2B962.50408@janh.de> <alpine.GSO.1.10.1005181201010.29136@multics.mit.edu> <4BF2D461.8030302@janh.de> <AANLkTin_RGNLOzP2X2WaJwwq8Ae7pKJQbA6P6Wbth4U3@mail.gmail.com> <4BF56F38.4040008@janh.de> <alpine.GSO.1.10.1005201320510.29136@multics.mit.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On 05/20/10 19:24, Benjamin Kaduk wrote:
> On Thu, 20 May 2010, Jan Henrik Sylvester wrote:
>
>> I would have tried to narrow down the circumstances that cause
>> deadlocks, if both of the machines in front of me were not crashing so
>> often...
>
> I fixed a bug that was causing very quick deadlocks on my system, in
> revision 42a280f50daf6e4dc65873150c4738aacf2c3a86 ( Wed, 19 May 2010
> 10:39:35 +0000 (03:39 -0700)). Now the most common failure mode I am
> seeing is kernel panics that seem to be due to some form of memory
> corruption.
I repeated my tests again with a build from half an hour ago, once on
SMP, once with kern.smp.disabled=1. Nothing changed: Copying a 2162
bytes file works fine, copying a 256004096 bytes file locks afs. cmdebug
does not return anything.
Trying to reboot, I get "init: some processes would not die; ps axl
advised". After "All buffers synced", afs got a message for me:
afs: WARM shutting down of: CB... afs... BkG... CTrunc... AFSDB...
RxEvent... UnmaskRxkSignals... RxListener... osi_StopListener:
rxk.ListenerPid ffffff0071a9c000 {2nd try: ffffff005cf3460}
WARNING: not all blocks freed: large 0 small 1
All allocated tables... done
Nothing new, unfortunately.
I could try different file sizes... maybe around the cache size -- would
that be interesting? (I have "/afs:/var/openafs/cache:100000" in
/usr/local/etc/openafs/cacheinfo. That is in KB, isn't it?)
Cheers,
Jan Henrik
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4BFBFA49.8040900>