mps0-troubles

[Kenneth D. Merry]

On Sun, Feb 20, 2011 at 04:10:19 +0100, Joachim Tingvold wrote:
> On Fri, Feb 18, 2011, at 18:16:19PM GMT+01:00, Kenneth D. Merry wrote:
> >I just committed a fix to -current for the out of chain buffers  
> >issue.  If
> >you've got time, pull that fix in and see whether it works for you.
> 
> 	[jocke at filserver /storage/tmp]$ i=0;while [ $i -le 10 ]; do dd if=/ 
> dev/zero of=foo bs=1M count=204800; let "i += 1"; done
> 
> I let it run for a while, and checked back later.
> 
> <http://home.komsys.org/~jocke/dmesg_mps0_freebsd-scsi_6.txt>
> 
> 	hw.mps.0.debug_level: 0
> 	hw.mps.0.allow_multiple_tm_cmds: 0
> 	hw.mps.0.io_cmds_active: 0
> 	hw.mps.0.io_cmds_highwater: 970
> 	hw.mps.0.chain_free: 1024
> 	hw.mps.0.chain_free_lowwater: 0
> 	hw.mps.0.chain_alloc_fail: 786
> 
> As you can see, I didn't do anything about the 'MPS_CHAIN_FRAMES'- 
> value after I pulled the latest version. I could change it, but I  
> guess this only would get rid of the 'out of chain'-messages, and not  
> really solve the underlying problem.

Okay, good.  It looks like it is running as designed.

I'll probably just double the number of chain buffers and disable the
messages.

>From previous experiments, the problem is much less likely to occur when
you have 2048 chain buffers, correct?

What filesystem are you using by the way?

Ken
-- 
Kenneth Merry
ken at FreeBSD.ORG