scsi rests during load

[vlado]

Hello,

Sorry for this long message that follow, but I think it's relevant to
this list.
I posted a similar one in sevral newsgroups in hope that it would be
helpful :)

I've run into a problem with an aic7xxx driver giving resets under load
with linux

The log says :
------------
Aug 22 13:25:30 perun kernel: scsi : aborting command due to timeout :
pid 269474, scsi0, channel 0, id 0, lun 0 Write (10) 00 01 0e 1a 98 00
00 80 00
Aug 22 13:25:30 perun kernel: scsi : aborting command due to timeout :
pid 269475, scsi0, channel 0, id 0, lun 0 Write (10) 00 01 0e 1b 18 00
00 80 00
Aug 22 13:25:30 perun kernel: scsi : aborting command due to timeout :
pid 269476, scsi0, channel 0, id 0, lun 0 Write (10) 00 01 0e 1b 98 00
00 80 00
Aug 22 13:25:31 perun kernel: SCSI host 0 abort (pid 269475) timed out -

resetting
Aug 22 13:25:31 perun kernel: SCSI bus is being reset for host 0 channel

0.
Aug 22 13:25:34 perun kernel: (scsi0:0:0:0) Synchronous at 40.0
Mbyte/sec, offset 16.
------------ from   /var/log/messages

It is definitely not a problem with a specific kernel as I the error
persists on kernels 2.2.14-16 with several (  5.1.28 - 5.1.30 )
different aic7xxx patches.
It's not a distro error either both RH and Debian were tested

the error is received during the following test script:

-- test script --------
#!/bin/bash
/opt/home/tiobench-0.3.1/tiobench.pl --dir /group --block 2048 --block
204800 --threads 1  --threads 100  --threads 200  --size 1500
2>spectrum.err | tee spectrum.log
----------

the resets occur the 204800B block size with 1 thread run.

Ok. I understand that this is overkill, but  I need a robust system,
nfs/smb server -
so  this is part of the testing process. Actually I started these harsh
tests after scsi resets
during nfs and smb test loads - copying files :(

Since I don't have the sufficient knowledge and information to start
tampering with the  scsi codes/erorrs and I don't have a similar
hardware to run equivalent tests on it I can't identify the source of
the error :

    - kernel/aic7xxx driver bug ?
    - hardware (scsi controller) error ?
    - cabling error sources are eliminated by several ( 5 ) tests with
different cables, overall bus length is under 2m.
    - the "HD" error /buggy firmware chance is quite low - this is a
second box of this type I'm testing (with different firmware vers.
though )

Hardware details
------------
Motherboard - ASUS P2B-DS SCPI Rev1012
                           Award bios v4.51p6
                          with onboard Adaptec 7890 bios v.2.11

"HD"    - ZERO-D  400 IDE-RAID  system, firmware v.2.2W

Thanks,
Vlado





To Unsubscribe: send mail to majordomo at FreeBSD.org
with "unsubscribe aic7xxx" in the body of the message