AIC-7902 problems

[Hall J D (ISeLS)]

Thanks for the further information Justin.

My problem is I've got a X5DE8 motherboard which is ServerWorks Grand
Champion SL based so I don't think it's got a P64H2. At least there's no
mention of one in the motherboard manual and I can't find one on the board.

Don's suggestion of turning off packetisation has helped with the problem
and I've now managed to install 4.8-RELEASE. Uptime is 17 hours so far with
no SCSI errors reported. Although everything is now running at Ultra160
instead of Ultra320. I don't know if this gives you any clues as to where
the problem might lie.

Jonathan

-----Original Message-----
From: Justin T. Gibbs [mailto:gibbs at scsiguy.com]
Sent: 09 May 2003 05:10
To: Don Bowman; 'Kenneth D. Merry'; Hall J D (ISeLS)
Cc: 'freebsd-scsi at freebsd.org'
Subject: RE: AIC-7902 problems


> FWIW, we have just run into the same issue with new supermicro
> X5 motherboards as well. Downgrading the system bios to c182 from
> 1.3c (no idea what these mean:) made the problem go away.
> We are working with supermicro support on it right now.

As I mentioned to Don in private email, I believe that this problem
is caused by an errata in the P64H2 PCI-X hub.  Information about
this errata can be found here:

http://developer.intel.com/design/chipsets/e7500/specupdt/290735.htm

My suspicion is that the SCBs for some transactions are getting
partially corrupted so the command that we think we are sending
to the drive is not really getting sent.  A good way to verify this
is to force the PCI-X bus into PCI mode.  The problem is supposed
to only occur if there is another device on the PCI-X bus sharing
the 7902, but it would not surprise me if the 7902 could also be
hit with an unexpected GNT.

I will be working with Supermicro to see if I can get both a PCI-X
and SCSI bus trace to help isolate the problem.

--
Justin