On Jan 21, 15:29, "Michael L. VanLoon -- Iowa State University" wrote:
} Subject: NCR PCI SCSI bugs/fixes?  Help!

} I'm using a Dell OmniPlex 66MHz Pentium machine at my new job with a
} NCR PCI SCSI (53c810) card in it and two SCSI drives.  I'm running
} NetBSD-1.0 that was supped off sun-lamp before the tree was turned
} back into current, several weeks ago (sources and binaries from my own
} home machine).  The machine was working wonderfully under light load.
} We are running a news server on it, and ran a limited feed to it to
} test things out.  Things were working great.
} 
} But, when we opened up the full feed, sending literally thousands of
} articles an hour, things aren't working as well.  Twice in the last
} two days the machine has died.

Sorry to hear about your problems.

There have been no fixes recently, and the code
works well for us under light and high load, so
there might be some other problem, not related 
to the driver (of course we take this seriously 
and will look into this !).

(There have been reports about panics under high 
load on two systems using the NCR with FreeBSD-2.0,
but it is not obvious at all that the NCR driver
is at fault. We need more input to start looking
for a possible bug ...)

Could you please send us as much information as
you have available:

1) Error messages (best send the complete log file, gzipped, uudecoded)

2) Hardware description (number of drives, models, internal/extrernal, ...)

I've seen the kind of behaviour on our Sun Sparc 
file server before, and it was related to problems 
with long cables, which made one particular device
act strange and interfere with the other devices
operation (i.e. we didn't use the tape drive that 
caused the problem, but it made our disks loose 
files ...).

The other case where I heard of problems under high 
load it turned out to have been an overloaded power
supply, that just couldn't support all drives seeking
simultanously ...

It's not easy to distinguish such problems from 
driver bugs, since in both cases things just get 
confused ...

Send the detailed data to Wolfgang Stanglmeier and me,
and wwe'll look into it immediately.

Regards, STefan

-- 
 Stefan Esser				Internet:	<se@ZPR.Uni-Koeln.DE>
 Zentrum fuer Paralleles Rechnen	Tel:		+49 221 4706019
 Universitaet zu Koeln			FAX:		+49 221 4705160
 Weyertal 80
 50931 Koeln