[vortex] Problem with built-in dual 3c982, Dual AMD S2462 MB.

[Heflin, Roger A.]

On Linux 2.2.20 the 3c982 driver that comes with it (18Feb01), appears
to work correctly.   The driver that we originally had with 2.2.19 fails
to find the 
card (19oct00), this says we did not have a clean 2.2.19 set as 2.2.19
should have had the later driver that works (the 18Feb01).   We probably
missed a diff somewhere that contained that change.

The 99U and 99UC drivers fail on both 2.2.19 and 2.2.20 (2.2.20 is a
clean
build - not a patched version of 2.2.19) with the eth0 timeout
errors after several hours of high traffic.

All tests were done with APIC off.

				Roger

> -----Original Message-----
> From:	Heflin, Roger A. 
> Sent:	 1/ 18/ 2002 4:01 PM
> To:	Donald Becker
> Cc:	Bogdan Costescu; vortex@scyld.com
> Subject:	RE: [vortex] Problem with built-in dual 3c982, Dual AMD
> S2462 MB.
> 
> Donald,
> 
> The machines having the issue are both AAPRO and Racksaver
> machines.   They were delivered early december, they are running
> 2.2.19 (1-2 extra patches) what are your customers running?
> 
> So far with noapic it has only failed once on 1 machine, we don't
> know absolutely for sure that it was the same thing, but the 
> log messages look consistant.
> 
> I will setup a larger test with both more machines and more 
> iterations of the test to determine if this 1 machine was a
> fluke and has some other issue, or if their is truly a problem.
> 
> 		Roger
> 
> > -----Original Message-----
> > From:	Donald Becker [SMTP:becker@scyld.com]
> > Sent:	 1/ 18/ 2002 2:53 PM
> > To:	Heflin, Roger A.
> > Cc:	Bogdan Costescu; vortex@scyld.com
> > Subject:	RE: [vortex] Problem with built-in dual 3c982, Dual AMD
> > S2462 MB.
> > 
> > On Fri, 18 Jan 2002, Heflin, Roger A. wrote:
> > 
> > > We have confirmed that noapic is now turned on, noapic did
> > > improve things, it appears that the situation is probably 10
> > > times better, but is still a significant problem.
> > 
> > We always enable 'noapic' by default to avoid this problem.  It's a
> > slight performance hit on SMP, but the reliability is well worth it.
> > 
> > > And from what I can tell all of the recent tyan dual
> > > cpu AMD boards S2462 with the 3com's on them
> > > seem to  do this.   I have a very large number of 
> > > the machines (100+) and the all appear to 
> > > have the issue, or at the very least alot of them 
> > > appear to have the issue.
> > 
> > Hmmm, customers have used those on clusters running our Beowulf
> system
> > without reported prolems.  We have one in-house, loaned by APPro for
> > release testing, that doesn't exhibit a problem.
> > 
> > Donald Becker				becker@scyld.com
> > Scyld Computing Corporation		http://www.scyld.com
> > 410 Severn Ave. Suite 210		Second Generation Beowulf
> > Clusters
> > Annapolis MD 21403			410-990-9993
> > 
> _______________________________________________
> vortex mailing list
> vortex@scyld.com
> http://www.scyld.com/mailman/listinfo/vortex