Re: linux-2.4.0 scsi problems on NetFinity servers

Tim Wright (timw@splhi.com)
Wed, 10 Jan 2001 12:49:58 -0800


Hmmm...
it's actually not quite that simple. The card on it's own doesn't cause any
problems. It's when the NMI watchdog stuff is enabled that all hell breaks
loose at least on my 8500R. Basically, every CPU in the system gets hammered
with NMIs (1000's per second). The system is slower than it should be, and in
my case it hangs after ~45 minutes (~256,000 NMIs per cpu). Booting with
nmi_watchdog=0 makes the problem go away and the machine is stable, so there's
some kind of nasty interaction with the card.

It seems a little unlikely that this is related to SCSI problems, but I could
be wrong. Anyway, I am trying to find more information on the adapter to find
out where the problem may lie.

Regards,

Tim

On Tue, Jan 09, 2001 at 03:08:03PM -0600, JP Navarro wrote:
> One possibility:
>
> When we first tested 2.4.0-test8 on NetFinity 7000s we had random crashes,
> typically within an hour of booting. The problem was identified as a Wiseman
> Systems Management adapter generated hardware interrupt that 2.4 doesn't handle
> (this was not a problem with 2.2.x).
>
> If you have these adapters installed, remove them.
>
> JP Navarro
> --
> John-Paul Navarro (630) 252-1233
> Mathematics & Computer Science Division
> Argonne National Laboratory navarro@mcs.anl.gov
> Argonne, IL 60439 http://www.mcs.anl.gov/~navarro
>
>
> Ken Brunsen/Iris wrote:
> >
> > Hello all,
> >
> > I've been sorta pulling the 2.4 kernel and testing with it now for
> > awhile on my IBM NetFinity 5500 and since the test12 I've been having a
> > continuous issue with crashing the OS during a pull of source code across
> > the network (>1Gb files). I've been trying to figure out what it may be
> > related to, but I'm relatively new with debugging the kernel so thought I'd
> > see if y'all could help. From looking at the archives, I did not see that
> > anyone else had been seeing these issues either. Basically, I've got 2
> > different machines which I'm working with - a NetFinity Quad CPU 5500 M20
> > with 2Gb Ram and Raid and a NetFinity Dual CPU 5500 M10 with 1Gb Ram and
> > Raid. Both machines exhibit the same behavior. Initially, both machines
> > had RH 6.0, now one is RH 7.0 (and I know about the compiler issue) and the
> > other is SuSE 7.0. I downloaded the 2.4.0 release and still got the issue,
> > so thought it was time to bring it here. Here is a stack of one crash:
> >
> > Started getting Scsi errors on controller during NFS transfer of >1Gb
> > worth of files
> >
> > SCSI disk error : host 0 channel 0 id 0 lun 0 return code = 70000
> > I/O error: dev 08:05, sector 31731256
> > SCSI disk error : host 0 channel 0 id 0 lun 0 return code = 70000
> > I/O error: dev 08:05, sector 31731264
> > SCSI disk error : host 0 channel 0 id 0 lun 0 return code = 70000
> > I/O error: dev 08:05, sector 31731272
> > SCSI disk error : host 0 channel 0 id 0 lun 0 return code = 70000
> > I/O error: dev 08:05, sector 31731280
> > .
> > .
> > .
> >
> > (the sector varies from run to run, is never consistent), and then
> > kernel panics with the following
> >
> > (ips0) Resetting controller.
> > NMI Watchdog detected LOCKUP on CPU1, registers:
> > CPU: 1
> > EIP: 0010:[<c0246544>]
> > EFLAGS: 00000002
> > eax: 003e240 ebx: 000612b0 ecx: 5a21a2f5 edx: 00000063
> > esi: 00000004 edi: 00000000 ebp:f7de2a78 esp: f7ddbf00
> > ds: 0018 es: 0018 ss: 0018
> > Process scsi_eh_0 (pid: 8, stackpage=f7ddb000)
> > Stack: 000003e6 c0246587 000612b0 c02465f5 000612b0 c01df470 00418570
> > ffffffff
> > f7de2a78 00000082 00000001 200012b0 f7ddbf36 000612b0 c01dfa7c
> > f7de2a78
> > f7de2ab8 f7de2a78 f7db1400 f7de2ab8 c01dc4ae f7de2a78 c0296220
> > c0295c67
> > Call Trace: [<c0246587>] [<c02465f5>] [<c01df470>] [<c01dfa7c>]
> > [<c01dc4ae>]
> > [<c01bda9c>] [<c01be1db>] [<c01be4e6>] [<c01074c4>]
> >
> > Code: 39 d8 72 f8 5b c3 89 f6 8b 44 24 04 eb 0e 8d b4 26 00 00 00
> > console shuts up ...
> >
> > Thinking it could be memory related - since I see the Cache fill up and the
> > system go to just over 1mb free prior to crash - i disabled highmem
> > support. I then disabled NFSv3 and automounter v4 support, jic. In the
> > last test, I disabled swap - since one thing I've noticed is that the 2.4
> > kernel never touches my swap at all. None of these changes have affected
> > the outcome; the closest I've gotten is by contintually doing "sync" in
> > another window which sometimes keeps it from crashing on a run, although
> > I'll still end up with a few of the SCSI disk error messages (although not
> > nearly as many as I get before a failure). Since this happens on multiple
> > machines, I do not believe it is. We're also seeing failures of this same
> > type when we try to do heavy database loading on the machine, ie., intense
> > disk accesses. Any help would be greatly appreciated, as we are really
> > needing to get this 2.4 kernel working
> >
> > Since I only get the archive list, please CC me with any responses!
> >
> > Thanks!
> >
> > kenbo
> >
> > ______________________
> > Firebirds rule, `stangs serve!
> >
> > Kenneth "kenbo" Brunsen
> > Iris Associates
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> Please read the FAQ at http://www.tux.org/lkml/

-- 
Tim Wright - timw@splhi.com or timw@aracnet.com or twright@us.ibm.com
IBM Linux Technology Center, Beaverton, Oregon
"Nobody ever said I was charming, they said "Rimmer, you're a git!"" RD VI
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/