[9fans] SuperMicro P6DGU

[9fans] SuperMicro P6DGU

[Steve Simon]

Hi,

Trying a SuperMicro P6DGU dual PIII 600 motherboard
in my all-in-one fossil/venti/cpu/auth server and
have had a few(3) unexpected reboots when running
with SMP enabled; the one panic I saw was an fp underflow
in fossil.

I am running it single cpu at the moment and its seems
to be reliable (so far). It could have been a cosmic ray strike,
or perhaps I have a duff motherboard (it was free).

Anyone running multi cpu x86 servers? if so and they are cpu
servers, anyone any idea if the all-in-one architecture might
prevoke problems?

-Steve

Re: [9fans] SuperMicro P6DGU

[Jim McKie]

Our main fileserver is a SuperMicro dual-pIII, I'm out of
town until Friday but I think it is a 370DLE, 933MHz. It has
both venti+fossil running and boots standalone with its own
fossil partition. We have lots of SMP systems.

Make sure you have binaries (including libraries) made recently,
there was a compiler bug over the summer that could perhaps cause
weird things like fp stack underflow.

--jim

----- Original Message ----- 
From: "Steve Simon" <steve@quintile.net>
To: <9fans@cse.psu.edu>
Sent: Monday, November 08, 2004 6:32 AM
Subject: [9fans] SuperMicro P6DGU


> Hi,
> 
> Trying a SuperMicro P6DGU dual PIII 600 motherboard
> in my all-in-one fossil/venti/cpu/auth server and
> have had a few(3) unexpected reboots when running
> with SMP enabled; the one panic I saw was an fp underflow
> in fossil.
> 
> I am running it single cpu at the moment and its seems
> to be reliable (so far). It could have been a cosmic ray strike,
> or perhaps I have a duff motherboard (it was free).
> 
> Anyone running multi cpu x86 servers? if so and they are cpu
> servers, anyone any idea if the all-in-one architecture might
> prevoke problems?
> 
> -Steve
> 

Re: [9fans] SuperMicro P6DGU

[Steve Simon]

Ok,

I understand a bit better now, I now have had 3 panics
and two of them where prevoked. Both these times I just rebooted
without halting - very early, just as fossil was comming up.

Both times I got an fp overflow and a panic, when I rebooted
again afterwards the system came up cleanly.

	panic (fp overflow) fppc=0x22aab status=0x80c1 pc=0x00022aa9

The problem seems to occur after fossil has checked a dirty filesystem.

0x22aa9 in fossil relates to /sys/src/libc/386/vlrt.c:105 which is _v2d().

Suffice to say my box seems happy now, I have one unexplained crash but
I would be happy to put that down to a mains dropout.

The above info might help somone else if this bites more severely.

-Steve

Re: [9fans] SuperMicro P6DGU

[C H Forsyth]

[-- Attachment #1: Type: text/plain, Size: 154 bytes --]

i'd have said that was an old bug fixed by adding fpclear to l.s and main.c
and that kernel is somehow out of date, unless there's a new manifestation.

[-- Attachment #2: Type: message/rfc822, Size: 2947 bytes --]

From: "Steve Simon" <steve@quintile.net>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] SuperMicro P6DGU
Date: Wed, 10 Nov 2004 13:02:26 0000
Message-ID: <48fa91a91dec3951895c5ab89a975502@quintile.net>

Ok,

I understand a bit better now, I now have had 3 panics
and two of them where prevoked. Both these times I just rebooted
without halting - very early, just as fossil was comming up.

Both times I got an fp overflow and a panic, when I rebooted
again afterwards the system came up cleanly.

	panic (fp overflow) fppc=0x22aab status=0x80c1 pc=0x00022aa9

The problem seems to occur after fossil has checked a dirty filesystem.

0x22aa9 in fossil relates to /sys/src/libc/386/vlrt.c:105 which is _v2d().

Suffice to say my box seems happy now, I have one unexplained crash but
I would be happy to put that down to a mains dropout.

The above info might help somone else if this bites more severely.

-Steve