2006-01-16 02:05:56

by Gilles May

[permalink] [raw]
Subject: SMP trouble

Hello,

I got a wierd problem with my dual Athlon box.
The board is a K7D Master-L with 2 Athlon-MP 2800+ processors.
Running it with SMP enabled in the kernel makes it freeze on heavy
activity. I can always reproduce a freeze
by watching a movie while copying files to/from USB disk, or on ping -f
to a box on my LAN. Without SMP
support in the kernel I can do this for hours and no freeze.
The kernels I tried are ranging from 2.6.11-1.1369 (FC4) to 2.6.15
vanilla kernel. Running from console
with no X nor any proprietary modules loaded.

I already tried diffent things to no avail like:
Different preemtion models
acpi=off on boot
Enable kernel irq balancing on/off
Lots of different BIOS setting (Using fail-safe most of the time though)

As my HW monitors tell me there's no overheating going on on any of the
CPU's (at 53C and 54C now).

I'd be really grateful for any ideas / workarounds as I really don't
know what to try anymore and a new machine is
out of the question financially.

Regards, Gilles May


2006-01-19 05:11:14

by Zwane Mwaikambo

[permalink] [raw]
Subject: Re: SMP trouble

On Mon, 16 Jan 2006, Gilles May wrote:

> I got a wierd problem with my dual Athlon box.
> The board is a K7D Master-L with 2 Athlon-MP 2800+ processors.
> Running it with SMP enabled in the kernel makes it freeze on heavy activity. I
> can always reproduce a freeze
> by watching a movie while copying files to/from USB disk, or on ping -f to a
> box on my LAN. Without SMP
> support in the kernel I can do this for hours and no freeze.
> The kernels I tried are ranging from 2.6.11-1.1369 (FC4) to 2.6.15 vanilla
> kernel. Running from console
> with no X nor any proprietary modules loaded.

Try booting the SMP kernel with 'noapic' kernel parameter and then send
the kernel bootlog.

2006-01-19 18:21:10

by Gilles May

[permalink] [raw]
Subject: Re: SMP trouble

Zwane Mwaikambo wrote:

>On Mon, 16 Jan 2006, Gilles May wrote:
>
>
>
>>I got a wierd problem with my dual Athlon box.
>>The board is a K7D Master-L with 2 Athlon-MP 2800+ processors.
>>Running it with SMP enabled in the kernel makes it freeze on heavy activity. I
>>can always reproduce a freeze
>>by watching a movie while copying files to/from USB disk, or on ping -f to a
>>box on my LAN. Without SMP
>>support in the kernel I can do this for hours and no freeze.
>>The kernels I tried are ranging from 2.6.11-1.1369 (FC4) to 2.6.15 vanilla
>>kernel. Running from console
>>with no X nor any proprietary modules loaded.
>>
>>
>
>Try booting the SMP kernel with 'noapic' kernel parameter and then send
>the kernel bootlog.
>
>
Hi and thanks for answering..
Attached the bootlog with noapic parameter passed to the kernel. It
still freezes though. :(
What I do exactly to make it freeze is after boot:

In one console I do a ping -f to a box on my local network using the
e100 card. (integrated on the motherboard)

In another console I copy a 2.5 GB file from my USB HDD to the IDE HDD
in a while loop (or do a readcd from USB DVD Writer to a file on IDE HDD)

In a third one I play an MP3 using mplayer (or mpg123) also in a while [
1 ] loop

This guarantees me a nice freeze after at most 20 Minutes. The same
thing runs very well for hours without SMP support compiled in.

Regards, Gilles May


Attachments:
noapic-2.6.15.log (13.55 kB)

2006-01-19 19:27:40

by Zwane Mwaikambo

[permalink] [raw]
Subject: Re: SMP trouble

On Thu, 19 Jan 2006, Gilles May wrote:

> Attached the bootlog with noapic parameter passed to the kernel. It still
> freezes though. :(
> What I do exactly to make it freeze is after boot:
>
> In one console I do a ping -f to a box on my local network using the e100
> card. (integrated on the motherboard)
>
> In another console I copy a 2.5 GB file from my USB HDD to the IDE HDD in a
> while loop (or do a readcd from USB DVD Writer to a file on IDE HDD)

Can you try it whilst copying from say SCSI disk to IDE disk?

2006-01-19 19:46:03

by Gilles May

[permalink] [raw]
Subject: Re: SMP trouble

Zwane Mwaikambo wrote:

>On Thu, 19 Jan 2006, Gilles May wrote:
>
>
>
>>Attached the bootlog with noapic parameter passed to the kernel. It still
>>freezes though. :(
>>What I do exactly to make it freeze is after boot:
>>
>>In one console I do a ping -f to a box on my local network using the e100
>>card. (integrated on the motherboard)
>>
>>In another console I copy a 2.5 GB file from my USB HDD to the IDE HDD in a
>>while loop (or do a readcd from USB DVD Writer to a file on IDE HDD)
>>
>>
>
>Can you try it whilst copying from say SCSI disk to IDE disk?
>
>
I don't think it has something to do with the USB card, nor the HDD oder
the DVD writer connected to it..
Just to be sure I bought a new USB card with a different chip even,
hangs with both controllers..
Besides it freezes aswell if I do the ping and IDE to IDE copies and
listening music.. Looks like high
IO loads brings it down, no matter where it comes from..
The wierd part is that it's only with Linux SMP, not with UP, and no
problems like that on WindowsXP SP2..

This starts giving me serious headaches.. ;)

Regards, Gilles May

2006-01-20 02:23:01

by Zwane Mwaikambo

[permalink] [raw]
Subject: Re: SMP trouble

On Thu, 19 Jan 2006, Gilles May wrote:

> I don't think it has something to do with the USB card, nor the HDD oder the
> DVD writer connected to it..
> Just to be sure I bought a new USB card with a different chip even, hangs with
> both controllers..
> Besides it freezes aswell if I do the ping and IDE to IDE copies and listening
> music.. Looks like high
> IO loads brings it down, no matter where it comes from..
> The wierd part is that it's only with Linux SMP, not with UP, and no problems
> like that on WindowsXP SP2..
>
> This starts giving me serious headaches.. ;)

Trying to isolate things here, do you need the ping/network load to
trigger it? How about only network load?

2006-01-20 02:58:40

by Gilles May

[permalink] [raw]
Subject: Re: SMP trouble

Zwane Mwaikambo wrote:

>On Thu, 19 Jan 2006, Gilles May wrote:
>
>
>
>>I don't think it has something to do with the USB card, nor the HDD oder the
>>DVD writer connected to it..
>>Just to be sure I bought a new USB card with a different chip even, hangs with
>>both controllers..
>>Besides it freezes aswell if I do the ping and IDE to IDE copies and listening
>>music.. Looks like high
>>IO loads brings it down, no matter where it comes from..
>>The wierd part is that it's only with Linux SMP, not with UP, and no problems
>>like that on WindowsXP SP2..
>>
>>This starts giving me serious headaches.. ;)
>>
>>
>
>Trying to isolate things here, do you need the ping/network load to
>trigger it? How about only network load?
>
>
Hmm good question, I'll do further tests, but from my past experiences I
got the feeling that it's rather the sound that is needed to trigger the
freeze, not the network load.

A few lines from dmesg puzzle me too, like:

spurious 8259A interrupt: IRQ7. -> What is that, and why?

mtrr: your CPUs had inconsistent fixed MTRR settings
mtrr: probably your BIOS does not setup all CPUs.
mtrr: corrected configuration. -> Maybe not corrected correctly?

BIOS failed to enable PCI standards compliance, fixing this error. ->
Maybe not really fixed?

Thanks for your effort,
Gilles May

PS: Am I the only one with a K7D Master-L and problems like that?