tg3 fails with x86_64 (but works with i386)

tg3 fails with x86_64 (but works with i386)

[Frank Steiner]

Hi,

I've two 3com 3C996B-T network cards (lspci says it's
a Broadcom BCM5701 chip) in Asus A8V boards with an
AMD64 4000+ cpu. Booting a x86_64 version of the
2.6.12.2 kernel, the tg3 module complains

   tg3.c: v3.31 (June 8, 2005)
   tg3_test_dma() Write the buffer failed -19
   tg3: DMA engine test failed, aborting.

and no eth0 is available. However, when booting a 32bit
version of 2.6.12.2, the cards work well with the tg3
module. So it must be some problem in the x86_64
specific kernel code.

Any ideas what I could do? What further information can
I provide?

cu,
Frank

-- 
Dipl.-Inform. Frank Steiner   Web:  http://www.bio.ifi.lmu.de/~steiner/
Lehrstuhl f. Bioinformatik    Mail: http://www.bio.ifi.lmu.de/~steiner/m/
LMU, Amalienstr. 17           Phone: +49 89 2180-4049
80333 Muenchen, Germany       Fax:   +49 89 2180-99-4049
* Rekursion kann man erst verstehen, wenn man Rekursion verstanden hat. *

Re: tg3 fails with x86_64 (but works with i386)

[Frank Steiner]

Frank Steiner wrote

> Hi,
> 
> I've two 3com 3C996B-T network cards (lspci says it's
> a Broadcom BCM5701 chip) in Asus A8V boards with an
> AMD64 4000+ cpu. Booting a x86_64 version of the
> 2.6.12.2 kernel, the tg3 module complains
> 
>    tg3.c: v3.31 (June 8, 2005)
>    tg3_test_dma() Write the buffer failed -19
>    tg3: DMA engine test failed, aborting.


I could solve this: The A8V boards have (like all 64bit boards I guess?)
an option for memory remapping around the memory hole at 3.5GB. When
this remapping is activated, the tg3 driver fails with the error message
above. If the remapping is deactivated, the driver and the card work fine
(but linux can only work with 3.7 of the 4GB). But I need to have the
remapping deactivated anyway, because the fglrx driver won't work
either otherwise.

cu,
FRank


-- 
Dipl.-Inform. Frank Steiner   Web:  http://www.bio.ifi.lmu.de/~steiner/
Lehrstuhl f. Bioinformatik    Mail: http://www.bio.ifi.lmu.de/~steiner/m/
LMU, Amalienstr. 17           Phone: +49 89 2180-4049
80333 Muenchen, Germany       Fax:   +49 89 2180-99-4049
* Rekursion kann man erst verstehen, wenn man Rekursion verstanden hat. *

Re: tg3 fails with x86_64 (but works with i386)

[Arjan van de Ven]

On Tue, 2005-07-12 at 10:05 +0200, Frank Steiner wrote:
> Frank Steiner wrote
> 
> > Hi,
> > 
> > I've two 3com 3C996B-T network cards (lspci says it's
> > a Broadcom BCM5701 chip) in Asus A8V boards with an
> > AMD64 4000+ cpu. Booting a x86_64 version of the
> > 2.6.12.2 kernel, the tg3 module complains
> > 
> >    tg3.c: v3.31 (June 8, 2005)
> >    tg3_test_dma() Write the buffer failed -19
> >    tg3: DMA engine test failed, aborting.
> 
> 
> I could solve this: The A8V boards have (like all 64bit boards I guess?)
> an option for memory remapping around the memory hole at 3.5GB. When
> this remapping is activated, the tg3 driver fails with the error message
> above. If the remapping is deactivated, the driver and the card work fine
> (but linux can only work with 3.7 of the 4GB). But I need to have the
> remapping deactivated anyway, because the fglrx driver won't work
> either otherwise.


you were using a binary kernel module but didn't mention it in your
report? tsk tsk...

Re: tg3 fails with x86_64 (but works with i386)

[Frank Steiner]

Arjan van de Ven wrote

> you were using a binary kernel module but didn't mention it in your
> report? tsk tsk...

No, for testing fglrx that was a SuSE kernel, while I was using a
vanilla 2.6.12.2 kernel for debugging the tg3 module. Come on, first
thing I do for debugging a problem is using an untainted kernel :-)
I've learned this lesson from the LKML already :-)

-- 
Dipl.-Inform. Frank Steiner   Web:  http://www.bio.ifi.lmu.de/~steiner/
Lehrstuhl f. Bioinformatik    Mail: http://www.bio.ifi.lmu.de/~steiner/m/
LMU, Amalienstr. 17           Phone: +49 89 2180-4049
80333 Muenchen, Germany       Fax:   +49 89 2180-99-4049
* Rekursion kann man erst verstehen, wenn man Rekursion verstanden hat. *