[tulip] Tulip and PowerPC endian issue?

Jim Rowe rowejames@acmsystems.com
Thu Jan 30 15:16:01 2003


On Thu, 2003-01-30 at 11:44, Donald Becker wrote:
> On 30 Jan 2003, Jim Rowe wrote:
> > Excellent! That was the trick. Thanks for your help Donald. 
> 
> OK, just to summarize:
> 
>   There is a bug in the kernel's tulip driver that byte-swaps the
>   station address when used with the Comet chip.
	
	Yes, the tulip driver in kernel 2.4.19 will byte swap the hardware
address when running on a CardBus interface on the PPC. It may swap
other data, but for sure it swaps the MAC.
> 
>   The bug may manifest itself as the ability to ping but not communicate
>   with TCP or UDP.  The exact symptoms depend on the station address and
>   the local router configuration.

	I was not able to ping or communicate with TCP or UDP. 
> 
>   The driver from Scyld works correctly, but does require that the
>   interface go through a hard-power-off if the flawed kernel driver is
>   run first.

	Yes. If the tulip driver in the kernel tree is loaded first power needs
to be shut off, and the Scyld driver loaded. The MAC address is then
reported correctly.

Note I still haven't had any success pinging or communicating with the
cardbus interface, however the information that is reported by
/var/log/messages and ifconfig now looks correct. 

Has this driver been known to work correctly with a CardBus interface on
the PPC? I'm working with some custom hardware and it would be helpful
to know if the driver has been tested on such a setup. 

Thanks,
Jim

> Please confirm that this is the case.

-- 
Jim Rowe
Advanced CounterMeasure Systems
Email: jrowe@acmsystems.com