eepro100 broken on ARM by 2.6.11-rc1 changes

eepro100 broken on ARM by 2.6.11-rc1 changes

[Lennert Buytenhek]

Hi,

eepro100 seems to have stopped working between 2.6.10 and 2.6.11-rc2
on an embedded ARM platform.  In 2.6.10 I would see this:

eepro100.c:v1.09j-t 9/29/99 Donald Becker http://www.scyld.com/network/eepro100.html
eepro100.c: $Revision: 1.36 $ 2000/11/17 Modified by Andrey V. Savochkin <saw@saw.sw.com.sg> and others
eth0: 0000:01:00.0, 00:00:50:14:56:11, IRQ 40.
  Receiver lock-up bug exists -- enabling work-around.
  Board assembly 698247-003, Physical connectors present: RJ45
  Primary interface chip i82555 PHY #1.
  General self-test: passed.
  Serial sub-system self-test: passed.
  Internal registers self-test: passed.
  ROM checksum self-test: passed (0x8b51f404).
  Receiver lock-up workaround activated.
DLCI driver v0.35, 4 Jan 1997, mike.mclagan@linux.org.
[...]

But now I just see this:

eepro100.c:v1.09j-t 9/29/99 Donald Becker http://www.scyld.com/network/eepro100.html
eepro100.c: $Revision: 1.36 $ 2000/11/17 Modified by Andrey V. Savochkin <saw@saw.sw.com.sg> and others
DLCI driver v0.35, 4 Jan 1997, mike.mclagan@linux.org.
[...]

Copying the eepro100 driver from 2.6.10 into 2.6.11-rc4 makes things
work again on 2.6.11-rc4.

Anyone else seeing this?


cheers,
Lennert

Re: eepro100 broken on ARM by 2.6.11-rc1 changes

[David S. Miller]

On Tue, 15 Feb 2005 22:10:20 +0100
Lennert Buytenhek <buytenh@wantstofly.org> wrote:

> eepro100 seems to have stopped working between 2.6.10 and 2.6.11-rc2
> on an embedded ARM platform.  In 2.6.10 I would see this:

Please check to make sure ARM implements pci_iomap(), ioread*()/iowrite*()
and friends correctly.  eepro100 is one of the few net drivers that makes
use of this newer API, and the conversion to using this new API occured
post 2.6.10

Re: eepro100 broken on ARM by 2.6.11-rc1 changes

[Lennert Buytenhek]

On Tue, Feb 15, 2005 at 01:20:47PM -0800, David S. Miller wrote:

> > eepro100 seems to have stopped working between 2.6.10 and 2.6.11-rc2
> > on an embedded ARM platform.  In 2.6.10 I would see this:
> 
> Please check to make sure ARM implements pci_iomap(),
> ioread*()/iowrite*() and friends correctly.

It doesn't.  The generic implementation (lib/iomap.c) assumes that
cookies below 0x10000 are I/O space, and this doesn't hold on the ARM
in general.

In my particular case, the e100 has its I/O space start at 0x100000.

Discussion taken to linux-arm-kernel.


thanks,
Lennert