Am Montag 10 September 2007 schrieb Herbert Xu:

Hans-J??rgen Koch [off-list ref] wrote:

quoted

The following patch fixes it. Tested on an AT91SAM9263-EK board, kernel
2.6.23-rc4 and -rc3-mm1.

Could you please audit all instances of physdev->lock and add
_bh where necessary?  I can see that at least phys_stop also
needs the _bh.

I think the patch does all that's necessary. At least, there're no error
messages in the logs anymore. I didn't check if there's an error on
unload, though.

We should also consider whether it makes sense to move the
timer into a work queue.

Somebody who's more involved in that PHY stuff should probably do that.
I'm on holidays, sitting in a hotel room, and will probably not have the
time to have a deeper look into that.

Thanks,
Hans