bugfix patch to arch/ppc/8260_io/fcc_enet.c

[Dan Malek]

David Schleef wrote:

> Check again.  The unsigned arithmetic works correctly.

OK.


>   (starting with a full queue)

This one can't happen because of the race conditions are avoided
by the use of cep->lock.

> The other (and more likely) is:
>
>   - fcc_enet_start_xmit() fills up the TX ring, thus cep->skb_cur
>     == cep->skb->dirty.


What driver are you using?  That line of code hasn't been in a driver
for a very long time.  This was a very obvious bug that has been
corrected long ago.  You should be using the FSM Labs linuxppc_2_5
bitkeeper tree for any of the 8xx/82xx development.  There are lots
of things that were placed in the 2_5 tree when the 2.4 tree was frozen.
We are now trying to merge them back, but there are still some pieces
left to go.


	-- Dan

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/