Anton Vorontsov [off-list ref] wrote on 26/03/2009 19:05:43:

On Thu, Mar 26, 2009 at 05:43:25PM +0100, Joakim Tjernlund wrote:
[...]

quoted

quoted

bd == ugeth->confBd[txQ]
and
!in_be32((u32 __iomem *)(bd+4))

Are not equivalent wrt. speed. MMIO accessors should be rather
slow comparing to normal memory.

Yes, I know. I did it this way because I something broke under stress
when ugeth->confBd[txQ] instead. The ucc_geth_tx() and 
ucc_geth_start_xmit()
gets out of sync somehow.

Would be great if you could investigate it more. Maybe there is
a serious bug somewhere, or maybe you're introducing another
(yet hidden) bug...

I have spent some time on it and didn't see how to do it. I don't think
the current method hides any bugs. I have used this method on 8xx for
many years and it works well.

Actually the current method might work without the 
spin_lock_irq(&ugeth->lock)/spin_unlock_irq(&ugeth->lock)
in ucc_geth_start_xmit(). I can't break it, can you?

 Jocke