Subject: Re: wi performance hack
To: David Young <firstname.lastname@example.org>
From: Charles M. Hannum <email@example.com>
Date: 07/20/2004 00:18:39
On Monday 19 July 2004 22:37, David Young wrote:
> On Mon, Jul 19, 2004 at 04:28:31PM -0500, David Young wrote:
> > On Mon, Jul 19, 2004 at 07:56:44PM +0000, Charles M. Hannum wrote:
> > > On Monday 19 July 2004 18:52, Holger Weiss wrote:
> > > > * Charles M. Hannum <firstname.lastname@example.org> [2004-07-19 16:49]:
> > > > > Here's a new patch that adds some more diagnostics. Give it a
> > > > > whirl and let me know what it outputs...
> > > >
> > > > wi0: bad alloc 1f7 != 1f9, alloc 2 queue 1 start 1 queued 0 started 2
> > > > wi0: bad alloc 1f8 != 1f9, alloc 2 queue 2 start 2 queued 0 started 3
> > > > wi0: bad alloc 1f9 != 1f7, alloc 0 queue 0 start 0 queued 0 started 3
> > >
> > > Hm, I have the suspicion that the card is simply not transmitting
> > > packets in the order we told it to. What happens with the following
> > > patch instead?
> > Broadcast/multicast packets are sometimes transmitted out of order.
> What I meant was, wi firmware puts unicast packets and broadcast/multicast
> packets on different queues. So a unicast packet that is queued after
> a multicast packet may nevertheless transmit first.
I can't seem to cause this to happen on either an Intersil or Symbol client.
At any rate, getting the FIDs back in a different order really shouldn't
matter to us. "It's just a number." As long as the card doesn't give us
duplicate FIDs, we're fine -- and I don't think it's worth bothering to even
check; the Agere driver doesn't.