On 10 Sep 2004 09:33:35 -0400
jamal <hadi@xxxxxxxxxx> wrote:
> On Wed, 2004-09-08 at 16:47, David S. Miller wrote:
> > We are merely moving the sch_generic.c locking logic into the
> > drivers. The behavior is entirely equivalent except that one
> > level of unnecessary locking has been removed.
> > I think his change is valid, will not break existing drivers (as
> > you mentioned as well Jamal), and works well for the cases he has
> > shown patches of. So I'm going to apply his patch.
> > BTW, if we are really concerned about some existing driver returning
> > -1 from hard_start_xmit() without the new feature flag being enabled,
> > we can test for that and log a debugging message if it happens.
> I am not 100% happy but let me do some testing on it. Would the best
> image be the latest bk snapshot?
Yes, Andi's code plus e1000 and tg3 driver uses are in current
2.6.x BK. Just grep for NETIF_F_LLTX in drivers/net/tg3.c to
be sure you've got it in your tree.