Re: Fix for kern/20700

[David Laight <david%l8s.co.uk@localhost>]

On Thu, Dec 01, 2011 at 03:25:51PM -0600, David Young wrote:
> On Thu, Dec 01, 2011 at 01:15:54PM -0800, Brian Buhrow wrote:
> >     Hello. I've written a patch which addresses this bug and submitted it 
> > to the pr.
> > If no one objects, I'll commit this patch in a few days and request a
> > pullup to the 5.x branch.  If there are objections, I'm happy to discuss
> > and/or incorporate a better fix.  This fix works quite well in my case.
> >     If you're interested in a greater analysis and discussion of
> > thisissue, see:
> > http://mail-index.NetBSD.org/tech-kern/2011/11/30/msg012025.html
> 
> In the kern/20700 discussion, Manuel Bouyer mentions that 'media
> 1000baseT' behaves like 'media auto', so NetBSD may negotiate '1000baseT
> full-duplex' or '100baseTX full-duplex'.  The latter result would be a
> very big surprise.  Maybe someone can help fix that before the patch is
> committed.

At a guess it isn't setting the ANAR (autonegotiate advertise register)
correctly.

I've long been an advocate of constraining the 'media' by limiting
the ANAR values - rather then fixing the mode.
IIRC the ANAR register is one of the standard 8 registers (if the
original definition has enough bits for Ge and 10Ge).

Of course, if you connect to a 10M hub or a 10/100M autodetecting
hub the phy will connect anyway. This can be borked because the
MAC driver has difficulty determining the actual mode the PHY is
using (not in a standard register!).

        David

-- 
David Laight: david%l8s.co.uk@localhost