Jason R Thorpe wrote:
> 
> On Sat, Jan 19, 2002 at 04:52:36AM +0900, Izumi Tsutsui wrote:
> 
>  > PCI hme and nsphy on my Ultra5 sometimes fails autonegotiation
>  > between switching hub during boot. The attached patch from OpenBSD
>  > seems to fix the problem, but does anyone else see the similar problem
>  > on U5 and PCI hme?
> 
> Uh.  This seems totally bogus.  The ANAR is set to the capabilities
> from the BMSR in mii_phy_auto().
> 
> I guess the real problem is that the MII code only sets "autonegotiate" if
> BMCR_AUTOEN is not already set.  Perhaps it should also check to see if
> the current ANAR is different from the ANAR that would be sent if we re-
> started autonegotiation?
> 

FWIW I've seen autonegotiate/autodetect problems with the 'hme' hardware
under Solaris.  You need a 10M hub, 100M hub and 10/100 autonegotiate
hub and maybe something that will do 100M FDX.  Move the cable around
and on one transition you lose the link.  Unfortunatley I've forgotten
which transition :-)  I didn't try to fix the driver at the time though.

I've never enabled FDX in any drivers I've written - too hard to keep
the MAC in sync with the phy...

	David