On Wed, Jun 30, 2004 at 08:39:59PM +0200, Martin Husemann wrote:
> On Wed, Jun 30, 2004 at 12:39:56PM -0400, der Mouse wrote:
> > Neither am I, but I _have_ seen networking hardware work at 100 but not
> > at 10.  (It was a 100/1000 gigabit interface.)
> 
> I have seen hme fail at 10baseT and work at 100baseTX-fdx ;-)
> 
> I use my QFE as a bridge - which in our current implementation only works
> correctly for IFF_SIMPLEX interfaces. Hmes apparently have the simplex
> property only if in full-duplex mode. I had to hard-wire all 10baseT ports
> of my QFE to full-duplex to make bridge0 happy.

Some experiments with the sun FEPS chipset (sbus Fast Ethernet, Parrallel,
Scsi) I did many years ago showed that it received its own transmitions
at 10M HDX, but not at 100M HDX).  To make things work I made the MAC
part discard packets to its own address, and the higher layer generate the
inbound traffic for transmits.

From what I remember there was also an undocumented 'feature' that the
receive buffer had to be (about) 32 bytes longer than the longest frame
you wanted to receive.

I donn't know the netbsd or sun driver names for that chipset though!

	David

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