Subject: Re: chapeter II: serious problem with hme0
To: Rafal Boni <rafal@attbi.com>
From: David Laight <david@l8s.co.uk>
List: port-sparc64
Date: 02/09/2003 18:38:12
> -> 
> -> The switch is:
> -> 
> -> D-Link DSS-8+ 10/100 switch.
> 
> Curse, mumble, curse!  That's the same one the on-board hme in my U5 
> *doesn't* work with!  I think I also had a similar issue with another
> unmanaged DLink 16 or 24 port switch at work.. What phy does your hme
> use?

A thought that might be relevant...

It is possible that the switch isn't doing auto-negotiation.
This means that the ANAx MII registers won't be setup up with
information about the far end.

The switch could be doing 'autodetect' instead.  This is based
on whether you receive single link test pulses or short bursts
of pulses - the latter indicating support for 100M HDX.

I did look at the PHY register seg a few years ago (playing with
a driver a suns sbus fast ethernet card (not sure what solaris
calls it though)  I couldn't see an easy method of determining
what  the transceiver had negotiated.  I just disabled FDX in
the advertise register and set the MAC to HDX.

(The only gotcha then is that the card will return tx traffic
(to its own address) at 10M but not at 100M....)

	David

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