Subject: Re: Intel Gigabit driver wm(4) fiber phy
To: Matthias Scheler <firstname.lastname@example.org>
From: Tom Ivar Helbekkmo <email@example.com>
Date: 09/13/2005 17:21:25
firstname.lastname@example.org (Matthias Scheler) writes:
> In article <email@example.com>,
> Tom Ivar Helbekkmo <firstname.lastname@example.org> writes:
>> I haven't tried to get it to do that, but I have another weird problem
>> with wm on my IBM T42p laptop: when it boots, the ethernet interface
>> (which is hardcoded to 100baseTX FDX because it takes so long to do an
>> autonegotiation against our gigabit capable switch) fails to come up.
> Which is no surprise. You cannot turn off auto negotiation on the host
> while leaving it turned on on the switch and expect that to work. You
> might be lucky and it will with some setups. But there is absolutely
> no guarantee.
Possibly so -- in my experience, it will work, except that you can't
know, without knowing the characteristics of the particular type of
switch you've got, whether you'll get half or full duplex.
Anyway, I should maybe have pointed out that I've got the laptop *and*
the switch configured for 100baseTX FDX. That works just fine, except
that the laptop's wm interface is dead (lamps dark) until I pull and
reinsert the cable, *after* having ifconfig'ed it up.
What happens is: when the laptop is powered on, the interface leds
light up, and it gets carrier going with the switch. Then, when the
NetBSD kernel recognizes the interface, the light goes out, and
carrier drops. The ifconfig doesn't help, but pulling and reinserting
the cable after the ifconfig lights up leds and establishes carrier.
> About the autonegotiation time: Ask the admin to deactive spanning
> tree support on at least the ports used by host (and not other
It's not the 30 second spanning tree delay, but a two minute gigabit
Don't ascribe to stupidity what can be adequately explained by ignorance.