On Wed, Jul 18, 2001 at 12:13:13AM -0500, Donald Lee wrote:
> I tried it with 1.5.1 stock kernel, and it works fine.  It negotiated
> 100 base T full duplex right out of the box.
> 
> For the docs: It *is* the rtk driver (not vr). relevant syslog:
> 
> Jul 18 00:09:48 charm /netbsd: rtk0 at pci0 dev 15 function 0: D-Link Systems DFE 530TX+
> Jul 18 00:09:48 charm /netbsd: rtk0: interrupting at irq 25
> Jul 18 00:09:48 charm /netbsd: rtk0: Ethernet address 00:50:ba:55:02:51
> Jul 18 00:09:48 charm /netbsd: ukphy0 at rtk0 phy 7: Generic IEEE 802.3u media interface
> Jul 18 00:09:49 charm /netbsd: ukphy0: OUI 0x000000, model 0x0000, rev. 0
> Jul 18 00:09:49 charm /netbsd: ukphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto

Yep. I've got two of those in the Intel box next to me. The fact
that they weren't attached in 1.5 (or, actually, 1.5T or so) is
the reason they are in 1.5.1.

> I'll have to do
> some benchmarking to see if it can drive a full 100 Mbits, but
> I see no console messages or other "issues".  Cool!

Don't get too excited yet. While you can probably get close to 100
Mbits, that card was cheap because the RealTek chip is cheap. I've
not been able to make mine overflow yet or anything, but they're not
exactly top-shelf quality, as I understand it.

> P.S. My asante card
> with the de driver can't drive 100 Mbits in 1.5 without lots of
> hiccups and whining.  It also does not do the auto 100 with the
> switch reliably.  I'll have to fire it up with 1.5.1 and see if this
> is all fixed up...(!)

There may be some stuff that's fixed things up. Have a look at
http://cvsweb.netbsd.org/bsdweb.cgi/syssrc/sys/dev/pci/if_de.c. None
of that screams "performance improvement" to me, though.

-- 
       ~ g r @ eclipsed.net