On Mon, Nov 26, 2001, Ken Key <key@cips.nokia.com> wrote:

>Thanks for the info, Ben.  The ATA4 wd0 lost interrupt problem was fatal 
>for me this weekend.  After it first happened, it happened with any 
>substantial disk activity.  I couldn't even CVS down the kernel sources 
>to try building a new kernel without locking up.  I'll try again later 
>when I have more time, but NetBSD isn't my primary purpose on this
>box yet.

Apple recently updated significantly the ATA controller drivers in
Darwin sources. The new timing code seem to be robust and finally
match the timings used by MacOS 9. I've re-implemented the linux
driver to use similar timing calculations and have no problem so
far.

>Possibly interesting side-note on the gm0.  When I boot plugged into a
>NetGear FS108, the gm0 shows Media: Ethernet autoselect (none) and
>is running 10M.  If I drop link (unplug back of PB), it comes back
>at 100baseTX full-duplex.

That's interesting, I noticed similar problem with my linux driver,
the link occasionally comes up at 10bps. Unplugging/plugging back
or just restarting autoneg (via an ioctl to the driver) is usually
enough to get it back full speed. Maybe some weirdness with that new
PHY autoneg process, I'll investigate on my side and let you know if
I find something.

Ben.