[24] in netbsd-help mailing list archive
Re: 3C503
daemon@ATHENA.MIT.EDU (Greg Hudson)
Sun Jan 22 00:22:34 1995
To: yoav@MIT.EDU
Cc: ghudson@MIT.EDU, sdfranks@MIT.EDU, netbsd-help@MIT.EDU
In-Reply-To: Your message of "Sat, 21 Jan 1995 21:52:29 EST."
<199501220252.VAA22942@I-See-Everything-Twice.MIT.EDU>
Date: Sun, 22 Jan 1995 00:22:21 EST
From: Greg Hudson <ghudson@MIT.EDU>
> Any guess as to what happens if you tell the kernel exactly where
> the card is? (i.e., give it the IRQ and address etc?).. would it
> just fail to find the card? or would there be a better chance of
> success?
I would guess that the problem Stephen is complaining most about is
not the autodetection problem (which I've had workarounds for all
along), but the problem where the driver stops receiving packets. I
and mycroft fixed the autodetction problem, and before it was fixed my
installation script would install a kernel on machines with 3c509
cards that only included the 3c509 driver, which was a successful
workaround.
Unfortunately, we don't yet have a solution to the falling-off-the-net
problem. The Athena kernels include a patch from Christos that seems
to help a bit, but glacier still falls off the net sometimes when
there's high traffic. This seems to be dependent on the particular
chipset in the card, since it doesn't occur on granola using either
BNC or AUI. I'm going to ask on Monday if anyone minds if we
temporarily switch the 3c509 cards in glacier and granola so that
mycroft can try to debug the problem. (Since I need glacier working
for the class, this won't actually happen for a week, of course.)
