Re: via rhine bug? (timeouts and resets)

[Roger Luethi]

On Tue, 11 Feb 2003 21:31:10 +0100, Henrik Persson wrote:
> And look what came up when I stressed the net a bit.. Worked fine at
> first, though.. But I guess that depends on other things.. Sunset and
> all.. Heh ;)

It's pretty easy to trigger, actually. Just have some heavy traffic going
in _and_ out, e.g. netcat blowing iso images both ways. It will last a
couple of seconds at most.

> Well.. dmesg attached..

# eth0: Setting full-duplex based on MII #1 link partner capability of 45e1.
# eth0: Done via_rhine_open(), status 0c1a MII status: 782d.
# eth0: no IPv6 routers present
# eth0: Transmit error, Tx status 00008800.
# eth0: Transmitter underrun, Tx threshold now 40.
# eth0: Transmit error, Tx status 00008800.
# eth0: Transmitter underrun, Tx threshold now 60.
# eth0: Transmit error, Tx status 00008800.
# eth0: Transmitter underrun, Tx threshold now 80.
# Gotcha: 0x2 0x8 0x0
# Gotcha: 0x1 0x8 0x0
# Gotcha: 0x1 0x8 0x0
# NETDEV WATCHDOG: eth0: transmit timed out
# eth0: Transmit timed out, status 0000, PHY status 782d, resetting...
# eth0: Reset succeeded.

As expected. Now comes the punch line: I don't know how to fix this. I
locked my machine up solid a couple of times trying. It seems that
particular flag doesn't want to be cleared. Of course I could simply reset
the chip, but that's a) less than elegant and b) would make cleaning up the
force_media mess kind of urgent. And doing that properly is a rather
non-trivial change. Also, I need to investigate the implications for
Rhine-III and have somebody test Rhine-I.

Thanks for the logs, though; at least now I know that many more would hit
that problem if they weren't using a driver that breaks down way earlier.
If the problem bothers you I can send you a dirty hack. I need to whack
some registers before writing a proper fix, and I don't know when that will
happen.

Roger