[3205] in Release_7.7_team
Re: some GX150 NIC's may have been stopped completely by the recent
daemon@ATHENA.MIT.EDU (Bill Cattey)
Thu Mar 28 03:22:22 2002
Date: Thu, 28 Mar 2002 03:22:20 -0500 (EST)
From: Bill Cattey <wdc@MIT.EDU>
To: Alex Coventry <alex_c@mit.edu>
cc: <hotline@mit.edu>, <release-team@mit.edu>
In-Reply-To: <1017207483.5247.307.camel@tokata.mit.edu>
Message-ID: <Pine.LNX.4.30L.0203280315330.23766-100000@tokata.mit.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Lou Isgur and I did extensive testing of GX150's in 1-142 today.
I have come to the conclusion that "We thought we understood
the network problem. We dont."
A temporary work around:
Reboot the system single user, and start the network driver forcing
10base-T as follows starting at the "boot: " prompt type:
linux single
insmod 3c59x options=2
^D
(The <ctrl>D will finish bringing up the Athena system. The machine
will be happy until the next reboot at which point it will again fail
to autonegotiate.)
The new EEPROM values CLEARLY undid a problem that was causing
the 3c59x driver to get confused and find no media interfaces, and so
a botched 100Mb/sec negotiation was quite predictable.
We've turned a situation of "all systems mis-negotiate to a slow network
in a completely predictable way" to one of "systems randomly autonegotiate
10baseT and work, 100baseT-FullDuplex and work, 100baseT-HalfDuplex and
get slow network, or some other thing that takes them off the net
entirely."
I am working VERY hard to understand this problem. So far I am STUMPED.
-wdc