[3209] in Release_7.7_team
Re: Progress on 3com bug!
daemon@ATHENA.MIT.EDU (Bill Cattey)
Thu Mar 28 18:55:54 2002
From: Bill Cattey <wdc@MIT.EDU>
To: Bill Cattey <wdc@mit.edu>
Cc: release-team@mit.edu, ops@mit.edu, hotline@mit.edu, alex_c@mit.edu,
csnowden@mit.edu
In-Reply-To: <1017353863.31009.163.camel@m1-142-14.mit.edu>
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
Date: 28 Mar 2002 18:55:50 -0500
Message-Id: <1017359751.1197.13.camel@m2-225-9.mit.edu>
Mime-Version: 1.0
Lou and I went to building 2 and I refined my understanding of the
situation.
The reason why machines got slow on building 2 was that we were setting
EEPROM value 0x000d to 0x8010 when it should be 0x0010. This is the
full-duplex issue I was studying yesterday.
The CORRECT settings for systems (as verified by my walking to every
system in 1-142 and 2-225) is:
use the new vortex-diag:
vortex-diag -x 2 -w
then reset the chip. (Note there is a program, mii-tool that seems
to be an early version of mii-diag. If you have a dead system you can
use it!)
mii-tool -R
Then you MUST Reboot.
Test to confirm that the machine is happy and fast.
----
Simpler procedure for recovering dead systems:
boot: linux single
insmod 3c59x
etc/init.d/network start
mii-tool -R