[2720] in Release_7.7_team
Re: Athena 8.4.23/24/25 patch release due tomorrow evening
daemon@ATHENA.MIT.EDU (Greg Hudson)
Mon Apr 23 13:01:55 2001
Message-Id: <200104231701.NAA06562@egyptian-gods.MIT.EDU>
To: Tom Cavin <tec@ai.mit.edu>
cc: release-team@MIT.EDU
In-Reply-To: Your message of "Mon, 23 Apr 2001 12:53:17 EDT."
<15076.24061.351213.948899@vivaldi.mit.edu>
Date: Mon, 23 Apr 2001 13:01:46 -0400
From: Greg Hudson <ghudson@MIT.EDU>
> Does the Linux sysinfo change have any effect on the Pentium IV /
> i?86 update problem? Or the rpm changes?
I'm afraid not. The problem is in the kernel. (Or rather, the
problem is with the Pentium IV itself--Intel seems to have changed CPU
family numbering conventions for no reason, and the kernel needs a
workaround for that. Or that's what the Linux kernel people said,
anyway.)
> If so, great. If not, do you have any partial solutions or interrim
> fixes for it?
I looked into this a bit last Wednesday but was unsuccessful (and
didn't wind up sending you mail about my lack of success). Basically,
you need kernel version 2.2.18 or better to recognize your processor
version. Red Hat has released update RPMs for kernel version 2.2.19,
and ordinarily you could just locally upgrade your kernel and your
problems would be solved. Unfortunately, AFS doesn't currently work
with kernel version 2.2.19, so you can't do that.
I also looked for a way to override what the kernel reports via uname
(I figured maybe something under /proc/sys), but there doesn't seem to
be any such knob.
You can get rpm to ignore the architecture mismatch, using something
like "rpm -U --ignorearch newrpm". update_ws (and its helper program,
rpmupdate) don't currently have any switches to do that, though.
If you feel up to it, you could do "update_ws -n", use rpm -U
--ignorearch to manually update to the newest version of the listed
RPMs (from /afs/athena.mit.edu/system/rhlinux/athena-8.4/free/RPMS and
/afs/athena.mit.edu/system/rhlinux/redhat-6.2-updates/RPMS), and then
add a line to /etc/athena/version in the proper format. That would
have the same effect as an update (assuming there's no kernel upgrade,
which I believe is the case). I realize that's somewhat complicated,
though.
