Red Hat Bugzilla – Bug 65104
RH 7.2 install on K7 motherboards dies on reboot with GPF
Last modified: 2007-03-26 23:53:21 EDT
Description of Problem:
Installing on a K7 with RH 7.2. Everything goes fine until you reboot. The
kernel that is installed does a rdmsr on MSRs that do not exist on the K7.
The PIII, for non-existent MSRs, will actually return a bad value, contrary
to what the PII did and what the PIII should do. The K7, on the other hand,
will take a GPF.
The result is that a user can install RH 7.2 on a K7 box, will watch it all
work, will reboot, and see his kernel come up and get a GPF and die horribly.
We have worked around this at LANL but I was reminded of it by a user from
another site just today. David Woodhouse told me to report it.
Version-Release number of selected component (if applicable):
Buy an off-the-shelf system with a (e.g.) PCCHIPS M810LMR motherboard and
install with RH 7.2. Wait for the reboot after install. BOOM!
Steps to Reproduce:
Ron, this is not the way to report it, I am sorry to say.
Surely, as LinuxBIOS leader you understand that we'd like
to see the oops trace. The approved way to capture it is
to use a serial console, but some people do get creative.
I saw them attaching JPEGs of screen images...
Actually, cancel that - the description is good, but I am not
sufficiently familar with that part of the code to cook up a fix
right away. And I still like to see the trace so that I know
what file to open in vi :)
Did you try to poke linux-kernel on your own with this problem?
Sorry for the lack of an oops trace. I don't have serial console (this is the
first boot after an install) and it never occured to me to take a picture.
Usually that text is borderline-readable.
I think the issue is that you are shipping a PIII kernel with the PIII serial #
feature enabled in .config. When the K7 hits that code, you get the GPF as there
is no such MSR in the K7.
I'll have to see if this is fixed in 7.3 when I get time.