A while ago I mentioned in passing that I'd been plagued by system lock-ups
(sometimes with the flashing keyboard lights that I believe indicate a
kernel panic, most of the time not) that I thought were caused by faulty
hardware, though I couldn't really figure out which piece of hardware.

Well, just as an update, it now seems that the problem is hardware and
software related, in the sense that some subset of my particular
mobo/chipset/cpu combo has a problem with the kernel's implementation of
cpufreq.  The technical explanations go way over my head, but from I've been
able to glean from numerous bug reports in various distros and Linux kernel
lists, the problem is with a flaw in certain chipsets that is exposed when
the powernow-k8 kernel module implements cpu frequency stepping.  This seems
to be an issue only with SMP kernels on amd-64 systems, and it doesn't seem
to matter whether you are running a 32- or 64-bit kernel or userland.  I
have used both and had the problem with both.  VIA chipsets seem to be
particularly vulnerable.  I think others are affected as well, but it might
just be other motherboards with VIA chipsets.  I have an ASUS-A8V mobo with
a VIA chipset.

As usual when trying to get to the root of a puzzling problem, there are a
lot of sidetracks that ultimately don't lead anywhere, and I still haven't
sorted them all out.  (It's especially difficult when you don't really know
what you're Googling for, beyond "random system freeze.")  Some people
thought various video cards and proprietary drivers were involved, but that
doesn't seem to be the case.  Some people reported that disabling AMD's
"Cool 'n Quiet" feature in the BIOS solved the problem, others said that
didn't solve the problem.  I have not tried that method yet.  Some people
said updating the BIOS helped, others said they'd done that and the problem
persisted.  Some people said the problem was with powernowd, but I never had
that daemon running and I still had system lock-ups.

What I did was prevent the kernel module from loading, about 2 weeks ago,
and I have not had a freeze since.  I hesitate to say, definitively, that
the problem is solved, because the last thing I tried worked for a few weeks
and then the problem returned.  So time will tell.  Still, it's been 10 days
since my last reboot -- I can't express what a relief that is after having
the machine lock-up on me as often as two or three times a day, seemingly at
random, for weeks on end.

I'm not sure yet if any BSD OS's are affected, but it doesn't seem to affect
Windows, or at least that's what a few people who dual-boot said.  My
machine is definitely noisier that it was with the module loaded, so I'm
probably sucking down more electricity and running hotter.  I'll keep
looking for other possible workarounds that don't require disabling cpufreq
altogether.  I might give FreeBSD another shot, after I look into whether
this is an issue that affects the BSD kernel.  I don't know when, or if,
Linux kernel developers are going to come up with a solution.

Michael M.

-- 
"Rightful liberty is unobstructed action according to our will within limits
drawn around us by the equal rights of others. I do not add 'within the
limits of the law' because law is often but the tyrant's will, and always so
when it violates the rights of the individual." --Thomas Jefferson
_______________________________________________
PLUG mailing list
[email protected]
http://lists.pdxlinux.org/mailman/listinfo/plug

Reply via email to