Re: panic: vm_page_t 0xc1c92660 phys_addr mismatch 0000000000b6d000 000000002bb6d405

On Tue, 2004-07-06 at 15:54, Doug White wrote:
> On Mon, 5 Jul 2004, eta wrote:
> 
> > Got a panic on the new little server I'm testing for use at our school.
> >
> > panic: vm_page_t 0xc1c92660 phys_addr mismatch 0000000000b6d000
> > 000000002bb6d405.
> >
> > Attached is some gdb and dmesg output.  It's -current from 2004-06-25,
> > and I was in the middle of a gnome-upgrade.sh.  The kernel config is
> > generic minus WITNESS plus COMPAT_LINUX.
> >
> > Anyone seen this?
> 
> Unfortunately the backtrace is corrupted by the secondary panic.  This
> might be bad memory, however; the pattern is kind of interesting. Can't
> say I've seen that on my dual Opteron, but I haven't built -current on it
> in a while.

Yeah, what was strange was that it was two of the same type of panic
while running the same task in short succession.  That's why I decided
to send it instead of sitting on it longer.  Also, it's been stable when
running the buildworlds and buildkernels I've done outside of the
chroot, but then those are much shorter than a gnome-upgrade.sh.  There
haven't been any strange signals when building things, which I would
think would happen more often in a bad memory situation.

I got another panic today running a 20040705 in a different part of the
kernel (taking a fault while doing pmap stuff and recursing on a lockmgr
lock).  So maybe it is a hardware issue.  Will have to stress the
machine some more in different ways.

-- 
Eric Anholt                                eta_at_lclark.edu          
http://people.freebsd.org/~anholt/         anholt_at_FreeBSD.org