paul@rasty.ph.unimelb.edu.au (Paul Gortmaker) writes:
> [ Ok, I just saw 102 is out, but haven't looked at it yet...]
>
> 2.1.101 (and 102 pre1) will reliably and consistently panic with:
>
> Releasing swap cache pages at 0xNNNNN
Nasty --- I haven't seen this one come up before. It may well be
related to the "writable dirty swap cache page" errors we are seeing.
The problem is that I introduced these new sanity checks while adding
new swap code, so it's hard to tell whether or not the problem was
already there but was being silently ignored beforehand.
> It happens about once every 45mins to an hour while trying to compile
> 2.0.34preX. (Needless to say, the disk light never goes out with 2.1.x
> on a 4MB box).
My trouble is that my VM test box (which I regularly bring down to 4 or
6 MB available memory at boot time) brings a lot of stuff in over NFS,
and fragmentation on 2.1 is so bad that NFS lockups throttle performance
in a big way. It just never sustains enough load to really saturate the
swapper. I've got a constantly increasing set of VM stress testers
which don't rely on NFS, however, and I'm hoping we can nail this stuff
soon.
> If it was a hardware flaw causing memory corruption, I'd expect
> semi-random oopses and not expect the exact same panic five times in a
> row.
Agreed. I'll look out for this one.
--Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu