On Wed, 2007-04-11 at 21:55 -0700, David Miller wrote: > From: David Miller <davem@xxxxxxxxxxxxx> > Date: Tue, 10 Apr 2007 14:04:58 -0700 (PDT) > > > From: BERTRAND Joël <joel.bertrand@xxxxxxxxxxx> > > Date: Tue, 20 Mar 2007 21:53:26 +0100 > > > > > I have tried the last 2.6.21-rc4. Same bug. > > > > The good news is that I can reproduce this problem now on my > > Ultra2, I'll try to figure out what's wrong. > > > > Running "dbench 32" a few time is enough to trigger it. > > Ok, I think I might have killed at least the bug I was able to hit. > > I make no promises that this will fix the issues everyone else was > seeing, but let's cross our fingers :-))) > > Please give the following patch some testing if you can. And please > be careful, messing with the DMA mapping code can do things like eat > your disk :-) > > If all goes well I'll push this bug fix around, and I have a secret > treat for your poor sparc64 SBUS users once we get past this. > > Thanks! > It's impossible to tell when a bug like this is fixed, of course, but I've been running U2(2x400)+kernel-2.6.20-r5+<this patch> since 26th June, and I have not seen a failure. Now, I have not subjected the system to any particular stress tests, but before, the daily updatedb/makewhatis/update-eix cycle plus the daily (gentoo) --sync operations were enough to kill this system pretty quickly. So right now, at least, I think you've probably killed the bug. Regards, Ferris -- Ferris McCormick (P44646, MI) <fmccor@xxxxxxxxxx> Developer, Gentoo Linux (Devrel, Sparc)
Attachment:
signature.asc
Description: This is a digitally signed message part
