> > (Linux version 2.2.10 (root@blabla) (gcc version egcs-2.91.66 19990314
> > (egcs-1.1.2 release)) #7 Mit Jun 23 17:00:19 MEST 1999 )
> >
> > ...with the same kernel version BUT with quota support on, I got
> >
> > Jun 23 12:47:01 blabla kernel: 08:12: rw=0, want=1484339577, limit=3076447
> > Jun 23 12:47:02 blabla kernel: attempt to access beyond end of device
> > Jun 23 12:47:02 blabla kernel: 08:12: rw=0, want=912684654, limit=3076447
>
> Bingo. Thats the sort of report that helps a lot. Do you have quotas enabled
> or just compiled in ?
Are you still trying to track this down? If not, ignore the rest of this
email. :)
We have about 30 heavily-loaded web-hosting-company related servers here
now running 2.2.10-ac5 ranging from single P2 233 to dual P3 450s, all
with aic7xxx SCSI (several different revisions of chipsets on-board), and
quota enabled on about 10 of these machines (two of which are UP while the
rest are SMP). The two that are UP are actually compiled as SMP out of
laziness (they're just the customer webserver machines), but the other UP
machines are compiled with UP kernels (no UP-compiled UP machines without
quota, though).
Everything compiled with plain old GCC-2.7.2.3, binutils 2.9.1.0.15.
No signs of corruption anywhere; however, I just noticed that I have
compiled every kernel with the following patch posted to linux-kernel a
while back to avoid the sync race which was one of the reasons that 2.2.8
had FS corruption problems:
---