Messages in this thread | | | Date | Tue, 22 Jun 1999 08:57:53 +0200 | From | Mike Thomi <> | Subject | Massive e2fs corruption with 2.2.9/10 uniproc <-> SMP |
| |
About the latest "Massive e2fs corruption with 2.2.9/10": I have the same probs on uniprocessor but *NOT* on SMP-kernel
SMP PII 266, Asus P2L97-ds, 128MB --------------------------------- 2 (130Mb) swap partitions on different hds Adaptec AIC-7880 Ultra SCSI host adapter Ultra Wide Controller onboard some uw-hds Linux version 2.2.9 (root@blabla) (gcc version 2.7.2.1) #2 SMP Sun Jun 6 21:54:33 MEST 1999
this box works great with 2.2.9/10, I can cp/cmp/bonnie tons of gigas without a problem.
P233, Asus tx97, 128Mb ---------------------- 2 (130Mb) swap partitions on different hds Adaptec AHA-294X Ultra2 SCSI host adapter Ultra-2 LVD/SE Wide Controller some u2w-hds Linux version 2.2.10 (root@blablaba) (gcc version egcs-2.91.66 19990314 (egcs-1.1.2 release)) #2 Mon Jun 21 12:49:19 MEST 1999 added "SCpnt->cmd_len = 0;" to the drivers/scsi/sd.c
it get sometimes the following kernel messages:
08:12: rw=0, want=1664051300, limit=3076447 attempt to access beyond end of device 08:12: rw=0, want=1664051300, limit=3076447
And it is not directly reproducible
file1=50Mb
when I "cp mountpoint1/file1 /mountpoint2/"; cmp -l mountpoint1/file1 /mountpoint2/file1 it works
BUT:
when I run more cp processes:
"cp mountpoint1/file1 /mountpoint2/" & "cp mountpoint1/file1 /mountpoint3/" & "cp mountpoint1/file1 /mountpoint4/" &
I often get "08:12: rw=0, want=1664051300, limit=3076447 attempt to access beyond end of device" and sometimes it works.
With files smaller than 10Mb there aren't any probs.
Could it be: - a buffer problem? - something about the changes with cache handling? page cache instead of buffer cache since 2.2.7? - something with "egcs-1.1.2 release"
What should I do?
- downgrading to 2.2.6 or 2.2.5? - which kernel is more stable?
mike
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |