lkml.org 
[lkml]   [1999]   [Jul]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: need multiple e2fsck runs
On Jul 12, Andrea Arcangeli wrote:

> On Thu, 8 Jul 1999, Rogier Wolff wrote:
>
> >It seems that there are bits falling over in memory on your
> >machine. So fsck counts the bits in a bitmap block, and comes one
>
> Harald never got Oops and from his feedback it seems he is not
> experiencing plain MM corruption. He see only fs corruption so maybe
> something is wrong in the NCR/scsi controller side?


long tests during the weekend showed that it's bad memory in my case (or other
hardware problem, I'll replace memory tomorrow and we'll know for sure:-(.


thanks to Daniel Kobras <daniel.kobras@student.uni-tuebingen.de> I got a
nice small hack for testing ram which has only _very_ rare bit errors like in
my case. it allocates and reserves (almost) all memory, fills it with
all 1s (in my case) and then only sits there and tests the memory for changes
from time to time.

in my case, in bytes with address regexp pattern 0x07[01].[159d]c68 and 0x07[89].[159d]468
(just below 128M) sometimes bit 5 got cleared (0xff -> 0xbf) after many minutes
(usually many 10 minutes to hours).


this also matches the pattern which I found once when the emacs binary
was corrupted in buffer cache (do the decimal/octal/hex math yourself
and keep in mind that `cmp' starts counting at 1!) :

# cmp -l /usr/bin/emacs.good /usr/bin/emacs.bad
300137 355 255

and dropping bits in the ext2fs allocation bitmap pages can explain the `duplicate block's
which I got as only fs curruption pattern.


Harald

PS: memtest86 really should be improved to do such a test "wait _long_ time for bit flips"
since it's now the 3rd time where I see/know such problems.
--
All SCSI disks will from now on ___ _____
be required to send an email notice 0--,| /OOOOOOO\
24 hours prior to complete hardware failure! <_/ / /OOOOOOOOOOO\
\ \/OOOOOOOOOOOOOOO\
\ OOOOOOOOOOOOOOOOO|//
Harald Koenig, \/\/\/\/\/\/\/\/\/
Inst.f.Theoret.Astrophysik // / \\ \
koenig@tat.physik.uni-tuebingen.de ^^^^^ ^^^^^

-
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/

\
 
 \ /
  Last update: 2005-03-22 13:52    [W:0.434 / U:0.080 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site