Subject: Re: Data Corruption
To: None <dc@microbits.com.au>
From: Johan A.van Zanten <johan@giantfoo.org>
List: netbsd-users
Date: 09/27/2005 01:27:00
"Daniel Cox" <dc@microbits.com.au> wrote:
> # cmp a.tgz b.tgz
> a.tgz b.tgz differ: char 4067329, line 15069
> 
> Nearly every time I copy a big file I get a different checksum.
> Same results with FFS1, FFS2 with and without softdep.
> 
> I have found that anything less than 10MB appears to work.

Very strange. I grabbed the first big file i found and started doing cp's
and cmp's. I can't reproduce this, on SPARC or Alpha architecture
machines, SCSI and IDE drives.  I've tried comparing the files when both
copies are on FFS or one on MFS, on SCSI disk (sparc) or EIDE (Alpha).

brahma:/tmp $ uname -a
NetBSD brahma 2.0.2_STABLE NetBSD 2.0.2_STABLE (MANGOLASSI.MP) #1: Sat May
7 00:21:31 CDT 2005 toor@sarasvati:/local/NetBSD/src/NetBSD-2.0/src/sys/arch/sparc/compile/obj.sparc/MANGOLASSI.MP sparc

brahma:/tmp $ df -kl
Filesystem  1K-blocks     Used     Avail Capacity  Mounted on
/dev/sd0a      123887    23598     94094    20%    /
/dev/sd0f      987571    47637    890555     5%    /var
/dev/sd0g     1017547   812829    153840    84%    /usr
/dev/sd0h    32455712  2974162  27858764     9%    /local
mfs:252         63471    36597     23700    60%    /tmp
brahma:/tmp $ cp 35MB-of-pdfs.tar.gz /local/tmp/
brahma:/tmp $ cmp 35MB-of-pdfs.tar.gz /local/tmp/35MB-of-pdfs.tar.gz
brahma:/tmp $ echo $?
0
brahma:/tmp $ ls -l 35MB-of-pdfs.tar.gz
-rw-r--r--  1 johan  wheel  37397921 Sep 27 01:16 35MB-of-pdfs.tar.gz


 -johan