On Sat, Apr 07, 2001 at 08:25:25PM -0500, Steve Lord wrote:
> > XFS likes me, I just know it. I get heaps and heaps of this while doing a
> > "dd if=largefile of=copyoflargefile bs=32768k".
> >
> >
> > Bad write on page 0xc1010130
> > Bad write on page 0xc10331ec
>
> This error message relates to flushing pages out beyond the end of file,
> the only place I saw it was via NFS, the code checked in on the 5th
> stopped it happening for me.
I am running the XFS CVS checkout from 20010407 (against 2.4.3), compiled
with RH7 kgcc.
> Presuming this is a current kernel, then we need more information
> about what you were doing to create the problem.
>
> a) was NFS involved, if so mount options and nfsd config options used,
Nope, no NFS involved. Just a dd of a file on a local XFS filesystem to a
file on the same local XFS filesystem. The block size given to dd seems to be
important: after this happened yesterday I retried with 1MB block size, and
that didn't give any kernel messages (whereas it initially did with 32MB block
size).
The filesystem in question is exported over NFS, but hasn't been mounted by
any client.
> b) are you still filling the filesystem
I'm sorry?
[buytenh@mara buytenh]$ df /data
Filesystem 1k-blocks Used Available Use% Mounted on
/dev/hda6 7058536 5539748 1518788 79% /data
> c) how big was the file
2GB. I just tried it again; the spam starts when about 800/900MB of the file
has been copied.
> d) machine configuration
Athlon 750, 128MB memory, 10GB system disk, 80GB data disk, all filesystems
XFS, usrquota. Anything more specific?
My coworker's machine does it too. Athlon 800, 128MB memory, 10GB system disk,
80GB data disk, all filesystems XFS, no usrquota. A "dd if=/dev/zero of=foo
bs=32768k count=128" on his 80GB partition gives no kernel messages, but a
subsequent "dd if=foo of=bar bs=32768k" starts spamming after about 256MB.
[buytenh@yoda buytenh]$ df /attic
Filesystem 1k-blocks Used Available Use% Mounted on
/dev/hdb1 80029796 10396888 69632908 13% /attic
cheers,
Lennert
|