Hi Eric.
I suppose I should have explained exactly what I did. It went
something like this:
- System is currently ext3 on /, reiserfs on /var, /usr, /home
- Booted to single-user mode
- Backed up /var filesystem (to /usr/something-or-other). /var is on
device /dev/hda7.
- Unmounted original /var
- Ran 'mkfs -t xfs -f /dev/hda7' (it complained if I left out '-f'),
and this completed with no errors
- mount -t xfs /dev/hda7 /var. No problems.
- Copied contents of /usr/whatever-it-was back to /var
At this point, almost any program I ran would exit with the
preempt_count error (I remember it happened with ls at least, but
others did the same. I forget which). If you want me to run some more
tests, I can probably do some tonight.
I have seen some of these errors in my system log, and once or
twice when shutting the system down, but they were always related to
nfs and rpciod (? memory failure: replace brain). They were also rare;
it was only when I created and used an xfs filesystem that the errors
appeared on lots of other things, and far more often. Furthermore,
they went away when I switched back to reiserfs on /var.
.....Ron
At 15 Jul 2002 11:41:22 -0500,
Eric Sandeen <sandeen@xxxxxxx> wrote:
>
> Hi Ron -
>
> I tested pre-empt with xfs a week or two ago, and I did get some of
> these "<process> exited with preempt_count x" messages.
>
> On the other hand, I then tested with a vanilla ext2 kernel + pre-empt,
> and got the same messages.
>
> So, I don't think the problem is unique to, or necessarily even caused
> by, XFS. What processes complained of non-zero preempt_counts? Can you
> run these same tests on an ext[2,3]-only box, without xfs patches?
>
> -Eric
--
Ron Murray (rjmx@xxxxxxxx)
http://www.rjmx.net/~ron
GPG Public Key Fingerprint: F2C1 FC47 5EF7 0317 133C D66B 8ADA A3C4 D86C 74DE
|