> > The following oops is more-or-less-deterministically reproducible
> > on my dual-processor SPARCstation 10 with 160MB RAM. It tends to send
> > the system down under heavy load caused by either sendmail/procmail
> > or apache. I first came across the bug at around 2.4.17, though it's
> > probably been lurking in the kernel much longer. I've gone through
> > quite a bit of trouble attempting to get the oops barf at me in 2.2.x
> > in case it's my hw config that's behind the whole problem, but I haven't
> > run into any breakdowns, 2.2.21-rc2 included.
>
> Assuming that you can handle your load, at least briefly, with a single
> CPU, have you tried booting with 'nosmp' on this machine? A serious test
> would build without SMP to get the whole locking stuff to go away, but
> this is quick and dirty.
>
> I'm convinced that there are evils still lurking in SMP after all these
> years.
Bad news, everyone. I'm definitely able to reproduce the very same fatal oops
on a UP kernel as well. Couldn't find a SPARC32 equivalent of i386's "nosmp,"
so I was forced to recompile the kernel (2.4.18-rc4) with CONFIG_SMP undefined
and do a "serious test" anyway.
Running something like the following kills the machine flat:
#!/bin/sh
i=0; while [ ! "$i" = "500" ]; do
/bin/echo testmail.| /bin/mail -s test autoreply@localhost
i=$(/usr/bin/expr $i + 1)
done
For a ksymoops decode of the oops in question and more, please look up my
former post in this thread.
Looks like it's back to 2.2 for all SPARC/Linux users for the time being.
-Tomas Szepe
pub 1024d/8e316a84 2002-01-29 tomas szepe <kala@pinerecords.com>
openpgp f/print 2955 2eea c4b8 b09e 7ae1 4d5d 68e3 d606 8e31 6a84
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/