Hi, Linus. FYI: 2.5.3-pre2 hangs during boot. The last couple of
messages I get are:
Based upon Swansea University Computer Society NET3.039
apm: BIOS version 1.2 Flags 0x03 (Driver version 1.15)
apm: disabled - APM is not SMP safe (power off active).
Going back to 2.5.2, the next boot line is:
Starting kswapd
Regards,
Richard....
Permanent: [email protected]
Current: [email protected]
On Sun, Jan 20, 2002 at 11:26:42PM -0700, Richard Gooch wrote:
> Hi, Linus. FYI: 2.5.3-pre2 hangs during boot. The last couple of
> messages I get are:
> Based upon Swansea University Computer Society NET3.039
> apm: BIOS version 1.2 Flags 0x03 (Driver version 1.15)
> apm: disabled - APM is not SMP safe (power off active).
Ingo's scheduler patches solved this one for me (and others).
--
| Dave Jones. http://www.codemonkey.org.uk
| SuSE Labs
On Mon, 21 Jan 2002, Dave Jones wrote:
> > Hi, Linus. FYI: 2.5.3-pre2 hangs during boot. The last couple of
> > messages I get are:
> > Based upon Swansea University Computer Society NET3.039
> > apm: BIOS version 1.2 Flags 0x03 (Driver version 1.15)
> > apm: disabled - APM is not SMP safe (power off active).
>
> Ingo's scheduler patches solved this one for me (and others).
it's the migration-fixes-2.5.3-pre2-A1 sub-patch of the -J2 scheduler
patch, just in case someone wants to apply a minimal patch.
Ingo
On Mon, 21 Jan 2002, Dave Jones wrote:
> > Hi, Linus. FYI: 2.5.3-pre2 hangs during boot. The last couple of
> > messages I get are:
> > Based upon Swansea University Computer Society NET3.039
> > apm: BIOS version 1.2 Flags 0x03 (Driver version 1.15)
> > apm: disabled - APM is not SMP safe (power off active).
>
> Ingo's scheduler patches solved this one for me (and others).
the bug is the following: the 2.5.3-pre2 kernel in essence includes the
-I3 patch. The -I3 scheduler doesnt have the task migration fixes yet, but
has the improved load-balancer. And exactly this improved load-balancer
makes task-migration much more likely to happen during bootup =>
triggering it for ksoftirqd migration, resulting in a hang.
older versions had the migration bug as well, but the first iteration of
the load-balancer was much less accurate at balancing runqueues, so it
didnt trigger the bug.
Ingo