Re: Wolk 4.9 smp cpu detection and initialization problems
KORN Andras <[email protected]>
| Newsgroups | gmane.linux.wolk.devel |
|---|---|
| Message-ID | <[email protected]> |
On Mon, Sep 15, 2003 at 12:45:06AM +0200, Olaf Unterkofler wrote:
> Hard lock - the beasts stop dead - about 98% of the time (black screen, no
> log, no console output), or every once in a while I have a reboot. As
> soon as I disable ACPI (wich happens to disable the second CPU) or the
> second CPU directly the system seems to work fine (no lockup, no reboot).
> I have the subtle feeling that the kernel tries to pass some
> task/interrupt handler or whatever to the second/third/fourth CPU and then
> gets stuck. Changing the maximum number of CPUs to more than 4 causes the
> system to reboot during kernel initialization - again, without any output
> on the console. The machines simply loop in that state until I press the
> reset switch.
I seem to recall that there exists a hyperthreading scheduler patch from
mingo. Maybe it would help, maybe not, but it probably is worth a try.
http://people.redhat.com/mingo/nptl-patches/nptl-2.4.22-ac1-A1
http://people.redhat.com/mingo/O(1)-scheduler/sched-HT-2.4.22-ac1-A0
I imaginge they'll require some manual conflict resolution to apply to wolk.
Andrew
--
Andrew Korn (Korn Andras) <korn at chardonnay.math.bme.hu>
Finger korn at chardonnay.math.bme.hu for pgp key. QOTD:
A language is a dialect with an army behind it.
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
WOLK - Working Overloaded Linux Kernel
[email protected]
https://lists.sourceforge.net/lists/listinfo/wolk-devel