Re: 2.6.17-rc1: kernel only boots one CPU on HT system

From: Shaohua Li
Date: Mon Apr 24 2006 - 21:55:07 EST


On Mon, 2006-04-24 at 19:26 +0200, Kevin Baradon wrote:
> Le Lundi 24 Avril 2006 03:57, Shaohua Li a Ãcrit :
> > Hi,
>
> Hello,
> >
> > On Sun, 2006-04-23 at 14:15 -0700, Andrew Morton wrote:
> > > Kevin Baradon <kevin.baradon@xxxxxxxxx> wrote:
> > > > Hello,
> > > >
> > > > Starting with kernel 2.6.17-rc1 (also happens with 2.6.17-rc2), second
> > > > logical-CPU of my Hyperthreading system no longer boots.
> > > >
> > > > I tracked up changes in APIC code, and it appears reverting commit
> > > > 7c5c1e427b5e83807fd05419d1cf6991b9d87247 fixes this bug.
> > >
> > > That helps heaps, thanks.
> >
> > The commit doesn't look like the root cause to me. BIOS already assigns
> > unique id to ioapic, and the cpu family is 15, so with/without the patch
> > the code path hasn't any difference. Kevin, can you please make a clean
> > build and check if the patch is the real cause?
> >
>
> You were right. Reverting this commit helps sometimes, but doesn't work
> reliably. When my computer booted this morning, I've had only one CPU
> detected. I've tried booting several times, even with a complete power down.
> Nothing changed.
>
> I've also tried with kernel 2.6.16-rc6, which booted fine and detected two
> CPUs.
>
> I've applied your small patch. Debugging output is attached.
>
> If you want, I can apply this small patch also to kernel 2.6.16-rc6, and send
> you debugging output.
>
> > If it still doesn't work, you might apply a small change below to
> > include/asm-i386/apic.h, and attach the dmesg, so we could analyze it.
> >
>
> File attached.
The CPU doesn't startup. Sometime ago somebody reported a similar issue
for 2.6.16.1. But the failure isn't reliably triggered. IIRC, there
isn't a solution.

Thanks,
Shaohua

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/