APIC Error when doing apic_pm_suspend

APIC Error when doing apic_pm_suspend

[Zwane Mwaikambo]

I get an APIC error 0x40 when resuming from an apm -s. If i'm correct
that would be an illegal register access wouldn't it? I tried putting
enter/exit printks in the apic_pm_resume/suspend functions and it showed
that both returned before the APIC error printk. Is there anyway of finding out
which register access it was? I "thought" it would be one of the
apic_writes in the pm functions but looks like i might be wrong.

The kernel is compiled with local APIC and gets detected and enabled on
boot (UP machine).

Thanks,
	Zwane Mwaikambo

Re: APIC Error when doing apic_pm_suspend

[Arjan van de Ven]

Zwane Mwaikambo wrote:
> 
> I get an APIC error 0x40 when resuming from an apm -s. If i'm correct
> that would be an illegal register access wouldn't it? I tried putting
> enter/exit printks in the apic_pm_resume/suspend functions and it showed
> that both returned before the APIC error printk. Is there anyway of finding out
> which register access it was? I "thought" it would be one of the
> apic_writes in the pm functions but looks like i might be wrong.
> 
> The kernel is compiled with local APIC and gets detected and enabled on
> boot (UP machine).

Just about all bioses that support suspend do not have the knowledge
that an
operating system would use apics, since windows95 doesn't do that. The
fact
that it appears to mostly work for you is RARE. You're very very lucky
with
an almost not broken bios..... UP APIC and Suspend are usually very
exclusive.
(well, actually, the suspend often works, it's the resume that hurts)

Greetings,
   Arjan van de Ven

Re: APIC Error when doing apic_pm_suspend

[Zwane Mwaikambo]

On Wed, 5 Dec 2001, Arjan van de Ven wrote:
> Just about all bioses that support suspend do not have the knowledge
> that an
> operating system would use apics, since windows95 doesn't do that. The
> fact
> that it appears to mostly work for you is RARE. You're very very lucky
> with
> an almost not broken bios..... UP APIC and Suspend are usually very
> exclusive.
> (well, actually, the suspend often works, it's the resume that hurts)

Hmm that puts things in a different light... don't you just the love PC
compatible ;)

Thanks,
	Zwane Mwaikambo

Re: APIC Error when doing apic_pm_suspend

[Mikael Pettersson]

Zwane Mwaikambo writes:
 > I get an APIC error 0x40 when resuming from an apm -s. If i'm correct
 > that would be an illegal register access wouldn't it? I tried putting
 > enter/exit printks in the apic_pm_resume/suspend functions and it showed
 > that both returned before the APIC error printk. Is there anyway of finding out
 > which register access it was? I "thought" it would be one of the
 > apic_writes in the pm functions but looks like i might be wrong.
 > 
 > The kernel is compiled with local APIC and gets detected and enabled on
 > boot (UP machine).

No, 0x40 is an illegal vector error. It's a (semi-) known quirk in the P6 family
of processors that you get this error when writing a null vector to any of the
LVT entries, even if you are also setting the mask bit at the same time.
Both the clear_local_APIC() call at PM suspend and the reinitialisation at PM
resume can trigger this.

The "error" is mostly harmless. Ignore it for now, I'll do a patch to silence it later.

/Mikael

Re: APIC Error when doing apic_pm_suspend

[Zwane Mwaikambo]

On Wed, 5 Dec 2001, Mikael Pettersson wrote:

> No, 0x40 is an illegal vector error. It's a (semi-) known quirk in the P6 family
> of processors that you get this error when writing a null vector to any of the
> LVT entries, even if you are also setting the mask bit at the same time.
> Both the clear_local_APIC() call at PM suspend and the reinitialisation at PM
> resume can trigger this.
>
> The "error" is mostly harmless. Ignore it for now, I'll do a patch to silence it later.

Could you please CC me the patch.

Thanks,
	Zwane Mwaikambo