Re: [patch 6/6] x86: add c1e aware idle function

[Thomas Gleixner]

On Thu, 19 Jun 2008, Rafael J. Wysocki wrote:
> On Wednesday, 18 of June 2008, Thomas Gleixner wrote:
> > On Wed, 18 Jun 2008, Rafael J. Wysocki wrote:
> > > On Wednesday, 18 of June 2008, Pavel Machek wrote:
> > > > On Thu 2008-06-12 10:29:00, Thomas Gleixner wrote:
> > > > > C1E on AMD machines is like C3 but without control from the OS. Up to
> > > > > now we disabled the local apic timer for those machines as it stops
> > > > > when the CPU goes into C1E. This excludes those machines from high
> > > > > resolution timers / dynamic ticks, which hurts especially the X2 based
> > > > > laptops.
> > > > > 
> > > > > The current boot time C1E detection has another more serious flaw:
> > > > > some BIOSes do not enable C1E until the ACPI processor module is
> > > > > loaded. This causes systems to stop working after that point.
> > > > > 
> > > > > To work nicely with C1E enabled machines we use a separate idle
> > > > > function, which checks on idle entry whether C1E was enabled in the
> > > > > Interrupt Pending Message MSR. This allows us to do timer broadcasting
> > > > 
> > > > Entering idle is quite a common operation, and reading MSR is quite
> > > > slow. Is it possible to do better here?
> > > > 
> > > > What happens if ACPI BIOS toggles MSR on all cpus *while* we are
> > > > entering idle? This seems inherently racy...
> > 
> > Well, on most of the machines we see the C1E bit on the first idle
> > entry either on CPU0 or on CPU1.
> > 
> > I know of exactly one machine which has the C1E thing enabled late
> > when the ACPI stuff runs.
> >  
> > > Yes, and that fits the picture I'm observing on the nx6325 (see the
> > > "linux-next: Tree for June 13: IO APIC breakage on HP nx6325" thread).
> > 
> > Hmm, when is the C1E detected on current mainline ? Does the boot CPU
> > have it or is it when the second CPU comes up ?
> 
> How can I check that?

When the boot cpu has it then it prints:

"Disabling APIC timer"

when the secondary CPU has it then it prints:

"AMD C1E detected late. Force timer broadcast."

Current mainline does not check for the late ACPI case so I'm pretty
sure that on your box it is detected during early boot way before we
can switch to highres/dyntick.

Thanks,

	tglx
--
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/