Re: x86: tsc: v2 make TSC calibration more immune to interrupts

[Kasper Pedersen]

On 04/21/2011 12:39 AM, Josh Triplett wrote:
> 
> Have you considered disabling interrupts while calibrating?  That would
> ensure that you only have to care about SMIs, not arbitrary interrupts.
>
> Also, on more recent x86 systems you could look at MSR_SMI_COUNT (MSR
> 0x34) to detect if any SMIs have occurred during the sample period.
> rdmsr, start sample period, stop sample period, rdmsr, if delta of 0
> then no SMIs occurred.  Exists on Nehalem and newer, at least.


I have now tested this, and it is worth doing.

+		local_irq_save(flags);
		t1 = get_cycles();
		if (hpet)
			tp = hpet_readl(HPET_COUNTER) & 0xFFFFFFFF;
		else
			tp = acpi_pm_read_early();
		t2 = get_cycles();
+		local_irq_restore(flags);

On P3, when no SMI occur, the sample is now always perfect.
On Core2 there is still very rare variation - up to 8000 clock,
maybe bus contention?
Thinking it was cache I tried keeping irqs disabled for 5M
iterations, and that makes it occur _more_ often.

Atom shows the same as Core2, but even more rarely.

So, even if we have a Nehalem with SMI counter, the 5 samples
will still have benefit.


/Kasper Pedersen
-- 
This is my second patch ever, tell me when I do something wrong.
--
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/