Re: [RFC patch 0/4] TSC calibration improvements

[Ingo Molnar]

* Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:

> +		unsigned char expect = 0xfe;
> +
> +		t1 = get_cycles();
> +		for (expect = 0xfe, i = 0; i < QUICK_PIT_ITERATIONS; i++, expect--) {
> +			if (!pit_expect_msb(expect))
> +				goto failed;
> +		}
> +		t2 = get_cycles();

i suspect you didnt want the second initialization of 'expect' to 0xfe? 
I fixed that up by removing the second one.

( although i think it would all be even clearer if we initialized 
  'expect' to 0xff before the first call of pit_expect_msb() higher up, 
  and then doing an expect-- before doing this loop - but that is really 
  painting the bikeshed i guess. )

> @@ -189,9 +300,15 @@ unsigned long native_calibrate_tsc(void)
>  {
>  	u64 tsc1, tsc2, delta, pm1, pm2, hpet1, hpet2;
>  	unsigned long tsc_pit_min = ULONG_MAX, tsc_ref_min = ULONG_MAX;
> -	unsigned long flags;
> +	unsigned long flags, fast_calibrate;
>  	int hpet = is_hpet_enabled(), i;
>  
> +	local_irq_save(flags);
> +	fast_calibrate = quick_pit_calibrate();
> +	local_irq_restore(flags);
> +	if (fast_calibrate)
> +		return fast_calibrate;
> +
>  	/*
>  	 * Run 5 calibration loops to get the lowest frequency value
>  	 * (the best estimate). We use two different calibration modes

these bits collided with Thomas's textually but not conceptually: Thomas 
improved the slowpath calibration while you introduced a new fastpath. I 
kept both and applied your patch to tip/x86/tsc.

if you are happy with the commit below, could you please give your 
Signed-off-by for it?

	Ingo

-------------->