Re: PowerOP 2/3: Intel Centrino support

[Ingo Molnar <mingo@elte.hu>]

* Todd Poynor <tpoynor@mvista.com> wrote:

> +static int
> +powerop_centrino_get_point(struct powerop_point *point)
> +{
> +	unsigned l, h;
> +	unsigned cpu_freq;
> +
> +	rdmsr(MSR_IA32_PERF_STATUS, l, h);
> +	if (unlikely((cpu_freq = ((l >> 8) & 0xff) * 100) == 0)) {
> +		/*
> +		 * On some CPUs, we can see transient MSR values (which are
> +		 * not present in _PSS), while CPU is doing some automatic
> +		 * P-state transition (like TM2). Get the last freq set 
> +		 * in PERF_CTL.
> +		 */
> +		rdmsr(MSR_IA32_PERF_CTL, l, h);
> +		cpu_freq = ((l >> 8) & 0xff) * 100;
> +	}
> +
> +	point->param[POWEROP_CPU + smp_processor_id()] = cpu_freq;
> +	point->param[POWEROP_V + smp_processor_id()] = ((l & 0xff) * 16) + 700;
> +	return 0;
> +}

doesnt seem to be SMP-safe, nor PREEMPT-safe. You probably want to do 
the locking in the highlevel functions.

	Ingo