[PATCH 1/5] arm: provide a mechanism to reserve performance counters

[Jamie Iles]

On Wed, Jan 06, 2010 at 01:00:56PM +0100, Michał Nazarewicz wrote:
>> +#else /* CONFIG_CPU_HAS_PMU */
>> +
>> +static inline const struct pmu_irqs *
>> +reserve_pmu(void)
>> +{
>> +	ERR_PTR(-ENODEV);
>
> -	ERR_PTR(-ENODEV);
> +	return ERR_PTR(-ENODEV);
>
>> +}
>> +
>> +static inline int
>> +release_pmu(const struct pmu_irqs *irqs)
>> +{
>
> +	return -ENODEV;
>
>> +}
>> +
>> +static inline int
>> +init_pmu(void)
>> +{
>> +	return -ENODEV;
>> +}
>> +
>> +#endif /* CONFIG_CPU_HAS_PMU */
>> +
>> +#endif /* __ARM_PMU_H__ */
Thanks, well spotted!
>> +static const struct pmu_irqs pmu_irqs = {
>> +	.irqs	    = irqs,
>> +	.num_irqs   = ARRAY_SIZE(irqs),
>> +};
>> +
>> +static DECLARE_MUTEX(pmu_mutex);
>
> Isn't mutex an overkill? A bit field would be enough:
>
> -static DECLARE_MUTEX(pmu_mutex);
> +static volatile long pmu_mutex;
Yes, it probably is. I don't think performance is important here but that's a
simpler solution so I'll make that change.

Thanks,

Jamie