Re: [PATCH v2] CPU hotplug, freezer: Fix bugs in CPU hotplug callpath

[Peter Zijlstra]

On Tue, 2011-10-04 at 18:58 +0530, Srivatsa S. Bhat wrote:
> +static int tasks_frozen;
> +
> +void set_tasks_frozen_flag(void)
> +{
> +       tasks_frozen = 1;
> +       smp_mb();
> +}
> +
> +void clear_tasks_frozen_flag(void)
> +{
> +       tasks_frozen = 0;
> +       smp_mb();
> +}
> +
> +int tasks_are_frozen(void)
> +{
> +       return tasks_frozen;
> +} 

See Documentation/memory-barriers.txt, memory barriers always come in
pairs, furthermore memory barriers always should have a comment
explaining the ordering and referring to the pair match.

I think you want at least an smp_rmb() before reading tasks_frozen,
possible you also want to use ACCESS_ONCE() to force the compiler to
emit the read.

Furthermore, do you really need this? isn't it both set and read from
the same task context, all under pm_mutex?
--
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/