Re: 3.15-rc2: longhaul cpufreq stalls tasks for 120s+

[Meelis Roos <mroos@linux.ee>]

> On 24 April 2014 15:46, Meelis Roos <mroos@linux.ee> wrote:
> > I can add debug to where needed and try it.
> 
> I am quite sure below wouldn't fix it, but just wanted to check
> for the corner case :(
> 
> Can you please try attached patch (mail's content would be
> broken):

Did not help. Will try lock debugging now.

> 
> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> index abda660..7198e23 100644
> --- a/drivers/cpufreq/cpufreq.c
> +++ b/drivers/cpufreq/cpufreq.c
> @@ -355,10 +355,12 @@ void cpufreq_freq_transition_begin(struct
> cpufreq_policy *policy,
>                 struct cpufreq_freqs *freqs)
>  {
>  wait:
> +       smp_mb();
>         wait_event(policy->transition_wait, !policy->transition_ongoing);
> 
>         spin_lock(&policy->transition_lock);
> 
> +       smp_mb();
>         if (unlikely(policy->transition_ongoing)) {
>                 spin_unlock(&policy->transition_lock);
>                 goto wait;
> @@ -380,8 +382,11 @@ void cpufreq_freq_transition_end(struct
> cpufreq_policy *policy,
> 
>         cpufreq_notify_post_transition(policy, freqs, transition_failed);
> 
> +       smp_mb();
> +
>         policy->transition_ongoing = false;
> 
> +       smp_mb();
>         wake_up(&policy->transition_wait);
>  }
>  EXPORT_SYMBOL_GPL(cpufreq_freq_transition_end);
> 

-- 
Meelis Roos (mroos@linux.ee)