Re: [patch] cpufreq: mark cpufreq_tsc() as core_initcall_sync

[Oleg Nesterov]

On 11/17, Paul E. McKenney wrote:
>
>  int srcu_read_lock(struct srcu_struct *sp)
>  {
>  	int idx;
> +	struct srcu_struct_array *sap;
>  
>  	preempt_disable();
>  	idx = sp->completed & 0x1;
> -	barrier();  /* ensure compiler looks -once- at sp->completed. */
> -	per_cpu_ptr(sp->per_cpu_ref, smp_processor_id())->c[idx]++;
> -	srcu_barrier();  /* ensure compiler won't misorder critical section. */
> +	sap = rcu_dereference(sp->per_cpu_ref);
> +	if (likely(sap != NULL)) {
> +		barrier();  /* ensure compiler looks -once- at sp->completed. */
> +		per_cpu_ptr(rcu_dereference(sap),
> +			    smp_processor_id())->c[idx]++;
> +		smp_mb();
> +		preempt_enable();
> +		return idx;
> +	}
> +	if (mutex_trylock(&sp->mutex)) {
> +		preempt_enable();
> +		if (sp->per_cpu_ref == NULL)
> +			sp->per_cpu_ref = alloc_srcu_struct_percpu();
> +		if (sp->per_cpu_ref == NULL) {
> +			atomic_inc(&sp->hardluckref);
> +			mutex_unlock(&sp->mutex);
> +			return -1;
> +		}
> +		mutex_unlock(&sp->mutex);
> +		return srcu_read_lock(sp);
> +	}
>  	preempt_enable();
> -	return idx;
> +	atomic_inc(&sp->hardluckref);
> +	return -1;
>  }

This is a real nitpick, but in theory we have a possibility for the livelock.

Suppose that synchronize_srcu() takes sp->mutex and fails to allocate
sp->per_cpu_ref. If we have a flow of srcu_read_lock/srcu_read_unlock,
this loop in synchronize_srcu()

  	while (srcu_readers_active_idx(sp, idx))
  		schedule_timeout_interruptible(1);

may spin unpredictably long, because we use the same sp->hardluckref for
accounting.

Oleg.

-
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/