Re: OOPS in perf_mmap_close()

[Peter Zijlstra <peterz@infradead.org>]

On Mon, Jun 03, 2013 at 03:26:45PM +0200, Peter Zijlstra wrote:
> On Thu, May 30, 2013 at 08:51:00AM -0400, Vince Weaver wrote:
> > > I'll go prod, thanks again!
> 
> 
> OK the below builds and seems to survive both test cases and about 10 minutes
> of fuzzing -- fingers crossed.
> 
> ---
> @@ -3779,12 +3834,21 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma)
>  		return -EINVAL;
>  
>  	WARN_ON_ONCE(event->ctx->parent_ctx);
> +again:
>  	mutex_lock(&event->mmap_mutex);
>  	if (event->rb) {
> -		if (event->rb->nr_pages == nr_pages)
> -			atomic_inc(&event->rb->refcount);
> -		else
> +		if (event->rb->nr_pages != nr_pages)
		{
>  			ret = -EINVAL;
			goto unlock;
		}

Seems to cure it... I'm going to let it run over night.

> +
> +		if (!atomic_inc_not_zero(&event->rb->mmap_count)) {
> +			/*
> +			 * Raced against perf_mmap_close() through
> +			 * perf_event_set_output(). Try again, hope for better
> +			 * luck.
> +			 */
> +			mutex_unlock(&event->mmap_mutex);
> +			goto again;
> +		}
>  		goto unlock;
>  	}
>