Re: recursive slqb_lock

[Nick Piggin <npiggin@suse.de>]

On Sun, Mar 01, 2009 at 04:07:25PM +0100, Nick Piggin wrote:
> On Thu, Feb 26, 2009 at 01:34:50PM +0800, Wu Fengguang wrote:
> > Hi Nick,
> > 
> > I got a lockdep warning. It looks like the slqb_lock will be taken
> > twice in the call chain:
> > 
> >         s_start()  => take slqb_lock
> >           s_show()
> >             gather_stats()  => take slqb_lock again
> 
> Hey, thanks for this. The following patch should fix it.
> 
> --
> Fix the lockdep error reported by Fengguang where down_read slqb_lock
> is taken twice, with the possibility for a deadlock.
> 
> Signed-off-by: Nick Piggin <npiggin@suse.de>
> ---
> Index: linux-2.6/mm/slqb.c
> ===================================================================
> --- linux-2.6.orig/mm/slqb.c	2009-03-02 02:05:16.000000000 +1100
> +++ linux-2.6/mm/slqb.c	2009-03-02 02:05:45.000000000 +1100
> @@ -3079,7 +3079,9 @@ static void __gather_stats(void *arg)
>  	spin_unlock(&gather->lock);
>  }
>  
> -static void gather_stats(struct kmem_cache *s, struct stats_gather *stats)
> +/* must be called with slqb_lock held */
> +static void gather_stats_locked(struct kmem_cache *s,
> +				struct stats_gather *stats)
>  {
>  #ifdef CONFIG_NUMA
>  	int node;
> @@ -3089,8 +3091,6 @@ static void gather_stats(struct kmem_cac
>  	stats->s = s;
>  	spin_lock_init(&stats->lock);
>  
> -	down_read(&slqb_lock); /* hold off hotplug */
> -
>  	on_each_cpu(__gather_stats, stats, 1);
>  
>  #ifdef CONFIG_NUMA
> @@ -3119,10 +3119,15 @@ static void gather_stats(struct kmem_cac
>  	}
>  #endif
>  
> -	up_read(&slqb_lock);
> -
>  	stats->nr_objects = stats->nr_slabs * s->objects;
>  }
> +
> +static void gather_stats(struct kmem_cache *s, struct stats_gather *stats)
> +{
> +	down_read(&slqb_lock); /* hold off hotplug */
> +	gather_stats(s, stats);
> +	up_read(&slqb_lock);
> +}

Oops, I didn't test it. That call should be to gather_stats_locked of
course.