Re: possible recursive locking detected at kmem_cache_free

["Pallipadi, Venkatesh" <venkatesh.pallipadi@intel.com>]

On Fri, 2009-06-26 at 00:37 -0700, Pekka J Enberg wrote:
> Hi Venki,
> 
> On Thu, 25 Jun 2009, Pallipadi, Venkatesh wrote:
> > > > Not sure if this has been reported elsewhere. I started seeing this with git
> > > > kernel from this morning on a system with RedHat EL5 base installation with
> > > > ext3 root file system. Attached is the config I am using.
> > > > 
> > > > Let me know if you need any more details. This seems to be at every boot. I
> > > > am trying the bisect right now.
> > > > 
> > > 
> > > Thanks.  Let's cc Pekka.
> > > 
> > 
> > git bisect result.
> 
> Thanks! I think I just broke lockdep annotations for slab. Does this patch 
> fix things?
> 

Yes. This patch helps.

Thanks,
Venki

> 			Pekka
> 
> diff --git a/mm/slab.c b/mm/slab.c
> index e74a16e..37ee532 100644
> --- a/mm/slab.c
> +++ b/mm/slab.c
> @@ -1544,9 +1544,6 @@ void __init kmem_cache_init(void)
>  	}
>  
>  	g_cpucache_up = EARLY;
> -
> -	/* Annotate slab for lockdep -- annotate the malloc caches */
> -	init_lock_keys();
>  }
>  
>  void __init kmem_cache_init_late(void)
> @@ -1563,6 +1560,9 @@ void __init kmem_cache_init_late(void)
>  	/* Done! */
>  	g_cpucache_up = FULL;
>  
> +	/* Annotate slab for lockdep -- annotate the malloc caches */
> +	init_lock_keys();
> +
>  	/*
>  	 * Register a cpu startup notifier callback that initializes
>  	 * cpu_cache_get for all new cpus