Re: [2.6.32-rc3 kmemleak] WARNING: atkernel/lockdep.c:3161check_flags+0xbe/0x180()

[Catalin Marinas <catalin.marinas@arm.com>]

On Wed, 2009-10-14 at 20:55 +0900, Tetsuo Handa wrote:
> Tetsuo Handa wrote:
> > > This is the "DEBUG_LOCKS_WARN_ON(current->softirqs_enabled)" warning.
> > > I'm not sure why this happens but from the trace it seems that kmemleak
> > > is being called recursively via alloc_slabmgmt() which is caused by
> > > kmem_cache_alloc() called from create_object() in kmemleak.c.
> > If what my guess shown below is correct,
> > (object_cache->slabp_cache->flags & SLAB_NOLEAKTRACE) == 0
> > is triggering recursive calls.
> 
> I applied below patch
> 
> --- linux-2.6.32-rc4/mm/slab.c	2009-10-14 16:22:44.962007072 +0900
> +++ linux-2.6.32-rc4-ccs/mm/slab.c	2009-10-14 16:08:14.000000000 +0900
> @@ -2573,6 +2573,8 @@
>  	struct slab *slabp;
>  
>  	if (OFF_SLAB(cachep)) {
> +		BUG_ON((cachep->flags & SLAB_NOLEAKTRACE) &&
> +		       !(cachep->slabp_cache->flags & SLAB_NOLEAKTRACE));
>  		/* Slab management obj is off-slab. */
>  		slabp = kmem_cache_alloc_node(cachep->slabp_cache,
>  					      local_flags, nodeid);
> 
> and verified that (cachep->slabp_cache->flags & SLAB_NOLEAKTRACE) == 0 is
> triggering recursive call.
> This is not locking related problem. This is stack overflow problem

I was looking at the same code but in my platform, OFF_SLAB(cachep) for
the kmemleak caches is always 0. I only added:

	BUG_ON(cachep->flags & SLAB_NOLEAKTRACE);

and it never triggered but I need to enable more debugging features as
in your .config and re-run.

I assume you haven't modified the kmemleak.c file to increase the
MAX_TRACE size.

I don't think we should add SLAB_NOLEAKTRACE to slabmgmt since in theory
it should work. Is there a slabmgmt allocated for each
kmem_cache_alloc()? If yes, it will indeed get into infinite recursive
calls. If not, kmemleak should be able to cope though in some situations
it may still overflow.

Until I manage to reproduce the problem, could you please try the patch
below:

diff --git a/mm/slab.c b/mm/slab.c
index 7dfa481..f8f671b 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -2263,7 +2263,8 @@ kmem_cache_create (const char *name, size_t size, size_t align,
 	 * (bootstrapping cannot cope with offslab caches so don't do
 	 * it too early on.)
 	 */
-	if ((size >= (PAGE_SIZE >> 3)) && !slab_early_init)
+	if ((size >= (PAGE_SIZE >> 3)) && !slab_early_init &&
+	    !(cachep->flags & SLAB_NOLEAKTRACE))
 		/*
 		 * Size is large, assume best to place the slab management obj
 		 * off-slab (should allow better packing of objs).


Thanks.

-- 
Catalin