Re: [PATCH] slub: add missing kmem cgroup support to kmem_cache_free_bulk

[Jesper Dangaard Brouer <brouer@redhat.com>]

On Thu, 5 Nov 2015 11:38:43 +0300
Vladimir Davydov <vdavydov@virtuozzo.com> wrote:

> On Thu, Nov 05, 2015 at 02:06:21PM +0900, Joonsoo Kim wrote:
> > On Thu, Oct 29, 2015 at 02:05:31PM +0100, Jesper Dangaard Brouer wrote:
> > > Initial implementation missed support for kmem cgroup support
> > > in kmem_cache_free_bulk() call, add this.
> > > 
> > > If CONFIG_MEMCG_KMEM is not enabled, the compiler should
> > > be smart enough to not add any asm code.
> > > 
> > > Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
> > > ---
> > >  mm/slub.c |    3 +++
> > >  1 file changed, 3 insertions(+)
> > > 
> > > diff --git a/mm/slub.c b/mm/slub.c
> > > index 9be12ffae9fc..9875864ad7b8 100644
> > > --- a/mm/slub.c
> > > +++ b/mm/slub.c
> > > @@ -2845,6 +2845,9 @@ static int build_detached_freelist(struct kmem_cache *s, size_t size,
> > >  	if (!object)
> > >  		return 0;
> > >  
> > > +	/* Support for kmemcg */
> > > +	s = cache_from_obj(s, object);
> > > +
> > >  	/* Start new detached freelist */
> > >  	set_freepointer(s, object, NULL);
> > >  	df->page = virt_to_head_page(object);
> > 
> > Hello,
> > 
> > It'd better to add this 's = cache_from_obj()' on kmem_cache_free_bulk().
> > Not only build_detached_freelist() but also slab_free() need proper
> > cache.
> 
> Yeah, Joonsoo is right.

But cache_from_obj() takes an object as input and in kmem_cache_free_bulk()
that object is not directly available...  Could send "s" as a reference
(to build_detached_freelist) to allow re-assignment of "s" so
slab_free() gets the correct "s".  But it will not look pretty... 

Else we can get the object via: p[size -1] which also look a little
funny... but it might not be correct in-case NULL pointers in the input
p-array.


> Besides, there's a bug in kmem_cache_alloc_bulk:

Thanks for spotting this!!!

> > /* Note that interrupts must be enabled when calling this function. */
> > bool kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > 			   void **p)
> > {
> > 	struct kmem_cache_cpu *c;
> > 	int i;
> > 
> > 	/*
> > 	 * Drain objects in the per cpu slab, while disabling local
> > 	 * IRQs, which protects against PREEMPT and interrupts
> > 	 * handlers invoking normal fastpath.
> > 	 */
> > 	local_irq_disable();
> > 	c = this_cpu_ptr(s->cpu_slab);
> > 
> > 	for (i = 0; i < size; i++) {
> > 		void *object = c->freelist;
> > 
> > 		if (unlikely(!object)) {
> > 			/*
> > 			 * Invoking slow path likely have side-effect
> > 			 * of re-populating per CPU c->freelist
> > 			 */
> > 			p[i] = ___slab_alloc(s, flags, NUMA_NO_NODE,
> > 					    _RET_IP_, c);
> > 			if (unlikely(!p[i]))
> > 				goto error;
> > 
> > 			c = this_cpu_ptr(s->cpu_slab);
> > 			continue; /* goto for-loop */
> > 		}
> > 
> > 		/* kmem_cache debug support */
> > 		s = slab_pre_alloc_hook(s, flags);
> 
> slab_pre_alloc_hook expects a global cache and returns per memcg one, so
> calling this function from inside a kmemcg will result in hitting the
> VM_BUG_ON in __memcg_kmem_get_cache, not saying about mis-accounting of
> __slab_alloc.
> 
> memcg_kmem_get_cache should be called once, in the very beginning of
> kmem_cache_alloc_bulk, and it should be matched by memcg_kmem_put_cache
> when we are done.

To solve this correctly it looks like I need to pull out
memcg_kmem_put_cache(s) call in the slab_post_alloc_hook() call.

> 
> > 		if (unlikely(!s))
> > 			goto error;
> > 
> > 		c->freelist = get_freepointer(s, object);
> > 		p[i] = object;
> > 
> > 		/* kmem_cache debug support */
> > 		slab_post_alloc_hook(s, flags, object);
> > 	}
> > 	c->tid = next_tid(c->tid);
> > 	local_irq_enable();
> > 
> > 	/* Clear memory outside IRQ disabled fastpath loop */
> > 	if (unlikely(flags & __GFP_ZERO)) {
> > 		int j;
> > 
> > 		for (j = 0; j < i; j++)
> > 			memset(p[j], 0, s->object_size);
> > 	}
> > 
> > 	return true;
> > 
> > error:
> > 	__kmem_cache_free_bulk(s, i, p);
> > 	local_irq_enable();
> > 	return false;
> > }



-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  Author of http://www.iptv-analyzer.org
  LinkedIn: http://www.linkedin.com/in/brouer

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>