Re: [PATCH] mm/slab: avoid allocating slabobj_ext array from its own slab

[Harry Yoo <harry.yoo@oracle.com>]

On Mon, Jan 26, 2026 at 09:37:29AM +0100, Vlastimil Babka wrote:
> On 1/26/26 09:30, Harry Yoo wrote:
> > On Mon, Jan 26, 2026 at 08:36:16AM +0100, Vlastimil Babka wrote:
> > /*
> >  * Calculate the allocation size for slabobj_ext array.
> >  *
> >  * When memory allocation profiling is enabled, the obj_exts array
> >  * could be allocated from the same slab it's being allocated for.
> >  * This would prevent the slab from ever being freed because it would
> >  * always contain at least one allocated object (its own obj_exts array).
> >  *
> >  * To avoid this, increase the allocation size when we detect the array
> >  * would come from the same cache, forcing it to use a different cache.
> >  */
> > static inline size_t obj_exts_alloc_size(struct kmem_cache *s,
> >                                          struct slab *slab, gfp_t gfp)
> > {
> >         size_t sz = sizeof(struct slabobj_ext) * slab->objects;
> >         struct kmem_cache *obj_exts_cache;
> > 
> >         /*
> >          * slabobj_ext array for KMALLOC_CGROUP allocations
> >          * are served from KMALLOC_NORMAL caches.
> >          */
> >         if (!mem_alloc_profiling_enabled())
> >                 return sz;
> > 
> >         if (sz > KMALLOC_MAX_CACHE_SIZE)
> >                 return sz;
> > 
> >         if (!is_kmalloc_normal(s))
> >                 return sz;
> > 
> >         obj_exts_cache = kmalloc_slab(sz, NULL, gfp, 0);
> >         /*
> >          * Random kmalloc caches have multiple caches per size, and the cache
> 
> Maybe start with something like "We can't simply compare s with
> obj_exts_cache, because..."
> 
> >          * is selected by the caller address. Since caller address may differ
> >          * between kmalloc_slab() and actual allocation, bump size when both
> >          * are normal kmalloc caches of same size.
> 
> As we don't test the other for normal kmalloc(), anymore this now reads as
> if we forgot to.

Ok, something like this:

"We can't simply compare s with obj_exts_cache, because random kmalloc
caches have multiple caches per size, selected by caller address.
Since caller address may differ between kmalloc_slab() and actual
allocation, bump size when sizes are equal."

> >          */
> >         if (s->size == obj_exts_cache->size)
> >                 return s->object_size + 1;
> 
> Why switch to size from object_size for the checks? I'd be worried that due
> to debugging etc this can yield wrong results?

Oops, sorry. copied from wrong version of the patch.
Yeah I initially compared size and then switched to object_size for the
reason you mentioned.

> > 
> >         return sz;
> > }

-- 
Cheers,
Harry / Hyeonggon