Re: [7.0-rc1] codetag: kernel warning "alloc_tag was not set" during boot

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

On Mon, Feb 23, 2026 at 11:07:08PM -0800, Suren Baghdasaryan wrote:
> On Mon, Feb 23, 2026 at 6:28 PM David Wang <00107082@163.com> wrote:
> >
> >
> > At 2026-02-24 01:18:32, "Vlastimil Babka" <vbabka@suse.com> wrote:
> > >On 2/23/26 16:51, David Wang wrote:
> > >> Hi,
> > >>
> > >> When upgrade to 7.0.0-rc1, caught a kernel WARN during boot:
> > >
> > >Would this possibly help? (it probably shouldn't but let's see)
> > >https://lore.kernel.org/all/20260223075809.19265-1-harry.yoo@oracle.com/
> > >
> >
> > Hi,
> > I tried this patch, but sadly, it doesn't help with the warning on my system...
> 
> Unfortunately I can't boot far enough with your config to get that warning.
> 
> You might be able to track down which allocation does not get its tag
> by first checking for current->alloc_tag==NULL inside
> __alloc_tagging_slab_alloc_hook() and dump the stack if that happens.
> If there are no hits (which is probably what will happen) you can do
> the same while checking for obj_exts==NULL. I suspect that condition
> will hit multiple times and one of those will likely be the cause of
> this warning. I'll look into this more tomorrow if the cause is not
> found by then.

Just sharing updates on this before going to bed...

I was able to reproduce it on my machine I have a working fix
(nowhere close to upstream quality, though)

The reason why we're seeing "alloc_tag not was set" is
because SLUB allocates empty sheaves for kmalloc
with __GFP_NO_OBJ_EXT and it doesn't teach memory profiling
how to handle this when such sheaves are freed.

When __GFP_NO_OBJ_EXT is used in alloc_slab_obj_exts(), it later
avoids this "alloc_tag was not set" warning by marking alloc_tags
empty in free_slab_obj_exts(), just before freeing obj_exts.

And we don't handle this when allocating freeing sheaves
that were allocated with __GFP_NO_OBJ_EXT.

Passing __GFP_NO_OBJ_EXT skips 1) allocation of obj_exts,
and 2) alloc_tag_add() even when obj_exts is already allocated,
and this confuses memory profiling later.

I'm adding the fix I have now. (I guess Suren might have some preference
on how to solve it though)

  1. mark obj_exts allocation failure when slab has no obj_exts and
     gfp flag has __GFP_NO_OBJ_EXT, so that a later obj_exts allocation
     will mark alloc_tags empty.

  2. Set alloc_tag when obj_exts is allocated available,
     even when __GFP_NO_OBJ_EXT is set.

     Because it's already allocated, we don't have to worry about
     recursive allocation.

diff --git a/mm/slub.c b/mm/slub.c
index c7c8b660a994..4dbdfcd46771 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -2060,8 +2060,6 @@ static inline void mark_objexts_empty(struct slabobj_ext *obj_exts)
 			return;
 		}
 
-		/* codetag should be NULL here */
-		WARN_ON(ext->ref.ct);
 		set_codetag_empty(&ext->ref);
 		put_slab_obj_exts(slab_exts);
 	}
@@ -2361,6 +2359,7 @@ __alloc_tagging_slab_alloc_hook(struct kmem_cache *s, void *object, gfp_t flags)
 	unsigned long obj_exts;
 	struct slabobj_ext *obj_ext;
 	struct slab *slab;
+	unsigned int obj_idx;
 
 	if (!object)
 		return;
@@ -2368,26 +2367,36 @@ __alloc_tagging_slab_alloc_hook(struct kmem_cache *s, void *object, gfp_t flags)
 	if (s->flags & (SLAB_NO_OBJ_EXT | SLAB_NOLEAKTRACE))
 		return;
 
-	if (flags & __GFP_NO_OBJ_EXT)
-		return;
-
 	slab = virt_to_slab(object);
+	if (flags & __GFP_NO_OBJ_EXT) {
+		obj_exts = slab_obj_exts(slab);
+		if (!obj_exts) {
+			mark_failed_objexts_alloc(slab);
+			return;
+		} else {
+			goto tag_add;
+		}
+	}
+
 	obj_exts = prepare_slab_obj_exts_hook(s, slab, flags, object);
 	/*
 	 * Currently obj_exts is used only for allocation profiling.
 	 * If other users appear then mem_alloc_profiling_enabled()
 	 * check should be added before alloc_tag_add().
 	 */
-	if (obj_exts) {
-		unsigned int obj_idx = obj_to_index(s, slab, object);
-
-		get_slab_obj_exts(obj_exts);
-		obj_ext = slab_obj_ext(slab, obj_exts, obj_idx);
-		alloc_tag_add(&obj_ext->ref, current->alloc_tag, s->size);
-		put_slab_obj_exts(obj_exts);
-	} else {
+	if (!obj_exts) {
 		alloc_tag_set_inaccurate(current->alloc_tag);
+		return;
 	}
+
+tag_add:
+	obj_idx = obj_to_index(s, slab, object);
+
+	get_slab_obj_exts(obj_exts);
+	obj_ext = slab_obj_ext(slab, obj_exts, obj_idx);
+	alloc_tag_add(&obj_ext->ref, current->alloc_tag, s->size);
+	put_slab_obj_exts(obj_exts);
+	return;
 }
 
 static inline void

-- 
Cheers,
Harry / Hyeonggon