Re: [PATCH v5] slab: reset obj_ext when it is not actually valid during freeing

[Suren Baghdasaryan]

On Wed, Oct 15, 2025 at 9:37 AM Vlastimil Babka <vbabka@xxxxxxx> wrote:
>
> On 10/15/25 18:29, Suren Baghdasaryan wrote:
> > On Wed, Oct 15, 2025 at 7:17 AM Hao Ge <hao.ge@xxxxxxxxx> wrote:
> >>
> >> From: Hao Ge <gehao@xxxxxxxxxx>
> >>
> >> If obj_exts allocation failed, slab->obj_exts is set to OBJEXTS_ALLOC_FAIL,
> >> But we did not clear it when freeing the slab. Since OBJEXTS_ALLOC_FAIL and
> >> MEMCG_DATA_OBJEXTS currently share the same bit position, during the
> >> release of the associated folio, a VM_BUG_ON_FOLIO() check in
> >> folio_memcg_kmem() is triggered because it was mistakenly assumed that
> >> a valid folio->memcg_data was not cleared before freeing the folio.
> >>
> >> When freeing a slab, we clear slab->obj_exts and reset it to 0
> >> if the obj_ext array has been successfully allocated.
> >> So let's reset slab->obj_exts to 0 when freeing a slab if
> >> the obj_ext array allocated fail to allow them to be returned
> >> to the buddy system more smoothly.
> >>
> >> Signed-off-by: Hao Ge <gehao@xxxxxxxxxx>
> >> ---
> >> v5: Adopt the simpler solution proposed by Vlastimil;
> >>     Many thanks to him
> >> ---
> >>  mm/slub.c | 10 +++++++++-
> >>  1 file changed, 9 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/mm/slub.c b/mm/slub.c
> >> index b1f15598fbfd..2e4340c75be2 100644
> >> --- a/mm/slub.c
> >> +++ b/mm/slub.c
> >> @@ -2170,8 +2170,16 @@ static inline void free_slab_obj_exts(struct slab *slab)
> >>         struct slabobj_ext *obj_exts;
> >>
> >>         obj_exts = slab_obj_exts(slab);
> >> -       if (!obj_exts)
> >> +       if (!obj_exts) {
> >> +               /*
> >> +                * If obj_exts allocation failed, slab->obj_exts is set to OBJEXTS_ALLOC_FAIL,
> >> +                * In this case, we will end up here.
> >> +                * Therefore, we should clear the OBJEXTS_ALLOC_FAIL flag first when freeing a slab.
> >> +                * Then let's set it to 0 as below.
> >> +                */
> >> +               slab->obj_exts = 0;
> >>                 return;
> >> +       }
> >
> > How about this instead:
> >
> > static inline void free_slab_obj_exts(struct slab *slab)
> > {
> >         struct slabobj_ext *obj_exts;
> >
> >         obj_exts = slab_obj_exts(slab);
> > +        /*
> > +         * Reset obj_exts to ensure all bits including OBJEXTS_ALLOC_FAIL
> > +         * are always cleared.
> > +         */
> > +        slab->obj_exts = 0;
> >         if (!obj_exts)
> >                 return;
> >
> >         /*
> >          * obj_exts was created with __GFP_NO_OBJ_EXT flag, therefore its
> >          * corresponding extension will be NULL. alloc_tag_sub() will throw a
> >          * warning if slab has extensions but the extension of an object is
> >          * NULL, therefore replace NULL with CODETAG_EMPTY to indicate that
> >          * the extension for obj_exts is expected to be NULL.
> >          */
> >         mark_objexts_empty(obj_exts);
> >         kfree(obj_exts);
> > -        slab->obj_exts = 0;
>
> You have an older base, check current mainline, we evaluate slab->obj_exts
> later in the function

Ah, sorry about that. Yeah, then this looks good.

>
> > }
> >
> >>
> >>         /*
> >>          * obj_exts was created with __GFP_NO_OBJ_EXT flag, therefore its
> >> --
> >> 2.25.1
> >>
>