Re: [PATCH mm] kasan: stop leaking stack trace handles

[Andrew Morton]

On Wed, 27 Dec 2023 22:42:40 +0100 Andrey Konovalov <andreyknvl@xxxxxxxxx> wrote:

> On Wed, Dec 27, 2023 at 10:23 PM Andrew Morton
> <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > Thanks, I added this fix:
> >
> > --- a/mm/kasan/generic.c~kasan-stop-leaking-stack-trace-handles-fix
> > +++ a/mm/kasan/generic.c
> > @@ -503,7 +503,7 @@ void kasan_init_object_meta(struct kmem_
> >          */
> >  }
> >
> > -void release_alloc_meta(struct kasan_alloc_meta *meta)
> > +static void release_alloc_meta(struct kasan_alloc_meta *meta)
> >  {
> >         /* Evict the stack traces from stack depot. */
> >         stack_depot_put(meta->alloc_track.stack);
> > @@ -514,7 +514,7 @@ void release_alloc_meta(struct kasan_all
> >         __memset(meta, 0, sizeof(*meta));
> >  }
> >
> > -void release_free_meta(const void *object, struct kasan_free_meta *meta)
> > +static void release_free_meta(const void *object, struct kasan_free_meta *meta)
> >  {
> >         /* Check if free meta is valid. */
> >         if (*(u8 *)kasan_mem_to_shadow(object) != KASAN_SLAB_FREE_META)
> > _
> >
> 
> Could you mark them as "static inline" even?

That's rather old-fashioned.  Nowadays gcc is supposed to work out
whether or not to inline things, and we override that with noinline and
__always_inline.