Re: [PATCH v2] kasan: add test for SLAB_TYPESAFE_BY_RCU quarantine skipping

[Jann Horn]

On Thu, Aug 14, 2025 at 7:10 AM Andrey Konovalov <andreyknvl@xxxxxxxxx> wrote:
> On Tue, Jul 29, 2025 at 6:49 PM Jann Horn <jannh@xxxxxxxxxx> wrote:
> > Verify that KASAN does not quarantine objects in SLAB_TYPESAFE_BY_RCU slabs
> > if CONFIG_SLUB_RCU_DEBUG is off.
> >
> > Signed-off-by: Jann Horn <jannh@xxxxxxxxxx>
> > ---
> > changes in v2:
> >  - disable migration to ensure that all SLUB operations use the same
> >    percpu state (vbabka)
> >  - use EXPECT instead of ASSERT for pointer equality check so that
> >    expectation failure doesn't terminate the test with migration still
> >    disabled
> > ---
> >  mm/kasan/kasan_test_c.c | 38 ++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 38 insertions(+)
> >
> > diff --git a/mm/kasan/kasan_test_c.c b/mm/kasan/kasan_test_c.c
> > index 5f922dd38ffa..0d50402d492c 100644
> > --- a/mm/kasan/kasan_test_c.c
> > +++ b/mm/kasan/kasan_test_c.c
> > @@ -1073,6 +1073,43 @@ static void kmem_cache_rcu_uaf(struct kunit *test)
> >         kmem_cache_destroy(cache);
> >  }
> >
> > +/*
> > + * Check that SLAB_TYPESAFE_BY_RCU objects are immediately reused when
> > + * CONFIG_SLUB_RCU_DEBUG is off, and stay at the same address.
>
> Would be great to also add an explanation of why we want to test for
> this (or a reference to the related fix commit?).

Okay, I'll add a sentence here, will send v3 in a bit.

> > + */
> > +static void kmem_cache_rcu_reuse(struct kunit *test)
> > +{
> > +       char *p, *p2;
> > +       struct kmem_cache *cache;
> > +
> > +       KASAN_TEST_NEEDS_CONFIG_OFF(test, CONFIG_SLUB_RCU_DEBUG);
> > +
> > +       cache = kmem_cache_create("test_cache", 16, 0, SLAB_TYPESAFE_BY_RCU,
> > +                                 NULL);
> > +       KUNIT_ASSERT_NOT_ERR_OR_NULL(test, cache);
> > +
> > +       migrate_disable();
> > +       p = kmem_cache_alloc(cache, GFP_KERNEL);
> > +       if (!p) {
> > +               kunit_err(test, "Allocation failed: %s\n", __func__);
> > +               goto out;
> > +       }
> > +
> > +       kmem_cache_free(cache, p);
> > +       p2 = kmem_cache_alloc(cache, GFP_KERNEL);
> > +       if (!p2) {
> > +               kunit_err(test, "Allocation failed: %s\n", __func__);
> > +               goto out;
> > +       }
> > +       KUNIT_EXPECT_PTR_EQ(test, p, p2);
>
> I think this might fail for the HW_TAGS mode? The location will be
> reused, but the tag will be different.

No, it's a SLAB_TYPESAFE_BY_RCU cache, so the tag can't really be
different. poison_slab_object() will bail out, and assign_tag() will
reuse the already-assigned tag.