Avoid false KMSAN negatives with SLUB_DEBUG by allowing
kmsan_slab_free() to poison the freed memory, and by preventing
init_object() from unpoisoning new allocations by using __memset().
There are two alternatives to this approach. First, init_object()
can be marked with __no_sanitize_memory. This annotation should be used
with great care, because it drops all instrumentation from the
function, and any shadow writes will be lost. Even though this is not a
concern with the current init_object() implementation, this may change
in the future.
Second, kmsan_poison_memory() calls may be added after memset() calls.
The downside is that init_object() is called from
free_debug_processing(), in which case poisoning will erase the
distinction between simply uninitialized memory and UAF.
Signed-off-by: Ilya Leoshkevich <[email protected]>
---
mm/kmsan/hooks.c | 2 +-
mm/slub.c | 13 +++++++++----
2 files changed, 10 insertions(+), 5 deletions(-)
diff --git a/mm/kmsan/hooks.c b/mm/kmsan/hooks.c
index 267d0afa2e8b..26d86dfdc819 100644
--- a/mm/kmsan/hooks.c
+++ b/mm/kmsan/hooks.c
@@ -74,7 +74,7 @@ void kmsan_slab_free(struct kmem_cache *s, void *object)
return;
/* RCU slabs could be legally used after free within the RCU period */
- if (unlikely(s->flags & (SLAB_TYPESAFE_BY_RCU | SLAB_POISON)))
+ if (unlikely(s->flags & SLAB_TYPESAFE_BY_RCU))
return;
/*
* If there's a constructor, freed memory must remain in the same state
diff --git a/mm/slub.c b/mm/slub.c
index 1373ac365a46..4dd55cabe701 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -1139,7 +1139,12 @@ static void init_object(struct kmem_cache *s, void *object, u8 val)
unsigned int poison_size = s->object_size;
if (s->flags & SLAB_RED_ZONE) {
- memset(p - s->red_left_pad, val, s->red_left_pad);
+ /*
+ * Use __memset() here and below in order to avoid overwriting
+ * the KMSAN shadow. Keeping the shadow makes it possible to
+ * distinguish uninit-value from use-after-free.
+ */
+ __memset(p - s->red_left_pad, val, s->red_left_pad);
if (slub_debug_orig_size(s) && val == SLUB_RED_ACTIVE) {
/*
@@ -1152,12 +1157,12 @@ static void init_object(struct kmem_cache *s, void *object, u8 val)
}
if (s->flags & __OBJECT_POISON) {
- memset(p, POISON_FREE, poison_size - 1);
- p[poison_size - 1] = POISON_END;
+ __memset(p, POISON_FREE, poison_size - 1);
+ __memset(p + poison_size - 1, POISON_END, 1);
}
if (s->flags & SLAB_RED_ZONE)
- memset(p + poison_size, val, s->inuse - poison_size);
+ __memset(p + poison_size, val, s->inuse - poison_size);
}
static void restore_bytes(struct kmem_cache *s, char *message, u8 data,
--
2.45.1
Hi Ilya,
On Thu, 13 Jun 2024 17:34:14 +0200 Ilya Leoshkevich <[email protected]> wrote:
> Avoid false KMSAN negatives with SLUB_DEBUG by allowing
> kmsan_slab_free() to poison the freed memory, and by preventing
> init_object() from unpoisoning new allocations by using __memset().
>
> There are two alternatives to this approach. First, init_object()
> can be marked with __no_sanitize_memory. This annotation should be used
> with great care, because it drops all instrumentation from the
> function, and any shadow writes will be lost. Even though this is not a
> concern with the current init_object() implementation, this may change
> in the future.
>
> Second, kmsan_poison_memory() calls may be added after memset() calls.
> The downside is that init_object() is called from
> free_debug_processing(), in which case poisoning will erase the
> distinction between simply uninitialized memory and UAF.
>
> Signed-off-by: Ilya Leoshkevich <[email protected]>
> ---
> mm/kmsan/hooks.c | 2 +-
> mm/slub.c | 13 +++++++++----
> 2 files changed, 10 insertions(+), 5 deletions(-)
>
[...]
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -1139,7 +1139,12 @@ static void init_object(struct kmem_cache *s, void *object, u8 val)
> unsigned int poison_size = s->object_size;
>
> if (s->flags & SLAB_RED_ZONE) {
> - memset(p - s->red_left_pad, val, s->red_left_pad);
> + /*
> + * Use __memset() here and below in order to avoid overwriting
> + * the KMSAN shadow. Keeping the shadow makes it possible to
> + * distinguish uninit-value from use-after-free.
> + */
> + __memset(p - s->red_left_pad, val, s->red_left_pad);
I found my build test[1] fails with below error on latest mm-unstable branch.
'git bisect' points me this patch.
CC mm/slub.o
/mm/slub.c: In function 'init_object':
/mm/slub.c:1147:17: error: implicit declaration of function '__memset'; did you mean 'memset'? [-Werror=implicit-function-declaration]
1147 | __memset(p - s->red_left_pad, val, s->red_left_pad);
| ^~~~~~~~
| memset
cc1: some warnings being treated as errors
I haven't looked in deep, but reporting first. Do you have any idea?
[1] https://github.com/awslabs/damon-tests/blob/next/corr/tests/build_m68k.sh
Thanks,
SJ
[...]
On Thu, 2024-06-13 at 16:30 -0700, SeongJae Park wrote:
> Hi Ilya,
>
> On Thu, 13 Jun 2024 17:34:14 +0200 Ilya Leoshkevich
> <[email protected]> wrote:
>
> > Avoid false KMSAN negatives with SLUB_DEBUG by allowing
> > kmsan_slab_free() to poison the freed memory, and by preventing
> > init_object() from unpoisoning new allocations by using __memset().
> >
> > There are two alternatives to this approach. First, init_object()
> > can be marked with __no_sanitize_memory. This annotation should be
> > used
> > with great care, because it drops all instrumentation from the
> > function, and any shadow writes will be lost. Even though this is
> > not a
> > concern with the current init_object() implementation, this may
> > change
> > in the future.
> >
> > Second, kmsan_poison_memory() calls may be added after memset()
> > calls.
> > The downside is that init_object() is called from
> > free_debug_processing(), in which case poisoning will erase the
> > distinction between simply uninitialized memory and UAF.
> >
> > Signed-off-by: Ilya Leoshkevich <[email protected]>
> > ---
> > mm/kmsan/hooks.c | 2 +-
> > mm/slub.c | 13 +++++++++----
> > 2 files changed, 10 insertions(+), 5 deletions(-)
> >
> [...]
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -1139,7 +1139,12 @@ static void init_object(struct kmem_cache
> > *s, void *object, u8 val)
> > unsigned int poison_size = s->object_size;
> >
> > if (s->flags & SLAB_RED_ZONE) {
> > - memset(p - s->red_left_pad, val, s->red_left_pad);
> > + /*
> > + * Use __memset() here and below in order to avoid
> > overwriting
> > + * the KMSAN shadow. Keeping the shadow makes it
> > possible to
> > + * distinguish uninit-value from use-after-free.
> > + */
> > + __memset(p - s->red_left_pad, val, s-
> > >red_left_pad);
>
> I found my build test[1] fails with below error on latest mm-unstable
> branch.
> 'git bisect' points me this patch.
>
> CC mm/slub.o
> /mm/slub.c: In function 'init_object':
> /mm/slub.c:1147:17: error: implicit declaration of function
> '__memset'; did you mean 'memset'? [-Werror=implicit-function-
> declaration]
> 1147 | __memset(p - s->red_left_pad, val, s-
> >red_left_pad);
> | ^~~~~~~~
> | memset
> cc1: some warnings being treated as errors
>
> I haven't looked in deep, but reporting first. Do you have any idea?
>
> [1]
> https://github.com/awslabs/damon-tests/blob/next/corr/tests/build_m68k.sh
>
>
> Thanks,
> SJ
>
> [...]
Thanks for the report.
Apparently not all architectures have __memset(). We should probably go
back to memset_no_sanitize_memory() [1], but this time mark it with
noinline __maybe_unused __no_sanitize_memory, like it's done in, e.g.,
32/35.
Alexander, what do you think?
[1]
https://lore.kernel.org/lkml/[email protected]/