[PATCH v5 5/5] mm/kasan: Add object validation in ksize()

[Marco Elver]

ksize() has been unconditionally unpoisoning the whole shadow memory region
associated with an allocation. This can lead to various undetected bugs,
for example, double-kzfree().

Specifically, kzfree() uses ksize() to determine the actual allocation
size, and subsequently zeroes the memory. Since ksize() used to just
unpoison the whole shadow memory region, no invalid free was detected.

This patch addresses this as follows:

1. Add a check in ksize(), and only then unpoison the memory region.

2. Preserve kasan_unpoison_slab() semantics by explicitly unpoisoning
   the shadow memory region using the size obtained from __ksize().

Tested:
1. With SLAB allocator: a) normal boot without warnings; b) verified the
   added double-kzfree() is detected.
2. With SLUB allocator: a) normal boot without warnings; b) verified the
   added double-kzfree() is detected.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=199359
Signed-off-by: Marco Elver <elver@xxxxxxxxxx>
Acked-by: Kees Cook <keescook@xxxxxxxxxxxx>
Cc: Andrey Ryabinin <aryabinin@xxxxxxxxxxxxx>
Cc: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
Cc: Alexander Potapenko <glider@xxxxxxxxxx>
Cc: Andrey Konovalov <andreyknvl@xxxxxxxxxx>
Cc: Christoph Lameter <cl@xxxxxxxxx>
Cc: Pekka Enberg <penberg@xxxxxxxxxx>
Cc: David Rientjes <rientjes@xxxxxxxxxx>
Cc: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Cc: Mark Rutland <mark.rutland@xxxxxxx>
Cc: Kees Cook <keescook@xxxxxxxxxxxx>
Cc: kasan-dev@xxxxxxxxxxxxxxxx
Cc: linux-kernel@xxxxxxxxxxxxxxx
Cc: linux-mm@xxxxxxxxx
---
v4:
* Prefer WARN_ON_ONCE() instead of BUG_ON().
---
 include/linux/kasan.h |  7 +++++--
 mm/slab_common.c      | 22 +++++++++++++++++++++-
 2 files changed, 26 insertions(+), 3 deletions(-)

diff --git a/include/linux/kasan.h b/include/linux/kasan.h
index b40ea104dd36..cc8a03cc9674 100644
--- a/include/linux/kasan.h
+++ b/include/linux/kasan.h
@@ -76,8 +76,11 @@ void kasan_free_shadow(const struct vm_struct *vm);
 int kasan_add_zero_shadow(void *start, unsigned long size);
 void kasan_remove_zero_shadow(void *start, unsigned long size);
 
-size_t ksize(const void *);
-static inline void kasan_unpoison_slab(const void *ptr) { ksize(ptr); }
+size_t __ksize(const void *);
+static inline void kasan_unpoison_slab(const void *ptr)
+{
+	kasan_unpoison_shadow(ptr, __ksize(ptr));
+}
 size_t kasan_metadata_size(struct kmem_cache *cache);
 
 bool kasan_save_enable_multi_shot(void);
diff --git a/mm/slab_common.c b/mm/slab_common.c
index b7c6a40e436a..a09bb10aa026 100644
--- a/mm/slab_common.c
+++ b/mm/slab_common.c
@@ -1613,7 +1613,27 @@ EXPORT_SYMBOL(kzfree);
  */
 size_t ksize(const void *objp)
 {
-	size_t size = __ksize(objp);
+	size_t size;
+
+	if (WARN_ON_ONCE(!objp))
+		return 0;
+	/*
+	 * We need to check that the pointed to object is valid, and only then
+	 * unpoison the shadow memory below. We use __kasan_check_read(), to
+	 * generate a more useful report at the time ksize() is called (rather
+	 * than later where behaviour is undefined due to potential
+	 * use-after-free or double-free).
+	 *
+	 * If the pointed to memory is invalid we return 0, to avoid users of
+	 * ksize() writing to and potentially corrupting the memory region.
+	 *
+	 * We want to perform the check before __ksize(), to avoid potentially
+	 * crashing in __ksize() due to accessing invalid metadata.
+	 */
+	if (unlikely(objp == ZERO_SIZE_PTR) || !__kasan_check_read(objp, 1))
+		return 0;
+
+	size = __ksize(objp);
 	/*
 	 * We assume that ksize callers could use whole allocated area,
 	 * so we need to unpoison this area.
-- 
2.22.0.410.gd8fdbe21b5-goog