mm/slab: validate cache membership under freelist hardening
authorKees Cook <keescook@chromium.org>
Fri, 12 Jul 2019 03:53:23 +0000 (20:53 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 12 Jul 2019 18:05:41 +0000 (11:05 -0700)
Patch series "mm/slab: Improved sanity checking".

This adds defenses against slab cache confusion (as seen in real-world
exploits[1]) and gracefully handles type confusions when trying to look
up slab caches from an arbitrary page.  (Also is patch 3: new LKDTM
tests for these defenses as well as for the existing double-free
detection.

This patch (of 3):

When building under CONFIG_SLAB_FREELIST_HARDENING, it makes sense to
perform sanity-checking on the assumed slab cache during
kmem_cache_free() to make sure the kernel doesn't mix freelists across
slab caches and corrupt memory (as seen in the exploitation of flaws
like CVE-2018-9568[1]).  Note that the prior code might WARN() but still
corrupt memory (i.e.  return the assumed cache instead of the owned
cache).

There is no noticeable performance impact (changes are within noise).
Measuring parallel kernel builds, I saw the following with
CONFIG_SLAB_FREELIST_HARDENED, before and after this patch:

before:

Run times: 288.85 286.53 287.09 287.07 287.21
Min: 286.53 Max: 288.85 Mean: 287.35 Std Dev: 0.79

after:

Run times: 289.58 287.40 286.97 287.20 287.01
Min: 286.97 Max: 289.58 Mean: 287.63 Std Dev: 0.99

Delta: 0.1% which is well below the standard deviation

[1] https://github.com/ThomasKing2014/slides/raw/master/Building%20universal%20Android%20rooting%20with%20a%20type%20confusion%20vulnerability.pdf

Link: http://lkml.kernel.org/r/20190530045017.15252-2-keescook@chromium.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Alexander Popov <alex.popov@linux.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/slab.h

index 43ac818..4dafae2 100644 (file)
--- a/mm/slab.h
+++ b/mm/slab.h
@@ -310,7 +310,7 @@ static inline bool is_root_cache(struct kmem_cache *s)
 static inline bool slab_equal_or_root(struct kmem_cache *s,
                                      struct kmem_cache *p)
 {
-       return true;
+       return s == p;
 }
 
 static inline const char *cache_name(struct kmem_cache *s)
@@ -363,18 +363,16 @@ static inline struct kmem_cache *cache_from_obj(struct kmem_cache *s, void *x)
         * will also be a constant.
         */
        if (!memcg_kmem_enabled() &&
+           !IS_ENABLED(CONFIG_SLAB_FREELIST_HARDENED) &&
            !unlikely(s->flags & SLAB_CONSISTENCY_CHECKS))
                return s;
 
        page = virt_to_head_page(x);
        cachep = page->slab_cache;
-       if (slab_equal_or_root(cachep, s))
-               return cachep;
-
-       pr_err("%s: Wrong slab cache. %s but object is from %s\n",
-              __func__, s->name, cachep->name);
-       WARN_ON_ONCE(1);
-       return s;
+       WARN_ONCE(!slab_equal_or_root(cachep, s),
+                 "%s: Wrong slab cache. %s but object is from %s\n",
+                 __func__, s->name, cachep->name);
+       return cachep;
 }
 
 static inline size_t slab_ksize(const struct kmem_cache *s)