archive mirror
 help / color / mirror / Atom feed
Subject: [PATCH] mm: slub: re-initialize randomized freelist sequence in calculate_sizes
Date: Sat,  8 Aug 2020 13:50:30 +0400	[thread overview]
Message-ID: <> (raw)

From: Sahara <>

Slab cache flags are exported to sysfs and are allowed to get modified
from userspace. Some of those may call calculate_sizes function because
the changed flag can take an effect on slab object size and layout,
which means kmem_cache may have different order and objects.
The freelist pointer corruption occurs if some slab flags are modified

 $ echo 0 > /sys/kernel/slab/zs_handle/store_user
 $ echo 0 > /sys/kernel/slab/zspage/store_user
 $ mkswap /dev/block/zram0
 $ swapon /dev/block/zram0 -p 32758

 BUG zs_handle (Not tainted): Freepointer corrupt

 Disabling lock debugging due to kernel taint
 INFO: Slab 0xffffffbf29603600 objects=102 used=102 fp=0x0000000000000000 flags=0x0200
 INFO: Object 0xffffffca580d8d78 @offset=3448 fp=0xffffffca580d8ed0

 Redzone 00000000f3cddd6c: bb bb bb bb bb bb bb bb                          ........
 Object 0000000082d5d74e: 6b 6b 6b 6b 6b 6b 6b a5                          kkkkkkk.
 Redzone 000000008fd80359: bb bb bb bb bb bb bb bb                          ........
 Padding 00000000c7f56047: 5a 5a 5a 5a 5a 5a 5a 5a                          ZZZZZZZZ

In this example, an Android device tries to use zram as a swap and to
turn off store_user in order to reduce the slub object size.
When calculate_sizes is called in kmem_cache_open, size, order and
objects for zs_handle is:
 size:360, order:0, objects:22
However, if the SLAB_STORE_USER bit is cleared in store_user_store:
 size: 56, order:1, objects:73

All the size, order, and objects is changed by calculate_sizes(), but
the size of the random_seq array is still old value(22). As a result,
out-of-bound array access can occur at shuffle_freelist() when slab
allocation is requested.

This patch fixes the problem by re-allocating the random_seq array
with re-calculated correct objects value.

Fixes: 210e7a43fa905 ("mm: SLUB freelist randomization")
Reported-by: Ari-Pekka Verta <>
Reported-by: Timo Simola <>
Signed-off-by: Sahara <>
 mm/slub.c | 23 ++++++++++++++++-------
 1 file changed, 16 insertions(+), 7 deletions(-)

diff --git a/mm/slub.c b/mm/slub.c
index f226d66408ee..be1e4d6682b8 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -3704,7 +3704,22 @@ static int calculate_sizes(struct kmem_cache *s, int forced_order)
 	if (oo_objects(s->oo) > oo_objects(s->max))
 		s->max = s->oo;
-	return !!oo_objects(s->oo);
+	if (!oo_objects(s->oo))
+		return 0;
+	/*
+	 * Initialize the pre-computed randomized freelist if slab is up.
+	 * If the randomized freelist random_seq is already initialized,
+	 * free and re-initialize it with re-calculated value.
+	 */
+	if (slab_state >= UP) {
+		if (s->random_seq)
+			cache_random_seq_destroy(s);
+		if (init_cache_random_seq(s))
+			return 0;
+	}
+	return 1;
 static int kmem_cache_open(struct kmem_cache *s, slab_flags_t flags)
@@ -3748,12 +3763,6 @@ static int kmem_cache_open(struct kmem_cache *s, slab_flags_t flags)
 	s->remote_node_defrag_ratio = 1000;
-	/* Initialize the pre-computed randomized freelist if slab is up */
-	if (slab_state >= UP) {
-		if (init_cache_random_seq(s))
-			goto error;
-	}
 	if (!init_kmem_cache_nodes(s))
 		goto error;

             reply	other threads:[~2020-08-08  9:50 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-08-08  9:50 kpark3469 [this message]
2020-08-19 20:26 ` Andrew Morton
2020-08-19 20:52   ` Thomas Garnier
2020-08-19 20:29 ` Kees Cook

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \ \ \ \ \ \ \ \ \ \
    --subject='Re: [PATCH] mm: slub: re-initialize randomized freelist sequence in calculate_sizes' \

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).