[PATCH] fscache: fscache_set_key() - align alloc and usage

[PATCH] fscache: fscache_set_key() - align alloc and usage

[Tomas Bortoli]

The fscache_set_key() function allocates the buf pointer if index_key_len > sizeof(cookie->inline_key).
In such cases the allocated space might not be aligned with the pointer type.
This may result in an out-of-bound in the for-loop later in the same function, as the counter is rounded up.

Signed-off-by: Tomas Bortoli <tomasbortoli@gmail.com>
Reported-by: syzbot+a95b989b2dde8e806af8@syzkaller.appspotmail.com
---
 fs/fscache/cookie.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/fs/fscache/cookie.c b/fs/fscache/cookie.c
index 97137d7ec5ee..ed28bfb6a0fe 100644
--- a/fs/fscache/cookie.c
+++ b/fs/fscache/cookie.c
@@ -98,7 +98,7 @@ static int fscache_set_key(struct fscache_cookie *cookie,
 	cookie->key_len = index_key_len;
 
 	if (index_key_len > sizeof(cookie->inline_key)) {
-		buf = kzalloc(index_key_len, GFP_KERNEL);
+		buf = kzalloc(round_up(index_key_len, sizeof(u32)), GFP_KERNEL);
 		if (!buf)
 			return -ENOMEM;
 		cookie->key = buf;
-- 
2.11.0

Re: [PATCH] fscache: fscache_set_key() - align alloc and usage

[David Howells]

Tomas Bortoli <tomasbortoli@gmail.com> wrote:

> The fscache_set_key() function allocates the buf pointer if index_key_len >
> sizeof(cookie->inline_key).  In such cases the allocated space might not be
> aligned with the pointer type.  This may result in an out-of-bound in the
> for-loop later in the same function, as the counter is rounded up.

Yeah, it's good idea anyway since I should tell the allocator everything that
I expect to use - though kmalloc() will effectively rounds up the size to a
multiple of 8 anyway (ie. the smallest allocation granule is 8 bytes).

David

Re: [PATCH] fscache: fscache_set_key() - align alloc and usage

[Tomas Bortoli]

On 08/03/2018 03:49 PM, David Howells wrote:
> Tomas Bortoli <tomasbortoli@gmail.com> wrote:
> 
>> The fscache_set_key() function allocates the buf pointer if index_key_len >
>> sizeof(cookie->inline_key).  In such cases the allocated space might not be
>> aligned with the pointer type.  This may result in an out-of-bound in the
>> for-loop later in the same function, as the counter is rounded up.
> 
> Yeah, it's good idea anyway since I should tell the allocator everything that
> I expect to use - though kmalloc() will effectively rounds up the size to a
> multiple of 8 anyway (ie. the smallest allocation granule is 8 bytes).
> 
> David
> 

Yeah I forgot that :) at least KASAN won't complain anymore.

Tomas