mm/madvise: remove racy mm ownership check
authorMinchan Kim <minchan@kernel.org>
Wed, 9 Dec 2020 04:57:18 +0000 (20:57 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 9 Dec 2020 04:57:18 +0000 (20:57 -0800)
Jann spotted the security hole due to race of mm ownership check.

If the task is sharing the mm_struct but goes through execve() before
mm_access(), it could skip process_madvise_behavior_valid check.  That
makes *any advice hint* to reach into the remote process.

This patch removes the mm ownership check.  With it, it will lose the
ability that local process could give *any* advice hint with vector
interface for some reason (e.g., performance).  Since there is no
concrete example in upstream yet, it would be better to remove the
abiliity at this moment and need to review when such new advice comes
up.

Fixes: ecb8ac8b1f14 ("mm/madvise: introduce process_madvise() syscall: an external memory hinting API")
Reported-by: Jann Horn <jannh@google.com>
Suggested-by: Jann Horn <jannh@google.com>
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/madvise.c

index a8d8d48a57fe968fe837f6432c90280ac9f35f15..13f5677b93222ac15dc96167bc9a3d0db9acd4e2 100644 (file)
@@ -1204,8 +1204,7 @@ SYSCALL_DEFINE5(process_madvise, int, pidfd, const struct iovec __user *, vec,
                goto put_pid;
        }
 
-       if (task->mm != current->mm &&
-                       !process_madvise_behavior_valid(behavior)) {
+       if (!process_madvise_behavior_valid(behavior)) {
                ret = -EINVAL;
                goto release_task;
        }