[PATCH 4.9 35/51] media: vb2: vb2_mmap: move lock up

From: Greg Kroah-Hartman
Date: Mon Jan 21 2019 - 08:56:31 EST


4.9-stable review patch. If anyone has any objections, please let me know.

------------------

From: Hans Verkuil <hverkuil@xxxxxxxxx>

commit cd26d1c4d1bc947b56ae404998ae2276df7b39b7 upstream.

If a filehandle is dup()ped, then it is possible to close it from one fd
and call mmap from the other. This creates a race condition in vb2_mmap
where it is using queue data that __vb2_queue_free (called from close())
is in the process of releasing.

By moving up the mutex_lock(mmap_lock) in vb2_mmap this race is avoided
since __vb2_queue_free is called with the same mutex locked. So vb2_mmap
now reads consistent buffer data.

Signed-off-by: Hans Verkuil <hverkuil@xxxxxxxxx>
Reported-by: syzbot+be93025dd45dccd8923c@xxxxxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: Hans Verkuil <hansverk@xxxxxxxxx>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@xxxxxxxxxx>
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>

---
drivers/media/v4l2-core/videobuf2-core.c | 11 ++++++++---
1 file changed, 8 insertions(+), 3 deletions(-)

--- a/drivers/media/v4l2-core/videobuf2-core.c
+++ b/drivers/media/v4l2-core/videobuf2-core.c
@@ -1916,9 +1916,13 @@ int vb2_mmap(struct vb2_queue *q, struct
return -EINVAL;
}
}
+
+ mutex_lock(&q->mmap_lock);
+
if (vb2_fileio_is_active(q)) {
dprintk(1, "mmap: file io in progress\n");
- return -EBUSY;
+ ret = -EBUSY;
+ goto unlock;
}

/*
@@ -1926,7 +1930,7 @@ int vb2_mmap(struct vb2_queue *q, struct
*/
ret = __find_plane_by_offset(q, off, &buffer, &plane);
if (ret)
- return ret;
+ goto unlock;

vb = q->bufs[buffer];

@@ -1942,8 +1946,9 @@ int vb2_mmap(struct vb2_queue *q, struct
return -EINVAL;
}

- mutex_lock(&q->mmap_lock);
ret = call_memop(vb, mmap, vb->planes[plane].mem_priv, vma);
+
+unlock:
mutex_unlock(&q->mmap_lock);
if (ret)
return ret;