sock_init_data() assumes that the `struct socket` passed in input is
contained in a `struct socket_alloc` allocated with sock_alloc().
However, tap_open() and tun_chr_open() pass a `struct socket` embedded
in a `struct tap_queue` and `struct tun_file` respectively, both
allocated with sk_alloc().
This causes a type confusion when issuing a container_of() with
SOCK_INODE() in sock_init_data() which results in assigning a wrong
sk_uid to the `struct sock` in input.
Due to the type confusion, both sockets happen to have their uid set
to 0, i.e. root.
While it will be often correct, as tuntap devices require
CAP_NET_ADMIN, it may not always be the case.
Not sure how widespread is the impact of this, it seems the socket uid
may be used for network filtering and routing, thus tuntap sockets may
be incorrectly managed.
Additionally, it seems a socket with an incorrect uid may be returned
to the vhost driver when issuing a get_socket() on a tuntap device in
vhost_net_set_backend().
The proposed fix may not be the cleanest one, as it simply overrides
the incorrect uid after the type confusion in sock_init_data()
happens.
While minimal, this may not be solid in case more logic relying on
SOCK_INODE() is added to sock_init_data().
The alternative fix would be to pass a NULL sock, and manually perform
the assignments after the sock_init_data() call:
```
sk_set_socket(sk, sock);
// and
sk->sk_type = sock->type;
RCU_INIT_POINTER(sk->sk_wq, &sock->wq);
sock->sk = sk;
sk->sk_uid = SOCK_INODE(sock)->i_uid;
```
Signed-off-by: Pietro Borrello <[email protected]>
---
Changes in v2:
- Shorten and format comments
- Link to v1: https://lore.kernel.org/r/[email protected]
---
Pietro Borrello (2):
tun: tun_chr_open(): correctly initialize socket uid
tap: tap_open(): correctly initialize socket uid
drivers/net/tap.c | 4 ++++
drivers/net/tun.c | 5 +++++
2 files changed, 9 insertions(+)
---
base-commit: 6d796c50f84ca79f1722bb131799e5a5710c4700
change-id: 20230131-tuntap-sk-uid-78efc80f4b82
Best regards,
--
Pietro Borrello <[email protected]>
sock_init_data() assumes that the `struct socket` passed in input is
contained in a `struct socket_alloc` allocated with sock_alloc().
However, tap_open() passes a `struct socket` embedded in a `struct
tap_queue` allocated with sk_alloc().
This causes a type confusion when issuing a container_of() with
SOCK_INODE() in sock_init_data() which results in assigning a wrong
sk_uid to the `struct sock` in input.
On default configuration, the type confused field overlaps with
padding bytes between `int vnet_hdr_sz` and `struct tap_dev __rcu
*tap` in `struct tap_queue`, which makes the uid of all tap sockets 0,
i.e., the root one. Fix the assignment by overriding it with the
correct uid.
Fixes: 86741ec25462 ("net: core: Add a UID field to struct sock.")
Signed-off-by: Pietro Borrello <[email protected]>
---
drivers/net/tap.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/net/tap.c b/drivers/net/tap.c
index a2be1994b389..145c3f84fac4 100644
--- a/drivers/net/tap.c
+++ b/drivers/net/tap.c
@@ -538,6 +538,10 @@ static int tap_open(struct inode *inode, struct file *file)
q->sk.sk_destruct = tap_sock_destruct;
q->flags = IFF_VNET_HDR | IFF_NO_PI | IFF_TAP;
q->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
+ /* Assign sk_uid from the inode argument, since q->sock
+ * passed to sock_init_data() has no corresponding inode
+ */
+ q->sk.sk_uid = inode->i_uid;
/*
* so far only KVM virtio_net uses tap, enable zero copy between
--
2.25.1