Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] vsock: connect() corrupts state and bytes_unsent on swept established socket
--
commit 1f8c2b3f7e5cf0bf0bba26600e236377dd6b1fe3
Author: Bobby Eshleman [off-list ref]
vsock: add a netlink command to assign the g2h device to a netns
This commit introduces a vsock generic netlink family with
VSOCK_CMD_DEV_NETNS_SET and VSOCK_CMD_DEV_NETNS_GET to assign the
guest-to-host vsock device to a network namespace and query its assignment.
Sockets that can no longer reach the device after an assignment are closed
with ECONNRESET.
Link: https://lore.kernel.org/all/20200427142518.uwssa6dtasrp3bfc@steredhat/ (local)
quoted hunk ↗ jump to hunk
diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
index c9b28e34b1b3c..7140a3c708389 100644
--- a/net/vmw_vsock/af_vsock.c
+++ b/net/vmw_vsock/af_vsock.c
[ ... ]
quoted hunk ↗ jump to hunk
@@ -742,8 +868,269 @@ bool vsock_find_cid(unsigned int cid)
[ ... ]
quoted hunk ↗ jump to hunk
+/* Close every connected socket of @t that can no longer reach the g2h device.
+ */
+static void vsock_g2h_reset_unreachable(const struct vsock_transport *t)
+{
+ struct vsock_sock *vsk;
+ unsigned int i = 0;
+ struct sock *sk;
+ bool reset;
+
+ lockdep_assert_held(&vsock_g2h_assign_mutex);
+ lockdep_assert_not_held(&vsock_register_mutex);
+
+ /* lock_sock() cannot be taken under vsock_table_lock, so pick the
+ * sockets one at a time, then recheck each under both locks.
+ */
+ while ((sk = vsock_next_connected_socket(t, &i,
+ vsock_g2h_should_reset))) {
+ vsk = vsock_sk(sk);
+
+ lock_sock(sk);
+
+ reset = false;
+ spin_lock_bh(&vsock_table_lock);
+ /* It may have been closed since we dropped the table lock. */
+ if (__vsock_in_connected_table(vsk) && vsk->transport == t &&
+ vsock_g2h_should_reset(vsk)) {
+ __vsock_remove_connected(vsk);
+ reset = true;
+ }
+ spin_unlock_bh(&vsock_table_lock);
+
+ if (reset) {
+ sock_set_flag(sk, SOCK_DONE);
+ WRITE_ONCE(vsk->peer_shutdown, SHUTDOWN_MASK);
+ sk->sk_state = TCP_CLOSE;
+ sk->sk_err = ECONNRESET;
+ sk_error_report(sk);
+ }
+
+ release_sock(sk);
+ sock_put(sk);
+ }
+}
[ ... ]
quoted hunk ↗ jump to hunk
@@ -1705,6 +2093,16 @@ static void vsock_connect_timeout(struct work_struct *work)
sock_put(sk);
}
+/* Avoid breaking the socket when another thread has changed its sock->state
+ * while we were asleep.
+ */
+static void vsock_connect_close(struct socket *sock)
+{
+ sock->sk->sk_state = TCP_CLOSE;
+ if (sock->state == SS_CONNECTING)
+ sock->state = SS_UNCONNECTED;
+}
+
static int vsock_connect(struct socket *sock, struct sockaddr_unsized *addr,
int addr_len, int flags)
{@@ -1863,8 +2261,7 @@ static int vsock_connect(struct socket *sock, struct sockaddr_unsized *addr,
* sk_state == TCP_SYN_SENT, which hereby we break.
* In such case VIRTIO_VSOCK_OP_RST will follow.
*/
- sk->sk_state = TCP_CLOSE;
- sock->state = SS_UNCONNECTED;
+ vsock_connect_close(sock);
/* Try to cancel VIRTIO_VSOCK_OP_REQUEST skb sent out by
* transport->connect().
@@ -1878,10 +2275,8 @@ static int vsock_connect(struct socket *sock, struct sockaddr_unsized *addr,
}
err = sock_error(sk);
- if (err) {
- sk->sk_state = TCP_CLOSE;
- sock->state = SS_UNCONNECTED;
- }
+ if (err)
+ vsock_connect_close(sock);
[Severity: Medium]
What happens in vsock_connect() when a socket reaches TCP_ESTABLISHED while
sleeping in schedule_timeout(), and is then swept by
vsock_g2h_reset_unreachable() before vsock_connect() reacquires
lock_sock(sk)?
While vsock_connect() has released lock_sock(sk) in schedule_timeout(),
virtio_transport_recv_connecting() can receive VIRTIO_VSOCK_OP_RESPONSE,
transition sk->sk_state to TCP_ESTABLISHED, set sock->state to SS_CONNECTED,
and insert vsk into vsock_connected_table. Once established, a concurrent
vsock_connectible_sendmsg() on another thread can acquire credits via
virtio_transport_get_credit() (incrementing vvs->bytes_unsent) and enqueue
VIRTIO_VSOCK_OP_RW skbs onto send_pkt_queue.
If VSOCK_CMD_DEV_NETNS_SET then runs vsock_g2h_reset_unreachable() before
vsock_connect() reacquires lock_sock(sk), it sets sk->sk_state to TCP_CLOSE
and sk->sk_err to ECONNRESET while leaving sock->state as SS_CONNECTED. When
vsock_connect() wakes up, the once-established check:
if (sk->sk_state == TCP_ESTABLISHED ||
sk->sk_state == TCP_CLOSING) {
err = 0;
goto out_wait;
}
does not match because sk->sk_state is TCP_CLOSE rather than TCP_CLOSING (unlike
virtio_transport_do_close()) and sock->state == SS_CONNECTED is not checked.
If signal_pending(current) || timeout == 0 holds, vsock_connect() proceeds to
call vsock_transport_cancel_pkt(vsk) -> virtio_transport_purge_skbs(), which
purges the post-connection VIRTIO_VSOCK_OP_RW skbs without calling
virtio_transport_consume_skb_sent() to decrement vvs->bytes_unsent. Doesn't
this leak vvs->bytes_unsent and cause vsock_linger() on close() to block until
sk_lingertime expires, similar to the issue fixed by commit 002541ef650b
("vsock: Do not allow signal/timeout on established sockets")?
And if no signal or timeout is pending, vsock_connect() exits the wait loop,
consumes sk->sk_err via sock_error(sk), calls vsock_connect_close(sock), and
returns -ECONNRESET while leaving sock->state as SS_CONNECTED with sk->sk_err
cleared to 0.
Should vsock_g2h_reset_unreachable() set sk->sk_state = TCP_CLOSING like
virtio_transport_do_close(), or should vsock_connect() also check for
sock->state == SS_CONNECTED after schedule_timeout()?
--
Sashiko AI review · https://sashiko.dev/#/patchset/20261103-vsock-guest-ns-v5-0-48179b9b66b8@meta.com?part=4