On Wed, Nov 19, 2025 at 03:09:25PM +0100, Michal Luczaj wrote:
On 11/19/25 11:36, Stefano Garzarella wrote:
On Tue, Nov 18, 2025 at 11:02:17PM +0100, Michal Luczaj wrote:
On 11/18/25 10:51, Stefano Garzarella wrote:
On Mon, Nov 17, 2025 at 09:57:25PM +0100, Michal Luczaj wrote:
...
+static void vsock_reset_interrupted(struct sock *sk)
+{
+ struct vsock_sock *vsk = vsock_sk(sk);
+
+ /* Try to cancel VIRTIO_VSOCK_OP_REQUEST skb sent out by
+ * transport->connect().
+ */
+ vsock_transport_cancel_pkt(vsk);
+
+ /* Listener might have already responded with VIRTIO_VSOCK_OP_RESPONSE.
+ * Its handling expects our sk_state == TCP_SYN_SENT, which hereby we
+ * break. In such case VIRTIO_VSOCK_OP_RST will follow.
+ */
+ sk->sk_state = TCP_CLOSE;
+ sk->sk_socket->state = SS_UNCONNECTED;
+}
+
static int vsock_connect(struct socket *sock, struct sockaddr *addr,
int addr_len, int flags)
{
@@ -1661,18 +1678,33 @@ static int vsock_connect(struct socket *sock, struct
sockaddr *addr,
timeout = schedule_timeout(timeout);
lock_sock(sk);
+ /* Connection established. Whatever happens to socket once we
+ * release it, that's not connect()'s concern. No need to go
+ * into signal and timeout handling. Call it a day.
+ *
+ * Note that allowing to "reset" an already established socket
+ * here is racy and insecure.
+ */
+ if (sk->sk_state == TCP_ESTABLISHED)
+ break;
+
+ /* If connection was _not_ established and a signal/timeout came
+ * to be, we want the socket's state reset. User space may want
+ * to retry.
+ *
+ * sk_state != TCP_ESTABLISHED implies that socket is not on
+ * vsock_connected_table. We keep the binding and the transport
+ * assigned.
+ */
if (signal_pending(current)) {
err = sock_intr_errno(timeout);
- sk->sk_state = sk->sk_state == TCP_ESTABLISHED ?
TCP_CLOSING : TCP_CLOSE;
- sock->state = SS_UNCONNECTED;
- vsock_transport_cancel_pkt(vsk);
- vsock_remove_connected(vsk);
+ vsock_reset_interrupted(sk);
goto out_wait;
- } else if ((sk->sk_state != TCP_ESTABLISHED) && (timeout == 0))
{
+ }
+
+ if (timeout == 0) {
err = -ETIMEDOUT;
- sk->sk_state = TCP_CLOSE;
- sock->state = SS_UNCONNECTED;
- vsock_transport_cancel_pkt(vsk);
+ vsock_reset_interrupted(sk);
goto out_wait;
I'm fine with the change, but now both code blocks are the same, so
can we unify them?
I mean something like this:
if (signal_pending(current) || timeout == 0 {
err = timeout == 0 ? -ETIMEDOUT :
sock_intr_errno(timeout);
...
}
Maybe at that point we can also remove the vsock_reset_interrupted()
function and put the code right there.
BTW I don't have a strong opinion, what do you prefer?
Sure, no problem.
But I've realized invoking `sock_intr_errno(timeout)` is unnecessary.
`timeout` can't be MAX_SCHEDULE_TIMEOUT, so the call always evaluates to
-EINTR, right?
IIUC currently schedule_timeout() can return MAX_SCHEDULE_TIMEOUT only
if it was called with that parameter, and I think we never call it in
that way, so I'd agree with you.
My only concern is if it's true for all the stable branches we will
backport this patch.
I would probably touch it as little as possible and continue using
sock_intr_errno() for now, but if you verify that it has always been
that way, then it's fine to change it.
All right then, here's a v2 with minimum changes:
https://lore.kernel.org/netdev/[email protected]/
Thanks!
Note a detail though: should signal and timeout happen at the same time,
now it's the timeout errno returned.
Yeah, I thought about that, but I don't see any problems with that.
I mean, it's something that if it happens, it's still not deterministic,
so we're not really changing anything.
Thanks,
Stefano
