Re: [PATCH net-next v18 12/25] ovpn: implement TCP transport

On 03/02/2025 11:05, Sabrina Dubroca wrote:

2025-01-13, 10:31:31 +0100, Antonio Quartulli wrote:

quoted

+static void ovpn_tcp_rcv(struct strparser *strp, struct sk_buff *skb)
+{

[...]

quoted

+	/* we need the first byte of data to be accessible
+	 * to extract the opcode and the key ID later on
+	 */
+	if (!pskb_may_pull(skb, 1)) {

make sure we have 1B...

quoted

+		net_warn_ratelimited("%s: packet too small to fetch opcode for peer %u\n",
+				     netdev_name(peer->ovpn->dev), peer->id);
+		goto err;
+	}
+
+	/* DATA_V2 packets are handled in kernel, the rest goes to user space */
+	opcode = ovpn_opcode_from_skb(skb, 0);

but this reads a u32 (4B) from skb->data

ACK, hand-in-hand with the comment not updated.
Will fix this.

[...]

quoted

+void ovpn_tcp_socket_detach(struct ovpn_socket *ovpn_sock)
+{
+	struct ovpn_peer *peer = ovpn_sock->peer;
+	struct socket *sock = ovpn_sock->sock;
+
+	strp_stop(&peer->tcp.strp);
+
+	skb_queue_purge(&peer->tcp.user_queue);

+	/* restore CBs that were saved in ovpn_sock_set_tcp_cb() */
+	sock->sk->sk_data_ready = peer->tcp.sk_cb.sk_data_ready;
+	sock->sk->sk_write_space = peer->tcp.sk_cb.sk_write_space;
+	sock->sk->sk_prot = peer->tcp.sk_cb.prot;
+	sock->sk->sk_socket->ops = peer->tcp.sk_cb.ops;
+
+	/* drop reference to peer */

nit: not really :)

drop comment :)

quoted

+	rcu_assign_sk_user_data(sock->sk, NULL);
+
+	/* before canceling any ongoing work we must ensure that CBs
+	 * have been reset to prevent workers from being re-armed
+	 */
+	barrier();
+
+	cancel_work_sync(&peer->tcp.tx_work);
+	strp_done(&peer->tcp.strp);
+	skb_queue_purge(&peer->tcp.out_queue);

Also kfree_skb(peer->tcp.out_msg.skb)?

hm yeah, it could be we allocated one but did not finish filling it.

quoted

+	ovpn_peer_put(peer);
+}

[...]

quoted

+static int ovpn_tcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
+{

[...]

quoted

+	ret = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size);
+	if (ret) {
+		kfree_skb(skb);
+		net_err_ratelimited("%s: skb copy from iter failed: %d\n",
+				    netdev_name(sock->peer->ovpn->dev), ret);
+		goto peer_free;
+	}
+
+	ovpn_tcp_send_sock_skb(sock->peer, skb);

This isn't propagating MSG_DONTWAIT down to ovpn_tcp_send_sock?

patch 14/25 will add a new member to ovpn_cb which will be filled right 
before calling ovpn_tcp_send_sock_skb() and that gets picked later.

Cheers,


-- 
Antonio Quartulli
OpenVPN Inc.