Re: [RFC] vsock: notify server to shutdown when client has pending signal

On Mon, May 17, 2021 at 02:18:51AM +0000, Longpeng (Mike, Cloud Infrastructure Service Product Dept.) wrote:

Hi Stefano,

quoted

-----Original Message-----
From: Longpeng (Mike, Cloud Infrastructure Service Product Dept.)
[mailto:longpeng2@huawei.com]
Sent: Thursday, May 13, 2021 6:36 PM
To: Stefano Garzarella <sgarzare@redhat.com>
Cc: netdev@vger.kernel.org; linux-kernel@vger.kernel.org; Gonglei (Arei)
[off-list ref]; Subo (Subo, Cloud Infrastructure Service Product
Dept.) [off-list ref]; David S . Miller [off-list ref]; Jakub
Kicinski [off-list ref]; Jorgen Hansen [off-list ref]; Norbert
Slusarek [off-list ref]; Andra Paraschiv [off-list ref];
Colin Ian King [off-list ref]; David Brazdil
[off-list ref]; Alexander Popov [off-list ref];
lixianming (E) [off-list ref]
Subject: RE: [RFC] vsock: notify server to shutdown when client has pending
signal

Hi Stefano,

quoted

-----Original Message-----
From: Stefano Garzarella [mailto:sgarzare@redhat.com]
Sent: Thursday, May 13, 2021 5:42 PM
To: Longpeng (Mike, Cloud Infrastructure Service Product Dept.)
[off-list ref]
Cc: netdev@vger.kernel.org; linux-kernel@vger.kernel.org; Gonglei
(Arei) [off-list ref]; Subo (Subo, Cloud Infrastructure
Service Product
Dept.) [off-list ref]; David S . Miller [off-list ref];
Jakub Kicinski [off-list ref]; Jorgen Hansen [off-list ref];
Norbert Slusarek [off-list ref]; Andra Paraschiv
[off-list ref]; Colin Ian King [off-list ref];
David Brazdil [off-list ref]; Alexander Popov
[off-list ref]; lixianming (E) [off-list ref]
Subject: Re: [RFC] vsock: notify server to shutdown when client has
pending signal

Hi,
thanks for this patch, comments below...

On Tue, May 11, 2021 at 05:41:27PM +0800, Longpeng(Mike) wrote:

quoted

The client's sk_state will be set to TCP_ESTABLISHED if the server
replay the client's connect request.
However, if the client has pending signal, its sk_state will be set
to TCP_CLOSE without notify the server, so the server will hold the
corrupt connection.

           client                        server

1. sk_state=TCP_SYN_SENT         |
2. call ->connect()              |
3. wait reply                    |
                                | 4. sk_state=TCP_ESTABLISHED
                                | 5. insert to connected list
                                | 6. reply to the client
7. sk_state=TCP_ESTABLISHED      |
8. insert to connected list      |
9. *signal pending* <--------------------- the user kill client
10. sk_state=TCP_CLOSE           |
client is exiting...             |
11. call ->release()             |
    virtio_transport_close
     if (!(sk->sk_state == TCP_ESTABLISHED ||
      sk->sk_state == TCP_CLOSING))
	return true; <------------- return at here As a result, the server
cannot notice the connection is corrupt.
So the client should notify the peer in this case.

Cc: David S. Miller <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Stefano Garzarella <sgarzare@redhat.com>
Cc: Jorgen Hansen <redacted>
Cc: Norbert Slusarek <redacted>
Cc: Andra Paraschiv <redacted>
Cc: Colin Ian King <redacted>
Cc: David Brazdil <redacted>
Cc: Alexander Popov <redacted>
Signed-off-by: lixianming <redacted>
Signed-off-by: Longpeng(Mike) <redacted>
---
net/vmw_vsock/af_vsock.c | 1 +
1 file changed, 1 insertion(+)

diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c

index
92a72f0..d5df908 100644

--- a/net/vmw_vsock/af_vsock.c
+++ b/net/vmw_vsock/af_vsock.c

@@ -1368,6 +1368,7 @@ static int vsock_stream_connect(struct socket

*sock,

struct sockaddr *addr,

quoted

		lock_sock(sk);

		if (signal_pending(current)) {
+			vsock_send_shutdown(sk, SHUTDOWN_MASK);

I see the issue, but I'm not sure is okay to send the shutdown in any
case, think about the server didn't setup the connection.

Maybe is better to set TCP_CLOSING if the socket state was
TCP_ESTABLISHED, so the shutdown will be handled by the
transport->release() as usual.

What do you think?

Your method looks more gracefully, we'll try it and get back to you, 
thanks.

As your suggestion, the following code can solve the problem:

               if (signal_pending(current)) {
                       err = sock_intr_errno(timeout);
-                       sk->sk_state = TCP_CLOSE;
+                       sk->sk_state = TCP_CLOSING;
                       sock->state = SS_UNCONNECTED;
                       vsock_transport_cancel_pkt(vsk);
                       goto out_wait;

This will send shutdown to the server even if the connection is not established, but
I don't see any side effects yet, right ?

Should we set TCP_CLOSING only if sk_state was TCP_ESTABLISHED?

The problem is also in the timeout case, we should fix it together ?

I'm not sure, if we reach the timeout, it should mean that the other 
peer never answered, so why take care to notify it?

Thanks,
Stefano