Re: [PATCH net 1/2] vsock/virtio: discard packets if the transport changes

["Michael S. Tsirkin" <mst@redhat.com>]

On Thu, Jan 09, 2025 at 10:01:31AM +0100, Stefano Garzarella wrote:
> On Wed, Jan 08, 2025 at 02:31:19PM -0500, Hyunwoo Kim wrote:
> > On Wed, Jan 08, 2025 at 07:06:16PM +0100, Stefano Garzarella wrote:
> > > If the socket has been de-assigned or assigned to another transport,
> > > we must discard any packets received because they are not expected
> > > and would cause issues when we access vsk->transport.
> > > 
> > > A possible scenario is described by Hyunwoo Kim in the attached link,
> > > where after a first connect() interrupted by a signal, and a second
> > > connect() failed, we can find `vsk->transport` at NULL, leading to a
> > > NULL pointer dereference.
> > > 
> > > Fixes: c0cfa2d8a788 ("vsock: add multi-transports support")
> > > Reported-by: Hyunwoo Kim <v4bel@theori.io>
> > > Reported-by: Wongi Lee <qwerty@theori.io>
> > > Closes: https://lore.kernel.org/netdev/Z2LvdTTQR7dBmPb5@v4bel-B760M-AORUS-ELITE-AX/
> > > Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
> > > ---
> > >  net/vmw_vsock/virtio_transport_common.c | 7 +++++--
> > >  1 file changed, 5 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
> > > index 9acc13ab3f82..51a494b69be8 100644
> > > --- a/net/vmw_vsock/virtio_transport_common.c
> > > +++ b/net/vmw_vsock/virtio_transport_common.c
> > > @@ -1628,8 +1628,11 @@ void virtio_transport_recv_pkt(struct virtio_transport *t,
> > > 
> > >  	lock_sock(sk);
> > > 
> > > -	/* Check if sk has been closed before lock_sock */
> > > -	if (sock_flag(sk, SOCK_DONE)) {
> > > +	/* Check if sk has been closed or assigned to another transport before
> > > +	 * lock_sock (note: listener sockets are not assigned to any transport)
> > > +	 */
> > > +	if (sock_flag(sk, SOCK_DONE) ||
> > > +	    (sk->sk_state != TCP_LISTEN && vsk->transport != &t->transport)) {
> > 
> > If a race scenario with vsock_listen() is added to the existing
> > race scenario, the patch can be bypassed.
> > 
> > In addition to the existing scenario:
> > ```
> >                     cpu0                                                      cpu1
> > 
> >                                                               socket(A)
> > 
> >                                                               bind(A, {cid: VMADDR_CID_LOCAL, port: 1024})
> >                                                                 vsock_bind()
> > 
> >                                                               listen(A)
> >                                                                 vsock_listen()
> >  socket(B)
> > 
> >  connect(B, {cid: VMADDR_CID_LOCAL, port: 1024})
> >    vsock_connect()
> >      lock_sock(sk);
> >      virtio_transport_connect()
> >        virtio_transport_connect()
> >          virtio_transport_send_pkt_info()
> >            vsock_loopback_send_pkt(VIRTIO_VSOCK_OP_REQUEST)
> >              queue_work(vsock_loopback_work)
> >      sk->sk_state = TCP_SYN_SENT;
> >      release_sock(sk);
> >                                                               vsock_loopback_work()
> >                                                                 virtio_transport_recv_pkt(VIRTIO_VSOCK_OP_REQUEST)
> >                                                                   sk = vsock_find_bound_socket(&dst);
> >                                                                   virtio_transport_recv_listen(sk, skb)
> >                                                                     child = vsock_create_connected(sk);
> >                                                                     vsock_assign_transport()
> >                                                                       vvs = kzalloc(sizeof(*vvs), GFP_KERNEL);
> >                                                                     vsock_insert_connected(vchild);
> >                                                                       list_add(&vsk->connected_table, list);
> >                                                                     virtio_transport_send_response(vchild, skb);
> >                                                                       virtio_transport_send_pkt_info()
> >                                                                         vsock_loopback_send_pkt(VIRTIO_VSOCK_OP_RESPONSE)
> >                                                                           queue_work(vsock_loopback_work)
> > 
> >                                                               vsock_loopback_work()
> >                                                                 virtio_transport_recv_pkt(VIRTIO_VSOCK_OP_RESPONSE)
> >                                                                   sk = vsock_find_bound_socket(&dst);
> >                                                                   lock_sock(sk);
> >                                                                   case TCP_SYN_SENT:
> >                                                                   virtio_transport_recv_connecting()
> >                                                                     sk->sk_state = TCP_ESTABLISHED;
> >                                                                   release_sock(sk);
> > 
> >                                                               kill(connect(B));
> >      lock_sock(sk);
> >      if (signal_pending(current)) {
> >      sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
> >      sock->state = SS_UNCONNECTED;    // [1]
> >      release_sock(sk);
> > 
> >  connect(B, {cid: VMADDR_CID_HYPERVISOR, port: 1024})
> >    vsock_connect(B)
> >      lock_sock(sk);
> >      vsock_assign_transport()
> >        virtio_transport_release()
> >          virtio_transport_close()
> >            if (!(sk->sk_state == TCP_ESTABLISHED || sk->sk_state == TCP_CLOSING))
> >            virtio_transport_shutdown()
> >              virtio_transport_send_pkt_info()
> >                vsock_loopback_send_pkt(VIRTIO_VSOCK_OP_SHUTDOWN)
> >                  queue_work(vsock_loopback_work)
> >            schedule_delayed_work(&vsk->close_work, VSOCK_CLOSE_TIMEOUT);	// [5]
> >        vsock_deassign_transport()
> >          vsk->transport = NULL;
> >        return -ESOCKTNOSUPPORT;
> >      release_sock(sk);
> >                                                               vsock_loopback_work()
> >                                                                 virtio_transport_recv_pkt(VIRTIO_VSOCK_OP_SHUTDOWN)
> >                                                                   virtio_transport_recv_connected()
> >                                                                     virtio_transport_reset()
> >                                                                       virtio_transport_send_pkt_info()
> >                                                                         vsock_loopback_send_pkt(VIRTIO_VSOCK_OP_RST)
> >                                                                           queue_work(vsock_loopback_work)
> >  listen(B)
> >    vsock_listen()
> >      if (sock->state != SS_UNCONNECTED)    // [2]
> >      sk->sk_state = TCP_LISTEN;    // [3]
> > 
> >                                                               vsock_loopback_work()
> >                                                                 virtio_transport_recv_pkt(VIRTIO_VSOCK_OP_RST)
> > 								   if ((sk->sk_state != TCP_LISTEN && vsk->transport != &t->transport)) {    // [4]
> > 								   ...
> > 							
> >  virtio_transport_close_timeout()
> >    virtio_transport_do_close()
> >      vsock_stream_has_data()
> >        return vsk->transport->stream_has_data(vsk);    // null-ptr-deref
> > 
> > ```
> > (Yes, This is quite a crazy scenario, but it can actually be induced)
> > 
> > Since sock->state is set to SS_UNCONNECTED during the first connect()[1],
> > it can pass the sock->state check[2] in vsock_listen() and set
> > sk->sk_state to TCP_LISTEN[3].
> > If this happens, the check in the patch with
> > `sk->sk_state != TCP_LISTEN` will pass[4], and a null-ptr-deref can
> > still occur.)
> > 
> > More specifically, because the sk_state has changed to TCP_LISTEN,
> > virtio_transport_recv_disconnecting() will not be called by the
> > loopback worker. However, a null-ptr-deref may occur in
> > virtio_transport_close_timeout(), which is scheduled by
> > virtio_transport_close() called in the flow of the second connect()[5].
> > (The patch no longer cancels the virtio_transport_close_timeout() worker)
> > 
> > And even if the `sk->sk_state != TCP_LISTEN` check is removed from the
> > patch, it seems that a null-ptr-deref will still occur due to
> > virtio_transport_close_timeout().
> > It might be necessary to add worker cancellation at the
> > appropriate location.
> 
> Thanks for the analysis!
> 
> Do you have time to cook a proper patch to cover this scenario?
> Or we should mix this fix together with your patch (return 0 in
> vsock_stream_has_data()) while we investigate a better handling?
> 
> Thanks,
> Stefano

better combine them imho.