return len;
 }
 
+static int
+vhost_transport_cancel_pkt(struct vsock_sock *vsk)
+{
+       struct vhost_vsock *vsock;
+       struct virtio_vsock_pkt *pkt, *n;
+       int cnt = 0;
+       LIST_HEAD(freeme);
+
+       /* Find the vhost_vsock according to guest context id  */
+       vsock = vhost_vsock_get(vsk->remote_addr.svm_cid);
+       if (!vsock)
+               return -ENODEV;
+
+       spin_lock_bh(&vsock->send_pkt_list_lock);
+       list_for_each_entry_safe(pkt, n, &vsock->send_pkt_list, list) {
+               if (pkt->vsk != vsk)
+                       continue;
+               list_move(&pkt->list, &freeme);
+       }
+       spin_unlock_bh(&vsock->send_pkt_list_lock);
+
+       list_for_each_entry_safe(pkt, n, &freeme, list) {
+               if (pkt->reply)
+                       cnt++;
+               list_del(&pkt->list);
+               virtio_transport_free_pkt(pkt);
+       }
+
+       if (cnt) {
+               struct vhost_virtqueue *tx_vq = &vsock->vqs[VSOCK_VQ_TX];
+               int new_cnt;
+
+               new_cnt = atomic_sub_return(cnt, &vsock->queued_replies);
+               if (new_cnt + cnt >= tx_vq->num && new_cnt < tx_vq->num)
+                       vhost_poll_queue(&tx_vq->poll);
+       }
+
+       return 0;
+}
+
 static struct virtio_vsock_pkt *
 vhost_vsock_alloc_pkt(struct vhost_virtqueue *vq,
                      unsigned int out, unsigned int in)
                .release                  = virtio_transport_release,
                .connect                  = virtio_transport_connect,
                .shutdown                 = virtio_transport_shutdown,
+               .cancel_pkt               = vhost_transport_cancel_pkt,
 
                .dgram_enqueue            = virtio_transport_dgram_enqueue,
                .dgram_dequeue            = virtio_transport_dgram_dequeue,