On 10/19/2023 3:12 AM, Vadim Fedorenko wrote:

On 18/10/2023 19:32, Alexandru Matei wrote:

quoted

Once VQs are filled with empty buffers and we kick the host, it can send
connection requests. If 'the_virtio_vsock' is not initialized before,
replies are silently dropped and do not reach the host.

Fixes: 0deab087b16a ("vsock/virtio: use RCU to avoid use-after-free on the_virtio_vsock")
Signed-off-by: Alexandru Matei <redacted>
---
  net/vmw_vsock/virtio_transport.c | 7 ++++---
  1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
index e95df847176b..eae0867133f8 100644
--- a/net/vmw_vsock/virtio_transport.c
+++ b/net/vmw_vsock/virtio_transport.c

@@ -658,12 +658,13 @@ static int virtio_vsock_probe(struct virtio_device *vdev)

          vsock->seqpacket_allow = true;
        vdev->priv = vsock;
+    rcu_assign_pointer(the_virtio_vsock, vsock);
        ret = virtio_vsock_vqs_init(vsock);
-    if (ret < 0)
+    if (ret < 0) {
+        rcu_assign_pointer(the_virtio_vsock, NULL);
          goto out;
-
-    rcu_assign_pointer(the_virtio_vsock, vsock);
+    }
        mutex_unlock(&the_virtio_vsock_mutex);
  

Looks like virtio_vsock_restore() needs the same changes. But
virtio_vsock_vqs_init() can fail only in virtio_find_vqs(). Maybe it can be split into 2 functions to avoid second rcu_assign_pointer() in case
of error?

Thanks, it definitely looks better and it solves a few other issues by splitting the function in two.