On Fri, Mar 03, 2017 at 09:25:54AM +0800, Peng Tao wrote:

On Fri, Mar 3, 2017 at 5:13 AM, David Miller [off-list ref] wrote:

quoted

From: Peng Tao <redacted>
Date: Wed,  1 Mar 2017 11:56:24 +0800

quoted

So that we can cancel a queued pkt later if necessary.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Peng Tao <redacted>
---
 include/linux/virtio_vsock.h            | 2 ++
 net/vmw_vsock/virtio_transport_common.c | 7 +++++++
 2 files changed, 9 insertions(+)

diff --git a/include/linux/virtio_vsock.h b/include/linux/virtio_vsock.h
index 9638bfe..193ad3a 100644
--- a/include/linux/virtio_vsock.h
+++ b/include/linux/virtio_vsock.h

@@ -48,6 +48,7 @@ struct virtio_vsock_pkt {
      struct virtio_vsock_hdr hdr;
      struct work_struct work;
      struct list_head list;
+     void *cancel_token; /* only used for cancellation */

The type here is fixed, you only store vhost_sock object pointers
here, so don't use "void *" please.

It used to be "struct vhost_sock *" but no refcount is held. Stefan
suggested to use "void *cancel_token" to make the code harder to
misuse.

Quoting Stefan:
"This field is just an opaque token used for cancellation rather than
a struct vsock_sock pointer that we are allowed to dereference.  You
could change this field to void *cancel_token to make the code harder
to misuse."

Ref:
https://www.mail-archive.com/netdev@vger.kernel.org/msg142550.html

Yes, the key point is that it shouldn't be used as a struct vsock_sock
since we don't hold a reference.  It's purely a token so the queued
packet can be found later for cancellation.

Stefan