Re: [PATCH net-next v2 1/2] vsock: report pending receive data to io_uring

[Stefano Garzarella <sgarzare@redhat.com>]

On Wed, Sep 23, 2026 at 07:40:07PM +0100, David Carlier wrote:
> AF_VSOCK stream receives never fill msghdr.msg_inq, so io_uring cannot set
> IORING_CQE_F_SOCK_NONEMPTY and retries a multishot receive even after the
> queue has been drained.
> 
> Fill the hint at the common receive exit using the transport callback that
> SIOCINQ already uses, and report 1 once the connection is finished so the
> caller performs the receive which observes EOF, as TCP does after a FIN.
> 
> A vsock loopback ping-pong with io_uring multishot receive drops entries
> into __vsock_connectible_recvmsg from 1.97 to 1.00 per delivered message,
> and receiver CPU time by about 3% (25 runs of 50000 messages, p=0.006).
> 
> Signed-off-by: David Carlier <devnexen@gmail.com>
> ---
>  net/vmw_vsock/af_vsock.c | 35 +++++++++++++++++++++++++++++++++++
>  1 file changed, 35 insertions(+)
> 
> diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
> index f840498b58af..20d6f9ca6a96 100644
> --- a/net/vmw_vsock/af_vsock.c
> +++ b/net/vmw_vsock/af_vsock.c
> @@ -2543,6 +2543,35 @@ static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
>  	return err;
>  }
>  
> +/* Bytes a following receive can consume, 1 if it would only see EOF, or -1
> + * if the transport cannot tell.
> + *
> + * Called under the socket lock after a nonnegative stream receive, so a NULL
> + * transport implies SOCK_DONE.
> + */
> +static int vsock_stream_inq_hint(struct sock *sk)
> +{
> +	struct vsock_sock *vsk = vsock_sk(sk);
> +	s64 data;
> +
> +	if ((sk->sk_shutdown & RCV_SHUTDOWN) || !vsk->transport ||
> +	    (sock_flag(sk, SOCK_DONE) && sk->sk_state != TCP_ESTABLISHED))
> +		return 1;
> +
> +	data = vsock_stream_has_data(vsk);
> +	if (data < 0)
> +		return -1;
> +	if (data > 0)
> +		return min_t(s64, data, INT_MAX);
> +
> +	/* Empty but finished: keep the caller reading so it sees EOF. */
> +	if (sock_flag(sk, SOCK_DONE) ||
> +	    (READ_ONCE(vsk->peer_shutdown) & SEND_SHUTDOWN))
> +		return 1;
> +
> +	return 0;
> +}
> +
>  int
>  __vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
>  			    int flags)
> @@ -2606,6 +2635,12 @@ __vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
>  		err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
>  
>  out:
> +	/* Seqpacket has_data counts messages, while io_uring treats msg_inq as

Do you mean vsock_stream_has_data() on SOCK_SEQPACKET doesn't return 
bytes?

IMO names are confusing, but it should return number of bytes, we should 
fix it, but of course not in this patch.

LGTM:

Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>


> +	 * a byte length when sizing retries, so only streams report a hint.
> +	 */
> +	if (msg->msg_get_inq && err >= 0 && sk->sk_type == SOCK_STREAM)
> +		msg->msg_inq = vsock_stream_inq_hint(sk);
> +
>  	release_sock(sk);
>  	return err;
>  }
> -- 
> 2.55.0
>