Re: [RFC PATCH v2 5/9] vsock: SO_RCVLOWAT transport set callback

From: Stefano Garzarella
Date: Wed Jul 27 2022 - 08:24:34 EST


On Mon, Jul 25, 2022 at 08:05:28AM +0000, Arseniy Krasnov wrote:
This adds transport specific callback for SO_RCVLOWAT, because in some
transports it may be difficult to know current available number of bytes
ready to read. Thus, when SO_RCVLOWAT is set, transport may reject it.

Signed-off-by: Arseniy Krasnov <AVKrasnov@xxxxxxxxxxxxxx>
---
include/net/af_vsock.h | 1 +
net/vmw_vsock/af_vsock.c | 19 +++++++++++++++++++
2 files changed, 20 insertions(+)

diff --git a/include/net/af_vsock.h b/include/net/af_vsock.h
index f742e50207fb..eae5874bae35 100644
--- a/include/net/af_vsock.h
+++ b/include/net/af_vsock.h
@@ -134,6 +134,7 @@ struct vsock_transport {
u64 (*stream_rcvhiwat)(struct vsock_sock *);
bool (*stream_is_active)(struct vsock_sock *);
bool (*stream_allow)(u32 cid, u32 port);
+ int (*set_rcvlowat)(struct vsock_sock *, int);

/* SEQ_PACKET. */
ssize_t (*seqpacket_dequeue)(struct vsock_sock *vsk, struct msghdr *msg,
diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
index 63a13fa2686a..b7a286db4af1 100644
--- a/net/vmw_vsock/af_vsock.c
+++ b/net/vmw_vsock/af_vsock.c
@@ -2130,6 +2130,24 @@ vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
return err;
}

+static int vsock_set_rcvlowat(struct sock *sk, int val)
+{
+ const struct vsock_transport *transport;
+ struct vsock_sock *vsk;
+ int err = 0;
+
+ vsk = vsock_sk(sk);
+ transport = vsk->transport;

`transport` can be NULL if the user call SO_RCVLOWAT before we assign it, so we should check it.

I think if the transport implements `set_rcvlowat`, maybe we should set there sk->sk_rcvlowat, so I would do something like that:

if (transport && transport->set_rcvlowat)
err = transport->set_rcvlowat(vsk, val);
else
WRITE_ONCE(sk->sk_rcvlowat, val ? : 1);

return err;

In addition I think we should check that val does not exceed vsk->buffer_size, something similar of what tcp_set_rcvlowat() does.

Thanks,
Stefano