On 26.09.2023 15:56, Stefano Garzarella wrote:

On Fri, Sep 22, 2023 at 08:24:24AM +0300, Arseniy Krasnov wrote:

quoted

For AF_VSOCK, zerocopy tx mode depends on transport, so this option must
be set in AF_VSOCK implementation where transport is accessible (if
transport is not set during setting SO_ZEROCOPY: for example socket is
not connected, then SO_ZEROCOPY will be enabled, but once transport will
be assigned, support of this type of transmission will be checked).

To handle SO_ZEROCOPY, AF_VSOCK implementation uses SOCK_CUSTOM_SOCKOPT
bit, thus handling SOL_SOCKET option operations, but all of them except
SO_ZEROCOPY will be forwarded to the generic handler by calling
'sock_setsockopt()'.

Signed-off-by: Arseniy Krasnov <redacted>
---
Changelog:
v5(big patchset) -> v1:
 * Compact 'if' conditions.
 * Rename 'zc_val' to 'zerocopy'.
 * Use 'zerocopy' value directly in 'sock_valbool_flag()', without
   ?: operator.
 * Set 'SOCK_CUSTOM_SOCKOPT' bit for connectible sockets only, as
   suggested by Bobby Eshleman [off-list ref].

net/vmw_vsock/af_vsock.c | 46 ++++++++++++++++++++++++++++++++++++++--
1 file changed, 44 insertions(+), 2 deletions(-)

diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
index 482300eb88e0..c05a42e02a17 100644
--- a/net/vmw_vsock/af_vsock.c
+++ b/net/vmw_vsock/af_vsock.c

@@ -1406,8 +1406,16 @@ static int vsock_connect(struct socket *sock, struct sockaddr *addr,

            goto out;
        }

-        if (vsock_msgzerocopy_allow(transport))
+        if (vsock_msgzerocopy_allow(transport)) {
            set_bit(SOCK_SUPPORT_ZC, &sk->sk_socket->flags);
+        } else if (sock_flag(sk, SOCK_ZEROCOPY)) {
+            /* If this option was set before 'connect()',
+             * when transport was unknown, check that this
+             * feature is supported here.
+             */
+            err = -EOPNOTSUPP;
+            goto out;
+        }

        err = vsock_auto_bind(vsk);
        if (err)

@@ -1643,7 +1651,7 @@ static int vsock_connectible_setsockopt(struct socket *sock,

    const struct vsock_transport *transport;
    u64 val;

-    if (level != AF_VSOCK)
+    if (level != AF_VSOCK && level != SOL_SOCKET)
        return -ENOPROTOOPT;

#define COPY_IN(_v)                                       \

@@ -1666,6 +1674,34 @@ static int vsock_connectible_setsockopt(struct socket *sock,

    transport = vsk->transport;

+    if (level == SOL_SOCKET) {
+        int zerocopy;
+
+        if (optname != SO_ZEROCOPY) {
+            release_sock(sk);
+            return sock_setsockopt(sock, level, optname, optval, optlen);
+        }
+
+        /* Use 'int' type here, because variable to
+         * set this option usually has this type.
+         */
+        COPY_IN(zerocopy);
+
+        if (zerocopy < 0 || zerocopy > 1) {
+            err = -EINVAL;
+            goto exit;
+        }
+
+        if (transport && !vsock_msgzerocopy_allow(transport)) {
+            err = -EOPNOTSUPP;
+            goto exit;
+        }
+
+        sock_valbool_flag(sk, SOCK_ZEROCOPY,
+                  zerocopy);

it's not necessary to wrap this call.

Sorry, what do you mean ?

Thanks, Arseniy

quoted

+        goto exit;
+    }
+
    switch (optname) {
    case SO_VM_SOCKETS_BUFFER_SIZE:
        COPY_IN(val);

@@ -2322,6 +2358,12 @@ static int vsock_create(struct net *net, struct socket *sock,

        }
    }

+    /* SOCK_DGRAM doesn't have 'setsockopt' callback set in its
+     * proto_ops, so there is no handler for custom logic.
+     */
+    if (sock_type_connectible(sock->type))
+        set_bit(SOCK_CUSTOM_SOCKOPT, &sk->sk_socket->flags);
+
    vsock_insert_unbound(vsk);

    return 0;
-- 
2.25.1