bpf: Fix use of sk->sk_reuseport from sk_assign

In testing, we found that for request sockets the sk->sk_reuseport field
may yet be uninitialized, which caused bpf_sk_assign() to randomly
succeed or return -ESOCKTNOSUPPORT when handling the forward ACK in a
three-way handshake.

Fix it by only applying the reuseport check for full sockets.

Fixes: cf7fbe660f2d ("bpf: Add socket assign support")
Signed-off-by: Joe Stringer <[e-mail]>
Signed-off-by: Daniel Borkmann <[e-mail]>
Acked-by: Martin KaFai Lau <[e-mail]>
Link: https://lore.kernel.org/bpf/[e-mail]

net/core/filter.c

···
 		return -EOPNOTSUPP;
 	if (unlikely(dev_net(skb->dev) != sock_net(sk)))
 		return -ENETUNREACH;
-	if (unlikely(sk->sk_reuseport))
+	if (unlikely(sk_fullsock(sk) && sk->sk_reuseport))
 		return -ESOCKTNOSUPPORT;
 	if (sk_is_refcounted(sk) &&
 	    unlikely(!refcount_inc_not_zero(&sk->sk_refcnt)))