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: cf7fbe66 ("bpf: Add socket assign support")
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20200408033540.10339-1-joe@wand.net.nz

net/core/filter.c

@@ -5925,7 +5925,7 @@ BPF_CALL_3(bpf_sk_assign, struct sk_buff *, skb, struct sock *, sk, u64, flags)
 		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)))