bpf: fix out-of-bounds read in __bpf_skc_lookup

__bpf_skc_lookup takes a socket tuple and the length of the
tuple as an argument. Based on the length, it decides which
address family to pass to the helper function sk_lookup.

In case of AF_INET6, it fails to verify that the length
of the tuple is long enough. sk_lookup may therefore access
data past the end of the tuple.

Fixes: 6acc9b4 ("bpf: Add helper to retrieve socket in BPF")
Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

net/core/filter.c

@@ -5304,7 +5304,13 @@ __bpf_skc_lookup(struct sk_buff *skb, struct bpf_sock_tuple *tuple, u32 len,
 	struct net *net;
 	int sdif;
 
-	family = len == sizeof(tuple->ipv4) ? AF_INET : AF_INET6;
+	if (len == sizeof(tuple->ipv4))
+		family = AF_INET;
+	else if (len == sizeof(tuple->ipv6))
+		family = AF_INET6;
+	else
+		return NULL;
+
 	if (unlikely(family == AF_UNSPEC || flags ||
 		     !((s32)netns_id < 0 || netns_id <= S32_MAX)))
 		goto out;