bpf: tcp: Recv() should return 0 when the peer socket is closed

If the peer is closed, we will never get more data, so
tcp_bpf_wait_data will get stuck forever. In case we passed
MSG_DONTWAIT to recv(), we get EAGAIN but we should actually get
0.

>From man 2 recv:

    RETURN VALUE

    When a stream socket peer has performed an orderly shutdown, the
    return value will be 0 (the traditional "end-of-file" return).

This patch makes tcp_bpf_wait_data always return 1 when the peer
socket has been shutdown. Either we have data available, and it would
have returned 1 anyway, or there isn't, in which case we'll call
tcp_recvmsg which does the right thing in this situation.

Fixes: eaf66302e456 ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/26038a28c21fea5d04d4bd4744c5686d3f2e5504.1591784177.git.sd@queasysnail.net

diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c
index 2b915aafda428e5d3ee0fe5b61babadec888f57d..7aa68f4aae6c3471b86ed08d97f50622003865d8 100644 (file)
--- a/net/ipv4/tcp_bpf.c
+++ b/net/ipv4/tcp_bpf.c
@@ -245,6 +245,9 @@ static int tcp_bpf_wait_data(struct sock *sk, struct sk_psock *psock,
        DEFINE_WAIT_FUNC(wait, woken_wake_function);
        int ret = 0;
 
+       if (sk->sk_shutdown & RCV_SHUTDOWN)
+               return 1;
+
        if (!timeo)
                return ret;