tcp: no longer abort SYN_SENT when receiving some ICMP

Currently, non fatal ICMP messages received on behalf
of SYN_SENT sockets do call tcp_ld_RTO_revert()
to implement RFC 6069, but immediately call tcp_done(),
thus aborting the connect() attempt.

This violates RFC 1122 following requirement:

4.2.3.9  ICMP Messages
...
          o    Destination Unreachable -- codes 0, 1, 5

                 Since these Unreachable messages indicate soft error
                 conditions, TCP MUST NOT abort the connection, and it
                 SHOULD make the information available to the
                 application.

This patch makes sure non 'fatal' ICMP[v6] messages do not
abort the connection attempt.

It enables RFC 6069 for SYN_SENT sockets as a result.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: David Morley <morleyd@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 5f693bbd578d2261b78aa0be6bf69499bbd5117e..86cc6d36f8188ec6a761c3a949382c03e74c91f8 100644 (file)
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -482,6 +482,7 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
        const int code = icmp_hdr(skb)->code;
        struct sock *sk;
        struct request_sock *fastopen;
+       bool harderr = false;
        u32 seq, snd_una;
        int err;
        struct net *net = dev_net(skb->dev);
@@ -555,6 +556,7 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
                goto out;
        case ICMP_PARAMETERPROB:
                err = EPROTO;
+               harderr = true;
                break;
        case ICMP_DEST_UNREACH:
                if (code > NR_ICMP_UNREACH)
@@ -579,6 +581,7 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
                }
 
                err = icmp_err_convert[code].errno;
+               harderr = icmp_err_convert[code].fatal;
                /* check if this ICMP message allows revert of backoff.
                 * (see RFC 6069)
                 */
@@ -604,6 +607,9 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
 
                ip_icmp_error(sk, skb, err, th->dest, info, (u8 *)th);
 
+               if (!harderr)
+                       break;
+
                if (!sock_owned_by_user(sk)) {
                        WRITE_ONCE(sk->sk_err, err);
 

diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 937a02c2e5345390ed592b19faa661cd703a23f0..43deda49cc521278f0bd869135e70bbe7aed6d35 100644 (file)
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -381,7 +381,7 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
        struct tcp_sock *tp;
        __u32 seq, snd_una;
        struct sock *sk;
-       bool fatal;
+       bool harderr;
        int err;
 
        sk = __inet6_lookup_established(net, net->ipv4.tcp_death_row.hashinfo,
@@ -402,9 +402,9 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                return 0;
        }
        seq = ntohl(th->seq);
-       fatal = icmpv6_err_convert(type, code, &err);
+       harderr = icmpv6_err_convert(type, code, &err);
        if (sk->sk_state == TCP_NEW_SYN_RECV) {
-               tcp_req_err(sk, seq, fatal);
+               tcp_req_err(sk, seq, harderr);
                return 0;
        }
 
@@ -489,6 +489,9 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 
                ipv6_icmp_error(sk, skb, err, th->dest, ntohl(info), (u8 *)th);
 
+               if (!harderr)
+                       break;
+
                if (!sock_owned_by_user(sk)) {
                        WRITE_ONCE(sk->sk_err, err);
                        sk_error_report(sk);            /* Wake people up to see the error (see connect in sock.c) */