Skip to contentSkip to conversation

tcp: tcp_v4_err() cleanups

PATCH In review Eric Dumazet 3 participants 6 messages May 27, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover tcp: tcp_v4_err() cleanups

This series is a followup of patch 239174945dac ("tcp: tcp_v4_err() icmp skb is named icmp_skb").

Move the RFC 6069 code into a helper, and rename icmp_skb to standard skb name so that tcp_v4_err() and tcp_v6_err() are using consistent names.

Eric Dumazet (2):
tcp: add tcp_ld_RTO_revert() helper
tcp: rename tcp_v4_err() skb parameter

 net/ipv4/tcp_ipv4.c | 103 +++++++++++++++++++++++---------------------
 1 file changed, 54 insertions(+), 49 deletions(-)
Signature · 1 line
2.27.0.rc0.183.gde8f92d652-goog
#1 lore raw

PATCH 1/2 tcp: add tcp_ld_RTO_revert() helper

RFC 6069 logic has been implemented for IPv4 only so far, right in the middle of tcp_v4_err() and was error prone.

Move this code to one helper, to make tcp_v4_err() more readable and to eventually expand RFC 6069 to IPv6 in the future.

Also perform sock_owned_by_user() check a bit sooner.

---
 net/ipv4/tcp_ipv4.c | 85 ++++++++++++++++++++++++---------------------
 1 file changed, 45 insertions(+), 40 deletions(-)
net/ipv4/tcp_ipv4.c +45 -40
index 3ac0a7523923e0f1d959dfa65cf2b73bd6a4af15..8b257a92c98ffdb4618b8cde0937740ad5fe2e64 100644
@@ -403,6 +403,45 @@ void tcp_req_err(struct sock *sk, u32 seq, bool abort)
 }
 EXPORT_SYMBOL(tcp_req_err);
 
+/* TCP-LD (RFC 6069) logic */
+static void tcp_ld_RTO_revert(struct sock *sk, u32 seq)
+{
+	struct inet_connection_sock *icsk = inet_csk(sk);
+	struct tcp_sock *tp = tcp_sk(sk);
+	struct sk_buff *skb;
+	s32 remaining;
+	u32 delta_us;
+
+	if (sock_owned_by_user(sk))
+		return;
+
+	if (seq != tp->snd_una  || !icsk->icsk_retransmits ||
+	    !icsk->icsk_backoff)
+		return;
+
+	skb = tcp_rtx_queue_head(sk);
+	if (WARN_ON_ONCE(!skb))
+		return;
+
+	icsk->icsk_backoff--;
+	icsk->icsk_rto = tp->srtt_us ? __tcp_set_rto(tp) : TCP_TIMEOUT_INIT;
+	icsk->icsk_rto = inet_csk_rto_backoff(icsk, TCP_RTO_MAX);
+
+	tcp_mstamp_refresh(tp);
+	delta_us = (u32)(tp->tcp_mstamp - tcp_skb_timestamp_us(skb));
+	remaining = icsk->icsk_rto - usecs_to_jiffies(delta_us);
+
+	if (remaining > 0) {
+		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
+					  remaining, TCP_RTO_MAX);
+	} else {
+		/* RTO revert clocked out retransmission.
+		 * Will retransmit now.
+		 */
+		tcp_retransmit_timer(sk);
+	}
+}
+
 /*
  * This routine is called by the ICMP module when it gets some
  * sort of error condition.  If err < 0 then the socket should
@@ -423,17 +462,13 @@ int tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
 {
 	const struct iphdr *iph = (const struct iphdr *)icmp_skb->data;
 	struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2));
-	struct inet_connection_sock *icsk;
 	struct tcp_sock *tp;
 	struct inet_sock *inet;
 	const int type = icmp_hdr(icmp_skb)->type;
 	const int code = icmp_hdr(icmp_skb)->code;
 	struct sock *sk;
-	struct sk_buff *skb;
 	struct request_sock *fastopen;
 	u32 seq, snd_una;
-	s32 remaining;
-	u32 delta_us;
 	int err;
 	struct net *net = dev_net(icmp_skb->dev);
 
@@ -476,7 +511,6 @@ int tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
 		goto out;
 	}
 
-	icsk = inet_csk(sk);
 	tp = tcp_sk(sk);
 	/* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
 	fastopen = rcu_dereference(tp->fastopen_rsk);
@@ -521,41 +555,12 @@ int tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
 		}
 
 		err = icmp_err_convert[code].errno;
-		/* check if icmp_skb allows revert of backoff
-		 * (see draft-zimmermann-tcp-lcd) */
-		if (code != ICMP_NET_UNREACH && code != ICMP_HOST_UNREACH)
-			break;
-		if (seq != tp->snd_una  || !icsk->icsk_retransmits ||
-		    !icsk->icsk_backoff || fastopen)
-			break;
-
-		if (sock_owned_by_user(sk))
-			break;
-
-		skb = tcp_rtx_queue_head(sk);
-		if (WARN_ON_ONCE(!skb))
-			break;
-
-		icsk->icsk_backoff--;
-		icsk->icsk_rto = tp->srtt_us ? __tcp_set_rto(tp) :
-					       TCP_TIMEOUT_INIT;
-		icsk->icsk_rto = inet_csk_rto_backoff(icsk, TCP_RTO_MAX);
-
-
-		tcp_mstamp_refresh(tp);
-		delta_us = (u32)(tp->tcp_mstamp - tcp_skb_timestamp_us(skb));
-		remaining = icsk->icsk_rto -
-			    usecs_to_jiffies(delta_us);
-
-		if (remaining > 0) {
-			inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
-						  remaining, TCP_RTO_MAX);
-		} else {
-			/* RTO revert clocked out retransmission.
-			 * Will retransmit now */
-			tcp_retransmit_timer(sk);
-		}
-
+		/* check if this ICMP message allows revert of backoff.
+		 * (see RFC 6069)
+		 */
+		if (!fastopen &&
+		    (code == ICMP_NET_UNREACH || code == ICMP_HOST_UNREACH))
+			tcp_ld_RTO_revert(sk, seq);
 		break;
 	case ICMP_TIME_EXCEEDED:
 		err = EHOSTUNREACH;
Signature · 1 line
2.27.0.rc0.183.gde8f92d652-goog
maintainer Acked-by Tested-by on v1 1/2
#2 lore raw

Re: [PATCH net-next 1/2] tcp: add tcp_ld_RTO_revert() helper

Quoted 4 lines · Eric Dumazet

RFC 6069 logic has been implemented for IPv4 only so far, right in the middle of tcp_v4_err() and was error prone.

Quoted 8 lines

Move this code to one helper, to make tcp_v4_err() more readable and to eventually expand RFC 6069 to IPv6 in the future.

Also perform sock_owned_by_user() check a bit sooner.

---

Nice clean-up. Thanks, Eric! It will be great to have IPv6 RFC 6069 support as well.

thanks,
neal

ps: Tested with the packetdrill script below earlier this morning to verify that IPv4 RFC 6069 works before and after Eric's patch. Eric independently wrote a nicer test this morning.

// Establish a connection.

    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3

+0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0 +0 bind(3, ..., ...) = 0
+0 listen(3, 1) = 0

+0 < S 0:0(0) win 32792 <mss 1460,nop,wscale 7> +0 > S. 0:0(0) ack 1 <mss 1460,nop,wscale 8>
+.010 < . 1:1(0) ack 1 win 257
+0 accept(3, ..., ...) = 4

// Send 10 data segments.
+0 write(4, ..., 14600) = 14600
+0 > P. 1:14601(14600) ack 1

// ICMP says that the network is unreachable +.010 < icmp unreachable net_unreachable [1:1461(1460)]

// RTO retransmit.
+.216 > . 1:1461(1460) ack 1
+.010 < icmp unreachable net_unreachable [1:1461(1460)]

// RTO retransmit, without exponential backoff. +.216 > . 1:1461(1460) ack 1
+.010 < icmp unreachable net_unreachable [1:1461(1460)]

// RTO retransmit, without exponential backoff. +.216 > . 1:1461(1460) ack 1
+.010 < icmp unreachable net_unreachable [1:1461(1460)]

#3 lore raw

PATCH 2/2 tcp: rename tcp_v4_err() skb parameter

In reply to Eric Dumazet

This essentially reverts 4d1a2d9ec1c1 ("Revert Backoff [v3]:
Rename skb to icmp_skb in tcp_v4_err()")

Now we have tcp_ld_RTO_revert() helper, we can use the usual name for sk_buff parameter, so that tcp_v4_err() and tcp_v6_err() use similar names.

---
 net/ipv4/tcp_ipv4.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)
net/ipv4/tcp_ipv4.c +9 -9
index 8b257a92c98ffdb4618b8cde0937740ad5fe2e64..3a1e2becb1e8d1e0513e87bdfc0e1d5769ffc8e8 100644
@@ -458,23 +458,23 @@ static void tcp_ld_RTO_revert(struct sock *sk, u32 seq)
  *
  */
 
-int tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
+int tcp_v4_err(struct sk_buff *skb, u32 info)
 {
-	const struct iphdr *iph = (const struct iphdr *)icmp_skb->data;
-	struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2));
+	const struct iphdr *iph = (const struct iphdr *)skb->data;
+	struct tcphdr *th = (struct tcphdr *)(skb->data + (iph->ihl << 2));
 	struct tcp_sock *tp;
 	struct inet_sock *inet;
-	const int type = icmp_hdr(icmp_skb)->type;
-	const int code = icmp_hdr(icmp_skb)->code;
+	const int type = icmp_hdr(skb)->type;
+	const int code = icmp_hdr(skb)->code;
 	struct sock *sk;
 	struct request_sock *fastopen;
 	u32 seq, snd_una;
 	int err;
-	struct net *net = dev_net(icmp_skb->dev);
+	struct net *net = dev_net(skb->dev);
 
 	sk = __inet_lookup_established(net, &tcp_hashinfo, iph->daddr,
 				       th->dest, iph->saddr, ntohs(th->source),
-				       inet_iif(icmp_skb), 0);
+				       inet_iif(skb), 0);
 	if (!sk) {
 		__ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
 		return -ENOENT;
@@ -524,7 +524,7 @@ int tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
 	switch (type) {
 	case ICMP_REDIRECT:
 		if (!sock_owned_by_user(sk))
-			do_redirect(icmp_skb, sk);
+			do_redirect(skb, sk);
 		goto out;
 	case ICMP_SOURCE_QUENCH:
 		/* Just silently ignore these. */
@@ -578,7 +578,7 @@ int tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
 		if (fastopen && !fastopen->sk)
 			break;
 
-		ip_icmp_error(sk, icmp_skb, err, th->dest, info, (u8 *)th);
+		ip_icmp_error(sk, skb, err, th->dest, info, (u8 *)th);
 
 		if (!sock_owned_by_user(sk)) {
 			sk->sk_err = err;
Signature · 1 line
2.27.0.rc0.183.gde8f92d652-goog
maintainer Acked-by on v1 2/2
#4 lore raw

Re: [PATCH net-next 2/2] tcp: rename tcp_v4_err() skb parameter

Quoted 10 lines · Eric Dumazet

This essentially reverts 4d1a2d9ec1c1 ("Revert Backoff [v3]:
Rename skb to icmp_skb in tcp_v4_err()")

Now we have tcp_ld_RTO_revert() helper, we can use the usual name for sk_buff parameter, so that tcp_v4_err() and tcp_v6_err() use similar names.

---

Thanks, Eric!

neal

maintainer
#5 lore raw
In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>
Date: Tue, 26 May 2020 19:48:48 -0700

Quoted 5 lines

This series is a followup of patch 239174945dac ("tcp: tcp_v4_err() icmp skb is named icmp_skb").

Move the RFC 6069 code into a helper, and rename icmp_skb to standard skb name so that tcp_v4_err() and tcp_v6_err() are using consistent names.

Series applied, thanks Eric.