Re: [PATCH net 3/3] tcp: fix xmit timer to only be reset if data ACKed/SACKed

[Neal Cardwell <ncardwell@google.com>]

[-- Attachment #1: Type: text/plain, Size: 1366 bytes --]

On Fri, Aug 4, 2017 at 3:33 AM, maowenan <maowenan@huawei.com> wrote:
> [Mao Wenan]Follow previous mail, in lower version such as 3.10, I found
> there are many timer type, e.g:ICSK_TIME_EARLY_RETRANS, RTO,PTO
> are used. I'm not sure there exist some unknown problem if we don't check
> isck_pending here and below if branch? And could you please post one lower
> version patch such as 3.10?  Thanks a lot.
>
> #define ICSK_TIME_RETRANS       1       /* Retransmit timer */
> #define ICSK_TIME_DACK          2       /* Delayed ack timer */
> #define ICSK_TIME_PROBE0        3       /* Zero window probe timer */
> #define ICSK_TIME_EARLY_RETRANS 4       /* Early retransmit timer */
> #define ICSK_TIME_LOSS_PROBE    5       /* Tail loss probe timer */

I think you'll find that it you audit how these patches interact with
those other timer types you'll find that the behavior either matches
the existing code, or is improved. We ran these patches through our
packetdrill test suite and found that was the case for all our
existing tests. But please let us know if you find a specific scenario
where that is not the case.

I have attached patches for this fix rebased on to v3.10.107, the
latest stable release for 3.10. That's pretty far back in history, so
there were substantial conflict resolutions and adjustments required.
:-) Hope that helps.

thanks,
neal

[-- Attachment #2: 0001-tcp-introduce-tcp_rto_delta_us-helper-for-xmit-timer.patch --]
[-- Type: application/octet-stream, Size: 2261 bytes --]

From 6e5001de12d20fee1696fe50df51b05438278f4c Mon Sep 17 00:00:00 2001
From: Neal Cardwell <ncardwell@google.com>
Date: Thu, 27 Jul 2017 10:01:20 -0400
Subject: [PATCH 1/3] tcp: introduce tcp_rto_delta_us() helper for xmit timer
 fix

Pure refactor. This helper will be required in the xmit timer fix
later in the patch series. (Because the TLP logic will want to make
this calculation.)

[This version of the commit was compiled and briefly tested
based on top of v3.10.107.]

Change-Id: I1ccfba0b00465454bf5ce22e6fef5f7b5dd94d15
Fixes: 6ba8a3b19e76 ("tcp: Tail loss probe (TLP)")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Nandita Dukkipati <nanditad@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
---
 include/net/tcp.h    | 10 ++++++++++
 net/ipv4/tcp_input.c |  4 +---
 2 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/net/tcp.h b/include/net/tcp.h
index 79cd118d5994..c4db9acefa9c 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -1592,4 +1592,14 @@ struct tcp_request_sock_ops {
 extern void tcp_v4_init(void);
 extern void tcp_init(void);
 
+/* At how many jiffies into the future should the RTO fire? */
+static inline s32 tcp_rto_delta(const struct sock *sk)
+{
+	const struct sk_buff *skb = tcp_write_queue_head(sk);
+	const u32 rto = inet_csk(sk)->icsk_rto;
+	const u32 rto_time_stamp = TCP_SKB_CB(skb)->when + rto;
+
+	return (s32)(rto_time_stamp - tcp_time_stamp);
+}
+
 #endif	/* _TCP_H */
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 0680058fe693..a36b7c55b76b 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -2972,9 +2972,7 @@ void tcp_rearm_rto(struct sock *sk)
 		/* Offset the time elapsed after installing regular RTO */
 		if (icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
 		    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
-			struct sk_buff *skb = tcp_write_queue_head(sk);
-			const u32 rto_time_stamp = TCP_SKB_CB(skb)->when + rto;
-			s32 delta = (s32)(rto_time_stamp - tcp_time_stamp);
+			s32 delta = tcp_rto_delta(sk);
 			/* delta may not be positive if the socket is locked
 			 * when the retrans timer fires and is rescheduled.
 			 */
-- 
2.14.0.rc1.383.gd1ce394fe2-goog


[-- Attachment #3: 0002-tcp-enable-xmit-timer-fix-by-having-TLP-use-time-whe.patch --]
[-- Type: application/octet-stream, Size: 3106 bytes --]

From 50c4e5e3bd4a437a3fbc237e0c248f758aa2ab85 Mon Sep 17 00:00:00 2001
From: Neal Cardwell <ncardwell@google.com>
Date: Thu, 27 Jul 2017 10:09:30 -0400
Subject: [PATCH 2/3] tcp: enable xmit timer fix by having TLP use time when
 RTO should fire

Have tcp_schedule_loss_probe() base the TLP scheduling decision based
on when the RTO *should* fire. This is to enable the upcoming xmit
timer fix in this series, where tcp_schedule_loss_probe() cannot
assume that the last timer installed was an RTO timer (because we are
no longer doing the "rearm RTO, rearm RTO, rearm TLP" dance on every
ACK). So tcp_schedule_loss_probe() must independently figure out when
an RTO would want to fire.

In the new TLP implementation following in this series, we cannot
assume that icsk_timeout was set based on an RTO; after processing a
cumulative ACK the icsk_timeout we see can be from a previous TLP or
RTO. So we need to independently recalculate the RTO time (instead of
reading it out of icsk_timeout). Removing this dependency on the
nature of icsk_timeout makes things a little easier to reason about
anyway.

Note that the old and new code should be equivalent, since they are
both saying: "if the RTO is in the future, but at an earlier time than
the normal TLP time, then set the TLP timer to fire when the RTO would
have fired".

[This version of the commit was compiled and briefly tested
based on top of v3.10.107.]

Change-Id: I597ad6446edde15bf2cea8e56d603a2c52f8221b
Fixes: 6ba8a3b19e76 ("tcp: Tail loss probe (TLP)")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Nandita Dukkipati <nanditad@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
---
 net/ipv4/tcp_output.c | 14 +++++---------
 1 file changed, 5 insertions(+), 9 deletions(-)

diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 8729a934124f..135440283f21 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -1947,8 +1947,8 @@ bool tcp_schedule_loss_probe(struct sock *sk)
 {
 	struct inet_connection_sock *icsk = inet_csk(sk);
 	struct tcp_sock *tp = tcp_sk(sk);
-	u32 timeout, tlp_time_stamp, rto_time_stamp;
 	u32 rtt = tp->srtt >> 3;
+	u32 timeout, rto_delta;
 
 	if (WARN_ON(icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS))
 		return false;
@@ -1987,14 +1987,10 @@ bool tcp_schedule_loss_probe(struct sock *sk)
 				(rtt + (rtt >> 1) + TCP_DELACK_MAX));
 	timeout = max_t(u32, timeout, msecs_to_jiffies(10));
 
-	/* If RTO is shorter, just schedule TLP in its place. */
-	tlp_time_stamp = tcp_time_stamp + timeout;
-	rto_time_stamp = (u32)inet_csk(sk)->icsk_timeout;
-	if ((s32)(tlp_time_stamp - rto_time_stamp) > 0) {
-		s32 delta = rto_time_stamp - tcp_time_stamp;
-		if (delta > 0)
-			timeout = delta;
-	}
+	/* If the RTO formula yields an earlier time, then use that time. */
+	rto_delta = tcp_rto_delta(sk);  /* How far in future is RTO? */
+	if (rto_delta > 0)
+		timeout = min_t(u32, timeout, rto_delta);
 
 	inet_csk_reset_xmit_timer(sk, ICSK_TIME_LOSS_PROBE, timeout,
 				  TCP_RTO_MAX);
-- 
2.14.0.rc1.383.gd1ce394fe2-goog


[-- Attachment #4: 0003-tcp-fix-xmit-timer-to-only-be-reset-if-data-ACKed-SA.patch --]
[-- Type: application/octet-stream, Size: 7689 bytes --]

From c920720b6bd7fa0cc855b16799b9ea11ef95db76 Mon Sep 17 00:00:00 2001
From: Neal Cardwell <ncardwell@google.com>
Date: Wed, 26 Jul 2017 21:43:27 -0400
Subject: [PATCH 3/3] tcp: fix xmit timer to only be reset if data ACKed/SACKed

Fix a TCP loss recovery performance bug raised recently on the netdev
list, in two threads:

(i)  July 26, 2017: netdev thread "TCP fast retransmit issues"
(ii) July 26, 2017: netdev thread:
     "[PATCH V2 net-next] TLP: Don't reschedule PTO when there's one
     outstanding TLP retransmission"

The basic problem is that incoming TCP packets that did not indicate
forward progress could cause the xmit timer (TLP or RTO) to be rearmed
and pushed back in time. In certain corner cases this could result in
the following problems noted in these threads:

 - Repeated ACKs coming in with bogus SACKs corrupted by middleboxes
   could cause TCP to repeatedly schedule TLPs forever. We kept
   sending TLPs after every ~200ms, which elicited bogus SACKs, which
   caused more TLPs, ad infinitum; we never fired an RTO to fill in
   the holes.

 - Incoming data segments could, in some cases, cause us to reschedule
   our RTO or TLP timer further out in time, for no good reason. This
   could cause repeated inbound data to result in stalls in outbound
   data, in the presence of packet loss.

This commit fixes these bugs by changing the TLP and RTO ACK
processing to:

 (a) Only reschedule the xmit timer once per ACK.

 (b) Only reschedule the xmit timer if tcp_clean_rtx_queue() deems the
     ACK indicates sufficient forward progress (a packet was
     cumulatively ACKed, or we got a SACK for a packet that was sent
     before the most recent retransmit of the write queue head).

This brings us back into closer compliance with the RFCs, since, as
the comment for tcp_rearm_rto() notes, we should only restart the RTO
timer after forward progress on the connection. Previously we were
restarting the xmit timer even in these cases where there was no
forward progress.

As a side benefit, this commit simplifies and speeds up the TCP timer
arming logic. We had been calling inet_csk_reset_xmit_timer() three
times on normal ACKs that cumulatively acknowledged some data:

1) Once near the top of tcp_ack() to switch from TLP timer to RTO:
        if (icsk->icsk_pending == ICSK_TIME_LOSS_PROBE)
               tcp_rearm_rto(sk);

2) Once in tcp_clean_rtx_queue(), to update the RTO:
        if (flag & FLAG_ACKED) {
               tcp_rearm_rto(sk);

3) Once in tcp_ack() after tcp_fastretrans_alert() to switch from RTO
   to TLP:
        if (icsk->icsk_pending == ICSK_TIME_RETRANS)
               tcp_schedule_loss_probe(sk);

This commit, by only rescheduling the xmit timer once per ACK,
simplifies the code and reduces CPU overhead.

This commit was tested in an A/B test with Google web server
traffic. SNMP stats and request latency metrics were within noise
levels, substantiating that for normal web traffic patterns this is a
rare issue. This commit was also tested with packetdrill tests to
verify that it fixes the timer behavior in the corner cases discussed
in the netdev threads mentioned above.

This patch is a bug fix patch intended to be queued for -stable
relases.

[This version of the commit was compiled and briefly tested
based on top of v3.10.107.]

Change-Id: If0417380fd59290b65cf04a415373aa13dd1dad7
Fixes: 6ba8a3b19e76 ("tcp: Tail loss probe (TLP)")
Reported-by: Klavs Klavsen <kl@vsen.dk>
Reported-by: Mao Wenan <maowenan@huawei.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Nandita Dukkipati <nanditad@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
---
 net/ipv4/tcp_input.c  | 25 +++++++++++++++----------
 net/ipv4/tcp_output.c | 12 ------------
 2 files changed, 15 insertions(+), 22 deletions(-)

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index a36b7c55b76b..70f217cb3a1b 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -111,6 +111,7 @@ int sysctl_tcp_early_retrans __read_mostly = 3;
 #define FLAG_ORIG_SACK_ACKED	0x200 /* Never retransmitted data are (s)acked	*/
 #define FLAG_SND_UNA_ADVANCED	0x400 /* Snd_una was changed (!= FLAG_DATA_ACKED) */
 #define FLAG_DSACKING_ACK	0x800 /* SACK blocks contained D-SACK info */
+#define FLAG_SET_XMIT_TIMER	0x1000 /* Set TLP or RTO timer */
 #define FLAG_SACK_RENEGING	0x2000 /* snd_una advanced to a sacked seq */
 #define FLAG_UPDATE_TS_RECENT	0x4000 /* tcp_replace_ts_recent() */
 
@@ -3002,6 +3003,13 @@ void tcp_resume_early_retransmit(struct sock *sk)
 	tcp_xmit_retransmit_queue(sk);
 }
 
+/* Try to schedule a loss probe; if that doesn't work, then schedule an RTO. */
+static void tcp_set_xmit_timer(struct sock *sk)
+{
+	if (!tcp_schedule_loss_probe(sk))
+		tcp_rearm_rto(sk);
+}
+
 /* If we get here, the whole TSO packet has not been acked. */
 static u32 tcp_tso_acked(struct sock *sk, struct sk_buff *skb)
 {
@@ -3132,7 +3140,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets,
 		}
 
 		tcp_ack_update_rtt(sk, flag, seq_rtt);
-		tcp_rearm_rto(sk);
+		flag |= FLAG_SET_XMIT_TIMER;  /* set TLP or RTO timer */
 
 		if (tcp_is_reno(tp)) {
 			tcp_remove_reno_sacks(sk, pkts_acked);
@@ -3392,10 +3400,6 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 	if (after(ack, tp->snd_nxt))
 		goto invalid_ack;
 
-	if (icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
-	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE)
-		tcp_rearm_rto(sk);
-
 	if (after(ack, prior_snd_una))
 		flag |= FLAG_SND_UNA_ADVANCED;
 
@@ -3452,6 +3456,12 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 
 	pkts_acked = previous_packets_out - tp->packets_out;
 
+	if (tp->tlp_high_seq)
+		tcp_process_tlp_ack(sk, ack, flag);
+	/* If needed, reset TLP/RTO timer; RACK may later override this. */
+	if (flag & FLAG_SET_XMIT_TIMER)
+		tcp_set_xmit_timer(sk);
+
 	if (tcp_ack_is_dubious(sk, flag)) {
 		/* Advance CWND, if state allows this. */
 		if ((flag & FLAG_DATA_ACKED) && tcp_may_raise_cwnd(sk, flag))
@@ -3464,17 +3474,12 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 			tcp_cong_avoid(sk, ack, prior_in_flight);
 	}
 
-	if (tp->tlp_high_seq)
-		tcp_process_tlp_ack(sk, ack, flag);
-
 	if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP)) {
 		struct dst_entry *dst = __sk_dst_get(sk);
 		if (dst)
 			dst_confirm(dst);
 	}
 
-	if (icsk->icsk_pending == ICSK_TIME_RETRANS)
-		tcp_schedule_loss_probe(sk);
 	if (tp->srtt != prior_rtt || tp->snd_cwnd != prior_cwnd)
 		tcp_update_pacing_rate(sk);
 	return 1;
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 135440283f21..f5d670ccd403 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -1945,28 +1945,16 @@ repair:
 
 bool tcp_schedule_loss_probe(struct sock *sk)
 {
-	struct inet_connection_sock *icsk = inet_csk(sk);
 	struct tcp_sock *tp = tcp_sk(sk);
 	u32 rtt = tp->srtt >> 3;
 	u32 timeout, rto_delta;
 
-	if (WARN_ON(icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS))
-		return false;
-	/* No consecutive loss probes. */
-	if (WARN_ON(icsk->icsk_pending == ICSK_TIME_LOSS_PROBE)) {
-		tcp_rearm_rto(sk);
-		return false;
-	}
 	/* Don't do any loss probe on a Fast Open connection before 3WHS
 	 * finishes.
 	 */
 	if (sk->sk_state == TCP_SYN_RECV)
 		return false;
 
-	/* TLP is only scheduled when next timer event is RTO. */
-	if (icsk->icsk_pending != ICSK_TIME_RETRANS)
-		return false;
-
 	/* Schedule a loss probe in 2*RTT for SACK capable connections
 	 * in Open state, that are either limited by cwnd or application.
 	 */
-- 
2.14.0.rc1.383.gd1ce394fe2-goog