Merge branch 'tcp-RACK-fast-recovery'

Yuchung Cheng says:

====================
tcp: RACK fast recovery

The patch set enables RACK loss detection (draft-ietf-tcpm-rack-01)
to trigger fast recovery with a reordering timer.

Previously RACK has been running in auxiliary mode where it is
used to detect packet losses once the recovery has triggered by
other algorithms (e.g., FACK). By inspecting packet timestamps,
RACK can start ACK-driven repairs timely. A few similar heuristics
are no longer needed and are either removed or disabled to reduce
the complexity of the Linux TCP loss recovery engine:

  1. FACK (Forward Acknowledgement)
  2. Early Retransmit (RFC5827)
  3. thin_dupack (fast recovery on single DUPACK for thin-streams)
  4. NCR (Non-Congestion Robustness RFC4653) (RFC4653)
  5. Forward Retransmit

After this change, Linux's loss recovery algorithms consist of
  1. Conventional DUPACK threshold approach (RFC6675)
  2. RACK and Tail Loss Probe (draft-ietf-tcpm-rack-01)
  3. RTO plus F-RTO extension (RFC5682)

The patch set has been tested on Google servers extensively and
presented in several IETF meetings. The data suggests that RACK
successfully improves recovery performance:
https://www.ietf.org/proceedings/97/slides/slides-97-tcpm-draft-ietf-tcpm-rack-01.pdf
https://www.ietf.org/proceedings/96/slides/slides-96-tcpm-3.pdf
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/networking/ip-sysctl.txt

@@ -246,21 +246,12 @@ tcp_dsack - BOOLEAN
 	Allows TCP to send "duplicate" SACKs.
 
 tcp_early_retrans - INTEGER
-	Enable Early Retransmit (ER), per RFC 5827. ER lowers the threshold
-	for triggering fast retransmit when the amount of outstanding data is
-	small and when no previously unsent data can be transmitted (such
-	that limited transmit could be used). Also controls the use of
-	Tail loss probe (TLP) that converts RTOs occurring due to tail
-	losses into fast recovery (draft-dukkipati-tcpm-tcp-loss-probe-01).
+	Tail loss probe (TLP) converts RTOs occurring due to tail
+	losses into fast recovery (draft-ietf-tcpm-rack). Note that
+	TLP requires RACK to function properly (see tcp_recovery below)
 	Possible values:
-		0 disables ER
-		1 enables ER
-		2 enables ER but delays fast recovery and fast retransmit
-		  by a fourth of RTT. This mitigates connection falsely
-		  recovers when network has a small degree of reordering
-		  (less than 3 packets).
-		3 enables delayed ER and TLP.
-		4 enables TLP only.
+		0 disables TLP
+		3 or 4 enables TLP
 	Default: 3
 
 tcp_ecn - INTEGER
@@ -712,18 +703,6 @@ tcp_thin_linear_timeouts - BOOLEAN
 	Documentation/networking/tcp-thin.txt
 	Default: 0
 
-tcp_thin_dupack - BOOLEAN
-	Enable dynamic triggering of retransmissions after one dupACK
-	for thin streams. If set, a check is performed upon reception
-	of a dupACK to determine if the stream is thin (less than 4
-	packets in flight). As long as the stream is found to be thin,
-	data is retransmitted on the first received dupACK. This
-	improves retransmission latency for non-aggressive thin
-	streams, often found to be time-dependent.
-	For more information on thin streams, see
-	Documentation/networking/tcp-thin.txt
-	Default: 0
-
 tcp_limit_output_bytes - INTEGER
 	Controls TCP Small Queue limit per tcp socket.
 	TCP bulk sender tends to increase packets in flight until it

include/linux/tcp.h

@@ -207,6 +207,8 @@ struct tcp_sock {
 	/* Information of the most recently (s)acked skb */
 	struct tcp_rack {
 		struct skb_mstamp mstamp; /* (Re)sent time of the skb */
+		u32 rtt_us;  /* Associated RTT */
+		u32 end_seq; /* Ending TCP sequence of the skb */
 		u8 advanced; /* mstamp advanced since last lost marking */
 		u8 reord;    /* reordering detected */
 	} rack;
@@ -218,12 +220,11 @@ struct tcp_sock {
 		unused:5;
 	u8	nonagle     : 4,/* Disable Nagle algorithm?             */
 		thin_lto    : 1,/* Use linear timeouts for thin streams */
-		thin_dupack : 1,/* Fast retransmit on first dupack      */
+		unused1	    : 1,
 		repair      : 1,
 		frto        : 1;/* F-RTO (RFC5682) activated in CA_Loss */
 	u8	repair_queue;
-	u8	do_early_retrans:1,/* Enable RFC5827 early-retransmit  */
-		syn_data:1,	/* SYN includes data */
+	u8	syn_data:1,	/* SYN includes data */
 		syn_fastopen:1,	/* SYN includes Fast Open option */
 		syn_fastopen_exp:1,/* SYN includes Fast Open exp. option */
 		syn_data_acked:1,/* data in SYN is acked by SYN-ACK */
@@ -305,7 +306,6 @@ struct tcp_sock {
 					 */
 
 	int     lost_cnt_hint;
-	u32     retransmit_high;	/* L-bits may be on up to this seqno */
 
 	u32	prior_ssthresh; /* ssthresh saved at recovery start	*/
 	u32	high_seq;	/* snd_nxt at onset of congestion	*/

include/net/inet_connection_sock.h

@@ -144,6 +144,7 @@ struct inet_connection_sock {
 #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 */
+#define ICSK_TIME_REO_TIMEOUT	6	/* Reordering timer */
 
 static inline struct inet_connection_sock *inet_csk(const struct sock *sk)
 {
@@ -234,7 +235,8 @@ static inline void inet_csk_reset_xmit_timer(struct sock *sk, const int what,
 	}
 
 	if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0 ||
-	    what == ICSK_TIME_EARLY_RETRANS || what ==  ICSK_TIME_LOSS_PROBE) {
+	    what == ICSK_TIME_EARLY_RETRANS || what == ICSK_TIME_LOSS_PROBE ||
+	    what == ICSK_TIME_REO_TIMEOUT) {
 		icsk->icsk_pending = what;
 		icsk->icsk_timeout = jiffies + when;
 		sk_reset_timer(sk, &icsk->icsk_retransmit_timer, icsk->icsk_timeout);

include/net/tcp.h

@@ -143,6 +143,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo);
 #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
 					                 * for local resources.
 					                 */
+#define TCP_REO_TIMEOUT_MIN	(2000) /* Min RACK reordering timeout in usec */
 
 #define TCP_KEEPALIVE_TIME	(120*60*HZ)	/* two hours */
 #define TCP_KEEPALIVE_PROBES	9		/* Max of 9 keepalive probes	*/
@@ -261,6 +262,9 @@ extern int sysctl_tcp_slow_start_after_idle;
 extern int sysctl_tcp_thin_linear_timeouts;
 extern int sysctl_tcp_thin_dupack;
 extern int sysctl_tcp_early_retrans;
+extern int sysctl_tcp_recovery;
+#define TCP_RACK_LOSS_DETECTION  0x1 /* Use RACK to detect losses */
+
 extern int sysctl_tcp_limit_output_bytes;
 extern int sysctl_tcp_challenge_ack_limit;
 extern int sysctl_tcp_min_tso_segs;
@@ -397,6 +401,7 @@ struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb,
 int tcp_child_process(struct sock *parent, struct sock *child,
 		      struct sk_buff *skb);
 void tcp_enter_loss(struct sock *sk);
+void tcp_cwnd_reduction(struct sock *sk, int newly_acked_sacked, int flag);
 void tcp_clear_retrans(struct tcp_sock *tp);
 void tcp_update_metrics(struct sock *sk);
 void tcp_init_metrics(struct sock *sk);
@@ -541,6 +546,7 @@ int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb, int segs);
 void tcp_retransmit_timer(struct sock *sk);
 void tcp_xmit_retransmit_queue(struct sock *);
 void tcp_simple_retransmit(struct sock *);
+void tcp_enter_recovery(struct sock *sk, bool ece_ack);
 int tcp_trim_head(struct sock *, struct sk_buff *, u32);
 int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int, gfp_t);
 
@@ -559,7 +565,6 @@ void tcp_skb_collapse_tstamp(struct sk_buff *skb,
 			     const struct sk_buff *next_skb);
 
 /* tcp_input.c */
-void tcp_resume_early_retransmit(struct sock *sk);
 void tcp_rearm_rto(struct sock *sk);
 void tcp_synack_rtt_meas(struct sock *sk, struct request_sock *req);
 void tcp_reset(struct sock *sk);
@@ -1031,23 +1036,6 @@ static inline void tcp_enable_fack(struct tcp_sock *tp)
 	tp->rx_opt.sack_ok |= TCP_FACK_ENABLED;
 }
 
-/* TCP early-retransmit (ER) is similar to but more conservative than
- * the thin-dupack feature.  Enable ER only if thin-dupack is disabled.
- */
-static inline void tcp_enable_early_retrans(struct tcp_sock *tp)
-{
-	struct net *net = sock_net((struct sock *)tp);
-
-	tp->do_early_retrans = sysctl_tcp_early_retrans &&
-		sysctl_tcp_early_retrans < 4 && !sysctl_tcp_thin_dupack &&
-		net->ipv4.sysctl_tcp_reordering == 3;
-}
-
-static inline void tcp_disable_early_retrans(struct tcp_sock *tp)
-{
-	tp->do_early_retrans = 0;
-}
-
 static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
 {
 	return tp->sacked_out + tp->lost_out;
@@ -1856,17 +1844,11 @@ void tcp_v4_init(void);
 void tcp_init(void);
 
 /* tcp_recovery.c */
-
-/* Flags to enable various loss recovery features. See below */
-extern int sysctl_tcp_recovery;
-
-/* Use TCP RACK to detect (some) tail and retransmit losses */
-#define TCP_RACK_LOST_RETRANS  0x1
-
-extern int tcp_rack_mark_lost(struct sock *sk);
-
-extern void tcp_rack_advance(struct tcp_sock *tp,
-			     const struct skb_mstamp *xmit_time, u8 sacked);
+extern void tcp_rack_mark_lost(struct sock *sk, const struct skb_mstamp *now);
+extern void tcp_rack_advance(struct tcp_sock *tp, u8 sacked, u32 end_seq,
+			     const struct skb_mstamp *xmit_time,
+			     const struct skb_mstamp *ack_time);
+extern void tcp_rack_reo_timeout(struct sock *sk);
 
 /*
  * Save and compile IPv4 options, return a pointer to it

net/ipv4/inet_diag.c

@@ -215,7 +215,7 @@ int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
 	}
 
 	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
-	    icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
+	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
 	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
 		r->idiag_timer = 1;
 		r->idiag_retrans = icsk->icsk_retransmits;

net/ipv4/sysctl_net_ipv4.c

@@ -536,13 +536,6 @@ static struct ctl_table ipv4_table[] = {
 		.mode           = 0644,
 		.proc_handler   = proc_dointvec
 	},
-	{
-		.procname       = "tcp_thin_dupack",
-		.data           = &sysctl_tcp_thin_dupack,
-		.maxlen         = sizeof(int),
-		.mode           = 0644,
-		.proc_handler   = proc_dointvec
-	},
 	{
 		.procname	= "tcp_early_retrans",
 		.data		= &sysctl_tcp_early_retrans,

net/ipv4/tcp.c

@@ -406,7 +406,6 @@ void tcp_init_sock(struct sock *sk)
 	tp->mss_cache = TCP_MSS_DEFAULT;
 
 	tp->reordering = sock_net(sk)->ipv4.sysctl_tcp_reordering;
-	tcp_enable_early_retrans(tp);
 	tcp_assign_congestion_control(sk);
 
 	tp->tsoffset = 0;
@@ -2475,11 +2474,6 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 	case TCP_THIN_DUPACK:
 		if (val < 0 || val > 1)
 			err = -EINVAL;
-		else {
-			tp->thin_dupack = val;
-			if (tp->thin_dupack)
-				tcp_disable_early_retrans(tp);
-		}
 		break;
 
 	case TCP_REPAIR:
@@ -2969,8 +2963,9 @@ static int do_tcp_getsockopt(struct sock *sk, int level,
 	case TCP_THIN_LINEAR_TIMEOUTS:
 		val = tp->thin_lto;
 		break;
+
 	case TCP_THIN_DUPACK:
-		val = tp->thin_dupack;
+		val = 0;
 		break;
 
 	case TCP_REPAIR:

net/ipv4/tcp_ipv4.c

@@ -2229,7 +2229,7 @@ static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i)
 	int state;
 
 	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
-	    icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
+	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
 	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
 		timer_active	= 1;
 		timer_expires	= icsk->icsk_timeout;

net/ipv4/tcp_metrics.c

@@ -522,7 +522,6 @@ void tcp_init_metrics(struct sock *sk)
 	val = tcp_metric_get(tm, TCP_METRIC_REORDERING);
 	if (val && tp->reordering != val) {
 		tcp_disable_fack(tp);
-		tcp_disable_early_retrans(tp);
 		tp->reordering = val;
 	}
 

net/ipv4/tcp_minisocks.c

@@ -468,7 +468,6 @@ struct sock *tcp_create_openreq_child(const struct sock *sk,
 		newtp->sacked_out = 0;
 		newtp->fackets_out = 0;
 		newtp->snd_ssthresh = TCP_INFINITE_SSTHRESH;
-		tcp_enable_early_retrans(newtp);
 		newtp->tlp_high_seq = 0;
 		newtp->lsndtime = treq->snt_synack.stamp_jiffies;
 		newsk->sk_txhash = treq->txhash;

net/ipv4/tcp_output.c

@@ -76,10 +76,8 @@ static void tcp_event_new_data_sent(struct sock *sk, const struct sk_buff *skb)
 	tp->snd_nxt = TCP_SKB_CB(skb)->end_seq;
 
 	tp->packets_out += tcp_skb_pcount(skb);
-	if (!prior_packets || icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
-	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
+	if (!prior_packets || icsk->icsk_pending == ICSK_TIME_LOSS_PROBE)
 		tcp_rearm_rto(sk);
-	}
 
 	NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPORIGDATASENT,
 		      tcp_skb_pcount(skb));
@@ -2289,8 +2287,6 @@ bool tcp_schedule_loss_probe(struct sock *sk)
 	u32 timeout, tlp_time_stamp, rto_time_stamp;
 	u32 rtt = usecs_to_jiffies(tp->srtt_us >> 3);
 
-	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);
@@ -2309,8 +2305,9 @@ bool tcp_schedule_loss_probe(struct sock *sk)
 	/* Schedule a loss probe in 2*RTT for SACK capable connections
 	 * in Open state, that are either limited by cwnd or application.
 	 */
-	if (sysctl_tcp_early_retrans < 3 || !tp->packets_out ||
-	    !tcp_is_sack(tp) || inet_csk(sk)->icsk_ca_state != TCP_CA_Open)
+	if ((sysctl_tcp_early_retrans != 3 && sysctl_tcp_early_retrans != 4) ||
+	    !tp->packets_out || !tcp_is_sack(tp) ||
+	    icsk->icsk_ca_state != TCP_CA_Open)
 		return false;
 
 	if ((tp->snd_cwnd > tcp_packets_in_flight(tp)) &&
@@ -2831,36 +2828,6 @@ int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb, int segs)
 	return err;
 }
 
-/* Check if we forward retransmits are possible in the current
- * window/congestion state.
- */
-static bool tcp_can_forward_retransmit(struct sock *sk)
-{
-	const struct inet_connection_sock *icsk = inet_csk(sk);
-	const struct tcp_sock *tp = tcp_sk(sk);
-
-	/* Forward retransmissions are possible only during Recovery. */
-	if (icsk->icsk_ca_state != TCP_CA_Recovery)
-		return false;
-
-	/* No forward retransmissions in Reno are possible. */
-	if (tcp_is_reno(tp))
-		return false;
-
-	/* Yeah, we have to make difficult choice between forward transmission
-	 * and retransmission... Both ways have their merits...
-	 *
-	 * For now we do not retransmit anything, while we have some new
-	 * segments to send. In the other cases, follow rule 3 for
-	 * NextSeg() specified in RFC3517.
-	 */
-
-	if (tcp_may_send_now(sk))
-		return false;
-
-	return true;
-}
-
 /* This gets called after a retransmit timeout, and the initially
  * retransmitted data is acknowledged.  It tries to continue
  * resending the rest of the retransmit queue, until either
@@ -2875,24 +2842,16 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct sk_buff *skb;
 	struct sk_buff *hole = NULL;
-	u32 max_segs, last_lost;
+	u32 max_segs;
 	int mib_idx;
-	int fwd_rexmitting = 0;
 
 	if (!tp->packets_out)
 		return;
 
-	if (!tp->lost_out)
-		tp->retransmit_high = tp->snd_una;
-
 	if (tp->retransmit_skb_hint) {
 		skb = tp->retransmit_skb_hint;
-		last_lost = TCP_SKB_CB(skb)->end_seq;
-		if (after(last_lost, tp->retransmit_high))
-			last_lost = tp->retransmit_high;
 	} else {
 		skb = tcp_write_queue_head(sk);
-		last_lost = tp->snd_una;
 	}
 
 	max_segs = tcp_tso_segs(sk, tcp_current_mss(sk));
@@ -2915,31 +2874,14 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 		 */
 		segs = min_t(int, segs, max_segs);
 
-		if (fwd_rexmitting) {
-begin_fwd:
-			if (!before(TCP_SKB_CB(skb)->seq, tcp_highest_sack_seq(tp)))
-				break;
-			mib_idx = LINUX_MIB_TCPFORWARDRETRANS;
-
-		} else if (!before(TCP_SKB_CB(skb)->seq, tp->retransmit_high)) {
-			tp->retransmit_high = last_lost;
-			if (!tcp_can_forward_retransmit(sk))
+		if (tp->retrans_out >= tp->lost_out) {
 			break;
-			/* Backtrack if necessary to non-L'ed skb */
-			if (hole) {
-				skb = hole;
-				hole = NULL;
-			}
-			fwd_rexmitting = 1;
-			goto begin_fwd;
-
 		} else if (!(sacked & TCPCB_LOST)) {
 			if (!hole && !(sacked & (TCPCB_SACKED_RETRANS|TCPCB_SACKED_ACKED)))
 				hole = skb;
 			continue;
 
 		} else {
-			last_lost = TCP_SKB_CB(skb)->end_seq;
 			if (icsk->icsk_ca_state != TCP_CA_Loss)
 				mib_idx = LINUX_MIB_TCPFASTRETRANS;
 			else
@@ -2960,7 +2902,8 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 		if (tcp_in_cwnd_reduction(sk))
 			tp->prr_out += tcp_skb_pcount(skb);
 
-		if (skb == tcp_write_queue_head(sk))
+		if (skb == tcp_write_queue_head(sk) &&
+		    icsk->icsk_pending != ICSK_TIME_REO_TIMEOUT)
 			inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
 						  inet_csk(sk)->icsk_rto,
 						  TCP_RTO_MAX);

net/ipv4/tcp_recovery.c

 #include <linux/tcp.h>
 #include <net/tcp.h>
 
-int sysctl_tcp_recovery __read_mostly = TCP_RACK_LOST_RETRANS;
+int sysctl_tcp_recovery __read_mostly = TCP_RACK_LOSS_DETECTION;
 
-/* Marks a packet lost, if some packet sent later has been (s)acked.
+static void tcp_rack_mark_skb_lost(struct sock *sk, struct sk_buff *skb)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	tcp_skb_mark_lost_uncond_verify(tp, skb);
+	if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS) {
+		/* Account for retransmits that are lost again */
+		TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS;
+		tp->retrans_out -= tcp_skb_pcount(skb);
+		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPLOSTRETRANSMIT);
+	}
+}
+
+static bool tcp_rack_sent_after(const struct skb_mstamp *t1,
+				const struct skb_mstamp *t2,
+				u32 seq1, u32 seq2)
+{
+	return skb_mstamp_after(t1, t2) ||
+	       (t1->v64 == t2->v64 && after(seq1, seq2));
+}
+
+/* RACK loss detection (IETF draft draft-ietf-tcpm-rack-01):
+ *
+ * Marks a packet lost, if some packet sent later has been (s)acked.
  * The underlying idea is similar to the traditional dupthresh and FACK
  * but they look at different metrics:
  *
@@ -16,31 +39,26 @@ int sysctl_tcp_recovery __read_mostly = TCP_RACK_LOST_RETRANS;
  * is being more resilient to reordering by simply allowing some
  * "settling delay", instead of tweaking the dupthresh.
  *
- * The current version is only used after recovery starts but can be
- * easily extended to detect the first loss.
+ * When tcp_rack_detect_loss() detects some packets are lost and we
+ * are not already in the CA_Recovery state, either tcp_rack_reo_timeout()
+ * or tcp_time_to_recover()'s "Trick#1: the loss is proven" code path will
+ * make us enter the CA_Recovery state.
  */
-int tcp_rack_mark_lost(struct sock *sk)
+static void tcp_rack_detect_loss(struct sock *sk, const struct skb_mstamp *now,
+				 u32 *reo_timeout)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct sk_buff *skb;
-	u32 reo_wnd, prior_retrans = tp->retrans_out;
-
-	if (inet_csk(sk)->icsk_ca_state < TCP_CA_Recovery || !tp->rack.advanced)
-		return 0;
-
-	/* Reset the advanced flag to avoid unnecessary queue scanning */
-	tp->rack.advanced = 0;
+	u32 reo_wnd;
 
+	*reo_timeout = 0;
 	/* To be more reordering resilient, allow min_rtt/4 settling delay
 	 * (lower-bounded to 1000uS). We use min_rtt instead of the smoothed
 	 * RTT because reordering is often a path property and less related
 	 * to queuing or delayed ACKs.
-	 *
-	 * TODO: measure and adapt to the observed reordering delay, and
-	 * use a timer to retransmit like the delayed early retransmit.
 	 */
 	reo_wnd = 1000;
-	if (tp->rack.reord && tcp_min_rtt(tp) != ~0U)
+	if ((tp->rack.reord || !tp->lost_out) && tcp_min_rtt(tp) != ~0U)
 		reo_wnd = max(tcp_min_rtt(tp) >> 2, reo_wnd);
 
 	tcp_for_write_queue(skb, sk) {
@@ -54,20 +72,29 @@ int tcp_rack_mark_lost(struct sock *sk)
 		    scb->sacked & TCPCB_SACKED_ACKED)
 			continue;
 
-		if (skb_mstamp_after(&tp->rack.mstamp, &skb->skb_mstamp)) {
+		if (tcp_rack_sent_after(&tp->rack.mstamp, &skb->skb_mstamp,
+					tp->rack.end_seq, scb->end_seq)) {
+			/* Step 3 in draft-cheng-tcpm-rack-00.txt:
+			 * A packet is lost if its elapsed time is beyond
+			 * the recent RTT plus the reordering window.
+			 */
+			u32 elapsed = skb_mstamp_us_delta(now,
+							  &skb->skb_mstamp);
+			s32 remaining = tp->rack.rtt_us + reo_wnd - elapsed;
 
-			if (skb_mstamp_us_delta(&tp->rack.mstamp,
-						&skb->skb_mstamp) <= reo_wnd)
+			if (remaining < 0) {
+				tcp_rack_mark_skb_lost(sk, skb);
 				continue;
-
-			/* skb is lost if packet sent later is sacked */
-			tcp_skb_mark_lost_uncond_verify(tp, skb);
-			if (scb->sacked & TCPCB_SACKED_RETRANS) {
-				scb->sacked &= ~TCPCB_SACKED_RETRANS;
-				tp->retrans_out -= tcp_skb_pcount(skb);
-				NET_INC_STATS(sock_net(sk),
-					      LINUX_MIB_TCPLOSTRETRANSMIT);
 			}
+
+			/* Skip ones marked lost but not yet retransmitted */
+			if ((scb->sacked & TCPCB_LOST) &&
+			    !(scb->sacked & TCPCB_SACKED_RETRANS))
+				continue;
+
+			/* Record maximum wait time (+1 to avoid 0) */
+			*reo_timeout = max_t(u32, *reo_timeout, 1 + remaining);
+
 		} else if (!(scb->sacked & TCPCB_RETRANS)) {
 			/* Original data are sent sequentially so stop early
 			 * b/c the rest are all sent after rack_sent
@@ -75,20 +102,43 @@ int tcp_rack_mark_lost(struct sock *sk)
 			break;
 		}
 	}
-	return prior_retrans - tp->retrans_out;
 }
 
-/* Record the most recently (re)sent time among the (s)acked packets */
-void tcp_rack_advance(struct tcp_sock *tp,
-		      const struct skb_mstamp *xmit_time, u8 sacked)
+void tcp_rack_mark_lost(struct sock *sk, const struct skb_mstamp *now)
 {
+	struct tcp_sock *tp = tcp_sk(sk);
+	u32 timeout;
+
+	if (!tp->rack.advanced)
+		return;
+
+	/* Reset the advanced flag to avoid unnecessary queue scanning */
+	tp->rack.advanced = 0;
+	tcp_rack_detect_loss(sk, now, &timeout);
+	if (timeout) {
+		timeout = usecs_to_jiffies(timeout + TCP_REO_TIMEOUT_MIN);
+		inet_csk_reset_xmit_timer(sk, ICSK_TIME_REO_TIMEOUT,
+					  timeout, inet_csk(sk)->icsk_rto);
+	}
+}
+
+/* Record the most recently (re)sent time among the (s)acked packets
+ * This is "Step 3: Advance RACK.xmit_time and update RACK.RTT" from
+ * draft-cheng-tcpm-rack-00.txt
+ */
+void tcp_rack_advance(struct tcp_sock *tp, u8 sacked, u32 end_seq,
+		      const struct skb_mstamp *xmit_time,
+		      const struct skb_mstamp *ack_time)
+{
+	u32 rtt_us;
+
 	if (tp->rack.mstamp.v64 &&
-	    !skb_mstamp_after(xmit_time, &tp->rack.mstamp))
+	    !tcp_rack_sent_after(xmit_time, &tp->rack.mstamp,
+				 end_seq, tp->rack.end_seq))
 		return;
 
+	rtt_us = skb_mstamp_us_delta(ack_time, xmit_time);
 	if (sacked & TCPCB_RETRANS) {
-		struct skb_mstamp now;
-
 		/* If the sacked packet was retransmitted, it's ambiguous
 		 * whether the retransmission or the original (or the prior
 		 * retransmission) was sacked.
@@ -99,11 +149,35 @@ void tcp_rack_advance(struct tcp_sock *tp,
 		 * so it's at least one RTT (i.e., retransmission is at least
 		 * an RTT later).
 		 */
-		skb_mstamp_get(&now);
-		if (skb_mstamp_us_delta(&now, xmit_time) < tcp_min_rtt(tp))
+		if (rtt_us < tcp_min_rtt(tp))
 			return;
 	}
-
+	tp->rack.rtt_us = rtt_us;
 	tp->rack.mstamp = *xmit_time;
+	tp->rack.end_seq = end_seq;
 	tp->rack.advanced = 1;
 }
+
+/* We have waited long enough to accommodate reordering. Mark the expired
+ * packets lost and retransmit them.
+ */
+void tcp_rack_reo_timeout(struct sock *sk)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	struct skb_mstamp now;
+	u32 timeout, prior_inflight;
+
+	skb_mstamp_get(&now);
+	prior_inflight = tcp_packets_in_flight(tp);
+	tcp_rack_detect_loss(sk, &now, &timeout);
+	if (prior_inflight != tcp_packets_in_flight(tp)) {
+		if (inet_csk(sk)->icsk_ca_state != TCP_CA_Recovery) {
+			tcp_enter_recovery(sk, false);
+			if (!inet_csk(sk)->icsk_ca_ops->cong_control)
+				tcp_cwnd_reduction(sk, 1, 0);
+		}
+		tcp_xmit_retransmit_queue(sk);
+	}
+	if (inet_csk(sk)->icsk_pending != ICSK_TIME_RETRANS)
+		tcp_rearm_rto(sk);
+}

net/ipv4/tcp_timer.c

@@ -563,8 +563,8 @@ void tcp_write_timer_handler(struct sock *sk)
 	event = icsk->icsk_pending;
 
 	switch (event) {
-	case ICSK_TIME_EARLY_RETRANS:
-		tcp_resume_early_retransmit(sk);
+	case ICSK_TIME_REO_TIMEOUT:
+		tcp_rack_reo_timeout(sk);
 		break;
 	case ICSK_TIME_LOSS_PROBE:
 		tcp_send_loss_probe(sk);

net/ipv6/tcp_ipv6.c

@@ -1745,7 +1745,7 @@ static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
 	srcp  = ntohs(inet->inet_sport);
 
 	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
-	    icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
+	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
 	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
 		timer_active	= 1;
 		timer_expires	= icsk->icsk_timeout;