tcp: tracking packets with CE marks in BW rate sample

In order to track CE marks per rate sample (one round trip), TCP needs a
per-skb header field to record the tp->delivered_ce count when the skb
was sent. To make space, we replace the "last_in_flight" field which is
used exclusively for NV congestion control. The stat needed by NV can be
alternatively approximated by existing stats tcp_sock delivered and
mss_cache.

This patch counts the number of packets delivered which have CE marks in
the rate sample, using similar approach of delivery accounting.

Cc: Lawrence Brakmo <[email protected]>
Signed-off-by: Yuchung Cheng <[email protected]>
Acked-by: Neal Cardwell <[email protected]>
Signed-off-by: Eric Dumazet <[email protected]>
Signed-off-by: Luke Hsiao <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Signed-off-by: Juhyung Park <[email protected]>

include/net/tcp.h

@@ -871,10 +871,11 @@ struct tcp_skb_cb {
 	__u32		ack_seq;	/* Sequence number ACK'd	*/
 	union {
 		struct {
+#define TCPCB_DELIVERED_CE_MASK ((1U<<20) - 1)
 			/* There is space for up to 24 bytes */
-			__u32 in_flight:30,/* Bytes in flight at transmit */
-			      is_app_limited:1, /* cwnd not fully used? */
-			      unused:1;
+			__u32 is_app_limited:1, /* cwnd not fully used? */
+			      delivered_ce:20,
+			      unused:11;
 			/* pkts S/ACKed so far upon tx of skb, incl retrans: */
 			__u32 delivered;
 			/* start of send pipeline phase */
@@ -1051,7 +1052,9 @@ struct ack_sample {
 struct rate_sample {
 	u64  prior_mstamp; /* starting timestamp for interval */
 	u32  prior_delivered;	/* tp->delivered at "prior_mstamp" */
+	u32  prior_delivered_ce;/* tp->delivered_ce at "prior_mstamp" */
 	s32  delivered;		/* number of packets delivered over interval */
+	s32  delivered_ce;	/* number of packets delivered w/ CE marks*/
 	long interval_us;	/* time for tp->delivered to incr "delivered" */
 	u32 snd_interval_us;	/* snd interval for delivered packets */
 	u32 rcv_interval_us;	/* rcv interval for delivered packets */

net/ipv4/tcp_input.c

@@ -3269,7 +3269,6 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 	long seq_rtt_us = -1L;
 	long ca_rtt_us = -1L;
 	u32 pkts_acked = 0;
-	u32 last_in_flight = 0;
 	bool rtt_update;
 	int flag = 0;
 
@@ -3305,7 +3304,6 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 			if (!first_ackt)
 				first_ackt = last_ackt;
 
-			last_in_flight = TCP_SKB_CB(skb)->tx.in_flight;
 			if (before(start_seq, reord))
 				reord = start_seq;
 			if (!after(scb->end_seq, tp->high_seq))
@@ -3371,8 +3369,8 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 		seq_rtt_us = tcp_stamp_us_delta(tp->tcp_mstamp, first_ackt);
 		ca_rtt_us = tcp_stamp_us_delta(tp->tcp_mstamp, last_ackt);
 
-		if (pkts_acked == 1 && last_in_flight < tp->mss_cache &&
-		    last_in_flight && !prior_sacked && fully_acked &&
+		if (pkts_acked == 1 && fully_acked && !prior_sacked &&
+		    (tp->snd_una - prior_snd_una) < tp->mss_cache &&
 		    sack->rate->prior_delivered + 1 == tp->delivered &&
 		    !(flag & (FLAG_CA_ALERT | FLAG_SYN_ACKED))) {
 			/* Conservatively mark a delayed ACK. It's typically
@@ -3429,9 +3427,10 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 
 	if (icsk->icsk_ca_ops->pkts_acked) {
 		struct ack_sample sample = { .pkts_acked = pkts_acked,
-					     .rtt_us = sack->rate->rtt_us,
-					     .in_flight = last_in_flight };
+					     .rtt_us = sack->rate->rtt_us };
 
+		sample.in_flight = tp->mss_cache *
+			(tp->delivered - sack->rate->prior_delivered);
 		icsk->icsk_ca_ops->pkts_acked(sk, &sample);
 	}
 

net/ipv4/tcp_output.c

@@ -1260,8 +1260,6 @@ static int __tcp_transmit_skb(struct sock *sk, struct sk_buff *skb,
 	tp->tcp_wstamp_ns = max(tp->tcp_wstamp_ns, tp->tcp_clock_cache);
 	skb->skb_mstamp_ns = tp->tcp_wstamp_ns;
 	if (clone_it) {
-		TCP_SKB_CB(skb)->tx.in_flight = TCP_SKB_CB(skb)->end_seq
-			- tp->snd_una;
 		oskb = skb;
 
 		tcp_skb_tsorted_save(oskb) {

net/ipv4/tcp_rate.c

@@ -65,6 +65,7 @@ void tcp_rate_skb_sent(struct sock *sk, struct sk_buff *skb)
 	TCP_SKB_CB(skb)->tx.first_tx_mstamp	= tp->first_tx_mstamp;
 	TCP_SKB_CB(skb)->tx.delivered_mstamp	= tp->delivered_mstamp;
 	TCP_SKB_CB(skb)->tx.delivered		= tp->delivered;
+	TCP_SKB_CB(skb)->tx.delivered_ce	= tp->delivered_ce;
 	TCP_SKB_CB(skb)->tx.is_app_limited	= tp->app_limited ? 1 : 0;
 }
 
@@ -86,6 +87,7 @@ void tcp_rate_skb_delivered(struct sock *sk, struct sk_buff *skb,
 
 	if (!rs->prior_delivered ||
 	    after(scb->tx.delivered, rs->prior_delivered)) {
+		rs->prior_delivered_ce  = scb->tx.delivered_ce;
 		rs->prior_delivered  = scb->tx.delivered;
 		rs->prior_mstamp     = scb->tx.delivered_mstamp;
 		rs->is_app_limited   = scb->tx.is_app_limited;
@@ -138,6 +140,10 @@ void tcp_rate_gen(struct sock *sk, u32 delivered, u32 lost,
 	}
 	rs->delivered   = tp->delivered - rs->prior_delivered;
 
+	rs->delivered_ce = tp->delivered_ce - rs->prior_delivered_ce;
+	/* delivered_ce occupies less than 32 bits in the skb control block */
+	rs->delivered_ce &= TCPCB_DELIVERED_CE_MASK;
+
 	/* Model sending data and receiving ACKs as separate pipeline phases
 	 * for a window. Usually the ACK phase is longer, but with ACK
 	 * compression the send phase can be longer. To be safe we use the