git: e4ee2a39ade5 - stable/13 - tcp: Add PRR cwnd reduction for non-SACK loss

[Richard Scheffenegger]

The branch stable/13 has been updated by rscheff:

URL: https://cgit.FreeBSD.org/src/commit/?id=e4ee2a39ade5d6be38aed21f2fe1224c77f61af2

commit e4ee2a39ade5d6be38aed21f2fe1224c77f61af2
Author:     Richard Scheffenegger <rscheff at FreeBSD.org>
AuthorDate: 2021-06-19 17:06:48 +0000
Commit:     Richard Scheffenegger <rscheff at FreeBSD.org>
CommitDate: 2021-08-02 11:59:23 +0000

    tcp: Add PRR cwnd reduction for non-SACK loss
    
    This completes PRR cwnd reduction in all circumstances
    for the base TCP stack (SACK loss recovery, ECN window reduction,
    non-SACK loss recovery), preventing the arriving ACKs to
    clock out new data at the old, too high rate. This
    reduces the chance to induce additional losses while
    recovering from loss (during congested network conditions).
    
    For non-SACK loss recovery, each ACK is assumed to have
    one MSS delivered. In order to prevent ACK-split attacks,
    only one window worth of ACKs is considered to actually
    have delivered new data.
    
    MFC after: 6 weeks
    Reviewed By: rrs, #transport
    Sponsored by: NetApp, Inc.
    Differential Revision: https://reviews.freebsd.org/D29441
    
    (cherry picked from commit 74d7fc8753a76851db6db556ff0f09de1fbb26d5)
---
 sys/netinet/tcp_input.c | 56 ++++++++++++++++++++++++++++++++++++-------------
 1 file changed, 41 insertions(+), 15 deletions(-)

diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c
index 6d86e43f07dd..f7ca4ce05200 100644
--- a/sys/netinet/tcp_input.c
+++ b/sys/netinet/tcp_input.c
@@ -2605,8 +2605,7 @@ tcp_do_segment(struct mbuf *m, struct tcphdr *th, struct socket *so,
 					cc_ack_received(tp, th, nsegs,
 					    CC_DUPACK);
 					if (V_tcp_do_prr &&
-					    IN_FASTRECOVERY(tp->t_flags) &&
-					    (tp->t_flags & TF_SACK_PERMIT)) {
+					    IN_FASTRECOVERY(tp->t_flags)) {
 						tcp_do_prr_ack(tp, th, &to);
 					} else if ((tp->t_flags & TF_SACK_PERMIT) &&
 					    (to.to_flags & TOF_SACK) &&
@@ -2682,8 +2681,16 @@ enter_recovery:
 						 * snd_ssthresh is already updated by
 						 * cc_cong_signal.
 						 */
-						tp->sackhint.prr_delivered =
-						    tp->sackhint.sacked_bytes;
+						if ((tp->t_flags & TF_SACK_PERMIT) &&
+						    (to.to_flags & TOF_SACK)) {
+							tp->sackhint.prr_delivered =
+							    tp->sackhint.sacked_bytes;
+						} else {
+							tp->sackhint.prr_delivered =
+							    imin(tp->snd_max - tp->snd_una,
+							    imin(INT_MAX / 65536,
+								tp->t_dupacks) * maxseg);
+						}
 						tp->sackhint.recover_fs = max(1,
 						    tp->snd_nxt - tp->snd_una);
 					}
@@ -3961,11 +3968,23 @@ tcp_do_prr_ack(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to)
 	 * (del_data) and an estimate of how many bytes are in the
 	 * network.
 	 */
-	del_data = tp->sackhint.delivered_data;
-	if (V_tcp_do_rfc6675_pipe)
-		pipe = tcp_compute_pipe(tp);
-	else
-		pipe = (tp->snd_nxt - tp->snd_fack) + tp->sackhint.sack_bytes_rexmit;
+	if (((tp->t_flags & TF_SACK_PERMIT) &&
+	    (to->to_flags & TOF_SACK)) ||
+	    (IN_CONGRECOVERY(tp->t_flags) &&
+	     !IN_FASTRECOVERY(tp->t_flags))) {
+		del_data = tp->sackhint.delivered_data;
+		if (V_tcp_do_rfc6675_pipe)
+			pipe = tcp_compute_pipe(tp);
+		else
+			pipe = (tp->snd_nxt - tp->snd_fack) +
+				tp->sackhint.sack_bytes_rexmit;
+	} else {
+		if (tp->sackhint.prr_delivered < (tcprexmtthresh * maxseg +
+					     tp->snd_recover - tp->snd_una))
+			del_data = maxseg;
+		pipe = imax(0, tp->snd_max - tp->snd_una -
+			    imin(INT_MAX / 65536, tp->t_dupacks) * maxseg);
+	}
 	tp->sackhint.prr_delivered += del_data;
 	/*
 	 * Proportional Rate Reduction
@@ -3978,9 +3997,9 @@ tcp_do_prr_ack(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to)
 			    tp->snd_ssthresh, tp->sackhint.recover_fs) -
 			    tp->sackhint.prr_out;
 	} else {
-		if (V_tcp_do_prr_conservative)
+		if (V_tcp_do_prr_conservative || (del_data == 0))
 			limit = tp->sackhint.prr_delivered -
-			    tp->sackhint.prr_out;
+				tp->sackhint.prr_out;
 		else
 			limit = imax(tp->sackhint.prr_delivered -
 				    tp->sackhint.prr_out, del_data) +
@@ -3994,11 +4013,18 @@ tcp_do_prr_ack(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to)
 	 * accordingly.
 	 */
 	if (IN_FASTRECOVERY(tp->t_flags)) {
-		tp->snd_cwnd = imax(maxseg, tp->snd_nxt - tp->snd_recover +
-			tp->sackhint.sack_bytes_rexmit + (snd_cnt * maxseg));
+		if ((tp->t_flags & TF_SACK_PERMIT) &&
+		    (to->to_flags & TOF_SACK)) {
+			tp->snd_cwnd = tp->snd_nxt - tp->snd_recover +
+					    tp->sackhint.sack_bytes_rexmit +
+					    (snd_cnt * maxseg);
+		} else {
+			tp->snd_cwnd = (tp->snd_max - tp->snd_una) +
+					    (snd_cnt * maxseg);
+		}
 	} else if (IN_CONGRECOVERY(tp->t_flags))
-		tp->snd_cwnd = imax(maxseg, pipe - del_data +
-				    (snd_cnt * maxseg));
+		tp->snd_cwnd = pipe - del_data + (snd_cnt * maxseg);
+	tp->snd_cwnd = imax(maxseg, tp->snd_cwnd);
 }
 
 /*