Date: Sat, 16 Dec 2023 10:29:48 GMT From: Richard Scheffenegger <rscheff@FreeBSD.org> To: src-committers@FreeBSD.org, dev-commits-src-all@FreeBSD.org, dev-commits-src-branches@FreeBSD.org Subject: git: 1fdbb16f405f - stable/14 - tcp: add PRR 6937bis heuristic and retire prr_conservative sysctl Message-ID: <202312161029.3BGATmPS077967@gitrepo.freebsd.org>
next in thread | raw e-mail | index | archive | help
The branch stable/14 has been updated by rscheff: URL: https://cgit.FreeBSD.org/src/commit/?id=1fdbb16f405fb44beb204c06c3c573ff7e402f2a commit 1fdbb16f405fb44beb204c06c3c573ff7e402f2a Author: Richard Scheffenegger <rscheff@FreeBSD.org> AuthorDate: 2023-11-15 21:37:42 +0000 Commit: Richard Scheffenegger <rscheff@FreeBSD.org> CommitDate: 2023-12-15 08:25:07 +0000 tcp: add PRR 6937bis heuristic and retire prr_conservative sysctl Improve Proportional Rate Reduction (RFC6937) by using a heuristic, which automatically chooses between conservative CRB and more aggressive SSRB modes. Only when snd_una advances (a partial ACK), SSRB may be used. Also, that ACK must not have any indication of ongoing loss - using the addition of new holes into the scoreboard as proxy for such an event. MFC after: 4 weeks Reviewed By: #transport, kbowling, rrs Sponsored By: NetApp, Inc. Differential Revision: https://reviews.freebsd.org/D28822 (cherry picked from commit 49a6fbe38728173da74d5b497f700178e2a6c830) --- share/man/man4/tcp.4 | 8 +------- sys/netinet/tcp_input.c | 47 +++++++++++++++++++++++++++++++---------------- sys/netinet/tcp_sack.c | 32 ++++++++++++++++++++------------ sys/netinet/tcp_var.h | 13 ++++++++++--- 4 files changed, 62 insertions(+), 38 deletions(-) diff --git a/share/man/man4/tcp.4 b/share/man/man4/tcp.4 index 0ee57dcb0594..16c9e0ce84df 100644 --- a/share/man/man4/tcp.4 +++ b/share/man/man4/tcp.4 @@ -33,7 +33,7 @@ .\" .\" From: @(#)tcp.4 8.1 (Berkeley) 6/5/93 .\" -.Dd June 29, 2023 +.Dd November 17, 2023 .Dt TCP 4 .Os .Sh NAME @@ -480,12 +480,6 @@ This improves the effectiveness of retransmissions particular in environments with ACK thinning or burst loss events, as chances to run out of the ACK clock are reduced, preventing lengthy and performance reducing RTO based loss recovery (default is true). -.It Va do_prr_conservative -While doing Proportional Rate Reduction, remain strictly in a packet conserving -mode, sending only one new packet for each ACK received. -Helpful when a misconfigured token bucket traffic policer causes persistent -high losses leading to RTO, but reduces PRR effectiveness in more common settings -(default is false). .It Va do_tcpdrain Flush packets in the .Tn TCP diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c index 15b7311f4605..9f7e42b582c6 100644 --- a/sys/netinet/tcp_input.c +++ b/sys/netinet/tcp_input.c @@ -159,11 +159,6 @@ SYSCTL_INT(_net_inet_tcp, OID_AUTO, drop_synfin, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(drop_synfin), 0, "Drop TCP packets with SYN+FIN set"); -VNET_DEFINE(int, tcp_do_prr_conservative) = 0; -SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_prr_conservative, CTLFLAG_VNET | CTLFLAG_RW, - &VNET_NAME(tcp_do_prr_conservative), 0, - "Do conservative Proportional Rate Reduction"); - VNET_DEFINE(int, tcp_do_prr) = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_prr, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_do_prr), 1, @@ -1524,7 +1519,8 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, int drop_hdrlen, int tlen, uint8_t iptos) { uint16_t thflags; - int acked, ourfinisacked, needoutput = 0, sack_changed; + int acked, ourfinisacked, needoutput = 0; + sackstatus_t sack_changed; int rstreason, todrop, win, incforsyn = 0; uint32_t tiwin; uint16_t nsegs; @@ -1539,7 +1535,7 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, thflags = tcp_get_flags(th); tp->sackhint.last_sack_ack = 0; - sack_changed = 0; + sack_changed = SACK_NOCHANGE; nsegs = max(1, m->m_pkthdr.lro_nsegs); NET_EPOCH_ASSERT(); @@ -2582,7 +2578,7 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, */ if (th->th_ack != tp->snd_una || (tcp_is_sack_recovery(tp, &to) && - !sack_changed)) + (sack_changed == SACK_NOCHANGE))) break; else if (!tcp_timer_active(tp, TT_REXMT)) tp->t_dupacks = 0; @@ -2591,8 +2587,9 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, cc_ack_received(tp, th, nsegs, CC_DUPACK); if (V_tcp_do_prr && - IN_FASTRECOVERY(tp->t_flags)) { - tcp_do_prr_ack(tp, th, &to); + IN_FASTRECOVERY(tp->t_flags) && + (tp->t_flags & TF_SACK_PERMIT)) { + tcp_do_prr_ack(tp, th, &to, sack_changed); } else if (tcp_is_sack_recovery(tp, &to) && IN_FASTRECOVERY(tp->t_flags)) { int awnd; @@ -2667,8 +2664,12 @@ enter_recovery: * cc_cong_signal. */ if (tcp_is_sack_recovery(tp, &to)) { + /* + * Exclude Limited Transmit + * segments here + */ tp->sackhint.prr_delivered = - tp->sackhint.sacked_bytes; + maxseg; } else { tp->sackhint.prr_delivered = imin(tp->snd_max - tp->snd_una, @@ -2771,7 +2772,7 @@ enter_recovery: * counted as dupacks here. */ if (tcp_is_sack_recovery(tp, &to) && - sack_changed) { + (sack_changed != SACK_NOCHANGE)) { tp->t_dupacks++; /* limit overhead by setting maxseg last */ if (!IN_FASTRECOVERY(tp->t_flags) && @@ -2797,7 +2798,7 @@ resume_partialack: if (V_tcp_do_prr && to.to_flags & TOF_SACK) { tcp_timer_activate(tp, TT_REXMT, 0); tp->t_rtttime = 0; - tcp_do_prr_ack(tp, th, &to); + tcp_do_prr_ack(tp, th, &to, sack_changed); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); } else @@ -2811,7 +2812,11 @@ resume_partialack: if (V_tcp_do_prr) { tp->sackhint.delivered_data = BYTES_THIS_ACK(tp, th); tp->snd_fack = th->th_ack; - tcp_do_prr_ack(tp, th, &to); + /* + * During ECN cwnd reduction + * always use PRR-SSRB + */ + tcp_do_prr_ack(tp, th, &to, SACK_CHANGE); (void) tcp_output(tp); } } else @@ -3934,7 +3939,7 @@ tcp_mssopt(struct in_conninfo *inc) } void -tcp_do_prr_ack(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to) +tcp_do_prr_ack(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to, sackstatus_t sack_changed) { int snd_cnt = 0, limit = 0, del_data = 0, pipe = 0; int maxseg = tcp_maxseg(tp); @@ -3974,7 +3979,17 @@ tcp_do_prr_ack(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to) tp->snd_ssthresh, tp->sackhint.recover_fs) - tp->sackhint.prr_out + maxseg - 1; } else { - if (V_tcp_do_prr_conservative || (del_data == 0)) + /* + * PRR 6937bis heuristic: + * - A partial ack without SACK block beneath snd_recover + * indicates further loss. + * - An SACK scoreboard update adding a new hole indicates + * further loss, so be conservative and send at most one + * segment. + * - Prevent ACK splitting attacks, by being conservative + * when no new data is acked. + */ + if ((sack_changed == SACK_NEWLOSS) || (del_data == 0)) limit = tp->sackhint.prr_delivered - tp->sackhint.prr_out; else diff --git a/sys/netinet/tcp_sack.c b/sys/netinet/tcp_sack.c index fca4b99ee83b..05aaa0b4a662 100644 --- a/sys/netinet/tcp_sack.c +++ b/sys/netinet/tcp_sack.c @@ -544,21 +544,23 @@ tcp_sackhole_remove(struct tcpcb *tp, struct sackhole *hole) * Process cumulative ACK and the TCP SACK option to update the scoreboard. * tp->snd_holes is an ordered list of holes (oldest to newest, in terms of * the sequence space). - * Returns 1 if incoming ACK has previously unknown SACK information, - * 0 otherwise. + * Returns SACK_NEWLOSS if incoming ACK indicates ongoing loss (hole split, new hole), + * SACK_CHANGE if incoming ACK has previously unknown SACK information, + * SACK_NOCHANGE otherwise. */ -int +sackstatus_t tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) { struct sackhole *cur, *temp; struct sackblk sack, sack_blocks[TCP_MAX_SACK + 1], *sblkp; - int i, j, num_sack_blks, sack_changed; + int i, j, num_sack_blks; + sackstatus_t sack_changed; int delivered_data, left_edge_delta; INP_WLOCK_ASSERT(tptoinpcb(tp)); num_sack_blks = 0; - sack_changed = 0; + sack_changed = SACK_NOCHANGE; delivered_data = 0; left_edge_delta = 0; /* @@ -577,7 +579,7 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) if (SEQ_LT(tp->snd_fack, th_ack)) { delivered_data += th_ack - tp->snd_una; tp->snd_fack = th_ack; - sack_changed = 1; + sack_changed = SACK_CHANGE; } } /* @@ -667,7 +669,7 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) delivered_data += sblkp->end - sblkp->start; tp->snd_fack = sblkp->end; sblkp--; - sack_changed = 1; + sack_changed = SACK_NEWLOSS; } else { /* * Append a new SACK hole at the tail. If the @@ -681,7 +683,7 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) tp->snd_fack = sblkp->end; /* Go to the previous sack block. */ sblkp--; - sack_changed = 1; + sack_changed = SACK_CHANGE; } else { /* * We failed to add a new hole based on the current @@ -698,7 +700,12 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) SEQ_LT(tp->snd_fack, sblkp->end)) { delivered_data += sblkp->end - tp->snd_fack; tp->snd_fack = sblkp->end; - sack_changed = 1; + /* + * While the Scoreboard didn't change in + * size, we only ended up here because + * some SACK data had to be dismissed. + */ + sack_changed = SACK_NEWLOSS; } } } @@ -706,7 +713,7 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) /* fack is advanced. */ delivered_data += sblkp->end - tp->snd_fack; tp->snd_fack = sblkp->end; - sack_changed = 1; + sack_changed = SACK_CHANGE; } cur = TAILQ_LAST(&tp->snd_holes, sackhole_head); /* Last SACK hole. */ /* @@ -734,7 +741,7 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) (SEQ_MIN(cur->rxmit, cur->end) - cur->start); KASSERT(tp->sackhint.sack_bytes_rexmit >= 0, ("sackhint bytes rtx >= 0")); - sack_changed = 1; + sack_changed = SACK_CHANGE; if (SEQ_LEQ(sblkp->start, cur->start)) { /* Data acks at least the beginning of hole. */ if (SEQ_GEQ(sblkp->end, cur->end)) { @@ -771,6 +778,7 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) */ temp = tcp_sackhole_insert(tp, sblkp->end, cur->end, cur); + sack_changed = SACK_NEWLOSS; if (temp != NULL) { if (SEQ_GT(cur->rxmit, temp->rxmit)) { temp->rxmit = cur->rxmit; @@ -808,7 +816,7 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) * DupAck for this. Also required * for RFC6675 rescue retransmission. */ - sack_changed = 0; + sack_changed = SACK_NOCHANGE; tp->sackhint.delivered_data = delivered_data; tp->sackhint.sacked_bytes += delivered_data - left_edge_delta; KASSERT((delivered_data >= 0), ("delivered_data < 0")); diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h index cd5712132522..624de66e3e23 100644 --- a/sys/netinet/tcp_var.h +++ b/sys/netinet/tcp_var.h @@ -502,6 +502,13 @@ struct tcptemp { struct tcphdr tt_t; }; +/* SACK scoreboard update status */ +typedef enum { + SACK_NOCHANGE = 0, + SACK_CHANGE, + SACK_NEWLOSS +} sackstatus_t; + /* Enable TCP/UDP tunneling port */ #define TCP_TUNNELING_PORT_MIN 0 #define TCP_TUNNELING_PORT_MAX 65535 @@ -1301,7 +1308,6 @@ VNET_DECLARE(struct inpcbinfo, tcbinfo); #define V_tcp_do_lrd VNET(tcp_do_lrd) #define V_tcp_do_prr VNET(tcp_do_prr) -#define V_tcp_do_prr_conservative VNET(tcp_do_prr_conservative) #define V_tcp_do_newcwv VNET(tcp_do_newcwv) #define V_drop_synfin VNET(drop_synfin) #define V_path_mtu_discovery VNET(path_mtu_discovery) @@ -1481,7 +1487,8 @@ extern struct protosw tcp6_protosw; /* shared for TOE */ uint32_t tcp_new_ts_offset(struct in_conninfo *); tcp_seq tcp_new_isn(struct in_conninfo *); -int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); +sackstatus_t + tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); int tcp_dsack_block_exists(struct tcpcb *); void tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq); void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend); @@ -1489,7 +1496,7 @@ void tcp_clean_dsack_blocks(struct tcpcb *tp); void tcp_clean_sackreport(struct tcpcb *tp); void tcp_sack_adjust(struct tcpcb *tp); struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt); -void tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *); +void tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *, sackstatus_t); void tcp_lost_retransmission(struct tcpcb *, struct tcphdr *); void tcp_sack_partialack(struct tcpcb *, struct tcphdr *); void tcp_free_sackholes(struct tcpcb *tp);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?202312161029.3BGATmPS077967>