#include #include int sysctl_tcp_recovery __read_mostly = TCP_RACK_LOSS_DETECTION; 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: * * dupthresh: 3 OOO packets delivered (packet count) * FACK: sequence delta to highest sacked sequence (sequence space) * RACK: sent time delta to the latest delivered packet (time domain) * * The advantage of RACK is it applies to both original and retransmitted * packet and therefore is robust against tail losses. Another advantage * is being more resilient to reordering by simply allowing some * "settling delay", instead of tweaking the dupthresh. * * 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. */ 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; *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. */ reo_wnd = 1000; 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) { struct tcp_skb_cb *scb = TCP_SKB_CB(skb); if (skb == tcp_send_head(sk)) break; /* Skip ones already (s)acked */ if (!after(scb->end_seq, tp->snd_una) || scb->sacked & TCPCB_SACKED_ACKED) continue; 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 (remaining < 0) { tcp_rack_mark_skb_lost(sk, skb); continue; } /* 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 */ break; } } } 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 && !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) { /* If the sacked packet was retransmitted, it's ambiguous * whether the retransmission or the original (or the prior * retransmission) was sacked. * * If the original is lost, there is no ambiguity. Otherwise * we assume the original can be delayed up to aRTT + min_rtt. * the aRTT term is bounded by the fast recovery or timeout, * so it's at least one RTT (i.e., retransmission is at least * an RTT later). */ 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); } 35logplain -rw-r--r--bcm-ns2.h2915logplain -rw-r--r--bcm-nsp.h2148logplain -rw-r--r--bcm21664.h1984logplain -rw-r--r--bcm281xx.h2456logplain -rw-r--r--bcm2835-aux.h635logplain -rw-r--r--bcm2835.h1962logplain -rw-r--r--berlin2.h1034logplain -rw-r--r--berlin2q.h695logplain -rw-r--r--clps711x-clock.h718logplain -rw-r--r--efm32-cmu.h1112logplain -rw-r--r--exynos-audss-clk.h597logplain -rw-r--r--exynos3250.h9083logplain -rw-r--r--exynos4.h8284logplain -rw-r--r--exynos4415.h9828logplain -rw-r--r--exynos5250.h4616logplain -rw-r--r--exynos5260-clk.h14876logplain -rw-r--r--exynos5410.h1689logplain -rw-r--r--exynos5420.h6857logplain -rw-r--r--exynos5433.h45372logplain -rw-r--r--exynos5440.h1141logplain -rw-r--r--exynos7-clk.h5281logplain -rw-r--r--gxbb-aoclkc.h2866logplain -rw-r--r--gxbb-clkc.h592logplain -rw-r--r--hi3516cv300-clock.h1668logplain -rw-r--r--hi3519-clock.h1328logplain -rw-r--r--hi3620-clock.h4496logplain -rw-r--r--hi6220-clock.h4508logplain -rw-r--r--hip04-clock.h1137logplain -rw-r--r--histb-clock.h2012logplain -rw-r--r--hix5hd2-clock.h2415logplain -rw-r--r--imx1-clock.h1055logplain -rw-r--r--imx21-clock.h2461logplain -rw-r--r--imx27-clock.h3494logplain -rw-r--r--imx5-clock.h7212logplain -rw-r--r--imx6qdl-clock.h9593logplain -rw-r--r--imx6sl-clock.h5849logplain -rw-r--r--imx6sx-clock.h9099logplain -rw-r--r--imx6ul-clock.h8203logplain -rw-r--r--imx7d-clock.h15974logplain -rw-r--r--jz4740-cgu.h1028logplain -rw-r--r--jz4780-cgu.h2470logplain -rw-r--r--lpc18xx-ccu.h2134logplain -rw-r--r--lpc18xx-cgu.h1142logplain -rw-r--r--lpc32xx-clock.h1633logplain -rw-r--r--lsi,axm5516-clks.h974logplain -rw-r--r--marvell,mmp2.h2022logplain -rw-r--r--marvell,pxa168.h1654logplain -rw-r--r--marvell,pxa1928.h1535logplain -rw-r--r--marvell,pxa910.h1598logplain -rw-r--r--maxim,max77620.h632logplain -rw-r--r--maxim,max77686.h648logplain -rw-r--r--maxim,max77802.h630logplain -rw-r--r--meson8b-clkc.h523logplain -rw-r--r--microchip,pic32-clock.h1150logplain -rw-r--r--mpc512x-clock.h2236logplain -rw-r--r--mt2701-clk.h13832logplain -rw-r--r--mt8135-clk.h5641logplain -rw-r--r--mt8173-clk.h9293logplain -rw-r--r--oxsemi,ox810se.h1002logplain -rw-r--r--oxsemi,ox820.h1203logplain -rw-r--r--pistachio-clk.h4863logplain -rw-r--r--pxa-clock.h1715logplain -rw-r--r--qcom,gcc-apq8084.h12872logplain -rw-r--r--qcom,gcc-ipq4019.h5423logplain -rw-r--r--qcom,gcc-ipq806x.h8574logplain -rw-r--r--qcom,gcc-mdm9615.h9497logplain -rw-r--r--qcom,gcc-msm8660.h7932logplain -rw-r--r--qcom,gcc-msm8916.h6190logplain -rw-r--r--qcom,gcc-msm8960.h9342logplain -rw-r--r--qcom,gcc-msm8974.h12340logplain -rw-r--r--qcom,gcc-msm8994.h4858logplain -rw-r--r--qcom,gcc-msm8996.h12575logplain -rw-r--r--qcom,lcc-ipq806x.h899logplain -rw-r--r--qcom,lcc-mdm9615.h1701logplain -rw-r--r--qcom,lcc-msm8960.h1616logplain -rw-r--r--qcom,mmcc-apq8084.h5722logplain -rw-r--r--qcom,mmcc-msm8960.h4109logplain -rw-r--r--qcom,mmcc-msm8974.h5223logplain -rw-r--r--qcom,mmcc-msm8996.h9403logplain -rw-r--r--qcom,rpmcc.h2101logplain -rw-r--r--r7s72100-clock.h1218logplain -rw-r--r--r8a73a4-clock.h1596logplain -rw-r--r--r8a7740-clock.h1992logplain -rw-r--r--r8a7743-cpg-mssr.h1269logplain -rw-r--r--r8a7745-cpg-mssr.h1298logplain -rw-r--r--r8a7778-clock.h1855logplain -rw-r--r--r8a7779-clock.h1647logplain -rw-r--r--r8a7790-clock.h4367logplain -rw-r--r--r8a7791-clock.h4388logplain -rw-r--r--r8a7792-clock.h2562logplain -rw-r--r--r8a7793-clock.h4561logplain -rw-r--r--r8a7794-clock.h3679logplain -rw-r--r--r8a7795-cpg-mssr.h1890logplain -rw-r--r--r8a7796-cpg-mssr.h2066logplain -rw-r--r--renesas-cpg-mssr.h542logplain -rw-r--r--rk1108-cru.h6605logplain -rw-r--r--rk3036-cru.h4584logplain -rw-r--r--rk3066a-cru.h1068logplain -rw-r--r--rk3188-cru-common.h6105logplain -rw-r--r--rk3188-cru.h1435logplain