From 61d2bcae99f66a640b3dd9632180209143fb5512 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Mon, 1 Feb 2016 21:03:07 -0800 Subject: tcp: fastopen: accept data/FIN present in SYNACK message RFC 7413 (TCP Fast Open) 4.2.2 states that the SYNACK message MAY include data and/or FIN This patch adds support for the client side : If we receive a SYNACK with payload or FIN, queue the skb instead of ignoring it. Since we already support the same for SYN, we refactor the existing code and reuse it. Note we need to clone the skb, so this operation might fail under memory pressure. Sara Dickinson pointed out FreeBSD server Fast Open implementation was planned to generate such SYNACK in the future. The server side might be implemented on linux later. Reported-by: Sara Dickinson Signed-off-by: Eric Dumazet Signed-off-by: Yuchung Cheng Signed-off-by: Neal Cardwell Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 1c2a73406261..4add3eb40e58 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -5509,6 +5509,9 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack, tp->syn_data_acked = tp->syn_data; if (tp->syn_data_acked) NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPFASTOPENACTIVE); + + tcp_fastopen_add_skb(sk, synack); + return false; } -- cgit v1.2.1 From e3e17b773bfe45462b7f3fae20c550025975cb13 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sat, 6 Feb 2016 11:16:28 -0800 Subject: tcp: fastopen: call tcp_fin() if FIN present in SYNACK When we acknowledge a FIN, it is not enough to ack the sequence number and queue the skb into receive queue. We also have to call tcp_fin() to properly update socket state and send proper poll() notifications. It seems we also had the problem if we received a SYN packet with the FIN flag set, but it does not seem an urgent issue, as no known implementation can do that. Fixes: 61d2bcae99f6 ("tcp: fastopen: accept data/FIN present in SYNACK message") Signed-off-by: Eric Dumazet Cc: Yuchung Cheng Cc: Neal Cardwell Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 4add3eb40e58..8194a250a01e 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -3995,7 +3995,7 @@ void tcp_reset(struct sock *sk) * * If we are in FINWAIT-2, a received FIN moves us to TIME-WAIT. */ -static void tcp_fin(struct sock *sk) +void tcp_fin(struct sock *sk) { struct tcp_sock *tp = tcp_sk(sk); -- cgit v1.2.1 From e662ca40de846e0a2be6326a7c4668326ddb194c Mon Sep 17 00:00:00 2001 From: Yuchung Cheng Date: Tue, 2 Feb 2016 10:33:04 -0800 Subject: tcp: retransmit after recovery processing and congestion control The retransmission and F-RTO transmission currently happen inside recovery state processing (tcp_fastretrans_alert) but before congestion control. This refactoring moves the logic after both s.t. we can determine how much to send (cwnd) before deciding what to send. Signed-off-by: Yuchung Cheng Signed-off-by: Neal Cardwell Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 58 +++++++++++++++++++++++++++++++++++++++++----------- 1 file changed, 46 insertions(+), 12 deletions(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 8194a250a01e..84a4ab9c05d1 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -126,6 +126,10 @@ int sysctl_tcp_invalid_ratelimit __read_mostly = HZ/2; #define TCP_REMNANT (TCP_FLAG_FIN|TCP_FLAG_URG|TCP_FLAG_SYN|TCP_FLAG_PSH) #define TCP_HP_BITS (~(TCP_RESERVED_BITS|TCP_FLAG_PSH)) +#define REXMIT_NONE 0 /* no loss recovery to do */ +#define REXMIT_LOST 1 /* retransmit packets marked lost */ +#define REXMIT_NEW 2 /* FRTO-style transmit of unsent/new packets */ + /* Adapt the MSS value used to make delayed ack decision to the * real world. */ @@ -2662,7 +2666,8 @@ static void tcp_enter_recovery(struct sock *sk, bool ece_ack) /* Process an ACK in CA_Loss state. Move to CA_Open if lost data are * recovered or spurious. Otherwise retransmits more on partial ACKs. */ -static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack) +static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack, + int *rexmit) { struct tcp_sock *tp = tcp_sk(sk); bool recovered = !before(tp->snd_una, tp->high_seq); @@ -2684,10 +2689,15 @@ static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack) tp->frto = 0; /* Step 3.a. loss was real */ } else if (flag & FLAG_SND_UNA_ADVANCED && !recovered) { tp->high_seq = tp->snd_nxt; - __tcp_push_pending_frames(sk, tcp_current_mss(sk), - TCP_NAGLE_OFF); - if (after(tp->snd_nxt, tp->high_seq)) - return; /* Step 2.b */ + /* Step 2.b. Try send new data (but deferred until cwnd + * is updated in tcp_ack()). Otherwise fall back to + * the conventional recovery. + */ + if (tcp_send_head(sk) && + after(tcp_wnd_end(tp), tp->snd_nxt)) { + *rexmit = REXMIT_NEW; + return; + } tp->frto = 0; } } @@ -2706,7 +2716,7 @@ static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack) else if (flag & FLAG_SND_UNA_ADVANCED) tcp_reset_reno_sack(tp); } - tcp_xmit_retransmit_queue(sk); + *rexmit = REXMIT_LOST; } /* Undo during fast recovery after partial ACK. */ @@ -2756,7 +2766,7 @@ static bool tcp_try_undo_partial(struct sock *sk, const int acked, */ static void tcp_fastretrans_alert(struct sock *sk, const int acked, const int prior_unsacked, - bool is_dupack, int flag) + bool is_dupack, int flag, int *rexmit) { struct inet_connection_sock *icsk = inet_csk(sk); struct tcp_sock *tp = tcp_sk(sk); @@ -2831,7 +2841,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, } break; case TCP_CA_Loss: - tcp_process_loss(sk, flag, is_dupack); + tcp_process_loss(sk, flag, is_dupack, rexmit); if (icsk->icsk_ca_state != TCP_CA_Open && !(flag & FLAG_LOST_RETRANS)) return; @@ -2871,7 +2881,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, if (do_lost) tcp_update_scoreboard(sk, fast_rexmit); tcp_cwnd_reduction(sk, prior_unsacked, fast_rexmit, flag); - tcp_xmit_retransmit_queue(sk); + *rexmit = REXMIT_LOST; } /* Kathleen Nichols' algorithm for tracking the minimum value of @@ -3506,6 +3516,27 @@ static inline void tcp_in_ack_event(struct sock *sk, u32 flags) icsk->icsk_ca_ops->in_ack_event(sk, flags); } +/* Congestion control has updated the cwnd already. So if we're in + * loss recovery then now we do any new sends (for FRTO) or + * retransmits (for CA_Loss or CA_recovery) that make sense. + */ +static void tcp_xmit_recovery(struct sock *sk, int rexmit) +{ + struct tcp_sock *tp = tcp_sk(sk); + + if (rexmit == REXMIT_NONE) + return; + + if (unlikely(rexmit == 2)) { + __tcp_push_pending_frames(sk, tcp_current_mss(sk), + TCP_NAGLE_OFF); + if (after(tp->snd_nxt, tp->high_seq)) + return; + tp->frto = 0; + } + tcp_xmit_retransmit_queue(sk); +} + /* This routine deals with incoming acks, but not outgoing ones. */ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) { @@ -3520,6 +3551,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) int prior_packets = tp->packets_out; const int prior_unsacked = tp->packets_out - tp->sacked_out; int acked = 0; /* Number of packets newly acked */ + int rexmit = REXMIT_NONE; /* Flag to (re)transmit to recover losses */ sack_state.first_sackt.v64 = 0; @@ -3616,7 +3648,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) if (tcp_ack_is_dubious(sk, flag)) { is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP)); tcp_fastretrans_alert(sk, acked, prior_unsacked, - is_dupack, flag); + is_dupack, flag, &rexmit); } if (tp->tlp_high_seq) tcp_process_tlp_ack(sk, ack, flag); @@ -3634,13 +3666,14 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) if (icsk->icsk_pending == ICSK_TIME_RETRANS) tcp_schedule_loss_probe(sk); tcp_update_pacing_rate(sk); + tcp_xmit_recovery(sk, rexmit); return 1; no_queue: /* If data was DSACKed, see if we can undo a cwnd reduction. */ if (flag & FLAG_DSACKING_ACK) tcp_fastretrans_alert(sk, acked, prior_unsacked, - is_dupack, flag); + is_dupack, flag, &rexmit); /* If this ack opens up a zero window, clear backoff. It was * being used to time the probes, and is probably far higher than * it needs to be for normal retransmission. @@ -3664,7 +3697,8 @@ old_ack: flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una, &sack_state); tcp_fastretrans_alert(sk, acked, prior_unsacked, - is_dupack, flag); + is_dupack, flag, &rexmit); + tcp_xmit_recovery(sk, rexmit); } SOCK_DEBUG(sk, "Ack %u before %u:%u\n", ack, tp->snd_una, tp->snd_nxt); -- cgit v1.2.1 From 31ba0c10723e9eba378f96de1d1a9426129949e1 Mon Sep 17 00:00:00 2001 From: Yuchung Cheng Date: Tue, 2 Feb 2016 10:33:05 -0800 Subject: tcp: move cwnd reduction after recovery state procesing Currently the cwnd is reduced and increased in various different places. The reduction happens in various places in the recovery state processing (tcp_fastretrans_alert) while the increase happens afterward. A better sequence is to identify lost packets and update the congestion control state (icsk_ca_state) first. Then base on the new state, up/down the cwnd in one central place. It's more clear to reason cwnd changes. Signed-off-by: Yuchung Cheng Signed-off-by: Neal Cardwell Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 60 ++++++++++++++++++++++++---------------------------- 1 file changed, 28 insertions(+), 32 deletions(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 84a4ab9c05d1..dc810df53e90 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -2471,14 +2471,12 @@ static void tcp_init_cwnd_reduction(struct sock *sk) tcp_ecn_queue_cwr(tp); } -static void tcp_cwnd_reduction(struct sock *sk, const int prior_unsacked, - int fast_rexmit, int flag) +static void tcp_cwnd_reduction(struct sock *sk, int newly_acked_sacked, + int flag) { struct tcp_sock *tp = tcp_sk(sk); int sndcnt = 0; int delta = tp->snd_ssthresh - tcp_packets_in_flight(tp); - int newly_acked_sacked = prior_unsacked - - (tp->packets_out - tp->sacked_out); if (newly_acked_sacked <= 0 || WARN_ON_ONCE(!tp->prior_cwnd)) return; @@ -2496,7 +2494,8 @@ static void tcp_cwnd_reduction(struct sock *sk, const int prior_unsacked, } else { sndcnt = min(delta, newly_acked_sacked); } - sndcnt = max(sndcnt, (fast_rexmit ? 1 : 0)); + /* Force a fast retransmit upon entering fast recovery */ + sndcnt = max(sndcnt, (tp->prr_out ? 0 : 1)); tp->snd_cwnd = tcp_packets_in_flight(tp) + sndcnt; } @@ -2541,7 +2540,7 @@ static void tcp_try_keep_open(struct sock *sk) } } -static void tcp_try_to_open(struct sock *sk, int flag, const int prior_unsacked) +static void tcp_try_to_open(struct sock *sk, int flag) { struct tcp_sock *tp = tcp_sk(sk); @@ -2555,8 +2554,6 @@ static void tcp_try_to_open(struct sock *sk, int flag, const int prior_unsacked) if (inet_csk(sk)->icsk_ca_state != TCP_CA_CWR) { tcp_try_keep_open(sk); - } else { - tcp_cwnd_reduction(sk, prior_unsacked, 0, flag); } } @@ -2720,8 +2717,7 @@ static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack, } /* Undo during fast recovery after partial ACK. */ -static bool tcp_try_undo_partial(struct sock *sk, const int acked, - const int prior_unsacked, int flag) +static bool tcp_try_undo_partial(struct sock *sk, const int acked) { struct tcp_sock *tp = tcp_sk(sk); @@ -2736,10 +2732,8 @@ static bool tcp_try_undo_partial(struct sock *sk, const int acked, * can undo. Otherwise we clock out new packets but do not * mark more packets lost or retransmit more. */ - if (tp->retrans_out) { - tcp_cwnd_reduction(sk, prior_unsacked, 0, flag); + if (tp->retrans_out) return true; - } if (!tcp_any_retrans_done(sk)) tp->retrans_stamp = 0; @@ -2758,21 +2752,21 @@ static bool tcp_try_undo_partial(struct sock *sk, const int acked, * taking into account both packets sitting in receiver's buffer and * packets lost by network. * - * Besides that it does CWND reduction, when packet loss is detected - * and changes state of machine. + * Besides that it updates the congestion state when packet loss or ECN + * is detected. But it does not reduce the cwnd, it is done by the + * congestion control later. * * It does _not_ decide what to send, it is made in function * tcp_xmit_retransmit_queue(). */ static void tcp_fastretrans_alert(struct sock *sk, const int acked, - const int prior_unsacked, - bool is_dupack, int flag, int *rexmit) + bool is_dupack, int *ack_flag, int *rexmit) { struct inet_connection_sock *icsk = inet_csk(sk); struct tcp_sock *tp = tcp_sk(sk); + int fast_rexmit = 0, flag = *ack_flag; bool do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) && (tcp_fackets_out(tp) > tp->reordering)); - int fast_rexmit = 0; if (WARN_ON(!tp->packets_out && tp->sacked_out)) tp->sacked_out = 0; @@ -2819,8 +2813,10 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, /* Use RACK to detect loss */ if (sysctl_tcp_recovery & TCP_RACK_LOST_RETRANS && - tcp_rack_mark_lost(sk)) + tcp_rack_mark_lost(sk)) { flag |= FLAG_LOST_RETRANS; + *ack_flag |= FLAG_LOST_RETRANS; + } /* E. Process state. */ switch (icsk->icsk_ca_state) { @@ -2829,7 +2825,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, if (tcp_is_reno(tp) && is_dupack) tcp_add_reno_sack(sk); } else { - if (tcp_try_undo_partial(sk, acked, prior_unsacked, flag)) + if (tcp_try_undo_partial(sk, acked)) return; /* Partial ACK arrived. Force fast retransmit. */ do_lost = tcp_is_reno(tp) || @@ -2858,7 +2854,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, tcp_try_undo_dsack(sk); if (!tcp_time_to_recover(sk, flag)) { - tcp_try_to_open(sk, flag, prior_unsacked); + tcp_try_to_open(sk, flag); return; } @@ -2880,7 +2876,6 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, if (do_lost) tcp_update_scoreboard(sk, fast_rexmit); - tcp_cwnd_reduction(sk, prior_unsacked, fast_rexmit, flag); *rexmit = REXMIT_LOST; } @@ -3306,9 +3301,6 @@ static inline bool tcp_ack_is_dubious(const struct sock *sk, const int flag) /* Decide wheather to run the increase function of congestion control. */ static inline bool tcp_may_raise_cwnd(const struct sock *sk, const int flag) { - if (tcp_in_cwnd_reduction(sk)) - return false; - /* If reordering is high then always grow cwnd whenever data is * delivered regardless of its ordering. Otherwise stay conservative * and only grow cwnd on in-order delivery (RFC5681). A stretched ACK w/ @@ -3551,6 +3543,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) int prior_packets = tp->packets_out; const int prior_unsacked = tp->packets_out - tp->sacked_out; int acked = 0; /* Number of packets newly acked */ + int acked_sacked; /* Number of packets newly acked or sacked */ int rexmit = REXMIT_NONE; /* Flag to (re)transmit to recover losses */ sack_state.first_sackt.v64 = 0; @@ -3647,15 +3640,20 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) if (tcp_ack_is_dubious(sk, flag)) { is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP)); - tcp_fastretrans_alert(sk, acked, prior_unsacked, - is_dupack, flag, &rexmit); + tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit); } if (tp->tlp_high_seq) tcp_process_tlp_ack(sk, ack, flag); + acked_sacked = prior_unsacked - (tp->packets_out - tp->sacked_out); /* Advance cwnd if state allows */ - if (tcp_may_raise_cwnd(sk, flag)) + if (tcp_in_cwnd_reduction(sk)) { + /* Reduce cwnd if state mandates */ + tcp_cwnd_reduction(sk, acked_sacked, flag); + } else if (tcp_may_raise_cwnd(sk, flag)) { + /* Advance cwnd if state allows */ tcp_cong_avoid(sk, ack, acked); + } if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP)) { struct dst_entry *dst = __sk_dst_get(sk); @@ -3672,8 +3670,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) no_queue: /* If data was DSACKed, see if we can undo a cwnd reduction. */ if (flag & FLAG_DSACKING_ACK) - tcp_fastretrans_alert(sk, acked, prior_unsacked, - is_dupack, flag, &rexmit); + tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit); /* If this ack opens up a zero window, clear backoff. It was * being used to time the probes, and is probably far higher than * it needs to be for normal retransmission. @@ -3696,8 +3693,7 @@ old_ack: if (TCP_SKB_CB(skb)->sacked) { flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una, &sack_state); - tcp_fastretrans_alert(sk, acked, prior_unsacked, - is_dupack, flag, &rexmit); + tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit); tcp_xmit_recovery(sk, rexmit); } -- cgit v1.2.1 From ddf1af6fa00e772fdb67a7d22cb83fac2b8968a8 Mon Sep 17 00:00:00 2001 From: Yuchung Cheng Date: Tue, 2 Feb 2016 10:33:06 -0800 Subject: tcp: new delivery accounting This patch changes the accounting of how many packets are newly acked or sacked when the sender receives an ACK. The current approach basically computes newly_acked_sacked = (prior_packets - prior_sacked) - (tp->packets_out - tp->sacked_out) where prior_packets and prior_sacked out are snapshot at the beginning of the ACK processing. The new approach tracks the delivery information via a new TCP state variable "delivered" which monotically increases as new packets are delivered in order or out-of-order. The reason for this change is that the current approach is brittle that produces negative or inaccurate estimate. 1) For non-SACK connections, an ACK that advances the SND.UNA could reset the DUPACK counters (tp->sacked_out) in tcp_process_loss() or tcp_fastretrans_alert(). This inflates the inflight suddenly and causes under-estimate or even negative estimate. Here is a real example: before after (processing ACK) packets_out 75 73 sacked_out 23 0 ca state Loss Open The old approach computes (75-23) - (73 - 0) = -21 delivered while the new approach computes 1 delivered since it considers the 2nd-24th packets are delivered OOO. 2) MSS change would re-count packets_out and sacked_out so the estimate is in-accurate and can even become negative. E.g., the inflight is doubled when MSS is halved. 3) Spurious retransmission signaled by DSACK is not accounted The new approach is simpler and more robust. For SACK connections, tp->delivered increments as packets are being acked or sacked in SACK and ACK processing. For non-sack connections, it's done in tcp_remove_reno_sacks() and tcp_add_reno_sack(). When an ACK advances the SND.UNA, tp->delivered is incremented by the number of packets ACKed (less the current number of DUPACKs received plus one packet hole). Upon receiving a DUPACK, tp->delivered is incremented assuming one out-of-order packet is delivered. Upon receiving a DSACK, tp->delivered is incremtened assuming one retransmission is delivered in tcp_sacktag_write_queue(). Signed-off-by: Yuchung Cheng Signed-off-by: Neal Cardwell Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index dc810df53e90..2d690b3f0a7b 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -1214,6 +1214,7 @@ static u8 tcp_sacktag_one(struct sock *sk, sacked |= TCPCB_SACKED_ACKED; state->flag |= FLAG_DATA_SACKED; tp->sacked_out += pcount; + tp->delivered += pcount; /* Out-of-order packets delivered */ fack_count += pcount; @@ -1825,8 +1826,12 @@ static void tcp_check_reno_reordering(struct sock *sk, const int addend) static void tcp_add_reno_sack(struct sock *sk) { struct tcp_sock *tp = tcp_sk(sk); + u32 prior_sacked = tp->sacked_out; + tp->sacked_out++; tcp_check_reno_reordering(sk, 0); + if (tp->sacked_out > prior_sacked) + tp->delivered++; /* Some out-of-order packet is delivered */ tcp_verify_left_out(tp); } @@ -1838,6 +1843,7 @@ static void tcp_remove_reno_sacks(struct sock *sk, int acked) if (acked > 0) { /* One ACK acked hole. The rest eat duplicate ACKs. */ + tp->delivered += max_t(int, acked - tp->sacked_out, 1); if (acked - 1 >= tp->sacked_out) tp->sacked_out = 0; else @@ -3156,10 +3162,13 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, flag |= FLAG_ORIG_SACK_ACKED; } - if (sacked & TCPCB_SACKED_ACKED) + if (sacked & TCPCB_SACKED_ACKED) { tp->sacked_out -= acked_pcount; - else if (tcp_is_sack(tp) && !tcp_skb_spurious_retrans(tp, skb)) - tcp_rack_advance(tp, &skb->skb_mstamp, sacked); + } else if (tcp_is_sack(tp)) { + tp->delivered += acked_pcount; + if (!tcp_skb_spurious_retrans(tp, skb)) + tcp_rack_advance(tp, &skb->skb_mstamp, sacked); + } if (sacked & TCPCB_LOST) tp->lost_out -= acked_pcount; @@ -3541,9 +3550,9 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) bool is_dupack = false; u32 prior_fackets; int prior_packets = tp->packets_out; - const int prior_unsacked = tp->packets_out - tp->sacked_out; + u32 prior_delivered = tp->delivered; int acked = 0; /* Number of packets newly acked */ - int acked_sacked; /* Number of packets newly acked or sacked */ + u32 acked_sacked; /* Number of packets newly acked or sacked */ int rexmit = REXMIT_NONE; /* Flag to (re)transmit to recover losses */ sack_state.first_sackt.v64 = 0; @@ -3645,7 +3654,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) if (tp->tlp_high_seq) tcp_process_tlp_ack(sk, ack, flag); - acked_sacked = prior_unsacked - (tp->packets_out - tp->sacked_out); + acked_sacked = tp->delivered - prior_delivered; /* Advance cwnd if state allows */ if (tcp_in_cwnd_reduction(sk)) { /* Reduce cwnd if state mandates */ -- cgit v1.2.1 From 3ebd88710584d494b670e54b2c339e3be290956c Mon Sep 17 00:00:00 2001 From: Yuchung Cheng Date: Tue, 2 Feb 2016 10:33:07 -0800 Subject: tcp: refactor pkts acked accounting A small refactoring that gets number of packets cumulatively acked from tcp_clean_rtx_queue() directly. Signed-off-by: Yuchung Cheng Signed-off-by: Neal Cardwell Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 2d690b3f0a7b..40824b380ef8 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -3104,7 +3104,7 @@ static void tcp_ack_tstamp(struct sock *sk, struct sk_buff *skb, * arrived at the other end. */ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, - u32 prior_snd_una, + u32 prior_snd_una, int *acked, struct tcp_sacktag_state *sack) { const struct inet_connection_sock *icsk = inet_csk(sk); @@ -3277,6 +3277,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, } } #endif + *acked = pkts_acked; return flag; } @@ -3642,10 +3643,8 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) goto no_queue; /* See if we can take anything off of the retransmit queue. */ - acked = tp->packets_out; - flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una, + flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una, &acked, &sack_state); - acked -= tp->packets_out; if (tcp_ack_is_dubious(sk, flag)) { is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP)); -- cgit v1.2.1 From 2d14a4def4fc87cb2d2712f7841b45189d75e301 Mon Sep 17 00:00:00 2001 From: Yuchung Cheng Date: Tue, 2 Feb 2016 10:33:08 -0800 Subject: tcp: make congestion control more robust against reordering This change enables congestion control to update cwnd based on not only packet cumulatively acked but also packets delivered out-of-order. This makes congestion control robust against packet reordering because it may raise cwnd as long as packets are being delivered once reordering has been detected (i.e., it only cares the amount of packets delivered, not the ordering among them). Signed-off-by: Yuchung Cheng Signed-off-by: Neal Cardwell Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 40824b380ef8..d598ff408cb9 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -3660,7 +3660,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) tcp_cwnd_reduction(sk, acked_sacked, flag); } else if (tcp_may_raise_cwnd(sk, flag)) { /* Advance cwnd if state allows */ - tcp_cong_avoid(sk, ack, acked); + tcp_cong_avoid(sk, ack, acked_sacked); } if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP)) { -- cgit v1.2.1 From d452e6caf8367cc70cf940c24a6a6cc2d521d3c1 Mon Sep 17 00:00:00 2001 From: Yuchung Cheng Date: Tue, 2 Feb 2016 10:33:09 -0800 Subject: tcp: tcp_cong_control helper Refactor and consolidate cwnd and rate updates into a new function tcp_cong_control(). Signed-off-by: Yuchung Cheng Signed-off-by: Neal Cardwell Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 31 +++++++++++++++++++------------ 1 file changed, 19 insertions(+), 12 deletions(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index d598ff408cb9..596c1cb6759a 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -3323,6 +3323,24 @@ static inline bool tcp_may_raise_cwnd(const struct sock *sk, const int flag) return flag & FLAG_DATA_ACKED; } +/* The "ultimate" congestion control function that aims to replace the rigid + * cwnd increase and decrease control (tcp_cong_avoid,tcp_*cwnd_reduction). + * It's called toward the end of processing an ACK with precise rate + * information. All transmission or retransmission are delayed afterwards. + */ +static void tcp_cong_control(struct sock *sk, u32 ack, u32 acked_sacked, + int flag) +{ + if (tcp_in_cwnd_reduction(sk)) { + /* Reduce cwnd if state mandates */ + tcp_cwnd_reduction(sk, acked_sacked, flag); + } else if (tcp_may_raise_cwnd(sk, flag)) { + /* Advance cwnd if state allows */ + tcp_cong_avoid(sk, ack, acked_sacked); + } + tcp_update_pacing_rate(sk); +} + /* Check that window update is acceptable. * The function assumes that snd_una<=ack<=snd_next. */ @@ -3553,7 +3571,6 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) int prior_packets = tp->packets_out; u32 prior_delivered = tp->delivered; int acked = 0; /* Number of packets newly acked */ - u32 acked_sacked; /* Number of packets newly acked or sacked */ int rexmit = REXMIT_NONE; /* Flag to (re)transmit to recover losses */ sack_state.first_sackt.v64 = 0; @@ -3653,16 +3670,6 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) if (tp->tlp_high_seq) tcp_process_tlp_ack(sk, ack, flag); - acked_sacked = tp->delivered - prior_delivered; - /* Advance cwnd if state allows */ - if (tcp_in_cwnd_reduction(sk)) { - /* Reduce cwnd if state mandates */ - tcp_cwnd_reduction(sk, acked_sacked, flag); - } else if (tcp_may_raise_cwnd(sk, flag)) { - /* Advance cwnd if state allows */ - tcp_cong_avoid(sk, ack, acked_sacked); - } - if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP)) { struct dst_entry *dst = __sk_dst_get(sk); if (dst) @@ -3671,7 +3678,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) if (icsk->icsk_pending == ICSK_TIME_RETRANS) tcp_schedule_loss_probe(sk); - tcp_update_pacing_rate(sk); + tcp_cong_control(sk, ack, tp->delivered - prior_delivered, flag); tcp_xmit_recovery(sk, rexmit); return 1; -- cgit v1.2.1 From 12ed8244ed8b31b023ea6d2851fd8b15f2999e9b Mon Sep 17 00:00:00 2001 From: Nikolay Borisov Date: Wed, 3 Feb 2016 09:46:51 +0200 Subject: ipv4: Namespaceify tcp syncookies sysctl knob Signed-off-by: Nikolay Borisov Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 596c1cb6759a..b17aba42a368 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -6163,9 +6163,10 @@ static bool tcp_syn_flood_action(const struct sock *sk, struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue; const char *msg = "Dropping request"; bool want_cookie = false; + struct net *net = sock_net(sk); #ifdef CONFIG_SYN_COOKIES - if (sysctl_tcp_syncookies) { + if (net->ipv4.sysctl_tcp_syncookies) { msg = "Sending cookies"; want_cookie = true; NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDOCOOKIES); @@ -6174,7 +6175,7 @@ static bool tcp_syn_flood_action(const struct sock *sk, NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDROP); if (!queue->synflood_warned && - sysctl_tcp_syncookies != 2 && + net->ipv4.sysctl_tcp_syncookies != 2 && xchg(&queue->synflood_warned, 1) == 0) pr_info("%s: Possible SYN flooding on port %d. %s. Check SNMP counters.\n", proto, ntohs(tcp_hdr(skb)->dest), msg); @@ -6207,6 +6208,7 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, __u32 isn = TCP_SKB_CB(skb)->tcp_tw_isn; struct tcp_options_received tmp_opt; struct tcp_sock *tp = tcp_sk(sk); + struct net *net = sock_net(sk); struct sock *fastopen_sk = NULL; struct dst_entry *dst = NULL; struct request_sock *req; @@ -6217,7 +6219,7 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, * limitations, they conserve resources and peer is * evidently real one. */ - if ((sysctl_tcp_syncookies == 2 || + if ((net->ipv4.sysctl_tcp_syncookies == 2 || inet_csk_reqsk_queue_is_full(sk)) && !isn) { want_cookie = tcp_syn_flood_action(sk, skb, rsk_ops->slab_name); if (!want_cookie) @@ -6283,7 +6285,7 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, } } /* Kill the following clause, if you dislike this way. */ - else if (!sysctl_tcp_syncookies && + else if (!net->ipv4.sysctl_tcp_syncookies && (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) < (sysctl_max_syn_backlog >> 2)) && !tcp_peer_is_proven(req, dst, false, -- cgit v1.2.1 From 1043e25ff96a1efc7bd34d11f5f32203a28a3bd7 Mon Sep 17 00:00:00 2001 From: Nikolay Borisov Date: Wed, 3 Feb 2016 09:46:52 +0200 Subject: ipv4: Namespaceify tcp reordering sysctl knob Signed-off-by: Nikolay Borisov Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index b17aba42a368..5ee6fe0d152d 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -80,9 +80,7 @@ int sysctl_tcp_timestamps __read_mostly = 1; int sysctl_tcp_window_scaling __read_mostly = 1; int sysctl_tcp_sack __read_mostly = 1; int sysctl_tcp_fack __read_mostly = 1; -int sysctl_tcp_reordering __read_mostly = TCP_FASTRETRANS_THRESH; int sysctl_tcp_max_reordering __read_mostly = 300; -EXPORT_SYMBOL(sysctl_tcp_reordering); int sysctl_tcp_dsack __read_mostly = 1; int sysctl_tcp_app_win __read_mostly = 31; int sysctl_tcp_adv_win_scale __read_mostly = 1; @@ -1883,6 +1881,7 @@ void tcp_enter_loss(struct sock *sk) { const struct inet_connection_sock *icsk = inet_csk(sk); struct tcp_sock *tp = tcp_sk(sk); + struct net *net = sock_net(sk); struct sk_buff *skb; bool new_recovery = icsk->icsk_ca_state < TCP_CA_Recovery; bool is_reneg; /* is receiver reneging on SACKs? */ @@ -1933,9 +1932,9 @@ void tcp_enter_loss(struct sock *sk) * suggests that the degree of reordering is over-estimated. */ if (icsk->icsk_ca_state <= TCP_CA_Disorder && - tp->sacked_out >= sysctl_tcp_reordering) + tp->sacked_out >= net->ipv4.sysctl_tcp_reordering) tp->reordering = min_t(unsigned int, tp->reordering, - sysctl_tcp_reordering); + net->ipv4.sysctl_tcp_reordering); tcp_set_ca_state(sk, TCP_CA_Loss); tp->high_seq = tp->snd_nxt; tcp_ecn_queue_cwr(tp); @@ -2119,6 +2118,7 @@ static bool tcp_time_to_recover(struct sock *sk, int flag) { struct tcp_sock *tp = tcp_sk(sk); __u32 packets_out; + int tcp_reordering = sock_net(sk)->ipv4.sysctl_tcp_reordering; /* Trick#1: The loss is proven. */ if (tp->lost_out) @@ -2133,7 +2133,7 @@ static bool tcp_time_to_recover(struct sock *sk, int flag) */ packets_out = tp->packets_out; if (packets_out <= tp->reordering && - tp->sacked_out >= max_t(__u32, packets_out/2, sysctl_tcp_reordering) && + tp->sacked_out >= max_t(__u32, packets_out/2, tcp_reordering) && !tcp_may_send_now(sk)) { /* We have nothing to send. This connection is limited * either by receiver window or by application. @@ -3317,7 +3317,7 @@ static inline bool tcp_may_raise_cwnd(const struct sock *sk, const int flag) * new SACK or ECE mark may first advance cwnd here and later reduce * cwnd in tcp_fastretrans_alert() based on more states. */ - if (tcp_sk(sk)->reordering > sysctl_tcp_reordering) + if (tcp_sk(sk)->reordering > sock_net(sk)->ipv4.sysctl_tcp_reordering) return flag & FLAG_FORWARD_PROGRESS; return flag & FLAG_DATA_ACKED; -- cgit v1.2.1