X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=net%2Fipv4%2Ftcp_input.c;h=c26076fb890e5f935631dc4e45f96b90cee15521;hb=refs%2Fheads%2Fvserver;hp=ece320e60f710cce99173133d7e4a4a39dbff113;hpb=c7b5ebbddf7bcd3651947760f423e3783bbe6573;p=linux-2.6.git diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index ece320e60..c26076fb8 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -7,7 +7,7 @@ * * Version: $Id: tcp_input.c,v 1.243 2002/02/01 22:01:04 davem Exp $ * - * Authors: Ross Biro, + * Authors: Ross Biro * Fred N. van Kempen, * Mark Evans, * Corey Minyard @@ -42,7 +42,7 @@ * Andi Kleen : Moved open_request checking here * and process RSTs for open_requests. * Andi Kleen : Better prune_queue, and other fixes. - * Andrey Savochkin: Fix RTT measurements in the presnce of + * Andrey Savochkin: Fix RTT measurements in the presence of * timestamps. * Andrey Savochkin: Check sequence numbers correctly when * removing SACKs due to in sequence incoming @@ -61,47 +61,35 @@ * Panu Kuhlberg: Experimental audit of TCP (re)transmission * engine. Lots of bugs are found. * Pasi Sarolahti: F-RTO for dealing with spurious RTOs - * Angelo Dell'Aera: TCP Westwood+ support */ -#include #include #include #include #include #include #include - -int sysctl_tcp_timestamps = 1; -int sysctl_tcp_window_scaling = 1; -int sysctl_tcp_sack = 1; -int sysctl_tcp_fack = 1; -int sysctl_tcp_reordering = TCP_FASTRETRANS_THRESH; -int sysctl_tcp_ecn; -int sysctl_tcp_dsack = 1; -int sysctl_tcp_app_win = 31; -int sysctl_tcp_adv_win_scale = 2; - -int sysctl_tcp_stdurg; -int sysctl_tcp_rfc1337; -int sysctl_tcp_max_orphans = NR_FILE; -int sysctl_tcp_frto; -int sysctl_tcp_nometrics_save; -int sysctl_tcp_westwood; -int sysctl_tcp_vegas_cong_avoid; - -int sysctl_tcp_moderate_rcvbuf = 1; - -/* Default values of the Vegas variables, in fixed-point representation - * with V_PARAM_SHIFT bits to the right of the binary point. - */ -#define V_PARAM_SHIFT 1 -int sysctl_tcp_vegas_alpha = 1< +#include + +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_ecn __read_mostly; +int sysctl_tcp_dsack __read_mostly = 1; +int sysctl_tcp_app_win __read_mostly = 31; +int sysctl_tcp_adv_win_scale __read_mostly = 2; + +int sysctl_tcp_stdurg __read_mostly; +int sysctl_tcp_rfc1337 __read_mostly; +int sysctl_tcp_max_orphans __read_mostly = NR_FILE; +int sysctl_tcp_frto __read_mostly; +int sysctl_tcp_nometrics_save __read_mostly; + +int sysctl_tcp_moderate_rcvbuf __read_mostly = 1; +int sysctl_tcp_abc __read_mostly; #define FLAG_DATA 0x01 /* Incoming frame contained data. */ #define FLAG_WIN_UPDATE 0x02 /* Incoming ACK was a window update. */ @@ -118,28 +106,30 @@ int sysctl_tcp_bic_low_window = 14; #define FLAG_CA_ALERT (FLAG_DATA_SACKED|FLAG_ECE) #define FLAG_FORWARD_PROGRESS (FLAG_ACKED|FLAG_DATA_SACKED) -#define IsReno(tp) ((tp)->sack_ok == 0) -#define IsFack(tp) ((tp)->sack_ok & 2) -#define IsDSack(tp) ((tp)->sack_ok & 4) +#define IsReno(tp) ((tp)->rx_opt.sack_ok == 0) +#define IsFack(tp) ((tp)->rx_opt.sack_ok & 2) +#define IsDSack(tp) ((tp)->rx_opt.sack_ok & 4) #define TCP_REMNANT (TCP_FLAG_FIN|TCP_FLAG_URG|TCP_FLAG_SYN|TCP_FLAG_PSH) /* Adapt the MSS value used to make delayed ack decision to the * real world. */ -static __inline__ void tcp_measure_rcv_mss(struct tcp_opt *tp, struct sk_buff *skb) +static void tcp_measure_rcv_mss(struct sock *sk, + const struct sk_buff *skb) { - unsigned int len, lss; + struct inet_connection_sock *icsk = inet_csk(sk); + const unsigned int lss = icsk->icsk_ack.last_seg_size; + unsigned int len; - lss = tp->ack.last_seg_size; - tp->ack.last_seg_size = 0; + icsk->icsk_ack.last_seg_size = 0; /* skb->len may jitter because of SACKs, even if peer * sends good full-sized frames. */ - len = skb->len; - if (len >= tp->ack.rcv_mss) { - tp->ack.rcv_mss = len; + len = skb_shinfo(skb)->gso_size ?: skb->len; + if (len >= icsk->icsk_ack.rcv_mss) { + icsk->icsk_ack.rcv_mss = len; } else { /* Otherwise, we make more careful check taking into account, * that SACKs block is variable. @@ -159,41 +149,46 @@ static __inline__ void tcp_measure_rcv_mss(struct tcp_opt *tp, struct sk_buff *s * tcp header plus fixed timestamp option length. * Resulting "len" is MSS free of SACK jitter. */ - len -= tp->tcp_header_len; - tp->ack.last_seg_size = len; + len -= tcp_sk(sk)->tcp_header_len; + icsk->icsk_ack.last_seg_size = len; if (len == lss) { - tp->ack.rcv_mss = len; + icsk->icsk_ack.rcv_mss = len; return; } } - tp->ack.pending |= TCP_ACK_PUSHED; + if (icsk->icsk_ack.pending & ICSK_ACK_PUSHED) + icsk->icsk_ack.pending |= ICSK_ACK_PUSHED2; + icsk->icsk_ack.pending |= ICSK_ACK_PUSHED; } } -static void tcp_incr_quickack(struct tcp_opt *tp) +static void tcp_incr_quickack(struct sock *sk) { - unsigned quickacks = tp->rcv_wnd/(2*tp->ack.rcv_mss); + struct inet_connection_sock *icsk = inet_csk(sk); + unsigned quickacks = tcp_sk(sk)->rcv_wnd / (2 * icsk->icsk_ack.rcv_mss); if (quickacks==0) quickacks=2; - if (quickacks > tp->ack.quick) - tp->ack.quick = min(quickacks, TCP_MAX_QUICKACKS); + if (quickacks > icsk->icsk_ack.quick) + icsk->icsk_ack.quick = min(quickacks, TCP_MAX_QUICKACKS); } -void tcp_enter_quickack_mode(struct tcp_opt *tp) +void tcp_enter_quickack_mode(struct sock *sk) { - tcp_incr_quickack(tp); - tp->ack.pingpong = 0; - tp->ack.ato = TCP_ATO_MIN; + struct inet_connection_sock *icsk = inet_csk(sk); + tcp_incr_quickack(sk); + icsk->icsk_ack.pingpong = 0; + icsk->icsk_ack.ato = TCP_ATO_MIN; } /* Send ACKs quickly, if "quick" count is not exhausted * and the session is not interactive. */ -static __inline__ int tcp_in_quickack_mode(struct tcp_opt *tp) +static inline int tcp_in_quickack_mode(const struct sock *sk) { - return (tp->ack.quick && !tp->ack.pingpong); + const struct inet_connection_sock *icsk = inet_csk(sk); + return icsk->icsk_ack.quick && !icsk->icsk_ack.pingpong; } /* Buffer size and advertised window tuning. @@ -203,7 +198,7 @@ static __inline__ int tcp_in_quickack_mode(struct tcp_opt *tp) static void tcp_fixup_sndbuf(struct sock *sk) { - int sndmem = tcp_sk(sk)->mss_clamp + MAX_TCP_HEADER + 16 + + int sndmem = tcp_sk(sk)->rx_opt.mss_clamp + MAX_TCP_HEADER + 16 + sizeof(struct sk_buff); if (sk->sk_sndbuf < 3 * sndmem) @@ -231,21 +226,21 @@ static void tcp_fixup_sndbuf(struct sock *sk) * of receiver window. Check #2. * * The scheme does not work when sender sends good segments opening - * window and then starts to feed us spagetti. But it should work + * window and then starts to feed us spaghetti. But it should work * in common situations. Otherwise, we have to rely on queue collapsing. */ /* Slow part of check#2. */ -static int -__tcp_grow_window(struct sock *sk, struct tcp_opt *tp, struct sk_buff *skb) +static int __tcp_grow_window(const struct sock *sk, struct tcp_sock *tp, + const struct sk_buff *skb) { /* Optimize this! */ int truesize = tcp_win_from_space(skb->truesize)/2; - int window = tcp_full_space(sk)/2; + int window = tcp_win_from_space(sysctl_tcp_rmem[2])/2; while (tp->rcv_ssthresh <= window) { if (truesize <= skb->len) - return 2*tp->ack.rcv_mss; + return 2 * inet_csk(sk)->icsk_ack.rcv_mss; truesize >>= 1; window >>= 1; @@ -253,8 +248,8 @@ __tcp_grow_window(struct sock *sk, struct tcp_opt *tp, struct sk_buff *skb) return 0; } -static __inline__ void -tcp_grow_window(struct sock *sk, struct tcp_opt *tp, struct sk_buff *skb) +static void tcp_grow_window(struct sock *sk, struct tcp_sock *tp, + struct sk_buff *skb) { /* Check #1 */ if (tp->rcv_ssthresh < tp->window_clamp && @@ -272,7 +267,7 @@ tcp_grow_window(struct sock *sk, struct tcp_opt *tp, struct sk_buff *skb) if (incr) { tp->rcv_ssthresh = min(tp->rcv_ssthresh + incr, tp->window_clamp); - tp->ack.quick |= 1; + inet_csk(sk)->icsk_ack.quick |= 1; } } } @@ -281,11 +276,11 @@ tcp_grow_window(struct sock *sk, struct tcp_opt *tp, struct sk_buff *skb) static void tcp_fixup_rcvbuf(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); int rcvmem = tp->advmss + MAX_TCP_HEADER + 16 + sizeof(struct sk_buff); /* Try to select rcvbuf so that 4 mss-sized segments - * will fit to window and correspoding skbs will fit to our rcvbuf. + * will fit to window and corresponding skbs will fit to our rcvbuf. * (was 3; 4 is minimum to allow fast retransmit to work.) */ while (tcp_win_from_space(rcvmem) < tp->advmss) @@ -294,12 +289,12 @@ static void tcp_fixup_rcvbuf(struct sock *sk) sk->sk_rcvbuf = min(4 * rcvmem, sysctl_tcp_rmem[2]); } -/* 4. Try to fixup all. It is made iimediately after connection enters +/* 4. Try to fixup all. It is made immediately after connection enters * established state. */ static void tcp_init_buffer_space(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); int maxwin; if (!(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) @@ -330,51 +325,42 @@ static void tcp_init_buffer_space(struct sock *sk) tp->snd_cwnd_stamp = tcp_time_stamp; } -static void init_bictcp(struct tcp_opt *tp) +/* 5. Recalculate window clamp after socket hit its memory bounds. */ +static void tcp_clamp_window(struct sock *sk, struct tcp_sock *tp) { - tp->bictcp.cnt = 0; + struct inet_connection_sock *icsk = inet_csk(sk); - tp->bictcp.last_max_cwnd = 0; - tp->bictcp.last_cwnd = 0; - tp->bictcp.last_stamp = 0; + icsk->icsk_ack.quick = 0; + + if (sk->sk_rcvbuf < sysctl_tcp_rmem[2] && + !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) && + !tcp_memory_pressure && + atomic_read(&tcp_memory_allocated) < sysctl_tcp_mem[0]) { + sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc), + sysctl_tcp_rmem[2]); + } + if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) + tp->rcv_ssthresh = min(tp->window_clamp, 2U*tp->advmss); } -/* 5. Recalculate window clamp after socket hit its memory bounds. */ -static void tcp_clamp_window(struct sock *sk, struct tcp_opt *tp) -{ - struct sk_buff *skb; - unsigned int app_win = tp->rcv_nxt - tp->copied_seq; - int ofo_win = 0; - tp->ack.quick = 0; +/* Initialize RCV_MSS value. + * RCV_MSS is an our guess about MSS used by the peer. + * We haven't any direct information about the MSS. + * It's better to underestimate the RCV_MSS rather than overestimate. + * Overestimations make us ACKing less frequently than needed. + * Underestimations are more easy to detect and fix by tcp_measure_rcv_mss(). + */ +void tcp_initialize_rcv_mss(struct sock *sk) +{ + struct tcp_sock *tp = tcp_sk(sk); + unsigned int hint = min_t(unsigned int, tp->advmss, tp->mss_cache); - skb_queue_walk(&tp->out_of_order_queue, skb) { - ofo_win += skb->len; - } + hint = min(hint, tp->rcv_wnd/2); + hint = min(hint, TCP_MIN_RCVMSS); + hint = max(hint, TCP_MIN_MSS); - /* If overcommit is due to out of order segments, - * do not clamp window. Try to expand rcvbuf instead. - */ - if (ofo_win) { - if (sk->sk_rcvbuf < sysctl_tcp_rmem[2] && - !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) && - !tcp_memory_pressure && - atomic_read(&tcp_memory_allocated) < sysctl_tcp_mem[0]) - sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc), - sysctl_tcp_rmem[2]); - } - if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) { - app_win += ofo_win; - if (atomic_read(&sk->sk_rmem_alloc) >= 2 * sk->sk_rcvbuf) - app_win >>= 1; - if (app_win > tp->ack.rcv_mss) - app_win -= tp->ack.rcv_mss; - app_win = max(app_win, 2U*tp->advmss); - - if (!ofo_win) - tp->window_clamp = min(tp->window_clamp, app_win); - tp->rcv_ssthresh = min(tp->window_clamp, 2U*tp->advmss); - } + inet_csk(sk)->icsk_ack.rcv_mss = hint; } /* Receiver "autotuning" code. @@ -388,7 +374,7 @@ static void tcp_clamp_window(struct sock *sk, struct tcp_opt *tp) * though this reference is out of date. A new paper * is pending. */ -static void tcp_rcv_rtt_update(struct tcp_opt *tp, u32 sample, int win_dep) +static void tcp_rcv_rtt_update(struct tcp_sock *tp, u32 sample, int win_dep) { u32 new_sample = tp->rcv_rtt_est.rtt; long m = sample; @@ -403,8 +389,8 @@ static void tcp_rcv_rtt_update(struct tcp_opt *tp, u32 sample, int win_dep) * are stalled on filesystem I/O. * * Also, since we are only going for a minimum in the - * non-timestamp case, we do not smoothe things out - * else with timestamps disabled convergance takes too + * non-timestamp case, we do not smooth things out + * else with timestamps disabled convergence takes too * long. */ if (!win_dep) { @@ -413,7 +399,7 @@ static void tcp_rcv_rtt_update(struct tcp_opt *tp, u32 sample, int win_dep) } else if (m < new_sample) new_sample = m << 3; } else { - /* No previous mesaure. */ + /* No previous measure. */ new_sample = m << 3; } @@ -421,7 +407,7 @@ static void tcp_rcv_rtt_update(struct tcp_opt *tp, u32 sample, int win_dep) tp->rcv_rtt_est.rtt = new_sample; } -static inline void tcp_rcv_rtt_measure(struct tcp_opt *tp) +static inline void tcp_rcv_rtt_measure(struct tcp_sock *tp) { if (tp->rcv_rtt_est.time == 0) goto new_measure; @@ -436,12 +422,13 @@ new_measure: tp->rcv_rtt_est.time = tcp_time_stamp; } -static inline void tcp_rcv_rtt_measure_ts(struct tcp_opt *tp, struct sk_buff *skb) +static inline void tcp_rcv_rtt_measure_ts(struct sock *sk, const struct sk_buff *skb) { - if (tp->rcv_tsecr && + struct tcp_sock *tp = tcp_sk(sk); + if (tp->rx_opt.rcv_tsecr && (TCP_SKB_CB(skb)->end_seq - - TCP_SKB_CB(skb)->seq >= tp->ack.rcv_mss)) - tcp_rcv_rtt_update(tp, tcp_time_stamp - tp->rcv_tsecr, 0); + TCP_SKB_CB(skb)->seq >= inet_csk(sk)->icsk_ack.rcv_mss)) + tcp_rcv_rtt_update(tp, tcp_time_stamp - tp->rx_opt.rcv_tsecr, 0); } /* @@ -450,7 +437,7 @@ static inline void tcp_rcv_rtt_measure_ts(struct tcp_opt *tp, struct sk_buff *sk */ void tcp_rcv_space_adjust(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); int time; int space; @@ -471,7 +458,8 @@ void tcp_rcv_space_adjust(struct sock *sk) tp->rcvq_space.space = space; - if (sysctl_tcp_moderate_rcvbuf) { + if (sysctl_tcp_moderate_rcvbuf && + !(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) { int new_clamp = space; /* Receive space grows, normalize in order to @@ -511,43 +499,44 @@ new_measure: * each ACK we send, he increments snd_cwnd and transmits more of his * queue. -DaveM */ -static void tcp_event_data_recv(struct sock *sk, struct tcp_opt *tp, struct sk_buff *skb) +static void tcp_event_data_recv(struct sock *sk, struct tcp_sock *tp, struct sk_buff *skb) { + struct inet_connection_sock *icsk = inet_csk(sk); u32 now; - tcp_schedule_ack(tp); + inet_csk_schedule_ack(sk); - tcp_measure_rcv_mss(tp, skb); + tcp_measure_rcv_mss(sk, skb); tcp_rcv_rtt_measure(tp); now = tcp_time_stamp; - if (!tp->ack.ato) { + if (!icsk->icsk_ack.ato) { /* The _first_ data packet received, initialize * delayed ACK engine. */ - tcp_incr_quickack(tp); - tp->ack.ato = TCP_ATO_MIN; + tcp_incr_quickack(sk); + icsk->icsk_ack.ato = TCP_ATO_MIN; } else { - int m = now - tp->ack.lrcvtime; + int m = now - icsk->icsk_ack.lrcvtime; if (m <= TCP_ATO_MIN/2) { /* The fastest case is the first. */ - tp->ack.ato = (tp->ack.ato>>1) + TCP_ATO_MIN/2; - } else if (m < tp->ack.ato) { - tp->ack.ato = (tp->ack.ato>>1) + m; - if (tp->ack.ato > tp->rto) - tp->ack.ato = tp->rto; - } else if (m > tp->rto) { - /* Too long gap. Apparently sender falled to + icsk->icsk_ack.ato = (icsk->icsk_ack.ato >> 1) + TCP_ATO_MIN / 2; + } else if (m < icsk->icsk_ack.ato) { + icsk->icsk_ack.ato = (icsk->icsk_ack.ato >> 1) + m; + if (icsk->icsk_ack.ato > icsk->icsk_rto) + icsk->icsk_ack.ato = icsk->icsk_rto; + } else if (m > icsk->icsk_rto) { + /* Too long gap. Apparently sender failed to * restart window, so that we send ACKs quickly. */ - tcp_incr_quickack(tp); + tcp_incr_quickack(sk); sk_stream_mem_reclaim(sk); } } - tp->ack.lrcvtime = now; + icsk->icsk_ack.lrcvtime = now; TCP_ECN_check_ce(tp, skb); @@ -555,45 +544,6 @@ static void tcp_event_data_recv(struct sock *sk, struct tcp_opt *tp, struct sk_b tcp_grow_window(sk, tp, skb); } -/* When starting a new connection, pin down the current choice of - * congestion algorithm. - */ -void tcp_ca_init(struct tcp_opt *tp) -{ - if (sysctl_tcp_westwood) - tp->adv_cong = TCP_WESTWOOD; - else if (sysctl_tcp_bic) - tp->adv_cong = TCP_BIC; - else if (sysctl_tcp_vegas_cong_avoid) { - tp->adv_cong = TCP_VEGAS; - tp->vegas.baseRTT = 0x7fffffff; - tcp_vegas_enable(tp); - } -} - -/* Do RTT sampling needed for Vegas. - * Basically we: - * o min-filter RTT samples from within an RTT to get the current - * propagation delay + queuing delay (we are min-filtering to try to - * avoid the effects of delayed ACKs) - * o min-filter RTT samples from a much longer window (forever for now) - * to find the propagation delay (baseRTT) - */ -static inline void vegas_rtt_calc(struct tcp_opt *tp, __u32 rtt) -{ - __u32 vrtt = rtt + 1; /* Never allow zero rtt or baseRTT */ - - /* Filter to find propagation delay: */ - if (vrtt < tp->vegas.baseRTT) - tp->vegas.baseRTT = vrtt; - - /* Find the min RTT during the last RTT to find - * the current prop. delay + queuing delay: - */ - tp->vegas.minRTT = min(tp->vegas.minRTT, vrtt); - tp->vegas.cntRTT++; -} - /* Called to compute a smoothed rtt estimate. The data fed to this * routine either comes from timestamps, or from segments that were * known _not_ to have been retransmitted [see Karn/Partridge @@ -603,13 +553,11 @@ static inline void vegas_rtt_calc(struct tcp_opt *tp, __u32 rtt) * To save cycles in the RFC 1323 implementation it was better to break * it up into three procedures. -- erics */ -static void tcp_rtt_estimator(struct tcp_opt *tp, __u32 mrtt) +static void tcp_rtt_estimator(struct sock *sk, const __u32 mrtt) { + struct tcp_sock *tp = tcp_sk(sk); long m = mrtt; /* RTT */ - if (tcp_vegas_enabled(tp)) - vegas_rtt_calc(tp, mrtt); - /* The following amusing code comes from Jacobson's * article in SIGCOMM '88. Note that rtt and mdev * are scaled versions of rtt and mean deviation. @@ -621,7 +569,7 @@ static void tcp_rtt_estimator(struct tcp_opt *tp, __u32 mrtt) * * Funny. This algorithm seems to be very broken. * These formulae increase RTO, when it should be decreased, increase - * too slowly, when it should be incresed fastly, decrease too fastly + * too slowly, when it should be increased quickly, decrease too quickly * etc. I guess in BSD RTO takes ONE value, so that it is absolutely * does not matter how to _calculate_ it. Seems, it was trap * that VJ failed to avoid. 8) @@ -666,15 +614,14 @@ static void tcp_rtt_estimator(struct tcp_opt *tp, __u32 mrtt) tp->mdev_max = tp->rttvar = max(tp->mdev, TCP_RTO_MIN); tp->rtt_seq = tp->snd_nxt; } - - tcp_westwood_update_rtt(tp, tp->srtt >> 3); } /* Calculate rto without backoff. This is the second half of Van Jacobson's * routine referred to above. */ -static __inline__ void tcp_set_rto(struct tcp_opt *tp) +static inline void tcp_set_rto(struct sock *sk) { + const struct tcp_sock *tp = tcp_sk(sk); /* Old crap is replaced with new one. 8) * * More seriously: @@ -683,24 +630,24 @@ static __inline__ void tcp_set_rto(struct tcp_opt *tp) * at least by solaris and freebsd. "Erratic ACKs" has _nothing_ * to do with delayed acks, because at cwnd>2 true delack timeout * is invisible. Actually, Linux-2.4 also generates erratic - * ACKs in some curcumstances. + * ACKs in some circumstances. */ - tp->rto = (tp->srtt >> 3) + tp->rttvar; + inet_csk(sk)->icsk_rto = (tp->srtt >> 3) + tp->rttvar; /* 2. Fixups made earlier cannot be right. * If we do not estimate RTO correctly without them, * all the algo is pure shit and should be replaced - * with correct one. It is exaclty, which we pretend to do. + * with correct one. It is exactly, which we pretend to do. */ } /* NOTE: clamping at TCP_RTO_MIN is not required, current algo * guarantees that rto is higher. */ -static __inline__ void tcp_bound_rto(struct tcp_opt *tp) +static inline void tcp_bound_rto(struct sock *sk) { - if (tp->rto > TCP_RTO_MAX) - tp->rto = TCP_RTO_MAX; + if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX) + inet_csk(sk)->icsk_rto = TCP_RTO_MAX; } /* Save metrics learned by this TCP session. @@ -709,7 +656,7 @@ static __inline__ void tcp_bound_rto(struct tcp_opt *tp) */ void tcp_update_metrics(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); struct dst_entry *dst = __sk_dst_get(sk); if (sysctl_tcp_nometrics_save) @@ -718,9 +665,10 @@ void tcp_update_metrics(struct sock *sk) dst_confirm(dst); if (dst && (dst->flags&DST_HOST)) { + const struct inet_connection_sock *icsk = inet_csk(sk); int m; - if (tp->backoff || !tp->srtt) { + if (icsk->icsk_backoff || !tp->srtt) { /* This session failed to estimate rtt. Why? * Probably, no packets returned in time. * Reset our results. @@ -769,7 +717,7 @@ void tcp_update_metrics(struct sock *sk) tp->snd_cwnd > dst_metric(dst, RTAX_CWND)) dst->metrics[RTAX_CWND-1] = tp->snd_cwnd; } else if (tp->snd_cwnd > tp->snd_ssthresh && - tp->ca_state == TCP_CA_Open) { + icsk->icsk_ca_state == TCP_CA_Open) { /* Cong. avoidance phase, cwnd is reliable. */ if (!dst_metric_locked(dst, RTAX_SSTHRESH)) dst->metrics[RTAX_SSTHRESH-1] = @@ -797,24 +745,45 @@ void tcp_update_metrics(struct sock *sk) } /* Numbers are taken from RFC2414. */ -__u32 tcp_init_cwnd(struct tcp_opt *tp, struct dst_entry *dst) +__u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst) { __u32 cwnd = (dst ? dst_metric(dst, RTAX_INITCWND) : 0); if (!cwnd) { - if (tp->mss_cache_std > 1460) + if (tp->mss_cache > 1460) cwnd = 2; else - cwnd = (tp->mss_cache_std > 1095) ? 3 : 4; + cwnd = (tp->mss_cache > 1095) ? 3 : 4; } return min_t(__u32, cwnd, tp->snd_cwnd_clamp); } +/* Set slow start threshold and cwnd not falling to slow start */ +void tcp_enter_cwr(struct sock *sk) +{ + struct tcp_sock *tp = tcp_sk(sk); + + tp->prior_ssthresh = 0; + tp->bytes_acked = 0; + if (inet_csk(sk)->icsk_ca_state < TCP_CA_CWR) { + tp->undo_marker = 0; + tp->snd_ssthresh = inet_csk(sk)->icsk_ca_ops->ssthresh(sk); + tp->snd_cwnd = min(tp->snd_cwnd, + tcp_packets_in_flight(tp) + 1U); + tp->snd_cwnd_cnt = 0; + tp->high_seq = tp->snd_nxt; + tp->snd_cwnd_stamp = tcp_time_stamp; + TCP_ECN_queue_cwr(tp); + + tcp_set_ca_state(sk, TCP_CA_CWR); + } +} + /* Initialize metrics on socket. */ static void tcp_init_metrics(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); struct dst_entry *dst = __sk_dst_get(sk); if (dst == NULL) @@ -831,7 +800,7 @@ static void tcp_init_metrics(struct sock *sk) } if (dst_metric(dst, RTAX_REORDERING) && tp->reordering != dst_metric(dst, RTAX_REORDERING)) { - tp->sack_ok &= ~2; + tp->rx_opt.sack_ok &= ~2; tp->reordering = dst_metric(dst, RTAX_REORDERING); } @@ -847,7 +816,7 @@ static void tcp_init_metrics(struct sock *sk) * to make it more realistic. * * A bit of theory. RTT is time passed after "normal" sized packet - * is sent until it is ACKed. In normal curcumstances sending small + * is sent until it is ACKed. In normal circumstances sending small * packets force peer to delay ACKs and calculation is correct too. * The algorithm is adaptive and, provided we follow specs, it * NEVER underestimate RTT. BUT! If peer tries to make some clever @@ -863,9 +832,9 @@ static void tcp_init_metrics(struct sock *sk) tp->mdev = dst_metric(dst, RTAX_RTTVAR); tp->mdev_max = tp->rttvar = max(tp->mdev, TCP_RTO_MIN); } - tcp_set_rto(tp); - tcp_bound_rto(tp); - if (tp->rto < TCP_TIMEOUT_INIT && !tp->saw_tstamp) + tcp_set_rto(sk); + tcp_bound_rto(sk); + if (inet_csk(sk)->icsk_rto < TCP_TIMEOUT_INIT && !tp->rx_opt.saw_tstamp) goto reset; tp->snd_cwnd = tcp_init_cwnd(tp, dst); tp->snd_cwnd_stamp = tcp_time_stamp; @@ -876,15 +845,17 @@ reset: * supported, TCP will fail to recalculate correct * rtt, if initial rto is too small. FORGET ALL AND RESET! */ - if (!tp->saw_tstamp && tp->srtt) { + if (!tp->rx_opt.saw_tstamp && tp->srtt) { tp->srtt = 0; tp->mdev = tp->mdev_max = tp->rttvar = TCP_TIMEOUT_INIT; - tp->rto = TCP_TIMEOUT_INIT; + inet_csk(sk)->icsk_rto = TCP_TIMEOUT_INIT; } } -static void tcp_update_reordering(struct tcp_opt *tp, int metric, int ts) +static void tcp_update_reordering(struct sock *sk, const int metric, + const int ts) { + struct tcp_sock *tp = tcp_sk(sk); if (metric > tp->reordering) { tp->reordering = min(TCP_MAX_REORDERING, metric); @@ -899,14 +870,14 @@ static void tcp_update_reordering(struct tcp_opt *tp, int metric, int ts) NET_INC_STATS_BH(LINUX_MIB_TCPSACKREORDER); #if FASTRETRANS_DEBUG > 1 printk(KERN_DEBUG "Disorder%d %d %u f%u s%u rr%d\n", - tp->sack_ok, tp->ca_state, + tp->rx_opt.sack_ok, inet_csk(sk)->icsk_ca_state, tp->reordering, - tcp_get_pcount(&tp->fackets_out), - tcp_get_pcount(&tp->sacked_out), + tp->fackets_out, + tp->sacked_out, tp->undo_marker ? tp->undo_retrans : 0); #endif /* Disable FACK yet. */ - tp->sack_ok &= ~2; + tp->rx_opt.sack_ok &= ~2; } } @@ -961,34 +932,41 @@ static void tcp_update_reordering(struct tcp_opt *tp, int metric, int ts) static int tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_una) { - struct tcp_opt *tp = tcp_sk(sk); + const struct inet_connection_sock *icsk = inet_csk(sk); + struct tcp_sock *tp = tcp_sk(sk); unsigned char *ptr = ack_skb->h.raw + TCP_SKB_CB(ack_skb)->sacked; - struct tcp_sack_block *sp = (struct tcp_sack_block *)(ptr+2); + struct tcp_sack_block_wire *sp = (struct tcp_sack_block_wire *)(ptr+2); int num_sacks = (ptr[1] - TCPOLEN_SACK_BASE)>>3; - int reord = tcp_get_pcount(&tp->packets_out); + int reord = tp->packets_out; int prior_fackets; u32 lost_retrans = 0; int flag = 0; + int dup_sack = 0; int i; - /* So, SACKs for already sent large segments will be lost. - * Not good, but alternative is to resegment the queue. */ - if (sk->sk_route_caps & NETIF_F_TSO) { - sk->sk_route_caps &= ~NETIF_F_TSO; - sk->sk_no_largesend = 1; - tp->mss_cache = tp->mss_cache_std; - } - - if (!tcp_get_pcount(&tp->sacked_out)) - tcp_set_pcount(&tp->fackets_out, 0); - prior_fackets = tcp_get_pcount(&tp->fackets_out); - - for (i=0; istart_seq); - __u32 end_seq = ntohl(sp->end_seq); - int fack_count = 0; - int dup_sack = 0; + if (!tp->sacked_out) + tp->fackets_out = 0; + prior_fackets = tp->fackets_out; + + /* SACK fastpath: + * if the only SACK change is the increase of the end_seq of + * the first block then only apply that SACK block + * and use retrans queue hinting otherwise slowpath */ + flag = 1; + for (i = 0; i< num_sacks; i++) { + __u32 start_seq = ntohl(sp[i].start_seq); + __u32 end_seq = ntohl(sp[i].end_seq); + + if (i == 0){ + if (tp->recv_sack_cache[i].start_seq != start_seq) + flag = 0; + } else { + if ((tp->recv_sack_cache[i].start_seq != start_seq) || + (tp->recv_sack_cache[i].end_seq != end_seq)) + flag = 0; + } + tp->recv_sack_cache[i].start_seq = start_seq; + tp->recv_sack_cache[i].end_seq = end_seq; /* Check for D-SACK. */ if (i == 0) { @@ -996,13 +974,13 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ if (before(start_seq, ack)) { dup_sack = 1; - tp->sack_ok |= 4; + tp->rx_opt.sack_ok |= 4; NET_INC_STATS_BH(LINUX_MIB_TCPDSACKRECV); } else if (num_sacks > 1 && !after(end_seq, ntohl(sp[1].end_seq)) && !before(start_seq, ntohl(sp[1].start_seq))) { dup_sack = 1; - tp->sack_ok |= 4; + tp->rx_opt.sack_ok |= 4; NET_INC_STATS_BH(LINUX_MIB_TCPDSACKOFORECV); } @@ -1020,26 +998,92 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ if (before(ack, prior_snd_una - tp->max_window)) return 0; } + } + + if (flag) + num_sacks = 1; + else { + int j; + tp->fastpath_skb_hint = NULL; + + /* order SACK blocks to allow in order walk of the retrans queue */ + for (i = num_sacks-1; i > 0; i--) { + for (j = 0; j < i; j++){ + if (after(ntohl(sp[j].start_seq), + ntohl(sp[j+1].start_seq))){ + struct tcp_sack_block_wire tmp; + + tmp = sp[j]; + sp[j] = sp[j+1]; + sp[j+1] = tmp; + } + + } + } + } + + /* clear flag as used for different purpose in following code */ + flag = 0; + + for (i=0; istart_seq); + __u32 end_seq = ntohl(sp->end_seq); + int fack_count; + + /* Use SACK fastpath hint if valid */ + if (tp->fastpath_skb_hint) { + skb = tp->fastpath_skb_hint; + fack_count = tp->fastpath_cnt_hint; + } else { + skb = sk->sk_write_queue.next; + fack_count = 0; + } /* Event "B" in the comment above. */ if (after(end_seq, tp->high_seq)) flag |= FLAG_DATA_LOST; - sk_stream_for_retrans_queue(skb, sk) { - u8 sacked = TCP_SKB_CB(skb)->sacked; - int in_sack; + sk_stream_for_retrans_queue_from(skb, sk) { + int in_sack, pcount; + u8 sacked; + + tp->fastpath_skb_hint = skb; + tp->fastpath_cnt_hint = fack_count; /* The retransmission queue is always in order, so * we can short-circuit the walk early. */ - if(!before(TCP_SKB_CB(skb)->seq, end_seq)) + if (!before(TCP_SKB_CB(skb)->seq, end_seq)) break; - fack_count += tcp_skb_pcount(skb); - in_sack = !after(start_seq, TCP_SKB_CB(skb)->seq) && !before(end_seq, TCP_SKB_CB(skb)->end_seq); + pcount = tcp_skb_pcount(skb); + + if (pcount > 1 && !in_sack && + after(TCP_SKB_CB(skb)->end_seq, start_seq)) { + unsigned int pkt_len; + + in_sack = !after(start_seq, + TCP_SKB_CB(skb)->seq); + + if (!in_sack) + pkt_len = (start_seq - + TCP_SKB_CB(skb)->seq); + else + pkt_len = (end_seq - + TCP_SKB_CB(skb)->seq); + if (tcp_fragment(sk, skb, pkt_len, skb_shinfo(skb)->gso_size)) + break; + pcount = tcp_skb_pcount(skb); + } + + fack_count += pcount; + + sacked = TCP_SKB_CB(skb)->sacked; + /* Account D-SACK for retransmitted packet. */ if ((dup_sack && in_sack) && (sacked & TCPCB_RETRANS) && @@ -1079,8 +1123,11 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ */ if (sacked & TCPCB_LOST) { TCP_SKB_CB(skb)->sacked &= ~(TCPCB_LOST|TCPCB_SACKED_RETRANS); - tcp_dec_pcount(&tp->lost_out, skb); - tcp_dec_pcount(&tp->retrans_out, skb); + tp->lost_out -= tcp_skb_pcount(skb); + tp->retrans_out -= tcp_skb_pcount(skb); + + /* clear lost hint */ + tp->retransmit_skb_hint = NULL; } } else { /* New sack for not retransmitted frame, @@ -1092,16 +1139,19 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ if (sacked & TCPCB_LOST) { TCP_SKB_CB(skb)->sacked &= ~TCPCB_LOST; - tcp_dec_pcount(&tp->lost_out, skb); + tp->lost_out -= tcp_skb_pcount(skb); + + /* clear lost hint */ + tp->retransmit_skb_hint = NULL; } } TCP_SKB_CB(skb)->sacked |= TCPCB_SACKED_ACKED; flag |= FLAG_DATA_SACKED; - tcp_inc_pcount(&tp->sacked_out, skb); + tp->sacked_out += tcp_skb_pcount(skb); - if (fack_count > tcp_get_pcount(&tp->fackets_out)) - tcp_set_pcount(&tp->fackets_out, fack_count); + if (fack_count > tp->fackets_out) + tp->fackets_out = fack_count; } else { if (dup_sack && (sacked&TCPCB_RETRANS)) reord = min(fack_count, reord); @@ -1115,7 +1165,8 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ if (dup_sack && (TCP_SKB_CB(skb)->sacked&TCPCB_SACKED_RETRANS)) { TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS; - tcp_dec_pcount(&tp->retrans_out, skb); + tp->retrans_out -= tcp_skb_pcount(skb); + tp->retransmit_skb_hint = NULL; } } } @@ -1126,7 +1177,7 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ * we have to account for reordering! Ugly, * but should help. */ - if (lost_retrans && tp->ca_state == TCP_CA_Recovery) { + if (lost_retrans && icsk->icsk_ca_state == TCP_CA_Recovery) { struct sk_buff *skb; sk_stream_for_retrans_queue(skb, sk) { @@ -1139,12 +1190,15 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ (IsFack(tp) || !before(lost_retrans, TCP_SKB_CB(skb)->ack_seq + tp->reordering * - tp->mss_cache_std))) { + tp->mss_cache))) { TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS; - tcp_dec_pcount(&tp->retrans_out, skb); + tp->retrans_out -= tcp_skb_pcount(skb); + + /* clear lost hint */ + tp->retransmit_skb_hint = NULL; if (!(TCP_SKB_CB(skb)->sacked&(TCPCB_LOST|TCPCB_SACKED_ACKED))) { - tcp_inc_pcount(&tp->lost_out, skb); + tp->lost_out += tcp_skb_pcount(skb); TCP_SKB_CB(skb)->sacked |= TCPCB_LOST; flag |= FLAG_DATA_SACKED; NET_INC_STATS_BH(LINUX_MIB_TCPLOSTRETRANSMIT); @@ -1153,20 +1207,15 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ } } - tcp_set_pcount(&tp->left_out, - (tcp_get_pcount(&tp->sacked_out) + - tcp_get_pcount(&tp->lost_out))); + tp->left_out = tp->sacked_out + tp->lost_out; - if ((reord < tcp_get_pcount(&tp->fackets_out)) && - tp->ca_state != TCP_CA_Loss) - tcp_update_reordering(tp, - ((tcp_get_pcount(&tp->fackets_out) + 1) - - reord), 0); + if ((reord < tp->fackets_out) && icsk->icsk_ca_state != TCP_CA_Loss) + tcp_update_reordering(sk, ((tp->fackets_out + 1) - reord), 0); #if FASTRETRANS_DEBUG > 0 - BUG_TRAP((int)tcp_get_pcount(&tp->sacked_out) >= 0); - BUG_TRAP((int)tcp_get_pcount(&tp->lost_out) >= 0); - BUG_TRAP((int)tcp_get_pcount(&tp->retrans_out) >= 0); + BUG_TRAP((int)tp->sacked_out >= 0); + BUG_TRAP((int)tp->lost_out >= 0); + BUG_TRAP((int)tp->retrans_out >= 0); BUG_TRAP((int)tcp_packets_in_flight(tp) >= 0); #endif return flag; @@ -1178,17 +1227,18 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ */ void tcp_enter_frto(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + const struct inet_connection_sock *icsk = inet_csk(sk); + struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *skb; tp->frto_counter = 1; - if (tp->ca_state <= TCP_CA_Disorder || + if (icsk->icsk_ca_state <= TCP_CA_Disorder || tp->snd_una == tp->high_seq || - (tp->ca_state == TCP_CA_Loss && !tp->retransmits)) { - tp->prior_ssthresh = tcp_current_ssthresh(tp); - if (!tcp_westwood_ssthresh(tp)) - tp->snd_ssthresh = tcp_recalc_ssthresh(tp); + (icsk->icsk_ca_state == TCP_CA_Loss && !icsk->icsk_retransmits)) { + tp->prior_ssthresh = tcp_current_ssthresh(sk); + tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk); + tcp_ca_event(sk, CA_EVENT_FRTO); } /* Have to clear retransmission markers here to keep the bookkeeping @@ -1196,7 +1246,7 @@ void tcp_enter_frto(struct sock *sk) * If something was really lost, it is eventually caught up * in tcp_enter_frto_loss. */ - tcp_set_pcount(&tp->retrans_out, 0); + tp->retrans_out = 0; tp->undo_marker = tp->snd_una; tp->undo_retrans = 0; @@ -1205,7 +1255,7 @@ void tcp_enter_frto(struct sock *sk) } tcp_sync_left_out(tp); - tcp_set_ca_state(tp, TCP_CA_Open); + tcp_set_ca_state(sk, TCP_CA_Open); tp->frto_highmark = tp->snd_nxt; } @@ -1215,13 +1265,13 @@ void tcp_enter_frto(struct sock *sk) */ static void tcp_enter_frto_loss(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *skb; int cnt = 0; - tcp_set_pcount(&tp->sacked_out, 0); - tcp_set_pcount(&tp->lost_out, 0); - tcp_set_pcount(&tp->fackets_out, 0); + tp->sacked_out = 0; + tp->lost_out = 0; + tp->fackets_out = 0; sk_stream_for_retrans_queue(skb, sk) { cnt += tcp_skb_pcount(skb); @@ -1234,11 +1284,11 @@ static void tcp_enter_frto_loss(struct sock *sk) if (!after(TCP_SKB_CB(skb)->end_seq, tp->frto_highmark)) { TCP_SKB_CB(skb)->sacked |= TCPCB_LOST; - tcp_inc_pcount(&tp->lost_out, skb); + tp->lost_out += tcp_skb_pcount(skb); } } else { - tcp_inc_pcount(&tp->sacked_out, skb); - tcp_set_pcount(&tp->fackets_out, cnt); + tp->sacked_out += tcp_skb_pcount(skb); + tp->fackets_out = cnt; } } tcp_sync_left_out(tp); @@ -1251,21 +1301,21 @@ static void tcp_enter_frto_loss(struct sock *sk) tp->reordering = min_t(unsigned int, tp->reordering, sysctl_tcp_reordering); - tcp_set_ca_state(tp, TCP_CA_Loss); + tcp_set_ca_state(sk, TCP_CA_Loss); tp->high_seq = tp->frto_highmark; TCP_ECN_queue_cwr(tp); - init_bictcp(tp); + clear_all_retrans_hints(tp); } -void tcp_clear_retrans(struct tcp_opt *tp) +void tcp_clear_retrans(struct tcp_sock *tp) { - tcp_set_pcount(&tp->left_out, 0); - tcp_set_pcount(&tp->retrans_out, 0); + tp->left_out = 0; + tp->retrans_out = 0; - tcp_set_pcount(&tp->fackets_out, 0); - tcp_set_pcount(&tp->sacked_out, 0); - tcp_set_pcount(&tp->lost_out, 0); + tp->fackets_out = 0; + tp->sacked_out = 0; + tp->lost_out = 0; tp->undo_marker = 0; tp->undo_retrans = 0; @@ -1277,20 +1327,23 @@ void tcp_clear_retrans(struct tcp_opt *tp) */ void tcp_enter_loss(struct sock *sk, int how) { - struct tcp_opt *tp = tcp_sk(sk); + const struct inet_connection_sock *icsk = inet_csk(sk); + struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *skb; int cnt = 0; /* Reduce ssthresh if it has not yet been made inside this window. */ - if (tp->ca_state <= TCP_CA_Disorder || tp->snd_una == tp->high_seq || - (tp->ca_state == TCP_CA_Loss && !tp->retransmits)) { - tp->prior_ssthresh = tcp_current_ssthresh(tp); - tp->snd_ssthresh = tcp_recalc_ssthresh(tp); + if (icsk->icsk_ca_state <= TCP_CA_Disorder || tp->snd_una == tp->high_seq || + (icsk->icsk_ca_state == TCP_CA_Loss && !icsk->icsk_retransmits)) { + tp->prior_ssthresh = tcp_current_ssthresh(sk); + tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk); + tcp_ca_event(sk, CA_EVENT_LOSS); } tp->snd_cwnd = 1; tp->snd_cwnd_cnt = 0; tp->snd_cwnd_stamp = tcp_time_stamp; + tp->bytes_acked = 0; tcp_clear_retrans(tp); /* Push undo marker, if it was plain RTO and nothing @@ -1306,22 +1359,24 @@ void tcp_enter_loss(struct sock *sk, int how) if (!(TCP_SKB_CB(skb)->sacked&TCPCB_SACKED_ACKED) || how) { TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_ACKED; TCP_SKB_CB(skb)->sacked |= TCPCB_LOST; - tcp_inc_pcount(&tp->lost_out, skb); + tp->lost_out += tcp_skb_pcount(skb); } else { - tcp_inc_pcount(&tp->sacked_out, skb); - tcp_set_pcount(&tp->fackets_out, cnt); + tp->sacked_out += tcp_skb_pcount(skb); + tp->fackets_out = cnt; } } tcp_sync_left_out(tp); tp->reordering = min_t(unsigned int, tp->reordering, sysctl_tcp_reordering); - tcp_set_ca_state(tp, TCP_CA_Loss); + tcp_set_ca_state(sk, TCP_CA_Loss); tp->high_seq = tp->snd_nxt; TCP_ECN_queue_cwr(tp); + + clear_all_retrans_hints(tp); } -static int tcp_check_sack_reneging(struct sock *sk, struct tcp_opt *tp) +static int tcp_check_sack_reneging(struct sock *sk) { struct sk_buff *skb; @@ -1333,32 +1388,33 @@ static int tcp_check_sack_reneging(struct sock *sk, struct tcp_opt *tp) */ if ((skb = skb_peek(&sk->sk_write_queue)) != NULL && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)) { + struct inet_connection_sock *icsk = inet_csk(sk); NET_INC_STATS_BH(LINUX_MIB_TCPSACKRENEGING); tcp_enter_loss(sk, 1); - tp->retransmits++; + icsk->icsk_retransmits++; tcp_retransmit_skb(sk, skb_peek(&sk->sk_write_queue)); - tcp_reset_xmit_timer(sk, TCP_TIME_RETRANS, tp->rto); + inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, + icsk->icsk_rto, TCP_RTO_MAX); return 1; } return 0; } -static inline int tcp_fackets_out(struct tcp_opt *tp) +static inline int tcp_fackets_out(struct tcp_sock *tp) { - return IsReno(tp) ? tcp_get_pcount(&tp->sacked_out)+1 : - tcp_get_pcount(&tp->fackets_out); + return IsReno(tp) ? tp->sacked_out+1 : tp->fackets_out; } -static inline int tcp_skb_timedout(struct tcp_opt *tp, struct sk_buff *skb) +static inline int tcp_skb_timedout(struct sock *sk, struct sk_buff *skb) { - return (tcp_time_stamp - TCP_SKB_CB(skb)->when > tp->rto); + return (tcp_time_stamp - TCP_SKB_CB(skb)->when > inet_csk(sk)->icsk_rto); } -static inline int tcp_head_timedout(struct sock *sk, struct tcp_opt *tp) +static inline int tcp_head_timedout(struct sock *sk, struct tcp_sock *tp) { - return tcp_get_pcount(&tp->packets_out) && - tcp_skb_timedout(tp, skb_peek(&sk->sk_write_queue)); + return tp->packets_out && + tcp_skb_timedout(sk, skb_peek(&sk->sk_write_queue)); } /* Linux NewReno/SACK/FACK/ECN state machine. @@ -1454,13 +1510,12 @@ static inline int tcp_head_timedout(struct sock *sk, struct tcp_opt *tp) * Main question: may we further continue forward transmission * with the same cwnd? */ -static int -tcp_time_to_recover(struct sock *sk, struct tcp_opt *tp) +static int tcp_time_to_recover(struct sock *sk, struct tcp_sock *tp) { __u32 packets_out; /* Trick#1: The loss is proven. */ - if (tcp_get_pcount(&tp->lost_out)) + if (tp->lost_out) return 1; /* Not-A-Trick#2 : Classic rule... */ @@ -1476,9 +1531,9 @@ tcp_time_to_recover(struct sock *sk, struct tcp_opt *tp) /* Trick#4: It is still not OK... But will it be useful to delay * recovery more? */ - packets_out = tcp_get_pcount(&tp->packets_out); + packets_out = tp->packets_out; if (packets_out <= tp->reordering && - tcp_get_pcount(&tp->sacked_out) >= max_t(__u32, packets_out/2, sysctl_tcp_reordering) && + tp->sacked_out >= max_t(__u32, packets_out/2, sysctl_tcp_reordering) && !tcp_may_send_now(sk, tp)) { /* We have nothing to send. This connection is limited * either by receiver window or by application. @@ -1493,69 +1548,87 @@ tcp_time_to_recover(struct sock *sk, struct tcp_opt *tp) * in assumption of absent reordering, interpret this as reordering. * The only another reason could be bug in receiver TCP. */ -static void tcp_check_reno_reordering(struct tcp_opt *tp, int addend) +static void tcp_check_reno_reordering(struct sock *sk, const int addend) { + struct tcp_sock *tp = tcp_sk(sk); u32 holes; - holes = max(tcp_get_pcount(&tp->lost_out), 1U); - holes = min(holes, tcp_get_pcount(&tp->packets_out)); + holes = max(tp->lost_out, 1U); + holes = min(holes, tp->packets_out); - if ((tcp_get_pcount(&tp->sacked_out) + holes) > - tcp_get_pcount(&tp->packets_out)) { - tcp_set_pcount(&tp->sacked_out, - (tcp_get_pcount(&tp->packets_out) - holes)); - tcp_update_reordering(tp, - tcp_get_pcount(&tp->packets_out)+addend, - 0); + if ((tp->sacked_out + holes) > tp->packets_out) { + tp->sacked_out = tp->packets_out - holes; + tcp_update_reordering(sk, tp->packets_out + addend, 0); } } /* Emulate SACKs for SACKless connection: account for a new dupack. */ -static void tcp_add_reno_sack(struct tcp_opt *tp) +static void tcp_add_reno_sack(struct sock *sk) { - tcp_inc_pcount_explicit(&tp->sacked_out, 1); - tcp_check_reno_reordering(tp, 0); + struct tcp_sock *tp = tcp_sk(sk); + tp->sacked_out++; + tcp_check_reno_reordering(sk, 0); tcp_sync_left_out(tp); } /* Account for ACK, ACKing some data in Reno Recovery phase. */ -static void tcp_remove_reno_sacks(struct sock *sk, struct tcp_opt *tp, int acked) +static void tcp_remove_reno_sacks(struct sock *sk, struct tcp_sock *tp, int acked) { if (acked > 0) { /* One ACK acked hole. The rest eat duplicate ACKs. */ - if (acked-1 >= tcp_get_pcount(&tp->sacked_out)) - tcp_set_pcount(&tp->sacked_out, 0); + if (acked-1 >= tp->sacked_out) + tp->sacked_out = 0; else - tcp_dec_pcount_explicit(&tp->sacked_out, acked-1); + tp->sacked_out -= acked-1; } - tcp_check_reno_reordering(tp, acked); + tcp_check_reno_reordering(sk, acked); tcp_sync_left_out(tp); } -static inline void tcp_reset_reno_sack(struct tcp_opt *tp) +static inline void tcp_reset_reno_sack(struct tcp_sock *tp) { - tcp_set_pcount(&tp->sacked_out, 0); - tcp_set_pcount(&tp->left_out, tcp_get_pcount(&tp->lost_out)); + tp->sacked_out = 0; + tp->left_out = tp->lost_out; } /* Mark head of queue up as lost. */ -static void -tcp_mark_head_lost(struct sock *sk, struct tcp_opt *tp, int packets, u32 high_seq) +static void tcp_mark_head_lost(struct sock *sk, struct tcp_sock *tp, + int packets, u32 high_seq) { struct sk_buff *skb; - int cnt = packets; + int cnt; - BUG_TRAP(cnt <= tcp_get_pcount(&tp->packets_out)); + BUG_TRAP(packets <= tp->packets_out); + if (tp->lost_skb_hint) { + skb = tp->lost_skb_hint; + cnt = tp->lost_cnt_hint; + } else { + skb = sk->sk_write_queue.next; + cnt = 0; + } - sk_stream_for_retrans_queue(skb, sk) { - cnt -= tcp_skb_pcount(skb); - if (cnt < 0 || after(TCP_SKB_CB(skb)->end_seq, high_seq)) + sk_stream_for_retrans_queue_from(skb, sk) { + /* TODO: do this better */ + /* this is not the most efficient way to do this... */ + tp->lost_skb_hint = skb; + tp->lost_cnt_hint = cnt; + cnt += tcp_skb_pcount(skb); + if (cnt > packets || after(TCP_SKB_CB(skb)->end_seq, high_seq)) break; if (!(TCP_SKB_CB(skb)->sacked&TCPCB_TAGBITS)) { TCP_SKB_CB(skb)->sacked |= TCPCB_LOST; - tcp_inc_pcount(&tp->lost_out, skb); + tp->lost_out += tcp_skb_pcount(skb); + + /* clear xmit_retransmit_queue hints + * if this is beyond hint */ + if(tp->retransmit_skb_hint != NULL && + before(TCP_SKB_CB(skb)->seq, + TCP_SKB_CB(tp->retransmit_skb_hint)->seq)) { + + tp->retransmit_skb_hint = NULL; + } } } tcp_sync_left_out(tp); @@ -1563,10 +1636,10 @@ tcp_mark_head_lost(struct sock *sk, struct tcp_opt *tp, int packets, u32 high_se /* Account newly detected lost packet(s) */ -static void tcp_update_scoreboard(struct sock *sk, struct tcp_opt *tp) +static void tcp_update_scoreboard(struct sock *sk, struct tcp_sock *tp) { if (IsFack(tp)) { - int lost = tcp_get_pcount(&tp->fackets_out) - tp->reordering; + int lost = tp->fackets_out - tp->reordering; if (lost <= 0) lost = 1; tcp_mark_head_lost(sk, tp, lost, tp->high_seq); @@ -1579,16 +1652,31 @@ static void tcp_update_scoreboard(struct sock *sk, struct tcp_opt *tp) * Hence, we can detect timed out packets during fast * retransmit without falling to slow start. */ - if (tcp_head_timedout(sk, tp)) { + if (!IsReno(tp) && tcp_head_timedout(sk, tp)) { struct sk_buff *skb; - sk_stream_for_retrans_queue(skb, sk) { - if (tcp_skb_timedout(tp, skb) && - !(TCP_SKB_CB(skb)->sacked&TCPCB_TAGBITS)) { + skb = tp->scoreboard_skb_hint ? tp->scoreboard_skb_hint + : sk->sk_write_queue.next; + + sk_stream_for_retrans_queue_from(skb, sk) { + if (!tcp_skb_timedout(sk, skb)) + break; + + if (!(TCP_SKB_CB(skb)->sacked&TCPCB_TAGBITS)) { TCP_SKB_CB(skb)->sacked |= TCPCB_LOST; - tcp_inc_pcount(&tp->lost_out, skb); + tp->lost_out += tcp_skb_pcount(skb); + + /* clear xmit_retrans hint */ + if (tp->retransmit_skb_hint && + before(TCP_SKB_CB(skb)->seq, + TCP_SKB_CB(tp->retransmit_skb_hint)->seq)) + + tp->retransmit_skb_hint = NULL; } } + + tp->scoreboard_skb_hint = skb; + tcp_sync_left_out(tp); } } @@ -1596,36 +1684,33 @@ static void tcp_update_scoreboard(struct sock *sk, struct tcp_opt *tp) /* CWND moderation, preventing bursts due to too big ACKs * in dubious situations. */ -static __inline__ void tcp_moderate_cwnd(struct tcp_opt *tp) +static inline void tcp_moderate_cwnd(struct tcp_sock *tp) { tp->snd_cwnd = min(tp->snd_cwnd, tcp_packets_in_flight(tp)+tcp_max_burst(tp)); tp->snd_cwnd_stamp = tcp_time_stamp; } -/* Decrease cwnd each second ack. */ - -static void tcp_cwnd_down(struct tcp_opt *tp) +/* Lower bound on congestion window is slow start threshold + * unless congestion avoidance choice decides to overide it. + */ +static inline u32 tcp_cwnd_min(const struct sock *sk) { - int decr = tp->snd_cwnd_cnt + 1; - __u32 limit; + const struct tcp_congestion_ops *ca_ops = inet_csk(sk)->icsk_ca_ops; - /* - * TCP Westwood - * Here limit is evaluated as BWestimation*RTTmin (for obtaining it - * in packets we use mss_cache). If sysctl_tcp_westwood is off - * tcp_westwood_bw_rttmin() returns 0. In such case snd_ssthresh is - * still used as usual. It prevents other strange cases in which - * BWE*RTTmin could assume value 0. It should not happen but... - */ + return ca_ops->min_cwnd ? ca_ops->min_cwnd(sk) : tcp_sk(sk)->snd_ssthresh; +} - if (!(limit = tcp_westwood_bw_rttmin(tp))) - limit = tp->snd_ssthresh/2; +/* Decrease cwnd each second ack. */ +static void tcp_cwnd_down(struct sock *sk) +{ + struct tcp_sock *tp = tcp_sk(sk); + int decr = tp->snd_cwnd_cnt + 1; tp->snd_cwnd_cnt = decr&1; decr >>= 1; - if (decr && tp->snd_cwnd > limit) + if (decr && tp->snd_cwnd > tcp_cwnd_min(sk)) tp->snd_cwnd -= decr; tp->snd_cwnd = min(tp->snd_cwnd, tcp_packets_in_flight(tp)+1); @@ -1635,34 +1720,41 @@ static void tcp_cwnd_down(struct tcp_opt *tp) /* Nothing was retransmitted or returned timestamp is less * than timestamp of the first retransmission. */ -static __inline__ int tcp_packet_delayed(struct tcp_opt *tp) +static inline int tcp_packet_delayed(struct tcp_sock *tp) { return !tp->retrans_stamp || - (tp->saw_tstamp && tp->rcv_tsecr && - (__s32)(tp->rcv_tsecr - tp->retrans_stamp) < 0); + (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr && + (__s32)(tp->rx_opt.rcv_tsecr - tp->retrans_stamp) < 0); } /* Undo procedures. */ #if FASTRETRANS_DEBUG > 1 -static void DBGUNDO(struct sock *sk, struct tcp_opt *tp, const char *msg) +static void DBGUNDO(struct sock *sk, struct tcp_sock *tp, const char *msg) { - struct inet_opt *inet = inet_sk(sk); + struct inet_sock *inet = inet_sk(sk); printk(KERN_DEBUG "Undo %s %u.%u.%u.%u/%u c%u l%u ss%u/%u p%u\n", msg, NIPQUAD(inet->daddr), ntohs(inet->dport), - tp->snd_cwnd, tcp_get_pcount(&tp->left_out), + tp->snd_cwnd, tp->left_out, tp->snd_ssthresh, tp->prior_ssthresh, - tcp_get_pcount(&tp->packets_out)); + tp->packets_out); } #else #define DBGUNDO(x...) do { } while (0) #endif -static void tcp_undo_cwr(struct tcp_opt *tp, int undo) +static void tcp_undo_cwr(struct sock *sk, const int undo) { + struct tcp_sock *tp = tcp_sk(sk); + if (tp->prior_ssthresh) { - tp->snd_cwnd = max(tp->snd_cwnd, tp->snd_ssthresh<<1); + const struct inet_connection_sock *icsk = inet_csk(sk); + + if (icsk->icsk_ca_ops->undo_cwnd) + tp->snd_cwnd = icsk->icsk_ca_ops->undo_cwnd(sk); + else + tp->snd_cwnd = max(tp->snd_cwnd, tp->snd_ssthresh<<1); if (undo && tp->prior_ssthresh > tp->snd_ssthresh) { tp->snd_ssthresh = tp->prior_ssthresh; @@ -1673,24 +1765,28 @@ static void tcp_undo_cwr(struct tcp_opt *tp, int undo) } tcp_moderate_cwnd(tp); tp->snd_cwnd_stamp = tcp_time_stamp; + + /* There is something screwy going on with the retrans hints after + an undo */ + clear_all_retrans_hints(tp); } -static inline int tcp_may_undo(struct tcp_opt *tp) +static inline int tcp_may_undo(struct tcp_sock *tp) { return tp->undo_marker && (!tp->undo_retrans || tcp_packet_delayed(tp)); } /* People celebrate: "We love our President!" */ -static int tcp_try_undo_recovery(struct sock *sk, struct tcp_opt *tp) +static int tcp_try_undo_recovery(struct sock *sk, struct tcp_sock *tp) { if (tcp_may_undo(tp)) { /* Happy end! We did not retransmit anything * or our original transmission succeeded. */ - DBGUNDO(sk, tp, tp->ca_state == TCP_CA_Loss ? "loss" : "retrans"); - tcp_undo_cwr(tp, 1); - if (tp->ca_state == TCP_CA_Loss) + DBGUNDO(sk, tp, inet_csk(sk)->icsk_ca_state == TCP_CA_Loss ? "loss" : "retrans"); + tcp_undo_cwr(sk, 1); + if (inet_csk(sk)->icsk_ca_state == TCP_CA_Loss) NET_INC_STATS_BH(LINUX_MIB_TCPLOSSUNDO); else NET_INC_STATS_BH(LINUX_MIB_TCPFULLUNDO); @@ -1703,16 +1799,16 @@ static int tcp_try_undo_recovery(struct sock *sk, struct tcp_opt *tp) tcp_moderate_cwnd(tp); return 1; } - tcp_set_ca_state(tp, TCP_CA_Open); + tcp_set_ca_state(sk, TCP_CA_Open); return 0; } /* Try to undo cwnd reduction, because D-SACKs acked all retransmitted data */ -static void tcp_try_undo_dsack(struct sock *sk, struct tcp_opt *tp) +static void tcp_try_undo_dsack(struct sock *sk, struct tcp_sock *tp) { if (tp->undo_marker && !tp->undo_retrans) { DBGUNDO(sk, tp, "D-SACK"); - tcp_undo_cwr(tp, 1); + tcp_undo_cwr(sk, 1); tp->undo_marker = 0; NET_INC_STATS_BH(LINUX_MIB_TCPDSACKUNDO); } @@ -1720,22 +1816,23 @@ static void tcp_try_undo_dsack(struct sock *sk, struct tcp_opt *tp) /* Undo during fast recovery after partial ACK. */ -static int tcp_try_undo_partial(struct sock *sk, struct tcp_opt *tp, int acked) +static int tcp_try_undo_partial(struct sock *sk, struct tcp_sock *tp, + int acked) { /* Partial ACK arrived. Force Hoe's retransmit. */ - int failed = IsReno(tp) || tcp_get_pcount(&tp->fackets_out)>tp->reordering; + int failed = IsReno(tp) || tp->fackets_out>tp->reordering; if (tcp_may_undo(tp)) { /* Plain luck! Hole if filled with delayed * packet, rather than with a retransmit. */ - if (tcp_get_pcount(&tp->retrans_out) == 0) + if (tp->retrans_out == 0) tp->retrans_stamp = 0; - tcp_update_reordering(tp, tcp_fackets_out(tp)+acked, 1); + tcp_update_reordering(sk, tcp_fackets_out(tp) + acked, 1); DBGUNDO(sk, tp, "Hoe"); - tcp_undo_cwr(tp, 0); + tcp_undo_cwr(sk, 0); NET_INC_STATS_BH(LINUX_MIB_TCPPARTIALUNDO); /* So... Do not make Hoe's retransmit yet. @@ -1748,64 +1845,92 @@ static int tcp_try_undo_partial(struct sock *sk, struct tcp_opt *tp, int acked) } /* Undo during loss recovery after partial ACK. */ -static int tcp_try_undo_loss(struct sock *sk, struct tcp_opt *tp) +static int tcp_try_undo_loss(struct sock *sk, struct tcp_sock *tp) { if (tcp_may_undo(tp)) { struct sk_buff *skb; sk_stream_for_retrans_queue(skb, sk) { TCP_SKB_CB(skb)->sacked &= ~TCPCB_LOST; } + + clear_all_retrans_hints(tp); + DBGUNDO(sk, tp, "partial loss"); - tcp_set_pcount(&tp->lost_out, 0); - tcp_set_pcount(&tp->left_out, tcp_get_pcount(&tp->sacked_out)); - tcp_undo_cwr(tp, 1); + tp->lost_out = 0; + tp->left_out = tp->sacked_out; + tcp_undo_cwr(sk, 1); NET_INC_STATS_BH(LINUX_MIB_TCPLOSSUNDO); - tp->retransmits = 0; + inet_csk(sk)->icsk_retransmits = 0; tp->undo_marker = 0; if (!IsReno(tp)) - tcp_set_ca_state(tp, TCP_CA_Open); + tcp_set_ca_state(sk, TCP_CA_Open); return 1; } return 0; } -static __inline__ void tcp_complete_cwr(struct tcp_opt *tp) +static inline void tcp_complete_cwr(struct sock *sk) { - if (tcp_westwood_cwnd(tp)) - tp->snd_ssthresh = tp->snd_cwnd; - else - tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_ssthresh); + struct tcp_sock *tp = tcp_sk(sk); + tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_ssthresh); tp->snd_cwnd_stamp = tcp_time_stamp; + tcp_ca_event(sk, CA_EVENT_COMPLETE_CWR); } -static void tcp_try_to_open(struct sock *sk, struct tcp_opt *tp, int flag) +static void tcp_try_to_open(struct sock *sk, struct tcp_sock *tp, int flag) { - tcp_set_pcount(&tp->left_out, tcp_get_pcount(&tp->sacked_out)); + tp->left_out = tp->sacked_out; - if (tcp_get_pcount(&tp->retrans_out) == 0) + if (tp->retrans_out == 0) tp->retrans_stamp = 0; if (flag&FLAG_ECE) - tcp_enter_cwr(tp); + tcp_enter_cwr(sk); - if (tp->ca_state != TCP_CA_CWR) { + if (inet_csk(sk)->icsk_ca_state != TCP_CA_CWR) { int state = TCP_CA_Open; - if (tcp_get_pcount(&tp->left_out) || - tcp_get_pcount(&tp->retrans_out) || - tp->undo_marker) + if (tp->left_out || tp->retrans_out || tp->undo_marker) state = TCP_CA_Disorder; - if (tp->ca_state != state) { - tcp_set_ca_state(tp, state); + if (inet_csk(sk)->icsk_ca_state != state) { + tcp_set_ca_state(sk, state); tp->high_seq = tp->snd_nxt; } tcp_moderate_cwnd(tp); } else { - tcp_cwnd_down(tp); + tcp_cwnd_down(sk); } } +static void tcp_mtup_probe_failed(struct sock *sk) +{ + struct inet_connection_sock *icsk = inet_csk(sk); + + icsk->icsk_mtup.search_high = icsk->icsk_mtup.probe_size - 1; + icsk->icsk_mtup.probe_size = 0; +} + +static void tcp_mtup_probe_success(struct sock *sk, struct sk_buff *skb) +{ + struct tcp_sock *tp = tcp_sk(sk); + struct inet_connection_sock *icsk = inet_csk(sk); + + /* FIXME: breaks with very large cwnd */ + tp->prior_ssthresh = tcp_current_ssthresh(sk); + tp->snd_cwnd = tp->snd_cwnd * + tcp_mss_to_mtu(sk, tp->mss_cache) / + icsk->icsk_mtup.probe_size; + tp->snd_cwnd_cnt = 0; + tp->snd_cwnd_stamp = tcp_time_stamp; + tp->rcv_ssthresh = tcp_current_ssthresh(sk); + + icsk->icsk_mtup.search_low = icsk->icsk_mtup.probe_size; + icsk->icsk_mtup.probe_size = 0; + tcp_sync_mss(sk, icsk->icsk_pmtu_cookie); +} + + /* Process an event, which can update packets-in-flight not trivially. * Main goal of this function is to calculate new estimate for left_out, * taking into account both packets sitting in receiver's buffer and @@ -1821,16 +1946,17 @@ static void tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, int prior_packets, int flag) { - struct tcp_opt *tp = tcp_sk(sk); + struct inet_connection_sock *icsk = inet_csk(sk); + struct tcp_sock *tp = tcp_sk(sk); int is_dupack = (tp->snd_una == prior_snd_una && !(flag&FLAG_NOT_DUP)); /* Some technical things: * 1. Reno does not count dupacks (sacked_out) automatically. */ - if (!tcp_get_pcount(&tp->packets_out)) - tcp_set_pcount(&tp->sacked_out, 0); + if (!tp->packets_out) + tp->sacked_out = 0; /* 2. SACK counts snd_fack in packets inaccurately. */ - if (tcp_get_pcount(&tp->sacked_out) == 0) - tcp_set_pcount(&tp->fackets_out, 0); + if (tp->sacked_out == 0) + tp->fackets_out = 0; /* Now state machine starts. * A. ECE, hence prohibit cwnd undoing, the reduction is required. */ @@ -1838,15 +1964,15 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, tp->prior_ssthresh = 0; /* B. In all the states check for reneging SACKs. */ - if (tcp_get_pcount(&tp->sacked_out) && tcp_check_sack_reneging(sk, tp)) + if (tp->sacked_out && tcp_check_sack_reneging(sk)) return; /* C. Process data loss notification, provided it is valid. */ if ((flag&FLAG_DATA_LOST) && before(tp->snd_una, tp->high_seq) && - tp->ca_state != TCP_CA_Open && - tcp_get_pcount(&tp->fackets_out) > tp->reordering) { - tcp_mark_head_lost(sk, tp, tcp_get_pcount(&tp->fackets_out)-tp->reordering, tp->high_seq); + icsk->icsk_ca_state != TCP_CA_Open && + tp->fackets_out > tp->reordering) { + tcp_mark_head_lost(sk, tp, tp->fackets_out-tp->reordering, tp->high_seq); NET_INC_STATS_BH(LINUX_MIB_TCPLOSS); } @@ -1855,14 +1981,14 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, /* E. Check state exit conditions. State can be terminated * when high_seq is ACKed. */ - if (tp->ca_state == TCP_CA_Open) { + if (icsk->icsk_ca_state == TCP_CA_Open) { if (!sysctl_tcp_frto) - BUG_TRAP(tcp_get_pcount(&tp->retrans_out) == 0); + BUG_TRAP(tp->retrans_out == 0); tp->retrans_stamp = 0; } else if (!before(tp->snd_una, tp->high_seq)) { - switch (tp->ca_state) { + switch (icsk->icsk_ca_state) { case TCP_CA_Loss: - tp->retransmits = 0; + icsk->icsk_retransmits = 0; if (tcp_try_undo_recovery(sk, tp)) return; break; @@ -1871,8 +1997,8 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, /* CWR is to be held something *above* high_seq * is ACKed for CWR bit to reach receiver. */ if (tp->snd_una != tp->high_seq) { - tcp_complete_cwr(tp); - tcp_set_ca_state(tp, TCP_CA_Open); + tcp_complete_cwr(sk); + tcp_set_ca_state(sk, TCP_CA_Open); } break; @@ -1883,7 +2009,7 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, * catching for all duplicate ACKs. */ IsReno(tp) || tp->snd_una != tp->high_seq) { tp->undo_marker = 0; - tcp_set_ca_state(tp, TCP_CA_Open); + tcp_set_ca_state(sk, TCP_CA_Open); } break; @@ -1892,20 +2018,19 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, tcp_reset_reno_sack(tp); if (tcp_try_undo_recovery(sk, tp)) return; - tcp_complete_cwr(tp); + tcp_complete_cwr(sk); break; } } /* F. Process state. */ - switch (tp->ca_state) { + switch (icsk->icsk_ca_state) { case TCP_CA_Recovery: if (prior_snd_una == tp->snd_una) { if (IsReno(tp) && is_dupack) - tcp_add_reno_sack(tp); + tcp_add_reno_sack(sk); } else { - int acked = prior_packets - - tcp_get_pcount(&tp->packets_out); + int acked = prior_packets - tp->packets_out; if (IsReno(tp)) tcp_remove_reno_sacks(sk, tp, acked); is_dupack = tcp_try_undo_partial(sk, tp, acked); @@ -1913,13 +2038,13 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, break; case TCP_CA_Loss: if (flag&FLAG_DATA_ACKED) - tp->retransmits = 0; + icsk->icsk_retransmits = 0; if (!tcp_try_undo_loss(sk, tp)) { tcp_moderate_cwnd(tp); tcp_xmit_retransmit_queue(sk); return; } - if (tp->ca_state != TCP_CA_Open) + if (icsk->icsk_ca_state != TCP_CA_Open) return; /* Loss is undone; fall through to processing in Open state. */ default: @@ -1927,10 +2052,10 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, if (tp->snd_una != prior_snd_una) tcp_reset_reno_sack(tp); if (is_dupack) - tcp_add_reno_sack(tp); + tcp_add_reno_sack(sk); } - if (tp->ca_state == TCP_CA_Disorder) + if (icsk->icsk_ca_state == TCP_CA_Disorder) tcp_try_undo_dsack(sk, tp); if (!tcp_time_to_recover(sk, tp)) { @@ -1938,6 +2063,17 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, return; } + /* MTU probe failure: don't reduce cwnd */ + if (icsk->icsk_ca_state < TCP_CA_CWR && + icsk->icsk_mtup.probe_size && + tp->snd_una == tp->mtu_probe.probe_seq_start) { + tcp_mtup_probe_failed(sk); + /* Restores the reduction we did in tcp_mtup_probe() */ + tp->snd_cwnd++; + tcp_simple_retransmit(sk); + return; + } + /* Otherwise enter Recovery state */ if (IsReno(tp)) @@ -1948,32 +2084,31 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, tp->high_seq = tp->snd_nxt; tp->prior_ssthresh = 0; tp->undo_marker = tp->snd_una; - tp->undo_retrans = tcp_get_pcount(&tp->retrans_out); + tp->undo_retrans = tp->retrans_out; - if (tp->ca_state < TCP_CA_CWR) { + if (icsk->icsk_ca_state < TCP_CA_CWR) { if (!(flag&FLAG_ECE)) - tp->prior_ssthresh = tcp_current_ssthresh(tp); - tp->snd_ssthresh = tcp_recalc_ssthresh(tp); + tp->prior_ssthresh = tcp_current_ssthresh(sk); + tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk); TCP_ECN_queue_cwr(tp); } + tp->bytes_acked = 0; tp->snd_cwnd_cnt = 0; - tcp_set_ca_state(tp, TCP_CA_Recovery); + tcp_set_ca_state(sk, TCP_CA_Recovery); } if (is_dupack || tcp_head_timedout(sk, tp)) tcp_update_scoreboard(sk, tp); - tcp_cwnd_down(tp); + tcp_cwnd_down(sk); tcp_xmit_retransmit_queue(sk); } /* Read draft-ietf-tcplw-high-performance before mucking - * with this code. (Superceeds RFC1323) + * with this code. (Supersedes RFC1323) */ -static void tcp_ack_saw_tstamp(struct tcp_opt *tp, int flag) +static void tcp_ack_saw_tstamp(struct sock *sk, int flag) { - __u32 seq_rtt; - /* RTTM Rule: A TSecr value received in a segment is used to * update the averaged RTT measurement only if the segment * acknowledges some new data, i.e., only if it advances the @@ -1983,20 +2118,21 @@ static void tcp_ack_saw_tstamp(struct tcp_opt *tp, int flag) * 1998/04/10 Andrey V. Savochkin * * Changed: reset backoff as soon as we see the first valid sample. - * If we do not, we get strongly overstimated rto. With timestamps + * If we do not, we get strongly overestimated rto. With timestamps * samples are accepted even from very old segments: f.e., when rtt=1 * increases to 8, we retransmit 5 times and after 8 seconds delayed * answer arrives rto becomes 120 seconds! If at least one of segments * in window is lost... Voila. --ANK (010210) */ - seq_rtt = tcp_time_stamp - tp->rcv_tsecr; - tcp_rtt_estimator(tp, seq_rtt); - tcp_set_rto(tp); - tp->backoff = 0; - tcp_bound_rto(tp); + struct tcp_sock *tp = tcp_sk(sk); + const __u32 seq_rtt = tcp_time_stamp - tp->rx_opt.rcv_tsecr; + tcp_rtt_estimator(sk, seq_rtt); + tcp_set_rto(sk); + inet_csk(sk)->icsk_backoff = 0; + tcp_bound_rto(sk); } -static void tcp_ack_no_tstamp(struct tcp_opt *tp, u32 seq_rtt, int flag) +static void tcp_ack_no_tstamp(struct sock *sk, u32 seq_rtt, int flag) { /* We don't have a timestamp. Can only use * packets that are not retransmitted to determine @@ -2010,384 +2146,62 @@ static void tcp_ack_no_tstamp(struct tcp_opt *tp, u32 seq_rtt, int flag) if (flag & FLAG_RETRANS_DATA_ACKED) return; - tcp_rtt_estimator(tp, seq_rtt); - tcp_set_rto(tp); - tp->backoff = 0; - tcp_bound_rto(tp); + tcp_rtt_estimator(sk, seq_rtt); + tcp_set_rto(sk); + inet_csk(sk)->icsk_backoff = 0; + tcp_bound_rto(sk); } -static __inline__ void -tcp_ack_update_rtt(struct tcp_opt *tp, int flag, s32 seq_rtt) +static inline void tcp_ack_update_rtt(struct sock *sk, const int flag, + const s32 seq_rtt) { + const struct tcp_sock *tp = tcp_sk(sk); /* Note that peer MAY send zero echo. In this case it is ignored. (rfc1323) */ - if (tp->saw_tstamp && tp->rcv_tsecr) - tcp_ack_saw_tstamp(tp, flag); + if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) + tcp_ack_saw_tstamp(sk, flag); else if (seq_rtt >= 0) - tcp_ack_no_tstamp(tp, seq_rtt, flag); -} - -/* - * Compute congestion window to use. - * - * This is from the implementation of BICTCP in - * Lison-Xu, Kahaled Harfoush, and Injog Rhee. - * "Binary Increase Congestion Control for Fast, Long Distance - * Networks" in InfoComm 2004 - * Available from: - * http://www.csc.ncsu.edu/faculty/rhee/export/bitcp.pdf - * - * Unless BIC is enabled and congestion window is large - * this behaves the same as the original Reno. - */ -static inline __u32 bictcp_cwnd(struct tcp_opt *tp) -{ - /* orignal Reno behaviour */ - if (!tcp_is_bic(tp)) - return tp->snd_cwnd; - - if (tp->bictcp.last_cwnd == tp->snd_cwnd && - (s32)(tcp_time_stamp - tp->bictcp.last_stamp) <= (HZ>>5)) - return tp->bictcp.cnt; - - tp->bictcp.last_cwnd = tp->snd_cwnd; - tp->bictcp.last_stamp = tcp_time_stamp; - - /* start off normal */ - if (tp->snd_cwnd <= sysctl_tcp_bic_low_window) - tp->bictcp.cnt = tp->snd_cwnd; - - /* binary increase */ - else if (tp->snd_cwnd < tp->bictcp.last_max_cwnd) { - __u32 dist = (tp->bictcp.last_max_cwnd - tp->snd_cwnd) - / BICTCP_B; - - if (dist > BICTCP_MAX_INCREMENT) - /* linear increase */ - tp->bictcp.cnt = tp->snd_cwnd / BICTCP_MAX_INCREMENT; - else if (dist <= 1U) - /* binary search increase */ - tp->bictcp.cnt = tp->snd_cwnd * BICTCP_FUNC_OF_MIN_INCR - / BICTCP_B; - else - /* binary search increase */ - tp->bictcp.cnt = tp->snd_cwnd / dist; - } else { - /* slow start amd linear increase */ - if (tp->snd_cwnd < tp->bictcp.last_max_cwnd + BICTCP_B) - /* slow start */ - tp->bictcp.cnt = tp->snd_cwnd * BICTCP_FUNC_OF_MIN_INCR - / BICTCP_B; - else if (tp->snd_cwnd < tp->bictcp.last_max_cwnd - + BICTCP_MAX_INCREMENT*(BICTCP_B-1)) - /* slow start */ - tp->bictcp.cnt = tp->snd_cwnd * (BICTCP_B-1) - / (tp->snd_cwnd-tp->bictcp.last_max_cwnd); - else - /* linear increase */ - tp->bictcp.cnt = tp->snd_cwnd / BICTCP_MAX_INCREMENT; - } - return tp->bictcp.cnt; + tcp_ack_no_tstamp(sk, seq_rtt, flag); } -/* This is Jacobson's slow start and congestion avoidance. - * SIGCOMM '88, p. 328. - */ -static __inline__ void reno_cong_avoid(struct tcp_opt *tp) +static void tcp_cong_avoid(struct sock *sk, u32 ack, u32 rtt, + u32 in_flight, int good) { - if (tp->snd_cwnd <= tp->snd_ssthresh) { - /* In "safe" area, increase. */ - if (tp->snd_cwnd < tp->snd_cwnd_clamp) - tp->snd_cwnd++; - } else { - /* In dangerous area, increase slowly. - * In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd - */ - if (tp->snd_cwnd_cnt >= bictcp_cwnd(tp)) { - if (tp->snd_cwnd < tp->snd_cwnd_clamp) - tp->snd_cwnd++; - tp->snd_cwnd_cnt=0; - } else - tp->snd_cwnd_cnt++; - } - tp->snd_cwnd_stamp = tcp_time_stamp; -} - -/* This is based on the congestion detection/avoidance scheme described in - * Lawrence S. Brakmo and Larry L. Peterson. - * "TCP Vegas: End to end congestion avoidance on a global internet." - * IEEE Journal on Selected Areas in Communication, 13(8):1465--1480, - * October 1995. Available from: - * ftp://ftp.cs.arizona.edu/xkernel/Papers/jsac.ps - * - * See http://www.cs.arizona.edu/xkernel/ for their implementation. - * The main aspects that distinguish this implementation from the - * Arizona Vegas implementation are: - * o We do not change the loss detection or recovery mechanisms of - * Linux in any way. Linux already recovers from losses quite well, - * using fine-grained timers, NewReno, and FACK. - * o To avoid the performance penalty imposed by increasing cwnd - * only every-other RTT during slow start, we increase during - * every RTT during slow start, just like Reno. - * o Largely to allow continuous cwnd growth during slow start, - * we use the rate at which ACKs come back as the "actual" - * rate, rather than the rate at which data is sent. - * o To speed convergence to the right rate, we set the cwnd - * to achieve the right ("actual") rate when we exit slow start. - * o To filter out the noise caused by delayed ACKs, we use the - * minimum RTT sample observed during the last RTT to calculate - * the actual rate. - * o When the sender re-starts from idle, it waits until it has - * received ACKs for an entire flight of new data before making - * a cwnd adjustment decision. The original Vegas implementation - * assumed senders never went idle. - */ -static void vegas_cong_avoid(struct tcp_opt *tp, u32 ack, u32 seq_rtt) -{ - /* The key players are v_beg_snd_una and v_beg_snd_nxt. - * - * These are so named because they represent the approximate values - * of snd_una and snd_nxt at the beginning of the current RTT. More - * precisely, they represent the amount of data sent during the RTT. - * At the end of the RTT, when we receive an ACK for v_beg_snd_nxt, - * we will calculate that (v_beg_snd_nxt - v_beg_snd_una) outstanding - * bytes of data have been ACKed during the course of the RTT, giving - * an "actual" rate of: - * - * (v_beg_snd_nxt - v_beg_snd_una) / (rtt duration) - * - * Unfortunately, v_beg_snd_una is not exactly equal to snd_una, - * because delayed ACKs can cover more than one segment, so they - * don't line up nicely with the boundaries of RTTs. - * - * Another unfortunate fact of life is that delayed ACKs delay the - * advance of the left edge of our send window, so that the number - * of bytes we send in an RTT is often less than our cwnd will allow. - * So we keep track of our cwnd separately, in v_beg_snd_cwnd. - */ - - if (after(ack, tp->vegas.beg_snd_nxt)) { - /* Do the Vegas once-per-RTT cwnd adjustment. */ - u32 old_wnd, old_snd_cwnd; - - - /* Here old_wnd is essentially the window of data that was - * sent during the previous RTT, and has all - * been acknowledged in the course of the RTT that ended - * with the ACK we just received. Likewise, old_snd_cwnd - * is the cwnd during the previous RTT. - */ - old_wnd = (tp->vegas.beg_snd_nxt - tp->vegas.beg_snd_una) / - tp->mss_cache_std; - old_snd_cwnd = tp->vegas.beg_snd_cwnd; - - /* Save the extent of the current window so we can use this - * at the end of the next RTT. - */ - tp->vegas.beg_snd_una = tp->vegas.beg_snd_nxt; - tp->vegas.beg_snd_nxt = tp->snd_nxt; - tp->vegas.beg_snd_cwnd = tp->snd_cwnd; - - /* Take into account the current RTT sample too, to - * decrease the impact of delayed acks. This double counts - * this sample since we count it for the next window as well, - * but that's not too awful, since we're taking the min, - * rather than averaging. - */ - vegas_rtt_calc(tp, seq_rtt); - - /* We do the Vegas calculations only if we got enough RTT - * samples that we can be reasonably sure that we got - * at least one RTT sample that wasn't from a delayed ACK. - * If we only had 2 samples total, - * then that means we're getting only 1 ACK per RTT, which - * means they're almost certainly delayed ACKs. - * If we have 3 samples, we should be OK. - */ - - if (tp->vegas.cntRTT <= 2) { - /* We don't have enough RTT samples to do the Vegas - * calculation, so we'll behave like Reno. - */ - if (tp->snd_cwnd > tp->snd_ssthresh) - tp->snd_cwnd++; - } else { - u32 rtt, target_cwnd, diff; - - /* We have enough RTT samples, so, using the Vegas - * algorithm, we determine if we should increase or - * decrease cwnd, and by how much. - */ - - /* Pluck out the RTT we are using for the Vegas - * calculations. This is the min RTT seen during the - * last RTT. Taking the min filters out the effects - * of delayed ACKs, at the cost of noticing congestion - * a bit later. - */ - rtt = tp->vegas.minRTT; - - /* Calculate the cwnd we should have, if we weren't - * going too fast. - * - * This is: - * (actual rate in segments) * baseRTT - * We keep it as a fixed point number with - * V_PARAM_SHIFT bits to the right of the binary point. - */ - target_cwnd = ((old_wnd * tp->vegas.baseRTT) - << V_PARAM_SHIFT) / rtt; - - /* Calculate the difference between the window we had, - * and the window we would like to have. This quantity - * is the "Diff" from the Arizona Vegas papers. - * - * Again, this is a fixed point number with - * V_PARAM_SHIFT bits to the right of the binary - * point. - */ - diff = (old_wnd << V_PARAM_SHIFT) - target_cwnd; - - if (tp->snd_cwnd < tp->snd_ssthresh) { - /* Slow start. */ - if (diff > sysctl_tcp_vegas_gamma) { - /* Going too fast. Time to slow down - * and switch to congestion avoidance. - */ - tp->snd_ssthresh = 2; - - /* Set cwnd to match the actual rate - * exactly: - * cwnd = (actual rate) * baseRTT - * Then we add 1 because the integer - * truncation robs us of full link - * utilization. - */ - tp->snd_cwnd = min(tp->snd_cwnd, - (target_cwnd >> - V_PARAM_SHIFT)+1); - - } - } else { - /* Congestion avoidance. */ - u32 next_snd_cwnd; - - /* Figure out where we would like cwnd - * to be. - */ - if (diff > sysctl_tcp_vegas_beta) { - /* The old window was too fast, so - * we slow down. - */ - next_snd_cwnd = old_snd_cwnd - 1; - } else if (diff < sysctl_tcp_vegas_alpha) { - /* We don't have enough extra packets - * in the network, so speed up. - */ - next_snd_cwnd = old_snd_cwnd + 1; - } else { - /* Sending just as fast as we - * should be. - */ - next_snd_cwnd = old_snd_cwnd; - } - - /* Adjust cwnd upward or downward, toward the - * desired value. - */ - if (next_snd_cwnd > tp->snd_cwnd) - tp->snd_cwnd++; - else if (next_snd_cwnd < tp->snd_cwnd) - tp->snd_cwnd--; - } - } - - /* Wipe the slate clean for the next RTT. */ - tp->vegas.cntRTT = 0; - tp->vegas.minRTT = 0x7fffffff; - } - - /* The following code is executed for every ack we receive, - * except for conditions checked in should_advance_cwnd() - * before the call to tcp_cong_avoid(). Mainly this means that - * we only execute this code if the ack actually acked some - * data. - */ - - /* If we are in slow start, increase our cwnd in response to this ACK. - * (If we are not in slow start then we are in congestion avoidance, - * and adjust our congestion window only once per RTT. See the code - * above.) - */ - if (tp->snd_cwnd <= tp->snd_ssthresh) - tp->snd_cwnd++; - - /* to keep cwnd from growing without bound */ - tp->snd_cwnd = min_t(u32, tp->snd_cwnd, tp->snd_cwnd_clamp); - - /* Make sure that we are never so timid as to reduce our cwnd below - * 2 MSS. - * - * Going below 2 MSS would risk huge delayed ACKs from our receiver. - */ - tp->snd_cwnd = max(tp->snd_cwnd, 2U); - - tp->snd_cwnd_stamp = tcp_time_stamp; -} - -static inline void tcp_cong_avoid(struct tcp_opt *tp, u32 ack, u32 seq_rtt) -{ - if (tcp_vegas_enabled(tp)) - vegas_cong_avoid(tp, ack, seq_rtt); - else - reno_cong_avoid(tp); + const struct inet_connection_sock *icsk = inet_csk(sk); + icsk->icsk_ca_ops->cong_avoid(sk, ack, rtt, in_flight, good); + tcp_sk(sk)->snd_cwnd_stamp = tcp_time_stamp; } /* Restart timer after forward progress on connection. * RFC2988 recommends to restart timer to now+rto. */ -static __inline__ void tcp_ack_packets_out(struct sock *sk, struct tcp_opt *tp) +static void tcp_ack_packets_out(struct sock *sk, struct tcp_sock *tp) { - if (!tcp_get_pcount(&tp->packets_out)) { - tcp_clear_xmit_timer(sk, TCP_TIME_RETRANS); + if (!tp->packets_out) { + inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS); } else { - tcp_reset_xmit_timer(sk, TCP_TIME_RETRANS, tp->rto); + inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, inet_csk(sk)->icsk_rto, TCP_RTO_MAX); } } -/* There is one downside to this scheme. Although we keep the - * ACK clock ticking, adjusting packet counters and advancing - * congestion window, we do not liberate socket send buffer - * space. - * - * Mucking with skb->truesize and sk->sk_wmem_alloc et al. - * then making a write space wakeup callback is a possible - * future enhancement. WARNING: it is not trivial to make. - */ static int tcp_tso_acked(struct sock *sk, struct sk_buff *skb, __u32 now, __s32 *seq_rtt) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); struct tcp_skb_cb *scb = TCP_SKB_CB(skb); - __u32 mss = tcp_skb_mss(skb); - __u32 snd_una = tp->snd_una; - __u32 orig_seq, seq; - __u32 packets_acked = 0; + __u32 seq = tp->snd_una; + __u32 packets_acked; int acked = 0; /* If we get here, the whole TSO packet has not been * acked. */ - BUG_ON(!after(scb->end_seq, snd_una)); + BUG_ON(!after(scb->end_seq, seq)); - seq = orig_seq = scb->seq; - while (!after(seq + mss, snd_una)) { - packets_acked++; - seq += mss; - } - - if (tcp_trim_head(sk, skb, (seq - orig_seq))) + packets_acked = tcp_skb_pcount(skb); + if (tcp_trim_head(sk, skb, seq - scb->seq)) return 0; + packets_acked -= tcp_skb_pcount(skb); if (packets_acked) { __u8 sacked = scb->sacked; @@ -2396,18 +2210,15 @@ static int tcp_tso_acked(struct sock *sk, struct sk_buff *skb, if (sacked) { if (sacked & TCPCB_RETRANS) { if (sacked & TCPCB_SACKED_RETRANS) - tcp_dec_pcount_explicit(&tp->retrans_out, - packets_acked); + tp->retrans_out -= packets_acked; acked |= FLAG_RETRANS_DATA_ACKED; *seq_rtt = -1; } else if (*seq_rtt < 0) *seq_rtt = now - scb->when; if (sacked & TCPCB_SACKED_ACKED) - tcp_dec_pcount_explicit(&tp->sacked_out, - packets_acked); + tp->sacked_out -= packets_acked; if (sacked & TCPCB_LOST) - tcp_dec_pcount_explicit(&tp->lost_out, - packets_acked); + tp->lost_out -= packets_acked; if (sacked & TCPCB_URG) { if (tp->urg_mode && !before(seq, tp->snd_up)) @@ -2416,12 +2227,11 @@ static int tcp_tso_acked(struct sock *sk, struct sk_buff *skb, } else if (*seq_rtt < 0) *seq_rtt = now - scb->when; - if (tcp_get_pcount(&tp->fackets_out)) { - __u32 dval = min(tcp_get_pcount(&tp->fackets_out), - packets_acked); - tcp_dec_pcount_explicit(&tp->fackets_out, dval); + if (tp->fackets_out) { + __u32 dval = min(tp->fackets_out, packets_acked); + tp->fackets_out -= dval; } - tcp_dec_pcount_explicit(&tp->packets_out, packets_acked); + tp->packets_out -= packets_acked; BUG_ON(tcp_skb_pcount(skb) == 0); BUG_ON(!before(scb->seq, scb->end_seq)); @@ -2430,15 +2240,27 @@ static int tcp_tso_acked(struct sock *sk, struct sk_buff *skb, return acked; } +static u32 tcp_usrtt(struct timeval *tv) +{ + struct timeval now; + + do_gettimeofday(&now); + return (now.tv_sec - tv->tv_sec) * 1000000 + (now.tv_usec - tv->tv_usec); +} /* Remove acknowledged frames from the retransmission queue. */ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); + const struct inet_connection_sock *icsk = inet_csk(sk); struct sk_buff *skb; __u32 now = tcp_time_stamp; int acked = 0; __s32 seq_rtt = -1; + u32 pkts_acked = 0; + void (*rtt_sample)(struct sock *sk, u32 usrtt) + = icsk->icsk_ca_ops->rtt_sample; + struct timeval tv = { .tv_sec = 0, .tv_usec = 0 }; while ((skb = skb_peek(&sk->sk_write_queue)) && skb != sk->sk_send_head) { @@ -2450,7 +2272,8 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) * the other end. */ if (after(scb->end_seq, tp->snd_una)) { - if (tcp_skb_pcount(skb) > 1) + if (tcp_skb_pcount(skb) > 1 && + after(tp->snd_una, scb->seq)) acked |= tcp_tso_acked(sk, skb, now, &seq_rtt); break; @@ -2465,63 +2288,79 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) */ if (!(scb->flags & TCPCB_FLAG_SYN)) { acked |= FLAG_DATA_ACKED; + ++pkts_acked; } else { acked |= FLAG_SYN_ACKED; tp->retrans_stamp = 0; } + /* MTU probing checks */ + if (icsk->icsk_mtup.probe_size) { + if (!after(tp->mtu_probe.probe_seq_end, TCP_SKB_CB(skb)->end_seq)) { + tcp_mtup_probe_success(sk, skb); + } + } + if (sacked) { if (sacked & TCPCB_RETRANS) { if(sacked & TCPCB_SACKED_RETRANS) - tcp_dec_pcount(&tp->retrans_out, skb); + tp->retrans_out -= tcp_skb_pcount(skb); acked |= FLAG_RETRANS_DATA_ACKED; seq_rtt = -1; - } else if (seq_rtt < 0) + } else if (seq_rtt < 0) { seq_rtt = now - scb->when; + skb_get_timestamp(skb, &tv); + } if (sacked & TCPCB_SACKED_ACKED) - tcp_dec_pcount(&tp->sacked_out, skb); + tp->sacked_out -= tcp_skb_pcount(skb); if (sacked & TCPCB_LOST) - tcp_dec_pcount(&tp->lost_out, skb); + tp->lost_out -= tcp_skb_pcount(skb); if (sacked & TCPCB_URG) { if (tp->urg_mode && !before(scb->end_seq, tp->snd_up)) tp->urg_mode = 0; } - } else if (seq_rtt < 0) + } else if (seq_rtt < 0) { seq_rtt = now - scb->when; + skb_get_timestamp(skb, &tv); + } tcp_dec_pcount_approx(&tp->fackets_out, skb); tcp_packets_out_dec(tp, skb); - __skb_unlink(skb, skb->list); + __skb_unlink(skb, &sk->sk_write_queue); sk_stream_free_skb(sk, skb); + clear_all_retrans_hints(tp); } if (acked&FLAG_ACKED) { - tcp_ack_update_rtt(tp, acked, seq_rtt); + tcp_ack_update_rtt(sk, acked, seq_rtt); tcp_ack_packets_out(sk, tp); + if (rtt_sample && !(acked & FLAG_RETRANS_DATA_ACKED)) + (*rtt_sample)(sk, tcp_usrtt(&tv)); + + if (icsk->icsk_ca_ops->pkts_acked) + icsk->icsk_ca_ops->pkts_acked(sk, pkts_acked); } #if FASTRETRANS_DEBUG > 0 - BUG_TRAP((int)tcp_get_pcount(&tp->sacked_out) >= 0); - BUG_TRAP((int)tcp_get_pcount(&tp->lost_out) >= 0); - BUG_TRAP((int)tcp_get_pcount(&tp->retrans_out) >= 0); - if (!tcp_get_pcount(&tp->packets_out) && tp->sack_ok) { - if (tcp_get_pcount(&tp->lost_out)) { + BUG_TRAP((int)tp->sacked_out >= 0); + BUG_TRAP((int)tp->lost_out >= 0); + BUG_TRAP((int)tp->retrans_out >= 0); + if (!tp->packets_out && tp->rx_opt.sack_ok) { + const struct inet_connection_sock *icsk = inet_csk(sk); + if (tp->lost_out) { printk(KERN_DEBUG "Leak l=%u %d\n", - tcp_get_pcount(&tp->lost_out), - tp->ca_state); - tcp_set_pcount(&tp->lost_out, 0); + tp->lost_out, icsk->icsk_ca_state); + tp->lost_out = 0; } - if (tcp_get_pcount(&tp->sacked_out)) { + if (tp->sacked_out) { printk(KERN_DEBUG "Leak s=%u %d\n", - tcp_get_pcount(&tp->sacked_out), - tp->ca_state); - tcp_set_pcount(&tp->sacked_out, 0); + tp->sacked_out, icsk->icsk_ca_state); + tp->sacked_out = 0; } - if (tcp_get_pcount(&tp->retrans_out)) { + if (tp->retrans_out) { printk(KERN_DEBUG "Leak r=%u %d\n", - tcp_get_pcount(&tp->retrans_out), - tp->ca_state); - tcp_set_pcount(&tp->retrans_out, 0); + tp->retrans_out, icsk->icsk_ca_state); + tp->retrans_out = 0; } } #endif @@ -2531,40 +2370,43 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) static void tcp_ack_probe(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + const struct tcp_sock *tp = tcp_sk(sk); + struct inet_connection_sock *icsk = inet_csk(sk); /* Was it a usable window open? */ if (!after(TCP_SKB_CB(sk->sk_send_head)->end_seq, tp->snd_una + tp->snd_wnd)) { - tp->backoff = 0; - tcp_clear_xmit_timer(sk, TCP_TIME_PROBE0); + icsk->icsk_backoff = 0; + inet_csk_clear_xmit_timer(sk, ICSK_TIME_PROBE0); /* Socket must be waked up by subsequent tcp_data_snd_check(). * This function is not for random using! */ } else { - tcp_reset_xmit_timer(sk, TCP_TIME_PROBE0, - min(tp->rto << tp->backoff, TCP_RTO_MAX)); + inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, + min(icsk->icsk_rto << icsk->icsk_backoff, TCP_RTO_MAX), + TCP_RTO_MAX); } } -static __inline__ int tcp_ack_is_dubious(struct tcp_opt *tp, int flag) +static inline int tcp_ack_is_dubious(const struct sock *sk, const int flag) { return (!(flag & FLAG_NOT_DUP) || (flag & FLAG_CA_ALERT) || - tp->ca_state != TCP_CA_Open); + inet_csk(sk)->icsk_ca_state != TCP_CA_Open); } -static __inline__ int tcp_may_raise_cwnd(struct tcp_opt *tp, int flag) +static inline int tcp_may_raise_cwnd(const struct sock *sk, const int flag) { + const struct tcp_sock *tp = tcp_sk(sk); return (!(flag & FLAG_ECE) || tp->snd_cwnd < tp->snd_ssthresh) && - !((1<ca_state)&(TCPF_CA_Recovery|TCPF_CA_CWR)); + !((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_Recovery | TCPF_CA_CWR)); } /* Check that window update is acceptable. * The function assumes that snd_una<=ack<=snd_next. */ -static __inline__ int -tcp_may_update_window(struct tcp_opt *tp, u32 ack, u32 ack_seq, u32 nwin) +static inline int tcp_may_update_window(const struct tcp_sock *tp, const u32 ack, + const u32 ack_seq, const u32 nwin) { return (after(ack, tp->snd_una) || after(ack_seq, tp->snd_wl1) || @@ -2576,14 +2418,14 @@ tcp_may_update_window(struct tcp_opt *tp, u32 ack, u32 ack_seq, u32 nwin) * Window update algorithm, described in RFC793/RFC1122 (used in linux-2.2 * and in FreeBSD. NetBSD's one is even worse.) is wrong. */ -static int tcp_ack_update_window(struct sock *sk, struct tcp_opt *tp, +static int tcp_ack_update_window(struct sock *sk, struct tcp_sock *tp, struct sk_buff *skb, u32 ack, u32 ack_seq) { int flag = 0; u32 nwin = ntohs(skb->h.th->window); if (likely(!skb->h.th->syn)) - nwin <<= tp->snd_wscale; + nwin <<= tp->rx_opt.snd_wscale; if (tcp_may_update_window(tp, ack, ack_seq, nwin)) { flag |= FLAG_WIN_UPDATE; @@ -2595,11 +2437,12 @@ static int tcp_ack_update_window(struct sock *sk, struct tcp_opt *tp, /* Note, it is the only place, where * fast path is recovered for sending TCP. */ + tp->pred_flags = 0; tcp_fast_path_check(sk, tp); if (nwin > tp->max_window) { tp->max_window = nwin; - tcp_sync_mss(sk, tp->pmtu_cookie); + tcp_sync_mss(sk, inet_csk(sk)->icsk_pmtu_cookie); } } } @@ -2611,7 +2454,7 @@ static int tcp_ack_update_window(struct sock *sk, struct tcp_opt *tp, static void tcp_process_frto(struct sock *sk, u32 prior_snd_una) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); tcp_sync_left_out(tp); @@ -2639,254 +2482,16 @@ static void tcp_process_frto(struct sock *sk, u32 prior_snd_una) } /* F-RTO affects on two new ACKs following RTO. - * At latest on third ACK the TCP behavor is back to normal. + * At latest on third ACK the TCP behavior is back to normal. */ tp->frto_counter = (tp->frto_counter + 1) % 3; } -/* - * TCP Westwood+ - */ - -/* - * @init_westwood - * This function initializes fields used in TCP Westwood+. We can't - * get no information about RTTmin at this time so we simply set it to - * TCP_WESTWOOD_INIT_RTT. This value was chosen to be too conservative - * since in this way we're sure it will be updated in a consistent - * way as soon as possible. It will reasonably happen within the first - * RTT period of the connection lifetime. - */ - -static void init_westwood(struct sock *sk) -{ - struct tcp_opt *tp = tcp_sk(sk); - - tp->westwood.bw_ns_est = 0; - tp->westwood.bw_est = 0; - tp->westwood.accounted = 0; - tp->westwood.cumul_ack = 0; - tp->westwood.rtt_win_sx = tcp_time_stamp; - tp->westwood.rtt = TCP_WESTWOOD_INIT_RTT; - tp->westwood.rtt_min = TCP_WESTWOOD_INIT_RTT; - tp->westwood.snd_una = tp->snd_una; -} - -/* - * @westwood_do_filter - * Low-pass filter. Implemented using constant coeffients. - */ - -static inline __u32 westwood_do_filter(__u32 a, __u32 b) -{ - return (((7 * a) + b) >> 3); -} - -static void westwood_filter(struct sock *sk, __u32 delta) -{ - struct tcp_opt *tp = tcp_sk(sk); - - tp->westwood.bw_ns_est = - westwood_do_filter(tp->westwood.bw_ns_est, - tp->westwood.bk / delta); - tp->westwood.bw_est = - westwood_do_filter(tp->westwood.bw_est, - tp->westwood.bw_ns_est); -} - -/* - * @westwood_update_rttmin - * It is used to update RTTmin. In this case we MUST NOT use - * WESTWOOD_RTT_MIN minimum bound since we could be on a LAN! - */ - -static inline __u32 westwood_update_rttmin(const struct sock *sk) -{ - const struct tcp_opt *tp = tcp_sk(sk); - __u32 rttmin = tp->westwood.rtt_min; - - if (tp->westwood.rtt != 0 && - (tp->westwood.rtt < tp->westwood.rtt_min || !rttmin)) - rttmin = tp->westwood.rtt; - - return rttmin; -} - -/* - * @westwood_acked - * Evaluate increases for dk. - */ - -static inline __u32 westwood_acked(const struct sock *sk) -{ - const struct tcp_opt *tp = tcp_sk(sk); - - return tp->snd_una - tp->westwood.snd_una; -} - -/* - * @westwood_new_window - * It evaluates if we are receiving data inside the same RTT window as - * when we started. - * Return value: - * It returns 0 if we are still evaluating samples in the same RTT - * window, 1 if the sample has to be considered in the next window. - */ - -static int westwood_new_window(const struct sock *sk) -{ - const struct tcp_opt *tp = tcp_sk(sk); - __u32 left_bound; - __u32 rtt; - int ret = 0; - - left_bound = tp->westwood.rtt_win_sx; - rtt = max(tp->westwood.rtt, (u32) TCP_WESTWOOD_RTT_MIN); - - /* - * A RTT-window has passed. Be careful since if RTT is less than - * 50ms we don't filter but we continue 'building the sample'. - * This minimum limit was choosen since an estimation on small - * time intervals is better to avoid... - * Obvioulsy on a LAN we reasonably will always have - * right_bound = left_bound + WESTWOOD_RTT_MIN - */ - - if ((left_bound + rtt) < tcp_time_stamp) - ret = 1; - - return ret; -} - -/* - * @westwood_update_window - * It updates RTT evaluation window if it is the right moment to do - * it. If so it calls filter for evaluating bandwidth. - */ - -static void __westwood_update_window(struct sock *sk, __u32 now) -{ - struct tcp_opt *tp = tcp_sk(sk); - __u32 delta = now - tp->westwood.rtt_win_sx; - - if (delta) { - if (tp->westwood.rtt) - westwood_filter(sk, delta); - - tp->westwood.bk = 0; - tp->westwood.rtt_win_sx = tcp_time_stamp; - } -} - - -static void westwood_update_window(struct sock *sk, __u32 now) -{ - if (westwood_new_window(sk)) - __westwood_update_window(sk, now); -} - -/* - * @__tcp_westwood_fast_bw - * It is called when we are in fast path. In particular it is called when - * header prediction is successfull. In such case infact update is - * straight forward and doesn't need any particular care. - */ - -void __tcp_westwood_fast_bw(struct sock *sk, struct sk_buff *skb) -{ - struct tcp_opt *tp = tcp_sk(sk); - - westwood_update_window(sk, tcp_time_stamp); - - tp->westwood.bk += westwood_acked(sk); - tp->westwood.snd_una = tp->snd_una; - tp->westwood.rtt_min = westwood_update_rttmin(sk); -} - - -/* - * @westwood_dupack_update - * It updates accounted and cumul_ack when receiving a dupack. - */ - -static void westwood_dupack_update(struct sock *sk) -{ - struct tcp_opt *tp = tcp_sk(sk); - - tp->westwood.accounted += tp->mss_cache_std; - tp->westwood.cumul_ack = tp->mss_cache_std; -} - -static inline int westwood_may_change_cumul(struct tcp_opt *tp) -{ - return (tp->westwood.cumul_ack > tp->mss_cache_std); -} - -static inline void westwood_partial_update(struct tcp_opt *tp) -{ - tp->westwood.accounted -= tp->westwood.cumul_ack; - tp->westwood.cumul_ack = tp->mss_cache_std; -} - -static inline void westwood_complete_update(struct tcp_opt *tp) -{ - tp->westwood.cumul_ack -= tp->westwood.accounted; - tp->westwood.accounted = 0; -} - -/* - * @westwood_acked_count - * This function evaluates cumul_ack for evaluating dk in case of - * delayed or partial acks. - */ - -static inline __u32 westwood_acked_count(struct sock *sk) -{ - struct tcp_opt *tp = tcp_sk(sk); - - tp->westwood.cumul_ack = westwood_acked(sk); - - /* If cumul_ack is 0 this is a dupack since it's not moving - * tp->snd_una. - */ - if (!(tp->westwood.cumul_ack)) - westwood_dupack_update(sk); - - if (westwood_may_change_cumul(tp)) { - /* Partial or delayed ack */ - if (tp->westwood.accounted >= tp->westwood.cumul_ack) - westwood_partial_update(tp); - else - westwood_complete_update(tp); - } - - tp->westwood.snd_una = tp->snd_una; - - return tp->westwood.cumul_ack; -} - - -/* - * @__tcp_westwood_slow_bw - * It is called when something is going wrong..even if there could - * be no problems! Infact a simple delayed packet may trigger a - * dupack. But we need to be careful in such case. - */ - -void __tcp_westwood_slow_bw(struct sock *sk, struct sk_buff *skb) -{ - struct tcp_opt *tp = tcp_sk(sk); - - westwood_update_window(sk, tcp_time_stamp); - - tp->westwood.bk += westwood_acked_count(sk); - tp->westwood.rtt_min = westwood_update_rttmin(sk); -} - /* This routine deals with incoming acks, but not outgoing ones. */ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) { - struct tcp_opt *tp = tcp_sk(sk); + struct inet_connection_sock *icsk = inet_csk(sk); + struct tcp_sock *tp = tcp_sk(sk); u32 prior_snd_una = tp->snd_una; u32 ack_seq = TCP_SKB_CB(skb)->seq; u32 ack = TCP_SKB_CB(skb)->ack_seq; @@ -2903,6 +2508,14 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) if (before(ack, prior_snd_una)) goto old_ack; + if (sysctl_tcp_abc) { + if (icsk->icsk_ca_state < TCP_CA_CWR) + tp->bytes_acked += ack - prior_snd_una; + else if (icsk->icsk_ca_state == TCP_CA_Loss) + /* we assume just one segment left network */ + tp->bytes_acked += min(ack - prior_snd_una, tp->mss_cache); + } + if (!(flag&FLAG_SLOWPATH) && after(ack, prior_snd_una)) { /* Window is constant, pure forward advance. * No more checks are required. @@ -2910,9 +2523,10 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) */ tcp_update_wl(tp, ack, ack_seq); tp->snd_una = ack; - tcp_westwood_fast_bw(sk, skb); flag |= FLAG_WIN_UPDATE; + tcp_ca_event(sk, CA_EVENT_FAST_ACK); + NET_INC_STATS_BH(LINUX_MIB_TCPHPACKS); } else { if (ack_seq != TCP_SKB_CB(skb)->end_seq) @@ -2928,7 +2542,7 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) if (TCP_ECN_rcv_ecn_echo(tp, skb->h.th)) flag |= FLAG_ECE; - tcp_westwood_slow_bw(sk,skb); + tcp_ca_event(sk, CA_EVENT_SLOW_ACK); } /* We passed data and got it acked, remove any soft error @@ -2936,7 +2550,7 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) */ sk->sk_err_soft = 0; tp->rcv_tstamp = tcp_time_stamp; - prior_packets = tcp_get_pcount(&tp->packets_out); + prior_packets = tp->packets_out; if (!prior_packets) goto no_queue; @@ -2948,17 +2562,14 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) if (tp->frto_counter) tcp_process_frto(sk, prior_snd_una); - if (tcp_ack_is_dubious(tp, flag)) { - /* Advanve CWND, if state allows this. */ - if ((flag & FLAG_DATA_ACKED) && - (tcp_vegas_enabled(tp) || prior_in_flight >= tp->snd_cwnd) && - tcp_may_raise_cwnd(tp, flag)) - tcp_cong_avoid(tp, ack, seq_rtt); + if (tcp_ack_is_dubious(sk, flag)) { + /* Advance CWND, if state allows this. */ + if ((flag & FLAG_DATA_ACKED) && tcp_may_raise_cwnd(sk, flag)) + tcp_cong_avoid(sk, ack, seq_rtt, prior_in_flight, 0); tcp_fastretrans_alert(sk, prior_snd_una, prior_packets, flag); } else { - if ((flag & FLAG_DATA_ACKED) && - (tcp_vegas_enabled(tp) || prior_in_flight >= tp->snd_cwnd)) - tcp_cong_avoid(tp, ack, seq_rtt); + if ((flag & FLAG_DATA_ACKED)) + tcp_cong_avoid(sk, ack, seq_rtt, prior_in_flight, 1); } if ((flag & FLAG_FORWARD_PROGRESS) || !(flag&FLAG_NOT_DUP)) @@ -2967,7 +2578,7 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) return 1; no_queue: - tp->probes_out = 0; + icsk->icsk_probes_out = 0; /* If this ack opens up a zero window, clear backoff. It was * being used to time the probes, and is probably far higher than @@ -2991,14 +2602,14 @@ uninteresting_ack: * But, this can also be called on packets in the established flow when * the fast version below fails. */ -void tcp_parse_options(struct sk_buff *skb, struct tcp_opt *tp, int estab) +void tcp_parse_options(struct sk_buff *skb, struct tcp_options_received *opt_rx, int estab) { unsigned char *ptr; struct tcphdr *th = skb->h.th; int length=(th->doff*4)-sizeof(struct tcphdr); ptr = (unsigned char *)(th + 1); - tp->saw_tstamp = 0; + opt_rx->saw_tstamp = 0; while(length>0) { int opcode=*ptr++; @@ -3019,43 +2630,44 @@ void tcp_parse_options(struct sk_buff *skb, struct tcp_opt *tp, int estab) switch(opcode) { case TCPOPT_MSS: if(opsize==TCPOLEN_MSS && th->syn && !estab) { - u16 in_mss = ntohs(*(__u16 *)ptr); + u16 in_mss = ntohs(get_unaligned((__be16 *)ptr)); if (in_mss) { - if (tp->user_mss && tp->user_mss < in_mss) - in_mss = tp->user_mss; - tp->mss_clamp = in_mss; + if (opt_rx->user_mss && opt_rx->user_mss < in_mss) + in_mss = opt_rx->user_mss; + opt_rx->mss_clamp = in_mss; } } break; case TCPOPT_WINDOW: if(opsize==TCPOLEN_WINDOW && th->syn && !estab) if (sysctl_tcp_window_scaling) { - tp->wscale_ok = 1; - tp->snd_wscale = *(__u8 *)ptr; - if(tp->snd_wscale > 14) { + __u8 snd_wscale = *(__u8 *) ptr; + opt_rx->wscale_ok = 1; + if (snd_wscale > 14) { if(net_ratelimit()) - printk("tcp_parse_options: Illegal window " - "scaling value %d >14 received.", - tp->snd_wscale); - tp->snd_wscale = 14; + printk(KERN_INFO "tcp_parse_options: Illegal window " + "scaling value %d >14 received.\n", + snd_wscale); + snd_wscale = 14; } + opt_rx->snd_wscale = snd_wscale; } break; case TCPOPT_TIMESTAMP: if(opsize==TCPOLEN_TIMESTAMP) { - if ((estab && tp->tstamp_ok) || + if ((estab && opt_rx->tstamp_ok) || (!estab && sysctl_tcp_timestamps)) { - tp->saw_tstamp = 1; - tp->rcv_tsval = ntohl(*(__u32 *)ptr); - tp->rcv_tsecr = ntohl(*(__u32 *)(ptr+4)); + opt_rx->saw_tstamp = 1; + opt_rx->rcv_tsval = ntohl(get_unaligned((__be32 *)ptr)); + opt_rx->rcv_tsecr = ntohl(get_unaligned((__be32 *)(ptr+4))); } } break; case TCPOPT_SACK_PERM: if(opsize==TCPOLEN_SACK_PERM && th->syn && !estab) { if (sysctl_tcp_sack) { - tp->sack_ok = 1; - tcp_sack_reset(tp); + opt_rx->sack_ok = 1; + tcp_sack_reset(opt_rx); } } break; @@ -3063,9 +2675,17 @@ void tcp_parse_options(struct sk_buff *skb, struct tcp_opt *tp, int estab) case TCPOPT_SACK: if((opsize >= (TCPOLEN_SACK_BASE + TCPOLEN_SACK_PERBLOCK)) && !((opsize - TCPOLEN_SACK_BASE) % TCPOLEN_SACK_PERBLOCK) && - tp->sack_ok) { + opt_rx->sack_ok) { TCP_SKB_CB(skb)->sacked = (ptr - 2) - (unsigned char *)th; } +#ifdef CONFIG_TCP_MD5SIG + case TCPOPT_MD5SIG: + /* + * The MD5 Hash has already been + * checked (see tcp_v{4,6}_do_rcv()). + */ + break; +#endif }; ptr+=opsize-2; length-=opsize; @@ -3076,39 +2696,38 @@ void tcp_parse_options(struct sk_buff *skb, struct tcp_opt *tp, int estab) /* Fast parse options. This hopes to only see timestamps. * If it is wrong it falls back on tcp_parse_options(). */ -static __inline__ int tcp_fast_parse_options(struct sk_buff *skb, struct tcphdr *th, struct tcp_opt *tp) +static int tcp_fast_parse_options(struct sk_buff *skb, struct tcphdr *th, + struct tcp_sock *tp) { if (th->doff == sizeof(struct tcphdr)>>2) { - tp->saw_tstamp = 0; + tp->rx_opt.saw_tstamp = 0; return 0; - } else if (tp->tstamp_ok && + } else if (tp->rx_opt.tstamp_ok && th->doff == (sizeof(struct tcphdr)>>2)+(TCPOLEN_TSTAMP_ALIGNED>>2)) { - __u32 *ptr = (__u32 *)(th + 1); - if (*ptr == ntohl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) + __be32 *ptr = (__be32 *)(th + 1); + if (*ptr == htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP)) { - tp->saw_tstamp = 1; + tp->rx_opt.saw_tstamp = 1; ++ptr; - tp->rcv_tsval = ntohl(*ptr); + tp->rx_opt.rcv_tsval = ntohl(*ptr); ++ptr; - tp->rcv_tsecr = ntohl(*ptr); + tp->rx_opt.rcv_tsecr = ntohl(*ptr); return 1; } } - tcp_parse_options(skb, tp, 1); + tcp_parse_options(skb, &tp->rx_opt, 1); return 1; } -static __inline__ void -tcp_store_ts_recent(struct tcp_opt *tp) +static inline void tcp_store_ts_recent(struct tcp_sock *tp) { - tp->ts_recent = tp->rcv_tsval; - tp->ts_recent_stamp = xtime.tv_sec; + tp->rx_opt.ts_recent = tp->rx_opt.rcv_tsval; + tp->rx_opt.ts_recent_stamp = xtime.tv_sec; } -static __inline__ void -tcp_replace_ts_recent(struct tcp_opt *tp, u32 seq) +static inline void tcp_replace_ts_recent(struct tcp_sock *tp, u32 seq) { - if (tp->saw_tstamp && !after(seq, tp->rcv_wup)) { + if (tp->rx_opt.saw_tstamp && !after(seq, tp->rcv_wup)) { /* PAWS bug workaround wrt. ACK frames, the PAWS discard * extra check below makes sure this can only happen * for pure ACK frames. -DaveM @@ -3116,8 +2735,8 @@ tcp_replace_ts_recent(struct tcp_opt *tp, u32 seq) * Not only, also it occurs for expired timestamps. */ - if((s32)(tp->rcv_tsval - tp->ts_recent) >= 0 || - xtime.tv_sec >= tp->ts_recent_stamp + TCP_PAWS_24DAYS) + if((s32)(tp->rx_opt.rcv_tsval - tp->rx_opt.ts_recent) >= 0 || + xtime.tv_sec >= tp->rx_opt.ts_recent_stamp + TCP_PAWS_24DAYS) tcp_store_ts_recent(tp); } } @@ -3145,8 +2764,9 @@ tcp_replace_ts_recent(struct tcp_opt *tp, u32 seq) * up to bandwidth of 18Gigabit/sec. 8) ] */ -static int tcp_disordered_ack(struct tcp_opt *tp, struct sk_buff *skb) +static int tcp_disordered_ack(const struct sock *sk, const struct sk_buff *skb) { + struct tcp_sock *tp = tcp_sk(sk); struct tcphdr *th = skb->h.th; u32 seq = TCP_SKB_CB(skb)->seq; u32 ack = TCP_SKB_CB(skb)->ack_seq; @@ -3158,17 +2778,18 @@ static int tcp_disordered_ack(struct tcp_opt *tp, struct sk_buff *skb) ack == tp->snd_una && /* 3. ... and does not update window. */ - !tcp_may_update_window(tp, ack, seq, ntohs(th->window)<snd_wscale) && + !tcp_may_update_window(tp, ack, seq, ntohs(th->window) << tp->rx_opt.snd_wscale) && /* 4. ... and sits in replay window. */ - (s32)(tp->ts_recent - tp->rcv_tsval) <= (tp->rto*1024)/HZ); + (s32)(tp->rx_opt.ts_recent - tp->rx_opt.rcv_tsval) <= (inet_csk(sk)->icsk_rto * 1024) / HZ); } -static __inline__ int tcp_paws_discard(struct tcp_opt *tp, struct sk_buff *skb) +static inline int tcp_paws_discard(const struct sock *sk, const struct sk_buff *skb) { - return ((s32)(tp->ts_recent - tp->rcv_tsval) > TCP_PAWS_WINDOW && - xtime.tv_sec < tp->ts_recent_stamp + TCP_PAWS_24DAYS && - !tcp_disordered_ack(tp, skb)); + const struct tcp_sock *tp = tcp_sk(sk); + return ((s32)(tp->rx_opt.ts_recent - tp->rx_opt.rcv_tsval) > TCP_PAWS_WINDOW && + xtime.tv_sec < tp->rx_opt.ts_recent_stamp + TCP_PAWS_24DAYS && + !tcp_disordered_ack(sk, skb)); } /* Check segment sequence number for validity. @@ -3184,7 +2805,7 @@ static __inline__ int tcp_paws_discard(struct tcp_opt *tp, struct sk_buff *skb) * (borrowed from freebsd) */ -static inline int tcp_sequence(struct tcp_opt *tp, u32 seq, u32 end_seq) +static inline int tcp_sequence(struct tcp_sock *tp, u32 seq, u32 end_seq) { return !before(end_seq, tp->rcv_wup) && !after(seq, tp->rcv_nxt + tcp_receive_window(tp)); @@ -3229,9 +2850,9 @@ static void tcp_reset(struct sock *sk) */ static void tcp_fin(struct sk_buff *skb, struct sock *sk, struct tcphdr *th) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); - tcp_schedule_ack(tp); + inet_csk_schedule_ack(sk); sk->sk_shutdown |= RCV_SHUTDOWN; sock_set_flag(sk, SOCK_DONE); @@ -3241,7 +2862,7 @@ static void tcp_fin(struct sk_buff *skb, struct sock *sk, struct tcphdr *th) case TCP_ESTABLISHED: /* Move to CLOSE_WAIT */ tcp_set_state(sk, TCP_CLOSE_WAIT); - tp->ack.pingpong = 1; + inet_csk(sk)->icsk_ack.pingpong = 1; break; case TCP_CLOSE_WAIT: @@ -3280,8 +2901,8 @@ static void tcp_fin(struct sk_buff *skb, struct sock *sk, struct tcphdr *th) * Probably, we should reset in this case. For now drop them. */ __skb_queue_purge(&tp->out_of_order_queue); - if (tp->sack_ok) - tcp_sack_reset(tp); + if (tp->rx_opt.sack_ok) + tcp_sack_reset(&tp->rx_opt); sk_stream_mem_reclaim(sk); if (!sock_flag(sk, SOCK_DEAD)) { @@ -3296,8 +2917,7 @@ static void tcp_fin(struct sk_buff *skb, struct sock *sk, struct tcphdr *th) } } -static __inline__ int -tcp_sack_extend(struct tcp_sack_block *sp, u32 seq, u32 end_seq) +static inline int tcp_sack_extend(struct tcp_sack_block *sp, u32 seq, u32 end_seq) { if (!after(seq, sp->end_seq) && !after(sp->start_seq, end_seq)) { if (before(seq, sp->start_seq)) @@ -3309,24 +2929,24 @@ tcp_sack_extend(struct tcp_sack_block *sp, u32 seq, u32 end_seq) return 0; } -static __inline__ void tcp_dsack_set(struct tcp_opt *tp, u32 seq, u32 end_seq) +static void tcp_dsack_set(struct tcp_sock *tp, u32 seq, u32 end_seq) { - if (tp->sack_ok && sysctl_tcp_dsack) { + if (tp->rx_opt.sack_ok && sysctl_tcp_dsack) { if (before(seq, tp->rcv_nxt)) NET_INC_STATS_BH(LINUX_MIB_TCPDSACKOLDSENT); else NET_INC_STATS_BH(LINUX_MIB_TCPDSACKOFOSENT); - tp->dsack = 1; + tp->rx_opt.dsack = 1; tp->duplicate_sack[0].start_seq = seq; tp->duplicate_sack[0].end_seq = end_seq; - tp->eff_sacks = min(tp->num_sacks+1, 4-tp->tstamp_ok); + tp->rx_opt.eff_sacks = min(tp->rx_opt.num_sacks + 1, 4 - tp->rx_opt.tstamp_ok); } } -static __inline__ void tcp_dsack_extend(struct tcp_opt *tp, u32 seq, u32 end_seq) +static void tcp_dsack_extend(struct tcp_sock *tp, u32 seq, u32 end_seq) { - if (!tp->dsack) + if (!tp->rx_opt.dsack) tcp_dsack_set(tp, seq, end_seq); else tcp_sack_extend(tp->duplicate_sack, seq, end_seq); @@ -3334,14 +2954,14 @@ static __inline__ void tcp_dsack_extend(struct tcp_opt *tp, u32 seq, u32 end_seq static void tcp_send_dupack(struct sock *sk, struct sk_buff *skb) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); if (TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(skb)->seq && before(TCP_SKB_CB(skb)->seq, tp->rcv_nxt)) { NET_INC_STATS_BH(LINUX_MIB_DELAYEDACKLOST); - tcp_enter_quickack_mode(tp); + tcp_enter_quickack_mode(sk); - if (tp->sack_ok && sysctl_tcp_dsack) { + if (tp->rx_opt.sack_ok && sysctl_tcp_dsack) { u32 end_seq = TCP_SKB_CB(skb)->end_seq; if (after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) @@ -3356,7 +2976,7 @@ static void tcp_send_dupack(struct sock *sk, struct sk_buff *skb) /* These routines update the SACK block as out-of-order packets arrive or * in-order packets close up the sequence space. */ -static void tcp_sack_maybe_coalesce(struct tcp_opt *tp) +static void tcp_sack_maybe_coalesce(struct tcp_sock *tp) { int this_sack; struct tcp_sack_block *sp = &tp->selective_acks[0]; @@ -3365,16 +2985,16 @@ static void tcp_sack_maybe_coalesce(struct tcp_opt *tp) /* See if the recent change to the first SACK eats into * or hits the sequence space of other SACK blocks, if so coalesce. */ - for (this_sack = 1; this_sack < tp->num_sacks; ) { + for (this_sack = 1; this_sack < tp->rx_opt.num_sacks; ) { if (tcp_sack_extend(sp, swalk->start_seq, swalk->end_seq)) { int i; /* Zap SWALK, by moving every further SACK up by one slot. * Decrease num_sacks. */ - tp->num_sacks--; - tp->eff_sacks = min(tp->num_sacks+tp->dsack, 4-tp->tstamp_ok); - for(i=this_sack; i < tp->num_sacks; i++) + tp->rx_opt.num_sacks--; + tp->rx_opt.eff_sacks = min(tp->rx_opt.num_sacks + tp->rx_opt.dsack, 4 - tp->rx_opt.tstamp_ok); + for(i=this_sack; i < tp->rx_opt.num_sacks; i++) sp[i] = sp[i+1]; continue; } @@ -3382,7 +3002,7 @@ static void tcp_sack_maybe_coalesce(struct tcp_opt *tp) } } -static __inline__ void tcp_sack_swap(struct tcp_sack_block *sack1, struct tcp_sack_block *sack2) +static inline void tcp_sack_swap(struct tcp_sack_block *sack1, struct tcp_sack_block *sack2) { __u32 tmp; @@ -3397,9 +3017,9 @@ static __inline__ void tcp_sack_swap(struct tcp_sack_block *sack1, struct tcp_sa static void tcp_sack_new_ofo_skb(struct sock *sk, u32 seq, u32 end_seq) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); struct tcp_sack_block *sp = &tp->selective_acks[0]; - int cur_sacks = tp->num_sacks; + int cur_sacks = tp->rx_opt.num_sacks; int this_sack; if (!cur_sacks) @@ -3424,7 +3044,7 @@ static void tcp_sack_new_ofo_skb(struct sock *sk, u32 seq, u32 end_seq) */ if (this_sack >= 4) { this_sack--; - tp->num_sacks--; + tp->rx_opt.num_sacks--; sp--; } for(; this_sack > 0; this_sack--, sp--) @@ -3434,22 +3054,22 @@ new_sack: /* Build the new head SACK, and we're done. */ sp->start_seq = seq; sp->end_seq = end_seq; - tp->num_sacks++; - tp->eff_sacks = min(tp->num_sacks + tp->dsack, 4 - tp->tstamp_ok); + tp->rx_opt.num_sacks++; + tp->rx_opt.eff_sacks = min(tp->rx_opt.num_sacks + tp->rx_opt.dsack, 4 - tp->rx_opt.tstamp_ok); } /* RCV.NXT advances, some SACKs should be eaten. */ -static void tcp_sack_remove(struct tcp_opt *tp) +static void tcp_sack_remove(struct tcp_sock *tp) { struct tcp_sack_block *sp = &tp->selective_acks[0]; - int num_sacks = tp->num_sacks; + int num_sacks = tp->rx_opt.num_sacks; int this_sack; /* Empty ofo queue, hence, all the SACKs are eaten. Clear. */ - if (skb_queue_len(&tp->out_of_order_queue) == 0) { - tp->num_sacks = 0; - tp->eff_sacks = tp->dsack; + if (skb_queue_empty(&tp->out_of_order_queue)) { + tp->rx_opt.num_sacks = 0; + tp->rx_opt.eff_sacks = tp->rx_opt.dsack; return; } @@ -3470,9 +3090,9 @@ static void tcp_sack_remove(struct tcp_opt *tp) this_sack++; sp++; } - if (num_sacks != tp->num_sacks) { - tp->num_sacks = num_sacks; - tp->eff_sacks = min(tp->num_sacks+tp->dsack, 4-tp->tstamp_ok); + if (num_sacks != tp->rx_opt.num_sacks) { + tp->rx_opt.num_sacks = num_sacks; + tp->rx_opt.eff_sacks = min(tp->rx_opt.num_sacks + tp->rx_opt.dsack, 4 - tp->rx_opt.tstamp_ok); } } @@ -3481,7 +3101,7 @@ static void tcp_sack_remove(struct tcp_opt *tp) */ static void tcp_ofo_queue(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); __u32 dsack_high = tp->rcv_nxt; struct sk_buff *skb; @@ -3498,7 +3118,7 @@ static void tcp_ofo_queue(struct sock *sk) if (!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) { SOCK_DEBUG(sk, "ofo packet was already received \n"); - __skb_unlink(skb, skb->list); + __skb_unlink(skb, &tp->out_of_order_queue); __kfree_skb(skb); continue; } @@ -3506,7 +3126,7 @@ static void tcp_ofo_queue(struct sock *sk) tp->rcv_nxt, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq); - __skb_unlink(skb, skb->list); + __skb_unlink(skb, &tp->out_of_order_queue); __skb_queue_tail(&sk->sk_receive_queue, skb); tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq; if(skb->h.th->fin) @@ -3519,21 +3139,20 @@ static int tcp_prune_queue(struct sock *sk); static void tcp_data_queue(struct sock *sk, struct sk_buff *skb) { struct tcphdr *th = skb->h.th; - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); int eaten = -1; if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) goto drop; - th = skb->h.th; __skb_pull(skb, th->doff*4); TCP_ECN_accept_cwr(tp, skb); - if (tp->dsack) { - tp->dsack = 0; - tp->eff_sacks = min_t(unsigned int, tp->num_sacks, - 4 - tp->tstamp_ok); + if (tp->rx_opt.dsack) { + tp->rx_opt.dsack = 0; + tp->rx_opt.eff_sacks = min_t(unsigned int, tp->rx_opt.num_sacks, + 4 - tp->rx_opt.tstamp_ok); } /* Queue data for delivery to the user. @@ -3581,17 +3200,17 @@ queue_and_out: if(th->fin) tcp_fin(skb, sk, th); - if (skb_queue_len(&tp->out_of_order_queue)) { + if (!skb_queue_empty(&tp->out_of_order_queue)) { tcp_ofo_queue(sk); /* RFC2581. 4.2. SHOULD send immediate ACK, when * gap in queue is filled. */ - if (!skb_queue_len(&tp->out_of_order_queue)) - tp->ack.pingpong = 0; + if (skb_queue_empty(&tp->out_of_order_queue)) + inet_csk(sk)->icsk_ack.pingpong = 0; } - if (tp->num_sacks) + if (tp->rx_opt.num_sacks) tcp_sack_remove(tp); tcp_fast_path_check(sk, tp); @@ -3609,8 +3228,8 @@ queue_and_out: tcp_dsack_set(tp, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq); out_of_window: - tcp_enter_quickack_mode(tp); - tcp_schedule_ack(tp); + tcp_enter_quickack_mode(sk); + inet_csk_schedule_ack(sk); drop: __kfree_skb(skb); return; @@ -3620,7 +3239,7 @@ drop: if (!before(TCP_SKB_CB(skb)->seq, tp->rcv_nxt + tcp_receive_window(tp))) goto out_of_window; - tcp_enter_quickack_mode(tp); + tcp_enter_quickack_mode(sk); if (before(TCP_SKB_CB(skb)->seq, tp->rcv_nxt)) { /* Partial packet, seq < rcv_next < end_seq */ @@ -3649,7 +3268,7 @@ drop: /* Disable header prediction. */ tp->pred_flags = 0; - tcp_schedule_ack(tp); + inet_csk_schedule_ack(sk); SOCK_DEBUG(sk, "out of order segment: rcv_next %X seq %X - %X\n", tp->rcv_nxt, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq); @@ -3658,10 +3277,10 @@ drop: if (!skb_peek(&tp->out_of_order_queue)) { /* Initial out of order segment, build 1 SACK. */ - if (tp->sack_ok) { - tp->num_sacks = 1; - tp->dsack = 0; - tp->eff_sacks = 1; + if (tp->rx_opt.sack_ok) { + tp->rx_opt.num_sacks = 1; + tp->rx_opt.dsack = 0; + tp->rx_opt.eff_sacks = 1; tp->selective_acks[0].start_seq = TCP_SKB_CB(skb)->seq; tp->selective_acks[0].end_seq = TCP_SKB_CB(skb)->end_seq; @@ -3673,9 +3292,9 @@ drop: u32 end_seq = TCP_SKB_CB(skb)->end_seq; if (seq == TCP_SKB_CB(skb1)->end_seq) { - __skb_append(skb1, skb); + __skb_append(skb1, skb, &tp->out_of_order_queue); - if (!tp->num_sacks || + if (!tp->rx_opt.num_sacks || tp->selective_acks[0].end_seq != seq) goto add_sack; @@ -3717,13 +3336,13 @@ drop: tcp_dsack_extend(tp, TCP_SKB_CB(skb1)->seq, end_seq); break; } - __skb_unlink(skb1, skb1->list); + __skb_unlink(skb1, &tp->out_of_order_queue); tcp_dsack_extend(tp, TCP_SKB_CB(skb1)->seq, TCP_SKB_CB(skb1)->end_seq); __kfree_skb(skb1); } add_sack: - if (tp->sack_ok) + if (tp->rx_opt.sack_ok) tcp_sack_new_ofo_skb(sk, seq, end_seq); } } @@ -3734,18 +3353,19 @@ add_sack: * simplifies code) */ static void -tcp_collapse(struct sock *sk, struct sk_buff *head, - struct sk_buff *tail, u32 start, u32 end) +tcp_collapse(struct sock *sk, struct sk_buff_head *list, + struct sk_buff *head, struct sk_buff *tail, + u32 start, u32 end) { struct sk_buff *skb; - /* First, check that queue is collapsable and find + /* First, check that queue is collapsible and find * the point where collapsing can be useful. */ for (skb = head; skb != tail; ) { /* No new bits? It is possible on ofo queue. */ if (!before(start, TCP_SKB_CB(skb)->end_seq)) { struct sk_buff *next = skb->next; - __skb_unlink(skb, skb->list); + __skb_unlink(skb, list); __kfree_skb(skb); NET_INC_STATS_BH(LINUX_MIB_TCPRCVCOLLAPSED); skb = next; @@ -3774,8 +3394,7 @@ tcp_collapse(struct sock *sk, struct sk_buff *head, while (before(start, end)) { struct sk_buff *nskb; int header = skb_headroom(skb); - int copy = (PAGE_SIZE - sizeof(struct sk_buff) - - sizeof(struct skb_shared_info) - header - 31)&~15; + int copy = SKB_MAX_ORDER(header, 0); /* Too big header? This can happen with IPv6. */ if (copy < 0) @@ -3792,7 +3411,7 @@ tcp_collapse(struct sock *sk, struct sk_buff *head, nskb->mac.raw = nskb->head + (skb->mac.raw-skb->head); memcpy(nskb->cb, skb->cb, sizeof(skb->cb)); TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(nskb)->end_seq = start; - __skb_insert(nskb, skb->prev, skb, skb->list); + __skb_insert(nskb, skb->prev, skb, list); sk_stream_set_owner_r(nskb, sk); /* Copy data, releasing collapsed skbs. */ @@ -3800,7 +3419,7 @@ tcp_collapse(struct sock *sk, struct sk_buff *head, int offset = start - TCP_SKB_CB(skb)->seq; int size = TCP_SKB_CB(skb)->end_seq - start; - if (offset < 0) BUG(); + BUG_ON(offset < 0); if (size > 0) { size = min(copy, size); if (skb_copy_bits(skb, offset, skb_put(nskb, size), size)) @@ -3811,7 +3430,7 @@ tcp_collapse(struct sock *sk, struct sk_buff *head, } if (!before(start, TCP_SKB_CB(skb)->end_seq)) { struct sk_buff *next = skb->next; - __skb_unlink(skb, skb->list); + __skb_unlink(skb, list); __kfree_skb(skb); NET_INC_STATS_BH(LINUX_MIB_TCPRCVCOLLAPSED); skb = next; @@ -3827,7 +3446,7 @@ tcp_collapse(struct sock *sk, struct sk_buff *head, */ static void tcp_collapse_ofo_queue(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *skb = skb_peek(&tp->out_of_order_queue); struct sk_buff *head; u32 start, end; @@ -3847,7 +3466,8 @@ static void tcp_collapse_ofo_queue(struct sock *sk) if (skb == (struct sk_buff *)&tp->out_of_order_queue || after(TCP_SKB_CB(skb)->seq, end) || before(TCP_SKB_CB(skb)->end_seq, start)) { - tcp_collapse(sk, head, skb, start, end); + tcp_collapse(sk, &tp->out_of_order_queue, + head, skb, start, end); head = skb; if (skb == (struct sk_buff *)&tp->out_of_order_queue) break; @@ -3872,7 +3492,7 @@ static void tcp_collapse_ofo_queue(struct sock *sk) */ static int tcp_prune_queue(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); SOCK_DEBUG(sk, "prune_queue: c=%x\n", tp->copied_seq); @@ -3884,7 +3504,8 @@ static int tcp_prune_queue(struct sock *sk) tp->rcv_ssthresh = min(tp->rcv_ssthresh, 4U * tp->advmss); tcp_collapse_ofo_queue(sk); - tcp_collapse(sk, sk->sk_receive_queue.next, + tcp_collapse(sk, &sk->sk_receive_queue, + sk->sk_receive_queue.next, (struct sk_buff*)&sk->sk_receive_queue, tp->copied_seq, tp->rcv_nxt); sk_stream_mem_reclaim(sk); @@ -3896,9 +3517,8 @@ static int tcp_prune_queue(struct sock *sk) * This must not ever occur. */ /* First, purge the out_of_order queue. */ - if (skb_queue_len(&tp->out_of_order_queue)) { - NET_ADD_STATS_BH(LINUX_MIB_OFOPRUNED, - skb_queue_len(&tp->out_of_order_queue)); + if (!skb_queue_empty(&tp->out_of_order_queue)) { + NET_INC_STATS_BH(LINUX_MIB_OFOPRUNED); __skb_queue_purge(&tp->out_of_order_queue); /* Reset SACK state. A conforming SACK implementation will @@ -3906,8 +3526,8 @@ static int tcp_prune_queue(struct sock *sk) * is in a sad state like this, we care only about integrity * of the connection not performance. */ - if (tp->sack_ok) - tcp_sack_reset(tp); + if (tp->rx_opt.sack_ok) + tcp_sack_reset(&tp->rx_opt); sk_stream_mem_reclaim(sk); } @@ -3932,14 +3552,15 @@ static int tcp_prune_queue(struct sock *sk) */ void tcp_cwnd_application_limited(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); - if (tp->ca_state == TCP_CA_Open && + if (inet_csk(sk)->icsk_ca_state == TCP_CA_Open && sk->sk_socket && !test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { /* Limited by application or receiver window. */ - u32 win_used = max(tp->snd_cwnd_used, 2U); + u32 init_win = tcp_init_cwnd(tp, __sk_dst_get(sk)); + u32 win_used = max(tp->snd_cwnd_used, init_win); if (win_used < tp->snd_cwnd) { - tp->snd_ssthresh = tcp_current_ssthresh(tp); + tp->snd_ssthresh = tcp_current_ssthresh(sk); tp->snd_cwnd = (tp->snd_cwnd + win_used) >> 1; } tp->snd_cwnd_used = 0; @@ -3947,22 +3568,41 @@ void tcp_cwnd_application_limited(struct sock *sk) tp->snd_cwnd_stamp = tcp_time_stamp; } +static int tcp_should_expand_sndbuf(struct sock *sk, struct tcp_sock *tp) +{ + /* If the user specified a specific send buffer setting, do + * not modify it. + */ + if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) + return 0; + + /* If we are under global TCP memory pressure, do not expand. */ + if (tcp_memory_pressure) + return 0; + + /* If we are under soft global TCP memory pressure, do not expand. */ + if (atomic_read(&tcp_memory_allocated) >= sysctl_tcp_mem[0]) + return 0; + + /* If we filled the congestion window, do not expand. */ + if (tp->packets_out >= tp->snd_cwnd) + return 0; + + return 1; +} /* When incoming ACK allowed to free some skb from write_queue, - * we remember this event in flag sk->sk_queue_shrunk and wake up socket + * we remember this event in flag SOCK_QUEUE_SHRUNK and wake up socket * on the exit from tcp input handler. * * PROBLEM: sndbuf expansion does not work well with largesend. */ static void tcp_new_space(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); - if (tcp_get_pcount(&tp->packets_out) < tp->snd_cwnd && - !(sk->sk_userlocks & SOCK_SNDBUF_LOCK) && - !tcp_memory_pressure && - atomic_read(&tcp_memory_allocated) < sysctl_tcp_mem[0]) { - int sndmem = max_t(u32, tp->mss_clamp, tp->mss_cache_std) + + if (tcp_should_expand_sndbuf(sk, tp)) { + int sndmem = max_t(u32, tp->rx_opt.mss_clamp, tp->mss_cache) + MAX_TCP_HEADER + 16 + sizeof(struct sk_buff), demanded = max_t(unsigned int, tp->snd_cwnd, tp->reordering + 1); @@ -3975,32 +3615,19 @@ static void tcp_new_space(struct sock *sk) sk->sk_write_space(sk); } -static inline void tcp_check_space(struct sock *sk) +static void tcp_check_space(struct sock *sk) { - if (sk->sk_queue_shrunk) { - sk->sk_queue_shrunk = 0; + if (sock_flag(sk, SOCK_QUEUE_SHRUNK)) { + sock_reset_flag(sk, SOCK_QUEUE_SHRUNK); if (sk->sk_socket && test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) tcp_new_space(sk); } } -static void __tcp_data_snd_check(struct sock *sk, struct sk_buff *skb) -{ - struct tcp_opt *tp = tcp_sk(sk); - - if (after(TCP_SKB_CB(skb)->end_seq, tp->snd_una + tp->snd_wnd) || - tcp_packets_in_flight(tp) >= tp->snd_cwnd || - tcp_write_xmit(sk, tp->nonagle)) - tcp_check_probe_timer(sk, tp); -} - -static __inline__ void tcp_data_snd_check(struct sock *sk) +static inline void tcp_data_snd_check(struct sock *sk, struct tcp_sock *tp) { - struct sk_buff *skb = sk->sk_send_head; - - if (skb != NULL) - __tcp_data_snd_check(sk, skb); + tcp_push_pending_frames(sk, tp); tcp_check_space(sk); } @@ -4009,16 +3636,16 @@ static __inline__ void tcp_data_snd_check(struct sock *sk) */ static void __tcp_ack_snd_check(struct sock *sk, int ofo_possible) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); /* More than one full frame received... */ - if (((tp->rcv_nxt - tp->rcv_wup) > tp->ack.rcv_mss + if (((tp->rcv_nxt - tp->rcv_wup) > inet_csk(sk)->icsk_ack.rcv_mss /* ... and right edge of window advances far enough. * (tcp_recvmsg() will send ACK otherwise). Or... */ && __tcp_select_window(sk) >= tp->rcv_wnd) || /* We ACK each frame or... */ - tcp_in_quickack_mode(tp) || + tcp_in_quickack_mode(sk) || /* We have out of order data. */ (ofo_possible && skb_peek(&tp->out_of_order_queue))) { @@ -4030,10 +3657,9 @@ static void __tcp_ack_snd_check(struct sock *sk, int ofo_possible) } } -static __inline__ void tcp_ack_snd_check(struct sock *sk) +static inline void tcp_ack_snd_check(struct sock *sk) { - struct tcp_opt *tp = tcp_sk(sk); - if (!tcp_ack_scheduled(tp)) { + if (!inet_csk_ack_scheduled(sk)) { /* We sent a data segment already. */ return; } @@ -4042,7 +3668,7 @@ static __inline__ void tcp_ack_snd_check(struct sock *sk) /* * This routine is only called when we have urgent data - * signalled. Its the 'slow' part of tcp_urg. It could be + * signaled. Its the 'slow' part of tcp_urg. It could be * moved inline now as tcp_urg is only called from one * place. We handle URGent data wrong. We have to - as * BSD still doesn't use the correction from RFC961. @@ -4052,7 +3678,7 @@ static __inline__ void tcp_ack_snd_check(struct sock *sk) static void tcp_check_urg(struct sock * sk, struct tcphdr * th) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); u32 ptr = ntohs(th->urg_ptr); if (ptr && !sysctl_tcp_stdurg) @@ -4087,7 +3713,7 @@ static void tcp_check_urg(struct sock * sk, struct tcphdr * th) * urgent. To do this requires some care. We cannot just ignore * tp->copied_seq since we would read the last urgent byte again * as data, nor can we alter copied_seq until this data arrives - * or we break the sematics of SIOCATMARK (and thus sockatmark()) + * or we break the semantics of SIOCATMARK (and thus sockatmark()) * * NOTE. Double Dutch. Rendering to plain English: author of comment * above did something sort of send("A", MSG_OOB); send("B", MSG_OOB); @@ -4104,7 +3730,7 @@ static void tcp_check_urg(struct sock * sk, struct tcphdr * th) struct sk_buff *skb = skb_peek(&sk->sk_receive_queue); tp->copied_seq++; if (skb && !before(tp->copied_seq, TCP_SKB_CB(skb)->end_seq)) { - __skb_unlink(skb, skb->list); + __skb_unlink(skb, &sk->sk_receive_queue); __kfree_skb(skb); } } @@ -4119,7 +3745,7 @@ static void tcp_check_urg(struct sock * sk, struct tcphdr * th) /* This is the 'fast' part of urgent handling. */ static void tcp_urg(struct sock *sk, struct sk_buff *skb, struct tcphdr *th) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); /* Check if we get a new urgent pointer - normally not. */ if (th->urg) @@ -4144,7 +3770,7 @@ static void tcp_urg(struct sock *sk, struct sk_buff *skb, struct tcphdr *th) static int tcp_copy_to_iovec(struct sock *sk, struct sk_buff *skb, int hlen) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); int chunk = skb->len - hlen; int err; @@ -4165,9 +3791,9 @@ static int tcp_copy_to_iovec(struct sock *sk, struct sk_buff *skb, int hlen) return err; } -static int __tcp_checksum_complete_user(struct sock *sk, struct sk_buff *skb) +static __sum16 __tcp_checksum_complete_user(struct sock *sk, struct sk_buff *skb) { - int result; + __sum16 result; if (sock_owned_by_user(sk)) { local_bh_enable(); @@ -4179,13 +3805,56 @@ static int __tcp_checksum_complete_user(struct sock *sk, struct sk_buff *skb) return result; } -static __inline__ int -tcp_checksum_complete_user(struct sock *sk, struct sk_buff *skb) +static inline int tcp_checksum_complete_user(struct sock *sk, struct sk_buff *skb) { return skb->ip_summed != CHECKSUM_UNNECESSARY && __tcp_checksum_complete_user(sk, skb); } +#ifdef CONFIG_NET_DMA +static int tcp_dma_try_early_copy(struct sock *sk, struct sk_buff *skb, int hlen) +{ + struct tcp_sock *tp = tcp_sk(sk); + int chunk = skb->len - hlen; + int dma_cookie; + int copied_early = 0; + + if (tp->ucopy.wakeup) + return 0; + + if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list) + tp->ucopy.dma_chan = get_softnet_dma(); + + if (tp->ucopy.dma_chan && skb->ip_summed == CHECKSUM_UNNECESSARY) { + + dma_cookie = dma_skb_copy_datagram_iovec(tp->ucopy.dma_chan, + skb, hlen, tp->ucopy.iov, chunk, tp->ucopy.pinned_list); + + if (dma_cookie < 0) + goto out; + + tp->ucopy.dma_cookie = dma_cookie; + copied_early = 1; + + tp->ucopy.len -= chunk; + tp->copied_seq += chunk; + tcp_rcv_space_adjust(sk); + + if ((tp->ucopy.len == 0) || + (tcp_flag_word(skb->h.th) & TCP_FLAG_PSH) || + (atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1))) { + tp->ucopy.wakeup = 1; + sk->sk_data_ready(sk, 0); + } + } else if (chunk > 0) { + tp->ucopy.wakeup = 1; + sk->sk_data_ready(sk, 0); + } +out: + return copied_early; +} +#endif /* CONFIG_NET_DMA */ + /* * TCP receive function for the ESTABLISHED state. * @@ -4212,7 +3881,7 @@ tcp_checksum_complete_user(struct sock *sk, struct sk_buff *skb) int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, struct tcphdr *th, unsigned len) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); /* * Header prediction. @@ -4229,10 +3898,10 @@ int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, * We do checksum and copy also but from device to kernel. */ - tp->saw_tstamp = 0; + tp->rx_opt.saw_tstamp = 0; /* pred_flags is 0xS?10 << 16 + snd_wnd - * if header_predition is to be made + * if header_prediction is to be made * 'S' will always be tp->tcp_header_len >> 2 * '?' will be 0 for the fast path, otherwise pred_flags is 0 to * turn it off (when there are holes in the receive @@ -4251,21 +3920,21 @@ int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, /* Check timestamp */ if (tcp_header_len == sizeof(struct tcphdr) + TCPOLEN_TSTAMP_ALIGNED) { - __u32 *ptr = (__u32 *)(th + 1); + __be32 *ptr = (__be32 *)(th + 1); /* No? Slow path! */ - if (*ptr != ntohl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) + if (*ptr != htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP)) goto slow_path; - tp->saw_tstamp = 1; + tp->rx_opt.saw_tstamp = 1; ++ptr; - tp->rcv_tsval = ntohl(*ptr); + tp->rx_opt.rcv_tsval = ntohl(*ptr); ++ptr; - tp->rcv_tsecr = ntohl(*ptr); + tp->rx_opt.rcv_tsecr = ntohl(*ptr); /* If PAWS failed, check it more carefully in slow path */ - if ((s32)(tp->rcv_tsval - tp->ts_recent) < 0) + if ((s32)(tp->rx_opt.rcv_tsval - tp->rx_opt.ts_recent) < 0) goto slow_path; /* DO NOT update ts_recent here, if checksum fails @@ -4287,14 +3956,12 @@ int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, tp->rcv_nxt == tp->rcv_wup) tcp_store_ts_recent(tp); - tcp_rcv_rtt_measure_ts(tp, skb); - /* We know that such packets are checksummed * on entry. */ tcp_ack(sk, skb, 0); __kfree_skb(skb); - tcp_data_snd_check(sk); + tcp_data_snd_check(sk, tp); return 0; } else { /* Header too small */ TCP_INC_STATS_BH(TCP_MIB_INERRS); @@ -4302,14 +3969,23 @@ int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, } } else { int eaten = 0; + int copied_early = 0; - if (tp->ucopy.task == current && - tp->copied_seq == tp->rcv_nxt && - len - tcp_header_len <= tp->ucopy.len && - sock_owned_by_user(sk)) { - __set_current_state(TASK_RUNNING); + if (tp->copied_seq == tp->rcv_nxt && + len - tcp_header_len <= tp->ucopy.len) { +#ifdef CONFIG_NET_DMA + if (tcp_dma_try_early_copy(sk, skb, tcp_header_len)) { + copied_early = 1; + eaten = 1; + } +#endif + if (tp->ucopy.task == current && sock_owned_by_user(sk) && !copied_early) { + __set_current_state(TASK_RUNNING); - if (!tcp_copy_to_iovec(sk, skb, tcp_header_len)) { + if (!tcp_copy_to_iovec(sk, skb, tcp_header_len)) + eaten = 1; + } + if (eaten) { /* Predicted packet is in window by definition. * seq == rcv_nxt and rcv_wup <= rcv_nxt. * Hence, check seq<=rcv_wup reduces to: @@ -4320,13 +3996,14 @@ int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, tp->rcv_nxt == tp->rcv_wup) tcp_store_ts_recent(tp); - tcp_rcv_rtt_measure_ts(tp, skb); + tcp_rcv_rtt_measure_ts(sk, skb); __skb_pull(skb, tcp_header_len); tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq; NET_INC_STATS_BH(LINUX_MIB_TCPHPHITSTOUSER); - eaten = 1; } + if (copied_early) + tcp_cleanup_rbuf(sk, skb->len); } if (!eaten) { if (tcp_checksum_complete_user(sk, skb)) @@ -4341,7 +4018,7 @@ int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, tp->rcv_nxt == tp->rcv_wup) tcp_store_ts_recent(tp); - tcp_rcv_rtt_measure_ts(tp, skb); + tcp_rcv_rtt_measure_ts(sk, skb); if ((int)skb->truesize > sk->sk_forward_alloc) goto step5; @@ -4360,22 +4037,18 @@ int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, if (TCP_SKB_CB(skb)->ack_seq != tp->snd_una) { /* Well, only one small jumplet in fast path... */ tcp_ack(sk, skb, FLAG_DATA); - tcp_data_snd_check(sk); - if (!tcp_ack_scheduled(tp)) + tcp_data_snd_check(sk, tp); + if (!inet_csk_ack_scheduled(sk)) goto no_ack; } - if (eaten) { - if (tcp_in_quickack_mode(tp)) { - tcp_send_ack(sk); - } else { - tcp_send_delayed_ack(sk); - } - } else { - __tcp_ack_snd_check(sk, 0); - } - + __tcp_ack_snd_check(sk, 0); no_ack: +#ifdef CONFIG_NET_DMA + if (copied_early) + __skb_queue_tail(&sk->sk_async_wait_queue, skb); + else +#endif if (eaten) __kfree_skb(skb); else @@ -4391,8 +4064,8 @@ slow_path: /* * RFC1323: H1. Apply PAWS check first. */ - if (tcp_fast_parse_options(skb, th, tp) && tp->saw_tstamp && - tcp_paws_discard(tp, skb)) { + if (tcp_fast_parse_options(skb, th, tp) && tp->rx_opt.saw_tstamp && + tcp_paws_discard(sk, skb)) { if (!th->rst) { NET_INC_STATS_BH(LINUX_MIB_PAWSESTABREJECTED); tcp_send_dupack(sk, skb); @@ -4439,7 +4112,7 @@ step5: if(th->ack) tcp_ack(sk, skb, FLAG_SLOWPATH); - tcp_rcv_rtt_measure_ts(tp, skb); + tcp_rcv_rtt_measure_ts(sk, skb); /* Process urgent data. */ tcp_urg(sk, skb, th); @@ -4447,7 +4120,7 @@ step5: /* step 7: process the segment text */ tcp_data_queue(sk, skb); - tcp_data_snd_check(sk); + tcp_data_snd_check(sk, tp); tcp_ack_snd_check(sk); return 0; @@ -4462,10 +4135,11 @@ discard: static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, struct tcphdr *th, unsigned len) { - struct tcp_opt *tp = tcp_sk(sk); - int saved_clamp = tp->mss_clamp; + struct tcp_sock *tp = tcp_sk(sk); + struct inet_connection_sock *icsk = inet_csk(sk); + int saved_clamp = tp->rx_opt.mss_clamp; - tcp_parse_options(skb, tp, 0); + tcp_parse_options(skb, &tp->rx_opt, 0); if (th->ack) { /* rfc793: @@ -4482,8 +4156,8 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, if (TCP_SKB_CB(skb)->ack_seq != tp->snd_nxt) goto reset_and_undo; - if (tp->saw_tstamp && tp->rcv_tsecr && - !between(tp->rcv_tsecr, tp->retrans_stamp, + if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr && + !between(tp->rx_opt.rcv_tsecr, tp->retrans_stamp, tcp_time_stamp)) { NET_INC_STATS_BH(LINUX_MIB_PAWSACTIVEREJECTED); goto reset_and_undo; @@ -4520,8 +4194,6 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, */ TCP_ECN_rcv_synack(tp, th); - if (tp->ecn_flags&TCP_ECN_OK) - sk->sk_no_largesend = 1; tp->snd_wl1 = TCP_SKB_CB(skb)->seq; tcp_ack(sk, skb, FLAG_SLOWPATH); @@ -4538,13 +4210,13 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, tp->snd_wnd = ntohs(th->window); tcp_init_wl(tp, TCP_SKB_CB(skb)->ack_seq, TCP_SKB_CB(skb)->seq); - if (!tp->wscale_ok) { - tp->snd_wscale = tp->rcv_wscale = 0; + if (!tp->rx_opt.wscale_ok) { + tp->rx_opt.snd_wscale = tp->rx_opt.rcv_wscale = 0; tp->window_clamp = min(tp->window_clamp, 65535U); } - if (tp->saw_tstamp) { - tp->tstamp_ok = 1; + if (tp->rx_opt.saw_tstamp) { + tp->rx_opt.tstamp_ok = 1; tp->tcp_header_len = sizeof(struct tcphdr) + TCPOLEN_TSTAMP_ALIGNED; tp->advmss -= TCPOLEN_TSTAMP_ALIGNED; @@ -4553,24 +4225,29 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, tp->tcp_header_len = sizeof(struct tcphdr); } - if (tp->sack_ok && sysctl_tcp_fack) - tp->sack_ok |= 2; + if (tp->rx_opt.sack_ok && sysctl_tcp_fack) + tp->rx_opt.sack_ok |= 2; - tcp_sync_mss(sk, tp->pmtu_cookie); + tcp_mtup_init(sk); + tcp_sync_mss(sk, icsk->icsk_pmtu_cookie); tcp_initialize_rcv_mss(sk); /* Remember, tcp_poll() does not lock socket! * Change state from SYN-SENT only after copied_seq * is initialized. */ tp->copied_seq = tp->rcv_nxt; - mb(); + smp_mb(); tcp_set_state(sk, TCP_ESTABLISHED); + security_inet_conn_established(sk, skb); + /* Make sure socket is routed, for correct metrics. */ - tp->af_specific->rebuild_header(sk); + icsk->icsk_af_ops->rebuild_header(sk); tcp_init_metrics(sk); + tcp_init_congestion_control(sk); + /* Prevent spurious tcp_cwnd_restart() on first data * packet. */ @@ -4579,9 +4256,9 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, tcp_init_buffer_space(sk); if (sock_flag(sk, SOCK_KEEPOPEN)) - tcp_reset_keepalive_timer(sk, keepalive_time_when(tp)); + inet_csk_reset_keepalive_timer(sk, keepalive_time_when(tp)); - if (!tp->snd_wscale) + if (!tp->rx_opt.snd_wscale) __tcp_fast_path_on(tp, tp->snd_wnd); else tp->pred_flags = 0; @@ -4591,7 +4268,9 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, sk_wake_async(sk, 0, POLL_OUT); } - if (sk->sk_write_pending || tp->defer_accept || tp->ack.pingpong) { + if (sk->sk_write_pending || + icsk->icsk_accept_queue.rskq_defer_accept || + icsk->icsk_ack.pingpong) { /* Save one ACK. Data will be ready after * several ticks, if write_pending is set. * @@ -4599,12 +4278,13 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, * look so _wonderfully_ clever, that I was not able * to stand against the temptation 8) --ANK */ - tcp_schedule_ack(tp); - tp->ack.lrcvtime = tcp_time_stamp; - tp->ack.ato = TCP_ATO_MIN; - tcp_incr_quickack(tp); - tcp_enter_quickack_mode(tp); - tcp_reset_xmit_timer(sk, TCP_TIME_DACK, TCP_DELACK_MAX); + inet_csk_schedule_ack(sk); + icsk->icsk_ack.lrcvtime = tcp_time_stamp; + icsk->icsk_ack.ato = TCP_ATO_MIN; + tcp_incr_quickack(sk); + tcp_enter_quickack_mode(sk); + inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, + TCP_DELACK_MAX, TCP_RTO_MAX); discard: __kfree_skb(skb); @@ -4628,7 +4308,7 @@ discard: } /* PAWS check. */ - if (tp->ts_recent_stamp && tp->saw_tstamp && tcp_paws_check(tp, 0)) + if (tp->rx_opt.ts_recent_stamp && tp->rx_opt.saw_tstamp && tcp_paws_check(&tp->rx_opt, 0)) goto discard_and_undo; if (th->syn) { @@ -4638,8 +4318,8 @@ discard: */ tcp_set_state(sk, TCP_SYN_RECV); - if (tp->saw_tstamp) { - tp->tstamp_ok = 1; + if (tp->rx_opt.saw_tstamp) { + tp->rx_opt.tstamp_ok = 1; tcp_store_ts_recent(tp); tp->tcp_header_len = sizeof(struct tcphdr) + TCPOLEN_TSTAMP_ALIGNED; @@ -4658,10 +4338,9 @@ discard: tp->max_window = tp->snd_wnd; TCP_ECN_rcv_syn(tp, th); - if (tp->ecn_flags&TCP_ECN_OK) - sk->sk_no_largesend = 1; - tcp_sync_mss(sk, tp->pmtu_cookie); + tcp_mtup_init(sk); + tcp_sync_mss(sk, icsk->icsk_pmtu_cookie); tcp_initialize_rcv_mss(sk); @@ -4686,13 +4365,13 @@ discard: */ discard_and_undo: - tcp_clear_options(tp); - tp->mss_clamp = saved_clamp; + tcp_clear_options(&tp->rx_opt); + tp->rx_opt.mss_clamp = saved_clamp; goto discard; reset_and_undo: - tcp_clear_options(tp); - tp->mss_clamp = saved_clamp; + tcp_clear_options(&tp->rx_opt); + tp->rx_opt.mss_clamp = saved_clamp; return 1; } @@ -4707,10 +4386,11 @@ reset_and_undo: int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, struct tcphdr *th, unsigned len) { - struct tcp_opt *tp = tcp_sk(sk); + struct tcp_sock *tp = tcp_sk(sk); + struct inet_connection_sock *icsk = inet_csk(sk); int queued = 0; - tp->saw_tstamp = 0; + tp->rx_opt.saw_tstamp = 0; switch (sk->sk_state) { case TCP_CLOSE: @@ -4724,12 +4404,9 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, goto discard; if(th->syn) { - if(tp->af_specific->conn_request(sk, skb) < 0) + if (icsk->icsk_af_ops->conn_request(sk, skb) < 0) return 1; - init_westwood(sk); - init_bictcp(tp); - /* Now we have several options: In theory there is * nothing else in the frame. KA9Q has an option to * send data with the syn, BSD accepts data with the @@ -4744,16 +4421,15 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, * But, this leaves one open to an easy denial of * service attack, and SYN cookies can't defend * against this problem. So, we drop the data - * in the interest of security over speed. + * in the interest of security over speed unless + * it's still in use. */ - goto discard; + kfree_skb(skb); + return 0; } goto discard; case TCP_SYN_SENT: - init_westwood(sk); - init_bictcp(tp); - queued = tcp_rcv_synsent_state_process(sk, skb, th, len); if (queued >= 0) return queued; @@ -4761,12 +4437,12 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, /* Do step6 onward by hand. */ tcp_urg(sk, skb, th); __kfree_skb(skb); - tcp_data_snd_check(sk); + tcp_data_snd_check(sk, tp); return 0; } - if (tcp_fast_parse_options(skb, th, tp) && tp->saw_tstamp && - tcp_paws_discard(tp, skb)) { + if (tcp_fast_parse_options(skb, th, tp) && tp->rx_opt.saw_tstamp && + tcp_paws_discard(sk, skb)) { if (!th->rst) { NET_INC_STATS_BH(LINUX_MIB_PAWSESTABREJECTED); tcp_send_dupack(sk, skb); @@ -4810,7 +4486,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, case TCP_SYN_RECV: if (acceptable) { tp->copied_seq = tp->rcv_nxt; - mb(); + smp_mb(); tcp_set_state(sk, TCP_ESTABLISHED); sk->sk_state_change(sk); @@ -4825,7 +4501,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, tp->snd_una = TCP_SKB_CB(skb)->ack_seq; tp->snd_wnd = ntohs(th->window) << - tp->snd_wscale; + tp->rx_opt.snd_wscale; tcp_init_wl(tp, TCP_SKB_CB(skb)->ack_seq, TCP_SKB_CB(skb)->seq); @@ -4833,25 +4509,28 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, * and does not calculate rtt. * Fix it at least with timestamps. */ - if (tp->saw_tstamp && tp->rcv_tsecr && + if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr && !tp->srtt) - tcp_ack_saw_tstamp(tp, 0); + tcp_ack_saw_tstamp(sk, 0); - if (tp->tstamp_ok) + if (tp->rx_opt.tstamp_ok) tp->advmss -= TCPOLEN_TSTAMP_ALIGNED; /* Make sure socket is routed, for * correct metrics. */ - tp->af_specific->rebuild_header(sk); + icsk->icsk_af_ops->rebuild_header(sk); tcp_init_metrics(sk); + tcp_init_congestion_control(sk); + /* Prevent spurious tcp_cwnd_restart() on * first data packet. */ tp->lsndtime = tcp_time_stamp; + tcp_mtup_init(sk); tcp_initialize_rcv_mss(sk); tcp_init_buffer_space(sk); tcp_fast_path_on(tp); @@ -4880,9 +4559,9 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, return 1; } - tmo = tcp_fin_time(tp); + tmo = tcp_fin_time(sk); if (tmo > TCP_TIMEWAIT_LEN) { - tcp_reset_keepalive_timer(sk, tmo - TCP_TIMEWAIT_LEN); + inet_csk_reset_keepalive_timer(sk, tmo - TCP_TIMEWAIT_LEN); } else if (th->fin || sock_owned_by_user(sk)) { /* Bad case. We could lose such FIN otherwise. * It is not a big problem, but it looks confusing @@ -4890,7 +4569,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, * if it spins in bh_lock_sock(), but it is really * marginal case. */ - tcp_reset_keepalive_timer(sk, tmo); + inet_csk_reset_keepalive_timer(sk, tmo); } else { tcp_time_wait(sk, TCP_FIN_WAIT2, tmo); goto discard; @@ -4950,7 +4629,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, /* tcp_data could move socket to TIME-WAIT */ if (sk->sk_state != TCP_CLOSE) { - tcp_data_snd_check(sk); + tcp_data_snd_check(sk, tp); tcp_ack_snd_check(sk); } @@ -4963,7 +4642,7 @@ discard: EXPORT_SYMBOL(sysctl_tcp_ecn); EXPORT_SYMBOL(sysctl_tcp_reordering); -EXPORT_SYMBOL(tcp_cwnd_application_limited); EXPORT_SYMBOL(tcp_parse_options); EXPORT_SYMBOL(tcp_rcv_established); EXPORT_SYMBOL(tcp_rcv_state_process); +EXPORT_SYMBOL(tcp_initialize_rcv_mss);