diff options
Diffstat (limited to 'net/ipv4/tcp_input.c')
| -rw-r--r-- | net/ipv4/tcp_input.c | 597 |
1 files changed, 266 insertions, 331 deletions
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 887585045b27..dabbf1d392fb 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -76,25 +76,10 @@ #include <linux/ipsec.h> #include <asm/unaligned.h> #include <linux/errqueue.h> +#include <trace/events/tcp.h> +#include <linux/static_key.h> -int sysctl_tcp_fack __read_mostly; -int sysctl_tcp_max_reordering __read_mostly = 300; -int sysctl_tcp_dsack __read_mostly = 1; -int sysctl_tcp_app_win __read_mostly = 31; -int sysctl_tcp_adv_win_scale __read_mostly = 1; -EXPORT_SYMBOL(sysctl_tcp_adv_win_scale); - -/* rfc5961 challenge ack rate limiting */ -int sysctl_tcp_challenge_ack_limit = 1000; - -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 = 2; -int sysctl_tcp_min_rtt_wlen __read_mostly = 300; -int sysctl_tcp_moderate_rcvbuf __read_mostly = 1; -int sysctl_tcp_early_retrans __read_mostly = 3; -int sysctl_tcp_invalid_ratelimit __read_mostly = HZ/2; #define FLAG_DATA 0x01 /* Incoming frame contained data. */ #define FLAG_WIN_UPDATE 0x02 /* Incoming ACK was a window update. */ @@ -335,7 +320,7 @@ static void tcp_sndbuf_expand(struct sock *sk) sndmem *= nr_segs * per_mss; if (sk->sk_sndbuf < sndmem) - sk->sk_sndbuf = min(sndmem, sysctl_tcp_wmem[2]); + sk->sk_sndbuf = min(sndmem, sock_net(sk)->ipv4.sysctl_tcp_wmem[2]); } /* 2. Tuning advertised window (window_clamp, rcv_ssthresh) @@ -368,8 +353,8 @@ static int __tcp_grow_window(const struct sock *sk, const struct sk_buff *skb) { struct tcp_sock *tp = tcp_sk(sk); /* Optimize this! */ - int truesize = tcp_win_from_space(skb->truesize) >> 1; - int window = tcp_win_from_space(sysctl_tcp_rmem[2]) >> 1; + int truesize = tcp_win_from_space(sk, skb->truesize) >> 1; + int window = tcp_win_from_space(sk, sock_net(sk)->ipv4.sysctl_tcp_rmem[2]) >> 1; while (tp->rcv_ssthresh <= window) { if (truesize <= skb->len) @@ -394,7 +379,7 @@ static void tcp_grow_window(struct sock *sk, const struct sk_buff *skb) /* Check #2. Increase window, if skb with such overhead * will fit to rcvbuf in future. */ - if (tcp_win_from_space(skb->truesize) <= skb->len) + if (tcp_win_from_space(sk, skb->truesize) <= skb->len) incr = 2 * tp->advmss; else incr = __tcp_grow_window(sk, skb); @@ -420,11 +405,11 @@ static void tcp_fixup_rcvbuf(struct sock *sk) /* Dynamic Right Sizing (DRS) has 2 to 3 RTT latency * Allow enough cushion so that sender is not limited by our window */ - if (sysctl_tcp_moderate_rcvbuf) + if (sock_net(sk)->ipv4.sysctl_tcp_moderate_rcvbuf) rcvmem <<= 2; if (sk->sk_rcvbuf < rcvmem) - sk->sk_rcvbuf = min(rcvmem, sysctl_tcp_rmem[2]); + sk->sk_rcvbuf = min(rcvmem, sock_net(sk)->ipv4.sysctl_tcp_rmem[2]); } /* 4. Try to fixup all. It is made immediately after connection enters @@ -432,6 +417,7 @@ static void tcp_fixup_rcvbuf(struct sock *sk) */ void tcp_init_buffer_space(struct sock *sk) { + int tcp_app_win = sock_net(sk)->ipv4.sysctl_tcp_app_win; struct tcp_sock *tp = tcp_sk(sk); int maxwin; @@ -450,14 +436,14 @@ void tcp_init_buffer_space(struct sock *sk) if (tp->window_clamp >= maxwin) { tp->window_clamp = maxwin; - if (sysctl_tcp_app_win && maxwin > 4 * tp->advmss) + if (tcp_app_win && maxwin > 4 * tp->advmss) tp->window_clamp = max(maxwin - - (maxwin >> sysctl_tcp_app_win), + (maxwin >> tcp_app_win), 4 * tp->advmss); } /* Force reservation of one segment. */ - if (sysctl_tcp_app_win && + if (tcp_app_win && tp->window_clamp > 2 * tp->advmss && tp->window_clamp + tp->advmss > maxwin) tp->window_clamp = max(2 * tp->advmss, maxwin - tp->advmss); @@ -471,15 +457,16 @@ static void tcp_clamp_window(struct sock *sk) { struct tcp_sock *tp = tcp_sk(sk); struct inet_connection_sock *icsk = inet_csk(sk); + struct net *net = sock_net(sk); icsk->icsk_ack.quick = 0; - if (sk->sk_rcvbuf < sysctl_tcp_rmem[2] && + if (sk->sk_rcvbuf < net->ipv4.sysctl_tcp_rmem[2] && !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) && !tcp_under_memory_pressure(sk) && sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)) { sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc), - sysctl_tcp_rmem[2]); + net->ipv4.sysctl_tcp_rmem[2]); } if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) tp->rcv_ssthresh = min(tp->window_clamp, 2U * tp->advmss); @@ -610,7 +597,7 @@ void tcp_rcv_space_adjust(struct sock *sk) * <prev RTT . ><current RTT .. ><next RTT .... > */ - if (sysctl_tcp_moderate_rcvbuf && + if (sock_net(sk)->ipv4.sysctl_tcp_moderate_rcvbuf && !(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) { int rcvwin, rcvmem, rcvbuf; @@ -634,10 +621,11 @@ void tcp_rcv_space_adjust(struct sock *sk) } rcvmem = SKB_TRUESIZE(tp->advmss + MAX_TCP_HEADER); - while (tcp_win_from_space(rcvmem) < tp->advmss) + while (tcp_win_from_space(sk, rcvmem) < tp->advmss) rcvmem += 128; - rcvbuf = min(rcvwin / tp->advmss * rcvmem, sysctl_tcp_rmem[2]); + rcvbuf = min(rcvwin / tp->advmss * rcvmem, + sock_net(sk)->ipv4.sysctl_tcp_rmem[2]); if (rcvbuf > sk->sk_rcvbuf) { sk->sk_rcvbuf = rcvbuf; @@ -781,15 +769,6 @@ static void tcp_rtt_estimator(struct sock *sk, long mrtt_us) tp->srtt_us = max(1U, srtt); } -/* Set the sk_pacing_rate to allow proper sizing of TSO packets. - * Note: TCP stack does not yet implement pacing. - * FQ packet scheduler can be used to implement cheap but effective - * TCP pacing, to smooth the burst on large writes when packets - * in flight is significantly lower than cwnd (or rwin) - */ -int sysctl_tcp_pacing_ss_ratio __read_mostly = 200; -int sysctl_tcp_pacing_ca_ratio __read_mostly = 120; - static void tcp_update_pacing_rate(struct sock *sk) { const struct tcp_sock *tp = tcp_sk(sk); @@ -807,9 +786,9 @@ static void tcp_update_pacing_rate(struct sock *sk) * end of slow start and should slow down. */ if (tp->snd_cwnd < tp->snd_ssthresh / 2) - rate *= sysctl_tcp_pacing_ss_ratio; + rate *= sock_net(sk)->ipv4.sysctl_tcp_pacing_ss_ratio; else - rate *= sysctl_tcp_pacing_ca_ratio; + rate *= sock_net(sk)->ipv4.sysctl_tcp_pacing_ca_ratio; rate *= max(tp->snd_cwnd, tp->packets_out); @@ -863,60 +842,46 @@ __u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst) return min_t(__u32, cwnd, tp->snd_cwnd_clamp); } -/* - * Packet counting of FACK is based on in-order assumptions, therefore TCP - * disables it when reordering is detected - */ -void tcp_disable_fack(struct tcp_sock *tp) -{ - /* RFC3517 uses different metric in lost marker => reset on change */ - if (tcp_is_fack(tp)) - tp->lost_skb_hint = NULL; - tp->rx_opt.sack_ok &= ~TCP_FACK_ENABLED; -} - /* Take a notice that peer is sending D-SACKs */ static void tcp_dsack_seen(struct tcp_sock *tp) { tp->rx_opt.sack_ok |= TCP_DSACK_SEEN; + tp->rack.dsack_seen = 1; } -static void tcp_update_reordering(struct sock *sk, const int metric, - const int ts) +/* It's reordering when higher sequence was delivered (i.e. sacked) before + * some lower never-retransmitted sequence ("low_seq"). The maximum reordering + * distance is approximated in full-mss packet distance ("reordering"). + */ +static void tcp_check_sack_reordering(struct sock *sk, const u32 low_seq, + const int ts) { struct tcp_sock *tp = tcp_sk(sk); - int mib_idx; + const u32 mss = tp->mss_cache; + u32 fack, metric; - if (WARN_ON_ONCE(metric < 0)) + fack = tcp_highest_sack_seq(tp); + if (!before(low_seq, fack)) return; - if (metric > tp->reordering) { - tp->reordering = min(sysctl_tcp_max_reordering, metric); - + metric = fack - low_seq; + if ((metric > tp->reordering * mss) && mss) { #if FASTRETRANS_DEBUG > 1 pr_debug("Disorder%d %d %u f%u s%u rr%d\n", tp->rx_opt.sack_ok, inet_csk(sk)->icsk_ca_state, tp->reordering, - tp->fackets_out, + 0, tp->sacked_out, tp->undo_marker ? tp->undo_retrans : 0); #endif - tcp_disable_fack(tp); + tp->reordering = min_t(u32, (metric + mss - 1) / mss, + sock_net(sk)->ipv4.sysctl_tcp_max_reordering); } tp->rack.reord = 1; - /* This exciting event is worth to be remembered. 8) */ - if (ts) - mib_idx = LINUX_MIB_TCPTSREORDER; - else if (tcp_is_reno(tp)) - mib_idx = LINUX_MIB_TCPRENOREORDER; - else if (tcp_is_fack(tp)) - mib_idx = LINUX_MIB_TCPFACKREORDER; - else - mib_idx = LINUX_MIB_TCPSACKREORDER; - - NET_INC_STATS(sock_net(sk), mib_idx); + NET_INC_STATS(sock_net(sk), + ts ? LINUX_MIB_TCPTSREORDER : LINUX_MIB_TCPSACKREORDER); } /* This must be called before lost_out is incremented */ @@ -990,7 +955,6 @@ void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp, struct sk_buff *skb) * 3. Loss detection event of two flavors: * A. Scoreboard estimator decided the packet is lost. * A'. Reno "three dupacks" marks head of queue lost. - * A''. Its FACK modification, head until snd.fack is lost. * B. SACK arrives sacking SND.NXT at the moment, when the * segment was retransmitted. * 4. D-SACK added new rule: D-SACK changes any tag to S. @@ -1133,8 +1097,7 @@ static bool tcp_check_dsack(struct sock *sk, const struct sk_buff *ack_skb, } struct tcp_sacktag_state { - int reord; - int fack_count; + u32 reord; /* Timestamps for earliest and latest never-retransmitted segment * that was SACKed. RTO needs the earliest RTT to stay conservative, * but congestion control should still get an accurate delay signal. @@ -1143,6 +1106,7 @@ struct tcp_sacktag_state { u64 last_sackt; struct rate_sample *rate; int flag; + unsigned int mss_now; }; /* Check if skb is fully within the SACK block. In presence of GSO skbs, @@ -1192,7 +1156,8 @@ static int tcp_match_skb_to_sack(struct sock *sk, struct sk_buff *skb, if (pkt_len >= skb->len && !in_sack) return 0; - err = tcp_fragment(sk, skb, pkt_len, mss, GFP_ATOMIC); + err = tcp_fragment(sk, TCP_FRAG_IN_RTX_QUEUE, skb, + pkt_len, mss, GFP_ATOMIC); if (err < 0) return err; } @@ -1208,15 +1173,15 @@ static u8 tcp_sacktag_one(struct sock *sk, u64 xmit_time) { struct tcp_sock *tp = tcp_sk(sk); - int fack_count = state->fack_count; /* Account D-SACK for retransmitted packet. */ if (dup_sack && (sacked & TCPCB_RETRANS)) { if (tp->undo_marker && tp->undo_retrans > 0 && after(end_seq, tp->undo_marker)) tp->undo_retrans--; - if (sacked & TCPCB_SACKED_ACKED) - state->reord = min(fack_count, state->reord); + if ((sacked & TCPCB_SACKED_ACKED) && + before(start_seq, state->reord)) + state->reord = start_seq; } /* Nothing to do; acked frame is about to be dropped (was ACKed). */ @@ -1242,9 +1207,10 @@ static u8 tcp_sacktag_one(struct sock *sk, * which was in hole. It is reordering. */ if (before(start_seq, - tcp_highest_sack_seq(tp))) - state->reord = min(fack_count, - state->reord); + tcp_highest_sack_seq(tp)) && + before(start_seq, state->reord)) + state->reord = start_seq; + if (!after(end_seq, tp->high_seq)) state->flag |= FLAG_ORIG_SACK_ACKED; if (state->first_sackt == 0) @@ -1263,15 +1229,10 @@ static u8 tcp_sacktag_one(struct sock *sk, tp->sacked_out += pcount; tp->delivered += pcount; /* Out-of-order packets delivered */ - fack_count += pcount; - /* Lost marker hint past SACKed? Tweak RFC3517 cnt */ - if (!tcp_is_fack(tp) && tp->lost_skb_hint && + if (tp->lost_skb_hint && before(start_seq, TCP_SKB_CB(tp->lost_skb_hint)->seq)) tp->lost_cnt_hint += pcount; - - if (fack_count > tp->fackets_out) - tp->fackets_out = fack_count; } /* D-SACK. We can detect redundant retransmission in S|R and plain R @@ -1289,13 +1250,13 @@ static u8 tcp_sacktag_one(struct sock *sk, /* Shift newly-SACKed bytes from this skb to the immediately previous * already-SACKed sk_buff. Mark the newly-SACKed bytes as such. */ -static bool tcp_shifted_skb(struct sock *sk, struct sk_buff *skb, +static bool tcp_shifted_skb(struct sock *sk, struct sk_buff *prev, + struct sk_buff *skb, struct tcp_sacktag_state *state, unsigned int pcount, int shifted, int mss, bool dup_sack) { struct tcp_sock *tp = tcp_sk(sk); - struct sk_buff *prev = tcp_write_queue_prev(sk, skb); u32 start_seq = TCP_SKB_CB(skb)->seq; /* start of newly-SACKed */ u32 end_seq = start_seq + shifted; /* end of newly-SACKed */ @@ -1364,8 +1325,7 @@ static bool tcp_shifted_skb(struct sock *sk, struct sk_buff *skb, if (unlikely(TCP_SKB_CB(prev)->tx.delivered_mstamp)) TCP_SKB_CB(prev)->tx.delivered_mstamp = 0; - tcp_unlink_write_queue(skb, sk); - sk_wmem_free_skb(sk, skb); + tcp_rtx_queue_unlink_and_free(skb, sk); NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKMERGED); @@ -1415,9 +1375,9 @@ static struct sk_buff *tcp_shift_skb_data(struct sock *sk, struct sk_buff *skb, goto fallback; /* Can only happen with delayed DSACK + discard craziness */ - if (unlikely(skb == tcp_write_queue_head(sk))) + prev = skb_rb_prev(skb); + if (!prev) goto fallback; - prev = tcp_write_queue_prev(sk, skb); if ((TCP_SKB_CB(prev)->sacked & TCPCB_TAGBITS) != TCPCB_SACKED_ACKED) goto fallback; @@ -1496,18 +1456,17 @@ static struct sk_buff *tcp_shift_skb_data(struct sock *sk, struct sk_buff *skb, if (!skb_shift(prev, skb, len)) goto fallback; - if (!tcp_shifted_skb(sk, skb, state, pcount, len, mss, dup_sack)) + if (!tcp_shifted_skb(sk, prev, skb, state, pcount, len, mss, dup_sack)) goto out; /* Hole filled allows collapsing with the next as well, this is very * useful when hole on every nth skb pattern happens */ - if (prev == tcp_write_queue_tail(sk)) + skb = skb_rb_next(prev); + if (!skb) goto out; - skb = tcp_write_queue_next(sk, prev); if (!skb_can_shift(skb) || - (skb == tcp_send_head(sk)) || ((TCP_SKB_CB(skb)->sacked & TCPCB_TAGBITS) != TCPCB_SACKED_ACKED) || (mss != tcp_skb_seglen(skb))) goto out; @@ -1515,11 +1474,11 @@ static struct sk_buff *tcp_shift_skb_data(struct sock *sk, struct sk_buff *skb, len = skb->len; if (skb_shift(prev, skb, len)) { pcount += tcp_skb_pcount(skb); - tcp_shifted_skb(sk, skb, state, tcp_skb_pcount(skb), len, mss, 0); + tcp_shifted_skb(sk, prev, skb, state, tcp_skb_pcount(skb), + len, mss, 0); } out: - state->fack_count += pcount; return prev; noop: @@ -1539,13 +1498,10 @@ static struct sk_buff *tcp_sacktag_walk(struct sk_buff *skb, struct sock *sk, struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *tmp; - tcp_for_write_queue_from(skb, sk) { + skb_rbtree_walk_from(skb) { int in_sack = 0; bool dup_sack = dup_sack_in; - if (skb == tcp_send_head(sk)) - break; - /* queue is in-order => we can short-circuit the walk early */ if (!before(TCP_SKB_CB(skb)->seq, end_seq)) break; @@ -1594,34 +1550,48 @@ static struct sk_buff *tcp_sacktag_walk(struct sk_buff *skb, struct sock *sk, tcp_skb_pcount(skb), skb->skb_mstamp); tcp_rate_skb_delivered(sk, skb, state->rate); + if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) + list_del_init(&skb->tcp_tsorted_anchor); if (!before(TCP_SKB_CB(skb)->seq, tcp_highest_sack_seq(tp))) tcp_advance_highest_sack(sk, skb); } - - state->fack_count += tcp_skb_pcount(skb); } return skb; } -/* Avoid all extra work that is being done by sacktag while walking in - * a normal way - */ +static struct sk_buff *tcp_sacktag_bsearch(struct sock *sk, + struct tcp_sacktag_state *state, + u32 seq) +{ + struct rb_node *parent, **p = &sk->tcp_rtx_queue.rb_node; + struct sk_buff *skb; + + while (*p) { + parent = *p; + skb = rb_to_skb(parent); + if (before(seq, TCP_SKB_CB(skb)->seq)) { + p = &parent->rb_left; + continue; + } + if (!before(seq, TCP_SKB_CB(skb)->end_seq)) { + p = &parent->rb_right; + continue; + } + return skb; + } + return NULL; +} + static struct sk_buff *tcp_sacktag_skip(struct sk_buff *skb, struct sock *sk, struct tcp_sacktag_state *state, u32 skip_to_seq) { - tcp_for_write_queue_from(skb, sk) { - if (skb == tcp_send_head(sk)) - break; - - if (after(TCP_SKB_CB(skb)->end_seq, skip_to_seq)) - break; + if (skb && after(TCP_SKB_CB(skb)->seq, skip_to_seq)) + return skb; - state->fack_count += tcp_skb_pcount(skb); - } - return skb; + return tcp_sacktag_bsearch(sk, state, skip_to_seq); } static struct sk_buff *tcp_maybe_skipping_dsack(struct sk_buff *skb, @@ -1666,13 +1636,10 @@ tcp_sacktag_write_queue(struct sock *sk, const struct sk_buff *ack_skb, int first_sack_index; state->flag = 0; - state->reord = tp->packets_out; + state->reord = tp->snd_nxt; - if (!tp->sacked_out) { - if (WARN_ON(tp->fackets_out)) - tp->fackets_out = 0; + if (!tp->sacked_out) tcp_highest_sack_reset(sk); - } found_dup_sack = tcp_check_dsack(sk, ack_skb, sp_wire, num_sacks, prior_snd_una); @@ -1743,8 +1710,8 @@ tcp_sacktag_write_queue(struct sock *sk, const struct sk_buff *ack_skb, } } - skb = tcp_write_queue_head(sk); - state->fack_count = 0; + state->mss_now = tcp_current_mss(sk); + skb = NULL; i = 0; if (!tp->sacked_out) { @@ -1801,7 +1768,6 @@ tcp_sacktag_write_queue(struct sock *sk, const struct sk_buff *ack_skb, skb = tcp_highest_sack(sk); if (!skb) break; - state->fack_count = tp->fackets_out; cache++; goto walk; } @@ -1816,7 +1782,6 @@ tcp_sacktag_write_queue(struct sock *sk, const struct sk_buff *ack_skb, skb = tcp_highest_sack(sk); if (!skb) break; - state->fack_count = tp->fackets_out; } skb = tcp_sacktag_skip(skb, sk, state, start_seq); @@ -1836,9 +1801,8 @@ advance_sp: for (j = 0; j < used_sacks; j++) tp->recv_sack_cache[i++] = sp[j]; - if ((state->reord < tp->fackets_out) && - ((inet_csk(sk)->icsk_ca_state != TCP_CA_Loss) || tp->undo_marker)) - tcp_update_reordering(sk, tp->fackets_out - state->reord, 0); + if (inet_csk(sk)->icsk_ca_state != TCP_CA_Loss || tp->undo_marker) + tcp_check_sack_reordering(sk, state->reord, 0); tcp_verify_left_out(tp); out: @@ -1876,8 +1840,13 @@ static bool tcp_limit_reno_sacked(struct tcp_sock *tp) static void tcp_check_reno_reordering(struct sock *sk, const int addend) { struct tcp_sock *tp = tcp_sk(sk); - if (tcp_limit_reno_sacked(tp)) - tcp_update_reordering(sk, tp->packets_out + addend, 0); + + if (!tcp_limit_reno_sacked(tp)) + return; + + tp->reordering = min_t(u32, tp->packets_out + addend, + sock_net(sk)->ipv4.sysctl_tcp_max_reordering); + NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRENOREORDER); } /* Emulate SACKs for SACKless connection: account for a new dupack. */ @@ -1923,7 +1892,6 @@ void tcp_clear_retrans(struct tcp_sock *tp) tp->lost_out = 0; tp->undo_marker = 0; tp->undo_retrans = -1; - tp->fackets_out = 0; tp->sacked_out = 0; } @@ -1968,19 +1936,15 @@ void tcp_enter_loss(struct sock *sk) if (tcp_is_reno(tp)) tcp_reset_reno_sack(tp); - skb = tcp_write_queue_head(sk); + skb = tcp_rtx_queue_head(sk); is_reneg = skb && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED); if (is_reneg) { NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPSACKRENEGING); tp->sacked_out = 0; - tp->fackets_out = 0; } tcp_clear_all_retrans_hints(tp); - tcp_for_write_queue(skb, sk) { - if (skb == tcp_send_head(sk)) - break; - + skb_rbtree_walk_from(skb) { mark_lost = (!(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) || is_reneg); if (mark_lost) @@ -2014,7 +1978,7 @@ void tcp_enter_loss(struct sock *sk) * falsely raise the receive window, which results in repeated * timeouts and stop-and-go behavior. */ - tp->frto = sysctl_tcp_frto && + tp->frto = net->ipv4.sysctl_tcp_frto && (new_recovery || icsk->icsk_retransmits) && !inet_csk(sk)->icsk_mtup.probe_size; } @@ -2043,19 +2007,10 @@ static bool tcp_check_sack_reneging(struct sock *sk, int flag) return false; } -static inline int tcp_fackets_out(const struct tcp_sock *tp) -{ - return tcp_is_reno(tp) ? tp->sacked_out + 1 : tp->fackets_out; -} - /* Heurestics to calculate number of duplicate ACKs. There's no dupACKs * counter when SACK is enabled (without SACK, sacked_out is used for * that purpose). * - * Instead, with FACK TCP uses fackets_out that includes both SACKed - * segments up to the highest received SACK block so far and holes in - * between them. - * * With reordering, holes may still be in flight, so RFC3517 recovery * uses pure sacked_out (total number of SACKed segments) even though * it violates the RFC that uses duplicate ACKs, often these are equal @@ -2065,10 +2020,10 @@ static inline int tcp_fackets_out(const struct tcp_sock *tp) */ static inline int tcp_dupack_heuristics(const struct tcp_sock *tp) { - return tcp_is_fack(tp) ? tp->fackets_out : tp->sacked_out + 1; + return tp->sacked_out + 1; } -/* Linux NewReno/SACK/FACK/ECN state machine. +/* Linux NewReno/SACK/ECN state machine. * -------------------------------------- * * "Open" Normal state, no dubious events, fast path. @@ -2133,16 +2088,6 @@ static inline int tcp_dupack_heuristics(const struct tcp_sock *tp) * dynamically measured and adjusted. This is implemented in * tcp_rack_mark_lost. * - * FACK (Disabled by default. Subsumbed by RACK): - * It is the simplest heuristics. As soon as we decided - * that something is lost, we decide that _all_ not SACKed - * packets until the most forward SACK are lost. I.e. - * lost_out = fackets_out - sacked_out and left_out = fackets_out. - * It is absolutely correct estimate, if network does not reorder - * packets. And it loses any connection to reality when reordering - * takes place. We use FACK by default until reordering - * is suspected on the path to this destination. - * * If the receiver does not support SACK: * * NewReno (RFC6582): in Recovery we assume that one segment @@ -2191,7 +2136,7 @@ static bool tcp_time_to_recover(struct sock *sk, int flag) } /* Detect loss in event "A" above by marking head of queue up as lost. - * For FACK or non-SACK(Reno) senders, the first "packets" number of segments + * For non-SACK(Reno) senders, the first "packets" number of segments * are considered lost. For RFC3517 SACK, a segment is considered lost if it * has at least tp->reordering SACKed seqments above it; "packets" refers to * the maximum SACKed segments to pass before reaching this limit. @@ -2206,20 +2151,18 @@ static void tcp_mark_head_lost(struct sock *sk, int packets, int mark_head) const u32 loss_high = tcp_is_sack(tp) ? tp->snd_nxt : tp->high_seq; WARN_ON(packets > tp->packets_out); - if (tp->lost_skb_hint) { - skb = tp->lost_skb_hint; - cnt = tp->lost_cnt_hint; + skb = tp->lost_skb_hint; + if (skb) { /* Head already handled? */ - if (mark_head && skb != tcp_write_queue_head(sk)) + if (mark_head && after(TCP_SKB_CB(skb)->seq, tp->snd_una)) return; + cnt = tp->lost_cnt_hint; } else { - skb = tcp_write_queue_head(sk); + skb = tcp_rtx_queue_head(sk); cnt = 0; } - tcp_for_write_queue_from(skb, sk) { - if (skb == tcp_send_head(sk)) - break; + skb_rbtree_walk_from(skb) { /* TODO: do this better */ /* this is not the most efficient way to do this... */ tp->lost_skb_hint = skb; @@ -2229,12 +2172,12 @@ static void tcp_mark_head_lost(struct sock *sk, int packets, int mark_head) break; oldcnt = cnt; - if (tcp_is_fack(tp) || tcp_is_reno(tp) || + if (tcp_is_reno(tp) || (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)) cnt += tcp_skb_pcount(skb); if (cnt > packets) { - if ((tcp_is_sack(tp) && !tcp_is_fack(tp)) || + if (tcp_is_sack(tp) || (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) || (oldcnt >= packets)) break; @@ -2243,7 +2186,8 @@ static void tcp_mark_head_lost(struct sock *sk, int packets, int mark_head) /* If needed, chop off the prefix to mark as lost. */ lost = (packets - oldcnt) * mss; if (lost < skb->len && - tcp_fragment(sk, skb, lost, mss, GFP_ATOMIC) < 0) + tcp_fragment(sk, TCP_FRAG_IN_RTX_QUEUE, skb, + lost, mss, GFP_ATOMIC) < 0) break; cnt = packets; } @@ -2264,11 +2208,6 @@ static void tcp_update_scoreboard(struct sock *sk, int fast_rexmit) if (tcp_is_reno(tp)) { tcp_mark_head_lost(sk, 1, 1); - } else if (tcp_is_fack(tp)) { - int lost = tp->fackets_out - tp->reordering; - if (lost <= 0) - lost = 1; - tcp_mark_head_lost(sk, lost, 0); } else { int sacked_upto = tp->sacked_out - tp->reordering; if (sacked_upto >= 0) @@ -2327,16 +2266,16 @@ static bool tcp_any_retrans_done(const struct sock *sk) if (tp->retrans_out) return true; - skb = tcp_write_queue_head(sk); + skb = tcp_rtx_queue_head(sk); if (unlikely(skb && TCP_SKB_CB(skb)->sacked & TCPCB_EVER_RETRANS)) return true; return false; } -#if FASTRETRANS_DEBUG > 1 static void DBGUNDO(struct sock *sk, const char *msg) { +#if FASTRETRANS_DEBUG > 1 struct tcp_sock *tp = tcp_sk(sk); struct inet_sock *inet = inet_sk(sk); @@ -2358,10 +2297,8 @@ static void DBGUNDO(struct sock *sk, const char *msg) tp->packets_out); } #endif -} -#else -#define DBGUNDO(x...) do { } while (0) #endif +} static void tcp_undo_cwnd_reduction(struct sock *sk, bool unmark_loss) { @@ -2370,9 +2307,7 @@ static void tcp_undo_cwnd_reduction(struct sock *sk, bool unmark_loss) if (unmark_loss) { struct sk_buff *skb; - tcp_for_write_queue(skb, sk) { - if (skb == tcp_send_head(sk)) - break; + skb_rbtree_walk(skb, &sk->tcp_rtx_queue) { TCP_SKB_CB(skb)->sacked &= ~TCPCB_LOST; } tp->lost_out = 0; @@ -2417,6 +2352,8 @@ static bool tcp_try_undo_recovery(struct sock *sk) mib_idx = LINUX_MIB_TCPFULLUNDO; NET_INC_STATS(sock_net(sk), mib_idx); + } else if (tp->rack.reo_wnd_persist) { + tp->rack.reo_wnd_persist--; } if (tp->snd_una == tp->high_seq && tcp_is_reno(tp)) { /* Hold old state until something *above* high_seq @@ -2436,6 +2373,8 @@ static bool tcp_try_undo_dsack(struct sock *sk) struct tcp_sock *tp = tcp_sk(sk); if (tp->undo_marker && !tp->undo_retrans) { + tp->rack.reo_wnd_persist = min(TCP_RACK_RECOVERY_THRESH, + tp->rack.reo_wnd_persist + 1); DBGUNDO(sk, "D-SACK"); tcp_undo_cwnd_reduction(sk, false); NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDSACKUNDO); @@ -2616,9 +2555,7 @@ void tcp_simple_retransmit(struct sock *sk) struct sk_buff *skb; unsigned int mss = tcp_current_mss(sk); - tcp_for_write_queue(skb, sk) { - if (skb == tcp_send_head(sk)) - break; + skb_rbtree_walk(skb, &sk->tcp_rtx_queue) { if (tcp_skb_seglen(skb) > mss && !(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)) { if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS) { @@ -2712,7 +2649,7 @@ static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack, * is updated in tcp_ack()). Otherwise fall back to * the conventional recovery. */ - if (tcp_send_head(sk) && + if (!tcp_write_queue_empty(sk) && after(tcp_wnd_end(tp), tp->snd_nxt)) { *rexmit = REXMIT_NEW; return; @@ -2739,15 +2676,15 @@ static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack, } /* Undo during fast recovery after partial ACK. */ -static bool tcp_try_undo_partial(struct sock *sk, const int acked) +static bool tcp_try_undo_partial(struct sock *sk, u32 prior_snd_una) { struct tcp_sock *tp = tcp_sk(sk); if (tp->undo_marker && tcp_packet_delayed(tp)) { /* Plain luck! Hole if filled with delayed - * packet, rather than with a retransmit. + * packet, rather than with a retransmit. Check reordering. */ - tcp_update_reordering(sk, tcp_fackets_out(tp) + acked, 1); + tcp_check_sack_reordering(sk, prior_snd_una, 1); /* We are getting evidence that the reordering degree is higher * than we realized. If there are no retransmits out then we @@ -2774,7 +2711,7 @@ static void tcp_rack_identify_loss(struct sock *sk, int *ack_flag) struct tcp_sock *tp = tcp_sk(sk); /* Use RACK to detect loss */ - if (sysctl_tcp_recovery & TCP_RACK_LOSS_DETECTION) { + if (sock_net(sk)->ipv4.sysctl_tcp_recovery & TCP_RACK_LOSS_DETECTION) { u32 prior_retrans = tp->retrans_out; tcp_rack_mark_lost(sk); @@ -2783,6 +2720,14 @@ static void tcp_rack_identify_loss(struct sock *sk, int *ack_flag) } } +static bool tcp_force_fast_retransmit(struct sock *sk) +{ + struct tcp_sock *tp = tcp_sk(sk); + + return after(tcp_highest_sack_seq(tp), + tp->snd_una + tp->reordering * tp->mss_cache); +} + /* 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 @@ -2795,19 +2740,17 @@ static void tcp_rack_identify_loss(struct sock *sk, int *ack_flag) * It does _not_ decide what to send, it is made in function * tcp_xmit_retransmit_queue(). */ -static void tcp_fastretrans_alert(struct sock *sk, const int acked, +static void tcp_fastretrans_alert(struct sock *sk, const u32 prior_snd_una, bool is_dupack, int *ack_flag, int *rexmit) { struct inet_connection_sock *icsk = inet_csk(sk); struct tcp_sock *tp = tcp_sk(sk); int fast_rexmit = 0, flag = *ack_flag; bool do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) && - (tcp_fackets_out(tp) > tp->reordering)); + tcp_force_fast_retransmit(sk)); - if (WARN_ON(!tp->packets_out && tp->sacked_out)) + if (!tp->packets_out && tp->sacked_out) tp->sacked_out = 0; - if (WARN_ON(!tp->sacked_out && tp->fackets_out)) - tp->fackets_out = 0; /* Now state machine starts. * A. ECE, hence prohibit cwnd undoing, the reduction is required. */ @@ -2854,11 +2797,11 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, if (tcp_is_reno(tp) && is_dupack) tcp_add_reno_sack(sk); } else { - if (tcp_try_undo_partial(sk, acked)) + if (tcp_try_undo_partial(sk, prior_snd_una)) return; /* Partial ACK arrived. Force fast retransmit. */ do_lost = tcp_is_reno(tp) || - tcp_fackets_out(tp) > tp->reordering; + tcp_force_fast_retransmit(sk); } if (tcp_try_undo_dsack(sk)) { tcp_try_keep_open(sk); @@ -2873,6 +2816,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, (*ack_flag & FLAG_LOST_RETRANS))) return; /* Change state if cwnd is undone or retransmits are lost */ + /* fall through */ default: if (tcp_is_reno(tp)) { if (flag & FLAG_SND_UNA_ADVANCED) @@ -2913,8 +2857,8 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, static void tcp_update_rtt_min(struct sock *sk, u32 rtt_us) { + u32 wlen = sock_net(sk)->ipv4.sysctl_tcp_min_rtt_wlen * HZ; struct tcp_sock *tp = tcp_sk(sk); - u32 wlen = sysctl_tcp_min_rtt_wlen * HZ; minmax_running_min(&tp->rtt_min, wlen, tcp_jiffies32, rtt_us ? : jiffies_to_usecs(1)); @@ -3056,28 +3000,31 @@ static void tcp_ack_tstamp(struct sock *sk, struct sk_buff *skb, shinfo = skb_shinfo(skb); if (!before(shinfo->tskey, prior_snd_una) && - before(shinfo->tskey, tcp_sk(sk)->snd_una)) - __skb_tstamp_tx(skb, NULL, sk, SCM_TSTAMP_ACK); + before(shinfo->tskey, tcp_sk(sk)->snd_una)) { + tcp_skb_tsorted_save(skb) { + __skb_tstamp_tx(skb, NULL, sk, SCM_TSTAMP_ACK); + } tcp_skb_tsorted_restore(skb); + } } /* Remove acknowledged frames from the retransmission queue. If our packet * is before the ack sequence we can discard it as it's confirmed to have * arrived at the other end. */ -static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, - u32 prior_snd_una, int *acked, +static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack, + u32 prior_snd_una, struct tcp_sacktag_state *sack) { const struct inet_connection_sock *icsk = inet_csk(sk); u64 first_ackt, last_ackt; struct tcp_sock *tp = tcp_sk(sk); u32 prior_sacked = tp->sacked_out; - u32 reord = tp->packets_out; + u32 reord = tp->snd_nxt; /* lowest acked un-retx un-sacked seq */ + struct sk_buff *skb, *next; bool fully_acked = true; long sack_rtt_us = -1L; long seq_rtt_us = -1L; long ca_rtt_us = -1L; - struct sk_buff *skb; u32 pkts_acked = 0; u32 last_in_flight = 0; bool rtt_update; @@ -3085,8 +3032,9 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, first_ackt = 0; - while ((skb = tcp_write_queue_head(sk)) && skb != tcp_send_head(sk)) { + for (skb = skb_rb_first(&sk->tcp_rtx_queue); skb; skb = next) { struct tcp_skb_cb *scb = TCP_SKB_CB(skb); + const u32 start_seq = scb->seq; u8 sacked = scb->sacked; u32 acked_pcount; @@ -3103,8 +3051,6 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, break; fully_acked = false; } else { - /* Speedup tcp_unlink_write_queue() and next loop */ - prefetchw(skb->next); acked_pcount = tcp_skb_pcount(skb); } @@ -3119,7 +3065,8 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, first_ackt = last_ackt; last_in_flight = TCP_SKB_CB(skb)->tx.in_flight; - reord = min(pkts_acked, reord); + if (before(start_seq, reord)) + reord = start_seq; if (!after(scb->end_seq, tp->high_seq)) flag |= FLAG_ORIG_SACK_ACKED; } @@ -3156,12 +3103,12 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, if (!fully_acked) break; - tcp_unlink_write_queue(skb, sk); - sk_wmem_free_skb(sk, skb); + next = skb_rb_next(skb); if (unlikely(skb == tp->retransmit_skb_hint)) tp->retransmit_skb_hint = NULL; if (unlikely(skb == tp->lost_skb_hint)) tp->lost_skb_hint = NULL; + tcp_rtx_queue_unlink_and_free(skb, sk); } if (!skb) @@ -3197,16 +3144,12 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, int delta; /* Non-retransmitted hole got filled? That's reordering */ - if (reord < prior_fackets && reord <= tp->fackets_out) - tcp_update_reordering(sk, tp->fackets_out - reord, 0); + if (before(reord, prior_fack)) + tcp_check_sack_reordering(sk, reord, 0); - delta = tcp_is_fack(tp) ? pkts_acked : - prior_sacked - tp->sacked_out; + delta = prior_sacked - tp->sacked_out; tp->lost_cnt_hint -= min(tp->lost_cnt_hint, delta); } - - tp->fackets_out -= min(pkts_acked, tp->fackets_out); - } else if (skb && rtt_update && sack_rtt_us >= 0 && sack_rtt_us > tcp_stamp_us_delta(tp->tcp_mstamp, skb->skb_mstamp)) { /* Do not re-arm RTO if the sack RTT is measured from data sent @@ -3247,18 +3190,19 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, } } #endif - *acked = pkts_acked; return flag; } static void tcp_ack_probe(struct sock *sk) { - const struct tcp_sock *tp = tcp_sk(sk); struct inet_connection_sock *icsk = inet_csk(sk); + struct sk_buff *head = tcp_send_head(sk); + const struct tcp_sock *tp = tcp_sk(sk); /* Was it a usable window open? */ - - if (!after(TCP_SKB_CB(tcp_send_head(sk))->end_seq, tcp_wnd_end(tp))) { + if (!head) + return; + if (!after(TCP_SKB_CB(head)->end_seq, tcp_wnd_end(tp))) { icsk->icsk_backoff = 0; inet_csk_clear_xmit_timer(sk, ICSK_TIME_PROBE0); /* Socket must be waked up by subsequent tcp_data_snd_check(). @@ -3378,7 +3322,7 @@ static int tcp_ack_update_window(struct sock *sk, const struct sk_buff *skb, u32 tp->pred_flags = 0; tcp_fast_path_check(sk); - if (tcp_send_head(sk)) + if (!tcp_write_queue_empty(sk)) tcp_slow_start_after_idle_check(sk); if (nwin > tp->max_window) { @@ -3399,7 +3343,7 @@ static bool __tcp_oow_rate_limited(struct net *net, int mib_idx, if (*last_oow_ack_time) { s32 elapsed = (s32)(tcp_jiffies32 - *last_oow_ack_time); - if (0 <= elapsed && elapsed < sysctl_tcp_invalid_ratelimit) { + if (0 <= elapsed && elapsed < net->ipv4.sysctl_tcp_invalid_ratelimit) { NET_INC_STATS(net, mib_idx); return true; /* rate-limited: don't send yet! */ } @@ -3435,10 +3379,11 @@ static void tcp_send_challenge_ack(struct sock *sk, const struct sk_buff *skb) static u32 challenge_timestamp; static unsigned int challenge_count; struct tcp_sock *tp = tcp_sk(sk); + struct net *net = sock_net(sk); u32 count, now; /* First check our per-socket dupack rate limit. */ - if (__tcp_oow_rate_limited(sock_net(sk), + if (__tcp_oow_rate_limited(net, LINUX_MIB_TCPACKSKIPPEDCHALLENGE, &tp->last_oow_ack_time)) return; @@ -3446,16 +3391,16 @@ static void tcp_send_challenge_ack(struct sock *sk, const struct sk_buff *skb) /* Then check host-wide RFC 5961 rate limit. */ now = jiffies / HZ; if (now != challenge_timestamp) { - u32 half = (sysctl_tcp_challenge_ack_limit + 1) >> 1; + u32 ack_limit = net->ipv4.sysctl_tcp_challenge_ack_limit; + u32 half = (ack_limit + 1) >> 1; challenge_timestamp = now; - WRITE_ONCE(challenge_count, half + - prandom_u32_max(sysctl_tcp_challenge_ack_limit)); + WRITE_ONCE(challenge_count, half + prandom_u32_max(ack_limit)); } count = READ_ONCE(challenge_count); if (count > 0) { WRITE_ONCE(challenge_count, count - 1); - NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPCHALLENGEACK); + NET_INC_STATS(net, LINUX_MIB_TCPCHALLENGEACK); tcp_send_ack(sk); } } @@ -3553,18 +3498,17 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) u32 ack_seq = TCP_SKB_CB(skb)->seq; u32 ack = TCP_SKB_CB(skb)->ack_seq; bool is_dupack = false; - u32 prior_fackets; int prior_packets = tp->packets_out; u32 delivered = tp->delivered; u32 lost = tp->lost; - int acked = 0; /* Number of packets newly acked */ int rexmit = REXMIT_NONE; /* Flag to (re)transmit to recover losses */ + u32 prior_fack; sack_state.first_sackt = 0; sack_state.rate = &rs; - /* We very likely will need to access write queue head. */ - prefetchw(sk->sk_write_queue.next); + /* We very likely will need to access rtx queue. */ + prefetch(sk->tcp_rtx_queue.rb_node); /* If the ack is older than previous acks * then we can probably ignore it. @@ -3590,7 +3534,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) icsk->icsk_retransmits = 0; } - prior_fackets = tp->fackets_out; + prior_fack = tcp_is_sack(tp) ? tcp_highest_sack_seq(tp) : tp->snd_una; rs.prior_in_flight = tcp_packets_in_flight(tp); /* ts_recent update must be made after we are sure that the packet @@ -3646,8 +3590,9 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) goto no_queue; /* See if we can take anything off of the retransmit queue. */ - flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una, &acked, - &sack_state); + flag |= tcp_clean_rtx_queue(sk, prior_fack, prior_snd_una, &sack_state); + + tcp_rack_update_reo_wnd(sk, &rs); if (tp->tlp_high_seq) tcp_process_tlp_ack(sk, ack, flag); @@ -3657,7 +3602,8 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) if (tcp_ack_is_dubious(sk, flag)) { is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP)); - tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit); + tcp_fastretrans_alert(sk, prior_snd_una, is_dupack, &flag, + &rexmit); } if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP)) @@ -3673,13 +3619,13 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) no_queue: /* If data was DSACKed, see if we can undo a cwnd reduction. */ if (flag & FLAG_DSACKING_ACK) - tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit); + tcp_fastretrans_alert(sk, prior_snd_una, is_dupack, &flag, + &rexmit); /* If this ack opens up a zero window, clear backoff. It was * being used to time the probes, and is probably far higher than * it needs to be for normal retransmission. */ - if (tcp_send_head(sk)) - tcp_ack_probe(sk); + tcp_ack_probe(sk); if (tp->tlp_high_seq) tcp_process_tlp_ack(sk, ack, flag); @@ -3696,7 +3642,8 @@ old_ack: if (TCP_SKB_CB(skb)->sacked) { flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una, &sack_state); - tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit); + tcp_fastretrans_alert(sk, prior_snd_una, is_dupack, &flag, + &rexmit); tcp_xmit_recovery(sk, rexmit); } @@ -3721,6 +3668,21 @@ static void tcp_parse_fastopen_option(int len, const unsigned char *cookie, foc->exp = exp_opt; } +static void smc_parse_options(const struct tcphdr *th, + struct tcp_options_received *opt_rx, + const unsigned char *ptr, + int opsize) +{ +#if IS_ENABLED(CONFIG_SMC) + if (static_branch_unlikely(&tcp_have_smc)) { + if (th->syn && !(opsize & 1) && + opsize >= TCPOLEN_EXP_SMC_BASE && + get_unaligned_be32(ptr) == TCPOPT_SMC_MAGIC) + opt_rx->smc_ok = 1; + } +#endif +} + /* Look for tcp options. Normally only called on SYN and SYNACK packets. * But, this can also be called on packets in the established flow when * the fast version below fails. @@ -3828,6 +3790,9 @@ void tcp_parse_options(const struct net *net, tcp_parse_fastopen_option(opsize - TCPOLEN_EXP_FASTOPEN_BASE, ptr + 2, th->syn, foc, true); + else + smc_parse_options(th, opt_rx, ptr, + opsize); break; } @@ -3995,6 +3960,8 @@ static inline bool tcp_sequence(const struct tcp_sock *tp, u32 seq, u32 end_seq) /* When we get a reset we do this. */ void tcp_reset(struct sock *sk) { + trace_tcp_receive_reset(sk); + /* We want the right error as BSD sees it (and indeed as we do). */ switch (sk->sk_state) { case TCP_SYN_SENT: @@ -4117,7 +4084,7 @@ static void tcp_dsack_set(struct sock *sk, u32 seq, u32 end_seq) { struct tcp_sock *tp = tcp_sk(sk); - if (tcp_is_sack(tp) && sysctl_tcp_dsack) { + if (tcp_is_sack(tp) && sock_net(sk)->ipv4.sysctl_tcp_dsack) { int mib_idx; if (before(seq, tp->rcv_nxt)) @@ -4152,7 +4119,7 @@ static void tcp_send_dupack(struct sock *sk, const struct sk_buff *skb) NET_INC_STATS(sock_net(sk), LINUX_MIB_DELAYEDACKLOST); tcp_enter_quickack_mode(sk); - if (tcp_is_sack(tp) && sysctl_tcp_dsack) { + if (tcp_is_sack(tp) && sock_net(sk)->ipv4.sysctl_tcp_dsack) { u32 end_seq = TCP_SKB_CB(skb)->end_seq; if (after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) @@ -4268,11 +4235,6 @@ static void tcp_sack_remove(struct tcp_sock *tp) tp->rx_opt.num_sacks = num_sacks; } -enum tcp_queue { - OOO_QUEUE, - RCV_QUEUE, -}; - /** * tcp_try_coalesce - try to merge skb to prior one * @sk: socket @@ -4288,7 +4250,6 @@ enum tcp_queue { * Returns true if caller should free @from instead of queueing it */ static bool tcp_try_coalesce(struct sock *sk, - enum tcp_queue dest, struct sk_buff *to, struct sk_buff *from, bool *fragstolen) @@ -4313,10 +4274,7 @@ static bool tcp_try_coalesce(struct sock *sk, if (TCP_SKB_CB(from)->has_rxtstamp) { TCP_SKB_CB(to)->has_rxtstamp = true; - if (dest == OOO_QUEUE) - TCP_SKB_CB(to)->swtstamp = TCP_SKB_CB(from)->swtstamp; - else - to->tstamp = from->tstamp; + to->tstamp = from->tstamp; } return true; @@ -4341,7 +4299,7 @@ static void tcp_ofo_queue(struct sock *sk) p = rb_first(&tp->out_of_order_queue); while (p) { - skb = rb_entry(p, struct sk_buff, rbnode); + skb = rb_to_skb(p); if (after(TCP_SKB_CB(skb)->seq, tp->rcv_nxt)) break; @@ -4353,9 +4311,6 @@ static void tcp_ofo_queue(struct sock *sk) } p = rb_next(p); rb_erase(&skb->rbnode, &tp->out_of_order_queue); - /* Replace tstamp which was stomped by rbnode */ - if (TCP_SKB_CB(skb)->has_rxtstamp) - skb->tstamp = TCP_SKB_CB(skb)->swtstamp; if (unlikely(!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt))) { SOCK_DEBUG(sk, "ofo packet was already received\n"); @@ -4367,8 +4322,7 @@ static void tcp_ofo_queue(struct sock *sk) TCP_SKB_CB(skb)->end_seq); tail = skb_peek_tail(&sk->sk_receive_queue); - eaten = tail && tcp_try_coalesce(sk, RCV_QUEUE, - tail, skb, &fragstolen); + eaten = tail && tcp_try_coalesce(sk, tail, skb, &fragstolen); tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq); fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN; if (!eaten) @@ -4409,7 +4363,7 @@ static int tcp_try_rmem_schedule(struct sock *sk, struct sk_buff *skb, static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb) { struct tcp_sock *tp = tcp_sk(sk); - struct rb_node **p, *q, *parent; + struct rb_node **p, *parent; struct sk_buff *skb1; u32 seq, end_seq; bool fragstolen; @@ -4422,10 +4376,6 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb) return; } - /* Stash tstamp to avoid being stomped on by rbnode */ - if (TCP_SKB_CB(skb)->has_rxtstamp) - TCP_SKB_CB(skb)->swtstamp = skb->tstamp; - /* Disable header prediction. */ tp->pred_flags = 0; inet_csk_schedule_ack(sk); @@ -4453,7 +4403,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb) /* In the typical case, we are adding an skb to the end of the list. * Use of ooo_last_skb avoids the O(Log(N)) rbtree lookup. */ - if (tcp_try_coalesce(sk, OOO_QUEUE, tp->ooo_last_skb, + if (tcp_try_coalesce(sk, tp->ooo_last_skb, skb, &fragstolen)) { coalesce_done: tcp_grow_window(sk, skb); @@ -4472,7 +4422,7 @@ coalesce_done: parent = NULL; while (*p) { parent = *p; - skb1 = rb_entry(parent, struct sk_buff, rbnode); + skb1 = rb_to_skb(parent); if (before(seq, TCP_SKB_CB(skb1)->seq)) { p = &parent->rb_left; continue; @@ -4504,7 +4454,7 @@ coalesce_done: __kfree_skb(skb1); goto merge_right; } - } else if (tcp_try_coalesce(sk, OOO_QUEUE, skb1, + } else if (tcp_try_coalesce(sk, skb1, skb, &fragstolen)) { goto coalesce_done; } @@ -4517,9 +4467,7 @@ insert: merge_right: /* Remove other segments covered by skb. */ - while ((q = rb_next(&skb->rbnode)) != NULL) { - skb1 = rb_entry(q, struct sk_buff, rbnode); - + while ((skb1 = skb_rb_next(skb)) != NULL) { if (!after(end_seq, TCP_SKB_CB(skb1)->seq)) break; if (before(end_seq, TCP_SKB_CB(skb1)->end_seq)) { @@ -4534,7 +4482,7 @@ merge_right: tcp_drop(sk, skb1); } /* If there is no skb after us, we are the last_skb ! */ - if (!q) + if (!skb1) tp->ooo_last_skb = skb; add_sack: @@ -4556,7 +4504,7 @@ static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int __skb_pull(skb, hdrlen); eaten = (tail && - tcp_try_coalesce(sk, RCV_QUEUE, tail, + tcp_try_coalesce(sk, tail, skb, fragstolen)) ? 1 : 0; tcp_rcv_nxt_update(tcp_sk(sk), TCP_SKB_CB(skb)->end_seq); if (!eaten) { @@ -4720,7 +4668,7 @@ static struct sk_buff *tcp_skb_next(struct sk_buff *skb, struct sk_buff_head *li if (list) return !skb_queue_is_last(list, skb) ? skb->next : NULL; - return rb_entry_safe(rb_next(&skb->rbnode), struct sk_buff, rbnode); + return skb_rb_next(skb); } static struct sk_buff *tcp_collapse_one(struct sock *sk, struct sk_buff *skb, @@ -4741,7 +4689,7 @@ static struct sk_buff *tcp_collapse_one(struct sock *sk, struct sk_buff *skb, } /* Insert skb into rb tree, ordered by TCP_SKB_CB(skb)->seq */ -static void tcp_rbtree_insert(struct rb_root *root, struct sk_buff *skb) +void tcp_rbtree_insert(struct rb_root *root, struct sk_buff *skb) { struct rb_node **p = &root->rb_node; struct rb_node *parent = NULL; @@ -4749,7 +4697,7 @@ static void tcp_rbtree_insert(struct rb_root *root, struct sk_buff *skb) while (*p) { parent = *p; - skb1 = rb_entry(parent, struct sk_buff, rbnode); + skb1 = rb_to_skb(parent); if (before(TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb1)->seq)) p = &parent->rb_left; else @@ -4796,7 +4744,7 @@ restart: * overlaps to the next one. */ if (!(TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)) && - (tcp_win_from_space(skb->truesize) > skb->len || + (tcp_win_from_space(sk, skb->truesize) > skb->len || before(TCP_SKB_CB(skb)->seq, start))) { end_of_skbs = false; break; @@ -4868,26 +4816,19 @@ static void tcp_collapse_ofo_queue(struct sock *sk) { struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *skb, *head; - struct rb_node *p; u32 start, end; - p = rb_first(&tp->out_of_order_queue); - skb = rb_entry_safe(p, struct sk_buff, rbnode); + skb = skb_rb_first(&tp->out_of_order_queue); new_range: if (!skb) { - p = rb_last(&tp->out_of_order_queue); - /* Note: This is possible p is NULL here. We do not - * use rb_entry_safe(), as ooo_last_skb is valid only - * if rbtree is not empty. - */ - tp->ooo_last_skb = rb_entry(p, struct sk_buff, rbnode); + tp->ooo_last_skb = skb_rb_last(&tp->out_of_order_queue); return; } start = TCP_SKB_CB(skb)->seq; end = TCP_SKB_CB(skb)->end_seq; for (head = skb;;) { - skb = tcp_skb_next(skb, NULL); + skb = skb_rb_next(skb); /* Range is terminated when we see a gap or when * we are at the queue end. @@ -4930,14 +4871,14 @@ static bool tcp_prune_ofo_queue(struct sock *sk) do { prev = rb_prev(node); rb_erase(node, &tp->out_of_order_queue); - tcp_drop(sk, rb_entry(node, struct sk_buff, rbnode)); + tcp_drop(sk, rb_to_skb(node)); sk_mem_reclaim(sk); if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && !tcp_under_memory_pressure(sk)) break; node = prev; } while (node); - tp->ooo_last_skb = rb_entry(prev, struct sk_buff, rbnode); + tp->ooo_last_skb = rb_to_skb(prev); /* Reset SACK state. A conforming SACK implementation will * do the same at a timeout based retransmit. When a connection @@ -5112,7 +5053,7 @@ static void tcp_check_urg(struct sock *sk, const struct tcphdr *th) struct tcp_sock *tp = tcp_sk(sk); u32 ptr = ntohs(th->urg_ptr); - if (ptr && !sysctl_tcp_stdurg) + if (ptr && !sock_net(sk)->ipv4.sysctl_tcp_stdurg) ptr--; ptr += ntohl(th->seq); @@ -5532,20 +5473,13 @@ void tcp_finish_connect(struct sock *sk, struct sk_buff *skb) security_inet_conn_established(sk, skb); } - /* Make sure socket is routed, for correct metrics. */ - icsk->icsk_af_ops->rebuild_header(sk); - - tcp_init_metrics(sk); - tcp_call_bpf(sk, BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB); - tcp_init_congestion_control(sk); + tcp_init_transfer(sk, BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB); /* Prevent spurious tcp_cwnd_restart() on first data * packet. */ tp->lsndtime = tcp_jiffies32; - tcp_init_buffer_space(sk); - if (sock_flag(sk, SOCK_KEEPOPEN)) inet_csk_reset_keepalive_timer(sk, keepalive_time_when(tp)); @@ -5559,7 +5493,7 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack, struct tcp_fastopen_cookie *cookie) { struct tcp_sock *tp = tcp_sk(sk); - struct sk_buff *data = tp->syn_data ? tcp_write_queue_head(sk) : NULL; + struct sk_buff *data = tp->syn_data ? tcp_rtx_queue_head(sk) : NULL; u16 mss = tp->rx_opt.mss_clamp, try_exp = 0; bool syn_drop = false; @@ -5594,9 +5528,8 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack, tcp_fastopen_cache_set(sk, mss, cookie, syn_drop, try_exp); if (data) { /* Retransmit unacked data in SYN */ - tcp_for_write_queue_from(data, sk) { - if (data == tcp_send_head(sk) || - __tcp_retransmit_skb(sk, data, 1)) + skb_rbtree_walk_from(data) { + if (__tcp_retransmit_skb(sk, data, 1)) break; } tcp_rearm_rto(sk); @@ -5614,6 +5547,16 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack, return false; } +static void smc_check_reset_syn(struct tcp_sock *tp) +{ +#if IS_ENABLED(CONFIG_SMC) + if (static_branch_unlikely(&tcp_have_smc)) { + if (tp->syn_smc && !tp->rx_opt.smc_ok) + tp->syn_smc = 0; + } +#endif +} + static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, const struct tcphdr *th) { @@ -5709,10 +5652,6 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, tp->tcp_header_len = sizeof(struct tcphdr); } - if (tcp_is_sack(tp) && sysctl_tcp_fack) - tcp_enable_fack(tp); - - tcp_mtup_init(sk); tcp_sync_mss(sk, icsk->icsk_pmtu_cookie); tcp_initialize_rcv_mss(sk); @@ -5721,6 +5660,8 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, * is initialized. */ tp->copied_seq = tp->rcv_nxt; + smc_check_reset_syn(tp); + smp_mb(); tcp_finish_connect(sk, skb); @@ -5938,15 +5879,18 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb) if (req) { inet_csk(sk)->icsk_retransmits = 0; reqsk_fastopen_remove(sk, req, false); + /* Re-arm the timer because data may have been sent out. + * This is similar to the regular data transmission case + * when new data has just been ack'ed. + * + * (TFO) - we could try to be more aggressive and + * retransmitting any data sooner based on when they + * are sent out. + */ + tcp_rearm_rto(sk); } else { - /* Make sure socket is routed, for correct metrics. */ - icsk->icsk_af_ops->rebuild_header(sk); - tcp_call_bpf(sk, BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB); - tcp_init_congestion_control(sk); - - tcp_mtup_init(sk); + tcp_init_transfer(sk, BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB); tp->copied_seq = tp->rcv_nxt; - tcp_init_buffer_space(sk); } smp_mb(); tcp_set_state(sk, TCP_ESTABLISHED); @@ -5966,19 +5910,6 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb) if (tp->rx_opt.tstamp_ok) tp->advmss -= TCPOLEN_TSTAMP_ALIGNED; - if (req) { - /* Re-arm the timer because data may have been sent out. - * This is similar to the regular data transmission case - * when new data has just been ack'ed. - * - * (TFO) - we could try to be more aggressive and - * retransmitting any data sooner based on when they - * are sent out. - */ - tcp_rearm_rto(sk); - } else - tcp_init_metrics(sk); - if (!inet_csk(sk)->icsk_ca_ops->cong_control) tcp_update_pacing_rate(sk); @@ -6075,6 +6006,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb) case TCP_LAST_ACK: if (!before(TCP_SKB_CB(skb)->seq, tp->rcv_nxt)) break; + /* fall through */ case TCP_FIN_WAIT1: case TCP_FIN_WAIT2: /* RFC 793 says to queue data in these states, @@ -6183,6 +6115,9 @@ static void tcp_openreq_init(struct request_sock *req, ireq->ir_rmt_port = tcp_hdr(skb)->source; ireq->ir_num = ntohs(tcp_hdr(skb)->dest); ireq->ir_mark = inet_request_mark(sk, skb); +#if IS_ENABLED(CONFIG_SMC) + ireq->smc_ok = rx_opt->smc_ok; +#endif } struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops, @@ -6358,7 +6293,7 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, tcp_openreq_init_rwin(req, sk, dst); if (!want_cookie) { tcp_reqsk_record_syn(sk, req, skb); - fastopen_sk = tcp_try_fastopen(sk, skb, req, &foc); + fastopen_sk = tcp_try_fastopen(sk, skb, req, &foc, dst); } if (fastopen_sk) { af_ops->send_synack(fastopen_sk, dst, &fl, req, |