diff options
Diffstat (limited to 'net/ipv4/tcp_input.c')
| -rw-r--r-- | net/ipv4/tcp_input.c | 397 |
1 files changed, 194 insertions, 203 deletions
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index b2fc7163bd40..b54ee09cbcf7 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. */ @@ -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, 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); @@ -420,7 +405,7 @@ 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) @@ -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); @@ -610,7 +596,7 @@ 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; @@ -634,7 +620,7 @@ 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]); @@ -781,15 +767,6 @@ 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 +784,9 @@ 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); @@ -879,6 +856,7 @@ void tcp_disable_fack(struct tcp_sock *tp) 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, @@ -891,7 +869,7 @@ static void tcp_update_reordering(struct sock *sk, const int metric, return; if (metric > tp->reordering) { - tp->reordering = min(sysctl_tcp_max_reordering, metric); + tp->reordering = min(sock_net(sk)->ipv4.sysctl_tcp_max_reordering, metric); #if FASTRETRANS_DEBUG > 1 pr_debug("Disorder%d %d %u f%u s%u rr%d\n", @@ -1143,6 +1121,7 @@ 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 +1171,8 @@ 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; } @@ -1289,13 +1269,13 @@ 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 +1344,7 @@ 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 +1394,9 @@ 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 +1475,17 @@ 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; @@ -1515,7 +1493,8 @@ 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: @@ -1539,13 +1518,10 @@ 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; @@ -1594,6 +1570,8 @@ 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))) @@ -1605,23 +1583,44 @@ static struct sk_buff *tcp_sacktag_walk(struct sk_buff *skb, struct sock *sk, 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; + int unack_bytes; + + 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; + } + + state->fack_count = 0; + unack_bytes = TCP_SKB_CB(skb)->seq - tcp_sk(sk)->snd_una; + if (state->mss_now && unack_bytes > 0) + state->fack_count = unack_bytes / state->mss_now; + + 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, @@ -1743,8 +1742,9 @@ tcp_sacktag_write_queue(struct sock *sk, const struct sk_buff *ack_skb, } } - skb = tcp_write_queue_head(sk); + state->mss_now = tcp_current_mss(sk); state->fack_count = 0; + skb = NULL; i = 0; if (!tp->sacked_out) { @@ -1968,7 +1968,7 @@ 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); @@ -1977,10 +1977,7 @@ void tcp_enter_loss(struct sock *sk) } 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 +2011,7 @@ 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; } @@ -2206,20 +2203,18 @@ 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; @@ -2243,7 +2238,8 @@ 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; } @@ -2327,16 +2323,16 @@ 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 +2354,8 @@ 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 +2364,7 @@ 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 +2409,8 @@ 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 +2430,8 @@ 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); @@ -2617,9 +2613,7 @@ void tcp_simple_retransmit(struct sock *sk) unsigned int mss = tcp_current_mss(sk); u32 prior_lost = tp->lost_out; - 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) { @@ -2713,7 +2707,7 @@ 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; @@ -2775,7 +2769,7 @@ 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); @@ -2805,9 +2799,9 @@ static void tcp_fastretrans_alert(struct sock *sk, const int acked, bool do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) && (tcp_fackets_out(tp) > tp->reordering)); - 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)) + if (!tp->sacked_out && tp->fackets_out) tp->fackets_out = 0; /* Now state machine starts. @@ -2874,6 +2868,7 @@ 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) @@ -2914,8 +2909,8 @@ 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)); @@ -3057,8 +3052,11 @@ 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 @@ -3074,11 +3072,11 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, struct tcp_sock *tp = tcp_sk(sk); u32 prior_sacked = tp->sacked_out; u32 reord = tp->packets_out; + 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; @@ -3086,7 +3084,7 @@ 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); u8 sacked = scb->sacked; u32 acked_pcount; @@ -3104,8 +3102,6 @@ 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); } @@ -3157,12 +3153,12 @@ 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) @@ -3254,12 +3250,14 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, 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(). @@ -3379,7 +3377,7 @@ 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) { @@ -3400,7 +3398,7 @@ 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! */ } @@ -3436,10 +3434,11 @@ 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; @@ -3447,16 +3446,16 @@ 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); } } @@ -3564,8 +3563,8 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) 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. @@ -3650,6 +3649,8 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una, &acked, &sack_state); + tcp_rack_update_reo_wnd(sk, &rs); + if (tp->tlp_high_seq) tcp_process_tlp_ack(sk, ack, flag); /* If needed, reset TLP/RTO timer; RACK may later override this. */ @@ -3679,8 +3680,7 @@ no_queue: * 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); @@ -3722,6 +3722,21 @@ 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. @@ -3829,6 +3844,9 @@ 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; } @@ -3996,6 +4014,8 @@ 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: @@ -4118,7 +4138,7 @@ 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)) @@ -4153,7 +4173,7 @@ 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)) @@ -4269,11 +4289,6 @@ 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 @@ -4289,7 +4304,6 @@ 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) @@ -4314,10 +4328,7 @@ 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; @@ -4342,7 +4353,7 @@ 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; @@ -4354,9 +4365,6 @@ 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"); @@ -4368,8 +4376,7 @@ 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) @@ -4410,7 +4417,7 @@ 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; @@ -4423,10 +4430,6 @@ 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); @@ -4454,7 +4457,7 @@ 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); @@ -4473,7 +4476,7 @@ 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; @@ -4505,7 +4508,7 @@ 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; } @@ -4518,9 +4521,7 @@ 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)) { @@ -4535,7 +4536,7 @@ 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: @@ -4557,7 +4558,7 @@ 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) { @@ -4721,7 +4722,7 @@ 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, @@ -4742,7 +4743,7 @@ 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; @@ -4750,7 +4751,7 @@ 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 @@ -4797,7 +4798,7 @@ 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; @@ -4869,26 +4870,19 @@ 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. @@ -4931,14 +4925,14 @@ 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 @@ -5113,7 +5107,7 @@ 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); @@ -5533,20 +5527,13 @@ 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)); @@ -5560,7 +5547,7 @@ 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; @@ -5595,9 +5582,8 @@ 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); @@ -5615,6 +5601,16 @@ 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) { @@ -5710,10 +5706,9 @@ 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) + if (tcp_is_sack(tp) && sock_net(sk)->ipv4.sysctl_tcp_fack) tcp_enable_fack(tp); - tcp_mtup_init(sk); tcp_sync_mss(sk, icsk->icsk_pmtu_cookie); tcp_initialize_rcv_mss(sk); @@ -5722,6 +5717,8 @@ 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); @@ -5939,15 +5936,18 @@ 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); @@ -5967,19 +5967,6 @@ 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); @@ -6076,6 +6063,7 @@ 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, @@ -6184,6 +6172,9 @@ 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, @@ -6359,7 +6350,7 @@ 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, |