diff --git a/include/net/tcp.h b/include/net/tcp.h index b92bdc7c92a9..0a4ed6e85c6f 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -719,6 +719,34 @@ static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) icsk->icsk_ca_ops->cwnd_event(sk, event); } +/* These functions determine how the current flow behaves in respect of SACK + * handling. SACK is negotiated with the peer, and therefore it can vary + * between different flows. + * + * tcp_is_sack - SACK enabled + * tcp_is_reno - No SACK + * tcp_is_fack - FACK enabled, implies SACK enabled + */ +static inline int tcp_is_sack(const struct tcp_sock *tp) +{ + return tp->rx_opt.sack_ok; +} + +static inline int tcp_is_reno(const struct tcp_sock *tp) +{ + return !tcp_is_sack(tp); +} + +static inline int tcp_is_fack(const struct tcp_sock *tp) +{ + return tp->rx_opt.sack_ok & 2; +} + +static inline void tcp_enable_fack(struct tcp_sock *tp) +{ + tp->rx_opt.sack_ok |= 2; +} + static inline unsigned int tcp_left_out(const struct tcp_sock *tp) { return tp->sacked_out + tp->lost_out; diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 7e740112b238..aff31427f525 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2014,7 +2014,7 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info) if (tp->rx_opt.tstamp_ok) info->tcpi_options |= TCPI_OPT_TIMESTAMPS; - if (tp->rx_opt.sack_ok) + if (tcp_is_sack(tp)) info->tcpi_options |= TCPI_OPT_SACK; if (tp->rx_opt.wscale_ok) { info->tcpi_options |= TCPI_OPT_WSCALE; diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index f8af018dd224..faba9beb3613 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -111,10 +111,6 @@ int sysctl_tcp_abc __read_mostly; #define FLAG_FORWARD_PROGRESS (FLAG_ACKED|FLAG_DATA_SACKED) #define FLAG_ANY_PROGRESS (FLAG_FORWARD_PROGRESS|FLAG_SND_UNA_ADVANCED) -#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 IsSackFrto() (sysctl_tcp_frto == 0x2) #define TCP_REMNANT (TCP_FLAG_FIN|TCP_FLAG_URG|TCP_FLAG_SYN|TCP_FLAG_PSH) @@ -860,6 +856,21 @@ void tcp_enter_cwr(struct sock *sk, const int set_ssthresh) } } +/* + * Packet counting of FACK is based on in-order assumptions, therefore TCP + * disables it when reordering is detected + */ +static void tcp_disable_fack(struct tcp_sock *tp) +{ + tp->rx_opt.sack_ok &= ~2; +} + +/* Take a notice that peer is sending DSACKs */ +static void tcp_dsack_seen(struct tcp_sock *tp) +{ + tp->rx_opt.sack_ok |= 4; +} + /* Initialize metrics on socket. */ static void tcp_init_metrics(struct sock *sk) @@ -881,7 +892,7 @@ static void tcp_init_metrics(struct sock *sk) } if (dst_metric(dst, RTAX_REORDERING) && tp->reordering != dst_metric(dst, RTAX_REORDERING)) { - tp->rx_opt.sack_ok &= ~2; + tcp_disable_fack(tp); tp->reordering = dst_metric(dst, RTAX_REORDERING); } @@ -943,9 +954,9 @@ static void tcp_update_reordering(struct sock *sk, const int metric, /* This exciting event is worth to be remembered. 8) */ if (ts) NET_INC_STATS_BH(LINUX_MIB_TCPTSREORDER); - else if (IsReno(tp)) + else if (tcp_is_reno(tp)) NET_INC_STATS_BH(LINUX_MIB_TCPRENOREORDER); - else if (IsFack(tp)) + else if (tcp_is_fack(tp)) NET_INC_STATS_BH(LINUX_MIB_TCPFACKREORDER); else NET_INC_STATS_BH(LINUX_MIB_TCPSACKREORDER); @@ -957,8 +968,7 @@ static void tcp_update_reordering(struct sock *sk, const int metric, tp->sacked_out, tp->undo_marker ? tp->undo_retrans : 0); #endif - /* Disable FACK yet. */ - tp->rx_opt.sack_ok &= ~2; + tcp_disable_fack(tp); } } @@ -1020,7 +1030,7 @@ static int tcp_check_dsack(struct tcp_sock *tp, struct sk_buff *ack_skb, if (before(start_seq_0, TCP_SKB_CB(ack_skb)->ack_seq)) { dup_sack = 1; - tp->rx_opt.sack_ok |= 4; + tcp_dsack_seen(tp); NET_INC_STATS_BH(LINUX_MIB_TCPDSACKRECV); } else if (num_sacks > 1) { u32 end_seq_1 = ntohl(get_unaligned(&sp[1].end_seq)); @@ -1029,7 +1039,7 @@ static int tcp_check_dsack(struct tcp_sock *tp, struct sk_buff *ack_skb, if (!after(end_seq_0, end_seq_1) && !before(start_seq_0, start_seq_1)) { dup_sack = 1; - tp->rx_opt.sack_ok |= 4; + tcp_dsack_seen(tp); NET_INC_STATS_BH(LINUX_MIB_TCPDSACKOFORECV); } } @@ -1326,7 +1336,7 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb, u32 prior_snd_ continue; if ((TCP_SKB_CB(skb)->sacked&TCPCB_SACKED_RETRANS) && after(lost_retrans, TCP_SKB_CB(skb)->ack_seq) && - (IsFack(tp) || + (tcp_is_fack(tp) || !before(lost_retrans, TCP_SKB_CB(skb)->ack_seq + tp->reordering * tp->mss_cache))) { @@ -1526,7 +1536,7 @@ static void tcp_enter_frto_loss(struct sock *sk, int allowed_segments, int flag) tp->lost_out = 0; tp->retrans_out = 0; - if (IsReno(tp)) + if (tcp_is_reno(tp)) tcp_reset_reno_sack(tp); tcp_for_write_queue(skb, sk) { @@ -1668,7 +1678,7 @@ static int tcp_check_sack_reneging(struct sock *sk) static inline int tcp_fackets_out(struct tcp_sock *tp) { - return IsReno(tp) ? tp->sacked_out+1 : tp->fackets_out; + return tcp_is_reno(tp) ? tp->sacked_out+1 : tp->fackets_out; } static inline int tcp_skb_timedout(struct sock *sk, struct sk_buff *skb) @@ -1872,7 +1882,7 @@ static void tcp_update_scoreboard(struct sock *sk) { struct tcp_sock *tp = tcp_sk(sk); - if (IsFack(tp)) { + if (tcp_is_fack(tp)) { int lost = tp->fackets_out - tp->reordering; if (lost <= 0) lost = 1; @@ -1886,7 +1896,7 @@ static void tcp_update_scoreboard(struct sock *sk) * Hence, we can detect timed out packets during fast * retransmit without falling to slow start. */ - if (!IsReno(tp) && tcp_head_timedout(sk)) { + if (!tcp_is_reno(tp) && tcp_head_timedout(sk)) { struct sk_buff *skb; skb = tp->scoreboard_skb_hint ? tp->scoreboard_skb_hint @@ -1938,7 +1948,7 @@ static void tcp_cwnd_down(struct sock *sk, int flag) int decr = tp->snd_cwnd_cnt + 1; if ((flag&(FLAG_ANY_PROGRESS|FLAG_DSACKING_ACK)) || - (IsReno(tp) && !(flag&FLAG_NOT_DUP))) { + (tcp_is_reno(tp) && !(flag&FLAG_NOT_DUP))) { tp->snd_cwnd_cnt = decr&1; decr >>= 1; @@ -2029,7 +2039,7 @@ static int tcp_try_undo_recovery(struct sock *sk) NET_INC_STATS_BH(LINUX_MIB_TCPFULLUNDO); tp->undo_marker = 0; } - if (tp->snd_una == tp->high_seq && IsReno(tp)) { + if (tp->snd_una == tp->high_seq && tcp_is_reno(tp)) { /* Hold old state until something *above* high_seq * is ACKed. For Reno it is MUST to prevent false * fast retransmits (RFC2582). SACK TCP is safe. */ @@ -2059,7 +2069,7 @@ static int tcp_try_undo_partial(struct sock *sk, int acked) { struct tcp_sock *tp = tcp_sk(sk); /* Partial ACK arrived. Force Hoe's retransmit. */ - int failed = IsReno(tp) || tp->fackets_out>tp->reordering; + int failed = tcp_is_reno(tp) || tp->fackets_out>tp->reordering; if (tcp_may_undo(tp)) { /* Plain luck! Hole if filled with delayed @@ -2104,7 +2114,7 @@ static int tcp_try_undo_loss(struct sock *sk) NET_INC_STATS_BH(LINUX_MIB_TCPLOSSUNDO); inet_csk(sk)->icsk_retransmits = 0; tp->undo_marker = 0; - if (!IsReno(tp)) + if (tcp_is_sack(tp)) tcp_set_ca_state(sk, TCP_CA_Open); return 1; } @@ -2251,14 +2261,14 @@ tcp_fastretrans_alert(struct sock *sk, int pkts_acked, int flag) if (!tp->undo_marker || /* For SACK case do not Open to allow to undo * catching for all duplicate ACKs. */ - IsReno(tp) || tp->snd_una != tp->high_seq) { + tcp_is_reno(tp) || tp->snd_una != tp->high_seq) { tp->undo_marker = 0; tcp_set_ca_state(sk, TCP_CA_Open); } break; case TCP_CA_Recovery: - if (IsReno(tp)) + if (tcp_is_reno(tp)) tcp_reset_reno_sack(tp); if (tcp_try_undo_recovery(sk)) return; @@ -2271,7 +2281,7 @@ tcp_fastretrans_alert(struct sock *sk, int pkts_acked, int flag) switch (icsk->icsk_ca_state) { case TCP_CA_Recovery: if (!(flag & FLAG_SND_UNA_ADVANCED)) { - if (IsReno(tp) && is_dupack) + if (tcp_is_reno(tp) && is_dupack) tcp_add_reno_sack(sk); } else do_lost = tcp_try_undo_partial(sk, pkts_acked); @@ -2288,7 +2298,7 @@ tcp_fastretrans_alert(struct sock *sk, int pkts_acked, int flag) return; /* Loss is undone; fall through to processing in Open state. */ default: - if (IsReno(tp)) { + if (tcp_is_reno(tp)) { if (flag & FLAG_SND_UNA_ADVANCED) tcp_reset_reno_sack(tp); if (is_dupack) @@ -2316,7 +2326,7 @@ tcp_fastretrans_alert(struct sock *sk, int pkts_acked, int flag) /* Otherwise enter Recovery state */ - if (IsReno(tp)) + if (tcp_is_reno(tp)) NET_INC_STATS_BH(LINUX_MIB_TCPRENORECOVERY); else NET_INC_STATS_BH(LINUX_MIB_TCPSACKRECOVERY); @@ -2573,7 +2583,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) tcp_ack_update_rtt(sk, acked, seq_rtt); tcp_ack_packets_out(sk); - if (IsReno(tp)) + if (tcp_is_reno(tp)) tcp_remove_reno_sacks(sk, pkts_acked); if (ca_ops->pkts_acked) { @@ -2599,7 +2609,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) 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) { + if (!tp->packets_out && tcp_is_sack(tp)) { const struct inet_connection_sock *icsk = inet_csk(sk); if (tp->lost_out) { printk(KERN_DEBUG "Leak l=%u %d\n", @@ -2779,7 +2789,7 @@ static int tcp_process_frto(struct sock *sk, int flag) return 1; } - if (!IsSackFrto() || IsReno(tp)) { + if (!IsSackFrto() || tcp_is_reno(tp)) { /* RFC4138 shortcoming in step 2; should also have case c): * ACK isn't duplicate nor advances window, e.g., opposite dir * data, winupdate @@ -3263,7 +3273,7 @@ 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->rx_opt.sack_ok) + if (tcp_is_sack(tp)) tcp_sack_reset(&tp->rx_opt); sk_stream_mem_reclaim(sk); @@ -3293,7 +3303,7 @@ static inline int tcp_sack_extend(struct tcp_sack_block *sp, u32 seq, u32 end_se static void tcp_dsack_set(struct tcp_sock *tp, u32 seq, u32 end_seq) { - if (tp->rx_opt.sack_ok && sysctl_tcp_dsack) { + if (tcp_is_sack(tp) && sysctl_tcp_dsack) { if (before(seq, tp->rcv_nxt)) NET_INC_STATS_BH(LINUX_MIB_TCPDSACKOLDSENT); else @@ -3323,7 +3333,7 @@ static void tcp_send_dupack(struct sock *sk, struct sk_buff *skb) NET_INC_STATS_BH(LINUX_MIB_DELAYEDACKLOST); tcp_enter_quickack_mode(sk); - if (tp->rx_opt.sack_ok && sysctl_tcp_dsack) { + if (tcp_is_sack(tp) && sysctl_tcp_dsack) { u32 end_seq = TCP_SKB_CB(skb)->end_seq; if (after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) @@ -3639,7 +3649,7 @@ drop: if (!skb_peek(&tp->out_of_order_queue)) { /* Initial out of order segment, build 1 SACK. */ - if (tp->rx_opt.sack_ok) { + if (tcp_is_sack(tp)) { tp->rx_opt.num_sacks = 1; tp->rx_opt.dsack = 0; tp->rx_opt.eff_sacks = 1; @@ -3704,7 +3714,7 @@ drop: } add_sack: - if (tp->rx_opt.sack_ok) + if (tcp_is_sack(tp)) tcp_sack_new_ofo_skb(sk, seq, end_seq); } } @@ -3893,7 +3903,7 @@ 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->rx_opt.sack_ok) + if (tcp_is_sack(tp)) tcp_sack_reset(&tp->rx_opt); sk_stream_mem_reclaim(sk); } @@ -4594,8 +4604,8 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, tp->tcp_header_len = sizeof(struct tcphdr); } - if (tp->rx_opt.sack_ok && sysctl_tcp_fack) - tp->rx_opt.sack_ok |= 2; + if (tcp_is_sack(tp) && sysctl_tcp_fack) + tcp_enable_fack(tp); tcp_mtup_init(sk); tcp_sync_mss(sk, icsk->icsk_pmtu_cookie); diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c index fdfe89fe646b..b61b76847ad9 100644 --- a/net/ipv4/tcp_minisocks.c +++ b/net/ipv4/tcp_minisocks.c @@ -445,7 +445,7 @@ struct sock *tcp_create_openreq_child(struct sock *sk, struct request_sock *req, newtp->rx_opt.tstamp_ok = ireq->tstamp_ok; if ((newtp->rx_opt.sack_ok = ireq->sack_ok) != 0) { if (sysctl_tcp_fack) - newtp->rx_opt.sack_ok |= 2; + tcp_enable_fack(newtp); } newtp->window_clamp = req->window_clamp; newtp->rcv_ssthresh = req->rcv_wnd; diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index a92fad55cd32..a3679174e78a 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -737,7 +737,7 @@ int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len, unsigned int mss if (diff > 0) { /* Adjust Reno SACK estimate. */ - if (!tp->rx_opt.sack_ok) { + if (tcp_is_reno(tp)) { tcp_dec_pcount_approx_int(&tp->sacked_out, diff); tcp_verify_left_out(tp); } @@ -1728,7 +1728,7 @@ static void tcp_retrans_try_collapse(struct sock *sk, struct sk_buff *skb, int m if (TCP_SKB_CB(next_skb)->sacked&TCPCB_LOST) tp->lost_out -= tcp_skb_pcount(next_skb); /* Reno case is special. Sigh... */ - if (!tp->rx_opt.sack_ok && tp->sacked_out) + if (tcp_is_reno(tp) && tp->sacked_out) tcp_dec_pcount_approx(&tp->sacked_out, next_skb); /* Not quite right: it can be > snd.fack, but @@ -1976,7 +1976,7 @@ void tcp_xmit_retransmit_queue(struct sock *sk) return; /* No forward retransmissions in Reno are possible. */ - if (!tp->rx_opt.sack_ok) + if (tcp_is_reno(tp)) return; /* Yeah, we have to make difficult choice between forward transmission diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c index e9b151b3a598..d8970ecfcfc8 100644 --- a/net/ipv4/tcp_timer.c +++ b/net/ipv4/tcp_timer.c @@ -315,7 +315,7 @@ static void tcp_retransmit_timer(struct sock *sk) if (icsk->icsk_retransmits == 0) { if (icsk->icsk_ca_state == TCP_CA_Disorder || icsk->icsk_ca_state == TCP_CA_Recovery) { - if (tp->rx_opt.sack_ok) { + if (tcp_is_sack(tp)) { if (icsk->icsk_ca_state == TCP_CA_Recovery) NET_INC_STATS_BH(LINUX_MIB_TCPSACKRECOVERYFAIL); else