X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/b0d623f7f2ae71ed96e60569f61f9a9a27016e80..3e170ce000f1506b7b5d2c5c7faec85ceabb573d:/bsd/netinet/tcp_var.h diff --git a/bsd/netinet/tcp_var.h b/bsd/netinet/tcp_var.h index 0fa518d78..1ec0559ee 100644 --- a/bsd/netinet/tcp_var.h +++ b/bsd/netinet/tcp_var.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2000-2008 Apple Inc. All rights reserved. + * Copyright (c) 2000-2015 Apple Inc. All rights reserved. * * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * @@ -63,9 +63,11 @@ #ifndef _NETINET_TCP_VAR_H_ #define _NETINET_TCP_VAR_H_ +#include #include #include #include +#include #include #if defined(__LP64__) @@ -79,11 +81,76 @@ struct name { \ #define _TCPCB_LIST_HEAD(name, type) LIST_HEAD(name, type) #endif -#define TCP_RETRANSHZ 10 /* tcp retrans timer (100ms) per hz */ - #ifdef KERNEL_PRIVATE + +#define TCP_RETRANSHZ 1000 /* granularity of TCP timestamps, 1ms */ +/* Minimum time quantum within which the timers are coalesced */ +#define TCP_TIMER_10MS_QUANTUM (TCP_RETRANSHZ/100) /* every 10ms */ +#define TCP_TIMER_100MS_QUANTUM (TCP_RETRANSHZ/10) /* every 100ms */ +#define TCP_TIMER_500MS_QUANTUM (TCP_RETRANSHZ/2) /* every 500ms */ + +#define TCP_RETRANSHZ_TO_USEC 1000 + #define N_TIME_WAIT_SLOTS 128 /* must be power of 2 */ +/* Base RTT is stored for N_MIN_RTT_HISTORY slots. This is used to + * estimate expected minimum RTT for delay based congestion control + * algorithms. + */ +#define N_RTT_BASE 5 + +/* Always allow at least 4 packets worth of recv window when adjusting + * recv window using inter-packet arrival jitter. + */ +#define MIN_IAJ_WIN 4 + +/* A variation in delay of this many milliseconds is tolerable. This limit has to + * be low but greater than zero. We also use standard deviation on jitter to adjust + * this limit for different link and connection types. + */ +#define ALLOWED_IAJ 5 + +/* Ignore the first few packets on a connection until the ACK clock gets going + */ +#define IAJ_IGNORE_PKTCNT 40 + +/* Let the accumulated IAJ value increase by this threshold at most. This limit + * will control how many ALLOWED_IAJ measurements a receiver will have to see + * before opening the receive window + */ +#define ACC_IAJ_HIGH_THRESH 100 + +/* When accumulated IAJ reaches this value, the receiver starts to react by + * closing the window + */ +#define ACC_IAJ_REACT_LIMIT 200 + +/* If the number of small packets (smaller than IAJ packet size) seen on a + * connection is more than this threshold, reset the size and learn it again. + * This is needed because the sender might send smaller segments after PMTU + * discovery and the receiver has to learn the new size. + */ +#define RESET_IAJ_SIZE_THRESH 20 + +/* + * Adaptive timeout is a read/write timeout specified by the application to + * get a socket event when the transport layer detects a stall in data + * transfer. The value specified is the number of probes that can be sent + * to the peer before generating an event. Since it is not specified as + * a time value, the timeout will adjust based on the RTT seen on the link. + * The timeout will start only when there is an indication that the read/write + * operation is not making progress. + * + * If a write operation stalls, the probe will be retransmission of data. + * If a read operation stalls, the probe will be a keep-alive packet. + * + * The maximum value of adaptive timeout is set to 10 which will allow + * transmission of enough number of probes to the peer. + */ +#define TCP_ADAPTIVE_TIMEOUT_MAX 10 + +#define TCP_CONNECTIVITY_PROBES_MAX 5 + /* * Kernel variables for tcp. */ @@ -96,11 +163,6 @@ struct tseg_qent { struct mbuf *tqe_m; /* mbuf contains packet */ }; LIST_HEAD(tsegqe_head, tseg_qent); -extern int tcp_reass_maxseg; -extern int tcp_reass_qsize; -#ifdef MALLOC_DECLARE -MALLOC_DECLARE(M_TSEGQ); -#endif struct sackblk { tcp_seq start; /* start seq no. of sack block */ @@ -111,6 +173,7 @@ struct sackhole { tcp_seq start; /* start seq no. of hole */ tcp_seq end; /* end seq no. */ tcp_seq rxmit; /* next seq. no in hole to be retransmitted */ + u_int32_t rxmit_start; /* timestamp of first retransmission */ TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */ }; @@ -119,13 +182,66 @@ struct sackhint { int sack_bytes_rexmit; }; +struct tcp_rxt_seg { + tcp_seq rx_start; + tcp_seq rx_end; + u_int16_t rx_count; + u_int16_t rx_flags; +#define TCP_RXT_SPURIOUS 0x1 /* received DSACK notification */ +#define TCP_RXT_DSACK_FOR_TLP 0x2 + SLIST_ENTRY(tcp_rxt_seg) rx_link; +}; + struct tcptemp { u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */ struct tcphdr tt_t; }; +struct bwmeas { + tcp_seq bw_start; /* start of bw measurement */ + uint32_t bw_ts; /* timestamp when bw measurement started */ + uint32_t bw_size; /* burst size in bytes for this bw measurement */ + uint32_t bw_minsizepkts; /* Min burst size as segments */ + uint32_t bw_maxsizepkts; /* Max burst size as segments */ + uint32_t bw_minsize; /* Min size in bytes */ + uint32_t bw_maxsize; /* Max size in bytes */ + uint32_t bw_sndbw; /* Measured send bw */ +}; + +/* MPTCP Data sequence map entry */ +struct mpt_dsn_map { + uint64_t mpt_dsn; /* data seq num recvd */ + uint32_t mpt_sseq; /* relative subflow # */ + uint16_t mpt_len; /* length of mapping */ + uint16_t mpt_csum; /* checksum value if on */ +}; #define tcp6cb tcpcb /* for KAME src sync over BSD*'s */ +struct tcp_ccstate { + union { + struct tcp_cubic_state { + u_int32_t tc_last_max; /* cwnd at last loss */ + u_int32_t tc_epoch_start; /* TS of last loss */ + u_int32_t tc_origin_point; /* window at the start of an epoch */ + u_int32_t tc_tcp_win; /* computed tcp win */ + u_int32_t tc_tcp_bytes_acked; /* bytes acked */ + u_int32_t tc_target_win; /* cubic target win */ + u_int32_t tc_avg_lastmax; /* Average of last max */ + u_int32_t tc_mean_deviation; /* Mean absolute deviation */ + float tc_epoch_period; /* K parameter */ + } _cubic_state_; +#define cub_last_max __u__._cubic_state_.tc_last_max +#define cub_epoch_start __u__._cubic_state_.tc_epoch_start +#define cub_origin_point __u__._cubic_state_.tc_origin_point +#define cub_tcp_win __u__._cubic_state_.tc_tcp_win +#define cub_tcp_bytes_acked __u__._cubic_state_.tc_tcp_bytes_acked +#define cub_epoch_period __u__._cubic_state_.tc_epoch_period +#define cub_target_win __u__._cubic_state_.tc_target_win +#define cub_avg_lastmax __u__._cubic_state_.tc_avg_lastmax +#define cub_mean_dev __u__._cubic_state_.tc_mean_deviation + } __u__; +}; + /* * Tcp control block, one per tcp; fields: * Organized for 16 byte cacheline efficiency. @@ -133,13 +249,12 @@ struct tcptemp { struct tcpcb { struct tsegqe_head t_segq; int t_dupacks; /* consecutive dup acks recd */ - struct tcptemp *unused; /* unused now: was t_template */ - - int t_timer[TCPT_NTIMERS]; /* tcp timers */ + int t_state; /* state of this connection */ + uint32_t t_timer[TCPT_NTIMERS]; /* tcp timers */ + struct tcptimerentry tentry; /* entry in timer list */ struct inpcb *t_inpcb; /* back pointer to internet pcb */ - int t_state; /* state of this connection */ - u_int t_flags; + uint32_t t_flags; #define TF_ACKNOW 0x00001 /* ack peer immediately */ #define TF_DELACK 0x00002 /* ack, but try to delay it */ #define TF_NODELAY 0x00004 /* don't delay packets to coalesce */ @@ -157,23 +272,21 @@ struct tcpcb { #define TF_RCVD_CC 0x04000 /* a CC was received in SYN */ #define TF_SENDCCNEW 0x08000 /* send CCnew instead of CC in SYN */ #define TF_MORETOCOME 0x10000 /* More data to be appended to sock */ -#define TF_LQ_OVERFLOW 0x20000 /* UNUSED listen queue overflow */ +#define TF_LOCAL 0x20000 /* connection to a host on local link */ #define TF_RXWIN0SENT 0x40000 /* sent a receiver win 0 in response */ #define TF_SLOWLINK 0x80000 /* route is a on a modem speed link */ - - #define TF_LASTIDLE 0x100000 /* connection was previously idle */ #define TF_FASTRECOVERY 0x200000 /* in NewReno Fast Recovery */ #define TF_WASFRECOVERY 0x400000 /* was in NewReno Fast Recovery */ #define TF_SIGNATURE 0x800000 /* require MD5 digests (RFC2385) */ #define TF_MAXSEGSNT 0x1000000 /* last segment sent was a full segment */ -#define TF_SENDINPROG 0x2000000 /* send is in progress */ +#define TF_ENABLE_ECN 0x2000000 /* Enable ECN */ #define TF_PMTUD 0x4000000 /* Perform Path MTU Discovery for this connection */ #define TF_CLOSING 0x8000000 /* pending tcp close */ #define TF_TSO 0x10000000 /* TCP Segment Offloading is enable on this connection */ #define TF_BLACKHOLE 0x20000000 /* Path MTU Discovery Black Hole detection */ - - int t_force; /* 1 if forcing out a byte */ +#define TF_TIMER_ONLIST 0x40000000 /* pcb is on tcp_timer_list */ +#define TF_STRETCHACK 0x80000000 /* receiver is going to delay acks */ tcp_seq snd_una; /* send unacknowledged */ tcp_seq snd_max; /* highest sequence number sent; @@ -192,36 +305,39 @@ struct tcpcb { u_int32_t rcv_wnd; /* receive window */ tcp_seq rcv_up; /* receive urgent pointer */ - u_int32_t snd_wnd; /* send window */ - u_int32_t snd_cwnd; /* congestion-controlled window */ - u_int32_t snd_bwnd; /* bandwidth-controlled window */ - u_int32_t snd_ssthresh; /* snd_cwnd size threshold for + u_int32_t snd_wnd; /* send window */ + u_int32_t snd_cwnd; /* congestion-controlled window */ + u_int32_t snd_ssthresh; /* snd_cwnd size threshold for * for slow start exponential to * linear switch */ - u_int32_t snd_bandwidth; /* calculated bandwidth or 0 */ tcp_seq snd_recover; /* for use in NewReno Fast Recovery */ u_int t_maxopd; /* mss plus options */ - u_int32_t t_rcvtime; /* inactivity time */ - u_int32_t t_starttime; /* time connection was established */ - int t_rtttime; /* round trip time */ + u_int32_t t_rcvtime; /* time at which a packet was received */ + u_int32_t t_starttime; /* time connection was established */ + int t_rtttime; /* tcp clock when rtt calculation was started */ tcp_seq t_rtseq; /* sequence number being timed */ - int t_bw_rtttime; /* used for bandwidth calculation */ - tcp_seq t_bw_rtseq; /* used for bandwidth calculation */ + u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */ + u_int32_t rfbuf_cnt; /* recv buffer autoscaling byte count */ int t_rxtcur; /* current retransmit value (ticks) */ u_int t_maxseg; /* maximum segment size */ int t_srtt; /* smoothed round-trip time */ int t_rttvar; /* variance in round-trip time */ - int t_rxtshift; /* log(2) of rexmt exp. backoff */ - u_int t_rttmin; /* minimum rtt allowed */ - u_int t_rttbest; /* best rtt we've seen */ - u_int32_t t_rttupdated; /* number of times rtt sampled */ - u_int32_t max_sndwnd; /* largest window peer has offered */ + u_int64_t t_accsleep_ms; /* accumulated sleep time since last boot */ + u_int16_t t_reassqlen; /* length of reassembly queue */ + u_int16_t t_rxtshift; /* log(2) of rexmt exp. backoff */ + u_int32_t t_rttmin; /* minimum rtt allowed */ + u_int32_t t_rttbest; /* best rtt we've seen */ + u_int32_t t_rttcur; /* most recent value of rtt */ + u_int32_t t_rttupdated; /* number of times rtt sampled */ + u_int32_t t_rxt_conndroptime; /* retxmt conn gets dropped after this time, when set */ + u_int32_t t_rxtstart; /* time at which retransmission started */ + u_int32_t max_sndwnd; /* largest window peer has offered */ int t_softerror; /* possible error not yet reported */ /* out-of-band data */ @@ -230,78 +346,364 @@ struct tcpcb { #define TCPOOB_HAVEDATA 0x01 #define TCPOOB_HADDATA 0x02 /* RFC 1323 variables */ - u_char snd_scale; /* window scaling for send window */ - u_char rcv_scale; /* window scaling for recv window */ - u_char request_r_scale; /* pending window scaling */ - u_char requested_s_scale; - u_int32_t ts_recent; /* timestamp echo data */ - - u_int32_t ts_recent_age; /* when last updated */ + u_int8_t snd_scale; /* window scaling for send window */ + u_int8_t rcv_scale; /* window scaling for recv window */ + u_int8_t request_r_scale; /* pending window scaling */ + u_int8_t requested_s_scale; + u_int8_t tcp_cc_index; /* index of congestion control algorithm */ + u_int8_t t_adaptive_rtimo; /* Read timeout used as a multiple of RTT */ + u_int8_t t_adaptive_wtimo; /* Write timeout used as a multiple of RTT */ + u_int8_t t_stretchack_delayed; /* stretch ack delayed */ + +/* State for limiting early retransmits when SACK is not enabled */ + u_int16_t t_early_rexmt_count; /* count of early rexmts */ + u_int32_t t_early_rexmt_win; /* window for limiting early rexmts */ + + u_int32_t ts_recent; /* timestamp echo data */ + + u_int32_t ts_recent_age; /* when last updated */ tcp_seq last_ack_sent; /* RFC 1644 variables */ tcp_cc cc_send; /* send connection count */ tcp_cc cc_recv; /* receive connection count */ /* RFC 3465 variables */ - u_int32_t t_bytes_acked; /* ABC "bytes_acked" parameter */ -/* experimental */ - u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */ - u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */ - u_int32_t t_badrxtwin; /* window for retransmit recovery */ + u_int32_t t_bytes_acked; /* ABC "bytes_acked" parameter */ - int t_keepidle; /* keepalive idle timer (override global if > 0) */ int t_lastchain; /* amount of packets chained last time around */ - int t_unacksegs; /* received but unacked segments: used for delaying acks */ - - -/* 3529618 MSS overload prevention */ - u_int32_t rcv_reset; - u_int32_t rcv_pps; - u_int32_t rcv_byps; - u_int32_t rcv_maxbyps; - tcp_seq snd_high; /* for use in NewReno Fast Recovery */ - tcp_seq snd_high_prev; /* snd_high prior to retransmit */ - + u_int16_t t_unacksegs; /* received but unacked segments for delaying acks */ + u_int8_t t_rexmtthresh; /* duplicate ack threshold for entering fast recovery */ + u_int8_t t_rtimo_probes; /* number of adaptive rtimo probes sent */ + u_int32_t t_persist_timeout; /* ZWP persistence limit as set by PERSIST_TIMEOUT */ + u_int32_t t_persist_stop; /* persistence limit deadline if triggered by ZWP */ + u_int32_t t_notsent_lowat; /* Low water for not sent data */ + +/* Receiver state for stretch-ack algorithm */ + u_int32_t rcv_unackwin; /* to measure win for stretching acks */ + u_int32_t rcv_by_unackwin; /* bytes seen during the last ack-stretching win */ + u_int32_t rcv_nostrack_ts; /* timestamp when stretch ack was disabled automatically */ + u_int16_t rcv_waitforss; /* wait for packets during slow-start */ + u_int16_t ecn_flags; +#define TE_SETUPSENT 0x01 /* Indicate we have sent ECN-SETUP SYN or SYN-ACK */ +#define TE_SETUPRECEIVED 0x02 /* Indicate we have received ECN-SETUP SYN or SYN-ACK */ +#define TE_SENDIPECT 0x04 /* Indicate we haven't sent or received non-ECN-setup SYN or SYN-ACK */ +#define TE_SENDCWR 0x08 /* Indicate that the next non-retransmit should have the TCP CWR flag set */ +#define TE_SENDECE 0x10 /* Indicate that the next packet should have the TCP ECE flag set */ +#define TE_INRECOVERY 0x20 /* connection entered recovery after receiving ECE */ +#define TE_RECV_ECN_CE 0x40 /* Received IPTOS_ECN_CE marking atleast once */ +#define TE_RECV_ECN_ECE 0x80 /* Received ECE marking atleast once */ +#define TE_ECN_ON (TE_SETUPSENT | TE_SETUPRECEIVED) /* Indicate ECN was successfully negotiated on a connection) */ + +/* state for bad retransmit recovery */ + u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */ + u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */ tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */ - u_char snd_limited; /* segments limited transmitted */ -/* anti DoS counters */ - u_int32_t rcv_second; /* start of interval second */ -/* SACK related state */ - int sack_enable; /* enable SACK for this connection */ - int snd_numholes; /* number of holes seen by sender */ + int t_srtt_prev; /* srtt prior to retransmit */ + int t_rttvar_prev; /* rttvar prior to retransmit */ + u_int32_t t_badrexmt_time; /* bad rexmt detection time */ + +/* Packet reordering metric */ + u_int16_t t_reorderwin; /* Reordering late time offset */ +/* SACK related state */ + int16_t snd_numholes; /* number of holes seen by sender */ + tcp_seq sack_newdata; /* New data xmitted in this recovery + episode starts at this seq number */ TAILQ_HEAD(sackhole_head, sackhole) snd_holes; /* SACK scoreboard (sorted) */ tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/ int rcv_numsacks; /* # distinct sack blks present */ struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */ - tcp_seq sack_newdata; /* New data xmitted in this recovery - episode starts at this seq number */ struct sackhint sackhint; /* SACK scoreboard hint */ - int t_rttlow; /* smallest observerved RTT */ - u_long ecn_flags; -#define TE_SETUPSENT 0x01 /* Indicate we have sent ECN-SETUP SYN or SYN-ACK */ -#define TE_SETUPRECEIVED 0x02 /* Indicate we have received ECN-SETUP SYN or SYN-ACK */ -#define TE_SENDIPECT 0x04 /* Indicate we haven't sent or received non-ECN-setup SYN or SYN-ACK */ -#define TE_SENDCWR 0x08 /* Indicate that the next non-retransmit should have the TCP CWR flag set */ -#define TE_SENDECE 0x10 /* Indicate that the next packet should have the TCP ECE flag set */ -#if TRAFFIC_MGT - u_int32_t tot_recv_snapshot; /* snapshot of global total pkts received */ - u_int32_t bg_recv_snapshot; /* snapshot of global background pkts received */ -#endif /* TRAFFIC_MGT */ - u_int32_t t_pktlist_sentlen; /* total bytes in transmit chain */ struct mbuf *t_pktlist_head; /* First packet in transmit chain */ struct mbuf *t_pktlist_tail; /* Last packet in transmit chain */ + u_int32_t t_pktlist_sentlen; /* total bytes in transmit chain */ - int t_keepinit; /* connection timeout, i.e. idle time in SYN_SENT or SYN_RECV state */ - u_int32_t tso_max_segment_size; /* TCP Segment Offloading maximum segment unit for NIC */ - u_int t_pmtud_saved_maxopd; /* MSS saved before performing PMTU-D BlackHole detection */ + u_int32_t t_keepidle; /* keepalive idle timer (override global if > 0) */ + u_int32_t t_keepinit; /* connection timeout, i.e. idle time + in SYN_SENT or SYN_RECV state */ + u_int32_t t_keepintvl; /* interval between keepalives */ + u_int32_t t_keepcnt; /* number of keepalives before close */ + + u_int32_t tso_max_segment_size; /* TSO maximum segment unit for NIC */ + u_int16_t t_pmtud_lastseg_size; /* size of the last sent segment */ + u_int16_t t_pmtud_saved_maxopd; /* MSS saved before performing PMTU-D BlackHole detection */ + u_int32_t t_pmtud_start_ts; /* Time of PMTUD blackhole detection */ + + struct + { + u_int32_t rxduplicatebytes; + u_int32_t rxoutoforderbytes; + u_int32_t txretransmitbytes; + u_int8_t synrxtshift; + u_int8_t unused; + u_int16_t unused_pad_to_8; + } t_stat; + + /* Background congestion related state */ + uint32_t rtt_hist[N_RTT_BASE]; /* history of minimum RTT */ + uint32_t rtt_count; /* Number of RTT samples in recent base history */ + uint32_t bg_ssthresh; /* Slow start threshold until delay increases */ + uint32_t t_flagsext; /* Another field to accommodate more flags */ +#define TF_RXTFINDROP 0x1 /* Drop conn after retransmitting FIN 3 times */ +#define TF_RCVUNACK_WAITSS 0x2 /* set when the receiver should not stretch acks */ +#define TF_BWMEAS_INPROGRESS 0x4 /* Indicate BW meas is happening */ +#define TF_MEASURESNDBW 0x8 /* Measure send bw on this connection */ +#define TF_LRO_OFFLOADED 0x10 /* Connection LRO offloaded */ +#define TF_SACK_ENABLE 0x20 /* SACK is enabled */ +#define TF_RECOMPUTE_RTT 0x40 /* recompute RTT after spurious retransmit */ +#define TF_DETECT_READSTALL 0x80 /* Used to detect a stall during read operation */ +#define TF_RECV_THROTTLE 0x100 /* Input throttling active */ +#define TF_NOSTRETCHACK 0x200 /* ack every other packet */ +#define TF_STREAMEOW 0x400 /* Last packet was small indicating end of write */ +#define TF_NOTIMEWAIT 0x800 /* Avoid going into time-wait */ +#define TF_SENT_TLPROBE 0x1000 /* Sent data in PTO */ +#define TF_PKTS_REORDERED 0x2000 /* Detected reordering */ +#define TF_DELAY_RECOVERY 0x4000 /* delay fast recovery */ +#define TF_FORCE 0x8000 /* force 1 byte out */ +#define TF_DISABLE_STRETCHACK 0x10000 /* auto-disable stretch ack */ +#define TF_NOBLACKHOLE_DETECTION 0x20000 /* Disable PMTU blackhole detection */ +#define TF_DISABLE_DSACK 0x40000 /* Ignore DSACK due to n/w duplication */ +#define TF_RESCUE_RXT 0x80000 /* SACK rescue retransmit */ +#define TF_CWND_NONVALIDATED 0x100000 /* cwnd non validated */ +#define TF_PROBING 0x200000 /* Trigger probe timeout */ +#define TF_FASTOPEN 0x400000 /* TCP Fastopen is enabled */ + +#if TRAFFIC_MGT + /* Inter-arrival jitter related state */ + uint32_t iaj_rcv_ts; /* tcp clock when the first packet was received */ + uint16_t iaj_size; /* Size of packet for iaj measurement */ + uint8_t iaj_small_pkt; /* Count of packets smaller than iaj_size */ + uint8_t t_pipeack_ind; /* index for next pipeack sample */ + uint16_t iaj_pktcnt; /* packet count, to avoid throttling initially */ + uint16_t acc_iaj; /* Accumulated iaj */ + tcp_seq iaj_rwintop; /* recent max advertised window */ + uint32_t avg_iaj; /* Mean */ + uint32_t std_dev_iaj; /* Standard deviation */ +#endif /* TRAFFIC_MGT */ + struct bwmeas *t_bwmeas; /* State for bandwidth measurement */ + uint32_t t_lropktlen; /* Bytes in a LRO frame */ + tcp_seq t_idleat; /* rcv_nxt at idle time */ + TAILQ_ENTRY(tcpcb) t_twentry; /* link for time wait queue */ + struct tcp_ccstate *t_ccstate; /* congestion control related state */ +/* Tail loss probe related state */ + tcp_seq t_tlphighrxt; /* snd_nxt after PTO */ + u_int32_t t_tlpstart; /* timestamp at PTO */ +/* DSACK data receiver state */ + tcp_seq t_dsack_lseq; /* DSACK left sequence */ + tcp_seq t_dsack_rseq; /* DSACK right sequence */ +/* DSACK data sender state */ + SLIST_HEAD(tcp_rxt_seghead, tcp_rxt_seg) t_rxt_segments; + tcp_seq t_dsack_lastuna; /* snd_una when last recovery episode started */ +/* state for congestion window validation (draft-ietf-tcpm-newcwv-07) */ +#define TCP_PIPEACK_SAMPLE_COUNT 3 + u_int32_t t_pipeack_sample[TCP_PIPEACK_SAMPLE_COUNT]; /* pipeack, bytes acked within RTT */ + tcp_seq t_pipeack_lastuna; /* una when pipeack measurement started */ + u_int32_t t_pipeack; + u_int32_t t_lossflightsize; + +#if MPTCP + u_int32_t t_mpflags; /* flags for multipath TCP */ + +#define TMPF_PREESTABLISHED 0x00000001 /* conn in pre-established state */ +#define TMPF_SENT_KEYS 0x00000002 /* indicates that keys were sent */ +#define TMPF_MPTCP_TRUE 0x00000004 /* negotiated MPTCP successfully */ +#define TMPF_MPTCP_RCVD_KEY 0x00000008 /* state for 3-way handshake */ +#define TMPF_SND_MPPRIO 0x00000010 /* send priority of subflow */ +#define TMPF_SND_REM_ADDR 0x00000020 /* initiate address removal */ +#define TMPF_UNUSED 0x00000040 /* address addition acked by peer */ +#define TMPF_JOINED_FLOW 0x00000080 /* Indicates additional flow */ +#define TMPF_BACKUP_PATH 0x00000100 /* Indicates backup path */ +#define TMPF_MPTCP_ACKNOW 0x00000200 /* Send Data ACK */ +#define TMPF_SEND_DSN 0x00000400 /* Send DSN mapping */ +#define TMPF_SEND_DFIN 0x00000800 /* Send Data FIN */ +#define TMPF_RECV_DFIN 0x00001000 /* Recv Data FIN */ +#define TMPF_SENT_JOIN 0x00002000 /* Sent Join */ +#define TMPF_RECVD_JOIN 0x00004000 /* Received Join */ +#define TMPF_RESET 0x00008000 /* Send RST */ +#define TMPF_TCP_FALLBACK 0x00010000 /* Fallback to TCP */ +#define TMPF_FASTCLOSE 0x00020000 /* Send Fastclose option */ +#define TMPF_EMBED_DSN 0x00040000 /* tp has DSN mapping */ +#define TMPF_MPTCP_READY 0x00080000 /* Can send DSS options on data */ +#define TMPF_INFIN_SENT 0x00100000 /* Sent infinite mapping */ +#define TMPF_SND_MPFAIL 0x00200000 /* Received mapping csum failure */ +#define TMPF_FASTJOIN_SEND 0x00400000 /* Fast join early data send */ +#define TMPF_FASTJOINBY2_SEND 0x00800000 /* Fast join send after 3 WHS */ +#define TMPF_MPCAP_RETRANSMIT 0x01000000 /* Retransmission of 3rd ACK */ + + tcp_seq t_mpuna; /* unacknowledged sequence */ + void *t_mptcb; /* pointer to MPTCP TCB */ + struct mpt_dsn_map t_rcv_map; /* Receive mapping list */ + u_int8_t t_local_aid; /* Addr Id for authentication */ + u_int8_t t_rem_aid; /* Addr ID of another subflow */ + u_int8_t t_mprxtshift; /* join retransmission */ +#endif /* MPTCP */ + +#define TFO_F_OFFER_COOKIE 0x01 /* We will offer a cookie */ +#define TFO_F_COOKIE_VALID 0x02 /* The received cookie is valid */ +#define TFO_F_COOKIE_REQ 0x04 /* Client requested a new cookie */ +#define TFO_F_COOKIE_SENT 0x08 /* Client did send a cookie in the SYN */ +#define TFO_F_SYN_LOSS 0x10 /* A SYN-loss triggered a fallback to regular TCP on the client-side */ +#define TFO_F_NO_RCVPROBING 0x20 /* This network is guaranteed to support TFO in the downstream direction */ +#define TFO_F_NO_SNDPROBING 0x40 /* This network is guaranteed to support TFO in the upstream direction */ + u_int8_t t_tfo_flags; +#define TFO_S_SYNDATA_RCV 0x01 /* SYN+data has been received */ +#define TFO_S_COOKIEREQ_RECV 0x02 /* TFO-cookie request received */ +#define TFO_S_COOKIE_SENT 0x04 /* TFO-cookie announced in SYN/ACK */ +#define TFO_S_COOKIE_INVALID 0x08 /* Received TFO-cookie is invalid */ +#define TFO_S_COOKIE_REQ 0x10 /* TFO-cookie requested within the SYN */ +#define TFO_S_COOKIE_RCV 0x20 /* TFO-cookie received in SYN/ACK */ +#define TFO_S_SYN_DATA_SENT 0x40 /* SYN+data sent */ +#define TFO_S_SYN_DATA_ACKED 0x80 /* SYN+data has been acknowledged in SYN/ACK */ +#define TFO_S_SYN_LOSS 0x0100 /* SYN+TFO has been lost - fallback to regular TCP */ + u_int16_t t_tfo_stats; + + u_int8_t t_tfo_probes; /* TFO-probes we did send */ +/* + * This here is the TFO-probing state-machine. Transitions are as follows: + * + * Current state: PROBE_NONE + * Event: SYN+DATA acknowledged + * Action: Transition to PROBE_PROBING and set keepalive-timer + * + * Current state: PROBE_PROBING (initial state) + * Event: Receive data + * Action: Transition to PROBE_NONE and cancel keepalive-timer + * Event: Receive ACK that does not indicate a hole + * Action: Transition to PROBE_NONE and cancel keepalive-timer + * Event: Receive ACK that indicates a hole + * Action: Transition to PROBE_WAIT_DATA and set a short timer + * to wait for the final segment. + * Event: Keepalive-timeout (did not receive any segment) + * Action: Signal ETIMEDOUT as with regular keepalive-timers + * + * Current state: PROBE_WAIT_DATA + * Event: Receive data + * Action: Transition to PROBE_NONE and cancel keepalive-timer + * Event: Data-timeout (did not receive the expected data) + * Action: Signal ENODATA up to the app and close everything. + */ +#define TFO_PROBE_NONE 0 /* Not probing now */ +#define TFO_PROBE_PROBING 1 /* Sending out TCP-keepalives waiting for reply */ +#define TFO_PROBE_WAIT_DATA 2 /* Received reply, waiting for data */ + u_int8_t t_tfo_probe_state; }; #define IN_FASTRECOVERY(tp) (tp->t_flags & TF_FASTRECOVERY) -#define ENTER_FASTRECOVERY(tp) tp->t_flags |= TF_FASTRECOVERY -#define EXIT_FASTRECOVERY(tp) tp->t_flags &= ~TF_FASTRECOVERY +#define SACK_ENABLED(tp) (tp->t_flagsext & TF_SACK_ENABLE) + +/* + * If the connection is in a throttled state due to advisory feedback from + * the interface output queue, reset that state. We do this in favor + * of entering recovery because the data transfer during recovery + * should be just a trickle and it will help to improve performance. + * We also do not want to back off twice in the same RTT. + */ +#define ENTER_FASTRECOVERY(_tp_) do { \ + (_tp_)->t_flags |= TF_FASTRECOVERY; \ + if (INP_IS_FLOW_CONTROLLED((_tp_)->t_inpcb)) \ + inp_reset_fc_state((_tp_)->t_inpcb); \ + if (!SLIST_EMPTY(&tp->t_rxt_segments)) \ + tcp_rxtseg_clean(tp); \ +} while(0) + +#define EXIT_FASTRECOVERY(_tp_) do { \ + (_tp_)->t_flags &= ~TF_FASTRECOVERY; \ + (_tp_)->t_dupacks = 0; \ + (_tp_)->t_rexmtthresh = tcprexmtthresh; \ + (_tp_)->t_bytes_acked = 0; \ + (_tp_)->ecn_flags &= ~TE_INRECOVERY; \ + (_tp_)->t_timer[TCPT_PTO] = 0; \ + (_tp_)->t_flagsext &= ~TF_RESCUE_RXT; \ + (_tp_)->t_lossflightsize = 0; \ +} while(0) + +/* + * When the number of duplicate acks received is less than + * the retransmit threshold, use Limited Transmit algorithm + */ +extern int tcprexmtthresh; +#define ALLOW_LIMITED_TRANSMIT(_tp_) \ + ((_tp_)->t_dupacks > 0 && \ + (_tp_)->t_dupacks < (_tp_)->t_rexmtthresh && \ + ((_tp_)->t_flagsext & (TF_PKTS_REORDERED|TF_DELAY_RECOVERY)) \ + != (TF_PKTS_REORDERED|TF_DELAY_RECOVERY)) + +/* + * This condition is true is timestamp option is supported + * on a connection. + */ +#define TSTMP_SUPPORTED(_tp_) \ + (((_tp_)->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP)) == \ + (TF_REQ_TSTMP|TF_RCVD_TSTMP)) + +/* + * This condition is true if window scale option is supported + * on a connection + */ +#define TCP_WINDOW_SCALE_ENABLED(_tp_) \ + (((_tp_)->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == \ + (TF_RCVD_SCALE|TF_REQ_SCALE)) + +/* Is ECN enabled end-to-end */ +#define TCP_ECN_ENABLED(_tp_) \ + (((_tp_)->ecn_flags & (TE_ECN_ON)) == (TE_ECN_ON)) + +/* + * Gives number of bytes acked by this ack + */ +#define BYTES_ACKED(_th_, _tp_) \ + ((_th_)->th_ack - (_tp_)->snd_una) + +/* Returns true if a DSACK option should be sent */ +#define TCP_SEND_DSACK_OPT(_tp_) \ + ((_tp_)->t_dsack_lseq > 0 && (_tp_)->t_dsack_rseq > 0) +/* Check if DSACK option should be processed */ +#define TCP_DSACK_ENABLED(tp) (tcp_dsack_enable == 1 && \ + !(tp->t_flagsext & TF_DISABLE_DSACK)) + +/* + * Returns true if a DSACK sequence is within the max send window that will + * be accepted. In order to set a window to validate sequence numbers, the + * max send window within which a DSACK option is processed is limited. + * + * We need to choose a maximum window to check if the sequence number is + * within the window. One arbitrary choice is 256 * MSS because if the + * window is as large as 256 segments it might be big enough to ignore the + * DSACK option. Choosing a much larger limit means that the memory for + * retransmit segments can be held for a longer time. + */ +#define TCP_DSACK_MAX_SEND_WINDOW(_tp_) ((_tp_)->t_maxseg << 8) +#define TCP_DSACK_SEQ_IN_WINDOW(_tp_, _seq_, _una_) \ + (SEQ_LEQ((_seq_), (_tp_)->snd_max) && \ + SEQ_GEQ((_seq_), ((_una_) - TCP_DSACK_MAX_SEND_WINDOW(_tp_)))) + + +enum tcp_cc_event { + TCP_CC_CWND_INIT, /* 0 */ + TCP_CC_INSEQ_ACK_RCVD, /* 1 */ + TCP_CC_ACK_RCVD, /* 2 */ + TCP_CC_ENTER_FASTRECOVERY, /* 3 */ + TCP_CC_IN_FASTRECOVERY, /* 4 */ + TCP_CC_EXIT_FASTRECOVERY, /* 5 */ + TCP_CC_PARTIAL_ACK, /* 6 */ + TCP_CC_IDLE_TIMEOUT, /* 7 */ + TCP_CC_REXMT_TIMEOUT, /* 8 */ + TCP_CC_ECN_RCVD, /* 9 */ + TCP_CC_BAD_REXMT_RECOVERY, /* 10 */ + TCP_CC_OUTPUT_ERROR, /* 11 */ + TCP_CC_CHANGE_ALGO, /* 12 */ + TCP_CC_FLOW_CONTROL, /* 13 */ + TCP_CC_SUSPEND, /* 14 */ + TCP_CC_LIMITED_TRANSMIT, /* 15 */ + TCP_CC_EARLY_RETRANSMIT, /* 16 */ + TCP_CC_TLP_RECOVERY, /* 17 */ + TCP_CC_TLP_RECOVER_LASTPACKET, /* 18 */ + TCP_CC_DELAY_FASTRECOVERY, /* 19 */ + TCP_CC_TLP_IN_FASTRECOVERY, /* 20 */ + TCP_CC_DSACK_BAD_REXMT /* 21 */ +}; /* * Structure to hold TCP options that are only used during segment @@ -317,12 +719,16 @@ struct tcpopt { #define TOF_SIGNATURE 0x0040 /* signature option present */ #define TOF_SIGLEN 0x0080 /* signature length valid (RFC2385) */ #define TOF_SACK 0x0100 /* Peer sent SACK option */ +#define TOF_MPTCP 0x0200 /* MPTCP options to be dropped */ +#define TOF_TFO 0x0400 /* TFO cookie option present */ +#define TOF_TFOREQ 0x0800 /* TFO cookie request present */ u_int32_t to_tsval; u_int32_t to_tsecr; u_int16_t to_mss; u_int8_t to_requested_s_scale; u_int8_t to_nsacks; /* number of SACK blocks */ u_char *to_sacks; /* pointer to the first SACK blocks */ + u_char *to_tfo; /* pointer to the TFO cookie */ }; /* @@ -346,20 +752,26 @@ struct rmxp_tao { #define sototcpcb(so) (intotcpcb(sotoinpcb(so))) /* - * The smoothed round-trip time and estimated variance + * The rtt measured is in milliseconds as the timestamp granularity is + * a millisecond. The smoothed round-trip time and estimated variance * are stored as fixed point numbers scaled by the values below. * For convenience, these scales are also used in smoothing the average * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed). - * With these scales, srtt has 3 bits to the right of the binary point, - * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the + * With these scales, srtt has 5 bits to the right of the binary point, + * and thus an "ALPHA" of 0.875. rttvar has 4 bits to the right of the * binary point, and is smoothed with an ALPHA of 0.75. */ #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */ -#define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */ -#define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */ -#define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */ +#define TCP_RTT_SHIFT 5 /* shift for srtt; 5 bits frac. */ +#define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 4 bits */ +#define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 4 bits */ #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */ +/* TFO-specific defines */ +#define TFO_COOKIE_LEN_MIN 4 +#define TFO_COOKIE_LEN_DEFAULT 8 +#define TFO_COOKIE_LEN_MAX 16 + /* * The initial retransmission should happen at rtt + 4 * rttvar. * Because of the way we do the smoothing, srtt and rttvar @@ -399,7 +811,7 @@ struct tcpcb { int t_dupacks; /* consecutive dup acks recd */ u_int32_t unused; /* unused now: was t_template */ - int t_timer[TCPT_NTIMERS]; /* tcp timers */ + int t_timer[TCPT_NTIMERS_EXT]; /* tcp timers */ _TCPCB_PTR(struct inpcb *) t_inpcb; /* back pointer to internet pcb */ int t_state; /* state of this connection */ @@ -452,7 +864,7 @@ struct tcpcb { */ u_int t_maxopd; /* mss plus options */ - u_int32_t t_rcvtime; /* inactivity time */ + u_int32_t t_rcvtime; /* time at which a packet was received */ u_int32_t t_starttime; /* time connection was established */ int t_rtttime; /* round trip time */ tcp_seq t_rtseq; /* sequence number being timed */ @@ -492,6 +904,9 @@ struct tcpcb { u_int32_t t_badrxtwin; /* window for retransmit recovery */ }; +#define tcps_ecn_setup tcps_ecn_client_success +#define tcps_sent_cwr tcps_ecn_recv_ece +#define tcps_sent_ece tcps_ecn_sent_ece /* * TCP statistics. @@ -589,15 +1004,138 @@ struct tcpstat { /* SACK related stats */ u_int32_t tcps_sack_recovery_episode; /* SACK recovery episodes */ - u_int32_t tcps_sack_rexmits; /* SACK rexmit segments */ - u_int32_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */ - u_int32_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */ - u_int32_t tcps_sack_send_blocks; /* SACK blocks (options) sent */ - u_int32_t tcps_sack_sboverflow; /* SACK sendblock overflow */ + u_int32_t tcps_sack_rexmits; /* SACK rexmit segments */ + u_int32_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */ + u_int32_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */ + u_int32_t tcps_sack_send_blocks; /* SACK blocks (options) sent */ + u_int32_t tcps_sack_sboverflow; /* SACK sendblock overflow */ -#if TRAFFIC_MGT u_int32_t tcps_bg_rcvtotal; /* total background packets received */ -#endif /* TRAFFIC_MGT */ + u_int32_t tcps_rxtfindrop; /* drop conn after retransmitting FIN */ + u_int32_t tcps_fcholdpacket; /* packets withheld because of flow control */ + + /* LRO related stats */ + u_int32_t tcps_coalesced_pack; /* number of coalesced packets */ + u_int32_t tcps_flowtbl_full; /* times flow table was full */ + u_int32_t tcps_flowtbl_collision; /* collisions in flow tbl */ + u_int32_t tcps_lro_twopack; /* 2 packets coalesced */ + u_int32_t tcps_lro_multpack; /* 3 or 4 pkts coalesced */ + u_int32_t tcps_lro_largepack; /* 5 or more pkts coalesced */ + + u_int32_t tcps_limited_txt; /* Limited transmit used */ + u_int32_t tcps_early_rexmt; /* Early retransmit used */ + u_int32_t tcps_sack_ackadv; /* Cumulative ack advanced along with sack */ + + /* Checksum related stats */ + u_int32_t tcps_rcv_swcsum; /* tcp swcksum (inbound), packets */ + u_int32_t tcps_rcv_swcsum_bytes; /* tcp swcksum (inbound), bytes */ + u_int32_t tcps_rcv6_swcsum; /* tcp6 swcksum (inbound), packets */ + u_int32_t tcps_rcv6_swcsum_bytes; /* tcp6 swcksum (inbound), bytes */ + u_int32_t tcps_snd_swcsum; /* tcp swcksum (outbound), packets */ + u_int32_t tcps_snd_swcsum_bytes; /* tcp swcksum (outbound), bytes */ + u_int32_t tcps_snd6_swcsum; /* tcp6 swcksum (outbound), packets */ + u_int32_t tcps_snd6_swcsum_bytes; /* tcp6 swcksum (outbound), bytes */ + u_int32_t tcps_msg_unopkts; /* unordered packet on TCP msg stream */ + u_int32_t tcps_msg_unoappendfail; /* failed to append unordered pkt */ + u_int32_t tcps_msg_sndwaithipri; /* send is waiting for high priority data */ + + /* MPTCP Related stats */ + u_int32_t tcps_invalid_mpcap; /* Invalid MPTCP capable opts */ + u_int32_t tcps_invalid_joins; /* Invalid MPTCP joins */ + u_int32_t tcps_mpcap_fallback; /* TCP fallback in primary */ + u_int32_t tcps_join_fallback; /* No MPTCP in secondary */ + u_int32_t tcps_estab_fallback; /* DSS option dropped */ + u_int32_t tcps_invalid_opt; /* Catchall error stat */ + u_int32_t tcps_mp_outofwin; /* Packet lies outside the + shared recv window */ + u_int32_t tcps_mp_reducedwin; /* Reduced subflow window */ + u_int32_t tcps_mp_badcsum; /* Bad DSS csum */ + u_int32_t tcps_mp_oodata; /* Out of order data */ + u_int32_t tcps_mp_switches; /* number of subflow switch */ + u_int32_t tcps_mp_rcvtotal; /* number of rcvd packets */ + u_int32_t tcps_mp_rcvbytes; /* number of bytes received */ + u_int32_t tcps_mp_sndpacks; /* number of data packs sent */ + u_int32_t tcps_mp_sndbytes; /* number of bytes sent */ + u_int32_t tcps_join_rxmts; /* join ack retransmits */ + u_int32_t tcps_tailloss_rto; /* RTO due to tail loss */ + u_int32_t tcps_reordered_pkts; /* packets reorderd */ + u_int32_t tcps_recovered_pkts; /* recovered after loss */ + u_int32_t tcps_pto; /* probe timeout */ + u_int32_t tcps_rto_after_pto; /* RTO after a probe */ + u_int32_t tcps_tlp_recovery; /* TLP induced fast recovery */ + u_int32_t tcps_tlp_recoverlastpkt; /* TLP recoverd last pkt */ + u_int32_t tcps_ecn_client_success; /* client-side connection negotiated ECN */ + u_int32_t tcps_ecn_recv_ece; /* ECE received, sent CWR */ + u_int32_t tcps_ecn_sent_ece; /* Sent ECE notification */ + u_int32_t tcps_detect_reordering; /* Detect pkt reordering */ + u_int32_t tcps_delay_recovery; /* Delay fast recovery */ + u_int32_t tcps_avoid_rxmt; /* Retransmission was avoided */ + u_int32_t tcps_unnecessary_rxmt; /* Retransmission was not needed */ + u_int32_t tcps_nostretchack; /* disabled stretch ack algorithm on a connection */ + u_int32_t tcps_rescue_rxmt; /* SACK rescue retransmit */ + u_int32_t tcps_pto_in_recovery; /* rescue retransmit in fast recovery */ + u_int32_t tcps_pmtudbh_reverted; /* PMTU Blackhole detection, segment size reverted */ + + /* DSACK related statistics */ + u_int32_t tcps_dsack_disable; /* DSACK disabled due to n/w duplication */ + u_int32_t tcps_dsack_ackloss; /* ignore DSACK due to ack loss */ + u_int32_t tcps_dsack_badrexmt; /* DSACK based bad rexmt recovery */ + u_int32_t tcps_dsack_sent; /* Sent DSACK notification */ + u_int32_t tcps_dsack_recvd; /* Received a valid DSACK option */ + u_int32_t tcps_dsack_recvd_old; /* Received an out of window DSACK option */ + + /* MPTCP Subflow selection stats */ + u_int32_t tcps_mp_sel_symtomsd; /* By symptomsd */ + u_int32_t tcps_mp_sel_rtt; /* By RTT comparison */ + u_int32_t tcps_mp_sel_rto; /* By RTO comparison */ + u_int32_t tcps_mp_sel_peer; /* By peer's output pattern */ + u_int32_t tcps_mp_num_probes; /* Number of probes sent */ + u_int32_t tcps_mp_verdowngrade; /* MPTCP version downgrade */ + u_int32_t tcps_drop_after_sleep; /* drop after long AP sleep */ + u_int32_t tcps_probe_if; /* probe packets after interface availability */ + u_int32_t tcps_probe_if_conflict; /* Can't send probe packets for interface */ + + u_int32_t tcps_ecn_client_setup; /* Attempted ECN setup from client side */ + u_int32_t tcps_ecn_server_setup; /* Attempted ECN setup from server side */ + u_int32_t tcps_ecn_server_success; /* server-side connection negotiated ECN */ + u_int32_t tcps_ecn_lost_synack; /* Lost SYN-ACK with ECN setup */ + u_int32_t tcps_ecn_lost_syn; /* Lost SYN with ECN setup */ + u_int32_t tcps_ecn_not_supported; /* Server did not support ECN setup */ + u_int32_t tcps_ecn_recv_ce; /* Received CE from the network */ + u_int32_t tcps_ecn_conn_recv_ce; /* Number of connections received CE atleast once */ + u_int32_t tcps_ecn_conn_recv_ece; /* Number of connections received ECE atleast once */ + u_int32_t tcps_ecn_conn_plnoce; /* Number of connections that received no CE and sufferred packet loss */ + u_int32_t tcps_ecn_conn_pl_ce; /* Number of connections that received CE and sufferred packet loss */ + u_int32_t tcps_ecn_conn_nopl_ce; /* Number of connections that received CE and sufferred no packet loss */ + + /* TFO-related statistics */ + u_int32_t tcps_tfo_syn_data_rcv; /* Received a SYN+data with valid cookie */ + u_int32_t tcps_tfo_cookie_req_rcv;/* Received a TFO cookie-request */ + u_int32_t tcps_tfo_cookie_sent; /* Offered a TFO-cookie to the client */ + u_int32_t tcps_tfo_cookie_invalid;/* Received an invalid TFO-cookie */ + u_int32_t tcps_tfo_cookie_req; /* Cookie requested with the SYN */ + u_int32_t tcps_tfo_cookie_rcv; /* Cookie received in a SYN/ACK */ + u_int32_t tcps_tfo_syn_data_sent; /* SYN+data+cookie sent */ + u_int32_t tcps_tfo_syn_data_acked;/* SYN+data has been acknowledged */ + u_int32_t tcps_tfo_syn_loss; /* SYN+TFO has been lost and we fallback */ + u_int32_t tcps_tfo_blackhole; /* TFO got blackholed by a middlebox. */ +}; + +struct tcpstat_local { + u_int64_t badformat; + u_int64_t unspecv6; + u_int64_t synfin; + u_int64_t badformatipsec; + u_int64_t noconnnolist; + u_int64_t noconnlist; + u_int64_t listbadsyn; + u_int64_t icmp6unreach; + u_int64_t deprecate6; + u_int64_t ooopacket; + u_int64_t rstinsynrcv; + u_int64_t dospacket; + u_int64_t cleanup; + u_int64_t synwindow; }; #pragma pack(4) @@ -624,7 +1162,6 @@ struct xtcpcb { u_quad_t xt_alignment_hack; }; -#if !CONFIG_EMBEDDED struct xtcpcb64 { u_int32_t xt_len; @@ -633,7 +1170,7 @@ struct xtcpcb64 { u_int64_t t_segq; int t_dupacks; /* consecutive dup acks recd */ - int t_timer[TCPT_NTIMERS]; /* tcp timers */ + int t_timer[TCPT_NTIMERS_EXT]; /* tcp timers */ int t_state; /* state of this connection */ u_int t_flags; @@ -665,7 +1202,7 @@ struct xtcpcb64 { */ u_int t_maxopd; /* mss plus options */ - u_int32_t t_rcvtime; /* inactivity time */ + u_int32_t t_rcvtime; /* time at which a packet was received */ u_int32_t t_starttime; /* time connection was established */ int t_rtttime; /* round trip time */ tcp_seq t_rtseq; /* sequence number being timed */ @@ -705,7 +1242,86 @@ struct xtcpcb64 { u_quad_t xt_alignment_hack; }; -#endif /* !CONFIG_EMBEDDED */ + +#ifdef PRIVATE + +struct xtcpcb_n { + u_int32_t xt_len; + u_int32_t xt_kind; /* XSO_TCPCB */ + + u_int64_t t_segq; + int t_dupacks; /* consecutive dup acks recd */ + + int t_timer[TCPT_NTIMERS_EXT]; /* tcp timers */ + + int t_state; /* state of this connection */ + u_int t_flags; + + int t_force; /* 1 if forcing out a byte */ + + tcp_seq snd_una; /* send unacknowledged */ + tcp_seq snd_max; /* highest sequence number sent; + * used to recognize retransmits + */ + tcp_seq snd_nxt; /* send next */ + tcp_seq snd_up; /* send urgent pointer */ + + tcp_seq snd_wl1; /* window update seg seq number */ + tcp_seq snd_wl2; /* window update seg ack number */ + tcp_seq iss; /* initial send sequence number */ + tcp_seq irs; /* initial receive sequence number */ + + tcp_seq rcv_nxt; /* receive next */ + tcp_seq rcv_adv; /* advertised window */ + u_int32_t rcv_wnd; /* receive window */ + tcp_seq rcv_up; /* receive urgent pointer */ + + u_int32_t snd_wnd; /* send window */ + u_int32_t snd_cwnd; /* congestion-controlled window */ + u_int32_t snd_ssthresh; /* snd_cwnd size threshold for + * for slow start exponential to + * linear switch + */ + u_int t_maxopd; /* mss plus options */ + + u_int32_t t_rcvtime; /* time at which a packet was received */ + u_int32_t t_starttime; /* time connection was established */ + int t_rtttime; /* round trip time */ + tcp_seq t_rtseq; /* sequence number being timed */ + + int t_rxtcur; /* current retransmit value (ticks) */ + u_int t_maxseg; /* maximum segment size */ + int t_srtt; /* smoothed round-trip time */ + int t_rttvar; /* variance in round-trip time */ + + int t_rxtshift; /* log(2) of rexmt exp. backoff */ + u_int t_rttmin; /* minimum rtt allowed */ + u_int32_t t_rttupdated; /* number of times rtt sampled */ + u_int32_t max_sndwnd; /* largest window peer has offered */ + + int t_softerror; /* possible error not yet reported */ + /* out-of-band data */ + char t_oobflags; /* have some */ + char t_iobc; /* input character */ + /* RFC 1323 variables */ + u_char snd_scale; /* window scaling for send window */ + u_char rcv_scale; /* window scaling for recv window */ + u_char request_r_scale; /* pending window scaling */ + u_char requested_s_scale; + u_int32_t ts_recent; /* timestamp echo data */ + + u_int32_t ts_recent_age; /* when last updated */ + tcp_seq last_ack_sent; + /* RFC 1644 variables */ + tcp_cc cc_send; /* send connection count */ + tcp_cc cc_recv; /* receive connection count */ + tcp_seq snd_recover; /* for use in fast recovery */ + /* experimental */ + u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */ + u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */ +}; + +#endif /* PRIVATE */ #pragma pack() @@ -727,7 +1343,9 @@ struct xtcpcb64 { #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */ #define TCPCTL_MAXID 14 -#ifdef KERNEL_PRIVATE +#ifdef BSD_KERNEL_PRIVATE +#include + #define TCP_PKTLIST_CLEAR(tp) { \ (tp)->t_pktlist_head = (tp)->t_pktlist_tail = NULL; \ (tp)->t_lastchain = (tp)->t_pktlist_sentlen = 0; \ @@ -759,21 +1377,47 @@ extern struct inpcbinfo tcbinfo; extern struct tcpstat tcpstat; /* tcp statistics */ extern int tcp_mssdflt; /* XXX */ extern int tcp_minmss; -extern int tcp_minmssoverload; -extern int tcp_do_newreno; +#define TCP_FASTOPEN_SERVER 0x01 +#define TCP_FASTOPEN_CLIENT 0x02 + +extern int tcp_tfo_halfcnt; +extern int tcp_tfo_backlog; +extern int tcp_fastopen; +extern int tcp_tfo_fallback_min; extern int ss_fltsz; extern int ss_fltsz_local; -#ifdef __APPLE__ +extern int tcp_do_rfc3390; /* Calculate ss_fltsz according to RFC 3390 */ +extern int tcp_do_rfc1323; +extern int target_qdelay; extern u_int32_t tcp_now; /* for RFC 1323 timestamps */ +extern struct timeval tcp_uptime; +extern lck_spin_t *tcp_uptime_lock; extern int tcp_delack_enabled; -#endif /* __APPLE__ */ - extern int tcp_do_sack; /* SACK enabled/disabled */ +extern int tcp_do_rfc3465; +extern int tcp_do_rfc3465_lim2; +extern int maxseg_unacked; +extern int tcp_use_newreno; +extern struct zone *tcp_reass_zone; +extern struct zone *tcp_rxt_seg_zone; +extern int tcp_ecn_outbound; +extern int tcp_ecn_inbound; + #if CONFIG_IFEF_NOWINDOWSCALE extern int tcp_obey_ifef_nowindowscale; #endif +struct protosw; +struct domain; + +struct tcp_respond_args { + unsigned int ifscope; + unsigned int nocell:1, + noexpensive:1, + awdl_unrestricted:1; +}; + void tcp_canceltimers(struct tcpcb *); struct tcpcb * tcp_close(struct tcpcb *); @@ -782,10 +1426,10 @@ int tcp_ctloutput(struct socket *, struct sockopt *); struct tcpcb * tcp_drop(struct tcpcb *, int); void tcp_drain(void); -void tcp_fasttimo(void *); +void tcp_getrt_rtt(struct tcpcb *tp, struct rtentry *rt); struct rmxp_tao * tcp_gettaocache(struct inpcb *); -void tcp_init(void) __attribute__((section("__TEXT, initcode"))); +void tcp_init(struct protosw *, struct domain *); void tcp_input(struct mbuf *, int); void tcp_mss(struct tcpcb *, int, unsigned int); int tcp_mssopt(struct tcpcb *); @@ -794,20 +1438,27 @@ void tcp_mtudisc(struct inpcb *, int); struct tcpcb * tcp_newtcpcb(struct inpcb *); int tcp_output(struct tcpcb *); -void tcp_respond(struct tcpcb *, void *, - struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int, - unsigned int); -struct rtentry *tcp_rtlookup(struct inpcb *, unsigned int); +void tcp_respond(struct tcpcb *, void *, struct tcphdr *, struct mbuf *, + tcp_seq, tcp_seq, int, struct tcp_respond_args *); +struct rtentry * + tcp_rtlookup(struct inpcb *, unsigned int); void tcp_setpersist(struct tcpcb *); -void tcp_slowtimo(void); -struct tcptemp * - tcp_maketemplate(struct tcpcb *); +void tcp_gc(struct inpcbinfo *); +void tcp_itimer(struct inpcbinfo *ipi); +void tcp_check_timer_state(struct tcpcb *tp); +void tcp_run_timerlist(void *arg1, void *arg2); + +struct tcptemp *tcp_maketemplate(struct tcpcb *); void tcp_fillheaders(struct tcpcb *, void *, void *); -struct tcpcb * - tcp_timers(struct tcpcb *, int); +struct tcpcb *tcp_timers(struct tcpcb *, int); void tcp_trace(int, int, struct tcpcb *, void *, struct tcphdr *, int); -void tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); +void tcp_sack_doack(struct tcpcb *, struct tcpopt *, struct tcphdr *, + u_int32_t *); +extern boolean_t tcp_sack_process_dsack(struct tcpcb *, struct tcpopt *, + struct tcphdr *); +int tcp_detect_bad_rexmt(struct tcpcb *, struct tcphdr *, struct tcpopt *, + u_int32_t rxtime); void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend); void tcp_clean_sackreport(struct tcpcb *tp); void tcp_sack_adjust(struct tcpcb *tp); @@ -816,10 +1467,35 @@ void tcp_sack_partialack(struct tcpcb *, struct tcphdr *); void tcp_free_sackholes(struct tcpcb *tp); int32_t tcp_sbspace(struct tcpcb *tp); void tcp_set_tso(struct tcpcb *tp, struct ifnet *ifp); +void tcp_reset_stretch_ack(struct tcpcb *tp); +extern void tcp_get_ports_used(u_int32_t, int, u_int32_t, bitstr_t *); +uint32_t tcp_count_opportunistic(unsigned int ifindex, u_int32_t flags); +uint32_t tcp_find_anypcb_byaddr(struct ifaddr *ifa); +void tcp_set_max_rwinscale(struct tcpcb *tp, struct socket *so); +struct bwmeas* tcp_bwmeas_alloc(struct tcpcb *tp); +void tcp_bwmeas_free(struct tcpcb *tp); +extern int32_t timer_diff(uint32_t t1, uint32_t toff1, uint32_t t2, uint32_t toff2); + +extern void tcp_set_background_cc(struct socket *); +extern void tcp_set_foreground_cc(struct socket *); +extern void tcp_set_recv_bg(struct socket *); +extern void tcp_clear_recv_bg(struct socket *); +extern boolean_t tcp_sack_byte_islost(struct tcpcb *tp); +#define IS_TCP_RECV_BG(_so) \ + ((_so)->so_traffic_mgt_flags & TRAFFIC_MGT_TCP_RECVBG) +#if TRAFFIC_MGT +#define CLEAR_IAJ_STATE(_tp_) (_tp_)->iaj_rcv_ts = 0 +void reset_acc_iaj(struct tcpcb *tp); +#endif /* TRAFFIC_MGT */ int tcp_lock (struct socket *, int, void *); int tcp_unlock (struct socket *, int, void *); +void calculate_tcp_clock(void); + +extern void tcp_keepalive_reset(struct tcpcb *); +extern uint32_t get_base_rtt(struct tcpcb *tp); + #ifdef _KERN_LOCKS_H_ lck_mtx_t * tcp_getlock (struct socket *, int); #else @@ -832,6 +1508,34 @@ extern u_int32_t tcp_sendspace; extern u_int32_t tcp_recvspace; tcp_seq tcp_new_isn(struct tcpcb *); -#endif /* KERNEL_RPIVATE */ +extern int tcp_input_checksum(int, struct mbuf *, struct tcphdr *, int, int); +extern void tcp_getconninfo(struct socket *, struct conninfo_tcp *); +extern void add_to_time_wait(struct tcpcb *, uint32_t delay); +extern void tcp_pmtud_revert_segment_size(struct tcpcb *tp); +extern void tcp_rxtseg_insert(struct tcpcb *, tcp_seq, tcp_seq); +extern struct tcp_rxt_seg *tcp_rxtseg_find(struct tcpcb *, tcp_seq, tcp_seq); +extern void tcp_rxtseg_clean(struct tcpcb *); +extern boolean_t tcp_rxtseg_detect_bad_rexmt(struct tcpcb *, tcp_seq); +extern boolean_t tcp_rxtseg_dsack_for_tlp(struct tcpcb *); +extern u_int32_t tcp_rxtseg_total_size(struct tcpcb *tp); +extern void tcp_rexmt_save_state(struct tcpcb *tp); +extern void tcp_interface_send_probe(u_int16_t if_index_available); +extern void tcp_probe_connectivity(struct ifnet *ifp, u_int32_t enable); +extern void tcp_get_connectivity_status(struct tcpcb *, + struct tcp_conn_status *); + +extern boolean_t tfo_enabled(const struct tcpcb *tp); +extern void tcp_disable_tfo(struct tcpcb *tp); +extern void tcp_tfo_gen_cookie(struct inpcb *inp, u_char *out, size_t blk_size); +#define TCP_FASTOPEN_KEYLEN 16 + +#if MPTCP +extern int mptcp_input_preproc(struct tcpcb *, struct mbuf *, int); +extern void mptcp_output_csum(struct tcpcb *, struct mbuf *, int32_t, unsigned, + u_int64_t, u_int32_t *); +extern int mptcp_adj_mss(struct tcpcb *, boolean_t); +extern void mptcp_insert_rmap(struct tcpcb *, struct mbuf *); +#endif +#endif /* BSD_KERNEL_RPIVATE */ #endif /* _NETINET_TCP_VAR_H_ */