]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet/tcp_var.h
xnu-3247.1.106.tar.gz
[apple/xnu.git] / bsd / netinet / tcp_var.h
index bf9fb3f20cf4f5d3a99586945e05a6dd64e8cd34..1ec0559eeaf32ee6c9b577f05a6a93cc1fe10789 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2014 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2015 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
 
 #ifndef _NETINET_TCP_VAR_H_
 #define _NETINET_TCP_VAR_H_
+#include <sys/types.h>
 #include <sys/appleapiopts.h>
 #include <sys/queue.h>
 #include <netinet/in_pcb.h>
+#include <netinet/tcp.h>
 #include <netinet/tcp_timer.h>
 
 #if defined(__LP64__)
@@ -145,7 +147,9 @@ struct name {                               \
  * 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_ADAPTIVE_TIMEOUT_MAX        10
+
+#define        TCP_CONNECTIVITY_PROBES_MAX     5
 
 /*
  * Kernel variables for tcp.
@@ -178,6 +182,16 @@ 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;
@@ -235,11 +249,11 @@ struct tcp_ccstate {
 struct tcpcb {
        struct  tsegqe_head t_segq;
        int     t_dupacks;              /* consecutive dup acks recd */
+       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 */
        uint32_t        t_flags;
 #define        TF_ACKNOW       0x00001         /* ack peer immediately */
 #define        TF_DELACK       0x00002         /* ack, but try to delay it */
@@ -291,9 +305,9 @@ 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_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
                                         */
@@ -314,15 +328,16 @@ struct tcpcb {
        int     t_srtt;                 /* smoothed round-trip time */
        int     t_rttvar;               /* variance in round-trip time */
 
+       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_int   t_rttmin;               /* minimum rtt allowed */
-       u_int   t_rttbest;              /* best rtt we've seen */
-       u_int   t_rttcur;               /* most recent value of rtt */
-       u_int32_t       t_rttupdated;           /* number of times rtt sampled */
+       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 */
+       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 */
@@ -331,9 +346,9 @@ struct tcpcb {
 #define        TCPOOB_HAVEDATA 0x01
 #define        TCPOOB_HADDATA  0x02
 /* RFC 1323 variables */
-       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        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 */
@@ -344,21 +359,21 @@ struct tcpcb {
        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;      /* timestamp echo data */
 
-       u_int32_t       ts_recent_age;          /* when last updated */
+       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 */
+       u_int32_t       t_bytes_acked;  /* ABC "bytes_acked" parameter */
 
        int     t_lastchain;            /* amount of packets chained last time around */
        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_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 */
 
@@ -374,6 +389,8 @@ struct tcpcb {
 #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 */
@@ -389,18 +406,18 @@ struct tcpcb {
 
 /* 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 */
        
-       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 */
 
        u_int32_t       t_keepidle;     /* keepalive idle timer (override global if > 0) */
        u_int32_t       t_keepinit;     /* connection timeout, i.e. idle time 
@@ -409,7 +426,8 @@ struct tcpcb {
        u_int32_t       t_keepcnt;      /* number of keepalives before close */
 
        u_int32_t       tso_max_segment_size;   /* TSO maximum segment unit for NIC */
-       u_int32_t       t_pmtud_saved_maxopd;   /* MSS saved before performing PMTU-D BlackHole detection */
+       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
@@ -445,12 +463,18 @@ struct tcpcb {
 #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 */
-       uint16_t        iaj_small_pkt;          /* Count of packets smaller than iaj_size */
+       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 */
@@ -465,6 +489,19 @@ struct tcpcb {
 /* 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 */
 
@@ -492,14 +529,64 @@ struct tcpcb {
 #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 */
 
-       void                    *t_mptcb;       /* pointer to MPTCP TCB */
        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)
@@ -516,6 +603,8 @@ struct tcpcb {
        (_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 {           \
@@ -525,6 +614,8 @@ struct tcpcb {
        (_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)
 
 /*
@@ -546,34 +637,72 @@ extern int tcprexmtthresh;
        (((_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,
-       TCP_CC_INSEQ_ACK_RCVD,
-       TCP_CC_ACK_RCVD,
-       TCP_CC_ENTER_FASTRECOVERY,
-       TCP_CC_IN_FASTRECOVERY,
-       TCP_CC_EXIT_FASTRECOVERY,
-       TCP_CC_PARTIAL_ACK,
-       TCP_CC_IDLE_TIMEOUT,
-       TCP_CC_REXMT_TIMEOUT,
-       TCP_CC_ECN_RCVD,
-       TCP_CC_BAD_REXMT_RECOVERY,
-       TCP_CC_OUTPUT_ERROR,
-       TCP_CC_CHANGE_ALGO,
-       TCP_CC_FLOW_CONTROL,
-       TCP_CC_SUSPEND,
-       TCP_CC_LIMITED_TRANSMIT,
-       TCP_CC_EARLY_RETRANSMIT,
-       TCP_CC_TLP_RECOVERY,
-       TCP_CC_TLP_RECOVER_LASTPACKET,
-       TCP_CC_DELAY_FASTRECOVERY,
-       TCP_CC_TLP_IN_FASTRECOVERY
+       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 */
 };
 
 /*
@@ -591,12 +720,15 @@ struct tcpopt {
 #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 */
 };
 
 /*
@@ -635,6 +767,11 @@ struct rmxp_tao {
 #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
@@ -767,6 +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.
@@ -924,17 +1064,61 @@ struct   tcpstat {
        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_setup;         /* connection negotiated ECN */
-       u_int32_t       tcps_sent_cwr;          /* Sent CWR, ECE received */
-       u_int32_t       tcps_sent_ece;          /* Sent ECE notification */
+       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;   /* PTO during fast recovery */
-       u_int32_t       tcps_pmtudbh_reverted;  /* PMTU Blackhole detection, segement size reverted */
+       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 {
@@ -1188,21 +1372,22 @@ struct  xtcpcb_n {
 SYSCTL_DECL(_net_inet_tcp);
 #endif /* SYSCTL_DECL */
 
-/*
- * Flags for TCP's connectx(2) user-protocol request routine.
- */
-#if MPTCP
-#define        TCP_CONNREQF_MPTCP      0x1     /* called internally by MPTCP */
-#endif /* MPTCP */
-
 extern struct inpcbhead tcb;           /* head of queue of active tcpcb's */
 extern struct inpcbinfo tcbinfo;
 extern struct tcpstat tcpstat; /* tcp statistics */
 extern int tcp_mssdflt;        /* XXX */
 extern int tcp_minmss;
+#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;
 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;
@@ -1213,6 +1398,10 @@ 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
@@ -1255,6 +1444,7 @@ struct rtentry *
         tcp_rtlookup(struct inpcb *, unsigned int);
 void    tcp_setpersist(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);
 
@@ -1265,6 +1455,8 @@ void       tcp_trace(int, int, struct tcpcb *, void *, struct tcphdr *, int);
 
 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);
@@ -1320,8 +1512,25 @@ 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 uint16_t mptcp_input_csum(struct tcpcb *, struct mbuf *, int);
+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);