]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet/tcp_var.h
xnu-2782.1.97.tar.gz
[apple/xnu.git] / bsd / netinet / tcp_var.h
index afd802f6196521cd4590dca396878dfcfa9f27e8..bf9fb3f20cf4f5d3a99586945e05a6dd64e8cd34 100644 (file)
@@ -1,16 +1,19 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2014 Apple Inc. All rights reserved.
  *
- * @APPLE_LICENSE_HEADER_START@
- * 
- * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
- * compliance with the License. Please obtain a copy of the License at
- * http://www.opensource.apple.com/apsl/ and read it before using this
- * file.
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
+ * 
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
  * 
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
@@ -20,7 +23,7 @@
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * 
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * Copyright (c) 1982, 1986, 1993, 1994, 1995
 #ifndef _NETINET_TCP_VAR_H_
 #define _NETINET_TCP_VAR_H_
 #include <sys/appleapiopts.h>
+#include <sys/queue.h>
+#include <netinet/in_pcb.h>
 #include <netinet/tcp_timer.h>
-#ifdef __APPLE_API_PRIVATE
 
-#define N_TIME_WAIT_SLOTS   128                /* must be power of 2 */
+#if defined(__LP64__)
+#define _TCPCB_PTR(x)                  u_int32_t
+#define _TCPCB_LIST_HEAD(name, type)   \
+struct name {                          \
+       u_int32_t       lh_first;       \
+};
+#else
+#define _TCPCB_PTR(x)                  x
+#define _TCPCB_LIST_HEAD(name, type)   LIST_HEAD(name, type)
+#endif
+
+#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
 
 /*
  * Kernel variables for tcp.
@@ -78,40 +159,88 @@ struct tseg_qent {
        struct  mbuf    *tqe_m;         /* mbuf contains packet */
 };
 LIST_HEAD(tsegqe_head, tseg_qent);
-#ifdef MALLOC_DECLARE
-MALLOC_DECLARE(M_TSEGQ);
-#endif
+
+struct sackblk {
+       tcp_seq start;          /* start seq no. of sack block */
+       tcp_seq end;            /* end seq no. */
+};
+
+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 */
+};
+
+struct sackhint {
+       struct sackhole *nexthole;
+       int     sack_bytes_rexmit;
+};
 
 struct tcptemp {
        u_char  tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
        struct  tcphdr tt_t;
 };
 
-#define tcp6cb         tcpcb  /* for KAME src sync over BSD*'s */
+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 */
+};
 
-#ifdef __APPLE__
-#define TCP_DELACK_BITSET(hash_elem)\
-delack_bitmask[((hash_elem) >> 5)] |= 1 << ((hash_elem) & 0x1F)
+/* 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 */
 
-#define DELACK_BITMASK_ON     1
-#define DELACK_BITMASK_THRESH 300
-#endif
+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.
  */
-#if KERNEL
 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 */
+       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 */
@@ -129,11 +258,21 @@ 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         /* 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 */
-
-       int     t_force;                /* 1 if forcing out a byte */
+#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_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 */
+#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;
@@ -149,31 +288,41 @@ struct tcpcb {
 
        tcp_seq rcv_nxt;                /* receive next */
        tcp_seq rcv_adv;                /* advertised window */
-       u_long  rcv_wnd;                /* receive window */
+       u_int32_t       rcv_wnd;                /* receive window */
        tcp_seq rcv_up;                 /* receive urgent pointer */
 
-       u_long  snd_wnd;                /* send window */
-       u_long  snd_cwnd;               /* congestion-controlled window */
-       u_long  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
                                         */
+       tcp_seq snd_recover;            /* for use in NewReno Fast Recovery */
+
        u_int   t_maxopd;               /* mss plus options */
 
-       u_long  t_rcvtime;              /* inactivity time */
-       u_long  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 */
 
+       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_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_long  t_rttupdated;           /* number of times rtt sampled */
-       u_long  max_sndwnd;             /* largest window peer has offered */
+       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_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 */
@@ -182,44 +331,352 @@ 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_long  ts_recent;              /* timestamp echo data */
+       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_long  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 */
-       tcp_seq snd_recover;            /* for use in fast recovery */
-/* experimental */
-       u_long  snd_cwnd_prev;          /* cwnd prior to retransmit */
-       u_long  snd_ssthresh_prev;      /* ssthresh prior to retransmit */
-       u_long  t_badrxtwin;            /* window for retransmit recovery */
+/* RFC 3465 variables */
+       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_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_ECN_ON              (TE_SETUPSENT | TE_SETUPRECEIVED) /* Indicate ECN was successfully negotiated on a connection) */
 
-       int     t_keepidle;             /* keepalive idle timer (override global if > 0) */
+/* 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 */
+       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 */
+       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_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_int32_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 */
+
+#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 */
+       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 */
+#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 */
+
+       void                    *t_mptcb;       /* pointer to MPTCP TCB */
+       tcp_seq                 t_mpuna;        /* unacknowledged sequence */
+       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 */
 };
-#else
 
-#define tcpcb otcpcb
+#define IN_FASTRECOVERY(tp)    (tp->t_flags & TF_FASTRECOVERY)
+#define SACK_ENABLED(tp)       (tp->t_flagsext & TF_SACK_ENABLE)
 
-#endif
+/*
+ * 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);            \
+} 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;          \
+} 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))
+
+/*
+ * Gives number of bytes acked by this ack
+ */
+#define BYTES_ACKED(_th_, _tp_) \
+       ((_th_)->th_ack - (_tp_)->snd_una)
+
+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
+};
+
+/*
+ * Structure to hold TCP options that are only used during segment
+ * processing (in tcp_input), but not held in the tcpcb.
+ * It's basically used to reduce the number of parameters
+ * to tcp_dooptions.
+ */
+struct tcpopt {
+       u_int32_t       to_flags;               /* which options are present */
+#define TOF_TS         0x0001          /* timestamp */
+#define        TOF_MSS         0x0010
+#define        TOF_SCALE       0x0020
+#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 */
+       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 */
+};
+
+/*
+ * The TAO cache entry which is stored in the protocol family specific
+ * portion of the route metrics.
+ */
+struct rmxp_tao {
+       tcp_cc  tao_cc;                 /* latest CC in valid SYN */
+       tcp_cc  tao_ccsent;             /* latest CC sent to peer */
+       u_short tao_mssopt;             /* peer's cached MSS */
+#ifdef notyet
+       u_short tao_flags;              /* cache status flags */
+#define        TAOF_DONT       0x0001          /* peer doesn't understand rfc1644 */
+#define        TAOF_OK         0x0002          /* peer does understand rfc1644 */
+#define        TAOF_UNDEF      0               /* we don't know yet */
+#endif /* notyet */
+};
+#define rmx_taop(r)    ((struct rmxp_tao *)(r).rmx_filler)
+
+#define        intotcpcb(ip)   ((struct tcpcb *)(ip)->inp_ppcb)
+#define        sototcpcb(so)   (intotcpcb(sotoinpcb(so)))
+
+/*
+ * 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 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; 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 */
 
+/*
+ * The initial retransmission should happen at rtt + 4 * rttvar.
+ * Because of the way we do the smoothing, srtt and rttvar
+ * will each average +1/2 tick of bias.  When we compute
+ * the retransmit timer, we want 1/2 tick of rounding and
+ * 1 extra tick because of +-1/2 tick uncertainty in the
+ * firing of the timer.  The bias will give us exactly the
+ * 1.5 tick we need.  But, because the bias is
+ * statistical, we have to test that we don't drop below
+ * the minimum feasible timer (which is 2 ticks).
+ * This version of the macro adapted from a paper by Lawrence
+ * Brakmo and Larry Peterson which outlines a problem caused
+ * by insufficient precision in the original implementation,
+ * which results in inappropriately large RTO values for very
+ * fast networks.
+ */
+#define        TCP_REXMTVAL(tp) \
+       max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
+         + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
 
 /*
  * Jaguar compatible TCP control block, for xtcpcb
  * Does not have the old fields
  */
 struct otcpcb {
+#else
+struct tseg_qent;
+_TCPCB_LIST_HEAD(tsegqe_head, tseg_qent);
+
+struct tcpcb {
+#endif /* KERNEL_PRIVATE */
+#if defined(KERNEL_PRIVATE)
+       u_int32_t t_segq;
+#else
        struct  tsegqe_head t_segq;
+#endif /* KERNEL_PRIVATE */
        int     t_dupacks;              /* consecutive dup acks recd */
-       struct  tcptemp *unused;        /* unused now: was t_template */
+       u_int32_t unused;               /* unused now: was t_template */
 
-       int     t_timer[TCPT_NTIMERS];  /* tcp timers */
+       int     t_timer[TCPT_NTIMERS_EXT];      /* tcp timers */
 
-       struct  inpcb *t_inpcb;         /* back pointer to internet pcb */
+       _TCPCB_PTR(struct inpcb *) t_inpcb;     /* back pointer to internet pcb */
        int     t_state;                /* state of this connection */
        u_int   t_flags;
 #define        TF_ACKNOW       0x00001         /* ack peer immediately */
@@ -259,19 +716,19 @@ struct otcpcb {
 
        tcp_seq rcv_nxt;                /* receive next */
        tcp_seq rcv_adv;                /* advertised window */
-       u_long  rcv_wnd;                /* receive window */
+       u_int32_t rcv_wnd;              /* receive window */
        tcp_seq rcv_up;                 /* receive urgent pointer */
 
-       u_long  snd_wnd;                /* send window */
-       u_long  snd_cwnd;               /* congestion-controlled window */
-       u_long  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_int   t_maxopd;               /* mss plus options */
 
-       u_long  t_rcvtime;              /* inactivity time */
-       u_long  t_starttime;            /* time connection was established */
+       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 */
 
@@ -282,8 +739,8 @@ struct otcpcb {
 
        int     t_rxtshift;             /* log(2) of rexmt exp. backoff */
        u_int   t_rttmin;               /* minimum rtt allowed */
-       u_long  t_rttupdated;           /* number of times rtt sampled */
-       u_long  max_sndwnd;             /* largest window peer has offered */
+       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 */
@@ -296,187 +753,393 @@ struct otcpcb {
        u_char  rcv_scale;              /* window scaling for recv window */
        u_char  request_r_scale;        /* pending window scaling */
        u_char  requested_s_scale;
-       u_long  ts_recent;              /* timestamp echo data */
+       u_int32_t ts_recent;            /* timestamp echo data */
 
-       u_long  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 */
        tcp_seq snd_recover;            /* for use in fast recovery */
 /* experimental */
-       u_long  snd_cwnd_prev;          /* cwnd prior to retransmit */
-       u_long  snd_ssthresh_prev;      /* ssthresh prior to retransmit */
-       u_long  t_badrxtwin;            /* window for retransmit recovery */
+       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 */
 };
 
 
 /*
- * Structure to hold TCP options that are only used during segment
- * processing (in tcp_input), but not held in the tcpcb.
- * It's basically used to reduce the number of parameters
- * to tcp_dooptions.
+ * TCP statistics.
+ * Many of these should be kept per connection,
+ * but that's inconvenient at the moment.
  */
-struct tcpopt {
-       u_long  to_flag;                /* which options are present */
-#define TOF_TS         0x0001          /* timestamp */
-#define TOF_CC         0x0002          /* CC and CCnew are exclusive */
-#define TOF_CCNEW      0x0004
-#define        TOF_CCECHO      0x0008
-       u_long  to_tsval;
-       u_long  to_tsecr;
-       tcp_cc  to_cc;          /* holds CC or CCnew */
-       tcp_cc  to_ccecho;
-       u_short         reserved;               /* unused now: was to_maxseg */
-};
+struct tcpstat {
+       u_int32_t       tcps_connattempt;       /* connections initiated */
+       u_int32_t       tcps_accepts;           /* connections accepted */
+       u_int32_t       tcps_connects;          /* connections established */
+       u_int32_t       tcps_drops;             /* connections dropped */
+       u_int32_t       tcps_conndrops;         /* embryonic connections dropped */
+       u_int32_t       tcps_closed;            /* conn. closed (includes drops) */
+       u_int32_t       tcps_segstimed;         /* segs where we tried to get rtt */
+       u_int32_t       tcps_rttupdated;        /* times we succeeded */
+       u_int32_t       tcps_delack;            /* delayed acks sent */
+       u_int32_t       tcps_timeoutdrop;       /* conn. dropped in rxmt timeout */
+       u_int32_t       tcps_rexmttimeo;        /* retransmit timeouts */
+       u_int32_t       tcps_persisttimeo;      /* persist timeouts */
+       u_int32_t       tcps_keeptimeo;         /* keepalive timeouts */
+       u_int32_t       tcps_keepprobe;         /* keepalive probes sent */
+       u_int32_t       tcps_keepdrops;         /* connections dropped in keepalive */
 
-/*
- * The TAO cache entry which is stored in the protocol family specific
- * portion of the route metrics.
- */
-struct rmxp_tao {
-       tcp_cc  tao_cc;                 /* latest CC in valid SYN */
-       tcp_cc  tao_ccsent;             /* latest CC sent to peer */
-       u_short tao_mssopt;             /* peer's cached MSS */
-#ifdef notyet
-       u_short tao_flags;              /* cache status flags */
-#define        TAOF_DONT       0x0001          /* peer doesn't understand rfc1644 */
-#define        TAOF_OK         0x0002          /* peer does understand rfc1644 */
-#define        TAOF_UNDEF      0               /* we don't know yet */
-#endif /* notyet */
-};
-#define rmx_taop(r)    ((struct rmxp_tao *)(r).rmx_filler)
+       u_int32_t       tcps_sndtotal;          /* total packets sent */
+       u_int32_t       tcps_sndpack;           /* data packets sent */
+       u_int32_t       tcps_sndbyte;           /* data bytes sent */
+       u_int32_t       tcps_sndrexmitpack;     /* data packets retransmitted */
+       u_int32_t       tcps_sndrexmitbyte;     /* data bytes retransmitted */
+       u_int32_t       tcps_sndacks;           /* ack-only packets sent */
+       u_int32_t       tcps_sndprobe;          /* window probes sent */
+       u_int32_t       tcps_sndurg;            /* packets sent with URG only */
+       u_int32_t       tcps_sndwinup;          /* window update-only packets sent */
+       u_int32_t       tcps_sndctrl;           /* control (SYN|FIN|RST) packets sent */
 
-#define        intotcpcb(ip)   ((struct tcpcb *)(ip)->inp_ppcb)
-#define        sototcpcb(so)   (intotcpcb(sotoinpcb(so)))
+       u_int32_t       tcps_rcvtotal;          /* total packets received */
+       u_int32_t       tcps_rcvpack;           /* packets received in sequence */
+       u_int32_t       tcps_rcvbyte;           /* bytes received in sequence */
+       u_int32_t       tcps_rcvbadsum;         /* packets received with ccksum errs */
+       u_int32_t       tcps_rcvbadoff;         /* packets received with bad offset */
+       u_int32_t       tcps_rcvmemdrop;        /* packets dropped for lack of memory */
+       u_int32_t       tcps_rcvshort;          /* packets received too short */
+       u_int32_t       tcps_rcvduppack;        /* duplicate-only packets received */
+       u_int32_t       tcps_rcvdupbyte;        /* duplicate-only bytes received */
+       u_int32_t       tcps_rcvpartduppack;    /* packets with some duplicate data */
+       u_int32_t       tcps_rcvpartdupbyte;    /* dup. bytes in part-dup. packets */
+       u_int32_t       tcps_rcvoopack;         /* out-of-order packets received */
+       u_int32_t       tcps_rcvoobyte;         /* out-of-order bytes received */
+       u_int32_t       tcps_rcvpackafterwin;   /* packets with data after window */
+       u_int32_t       tcps_rcvbyteafterwin;   /* bytes rcvd after window */
+       u_int32_t       tcps_rcvafterclose;     /* packets rcvd after "close" */
+       u_int32_t       tcps_rcvwinprobe;       /* rcvd window probe packets */
+       u_int32_t       tcps_rcvdupack;         /* rcvd duplicate acks */
+       u_int32_t       tcps_rcvacktoomuch;     /* rcvd acks for unsent data */
+       u_int32_t       tcps_rcvackpack;        /* rcvd ack packets */
+       u_int32_t       tcps_rcvackbyte;        /* bytes acked by rcvd acks */
+       u_int32_t       tcps_rcvwinupd;         /* rcvd window update packets */
+       u_int32_t       tcps_pawsdrop;          /* segments dropped due to PAWS */
+       u_int32_t       tcps_predack;           /* times hdr predict ok for acks */
+       u_int32_t       tcps_preddat;           /* times hdr predict ok for data pkts */
+       u_int32_t       tcps_pcbcachemiss;
+       u_int32_t       tcps_cachedrtt;         /* times cached RTT in route updated */
+       u_int32_t       tcps_cachedrttvar;      /* times cached rttvar updated */
+       u_int32_t       tcps_cachedssthresh;    /* times cached ssthresh updated */
+       u_int32_t       tcps_usedrtt;           /* times RTT initialized from route */
+       u_int32_t       tcps_usedrttvar;        /* times RTTVAR initialized from rt */
+       u_int32_t       tcps_usedssthresh;      /* times ssthresh initialized from rt*/
+       u_int32_t       tcps_persistdrop;       /* timeout in persist state */
+       u_int32_t       tcps_badsyn;            /* bogus SYN, e.g. premature ACK */
+       u_int32_t       tcps_mturesent;         /* resends due to MTU discovery */
+       u_int32_t       tcps_listendrop;        /* listen queue overflows */
 
-/*
- * 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
- * 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_DELTA_SHIFT         2       /* see tcp_input.c */
+       /* new stats from FreeBSD 5.4 sync up */
+       u_int32_t       tcps_minmssdrops;       /* average minmss too low drops */
+       u_int32_t       tcps_sndrexmitbad;      /* unnecessary packet retransmissions */
+       u_int32_t       tcps_badrst;            /* ignored RSTs in the window */
 
-/*
- * The initial retransmission should happen at rtt + 4 * rttvar.
- * Because of the way we do the smoothing, srtt and rttvar
- * will each average +1/2 tick of bias.  When we compute
- * the retransmit timer, we want 1/2 tick of rounding and
- * 1 extra tick because of +-1/2 tick uncertainty in the
- * firing of the timer.  The bias will give us exactly the
- * 1.5 tick we need.  But, because the bias is
- * statistical, we have to test that we don't drop below
- * the minimum feasible timer (which is 2 ticks).
- * This version of the macro adapted from a paper by Lawrence
- * Brakmo and Larry Peterson which outlines a problem caused
- * by insufficient precision in the original implementation,
- * which results in inappropriately large RTO values for very
- * fast networks.
- */
-#define        TCP_REXMTVAL(tp) \
-       max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
-         + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
-#endif /* __APPLE_API_PRIVATE */
+       u_int32_t       tcps_sc_added;          /* entry added to syncache */
+       u_int32_t       tcps_sc_retransmitted;  /* syncache entry was retransmitted */
+       u_int32_t       tcps_sc_dupsyn;         /* duplicate SYN packet */
+       u_int32_t       tcps_sc_dropped;        /* could not reply to packet */
+       u_int32_t       tcps_sc_completed;      /* successful extraction of entry */
+       u_int32_t       tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */
+       u_int32_t       tcps_sc_cacheoverflow;  /* syncache cache limit hit */
+       u_int32_t       tcps_sc_reset;          /* RST removed entry from syncache */
+       u_int32_t       tcps_sc_stale;          /* timed out or listen socket gone */
+       u_int32_t       tcps_sc_aborted;        /* syncache entry aborted */
+       u_int32_t       tcps_sc_badack;         /* removed due to bad ACK */
+       u_int32_t       tcps_sc_unreach;        /* ICMP unreachable received */
+       u_int32_t       tcps_sc_zonefail;       /* zalloc() failed */
+       u_int32_t       tcps_sc_sendcookie;     /* SYN cookie sent */
+       u_int32_t       tcps_sc_recvcookie;     /* SYN cookie received */
 
-#ifdef __APPLE_API_UNSTABLE
-/*
- * TCP statistics.
- * Many of these should be kept per connection,
- * but that's inconvenient at the moment.
- */
-struct tcpstat {
-       u_long  tcps_connattempt;       /* connections initiated */
-       u_long  tcps_accepts;           /* connections accepted */
-       u_long  tcps_connects;          /* connections established */
-       u_long  tcps_drops;             /* connections dropped */
-       u_long  tcps_conndrops;         /* embryonic connections dropped */
-       u_long  tcps_closed;            /* conn. closed (includes drops) */
-       u_long  tcps_segstimed;         /* segs where we tried to get rtt */
-       u_long  tcps_rttupdated;        /* times we succeeded */
-       u_long  tcps_delack;            /* delayed acks sent */
-       u_long  tcps_timeoutdrop;       /* conn. dropped in rxmt timeout */
-       u_long  tcps_rexmttimeo;        /* retransmit timeouts */
-       u_long  tcps_persisttimeo;      /* persist timeouts */
-       u_long  tcps_keeptimeo;         /* keepalive timeouts */
-       u_long  tcps_keepprobe;         /* keepalive probes sent */
-       u_long  tcps_keepdrops;         /* connections dropped in keepalive */
-
-       u_long  tcps_sndtotal;          /* total packets sent */
-       u_long  tcps_sndpack;           /* data packets sent */
-       u_long  tcps_sndbyte;           /* data bytes sent */
-       u_long  tcps_sndrexmitpack;     /* data packets retransmitted */
-       u_long  tcps_sndrexmitbyte;     /* data bytes retransmitted */
-       u_long  tcps_sndacks;           /* ack-only packets sent */
-       u_long  tcps_sndprobe;          /* window probes sent */
-       u_long  tcps_sndurg;            /* packets sent with URG only */
-       u_long  tcps_sndwinup;          /* window update-only packets sent */
-       u_long  tcps_sndctrl;           /* control (SYN|FIN|RST) packets sent */
-
-       u_long  tcps_rcvtotal;          /* total packets received */
-       u_long  tcps_rcvpack;           /* packets received in sequence */
-       u_long  tcps_rcvbyte;           /* bytes received in sequence */
-       u_long  tcps_rcvbadsum;         /* packets received with ccksum errs */
-       u_long  tcps_rcvbadoff;         /* packets received with bad offset */
-       u_long  tcps_rcvmemdrop;        /* packets dropped for lack of memory */
-       u_long  tcps_rcvshort;          /* packets received too short */
-       u_long  tcps_rcvduppack;        /* duplicate-only packets received */
-       u_long  tcps_rcvdupbyte;        /* duplicate-only bytes received */
-       u_long  tcps_rcvpartduppack;    /* packets with some duplicate data */
-       u_long  tcps_rcvpartdupbyte;    /* dup. bytes in part-dup. packets */
-       u_long  tcps_rcvoopack;         /* out-of-order packets received */
-       u_long  tcps_rcvoobyte;         /* out-of-order bytes received */
-       u_long  tcps_rcvpackafterwin;   /* packets with data after window */
-       u_long  tcps_rcvbyteafterwin;   /* bytes rcvd after window */
-       u_long  tcps_rcvafterclose;     /* packets rcvd after "close" */
-       u_long  tcps_rcvwinprobe;       /* rcvd window probe packets */
-       u_long  tcps_rcvdupack;         /* rcvd duplicate acks */
-       u_long  tcps_rcvacktoomuch;     /* rcvd acks for unsent data */
-       u_long  tcps_rcvackpack;        /* rcvd ack packets */
-       u_long  tcps_rcvackbyte;        /* bytes acked by rcvd acks */
-       u_long  tcps_rcvwinupd;         /* rcvd window update packets */
-       u_long  tcps_pawsdrop;          /* segments dropped due to PAWS */
-       u_long  tcps_predack;           /* times hdr predict ok for acks */
-       u_long  tcps_preddat;           /* times hdr predict ok for data pkts */
-       u_long  tcps_pcbcachemiss;
-       u_long  tcps_cachedrtt;         /* times cached RTT in route updated */
-       u_long  tcps_cachedrttvar;      /* times cached rttvar updated */
-       u_long  tcps_cachedssthresh;    /* times cached ssthresh updated */
-       u_long  tcps_usedrtt;           /* times RTT initialized from route */
-       u_long  tcps_usedrttvar;        /* times RTTVAR initialized from rt */
-       u_long  tcps_usedssthresh;      /* times ssthresh initialized from rt*/
-       u_long  tcps_persistdrop;       /* timeout in persist state */
-       u_long  tcps_badsyn;            /* bogus SYN, e.g. premature ACK */
-       u_long  tcps_mturesent;         /* resends due to MTU discovery */
-       u_long  tcps_listendrop;        /* listen queue overflows */
+       u_int32_t       tcps_hc_added;          /* entry added to hostcache */
+       u_int32_t       tcps_hc_bucketoverflow; /* hostcache per bucket limit hit */
+
+       /* 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_bg_rcvtotal;       /* total background packets received */
+       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_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_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 */
 };
-#endif /* __APPLE_API_UNSTABLE */
+
+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)
 
 /*
  * TCB structure exported to user-land via sysctl(3).
  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
  * included.  Not all of our clients do.
  */
-#if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
-struct xtcpcb {
-       size_t  xt_len;
-       struct  inpcb   xt_inp;
-#if KERNEL
-       struct  otcpcb  xt_tp;
+
+struct  xtcpcb {
+        u_int32_t       xt_len;
+#ifdef KERNEL_PRIVATE
+        struct  inpcb_compat    xt_inp;
 #else
-       struct  tcpcb   xt_tp;
+        struct  inpcb   xt_inp;
 #endif
-       struct  xsocket xt_socket;
-       u_quad_t        xt_alignment_hack;
-};
+#ifdef KERNEL_PRIVATE
+        struct  otcpcb  xt_tp;
+#else
+        struct  tcpcb   xt_tp;
 #endif
+        struct  xsocket xt_socket;
+        u_quad_t        xt_alignment_hack;
+};
+
+
+struct  xtcpcb64 {
+        u_int32_t                      xt_len;
+        struct xinpcb64                xt_inpcb;
+
+        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 */
+        u_int32_t t_badrxtwin;          /* window for retransmit recovery */
+
+        u_quad_t               xt_alignment_hack;
+};
+
+
+#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()
 
 /*
  * Names for TCP sysctl objects
@@ -496,6 +1159,14 @@ struct    xtcpcb {
 #define        TCPCTL_V6MSSDFLT        13      /* MSS default for IPv6 */
 #define        TCPCTL_MAXID            14
 
+#ifdef BSD_KERNEL_PRIVATE
+#include <sys/bitstring.h>
+
+#define        TCP_PKTLIST_CLEAR(tp) {                                         \
+       (tp)->t_pktlist_head = (tp)->t_pktlist_tail = NULL;             \
+       (tp)->t_lastchain = (tp)->t_pktlist_sentlen = 0;                \
+}
+
 #define TCPCTL_NAMES { \
        { 0, 0 }, \
        { "rfc1323", CTLTYPE_INT }, \
@@ -513,67 +1184,149 @@ struct  xtcpcb {
        { "v6mssdflt", CTLTYPE_INT }, \
 }
 
-#ifdef __APPLE_API_PRIVATE
-#ifdef KERNEL
 #ifdef SYSCTL_DECL
 SYSCTL_DECL(_net_inet_tcp);
-#endif
+#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_delack_enabled;
-extern int tcp_do_newreno;
+extern int tcp_minmss;
 extern int ss_fltsz;
 extern int ss_fltsz_local;
-#ifdef __APPLE__
-extern u_long tcp_now;         /* for RFC 1323 timestamps */ 
+extern         int tcp_do_rfc3390;             /* Calculate ss_fltsz according to RFC 3390 */
+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;
+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;
+
+
+#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 __P((struct tcpcb *));
+void    tcp_canceltimers(struct tcpcb *);
 struct tcpcb *
-        tcp_close __P((struct tcpcb *));
-void    tcp_ctlinput __P((int, struct sockaddr *, void *));
-int     tcp_ctloutput __P((struct socket *, struct sockopt *));
+        tcp_close(struct tcpcb *);
+void    tcp_ctlinput(int, struct sockaddr *, void *);
+int     tcp_ctloutput(struct socket *, struct sockopt *);
 struct tcpcb *
-        tcp_drop __P((struct tcpcb *, int));
-void    tcp_drain __P((void));
-void    tcp_fasttimo __P((void));
+        tcp_drop(struct tcpcb *, int);
+void    tcp_drain(void);
+void    tcp_getrt_rtt(struct tcpcb *tp, struct rtentry *rt);
 struct rmxp_tao *
-        tcp_gettaocache __P((struct inpcb *));
-void    tcp_init __P((void));
-void    tcp_input __P((struct mbuf *, int));
-void    tcp_mss __P((struct tcpcb *, int));
-int     tcp_mssopt __P((struct tcpcb *));
-void    tcp_drop_syn_sent __P((struct inpcb *, int));
-void    tcp_mtudisc __P((struct inpcb *, int));
+        tcp_gettaocache(struct inpcb *);
+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 *);
+void    tcp_drop_syn_sent(struct inpcb *, int);
+void    tcp_mtudisc(struct inpcb *, int);
 struct tcpcb *
-        tcp_newtcpcb __P((struct inpcb *));
-int     tcp_output __P((struct tcpcb *));
-void    tcp_quench __P((struct inpcb *, int));
-void    tcp_respond __P((struct tcpcb *, void *,
-           struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int));
+        tcp_newtcpcb(struct inpcb *);
+int     tcp_output(struct tcpcb *);
+void    tcp_respond(struct tcpcb *, void *, struct tcphdr *, struct mbuf *,
+    tcp_seq, tcp_seq, int, struct tcp_respond_args *);
 struct rtentry *
-        tcp_rtlookup __P((struct inpcb *));
-void    tcp_setpersist __P((struct tcpcb *));
-void    tcp_slowtimo __P((void));
-struct tcptemp *
-        tcp_maketemplate __P((struct tcpcb *));
-void    tcp_fillheaders __P((struct tcpcb *, void *, void *));
-struct tcpcb *
-        tcp_timers __P((struct tcpcb *, int));
-void    tcp_trace __P((int, int, struct tcpcb *, void *, struct tcphdr *,
-                       int));
+        tcp_rtlookup(struct inpcb *, unsigned int);
+void    tcp_setpersist(struct tcpcb *);
+void    tcp_gc(struct inpcbinfo *);
+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);
+void    tcp_trace(int, int, struct tcpcb *, void *, struct tcphdr *, int);
+
+void tcp_sack_doack(struct tcpcb *, struct tcpopt *, struct tcphdr *,
+    u_int32_t *);
+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);
+struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
+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
+void *  tcp_getlock (struct socket *, int);
+#endif
+
 
 extern struct pr_usrreqs tcp_usrreqs;
-extern u_long tcp_sendspace;
-extern u_long tcp_recvspace;
-tcp_seq tcp_new_isn __P((struct tcpcb *));
+extern u_int32_t tcp_sendspace;
+extern u_int32_t tcp_recvspace;
+tcp_seq tcp_new_isn(struct tcpcb *);
 
-#endif /* KERNEL */
-#endif /* __APPLE_API_PRIVATE */
+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);
+#if MPTCP
+extern uint16_t mptcp_input_csum(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_ */