xref: /xnu-12377.81.4/bsd/netinet/tcp_var.h (revision 043036a2b3718f7f0be807e2870f8f47d3fa0796)
1 /*
2  * Copyright (c) 2000-2024 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 /*
29  * Copyright (c) 1982, 1986, 1993, 1994, 1995
30  *	The Regents of the University of California.  All rights reserved.
31  *
32  * Redistribution and use in source and binary forms, with or without
33  * modification, are permitted provided that the following conditions
34  * are met:
35  * 1. Redistributions of source code must retain the above copyright
36  *    notice, this list of conditions and the following disclaimer.
37  * 2. Redistributions in binary form must reproduce the above copyright
38  *    notice, this list of conditions and the following disclaimer in the
39  *    documentation and/or other materials provided with the distribution.
40  * 3. All advertising materials mentioning features or use of this software
41  *    must display the following acknowledgement:
42  *	This product includes software developed by the University of
43  *	California, Berkeley and its contributors.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
61  * $FreeBSD: src/sys/netinet/tcp_var.h,v 1.56.2.8 2001/08/22 00:59:13 silby Exp $
62  */
63 
64 #ifndef _NETINET_TCP_VAR_H_
65 #define _NETINET_TCP_VAR_H_
66 #include <stdint.h>
67 #include <sys/types.h>
68 #include <sys/appleapiopts.h>
69 #include <sys/queue.h>
70 #include <netinet/in_pcb.h>
71 #include <netinet/tcp.h>
72 #include <netinet/tcp_timer.h>
73 #if !KERNEL
74 #include <TargetConditionals.h>
75 #endif
76 
77 #if defined(__LP64__)
78 #define _TCPCB_PTR(x)                   u_int32_t
79 #define _TCPCB_LIST_HEAD(name, type)    \
80 struct name {                           \
81 	u_int32_t	lh_first;       \
82 }
83 #else
84 #define _TCPCB_PTR(x)                   x
85 #define _TCPCB_LIST_HEAD(name, type)    LIST_HEAD(name, type)
86 #endif
87 
88 #ifdef KERNEL_PRIVATE
89 #include <sys/tree.h>
90 
91 #define TCP_RETRANSHZ   1000    /* granularity of TCP timestamps, 1ms */
92 /* Minimum time quantum within which the timers are coalesced */
93 #define TCP_TIMER_10MS_QUANTUM  (TCP_RETRANSHZ/100) /* every 10ms */
94 #define TCP_TIMER_100MS_QUANTUM   (TCP_RETRANSHZ/10) /* every 100ms */
95 #define TCP_TIMER_500MS_QUANTUM   (TCP_RETRANSHZ/2) /* every 500ms */
96 
97 #define TCP_RETRANSHZ_TO_USEC 1000
98 
99 #define N_TIME_WAIT_SLOTS   128         /* must be power of 2 */
100 
101 /* Always allow at least 16 packets worth of recv window when adjusting
102  * recv window using inter-packet arrival jitter.
103  */
104 #define MIN_IAJ_WIN 16
105 
106 /* A variation in delay of this many milliseconds is tolerable. This limit has to
107  * be low but greater than zero. We also use standard deviation on jitter to adjust
108  * this limit for different link and connection types.
109  */
110 #define ALLOWED_IAJ 5
111 
112 /* Ignore the first few packets on a connection until the ACK clock gets going
113  */
114 #define IAJ_IGNORE_PKTCNT 40
115 
116 /* Let the accumulated IAJ value increase by this threshold at most. This limit
117  * will control how many ALLOWED_IAJ measurements a receiver will have to see
118  * before opening the receive window
119  */
120 #define ACC_IAJ_HIGH_THRESH 100
121 
122 /* When accumulated IAJ reaches this value, the receiver starts to react by
123  * closing the window
124  */
125 #define ACC_IAJ_REACT_LIMIT 200
126 
127 /* If the number of small packets (smaller than IAJ packet size) seen on a
128  * connection is more than this threshold, reset the size and learn it again.
129  * This is needed because the sender might send smaller segments after PMTU
130  * discovery and the receiver has to learn the new size.
131  */
132 #define RESET_IAJ_SIZE_THRESH 20
133 
134 /*
135  * Adaptive timeout is a read/write timeout specified by the application to
136  * get a socket event when the transport layer detects a stall in data
137  * transfer. The value specified is the number of probes that can be sent
138  * to the peer before generating an event. Since it is not specified as
139  * a time value, the timeout will adjust based on the RTT seen on the link.
140  * The timeout will start only when there is an indication that the read/write
141  * operation is not making progress.
142  *
143  * If a write operation stalls, the probe will be retransmission of data.
144  * If a read operation stalls, the probe will be a keep-alive packet.
145  *
146  * The maximum value of adaptive timeout is set to 10 which will allow
147  * transmission of enough number of probes to the peer.
148  */
149 #define TCP_ADAPTIVE_TIMEOUT_MAX        10
150 
151 #define TCP_CONNECTIVITY_PROBES_MAX     5
152 
153 /* 3 bit mask used for Accurate ECN ACE field */
154 #define TCP_ACE_MASK (0x7)
155 /* Divisor used for Accurate ECN ACE field */
156 #define TCP_ACE_DIV (1 << 3)
157 
158 /* 24 bit mask used for Accurate ECN option */
159 #define TCP_ACO_MASK (0xFFFFFF)
160 /* Divisor used for Accurate ECN options field */
161 #define TCP_ACO_DIV (1 << 24)
162 
163 static __inline uint16_t
tcp_get_flags(const struct tcphdr * th)164 tcp_get_flags(const struct tcphdr *th)
165 {
166 	return (uint16_t)((th->th_x2 << 8) | th->th_flags);
167 }
168 
169 static __inline void
tcp_set_flags(struct tcphdr * th,uint16_t flags)170 tcp_set_flags(struct tcphdr *th, uint16_t flags)
171 {
172 	th->th_x2 = (flags >> 8) & 0x0f;
173 	th->th_flags = flags & 0xff;
174 }
175 
176 /*
177  * Kernel variables for tcp.
178  */
179 
180 /* TCP segment queue entry */
181 struct tseg_qent {
182 	LIST_ENTRY(tseg_qent) tqe_q;
183 	int     tqe_len;                /* TCP segment data length */
184 	struct  tcphdr *tqe_th;         /* a pointer to tcp header */
185 	struct  mbuf    *tqe_m;         /* mbuf contains packet */
186 };
187 LIST_HEAD(tsegqe_head, tseg_qent);
188 
189 struct sackblk {
190 	tcp_seq start;          /* start seq no. of sack block */
191 	tcp_seq end;            /* end seq no. */
192 };
193 
194 struct sackhole {
195 	tcp_seq start;          /* start seq no. of hole */
196 	tcp_seq end;            /* end seq no. */
197 	tcp_seq rxmit;          /* next seq. no in hole to be retransmitted */
198 	u_int32_t rxmit_start;  /* timestamp of first retransmission */
199 	TAILQ_ENTRY(sackhole) scblink;  /* scoreboard linkage */
200 };
201 
202 struct sackhint {
203 	struct sackhole *nexthole;
204 	int     sack_bytes_rexmit;
205 	int sack_bytes_acked;
206 };
207 
208 struct tcp_rxt_seg {
209 	tcp_seq rx_start;
210 	tcp_seq rx_end;
211 	u_int16_t rx_count;
212 	u_int16_t rx_flags;
213 #define TCP_RXT_SPURIOUS        0x1     /* received DSACK notification */
214 	SLIST_ENTRY(tcp_rxt_seg) rx_link;
215 };
216 
217 /* TCP state of a segment that was sent */
218 struct tcp_seg_sent {
219 	tcp_seq start_seq;
220 	tcp_seq end_seq;        /* last seq sent + 1 */
221 	uint32_t xmit_ts;
222 	uint8_t flags;
223 	uint8_t pad[3];
224 #define TCP_SEGMENT_SACKED                  0x1
225 #define TCP_SEGMENT_LOST                    0x2
226 #define TCP_RACK_RETRANSMITTED              0x4
227 #define TCP_SEGMENT_RETRANSMITTED_ATLEAST_ONCE      0x8 /* If a segment was retransmitted at least once; used for reordering detection and to avoid spurious inferences for RACK.rtt */
228 #define TCP_SEGMENT_RETRANSMITTED           (TCP_RACK_RETRANSMITTED | TCP_SEGMENT_RETRANSMITTED_ATLEAST_ONCE)
229 
230 	TAILQ_ENTRY(tcp_seg_sent) tx_link; /* Time ordered list of segments */
231 	RB_ENTRY(tcp_seg_sent)    seg_link; /* RB tree to keep track of S/ACKed segments */
232 	TAILQ_ENTRY(tcp_seg_sent) ack_link; /* List of ACKed segments to be deleted from time ordered list and RB tree */
233 	TAILQ_ENTRY(tcp_seg_sent) free_link; /* List of free segments from the pool */
234 };
235 
236 typedef struct tcp_seg_pool {
237 	TAILQ_HEAD(, tcp_seg_sent) free_segs;
238 	uint32_t free_segs_count;
239 	char pad[4];
240 } *tcp_seg_pool_t;
241 
242 #define TCP_SEG_POOL_MAX_ITEM_COUNT (64)
243 
244 int tcp_seg_cmp(const struct tcp_seg_sent *, const struct tcp_seg_sent *);
245 
246 RB_HEAD(tcp_seg_sent_tree_head, tcp_seg_sent);
247 RB_PROTOTYPE(tcp_seg_sent_tree_head, tcp_seg_sent, seg_link, tcp_seg_cmp)
248 
249 struct tcp_notify_ack_marker {
250 	tcp_seq notify_snd_una; /* Notify when snd_una crosses this seq */
251 	tcp_notify_ack_id_t notify_id;
252 	SLIST_ENTRY(tcp_notify_ack_marker) notify_next;
253 };
254 
255 struct tcptemp {
256 	u_char  tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
257 	struct  tcphdr tt_t;
258 };
259 
260 struct tcp_inp {
261 	struct socket *so;
262 	struct inpcb **inp;
263 	struct tcpcb **tp;
264 	struct mbuf *m;
265 	struct tcphdr *th;
266 	struct tcpopt *to;
267 	u_char *optp __counted_by(optlen);
268 	struct ip6_hdr *ip6;
269 	struct ip *ip;
270 	struct ifnet *ifp;
271 	struct proc *kernel_proc;
272 	tcp_seq iss;
273 	tcp_seq irs;
274 	uint32_t tiwin;
275 	uint32_t ts_offset;
276 	int optlen;
277 	unsigned int ifscope;
278 	uint16_t peer_mss;
279 	uint8_t peer_wscale;
280 	bool sackok;
281 	bool ecnok;
282 	uint8_t ip_ecn;
283 	bool isipv6;
284 };
285 
286 struct bwmeas {
287 	tcp_seq bw_start;       /* start of bw measurement */
288 	uint32_t bw_ts;         /* timestamp when bw measurement started */
289 	uint32_t bw_size;       /* burst size in bytes for this bw measurement */
290 	uint32_t bw_minsizepkts; /* Min burst size as segments */
291 	uint32_t bw_maxsizepkts; /* Max burst size as segments */
292 	uint32_t bw_minsize;    /* Min size in bytes */
293 	uint32_t bw_maxsize;    /* Max size in bytes */
294 	uint32_t bw_sndbw;      /* Measured send bandwidth */
295 	uint32_t bw_sndbw_max;  /* Max measured bandwidth */
296 	uint32_t bw_rcvbw_max;  /* Max receive bandwidth measured */
297 };
298 
299 /* MPTCP Data sequence map entry */
300 struct mpt_dsn_map {
301 	uint64_t                mpt_dsn;        /* data seq num recvd */
302 	uint32_t                mpt_sseq;       /* relative subflow # */
303 	uint16_t                mpt_len;        /* length of mapping */
304 	uint16_t                mpt_csum;       /* checksum value if on */
305 	uint8_t                 mpt_dfin;       /* It's a DATA_FIN */
306 };
307 #define tcp6cb          tcpcb  /* for KAME src sync over BSD*'s */
308 
309 struct tcp_ccstate {
310 	union {
311 		struct tcp_cubic_state {
312 			u_int32_t tc_last_max;  /* cwnd at last loss */
313 			u_int32_t tc_epoch_start; /* TS of last loss */
314 			u_int32_t tc_origin_point; /* window at the start of an epoch */
315 			u_int32_t tc_tcp_win; /* computed tcp win */
316 			u_int32_t tc_tcp_bytes_acked; /* bytes acked */
317 			u_int32_t tc_avg_lastmax; /* Average of last max */
318 			u_int32_t tc_mean_deviation; /* Mean absolute deviation */
319 			float     tc_epoch_period; /* K parameter */
320 		} _cubic_state_;
321 #define cub_last_max __u__._cubic_state_.tc_last_max
322 #define cub_epoch_start __u__._cubic_state_.tc_epoch_start
323 #define cub_origin_point __u__._cubic_state_.tc_origin_point
324 #define cub_tcp_win __u__._cubic_state_.tc_tcp_win
325 #define cub_tcp_bytes_acked __u__._cubic_state_.tc_tcp_bytes_acked
326 #define cub_epoch_period __u__._cubic_state_.tc_epoch_period
327 #define cub_avg_lastmax __u__._cubic_state_.tc_avg_lastmax
328 #define cub_mean_dev __u__._cubic_state_.tc_mean_deviation
329 		struct tcp_prague_state {
330 			uint16_t num_cong_events_loss;
331 			uint16_t num_cong_events_ce;
332 			uint32_t packets_acked;   /* cumulative packets ack'ed updated at round start for alpha */
333 			uint32_t packets_marked;  /* cumulative CE counts updated at round start for alpha */
334 			uint32_t ce_counter;      /* cumulative CE counts updated on every ACK,
335 			                           *                        used for calculating non-CE acked and performing any CE related action */
336 			uint32_t bytes_acked;     /* bytes acked in this ACK that were not CE marked, used for cwnd increase */
337 			uint32_t snd_nxt_alpha;   /* snd_nxt at round start to track 1 RTT elapsed for alpha */
338 			uint32_t snd_nxt_cwr;     /* snd_nxt at round start to track 1 RTT elapsed for CWR */
339 			uint8_t ever_saw_ce:1,       /* To test if there is a AQM on path */
340 			    in_loss:1,                       /* Are we in loss recovery */
341 			    reduced_due_to_ce:1,             /* Was the last reduction due to CE */
342 			    unused:5;
343 			uint8_t pad[3];
344 			uint64_t scaled_alpha;    /* Scaled value of DCTCP.alpha */
345 			uint64_t alpha_ai;        /* AI.alpha used after CE for additive increase */
346 			/* After reduction due to loss CUBIC window increase state */
347 			struct tcp_cubic_state cubic_state;
348 		} _prague_state_;
349 #define num_cong_events_loss __u__._prague_state_.num_cong_events_loss
350 #define num_cong_events_ce __u__._prague_state_.num_cong_events_ce
351 #define prague_packets_acked __u__._prague_state_.packets_acked
352 #define prague_packets_marked __u__._prague_state_.packets_marked
353 #define prague_ce_counter __u__._prague_state_.ce_counter
354 #define prague_bytes_acked __u__._prague_state_.bytes_acked
355 #define snd_nxt_alpha __u__._prague_state_.snd_nxt_alpha
356 #define snd_nxt_cwr __u__._prague_state_.snd_nxt_cwr
357 #define prague_scaled_alpha __u__._prague_state_.scaled_alpha
358 #define prague_alpha_ai __u__._prague_state_.alpha_ai
359 #define ever_saw_ce __u__._prague_state_.ever_saw_ce
360 #define in_loss __u__._prague_state_.in_loss
361 #define reduced_due_to_ce __u__._prague_state_.reduced_due_to_ce
362 #define cubic_K __u__._prague_state_.cubic_state.tc_epoch_period
363 #define cubic_epoch_start __u__._prague_state_.cubic_state.tc_epoch_start
364 #define cubic_origin_point __u__._prague_state_.cubic_state.tc_origin_point
365 #define cubic_W_max __u__._prague_state_.cubic_state.tc_last_max
366 #define reno_acked __u__._prague_state_.cubic_state.tc_tcp_bytes_acked
367 #define reno_cwnd __u__._prague_state_.cubic_state.tc_tcp_win
368 		struct tcp_ledbat_state {
369 			uint32_t num_slowdown_events;
370 			uint32_t slowdown_ts;
371 			uint32_t slowdown_begin;
372 			uint32_t md_bytes_acked;
373 		} _ledbat_state_;
374 #define ledbat_slowdown_events __u__._ledbat_state_.num_slowdown_events
375 #define ledbat_slowdown_ts __u__._ledbat_state_.slowdown_ts
376 #define ledbat_slowdown_begin __u__._ledbat_state_.slowdown_begin
377 #define ledbat_md_bytes_acked __u__._ledbat_state_.md_bytes_acked
378 	} __u__;
379 };
380 
381 struct tcp_rledbat_state {
382 	uint32_t num_slowdown_events;  /* Number of slowdown events until reset */
383 	uint32_t slowdown_ts;          /* Next slowdown timestamp */
384 	uint32_t slowdown_begin;       /* Slowdown begin time */
385 	uint32_t reduction_end;        /* Used to track rl_win reduction once per RTT */
386 	uint32_t rcvd_bytes;           /* bytes received used for byte counting */
387 	uint32_t md_rcvd_bytes;        /* bytes received during MD used for byte counting */
388 	uint32_t win;                  /* receive Ledbat window size */
389 	uint32_t ssthresh;             /* receive Ledbat ssthresh */
390 	uint32_t drained_bytes;        /* bytes drained from the flight-size */
391 	uint32_t win_ws;               /* receive Ledbat window after avoiding window shrinking */
392 };
393 
394 struct accecn {
395 	/* ACE CE packet counters */
396 	uint32_t        t_rcv_ce_packets;                  /* Number of CE received packets at the receiver */
397 	uint32_t        t_snd_ce_packets;                  /* Synced number of CE received feedback at the sender */
398 	uint32_t        t_delta_ce_packets;                /* Change in CE count between previous and current ACK */
399 	uint8_t         accecn_processed:1,                /* Accurate ECN feedback has already been processed */
400 	    unused:7;
401 
402 	/* AccECN option byte counters */
403 	uint64_t        t_rcv_ect1_bytes;                  /* ECT1 byte counter at the receiver, used for AccECN option feedback */
404 	uint64_t        t_rcv_ect0_bytes;                  /* ECT0 byte counter at the receiver, used for AccECN option feedback */
405 	uint64_t        t_rcv_ce_bytes;                    /* CE byte counter at the receiver, used for AccECN option feedback */
406 	uint64_t        t_snd_ect1_bytes;                  /* Synced ECT1 byte counter at the sender */
407 	uint64_t        t_snd_ect0_bytes;                  /* Synced ECT0 byte counter at the sender */
408 	uint64_t        t_snd_ce_bytes;                    /* Synced CE byte counter at the sender */
409 };
410 
411 struct pacer {
412 	uint64_t rate;
413 	uint32_t tso_burst_size; /* maximum allowed burst size, segments that fit in a burst have the same Tx timestamp */
414 	uint32_t current_size; /* track how many bytes have been accumulated in a burst */
415 	uint64_t packet_tx_time;
416 };
417 
418 /*
419  * Tcp control block, one per tcp; fields:
420  * Organized for 16 byte cacheline efficiency.
421  */
422 struct tcpcb {
423 	struct tsegqe_head t_segq;
424 	uint32_t t_dupacks;             /* consecutive dup acks recd */
425 	int      t_state;               /* state of this connection */
426 	uint32_t t_timer[TCPT_NTIMERS]; /* tcp timers */
427 	struct tcptimerentry tentry;    /* entry in timer list */
428 
429 	struct  inpcb *t_inpcb;         /* back pointer to internet pcb */
430 	uint32_t        t_flags;
431 #define TF_ACKNOW       0x00001         /* ack peer immediately */
432 #define TF_DELACK       0x00002         /* ack, but try to delay it */
433 #define TF_NODELAY      0x00004         /* don't delay packets to coalesce */
434 #define TF_NOOPT        0x00008         /* don't use tcp options */
435 #define TF_SENTFIN      0x00010         /* have sent FIN */
436 #define TF_REQ_SCALE    0x00020         /* have/will request window scaling */
437 #define TF_RCVD_SCALE   0x00040         /* other side has requested scaling */
438 #define TF_REQ_TSTMP    0x00080         /* have/will request timestamps */
439 #define TF_RCVD_TSTMP   0x00100         /* a timestamp was received in SYN */
440 #define TF_SACK_PERMIT  0x00200         /* other side said I could SACK */
441 #define TF_NEEDSYN      0x00400         /* send SYN (implicit state) - unused but needed for backwards compatibility */
442 #define TF_NEEDFIN      0x00800         /* send FIN (implicit state) */
443 #define TF_NOPUSH       0x01000         /* don't push */
444 #define TF_UNUSED1      0x02000         /* Unused */
445 #define TF_UNUSED2      0x04000         /* Unused */
446 #define TF_UNUSED3      0x08000         /* Unused */
447 #define TF_MORETOCOME   0x10000         /* More data to be appended to sock */
448 #define TF_LOCAL        0x20000         /* connection to a host on local link */
449 #define TF_RXWIN0SENT   0x40000         /* sent a receiver win 0 in response */
450 #define TF_SLOWLINK     0x80000         /* route is a on a modem speed link */
451 #define TF_LASTIDLE     0x100000        /* connection was previously idle */
452 #define TF_FASTRECOVERY 0x200000        /* in NewReno Fast Recovery */
453 #define TF_WASFRECOVERY 0x400000        /* was in NewReno Fast Recovery */
454 #define TF_SIGNATURE    0x800000        /* require MD5 digests (RFC2385) */
455 #define TF_MAXSEGSNT    0x1000000       /* last segment sent was a full segment */
456 /* Unused 0x2000000 */
457 #define TF_PMTUD        0x4000000       /* Perform Path MTU Discovery for this connection */
458 #define TF_CLOSING      0x8000000       /* pending tcp close */
459 #define TF_TSO          0x10000000      /* TCP Segment Offloading is enable on this connection */
460 #define TF_BLACKHOLE    0x20000000      /* Path MTU Discovery Black Hole detection */
461 #define TF_TIMER_ONLIST 0x40000000      /* pcb is on tcp_timer_list */
462 /* Unused 0x80000000 */
463 
464 	tcp_seq snd_una;                /* send unacknowledged */
465 	tcp_seq snd_max;                /* highest sequence number sent;
466 	                                 * used to recognize retransmits
467 	                                 */
468 	tcp_seq snd_nxt;                /* send next */
469 	tcp_seq snd_up;                 /* send urgent pointer */
470 
471 	tcp_seq snd_wl1;                /* window update seg seq number */
472 	tcp_seq snd_wl2;                /* window update seg ack number */
473 	tcp_seq iss;                    /* initial send sequence number */
474 	tcp_seq irs;                    /* initial receive sequence number */
475 
476 	tcp_seq rcv_nxt;                /* receive next */
477 	tcp_seq rcv_adv;                /* advertised window */
478 	uint32_t        rcv_wnd;        /* receive window */
479 	uint32_t        t_last_recwin;
480 	tcp_seq rcv_up;                 /* receive urgent pointer */
481 	uint32_t        t_latest_tx;    /* Most recent transmit scheduled (including potential pacing) */
482 
483 	uint32_t        snd_wnd;        /* send window */
484 	uint32_t        snd_cwnd;       /* congestion-controlled window */
485 	uint32_t        snd_ssthresh;   /* snd_cwnd size threshold for
486 	                                 * for slow start exponential to
487 	                                 * linear switch
488 	                                 */
489 	tcp_seq snd_recover;            /* for use in NewReno Fast Recovery */
490 
491 	uint32_t        t_maxopd;       /* mss plus options */
492 	uint32_t        t_rcvtime;      /* time at which a packet was received */
493 	uint32_t        t_sndtime;      /* time at which we last sent new data */
494 	uint32_t        t_starttime;    /* time connection was established */
495 	int     t_rtttime;              /* tcp clock when rtt calculation was started */
496 	tcp_seq t_rtseq;                /* sequence number being timed */
497 
498 	uint32_t rfbuf_ts;              /* recv buffer autoscaling timestamp */
499 	uint32_t rfbuf_cnt;             /* recv buffer autoscaling byte count */
500 	uint32_t rfbuf_space;           /* Current "ideal" estimate of the space */
501 
502 	int            t_rxtcur;        /* current retransmit value (ticks) */
503 	unsigned int   t_maxseg;        /* maximum segment size */
504 	int            t_srtt;          /* smoothed round-trip time */
505 	int            t_rttvar;        /* variance in round-trip time */
506 
507 	uint64_t t_accsleep_ms;         /* accumulated sleep time since last boot */
508 	uint32_t t_reassqlen;           /* length of reassembly queue */
509 	uint32_t t_reassq_mbcnt;        /* amount in bytes of mbuf space used */
510 	uint16_t t_rxtshift;            /* log(2) of rexmt exp. backoff */
511 	uint32_t t_rttmin;              /* minimum rtt allowed */
512 	uint32_t t_rttbest;             /* best rtt we've seen */
513 	uint32_t t_rttcur;              /* most recent value of rtt */
514 	uint32_t t_rttupdated;          /* number of times rtt sampled */
515 	uint32_t t_rxt_conndroptime;    /* retxmt conn gets dropped after this time, when set */
516 	uint32_t t_rxtstart;            /* time at which retransmission started */
517 	uint32_t max_sndwnd;            /* largest window peer has offered */
518 
519 	int     t_softerror;            /* possible error not yet reported */
520 /* out-of-band data */
521 	char    t_oobflags;             /* have some */
522 	char    t_iobc;                 /* input character */
523 #define TCPOOB_HAVEDATA 0x01
524 #define TCPOOB_HADDATA  0x02
525 /* RFC 1323 variables */
526 	uint8_t         snd_scale;      /* window scaling for send window */
527 	uint8_t         rcv_scale;      /* window scaling for recv window */
528 	uint8_t         request_r_scale; /* pending window scaling */
529 	uint8_t         requested_s_scale;
530 	uint8_t         tcp_cc_index;   /* index of congestion control algorithm */
531 	uint8_t         t_adaptive_rtimo;       /* Read timeout used as a multiple of RTT */
532 	uint8_t         t_adaptive_wtimo;       /* Write timeout used as a multiple of RTT */
533 
534 /* State for limiting early retransmits when SACK is not enabled */
535 	uint16_t        t_early_rexmt_count; /* count of early rexmts */
536 	uint32_t        t_early_rexmt_win; /* window for limiting early rexmts */
537 
538 	uint32_t        ts_recent;      /* timestamp echo data */
539 	uint32_t        t_ts_offset; /* Randomized timestamp offset to hide on-the-wire timestamp */
540 
541 	uint32_t        ts_recent_age;  /* when last updated */
542 	tcp_seq         last_ack_sent;
543 
544 	uint32_t        t_bytes_acked;  /* RFC 3465 variable for ABC, used by CCA only */
545 	uint32_t        total_ect_packets_marked; /* Cumulative count of total ECT packets marked */
546 	uint32_t        total_ect_packets_acked;  /* Cumulative count of total ECT packets acked */
547 
548 	int             t_lastchain;    /* amount of packets chained last time around */
549 	uint16_t        t_unacksegs;    /* received but unacked segments for delaying acks */
550 	uint16_t        t_unacksegs_ce; /* received but unacked segments/pure ACKs that were CE marked */
551 
552 	/*
553 	 * Pretty arbitrary value ;-)
554 	 * Goal is to make sure that some ACKs are being sent more frequently
555 	 * to allow the other side to ramp-up.
556 	 */
557 #define TCP_FORCED_ACKS_COUNT 16
558 	uint16_t        t_forced_acks;  /* count of pure ACKs that need to be forced out */
559 	uint8_t         t_rexmtthresh;  /* duplicate ack threshold for entering fast recovery */
560 	uint8_t         t_rtimo_probes; /* number of adaptive rtimo probes sent */
561 	uint32_t        t_persist_timeout; /* ZWP persistence limit as set by PERSIST_TIMEOUT */
562 	uint32_t        t_persist_stop;    /* persistence limit deadline if triggered by ZWP */
563 	uint32_t        t_notsent_lowat;   /* Low water for not sent data */
564 
565 /* ECN stats */
566 	uint32_t        ecn_flags;
567 #define TE_SETUPSENT            0x00000001  /* We have sent classic ECN-SETUP SYN or SYN-ACK */
568 #define TE_SETUPRECEIVED        0x00000002  /* We have received classic ECN-SETUP SYN or SYN-ACK */
569 #define TE_SENDIPECT            0x00000004  /* We haven't sent or received non-ECN-setup SYN or SYN-ACK, set IP ECT on outbound packet */
570 #define TE_SENDCWR              0x00000008  /* Next non-retransmit should have TCP CWR set, only used for classic ECN */
571 #define TE_SENDECE              0x00000010  /* Next packet should have TCP ECE set, only used for classic ECN */
572 #define TE_INRECOVERY           0x00000020  /* connection entered recovery after receiving ECE */
573 #define TE_RECV_ECN_CE          0x00000040  /* Received IPTOS_ECN_CE marking atleast once */
574 #define TE_RECV_ECN_ECE         0x00000080  /* Received ECE marking atleast once */
575 #define TE_LOST_SYN             0x00000100  /* Lost SYN with ECN setup */
576 #define TE_LOST_SYNACK          0x00000200  /* Lost SYN-ACK with ECN setup */
577 #define TE_ECN_MODE_ENABLE      0x00000400  /* Option ECN mode set to enable */
578 #define TE_ECN_MODE_DISABLE     0x00000800  /* Option ECN mode set to disable */
579 #define TE_ENABLE_ECN           0x00001000  /* Enable negotiation of ECN */
580 #define TE_ECN_ON               (TE_SETUPSENT | TE_SETUPRECEIVED) /* ECN was successfully negotiated on a connection */
581 #define TE_ECEHEURI_SET         0x00002000 /* We did our E/CE-probing at the beginning */
582 #define TE_CLIENT_SETUP         0x00004000  /* setup from client side */
583 #define TE_RCVD_SYN_RST         0x00008000  /* Received RST to the first ECN enabled SYN */
584 #define TE_ACE_SETUP_NON_ECT    0x00010000  /* Encode received non-ECT either for SYN-ACK (server) or final ACK (client) */
585 #define TE_ACE_SETUP_ECT1       0x00020000  /* Encode received ECT1 either for SYN-ACK (server) or final ACK (client) */
586 #define TE_ACE_SETUP_ECT0       0x00040000  /* Encode received ECT0 either for SYN-ACK (server) or final ACK (client) */
587 #define TE_ACE_SETUP_CE         0x00080000  /* Encode received CE either for SYN-ACK (server) or final ACK (client) */
588 #define TE_ACE_SETUPSENT        0x00100000  /* We have sent Accurate ECN setup SYN or SYN-ACK */
589 #define TE_ACE_SETUPRECEIVED    0x00200000  /* We have received Accurate ECN setup SYN or SYN-ACK */
590 #define TE_ACC_ECN_ON          (TE_ACE_SETUPSENT | TE_ACE_SETUPRECEIVED) /* Accurate ECN was negotiated */
591 #define TE_ACE_FINAL_ACK_3WHS   0x00400000  /* Client has received SYN-ACK and will now send final ACK of 3WHS, only used for AccECN */
592 #define TE_ACO_ECT1             0x00800000  /* ECT1 counter changed flag, used to decide ordering for AccECN option */
593 #define TE_ACO_ECT0             0x01000000  /* ECT0 counter changed flag, used to decide ordering for AccECN option */
594 #define TE_RETRY_WITHOUT_ACO    0x02000000  /* Data segment with AccECN option was not acknowledged, retry without AccECN option */
595 #define TE_FORCE_ECT1           0x40000000  /* Force setting ECT1 on outgoing packets for testing purpose */
596 #define TE_FORCE_ECT0           0x80000000  /* Force setting ECT0 on outgoing packets for testing purpose */
597 
598 	uint32_t        t_ecn_recv_ce;  /* Received CE from the network */
599 	uint32_t        t_ecn_recv_cwr; /* Packets received with CWR */
600 	uint32_t        t_client_accecn_state;    /* Client's Accurate ECN state */
601 	uint32_t        t_server_accecn_state;    /* Server's Accurate ECN state */
602 	uint64_t        t_ecn_capable_packets_sent;     /* Packets sent with ECT */
603 	uint64_t        t_ecn_capable_packets_acked;    /* Packets sent with ECT that were acked */
604 	uint64_t        t_ecn_capable_packets_marked;   /* Packets sent with ECT that were marked */
605 	uint64_t        t_ecn_capable_packets_lost;     /* Packets sent with ECT that were lost */
606 
607 	uint32_t        t_last_ack_tsecr;       /* TS Echo Reply for last ACK that acknowledged a data or control packet */
608 	uint16_t        t_prev_ace_flags;   /* ACE flags that were sent in previous packet, used for retransmitting after timeout */
609 	uint8_t         t_prev_ip_ecn;      /* IP ECN flag on the previous packet, used for change-triggered ACKs */
610 
611 	struct accecn   t_aecn;         /* AccECN related byte counters */
612 
613 	struct pacer    t_pacer;        /* Pacer state used to pace packets */
614 
615 /* state for bad retransmit recovery */
616 	uint32_t        snd_cwnd_prev;  /* cwnd prior to retransmit */
617 	uint32_t        snd_ssthresh_prev; /* ssthresh prior to retransmit */
618 	tcp_seq         snd_recover_prev;       /* snd_recover prior to retransmit */
619 	int             t_srtt_prev;            /* srtt prior to retransmit */
620 	int             t_rttvar_prev;          /* rttvar prior to retransmit */
621 	uint32_t        t_badrexmt_time; /* bad rexmt detection time */
622 
623 /* Packet reordering metric */
624 	uint32_t        t_reorderwin; /* Reordering late time offset */
625 
626 /* SACK related state */
627 	int16_t snd_numholes;           /* number of holes seen by sender */
628 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
629 	/* SACK scoreboard (sorted) */
630 	tcp_seq snd_fack;               /* last seq number(+1) sack'd by rcv'r*/
631 	int     rcv_numsacks;           /* # distinct sack blks present */
632 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
633 	struct sackhint sackhint;       /* SACK scoreboard hint */
634 
635 	struct mbuf     *t_pktlist_head; /* First packet in transmit chain */
636 	struct mbuf     *t_pktlist_tail; /* Last packet in transmit chain */
637 	uint32_t        t_pktlist_sentlen; /* total bytes in transmit chain */
638 
639 	uint32_t        t_keepidle;     /* keepalive idle timer (override global if > 0) */
640 	uint32_t        t_keepinit;     /* connection timeout, i.e. idle time
641 	                                 *  in SYN_SENT or SYN_RECV state */
642 	uint32_t        t_keepintvl;    /* interval between keepalives */
643 	uint32_t        t_keepcnt;      /* number of keepalives before close */
644 
645 	uint32_t        tso_max_segment_size;   /* TSO maximum segment unit for NIC */
646 	uint16_t        t_pmtud_lastseg_size;   /* size of the last sent segment */
647 	uint32_t        t_pmtud_saved_maxopd;   /* MSS saved before performing PMTU-D BlackHole detection */
648 	uint32_t        t_pmtud_start_ts;       /* Time of PMTUD blackhole detection */
649 
650 	struct{
651 		uint32_t        rxduplicatebytes;
652 		uint32_t        rxoutoforderbytes;
653 		uint32_t        txretransmitbytes;
654 		uint16_t        synrxtshift;
655 		uint16_t        rxmitsyns;
656 		uint16_t        unused_pad_to_8;
657 		uint32_t        rxmitpkts;
658 		uint32_t        delayed_acks_sent;
659 		uint32_t        acks_delayed;
660 		uint64_t        bytes_acked;
661 	} t_stat;
662 	uint8_t         t_syn_sent;
663 	uint8_t         t_syn_rcvd;
664 	uint8_t         t_notify_ack_count;
665 	uint8_t         t_ecn_recv_ce_pkt; /* Received data packet with CE bit set */
666 	uint8_t         t_ecn_recv_ece_pkt; /* Received ACK packet with ECE bit set */
667 	uint32_t        t_cached_maxopd; /* default for MSS adjustment using link status report */
668 
669 	uint32_t        bg_ssthresh;            /* Slow start threshold until delay increases */
670 	uint32_t        t_flagsext;             /* Another field to accommodate more flags */
671 #define TF_RXTFINDROP           0x1             /* Drop conn after retransmitting FIN 3 times */
672 /* Unused 0x2 */
673 #define TF_BWMEAS_INPROGRESS    0x4             /* Indicate BW meas is happening */
674 #define TF_MEASURESNDBW         0x8             /* Measure send bw on this connection */
675 #define TF_LAST_IS_PSH          0x10            /* Indicates whether the last packet in the rcv socket buffer had the PUSH-flag set */
676 #define TF_SACK_ENABLE          0x20            /* SACK is enabled */
677 #define TF_RECOMPUTE_RTT        0x40            /* recompute RTT after spurious retransmit */
678 #define TF_DETECT_READSTALL     0x80            /* Used to detect a stall during read operation */
679 #define TF_RECV_THROTTLE        0x100           /* Input throttling active */
680 #define TF_QUICKACK             0x200           /* Force-ACK every other packet */
681 #define TF_SYN_COOKIE_ENABLED   0x400           /* SYN cookie is enabled for listener when max backlog is reached */
682 #define TF_NOTIMEWAIT           0x800           /* Avoid going into time-wait */
683 #define TF_SENT_TLPROBE         0x1000          /* Sent data in PTO */
684 #define TF_PKTS_REORDERED       0x2000          /* Detected reordering */
685 #define TF_DELAY_RECOVERY       0x4000          /* delay fast recovery */
686 #define TF_FORCE                0x8000          /* force 1 byte out */
687 /* Unused 0x10000 */
688 #define TF_NOBLACKHOLE_DETECTION 0x20000        /* Disable PMTU blackhole detection */
689 #define TF_SYN_COOKIE_FORCE_ENABLED 0x40000     /* SYN cookie is enabled for listener unconditionally */
690 #define TF_RESCUE_RXT           0x80000         /* SACK rescue retransmit */
691 #define TF_CWND_NONVALIDATED    0x100000        /* cwnd non validated */
692 #define TF_IF_PROBING           0x200000        /* Trigger interface probe timeout */
693 #define TF_FASTOPEN             0x400000        /* TCP Fastopen is enabled */
694 #define TF_REASS_INPROG         0x800000        /* Reassembly is in progress */
695 #define TF_FASTOPEN_FORCE_ENABLE 0x1000000      /* Force-enable TCP Fastopen */
696 #define TF_USR_OUTPUT           0x2000000       /* In connect() or send() so tcp_output() can log */
697 #define TF_L4S_ENABLED          0x8000000       /* L4S was force enabled */
698 #define TF_L4S_DISABLED         0x10000000      /* L4S was force disabled */
699 #define TF_RACK_ENABLED         0x20000000      /* RACK is enabled */
700 #define TF_TLP_IS_RETRANS       0x40000000      /* Is TLP-probe a retransmission ? (field TLP.is_retrans in RFC 8985) */
701 
702 #if TRAFFIC_MGT
703 	/* Inter-arrival jitter related state */
704 	uint32_t        iaj_rcv_ts;             /* tcp clock when the first packet was received */
705 	int             iaj_size;               /* Size of packet for iaj measurement */
706 	uint8_t         iaj_small_pkt;          /* Count of packets smaller than iaj_size */
707 	uint8_t         t_pipeack_ind;          /* index for next pipeack sample */
708 	uint16_t        iaj_pktcnt;             /* packet count, to avoid throttling initially */
709 	uint32_t        acc_iaj;                /* Accumulated iaj */
710 	uint32_t        avg_iaj;                /* Mean */
711 	uint32_t        std_dev_iaj;            /* Standard deviation */
712 #endif /* TRAFFIC_MGT */
713 	struct bwmeas   *t_bwmeas;              /* State for bandwidth measurement */
714 	tcp_seq         t_idleat;               /* rcv_nxt at idle time */
715 	uint8_t         t_fin_sent;
716 	uint8_t         t_fin_rcvd;
717 	uint8_t         t_rst_sent;
718 	uint8_t         t_rst_rcvd;
719 	TAILQ_ENTRY(tcpcb) t_twentry;           /* link for time wait queue */
720 	struct tcp_ccstate      *t_ccstate;     /* congestion control related state */
721 	struct tcp_ccstate      _t_ccstate;     /* congestion control related state, non-allocated */
722 /* Tail loss probe related state */
723 	tcp_seq         t_tlphighrxt;           /* snd_nxt after PTO */
724 	tcp_seq         t_tlphightrxt_persist;  /* like t_tlphighrxt but persists over ACKs until DSACK (if any) is processed */
725 	uint32_t        t_tlpstart;             /* timestamp at PTO */
726 /* DSACK data receiver state */
727 	tcp_seq         t_dsack_lseq;           /* DSACK left sequence */
728 	tcp_seq         t_dsack_rseq;           /* DSACK right sequence */
729 /* DSACK data sender state */
730 	SLIST_HEAD(tcp_rxt_seghead, tcp_rxt_seg) t_rxt_segments;
731 	uint32_t        t_rxt_seg_count;
732 	uint32_t        t_rxt_seg_drop;
733 	tcp_seq         t_dsack_lastuna;        /* snd_una when last recovery episode started */
734 /* state for congestion window validation (draft-ietf-tcpm-newcwv-07) */
735 #define TCP_PIPEACK_SAMPLE_COUNT        3
736 	uint32_t        t_pipeack_sample[TCP_PIPEACK_SAMPLE_COUNT];     /* pipeack, bytes acked within RTT */
737 	tcp_seq         t_pipeack_lastuna; /* una when pipeack measurement started */
738 	uint32_t        t_pipeack;
739 	uint32_t        t_lossflightsize;
740 
741 #if MPTCP
742 	uint32_t        t_mpflags;              /* flags for multipath TCP */
743 
744 #define TMPF_PREESTABLISHED     0x00000001 /* conn in pre-established state */
745 #define TMPF_SND_KEYS           0x00000002 /* indicates that keys should be send */
746 #define TMPF_MPTCP_TRUE         0x00000004 /* negotiated MPTCP successfully */
747 /* UNUSED */
748 #define TMPF_SND_MPPRIO         0x00000010 /* send priority of subflow */
749 #define TMPF_SND_REM_ADDR       0x00000020 /* initiate address removal */
750 #define TMPF_RCVD_DACK          0x00000040 /* received a data-ack */
751 #define TMPF_JOINED_FLOW        0x00000080 /* Indicates additional flow */
752 #define TMPF_BACKUP_PATH        0x00000100 /* Indicates backup path */
753 #define TMPF_MPTCP_ACKNOW       0x00000200 /* Send Data ACK */
754 #define TMPF_SEND_DSN           0x00000400 /* Send DSN mapping */
755 #define TMPF_SEND_DFIN          0x00000800 /* Send Data FIN */
756 #define TMPF_RECV_DFIN          0x00001000 /* Recv Data FIN */
757 #define TMPF_SENT_JOIN          0x00002000 /* Sent Join */
758 #define TMPF_RECVD_JOIN         0x00004000 /* Received Join */
759 #define TMPF_RESET              0x00008000 /* Send RST */
760 #define TMPF_TCP_FALLBACK       0x00010000 /* Fallback to TCP */
761 #define TMPF_FASTCLOSERCV       0x00020000 /* Received Fastclose option */
762 #define TMPF_EMBED_DSN          0x00040000 /* tp has DSN mapping */
763 #define TMPF_MPTCP_READY        0x00080000 /* Can send DSS options on data */
764 #define TMPF_INFIN_SENT         0x00100000 /* Sent infinite mapping */
765 #define TMPF_SND_MPFAIL         0x00200000 /* Received mapping csum failure */
766 #define TMPF_SND_JACK           0x00400000 /* Send a Join-ACK */
767 #define TMPF_TFO_REQUEST        0x00800000 /* TFO Requested */
768 #define TMPF_MPTCP_ECHO_ADDR    0x01000000 /* MPTCP echoes add_addr */
769 
770 #define TMPF_MPTCP_SIGNALS      (TMPF_SND_MPPRIO | TMPF_SND_REM_ADDR | TMPF_SND_MPFAIL | TMPF_SND_JACK | TMPF_MPTCP_ECHO_ADDR)
771 
772 	tcp_seq                 t_mpuna;        /* unacknowledged sequence */
773 	struct mptcb            *t_mptcb;       /* pointer to MPTCP TCB */
774 	struct mptsub           *t_mpsub;       /* pointer to the MPTCP subflow */
775 	uint8_t                 t_local_aid;    /* Addr Id for authentication */
776 	uint8_t                 t_rem_aid;      /* Addr ID of another subflow */
777 	uint8_t                 t_mprxtshift;   /* join retransmission */
778 #endif /* MPTCP */
779 
780 #define TFO_F_OFFER_COOKIE      0x01 /* We will offer a cookie */
781 #define TFO_F_COOKIE_VALID      0x02 /* The received cookie is valid */
782 #define TFO_F_COOKIE_REQ        0x04 /* Client requested a new cookie */
783 #define TFO_F_COOKIE_SENT       0x08 /* Client did send a cookie in the SYN */
784 #define TFO_F_SYN_LOSS          0x10 /* A SYN-loss triggered a fallback to regular TCP on the client-side */
785 #define TFO_F_NO_SNDPROBING     0x20 /* This network is guaranteed to support TFO in the upstream direction */
786 #define TFO_F_HEURISTIC_DONE    0x40 /* We have already marked this network as bad */
787 	uint8_t                 t_tfo_flags;
788 #define TFO_S_SYNDATA_RCV       0x01 /* SYN+data has been received */
789 #define TFO_S_COOKIEREQ_RECV    0x02 /* TFO-cookie request received */
790 #define TFO_S_COOKIE_SENT       0x04 /* TFO-cookie announced in SYN/ACK */
791 #define TFO_S_COOKIE_INVALID    0x08 /* Received TFO-cookie is invalid */
792 #define TFO_S_COOKIE_REQ        0x10 /* TFO-cookie requested within the SYN */
793 #define TFO_S_COOKIE_RCV        0x20 /* TFO-cookie received in SYN/ACK */
794 #define TFO_S_SYN_DATA_SENT     0x40 /* SYN+data sent */
795 #define TFO_S_SYN_DATA_ACKED    0x80 /* SYN+data has been acknowledged in SYN/ACK */
796 #define TFO_S_SYN_LOSS          0x0100 /* SYN+TFO has been lost - fallback to regular TCP */
797 #define TFO_S_COOKIE_WRONG      0x0200 /* Cookie we sent in the SYN was wrong */
798 #define TFO_S_NO_COOKIE_RCV     0x0400 /* We asked for a cookie but didn't get one */
799 #define TFO_S_HEURISTICS_DISABLE 0x0800 /* TFO-heuristics disabled it for this connection */
800 #define TFO_S_SEND_BLACKHOLE    0x1000 /* TFO got blackholed in the send direction */
801 #define TFO_S_RECV_BLACKHOLE    0x2000 /* TFO got blackholed in the recv direction */
802 #define TFO_S_ONE_BYTE_PROXY    0x4000 /* TFO failed because of a proxy acknowledging just one byte */
803 	uint16_t                t_tfo_stats;
804 
805 	uint8_t                 t_tfo_probes; /* TFO-probes we did send */
806 /*
807  * This here is the TFO-probing state-machine. Transitions are as follows:
808  *
809  * Current state: PROBE_NONE
810  *		  Event: SYN+DATA acknowledged
811  *			 Action: Transition to PROBE_PROBING and set keepalive-timer
812  *
813  * Current state: PROBE_PROBING (initial state)
814  *		  Event: Receive data
815  *			 Action: Transition to PROBE_NONE and cancel keepalive-timer
816  *		  Event: Receive ACK that does not indicate a hole
817  *			 Action: Transition to PROBE_NONE and cancel keepalive-timer
818  *		  Event: Receive ACK that indicates a hole
819  *			 Action: Transition to PROBE_WAIT_DATA and set a short timer
820  *				 to wait for the final segment.
821  *		  Event: Keepalive-timeout (did not receive any segment)
822  *			 Action: Signal ETIMEDOUT as with regular keepalive-timers
823  *
824  * Current state: PROBE_WAIT_DATA
825  *		  Event: Receive data
826  *			 Action: Transition to PROBE_NONE and cancel keepalive-timer
827  *		  Event: Data-timeout (did not receive the expected data)
828  *			 Action: Signal ENODATA up to the app and close everything.
829  */
830 #define TFO_PROBE_NONE          0 /* Not probing now */
831 #define TFO_PROBE_PROBING       1 /* Sending out TCP-keepalives waiting for reply */
832 #define TFO_PROBE_WAIT_DATA     2 /* Received reply, waiting for data */
833 	uint8_t                t_tfo_probe_state;
834 
835 	uint32_t        t_rcvoopack;            /* out-of-order packets received */
836 	uint32_t        t_pawsdrop;             /* segments dropped due to PAWS */
837 	uint32_t        t_sack_recovery_episode; /* SACK recovery episodes */
838 	uint32_t        t_rack_recovery_episode; /* RACK recovery episodes */
839 	uint32_t        t_rack_reo_timeout_recovery_episode; /* RACK recovery triggered by reordering timeout */
840 	uint32_t        t_reordered_pkts;       /* packets reorderd */
841 	uint32_t        t_dsack_sent;           /* Sent DSACK notification */
842 	uint32_t        t_dsack_recvd;          /* Received a valid DSACK option */
843 	SLIST_HEAD(, tcp_notify_ack_marker) t_notify_ack; /* state for notifying data acknowledgements */
844 	uint32_t        t_recv_throttle_ts;     /* TS for start of recv throttle */
845 	uint32_t        t_rxt_minimum_timeout;  /* minimum retransmit timeout in ms */
846 	uint32_t        t_challengeack_last;    /* last time challenge ACK was sent per sec */
847 	uint32_t        t_challengeack_count;   /* # of challenge ACKs already sent per sec */
848 
849 	uint32_t        t_connect_time;         /* time when the connection started */
850 
851 	uint64_t        t_rcvwnd_limited_total_time;
852 	uint64_t        t_rcvwnd_limited_start_time;
853 
854 
855 	uint32_t        t_comp_rxmt_gencnt; /* Current compression generation-count for segments */
856 
857 	uint32_t        t_comp_ack_gencnt; /* Current compression generation-count for ACKs */
858 	uint32_t        t_comp_ack_lastinc; /* Last time the gen-count was changed - should change every TCP_COMP_CHANGE_RATE ms */
859 #define TCP_COMP_CHANGE_RATE    5 /* Intervals at which we change the gencnt. Means that worst-case we send one ACK every TCP_COMP_CHANGE_RATE ms */
860 
861 #define NCURR_RTT_HIST 4                         /* Number of current RTT samples (k) */
862 	uint32_t curr_rtt_hist[NCURR_RTT_HIST];  /* last k current RTT samples */
863 	uint32_t curr_rtt_min;                   /* Minimum current RTT from last k samples */
864 	uint32_t curr_rtt_index;                 /* Index for current RTT samples */
865 
866 	tcp_seq rcv_high;                   /* highest sequence number received */
867 	uint32_t tsv_high;                  /* timestamp value of highest received sequence number */
868 	struct tcp_rledbat_state t_rlstate; /* State used by rLedbat */
869 
870 	uint32_t rcv_srtt;                  /* receiver's SRTT, coarse when Timestamp not supported */
871 	uint32_t rcv_rtt_est_ts;            /* start of measurement for estimating RTT when Timestamp not supported */
872 	uint32_t rcv_rtt_est_seq;           /* expected sequence number for completion of 1RTT when Timestamp not supported */
873 
874 	TAILQ_HEAD(tcp_seg_sent_head, tcp_seg_sent) t_segs_sent; /* Time ordered segment list used for RACK */
875 	struct tcp_seg_sent_tree_head t_segs_sent_tree; /* RB tree to track S/ACKED segments used for RACK */
876 	TAILQ_HEAD(tcp_seg_acked_head, tcp_seg_sent) t_segs_acked; /* Temporary storage for ACKed segments used for RACK */
877 	struct tcp_seg_pool seg_pool;
878 
879 # define TCP_RACK_RECOVERY_PERSIST_MAX (16)
880 	struct tcp_rack {
881 		uint32_t        xmit_ts;
882 		uint32_t        end_seq;
883 		uint32_t        rtt; /* RTT of the most recently delivered segment that was not marked as invalid as a possible spurious retransmission. */
884 		tcp_seq         dsack_round_end;
885 		uint32_t        reo_wnd;
886 		uint8_t         reo_wnd_multi;
887 		uint8_t         reo_wnd_persist:5,
888 		    advanced:1,
889 		    dsack_round_seen:1,
890 		    segs_retransmitted:1;
891 	} rack;
892 
893 	/*
894 	 * Below counters are used to estmitate current flight size,
895 	 * hence these counters reflect current values instead of total
896 	 */
897 	uint32_t bytes_lost;
898 	uint32_t bytes_retransmitted;
899 	uint32_t bytes_sacked;
900 
901 	uint8_t l4s_enabled:1,
902 	    accurate_ecn_on:1,
903 	    _pad:6;
904 
905 	uuid_t          t_fsw_uuid;
906 	uuid_t          t_flow_uuid;
907 };
908 
909 __CCT_DECLARE_CONSTRAINED_PTR_TYPES(struct tcpcb, tcpcb);
910 
911 #define IN_FASTRECOVERY(tp)     (tp->t_flags & TF_FASTRECOVERY)
912 #define SACK_ENABLED(tp)        (tp->t_flagsext & TF_SACK_ENABLE)
913 #define TFO_ENABLED(tp)         (tp->t_flagsext & TF_FASTOPEN)
914 #define TCP_RACK_ENABLED(tp)    ((tp->t_flagsext & TF_RACK_ENABLED) && SACK_ENABLED(tp) && !TFO_ENABLED(tp))
915 
916 extern int tcp_syncookie;
917 extern int soqlimitcompat;
918 extern int mptcp_enable;
919 
920 #define LOOPBACK_INTERFACE(tp) (tp->t_inpcb->inp_boundifp != NULL && \
921     (tp->t_inpcb->inp_boundifp->if_flags & IFF_LOOPBACK))
922 
923 #define TCP_SYN_COOKIE_DISABLED(tp) (TFO_ENABLED(tp) || tp->t_mpflags & TMPF_MPTCP_TRUE || LOOPBACK_INTERFACE(tp))
924 
925 #define TCP_SYN_COOKIE_FORCE_ENABLED(tp) ((tp->t_flagsext & TF_SYN_COOKIE_FORCE_ENABLED) \
926     && !TCP_SYN_COOKIE_DISABLED(tp))
927 
928 #define TCP_SYN_COOKIE_ENABLED(tp) ((tp->t_flagsext & (TF_SYN_COOKIE_ENABLED | TF_SYN_COOKIE_FORCE_ENABLED)) \
929     && !TCP_SYN_COOKIE_DISABLED(tp))
930 
931 static __inline bool
tcp_can_send_syncookie(const struct socket * head,const struct tcpcb * tp,const uint8_t th_flags)932 tcp_can_send_syncookie(const struct socket *head, const struct tcpcb *tp, const uint8_t th_flags)
933 {
934 	int backlog = soqlimitcompat ? head->so_qlimit : (3 * head->so_qlimit / 2);
935 	backlog = backlog - (backlog >> 3);
936 	bool backlog_reached = head->so_incqlen >= (backlog >> 1) || head->so_qlen >= backlog;
937 	bool can_send_syncookie = (TCP_SYN_COOKIE_FORCE_ENABLED(tp) || (TCP_SYN_COOKIE_ENABLED(tp) && backlog_reached));
938 	can_send_syncookie &=  ((th_flags & (TH_RST | TH_ACK | TH_SYN)) == TH_SYN);
939 	return can_send_syncookie;
940 }
941 
942 static inline bool
tcp_sent_tlp_retrans(const struct tcpcb * tp)943 tcp_sent_tlp_retrans(const struct tcpcb *tp)
944 {
945 	return (tp->t_flagsext & (TF_SENT_TLPROBE | TF_TLP_IS_RETRANS)) == (TF_SENT_TLPROBE | TF_TLP_IS_RETRANS);
946 }
947 
948 /*
949  * If the connection is in a throttled state due to advisory feedback from
950  * the interface output queue, reset that state. We do this in favor
951  * of entering recovery because the data transfer during recovery
952  * should be just a trickle and it will help to improve performance.
953  * We also do not want to back off twice in the same RTT.
954  */
955 #define ENTER_FASTRECOVERY(_tp_) do {                                           \
956 	(_tp_)->t_flags |= TF_FASTRECOVERY;                                     \
957 	if (INP_IS_FLOW_CONTROLLED((_tp_)->t_inpcb))                            \
958 	        inp_reset_fc_state((_tp_)->t_inpcb);                            \
959 	if (!SLIST_EMPTY(&tp->t_rxt_segments))                                  \
960 	        tcp_rxtseg_clean(tp);                                           \
961 } while(0)
962 
963 #define EXIT_FASTRECOVERY(_tp_) do {                                            \
964 	(_tp_)->t_flags &= ~TF_FASTRECOVERY;                                    \
965 	(_tp_)->t_dupacks = 0;                                                  \
966 	(_tp_)->t_rexmtthresh = (uint8_t)tcprexmtthresh;                        \
967 	(_tp_)->t_bytes_acked = 0;                                              \
968 	(_tp_)->ecn_flags &= ~TE_INRECOVERY;                                    \
969 	(_tp_)->t_timer[TCPT_PTO] = 0;                                          \
970 	(_tp_)->t_flagsext &= ~TF_RESCUE_RXT;                                   \
971 	(_tp_)->t_lossflightsize = 0;                                           \
972 	(_tp_)->sackhint.sack_bytes_acked = 0;                                  \
973 } while(0)
974 
975 /*
976  * When the number of duplicate acks received is less than
977  * the retransmit threshold, use Limited Transmit algorithm
978  */
979 extern uint8_t tcprexmtthresh;
980 #define ALLOW_LIMITED_TRANSMIT(_tp_)                                            \
981 	((_tp_)->t_dupacks > 0 &&                                               \
982 	(_tp_)->t_dupacks < (_tp_)->t_rexmtthresh &&                            \
983 	((_tp_)->t_flagsext & (TF_PKTS_REORDERED|TF_DELAY_RECOVERY))            \
984 	    != (TF_PKTS_REORDERED|TF_DELAY_RECOVERY))
985 
986 /*
987  * This condition is true if timestamp option is supported
988  * on a connection.
989  */
990 #define TSTMP_SUPPORTED(_tp_) \
991 	(((_tp_)->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP)) == \
992 	        (TF_REQ_TSTMP|TF_RCVD_TSTMP))
993 
994 /*
995  * This condition is true if window scale option is supported
996  * on a connection
997  */
998 #define TCP_WINDOW_SCALE_ENABLED(_tp_) \
999 	(((_tp_)->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == \
1000 	        (TF_RCVD_SCALE|TF_REQ_SCALE))
1001 
1002 /* Is ECN negotiated end-to-end */
1003 #define TCP_ECN_ENABLED(_tp_) \
1004 	(((_tp_)->ecn_flags & (TE_ECN_ON)) == (TE_ECN_ON))
1005 
1006 extern int tcp_l4s;
1007 extern int tcp_l4s_developer;
1008 
1009 /* L4S Developer setting will override system setting */
1010 typedef enum {
1011 	tcp_l4s_system = 0,
1012 	tcp_l4s_developer_enable = 1,
1013 	tcp_l4s_developer_disable = 2
1014 } tcp_l4s_t;
1015 
1016 /*
1017  * Gives number of bytes acked by this ack
1018  */
1019 #define BYTES_ACKED(_th_, _tp_) \
1020 	((_th_)->th_ack - (_tp_)->snd_una)
1021 
1022 /* Returns true if a DSACK option should be sent */
1023 #define TCP_SEND_DSACK_OPT(_tp_) \
1024 	((_tp_)->t_dsack_lseq > 0 && (_tp_)->t_dsack_rseq > 0)
1025 
1026 /*
1027  * Returns true if a DSACK sequence is within the max send window that will
1028  * be accepted. In order to set a window to validate sequence numbers, the
1029  * max send window within which a DSACK option is processed is limited.
1030  *
1031  * We need to choose a maximum window to check if the sequence number is
1032  * within the window. One arbitrary choice is 256 * MSS because if the
1033  * window is as large as 256 segments it might be big enough to ignore the
1034  * DSACK option. Choosing a much larger limit means that the memory for
1035  * retransmit segments can be held for a longer time.
1036  */
1037 #define TCP_DSACK_MAX_SEND_WINDOW(_tp_) (MIN((_tp_)->snd_wnd, tcp_autosndbuf_max))
1038 #define TCP_DSACK_SEQ_IN_WINDOW(_tp_, _seq_, _una_) \
1039     (SEQ_LEQ((_seq_), (_tp_)->snd_max) && \
1040     SEQ_GEQ((_seq_), ((_una_) - TCP_DSACK_MAX_SEND_WINDOW(_tp_))))
1041 
1042 #define TCP_RESET_REXMT_STATE(_tp_) do { \
1043 	(_tp_)->t_rxtshift = 0; \
1044 	(_tp_)->t_rxtstart = 0; \
1045 	mptcp_reset_rexmit_state((_tp_)); \
1046 } while(0);
1047 
1048 #define TCP_IF_STATE_CHANGED(tp, probe_if_index)                        \
1049 	(probe_if_index > 0 && tp->t_inpcb->inp_last_outifp != NULL &&  \
1050 	probe_if_index == tp->t_inpcb->inp_last_outifp->if_index)
1051 
1052 #define TCP_RLEDBAT_ENABLED(_tp) \
1053     (tcp_rledbat == 1 && TSTMP_SUPPORTED(_tp))
1054 
1055 #define TCP_RECV_BG(_so) \
1056     (tcp_recv_bg == 1 || IS_TCP_RECV_BG(_so))
1057 
1058 #define TCP_USE_RLEDBAT(_tp, _so) \
1059     TCP_RLEDBAT_ENABLED(_tp) && TCP_RECV_BG(_so)
1060 
1061 /*
1062  * Structure to hold TCP options that are only used during segment
1063  * processing (in tcp_input), but not held in the tcpcb.
1064  * It's basically used to reduce the number of parameters
1065  * to tcp_dooptions.
1066  */
1067 struct tcpopt {
1068 	uint32_t        to_flags;   /* which options are present */
1069 #define TOF_TS          0x0001  /* timestamp */
1070 #define TOF_SACKPERM    0x0004  /* SACK permitted (only in SYN/ACK) */
1071 #define TOF_MSS         0x0010
1072 #define TOF_SCALE       0x0020
1073 #define TOF_SIGNATURE   0x0040  /* signature option present */
1074 #define TOF_SIGLEN      0x0080  /* signature length valid (RFC2385) */
1075 #define TOF_SACK        0x0100  /* SACK option present */
1076 #define TOF_MPTCP       0x0200  /* MPTCP options to be dropped */
1077 #define TOF_TFO         0x0400  /* TFO cookie option present */
1078 #define TOF_TFOREQ      0x0800  /* TFO cookie request present */
1079 #define TOF_MAXOPT      0x1000
1080 	uint32_t        to_tsval;
1081 	uint32_t        to_tsecr;
1082 	uint16_t        to_mss;
1083 	uint8_t         to_wscale;
1084 	uint8_t         to_nsacks;                                                              /* number of SACK blocks */
1085 	u_char          *to_sacks __sized_by(to_sacks_size);        /* pointer to the first SACK blocks */
1086 	uint32_t                to_sacks_size;                                                          /* boundary for to_sacks */
1087 	u_char          *to_tfo  __sized_by(to_tfo_size);                       /* pointer to the TFO cookie */
1088 	uint32_t                to_tfo_size;                                                            /* boundary for to_tfo */
1089 	uint8_t         to_num_accecn;                                                          /* number of Accurate ECN counters */
1090 	uint8_t         *to_accecn __sized_by(to_accecn_size);      /* pointer to the first Accurate ECN counter */
1091 	uint32_t                to_accecn_size;                                                         /* boundary for to_accecn */
1092 	uint8_t         to_accecn_order;                                                        /* Accurate ECN ordering */
1093 };
1094 
1095 #define intotcpcb(ip)   ((struct tcpcb *)(ip)->inp_ppcb)
1096 #define sototcpcb(so)   (intotcpcb(sotoinpcb(so)))
1097 
1098 /* TFO-specific defines */
1099 #define TFO_COOKIE_LEN_MIN      4
1100 #define TFO_COOKIE_LEN_DEFAULT  8
1101 #define TFO_COOKIE_LEN_MAX      16
1102 
1103 /*
1104  * Jaguar compatible TCP control block, for xtcpcb
1105  * Does not have the old fields
1106  */
1107 struct otcpcb {
1108 #else
1109 struct tseg_qent;
1110 _TCPCB_LIST_HEAD(tsegqe_head, tseg_qent);
1111 
1112 struct tcpcb {
1113 #endif /* KERNEL_PRIVATE */
1114 #if defined(KERNEL_PRIVATE)
1115 	u_int32_t t_segq;
1116 #else
1117 	struct  tsegqe_head t_segq;
1118 #endif /* KERNEL_PRIVATE */
1119 	int     t_dupacks;              /* consecutive dup acks recd */
1120 	u_int32_t unused;               /* unused now: was t_template */
1121 
1122 	int     t_timer[TCPT_NTIMERS_EXT];      /* tcp timers */
1123 
1124 	_TCPCB_PTR(struct inpcb *) t_inpcb;     /* back pointer to internet pcb */
1125 	int     t_state;                /* state of this connection */
1126 	u_int   t_flags;
1127 #define TF_ACKNOW       0x00001         /* ack peer immediately */
1128 #define TF_DELACK       0x00002         /* ack, but try to delay it */
1129 #define TF_NODELAY      0x00004         /* don't delay packets to coalesce */
1130 #define TF_NOOPT        0x00008         /* don't use tcp options */
1131 #define TF_SENTFIN      0x00010         /* have sent FIN */
1132 #define TF_REQ_SCALE    0x00020         /* have/will request window scaling */
1133 #define TF_RCVD_SCALE   0x00040         /* other side has requested scaling */
1134 #define TF_REQ_TSTMP    0x00080         /* have/will request timestamps */
1135 #define TF_RCVD_TSTMP   0x00100         /* a timestamp was received in SYN */
1136 #define TF_SACK_PERMIT  0x00200         /* other side said I could SACK */
1137 #define TF_NEEDSYN      0x00400         /* send SYN (implicit state) - unused but needed for backwards compatibility */
1138 #define TF_NEEDFIN      0x00800         /* send FIN (implicit state) */
1139 #define TF_NOPUSH       0x01000         /* don't push */
1140 #define TF_REQ_CC       0x02000         /* have/will request CC */
1141 #define TF_RCVD_CC      0x04000         /* a CC was received in SYN */
1142 #define TF_SENDCCNEW    0x08000         /* Not implemented */
1143 #define TF_MORETOCOME   0x10000         /* More data to be appended to sock */
1144 #define TF_LQ_OVERFLOW  0x20000         /* listen queue overflow */
1145 #define TF_RXWIN0SENT   0x40000         /* sent a receiver win 0 in response */
1146 #define TF_SLOWLINK     0x80000         /* route is a on a modem speed link */
1147 
1148 	int     t_force;                /* 1 if forcing out a byte */
1149 
1150 	tcp_seq snd_una;                /* send unacknowledged */
1151 	tcp_seq snd_max;                /* highest sequence number sent;
1152 	                                 * used to recognize retransmits
1153 	                                 */
1154 	tcp_seq snd_nxt;                /* send next */
1155 	tcp_seq snd_up;                 /* send urgent pointer */
1156 
1157 	tcp_seq snd_wl1;                /* window update seg seq number */
1158 	tcp_seq snd_wl2;                /* window update seg ack number */
1159 	tcp_seq iss;                    /* initial send sequence number */
1160 	tcp_seq irs;                    /* initial receive sequence number */
1161 
1162 	tcp_seq rcv_nxt;                /* receive next */
1163 	tcp_seq rcv_adv;                /* advertised window */
1164 	u_int32_t rcv_wnd;              /* receive window */
1165 	tcp_seq rcv_up;                 /* receive urgent pointer */
1166 
1167 	u_int32_t snd_wnd;              /* send window */
1168 	u_int32_t snd_cwnd;             /* congestion-controlled window */
1169 	u_int32_t snd_ssthresh;         /* snd_cwnd size threshold for
1170 	                                 * for slow start exponential to
1171 	                                 * linear switch
1172 	                                 */
1173 	u_int   t_maxopd;               /* mss plus options */
1174 
1175 	u_int32_t t_rcvtime;            /* time at which a packet was received */
1176 	u_int32_t t_starttime;          /* time connection was established */
1177 	int     t_rtttime;              /* round trip time */
1178 	tcp_seq t_rtseq;                /* sequence number being timed */
1179 
1180 	int     t_rxtcur;               /* current retransmit value (ticks) */
1181 	u_int   t_maxseg;               /* maximum segment size */
1182 	int     t_srtt;                 /* smoothed round-trip time */
1183 	int     t_rttvar;               /* variance in round-trip time */
1184 
1185 	int     t_rxtshift;             /* log(2) of rexmt exp. backoff */
1186 	u_int   t_rttmin;               /* minimum rtt allowed */
1187 	u_int32_t t_rttupdated;         /* number of times rtt sampled */
1188 	u_int32_t max_sndwnd;           /* largest window peer has offered */
1189 
1190 	int     t_softerror;            /* possible error not yet reported */
1191 /* out-of-band data */
1192 	char    t_oobflags;             /* have some */
1193 	char    t_iobc;                 /* input character */
1194 #define TCPOOB_HAVEDATA 0x01
1195 #define TCPOOB_HADDATA  0x02
1196 /* RFC 1323 variables */
1197 	u_char  snd_scale;              /* window scaling for send window */
1198 	u_char  rcv_scale;              /* window scaling for recv window */
1199 	u_char  request_r_scale;        /* pending window scaling */
1200 	u_char  requested_s_scale;
1201 	u_int32_t ts_recent;            /* timestamp echo data */
1202 
1203 	u_int32_t ts_recent_age;        /* when last updated */
1204 	tcp_seq last_ack_sent;
1205 /* RFC 1644 variables */
1206 	tcp_cc  cc_send;                /* send connection count */
1207 	tcp_cc  cc_recv;                /* receive connection count */
1208 	tcp_seq snd_recover;            /* for use in fast recovery */
1209 /* experimental */
1210 	u_int32_t snd_cwnd_prev;        /* cwnd prior to retransmit */
1211 	u_int32_t snd_ssthresh_prev;    /* ssthresh prior to retransmit */
1212 	u_int32_t t_badrxtwin;          /* window for retransmit recovery */
1213 };
1214 
1215 #define        tcps_ecn_setup  tcps_ecn_client_success
1216 #define        tcps_sent_cwr   tcps_ecn_recv_ece
1217 #define        tcps_sent_ece   tcps_ecn_sent_ece
1218 
1219 /*
1220  * TCP statistics.
1221  * Many of these should be kept per connection,
1222  * but that's inconvenient at the moment.
1223  */
1224 struct  tcpstat {
1225 	u_int32_t       tcps_connattempt;       /* connections initiated */
1226 	u_int32_t       tcps_accepts;           /* connections accepted */
1227 	u_int32_t       tcps_connects;          /* connections established */
1228 	u_int32_t       tcps_drops;             /* connections dropped */
1229 	u_int32_t       tcps_conndrops;         /* embryonic connections dropped */
1230 	u_int32_t       tcps_closed;            /* conn. closed (includes drops) */
1231 	u_int32_t       tcps_segstimed;         /* segs where we tried to get rtt */
1232 	u_int32_t       tcps_rttupdated;        /* times we succeeded */
1233 	u_int32_t       tcps_delack;            /* delayed acks sent */
1234 	u_int32_t       tcps_timeoutdrop;       /* conn. dropped in rxmt timeout */
1235 	u_int32_t       tcps_rexmttimeo;        /* retransmit timeouts */
1236 	u_int32_t       tcps_persisttimeo;      /* persist timeouts */
1237 	u_int32_t       tcps_keeptimeo;         /* keepalive timeouts */
1238 	u_int32_t       tcps_keepprobe;         /* keepalive probes sent */
1239 	u_int32_t       tcps_keepdrops;         /* connections dropped in keepalive */
1240 
1241 	u_int32_t       tcps_sndtotal;          /* total packets sent */
1242 	u_int32_t       tcps_sndpack;           /* data packets sent */
1243 	u_int32_t       tcps_sndbyte;           /* data bytes sent */
1244 	u_int32_t       tcps_sndrexmitpack;     /* data packets retransmitted */
1245 	u_int32_t       tcps_sndrexmitbyte;     /* data bytes retransmitted */
1246 	u_int32_t       tcps_sndacks;           /* ack-only packets sent */
1247 	u_int32_t       tcps_sndprobe;          /* window probes sent */
1248 	u_int32_t       tcps_sndurg;            /* packets sent with URG only */
1249 	u_int32_t       tcps_sndwinup;          /* window update-only packets sent */
1250 	u_int32_t       tcps_sndctrl;           /* control (SYN|FIN|RST) packets sent */
1251 
1252 	u_int32_t       tcps_rcvtotal;          /* total packets received */
1253 	u_int32_t       tcps_rcvpack;           /* packets received in sequence */
1254 	u_int32_t       tcps_rcvbyte;           /* bytes received in sequence */
1255 	u_int32_t       tcps_rcvbadsum;         /* packets received with ccksum errs */
1256 	u_int32_t       tcps_rcvbadoff;         /* packets received with bad offset */
1257 	u_int32_t       tcps_rcvmemdrop;        /* packets dropped for lack of memory */
1258 	u_int32_t       tcps_rcvshort;          /* packets received too short */
1259 	u_int32_t       tcps_rcvduppack;        /* duplicate-only packets received */
1260 	u_int32_t       tcps_rcvdupbyte;        /* duplicate-only bytes received */
1261 	u_int32_t       tcps_rcvpartduppack;    /* packets with some duplicate data */
1262 	u_int32_t       tcps_rcvpartdupbyte;    /* dup. bytes in part-dup. packets */
1263 	u_int32_t       tcps_rcvoopack;         /* out-of-order packets received */
1264 	u_int32_t       tcps_rcvoobyte;         /* out-of-order bytes received */
1265 	u_int32_t       tcps_rcvpackafterwin;   /* packets with data after window */
1266 	u_int32_t       tcps_rcvbyteafterwin;   /* bytes rcvd after window */
1267 	u_int32_t       tcps_rcvafterclose;     /* packets rcvd after "close" */
1268 	u_int32_t       tcps_rcvwinprobe;       /* rcvd window probe packets */
1269 	u_int32_t       tcps_rcvdupack;         /* rcvd duplicate acks */
1270 	u_int32_t       tcps_rcvacktoomuch;     /* rcvd acks for unsent data */
1271 	u_int32_t       tcps_rcvackpack;        /* rcvd ack packets */
1272 	u_int32_t       tcps_rcvackbyte;        /* bytes acked by rcvd acks */
1273 	u_int32_t       tcps_rcvwinupd;         /* rcvd window update packets */
1274 	u_int32_t       tcps_pawsdrop;          /* segments dropped due to PAWS */
1275 	u_int32_t       tcps_predack;           /* times hdr predict ok for acks */
1276 	u_int32_t       tcps_preddat;           /* times hdr predict ok for data pkts */
1277 	u_int32_t       tcps_cachedrtt;         /* times cached RTT in route updated */
1278 	u_int32_t       tcps_cachedrttvar;      /* times cached rttvar updated */
1279 	u_int32_t       tcps_cachedssthresh;    /* times cached ssthresh updated */
1280 	u_int32_t       tcps_usedrtt;           /* times RTT initialized from route */
1281 	u_int32_t       tcps_usedrttvar;        /* times RTTVAR initialized from rt */
1282 	u_int32_t       tcps_usedssthresh;      /* times ssthresh initialized from rt*/
1283 	u_int32_t       tcps_persistdrop;       /* timeout in persist state */
1284 	u_int32_t       tcps_badsyn;            /* bogus SYN, e.g. premature ACK */
1285 	u_int32_t       tcps_mturesent;         /* resends due to MTU discovery */
1286 	u_int32_t       tcps_listendrop;        /* listen queue overflows */
1287 	u_int32_t       tcps_synchallenge;      /* challenge ACK due to bad SYN */
1288 	u_int32_t       tcps_rstchallenge;      /* challenge ACK due to bad RST */
1289 
1290 	/* new stats from FreeBSD 5.4 sync up */
1291 	u_int32_t       tcps_minmssdrops;       /* average minmss too low drops */
1292 
1293 	u_int32_t       tcps_sndrexmitbad;      /* unnecessary packet retransmissions */
1294 	u_int32_t       tcps_badrst;            /* ignored RSTs in the window */
1295 
1296 	u_int32_t       tcps_sc_dropped;        /* could not reply to packet */
1297 	u_int32_t       tcps_sc_completed;      /* successful extraction of entry */
1298 	u_int32_t       tcps_sc_aborted;        /* syncache entry aborted */
1299 	u_int32_t       tcps_sc_sendcookie;     /* SYN cookie sent */
1300 	u_int32_t       tcps_sc_recvcookie;     /* SYN cookie received */
1301 
1302 	/* SACK related stats */
1303 	u_int32_t       tcps_sack_recovery_episode; /* SACK recovery episodes */
1304 	u_int32_t       tcps_sack_rexmits;          /* SACK rexmit segments   */
1305 	u_int32_t       tcps_sack_rexmit_bytes;     /* SACK rexmit bytes      */
1306 	u_int32_t       tcps_sack_rcv_blocks;       /* SACK blocks (options) received */
1307 	u_int32_t       tcps_sack_send_blocks;      /* SACK blocks (options) sent     */
1308 	u_int32_t       tcps_sack_sboverflow;       /* SACK sendblock overflow   */
1309 
1310 	/* RACK related stats */
1311 	u_int32_t       tcps_rack_recovery_episode; /* RACK recovery episodes */
1312 	u_int32_t       tcps_rack_reordering_timeout_recovery_episode; /* RACK recovery episodes due to reordering timeout */
1313 	u_int32_t       tcps_rack_rexmits;          /* RACK rexmit segments   */
1314 
1315 	u_int32_t       tcps_bg_rcvtotal;       /* total background packets received */
1316 	u_int32_t       tcps_rxtfindrop;        /* drop conn after retransmitting FIN */
1317 	u_int32_t       tcps_fcholdpacket;      /* packets withheld because of flow control */
1318 
1319 	u_int32_t       tcps_limited_txt;       /* Limited transmit used */
1320 	u_int32_t       tcps_early_rexmt;       /* Early retransmit used */
1321 	u_int32_t       tcps_sack_ackadv;       /* Cumulative ack advanced along with sack */
1322 
1323 	/* Checksum related stats */
1324 	u_int32_t       tcps_rcv_swcsum;        /* tcp swcksum (inbound), packets */
1325 	u_int32_t       tcps_rcv_swcsum_bytes;  /* tcp swcksum (inbound), bytes */
1326 	u_int32_t       tcps_rcv6_swcsum;       /* tcp6 swcksum (inbound), packets */
1327 	u_int32_t       tcps_rcv6_swcsum_bytes; /* tcp6 swcksum (inbound), bytes */
1328 	u_int32_t       tcps_snd_swcsum;        /* tcp swcksum (outbound), packets */
1329 	u_int32_t       tcps_snd_swcsum_bytes;  /* tcp swcksum (outbound), bytes */
1330 	u_int32_t       tcps_snd6_swcsum;       /* tcp6 swcksum (outbound), packets */
1331 	u_int32_t       tcps_snd6_swcsum_bytes; /* tcp6 swcksum (outbound), bytes */
1332 
1333 	/* MPTCP Related stats */
1334 	u_int32_t       tcps_invalid_mpcap;     /* Invalid MPTCP capable opts */
1335 	u_int32_t       tcps_invalid_joins;     /* Invalid MPTCP joins */
1336 	u_int32_t       tcps_mpcap_fallback;    /* TCP fallback in primary */
1337 	u_int32_t       tcps_join_fallback;     /* No MPTCP in secondary */
1338 	u_int32_t       tcps_estab_fallback;    /* DSS option dropped */
1339 	u_int32_t       tcps_invalid_opt;       /* Catchall error stat */
1340 	u_int32_t       tcps_mp_reducedwin;     /* Reduced subflow window */
1341 	u_int32_t       tcps_mp_badcsum;        /* Bad DSS csum */
1342 	u_int32_t       tcps_mp_oodata;         /* Out of order data */
1343 	u_int32_t       tcps_mp_switches;       /* number of subflow switch */
1344 	u_int32_t       tcps_mp_rcvtotal;       /* number of rcvd packets */
1345 	u_int32_t       tcps_mp_rcvbytes;       /* number of bytes received */
1346 	u_int32_t       tcps_mp_sndpacks;       /* number of data packs sent */
1347 	u_int32_t       tcps_mp_sndbytes;       /* number of bytes sent */
1348 	u_int32_t       tcps_join_rxmts;        /* join ack retransmits */
1349 	u_int32_t       tcps_tailloss_rto;      /* RTO due to tail loss */
1350 	u_int32_t       tcps_reordered_pkts;    /* packets reorderd */
1351 	u_int32_t       tcps_recovered_pkts;    /* recovered after loss */
1352 	u_int32_t       tcps_pto;               /* probe timeout */
1353 	u_int32_t       tcps_rto_after_pto;     /* RTO after a probe */
1354 	u_int32_t       tcps_tlp_recovery;      /* TLP induced fast recovery */
1355 	u_int32_t       tcps_tlp_recoverlastpkt; /* TLP recoverd last pkt */
1356 	u_int32_t       tcps_ecn_client_success; /* client-side connection negotiated ECN */
1357 	u_int32_t       tcps_ecn_recv_ece;      /* ECE received, sent CWR */
1358 	u_int32_t       tcps_ecn_sent_ece;      /* Sent ECE notification */
1359 	u_int32_t       tcps_detect_reordering; /* Detect pkt reordering */
1360 	u_int32_t       tcps_delay_recovery;    /* Delay fast recovery */
1361 	u_int32_t       tcps_avoid_rxmt;        /* Retransmission was avoided */
1362 	u_int32_t       tcps_pto_in_recovery;   /* rescue retransmit in fast recovery */
1363 	u_int32_t       tcps_pmtudbh_reverted;  /* PMTU Blackhole detection, segment size reverted */
1364 
1365 	/* DSACK related statistics */
1366 	u_int32_t       tcps_dsack_ackloss;     /* ignore DSACK due to ack loss */
1367 	u_int32_t       tcps_dsack_badrexmt;    /* DSACK based bad rexmt recovery */
1368 	u_int32_t       tcps_dsack_sent;        /* Sent DSACK notification */
1369 	u_int32_t       tcps_dsack_recvd;       /* Received a valid DSACK option */
1370 	u_int32_t       tcps_dsack_recvd_old;   /* Received an out of window DSACK option */
1371 
1372 	/* MPTCP Subflow selection stats */
1373 	u_int32_t       tcps_mp_sel_rtt;        /* By RTT comparison */
1374 	u_int32_t       tcps_mp_sel_rto;        /* By RTO comparison */
1375 	u_int32_t       tcps_mp_num_probes;     /* Number of probes sent */
1376 	u_int32_t       tcps_mp_verdowngrade;   /* MPTCP version downgrade */
1377 	u_int32_t       tcps_drop_after_sleep;  /* drop after long AP sleep */
1378 	u_int32_t       tcps_probe_if;          /* probe packets after interface availability */
1379 	u_int32_t       tcps_probe_if_conflict; /* Can't send probe packets for interface */
1380 
1381 	u_int32_t       tcps_ecn_client_setup;    /* Attempted ECN setup from client side */
1382 	u_int32_t       tcps_ecn_server_setup;    /* Attempted ECN setup from server side */
1383 	u_int32_t       tcps_ecn_server_success;  /* server-side connection negotiated ECN */
1384 	u_int32_t       tcps_ecn_ace_syn_not_ect; /* received AccECN SYN packet with Not-ECT */
1385 	u_int32_t       tcps_ecn_ace_syn_ect1;    /* received AccECN SYN packet with ECT1 */
1386 	u_int32_t       tcps_ecn_ace_syn_ect0;    /* received AccECN SYN packet with ECT0 */
1387 	u_int32_t       tcps_ecn_ace_syn_ce;      /* received AccECN SYN packet with CE */
1388 	u_int32_t       tcps_ecn_lost_synack;   /* Lost SYN-ACK with ECN setup */
1389 	u_int32_t       tcps_ecn_lost_syn;      /* Lost SYN with ECN setup */
1390 	u_int32_t       tcps_ecn_not_supported; /* Server did not support ECN setup */
1391 	u_int32_t       tcps_ecn_recv_ce;       /* Received CE from the network */
1392 	u_int32_t       tcps_ecn_ace_recv_ce;   /* CE count received in ACE field */
1393 	u_int32_t       tcps_ecn_conn_recv_ce;  /* Number of connections received CE atleast once */
1394 	u_int32_t       tcps_ecn_conn_recv_ece; /* Number of connections received ECE atleast once */
1395 	u_int32_t       tcps_ecn_conn_plnoce;   /* Number of connections that received no CE and sufferred packet loss */
1396 	u_int32_t       tcps_ecn_conn_pl_ce;    /* Number of connections that received CE and sufferred packet loss */
1397 	u_int32_t       tcps_ecn_conn_nopl_ce;  /* Number of connections that received CE and sufferred no packet loss */
1398 	u_int32_t       tcps_ecn_fallback_synloss; /* Number of times we did fall back due to SYN-Loss */
1399 	u_int32_t       tcps_ecn_fallback_reorder; /* Number of times we fallback because we detected the PAWS-issue */
1400 	u_int32_t       tcps_ecn_fallback_ce;   /* Number of times we fallback because we received too many CEs */
1401 
1402 	/* TFO-related statistics */
1403 	u_int32_t       tcps_tfo_syn_data_rcv;  /* Received a SYN+data with valid cookie */
1404 	u_int32_t       tcps_tfo_cookie_req_rcv;/* Received a TFO cookie-request */
1405 	u_int32_t       tcps_tfo_cookie_sent;   /* Offered a TFO-cookie to the client */
1406 	u_int32_t       tcps_tfo_cookie_invalid;/* Received an invalid TFO-cookie */
1407 	u_int32_t       tcps_tfo_cookie_req;    /* Cookie requested with the SYN */
1408 	u_int32_t       tcps_tfo_cookie_rcv;    /* Cookie received in a SYN/ACK */
1409 	u_int32_t       tcps_tfo_syn_data_sent; /* SYN+data+cookie sent */
1410 	u_int32_t       tcps_tfo_syn_data_acked;/* SYN+data has been acknowledged */
1411 	u_int32_t       tcps_tfo_syn_loss;      /* SYN+TFO has been lost and we fallback */
1412 	u_int32_t       tcps_tfo_blackhole;     /* TFO got blackholed by a middlebox. */
1413 	u_int32_t       tcps_tfo_cookie_wrong;  /* TFO-cookie we sent was wrong */
1414 	u_int32_t       tcps_tfo_no_cookie_rcv; /* We asked for a cookie but didn't get one */
1415 	u_int32_t       tcps_tfo_heuristics_disable; /* TFO got disabled due to heuristics */
1416 	u_int32_t       tcps_tfo_sndblackhole;  /* TFO got blackholed in the sending direction */
1417 	u_int32_t       tcps_mss_to_default;    /* Change MSS to default using link status report */
1418 	u_int32_t       tcps_mss_to_medium;     /* Change MSS to medium using link status report */
1419 	u_int32_t       tcps_mss_to_low;        /* Change MSS to low using link status report */
1420 	u_int32_t       tcps_ecn_fallback_droprst; /* ECN fallback caused by connection drop due to RST */
1421 	u_int32_t       tcps_ecn_fallback_droprxmt; /* ECN fallback due to drop after multiple retransmits */
1422 	u_int32_t       tcps_ecn_fallback_synrst; /* ECN fallback due to rst after syn */
1423 
1424 	u_int32_t       tcps_mptcp_rcvmemdrop;  /* MPTCP packets dropped for lack of memory */
1425 	u_int32_t       tcps_mptcp_rcvduppack;  /* MPTCP duplicate-only packets received */
1426 	u_int32_t       tcps_mptcp_rcvpackafterwin; /* MPTCP packets with data after window */
1427 
1428 	/* TCP timer statistics */
1429 	u_int32_t       tcps_timer_drift_le_1_ms;       /* Timer drift less or equal to 1 ms */
1430 	u_int32_t       tcps_timer_drift_le_10_ms;      /* Timer drift less or equal to 10 ms */
1431 	u_int32_t       tcps_timer_drift_le_20_ms;      /* Timer drift less or equal to 20 ms */
1432 	u_int32_t       tcps_timer_drift_le_50_ms;      /* Timer drift less or equal to 50 ms */
1433 	u_int32_t       tcps_timer_drift_le_100_ms;     /* Timer drift less or equal to 100 ms */
1434 	u_int32_t       tcps_timer_drift_le_200_ms;     /* Timer drift less or equal to 200 ms */
1435 	u_int32_t       tcps_timer_drift_le_500_ms;     /* Timer drift less or equal to 500 ms */
1436 	u_int32_t       tcps_timer_drift_le_1000_ms;    /* Timer drift less or equal to 1000 ms */
1437 	u_int32_t       tcps_timer_drift_gt_1000_ms;    /* Timer drift greater than 1000 ms */
1438 
1439 	u_int32_t       tcps_mptcp_handover_attempt;    /* Total number of MPTCP-attempts using handover mode */
1440 	u_int32_t       tcps_mptcp_interactive_attempt; /* Total number of MPTCP-attempts using interactive mode */
1441 	u_int32_t       tcps_mptcp_aggregate_attempt;   /* Total number of MPTCP-attempts using aggregate mode */
1442 	u_int32_t       tcps_mptcp_fp_handover_attempt; /* Same as previous three but only for first-party apps */
1443 	u_int32_t       tcps_mptcp_fp_interactive_attempt;
1444 	u_int32_t       tcps_mptcp_fp_aggregate_attempt;
1445 	u_int32_t       tcps_mptcp_heuristic_fallback;  /* Total number of MPTCP-connections that fell back due to heuristics */
1446 	u_int32_t       tcps_mptcp_fp_heuristic_fallback;       /* Same as previous but for first-party apps */
1447 	u_int32_t       tcps_mptcp_handover_success_wifi;       /* Total number of successfull handover-mode connections that *started* on WiFi */
1448 	u_int32_t       tcps_mptcp_handover_success_cell;       /* Total number of successfull handover-mode connections that *started* on Cell */
1449 	u_int32_t       tcps_mptcp_interactive_success;         /* Total number of interactive-mode connections that negotiated MPTCP */
1450 	u_int32_t       tcps_mptcp_aggregate_success;           /* Same as previous but for aggregate */
1451 	u_int32_t       tcps_mptcp_fp_handover_success_wifi;    /* Same as previous four, but for first-party apps */
1452 	u_int32_t       tcps_mptcp_fp_handover_success_cell;
1453 	u_int32_t       tcps_mptcp_fp_interactive_success;
1454 	u_int32_t       tcps_mptcp_fp_aggregate_success;
1455 	u_int32_t       tcps_mptcp_handover_cell_from_wifi;     /* Total number of connections that use cell in handover-mode (coming from WiFi) */
1456 	u_int32_t       tcps_mptcp_handover_wifi_from_cell;     /* Total number of connections that use WiFi in handover-mode (coming from cell) */
1457 	u_int32_t       tcps_mptcp_interactive_cell_from_wifi;  /* Total number of connections that use cell in interactive mode (coming from WiFi) */
1458 	u_int64_t       tcps_mptcp_handover_cell_bytes;         /* Total number of bytes sent on cell in handover-mode (on new subflows, ignoring initial one) */
1459 	u_int64_t       tcps_mptcp_interactive_cell_bytes;      /* Same as previous but for interactive */
1460 	u_int64_t       tcps_mptcp_aggregate_cell_bytes;
1461 	u_int64_t       tcps_mptcp_handover_all_bytes;          /* Total number of bytes sent in handover */
1462 	u_int64_t       tcps_mptcp_interactive_all_bytes;
1463 	u_int64_t       tcps_mptcp_aggregate_all_bytes;
1464 	u_int32_t       tcps_mptcp_back_to_wifi;        /* Total number of connections that succeed to move traffic away from cell (when starting on cell) */
1465 	u_int32_t       tcps_mptcp_wifi_proxy;          /* Total number of new subflows that fell back to regular TCP on cell */
1466 	u_int32_t       tcps_mptcp_cell_proxy;          /* Total number of new subflows that fell back to regular TCP on WiFi */
1467 
1468 	/* TCP offload statistics */
1469 	u_int32_t       tcps_ka_offload_drops;  /* Keep alive drops for timeout reported by firmware */
1470 
1471 	u_int32_t       tcps_mptcp_triggered_cell;      /* Total number of times an MPTCP-connection triggered cell bringup */
1472 
1473 	u_int32_t       tcps_fin_timeout_drops;
1474 
1475 	/* RST compression statistics */
1476 	u_int64_t       tcps_rst_dup_suppressed;
1477 	u_int64_t       tcps_rst_not_suppressed;
1478 };
1479 
1480 #ifdef PRIVATE
1481 #define HAS_TCPSTAT_RST_SUPPRESSION 1
1482 #endif /* PRIVATE */
1483 
1484 struct tcpstat_local {
1485 	u_int64_t badformat;
1486 	u_int64_t unspecv6;
1487 	u_int64_t synfin;
1488 	u_int64_t badformatipsec;
1489 	u_int64_t noconnnolist;
1490 	u_int64_t noconnlist;
1491 	u_int64_t listbadsyn;
1492 	u_int64_t icmp6unreach;
1493 	u_int64_t deprecate6;
1494 	u_int64_t ooopacket;
1495 	u_int64_t rstinsynrcv;
1496 	u_int64_t dospacket;
1497 	u_int64_t cleanup;
1498 	u_int64_t synwindow;
1499 	u_int64_t linkheur_stealthdrop;
1500 	u_int64_t linkheur_noackpri;
1501 	u_int64_t linkheur_comprxmt;
1502 	u_int64_t linkheur_synrxmt;
1503 	u_int64_t linkheur_rxmtfloor;
1504 };
1505 
1506 #pragma pack(4)
1507 
1508 /*
1509  * TCB structure exported to user-land via sysctl(3).
1510  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
1511  * included.  Not all of our clients do.
1512  */
1513 
1514 struct  xtcpcb {
1515 	u_int32_t       xt_len;
1516 #ifdef KERNEL_PRIVATE
1517 	struct  inpcb_compat    xt_inp;
1518 #else
1519 	struct  inpcb   xt_inp;
1520 #endif
1521 #ifdef KERNEL_PRIVATE
1522 	struct  otcpcb  xt_tp;
1523 #else
1524 	struct  tcpcb   xt_tp;
1525 #endif
1526 	struct  xsocket xt_socket;
1527 	u_quad_t        xt_alignment_hack;
1528 };
1529 
1530 #if XNU_TARGET_OS_OSX || KERNEL || !(TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
1531 
1532 struct  xtcpcb64 {
1533 	u_int32_t               xt_len;
1534 	struct xinpcb64         xt_inpcb;
1535 
1536 	u_int64_t t_segq;
1537 	int     t_dupacks;              /* consecutive dup acks recd */
1538 
1539 	int t_timer[TCPT_NTIMERS_EXT];  /* tcp timers */
1540 
1541 	int     t_state;                /* state of this connection */
1542 	u_int   t_flags;
1543 
1544 	int     t_force;                /* 1 if forcing out a byte */
1545 
1546 	tcp_seq snd_una;                /* send unacknowledged */
1547 	tcp_seq snd_max;                /* highest sequence number sent;
1548 	                                 * used to recognize retransmits
1549 	                                 */
1550 	tcp_seq snd_nxt;                /* send next */
1551 	tcp_seq snd_up;                 /* send urgent pointer */
1552 
1553 	tcp_seq snd_wl1;                /* window update seg seq number */
1554 	tcp_seq snd_wl2;                /* window update seg ack number */
1555 	tcp_seq iss;                    /* initial send sequence number */
1556 	tcp_seq irs;                    /* initial receive sequence number */
1557 
1558 	tcp_seq rcv_nxt;                /* receive next */
1559 	tcp_seq rcv_adv;                /* advertised window */
1560 	u_int32_t rcv_wnd;              /* receive window */
1561 	tcp_seq rcv_up;                 /* receive urgent pointer */
1562 
1563 	u_int32_t snd_wnd;              /* send window */
1564 	u_int32_t snd_cwnd;             /* congestion-controlled window */
1565 	u_int32_t snd_ssthresh;         /* snd_cwnd size threshold for
1566 	                                 * for slow start exponential to
1567 	                                 * linear switch
1568 	                                 */
1569 	u_int   t_maxopd;               /* mss plus options */
1570 
1571 	u_int32_t t_rcvtime;            /* time at which a packet was received */
1572 	u_int32_t t_starttime;          /* time connection was established */
1573 	int     t_rtttime;              /* round trip time */
1574 	tcp_seq t_rtseq;                /* sequence number being timed */
1575 
1576 	int     t_rxtcur;               /* current retransmit value (ticks) */
1577 	u_int   t_maxseg;               /* maximum segment size */
1578 	int     t_srtt;                 /* smoothed round-trip time */
1579 	int     t_rttvar;               /* variance in round-trip time */
1580 
1581 	int     t_rxtshift;             /* log(2) of rexmt exp. backoff */
1582 	u_int   t_rttmin;               /* minimum rtt allowed */
1583 	u_int32_t t_rttupdated;         /* number of times rtt sampled */
1584 	u_int32_t max_sndwnd;           /* largest window peer has offered */
1585 
1586 	int     t_softerror;            /* possible error not yet reported */
1587 /* out-of-band data */
1588 	char    t_oobflags;             /* have some */
1589 	char    t_iobc;                 /* input character */
1590 /* RFC 1323 variables */
1591 	u_char  snd_scale;              /* window scaling for send window */
1592 	u_char  rcv_scale;              /* window scaling for recv window */
1593 	u_char  request_r_scale;        /* pending window scaling */
1594 	u_char  requested_s_scale;
1595 	u_int32_t ts_recent;            /* timestamp echo data */
1596 
1597 	u_int32_t ts_recent_age;        /* when last updated */
1598 	tcp_seq last_ack_sent;
1599 /* RFC 1644 variables */
1600 	tcp_cc  cc_send;                /* send connection count */
1601 	tcp_cc  cc_recv;                /* receive connection count */
1602 	tcp_seq snd_recover;            /* for use in fast recovery */
1603 /* experimental */
1604 	u_int32_t snd_cwnd_prev;        /* cwnd prior to retransmit */
1605 	u_int32_t snd_ssthresh_prev;    /* ssthresh prior to retransmit */
1606 	u_int32_t t_badrxtwin;          /* window for retransmit recovery */
1607 
1608 	u_quad_t                xt_alignment_hack;
1609 };
1610 
1611 #endif /* XNU_TARGET_OS_OSX || KERNEL || !(TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR) */
1612 
1613 #ifdef PRIVATE
1614 
1615 struct  xtcpcb_n {
1616 	u_int32_t               xt_len;
1617 	u_int32_t                       xt_kind;                /* XSO_TCPCB */
1618 
1619 	u_int64_t t_segq;
1620 	int     t_dupacks;              /* consecutive dup acks recd */
1621 
1622 	int t_timer[TCPT_NTIMERS_EXT];  /* tcp timers */
1623 
1624 	int     t_state;                /* state of this connection */
1625 	u_int   t_flags;
1626 
1627 	int     t_force;                /* 1 if forcing out a byte */
1628 
1629 	tcp_seq snd_una;                /* send unacknowledged */
1630 	tcp_seq snd_max;                /* highest sequence number sent;
1631 	                                 * used to recognize retransmits
1632 	                                 */
1633 	tcp_seq snd_nxt;                /* send next */
1634 	tcp_seq snd_up;                 /* send urgent pointer */
1635 
1636 	tcp_seq snd_wl1;                /* window update seg seq number */
1637 	tcp_seq snd_wl2;                /* window update seg ack number */
1638 	tcp_seq iss;                    /* initial send sequence number */
1639 	tcp_seq irs;                    /* initial receive sequence number */
1640 
1641 	tcp_seq rcv_nxt;                /* receive next */
1642 	tcp_seq rcv_adv;                /* advertised window */
1643 	u_int32_t rcv_wnd;              /* receive window */
1644 	tcp_seq rcv_up;                 /* receive urgent pointer */
1645 
1646 	u_int32_t snd_wnd;              /* send window */
1647 	u_int32_t snd_cwnd;             /* congestion-controlled window */
1648 	u_int32_t snd_ssthresh;         /* snd_cwnd size threshold for
1649 	                                 * for slow start exponential to
1650 	                                 * linear switch
1651 	                                 */
1652 	u_int   t_maxopd;               /* mss plus options */
1653 
1654 	u_int32_t t_rcvtime;            /* time at which a packet was received */
1655 	u_int32_t t_starttime;          /* time connection was established */
1656 	int     t_rtttime;              /* round trip time */
1657 	tcp_seq t_rtseq;                /* sequence number being timed */
1658 
1659 	int     t_rxtcur;               /* current retransmit value (ticks) */
1660 	u_int   t_maxseg;               /* maximum segment size */
1661 	int     t_srtt;                 /* smoothed round-trip time */
1662 	int     t_rttvar;               /* variance in round-trip time */
1663 
1664 	int     t_rxtshift;             /* log(2) of rexmt exp. backoff */
1665 	u_int   t_rttmin;               /* minimum rtt allowed */
1666 	u_int32_t t_rttupdated;         /* number of times rtt sampled */
1667 	u_int32_t max_sndwnd;           /* largest window peer has offered */
1668 
1669 	int     t_softerror;            /* possible error not yet reported */
1670 	/* out-of-band data */
1671 	char    t_oobflags;             /* have some */
1672 	char    t_iobc;                 /* input character */
1673 	/* RFC 1323 variables */
1674 	u_char  snd_scale;              /* window scaling for send window */
1675 	u_char  rcv_scale;              /* window scaling for recv window */
1676 	u_char  request_r_scale;        /* pending window scaling */
1677 	u_char  requested_s_scale;
1678 	u_int32_t ts_recent;            /* timestamp echo data */
1679 
1680 	u_int32_t ts_recent_age;        /* when last updated */
1681 	tcp_seq last_ack_sent;
1682 	/* RFC 1644 variables */
1683 	tcp_cc  cc_send;                /* send connection count */
1684 	tcp_cc  cc_recv;                /* receive connection count */
1685 	tcp_seq snd_recover;            /* for use in fast recovery */
1686 	/* experimental */
1687 	u_int32_t snd_cwnd_prev;        /* cwnd prior to retransmit */
1688 	u_int32_t snd_ssthresh_prev;    /* ssthresh prior to retransmit */
1689 };
1690 
1691 /*
1692  * The rtt measured is in milliseconds as the timestamp granularity is
1693  * a millisecond. The smoothed round-trip time and estimated variance
1694  * are stored as fixed point numbers scaled by the values below.
1695  * For convenience, these scales are also used in smoothing the average
1696  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
1697  * With these scales, srtt has 5 bits to the right of the binary point,
1698  * and thus an "ALPHA" of 0.875.  rttvar has 4 bits to the right of the
1699  * binary point, and is smoothed with an ALPHA of 0.75.
1700  */
1701 #define TCP_RTT_SCALE           32      /* multiplier for srtt; 3 bits frac. */
1702 #define TCP_RTT_SHIFT           5       /* shift for srtt; 5 bits frac. */
1703 #define TCP_RTTVAR_SCALE        16      /* multiplier for rttvar; 4 bits */
1704 #define TCP_RTTVAR_SHIFT        4       /* shift for rttvar; 4 bits */
1705 #define TCP_DELTA_SHIFT         2       /* see tcp_input.c */
1706 
1707 struct tcpprobereq {
1708 	u_int64_t       ifindex;                /* Optional interface index for TCP keep-alive probing */
1709 	u_int64_t       enable;                 /* Flag to enable or disable probing (1=on, 0=off)*/
1710 	u_int64_t       filter_flags;           /* Optional flags for filtering interfaces per ntstat.h (NSTAT_IFNET_IS_*) */
1711 	u_int32_t       reserved;               /* Expansion */
1712 	u_int32_t       reserved2;              /* Expansion */
1713 };
1714 
1715 #endif /* PRIVATE */
1716 
1717 #pragma pack()
1718 
1719 /*
1720  * Names for TCP sysctl objects
1721  */
1722 #define TCPCTL_DO_RFC1323       1       /* use RFC-1323 extensions */
1723 #define TCPCTL_DO_RFC1644       2       /* use RFC-1644 extensions */
1724 #define TCPCTL_MSSDFLT          3       /* MSS default */
1725 #define TCPCTL_STATS            4       /* statistics (read-only) */
1726 #define TCPCTL_RTTDFLT          5       /* default RTT estimate */
1727 #define TCPCTL_KEEPIDLE         6       /* keepalive idle timer */
1728 #define TCPCTL_KEEPINTVL        7       /* interval to send keepalives */
1729 #define TCPCTL_SENDSPACE        8       /* send buffer space */
1730 #define TCPCTL_RECVSPACE        9       /* receive buffer space */
1731 #define TCPCTL_KEEPINIT         10      /* timeout for establishing syn */
1732 #define TCPCTL_PCBLIST          11      /* list of all outstanding PCBs */
1733 #define TCPCTL_DELACKTIME       12      /* time before sending delayed ACK */
1734 #define TCPCTL_V6MSSDFLT        13      /* MSS default for IPv6 */
1735 #define TCPCTL_MAXID            14
1736 
1737 #ifdef BSD_KERNEL_PRIVATE
1738 #include <sys/bitstring.h>
1739 
1740 #define TCP_PKTLIST_CLEAR(tp) {                                         \
1741 	(tp)->t_pktlist_head = (tp)->t_pktlist_tail = NULL;             \
1742 	(tp)->t_lastchain = (tp)->t_pktlist_sentlen = 0;                \
1743 }
1744 
1745 extern int tcp_TCPTV_MIN;
1746 
1747 #ifdef SYSCTL_DECL
1748 SYSCTL_DECL(_net_inet_tcp);
1749 #endif /* SYSCTL_DECL */
1750 
1751 extern  struct inpcbhead tcb;           /* head of queue of active tcpcb's */
1752 extern  struct inpcbinfo tcbinfo;
1753 extern  struct tcpstat tcpstat; /* tcp statistics */
1754 extern  int tcp_mssdflt;        /* XXX */
1755 extern  int tcp_minmss;
1756 #define TCP_FASTOPEN_SERVER 0x01
1757 #define TCP_FASTOPEN_CLIENT 0x02
1758 
1759 extern int tcp_tfo_halfcnt;
1760 extern int tcp_tfo_backlog;
1761 extern int tcp_fastopen;
1762 extern int ss_fltsz_local;
1763 extern int target_qdelay;
1764 extern uint32_t tcp_now;               /* for RFC 1323 timestamps */
1765 extern int tcp_delack_enabled;
1766 extern int maxseg_unacked;
1767 extern int tcp_ecn;
1768 extern uint32_t tcp_do_autorcvbuf;
1769 extern uint32_t tcp_autorcvbuf_max;
1770 extern int tcp_recv_bg;
1771 extern int tcp_do_ack_compression;
1772 extern int tcp_randomize_timestamps;
1773 extern int tcp_rledbat;
1774 extern int tcp_use_min_curr_rtt;
1775 extern int tcp_do_timestamps;
1776 
1777 /*
1778  * Feature flags for LQM heuristics
1779  * Can be useful for testing
1780  */
1781 #define TCP_LINK_HEUR_RXMT_COMP          0x0001
1782 #define TCP_LINK_HEUR_NOACKPRI           0x0002
1783 #define TCP_LINK_HEUR_SYNRMXT            0x0004
1784 #define TCP_LINK_HEUR_STEALTH            0x0008
1785 #define TCP_LINK_HEUR_RTOMIN             0x0010
1786 #define TCP_LINK_HEUR_NOTLP              0x0020
1787 
1788 #define TCP_LINK_HEURISTICS_DEFAULT (\
1789     TCP_LINK_HEUR_RXMT_COMP | \
1790     TCP_LINK_HEUR_NOACKPRI | \
1791     TCP_LINK_HEUR_SYNRMXT | \
1792     TCP_LINK_HEUR_STEALTH | \
1793     TCP_LINK_HEUR_RTOMIN | \
1794     TCP_LINK_HEUR_NOTLP)
1795 
1796 extern int32_t tcp_link_heuristics_flags;
1797 extern int32_t tcp_link_heuristics_rto_min;
1798 
1799 #define TCP_COMP_RXMT_GENCNT_MASK       0x80000000
1800 
1801 /* 3 seconds is conservative value (see RFC 2988 and RFC 6298) */
1802 #define TCP_DEFAULT_LINK_HEUR_RTOMIN 3000
1803 
1804 /*
1805  * Dummy value used for when there is no flow and we want to ensure that compression
1806  * can happen.
1807  */
1808 #define TCP_ACK_COMPRESSION_DUMMY 1
1809 
1810 extern struct tseg_qent *tcp_create_reass_qent(struct tcpcb *tp, struct mbuf *m,
1811     struct tcphdr *th, int len);
1812 extern struct mbuf *tcp_destroy_reass_qent(struct tcpcb *tp,
1813     struct tseg_qent *q);
1814 KALLOC_TYPE_DECLARE(tcp_reass_zone);
1815 extern struct tseg_qent *tcp_reass_qent_alloc(struct protosw *proto);
1816 extern void tcp_reass_qent_free(struct protosw *proto, struct tseg_qent *te);
1817 
1818 extern int tcp_rack;
1819 
1820 extern int tcp_reass_total_qlen;
1821 
1822 struct protosw;
1823 struct domain;
1824 
1825 struct tcp_respond_args {
1826 	unsigned int ifscope;
1827 	unsigned int nocell:1,
1828 	    noexpensive:1,
1829 	    awdl_unrestricted:1,
1830 	    intcoproc_allowed:1,
1831 	    keep_alive:1,
1832 	    noconstrained:1,
1833 	    management_allowed:1,
1834 	    ultra_constrained_allowed:1;
1835 };
1836 
1837 void     tcp_canceltimers(struct tcpcb *);
1838 uint8_t tcp_addoptions(struct tcpopt *to, u_char * __ended_by(optend) optp, u_char * optend);
1839 struct tcpcb *
1840 tcp_close(struct tcpcb *);
1841 void     tcp_ctlinput(int, struct sockaddr *, void *, struct ifnet *);
1842 int      tcp_ctloutput(struct socket *, struct sockopt *);
1843 struct tcpcb *
1844 tcp_drop(struct tcpcb *, int);
1845 void     tcp_drain(void);
1846 void     tcp_getrt_rtt(struct tcpcb *tp, struct rtentry *rt);
1847 void     tcp_init(struct protosw *, struct domain *);
1848 bool tcp_create_server_socket(struct tcp_inp *tpi, struct socket **so2, bool *syn_cookie_sent, int *dropsocket);
1849 bool tcp_setup_server_socket(struct tcp_inp *tpi, struct socket *so, bool syn_cookie_used);
1850 void     tcp_input(struct mbuf *, int);
1851 void     tcp_mss(struct tcpcb *, int, unsigned int);
1852 uint32_t tcp_ceil(double a);
1853 uint32_t tcp_round_to(uint32_t val, uint32_t round);
1854 uint32_t tcp_round_up(uint32_t val, uint32_t base);
1855 uint32_t ntoh24(u_char * p __sized_by(3));
1856 uint32_t tcp_packets_this_ack(struct tcpcb *tp, uint32_t acked);
1857 int      tcp_mssopt(struct tcpcb *);
1858 void     tcp_drop_syn_sent(struct inpcb *, int);
1859 uint32_t tcp_get_effective_mtu(struct rtentry *, uint32_t);
1860 void     tcp_mtudisc(struct inpcb *, int);
1861 struct tcpcb *
1862 tcp_newtcpcb(struct inpcb *);
1863 int      tcp_output(struct tcpcb *);
1864 void     tcp_respond(struct tcpcb *, void *ipgen __sized_by(ipgen_size), size_t ipgen_size, struct tcphdr *, struct mbuf *,
1865     tcp_seq, tcp_seq, uint32_t, uint16_t, struct tcpopt *, uint16_t, uint8_t, uint32_t, struct tcp_respond_args *, bool send_syncookie);
1866 struct rtentry *
1867 tcp_rtlookup(struct inpcb *, unsigned int);
1868 void     tcp_setpersist(struct tcpcb *);
1869 void     tcp_gc(struct inpcbinfo *);
1870 void     tcp_itimer(struct inpcbinfo *ipi);
1871 void     tcp_check_timer_state(struct tcpcb *tp);
1872 void     tcp_run_timerlist(void *arg1, void *arg2);
1873 void     tcp_sched_timers(struct tcpcb *tp);
1874 
1875 struct tcptemp *tcp_maketemplate(struct tcpcb *, struct mbuf **, struct sockaddr *, struct sockaddr *);
1876 void     tcp_fillheaders(struct mbuf *, struct tcpcb *, void *, void *, struct sockaddr *, struct sockaddr *);
1877 struct tcpcb *tcp_timers(struct tcpcb *, int);
1878 void     tcp_trace(int, int, struct tcpcb *, void *, struct tcphdr *, int);
1879 
1880 void tcp_fill_info(struct tcpcb *, struct tcp_info *);
1881 void tcp_sack_doack(struct tcpcb *, struct tcpopt *, struct tcphdr *,
1882     uint32_t *, uint32_t *);
1883 extern boolean_t tcp_sack_process_dsack(struct tcpcb *, struct tcpopt *,
1884     struct tcphdr *, boolean_t *);
1885 int tcp_detect_bad_rexmt(struct tcpcb *, struct tcphdr *, struct tcpopt *,
1886     u_int32_t rxtime);
1887 void     tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1888 void     tcp_clean_sackreport(struct tcpcb *tp);
1889 uint32_t tcp_sack_adjust(struct tcpcb *tp);
1890 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1891 void     tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1892 void     tcp_free_sackholes(struct tcpcb *tp);
1893 int32_t  tcp_sbspace(struct tcpcb *tp);
1894 void     tcp_set_tso(struct tcpcb *tp, struct ifnet *ifp);
1895 void     tcp_set_ecn(struct tcpcb *tp);
1896 void     tcp_set_l4s(struct tcpcb *tp, struct ifnet *ifp);
1897 void     tcp_set_accurate_ecn(struct tcpcb *tp);
1898 bool tcp_ecn_enabled(uint32_t ecn_flags);
1899 uint8_t  tcp_get_ace(struct tcphdr *th);
1900 uint32_t tcp_flight_size(struct tcpcb *tp);
1901 extern void tcp_get_ports_used(ifnet_t ifp, int, u_int32_t, bitstr_t *__counted_by(bitstr_size(IP_PORTRANGE_SIZE)));
1902 uint32_t tcp_count_opportunistic(unsigned int ifindex, u_int32_t flags);
1903 uint32_t tcp_find_anypcb_byaddr(struct ifaddr *ifa);
1904 uint8_t tcp_get_max_rwinscale(struct tcpcb *tp, struct socket *so);
1905 struct bwmeas* tcp_bwmeas_alloc(struct tcpcb *tp);
1906 void tcp_bwmeas_free(struct tcpcb *tp);
1907 extern int32_t timer_diff(uint32_t t1, uint32_t toff1, uint32_t t2, uint32_t toff2);
1908 
1909 /* RACK related functions */
1910 void tcp_rack_transmit_seg(struct tcpcb *tp, struct tcp_seg_sent *seg, tcp_seq start, tcp_seq end, uint32_t xmit_ts, uint8_t flags);
1911 void tcp_rack_update_reordering_window(struct tcpcb *tp, tcp_seq highest_acked_sacked);
1912 void tcp_rack_update_reordering_win_persist(struct tcpcb *tp);
1913 void tcp_rack_bad_rexmt_restore(struct tcpcb *tp);
1914 void tcp_rack_reset_segs_retransmitted(struct tcpcb *tp);
1915 void tcp_rack_update_segment_acked(struct tcpcb *tp, uint32_t tsecr, uint32_t xmit_ts, uint32_t end_seq, bool retransmitted);
1916 bool tcp_rack_detect_loss_and_arm_timer(struct tcpcb *tp, uint32_t dup_acks);
1917 void tcp_rack_reordering_timeout(struct tcpcb *tp, uint32_t dup_acks);
1918 struct tcp_seg_sent * tcp_rack_output(struct tcpcb *tp, uint32_t cwin, uint16_t *rack_seg_len);
1919 void tcp_rack_loss_on_rto(struct tcpcb *tp, bool in_rto);
1920 uint32_t tcp_rack_adjust(struct tcpcb *tp, uint32_t cwin);
1921 void tcp_rack_detect_reordering_dsack(struct tcpcb *tp, tcp_seq start, tcp_seq end);
1922 void tcp_rack_detect_reordering_acked(struct tcpcb *tp, struct tcp_seg_sent *seg);
1923 /* RACK segment related functions */
1924 uint32_t tcp_seg_len(struct tcp_seg_sent *seg);
1925 void tcp_seg_sent_insert(struct tcpcb *tp, struct tcp_seg_sent *seg, tcp_seq start, tcp_seq end, uint32_t xmit_ts, uint8_t flags);
1926 void tcp_segs_doack(struct tcpcb *tp, tcp_seq th_ack, struct tcpopt *to);
1927 void tcp_segs_dosack(struct tcpcb *tp, tcp_seq sblk_start, tcp_seq sblk_end, uint32_t tsecr, uint32_t *newbytes_sacked);
1928 void tcp_segs_clear_sacked(struct tcpcb *tp);
1929 void tcp_mark_seg_lost(struct tcpcb *tp, struct tcp_seg_sent *seg);
1930 void tcp_seg_delete(struct tcpcb *tp, struct tcp_seg_sent *seg);
1931 void tcp_segs_sent_clean(struct tcpcb *tp, bool free_segs);
1932 
1933 extern void tcp_set_background_cc(struct socket *);
1934 extern void tcp_set_foreground_cc(struct socket *);
1935 extern void tcp_set_recv_bg(struct socket *);
1936 extern void tcp_clear_recv_bg(struct socket *);
1937 extern boolean_t tcp_sack_byte_islost(struct tcpcb *tp);
1938 #define IS_TCP_RECV_BG(_so)     \
1939 	((_so)->so_flags1 & SOF1_TRAFFIC_MGT_TCP_RECVBG)
1940 
1941 #if TRAFFIC_MGT
1942 #define CLEAR_IAJ_STATE(_tp_) (_tp_)->iaj_rcv_ts = 0
1943 void     reset_acc_iaj(struct tcpcb *tp);
1944 #endif /* TRAFFIC_MGT */
1945 
1946 int      tcp_lock(struct socket *, int, void *);
1947 int      tcp_unlock(struct socket *, int, void *);
1948 void     calculate_tcp_clock(void);
1949 
1950 extern void tcp_keepalive_reset(struct tcpcb *);
1951 extern uint32_t get_base_rtt(struct tcpcb *tp);
1952 
1953 #ifdef _KERN_LOCKS_H_
1954 lck_mtx_t *      tcp_getlock(struct socket *, int);
1955 #else
1956 void *   tcp_getlock(struct socket *, int);
1957 #endif
1958 
1959 extern struct pr_usrreqs tcp_usrreqs;
1960 extern uint32_t tcp_sendspace;
1961 extern uint32_t tcp_recvspace;
1962 extern tcp_seq tcp_new_isn(struct tcpcb *tp);
1963 
1964 extern int tcp_input_checksum(int, struct mbuf *, struct tcphdr *, int, int);
1965 extern void tcp_getconninfo(struct socket *, struct conninfo_tcp *);
1966 extern void add_to_time_wait(struct tcpcb *, uint32_t delay);
1967 extern void add_to_time_wait_now(struct tcpcb *tp, uint32_t delay);
1968 extern void tcp_pmtud_revert_segment_size(struct tcpcb *tp);
1969 extern void tcp_rxtseg_insert(struct tcpcb *, tcp_seq, tcp_seq);
1970 extern struct tcp_rxt_seg *tcp_rxtseg_find(struct tcpcb *, tcp_seq, tcp_seq);
1971 extern void tcp_rxtseg_set_spurious(struct tcpcb *tp, tcp_seq start, tcp_seq end);
1972 extern void tcp_rxtseg_clean(struct tcpcb *);
1973 extern boolean_t tcp_rxtseg_detect_bad_rexmt(struct tcpcb *, tcp_seq);
1974 extern boolean_t tcp_rxtseg_dsack_for_tlp(struct tcpcb *);
1975 extern u_int32_t tcp_rxtseg_total_size(struct tcpcb *tp);
1976 extern void tcp_rexmt_save_state(struct tcpcb *tp);
1977 void tcp_local_congestion_notification(struct tcpcb *tp);
1978 void tcp_enter_fast_recovery(struct tcpcb *tp);
1979 extern void tcp_interface_send_probe(u_int16_t if_index_available);
1980 extern void tcp_probe_connectivity(struct ifnet *ifp, u_int32_t enable);
1981 extern void tcp_get_connectivity_status(struct tcpcb *,
1982     struct tcp_conn_status *);
1983 
1984 extern void tcp_clear_keep_alive_offload(struct socket *so);
1985 extern void tcp_fill_keepalive_offload_frames(struct ifnet *,
1986     struct ifnet_keepalive_offload_frame * frames_array __counted_by(frames_array_count), u_int32_t frames_array_count, size_t, u_int32_t *);
1987 extern int tcp_notify_kao_timeout(ifnet_t ifp,
1988     struct ifnet_keepalive_offload_frame *frame);
1989 
1990 extern void tcp_disable_tfo(struct tcpcb *tp);
1991 extern void tcp_tfo_gen_cookie(struct inpcb *inp, u_char *out __sized_by(blk_size), size_t blk_size);
1992 #define TCP_FASTOPEN_KEYLEN 16
1993 extern errno_t tcp_notify_ack_id_valid(struct tcpcb *, struct socket *, u_int32_t);
1994 extern errno_t tcp_add_notify_ack_marker(struct tcpcb *, u_int32_t);
1995 extern void tcp_notify_ack_free(struct tcpcb *);
1996 extern void tcp_notify_acknowledgement(struct tcpcb *, struct socket *);
1997 extern void tcp_get_notify_ack_count(struct tcpcb *,
1998     struct tcp_notify_ack_complete *);
1999 extern void tcp_get_notify_ack_ids(struct tcpcb *tp,
2000     struct tcp_notify_ack_complete *);
2001 extern void tcp_update_mss_locked(struct socket *, struct ifnet *);
2002 
2003 extern int get_tcp_inp_list(struct inpcb * __single * __counted_by(n), size_t n, inp_gen_t);
2004 extern bool tcp_notify_ack_active(struct socket *so);
2005 extern void tcp_set_finwait_timeout(struct tcpcb *);
2006 
2007 #if MPTCP
2008 extern uint16_t mptcp_output_csum(struct mbuf *m, uint64_t dss_val,
2009     uint32_t sseq, uint16_t dlen);
2010 extern int mptcp_adj_mss(struct tcpcb *, boolean_t);
2011 extern void mptcp_insert_rmap(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th);
2012 #endif
2013 
2014 extern uint32_t tcp_reass_qlen_space(struct socket *);
2015 
2016 __private_extern__ void tcp_update_stats_per_flow(
2017 	struct ifnet_stats_per_flow *, struct ifnet *);
2018 
2019 extern void tcp_set_rto(struct tcpcb *tp);
2020 extern void tcp_set_pto(struct tcpcb *tp);
2021 
2022 extern bool tcp_rst_rlc_compress(void *ipgen __sized_by(ipgen_size), size_t ipgen_size, struct tcphdr *th);
2023 
2024 extern struct mem_acct *tcp_memacct;
2025 
2026 #if SKYWALK
2027 void tcp_add_fsw_flow(struct tcpcb *, struct ifnet *);
2028 void tcp_del_fsw_flow(struct tcpcb *);
2029 #else /* !SKYWALK */
2030 #define tcp_add_fsw_flow(...)
2031 #define tcp_del_fsw_flow(...)
2032 #endif /* !SKYWALK */
2033 
2034 typedef void *__single lr_ref_t;
2035 #define TCP_INIT_LR_SAVED(lr) ((lr) == NULL                                             \
2036 	? __unsafe_forge_single(void *, __builtin_return_address(0))    \
2037 	: (lr))
2038 
2039 /*
2040  * The initial retransmission should happen at rtt + 4 * rttvar.
2041  * Because of the way we do the smoothing, srtt and rttvar
2042  * will each average +1/2 tick of bias.  When we compute
2043  * the retransmit timer, we want 1/2 tick of rounding and
2044  * 1 extra tick because of +-1/2 tick uncertainty in the
2045  * firing of the timer.  The bias will give us exactly the
2046  * 1.5 tick we need.  But, because the bias is
2047  * statistical, we have to test that we don't drop below
2048  * the minimum feasible timer (which is 2 ticks).
2049  * This version of the macro adapted from a paper by Lawrence
2050  * Brakmo and Larry Peterson which outlines a problem caused
2051  * by insufficient precision in the original implementation,
2052  * which results in inappropriately large RTO values for very
2053  * fast networks.
2054  */
2055 static inline uint32_t
tcp_rto_formula(uint32_t rttmin,uint32_t srtt,uint32_t rttvar)2056 tcp_rto_formula(uint32_t rttmin, uint32_t srtt, uint32_t rttvar)
2057 {
2058 	return max(rttmin,
2059 	           ((srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) + rttvar) >> TCP_DELTA_SHIFT);
2060 }
2061 
2062 static inline uint32_t
_tcp_offset_from_start(const struct tcpcb * tp,uint32_t offset,uint32_t tcp_now_var)2063 _tcp_offset_from_start(const struct tcpcb *tp, uint32_t offset,
2064     uint32_t tcp_now_var)
2065 {
2066 	return tcp_now_var + offset - tp->tentry.te_timer_start;
2067 }
2068 
2069 static inline uint32_t
tcp_offset_from_start(const struct tcpcb * tp,uint32_t offset)2070 tcp_offset_from_start(const struct tcpcb *tp, uint32_t offset)
2071 {
2072 	return _tcp_offset_from_start(tp, offset, tcp_now);
2073 }
2074 
2075 #define TCP_REXMTVAL(tp) tcp_rto_formula((tp)->t_rttmin, (tp)->t_srtt, (tp)->t_rttvar)
2076 #endif /* BSD_KERNEL_PRIVATE */
2077 
2078 #endif /* _NETINET_TCP_VAR_H_ */
2079