xref: /freebsd/sys/netinet/tcp_var.h (revision 5bb3134a8c21cb87b30e135ef168483f0333dabb)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1993, 1994, 1995
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
32  * $FreeBSD$
33  */
34 
35 #ifndef _NETINET_TCP_VAR_H_
36 #define _NETINET_TCP_VAR_H_
37 
38 #include <netinet/tcp.h>
39 #include <netinet/tcp_fsm.h>
40 
41 #ifdef _KERNEL
42 #include "opt_kern_tls.h"
43 #include <net/vnet.h>
44 #include <sys/mbuf.h>
45 #include <sys/ktls.h>
46 #endif
47 
48 #define TCP_END_BYTE_INFO 8	/* Bytes that makeup the "end information array" */
49 /* Types of ending byte info */
50 #define TCP_EI_EMPTY_SLOT	0
51 #define TCP_EI_STATUS_CLIENT_FIN	0x1
52 #define TCP_EI_STATUS_CLIENT_RST	0x2
53 #define TCP_EI_STATUS_SERVER_FIN	0x3
54 #define TCP_EI_STATUS_SERVER_RST	0x4
55 #define TCP_EI_STATUS_RETRAN		0x5
56 #define TCP_EI_STATUS_PROGRESS		0x6
57 #define TCP_EI_STATUS_PERSIST_MAX	0x7
58 #define TCP_EI_STATUS_KEEP_MAX		0x8
59 #define TCP_EI_STATUS_DATA_A_CLOSE	0x9
60 #define TCP_EI_STATUS_RST_IN_FRONT	0xa
61 #define TCP_EI_STATUS_2MSL		0xb
62 #define TCP_EI_STATUS_MAX_VALUE		0xb
63 
64 /************************************************/
65 /* Status bits we track to assure no duplicates,
66  * the bits here are not used by the code but
67  * for human representation. To check a bit we
68  * take and shift over by 1 minus the value (1-8).
69  */
70 /************************************************/
71 #define TCP_EI_BITS_CLIENT_FIN	0x001
72 #define TCP_EI_BITS_CLIENT_RST	0x002
73 #define TCP_EI_BITS_SERVER_FIN	0x004
74 #define TCP_EI_BITS_SERVER_RST	0x008
75 #define TCP_EI_BITS_RETRAN	0x010
76 #define TCP_EI_BITS_PROGRESS	0x020
77 #define TCP_EI_BITS_PRESIST_MAX	0x040
78 #define TCP_EI_BITS_KEEP_MAX	0x080
79 #define TCP_EI_BITS_DATA_A_CLO  0x100
80 #define TCP_EI_BITS_RST_IN_FR	0x200	/* a front state reset */
81 #define TCP_EI_BITS_2MS_TIMER	0x400	/* 2 MSL timer expired */
82 
83 #if defined(_KERNEL) || defined(_WANT_TCPCB)
84 /* TCP segment queue entry */
85 struct tseg_qent {
86 	TAILQ_ENTRY(tseg_qent) tqe_q;
87 	struct	mbuf   *tqe_m;		/* mbuf contains packet */
88 	struct  mbuf   *tqe_last;	/* last mbuf in chain */
89 	tcp_seq tqe_start;		/* TCP Sequence number start */
90 	int	tqe_len;		/* TCP segment data length */
91 	uint32_t tqe_flags;		/* The flags from the th->th_flags */
92 	uint32_t tqe_mbuf_cnt;		/* Count of mbuf overhead */
93 };
94 TAILQ_HEAD(tsegqe_head, tseg_qent);
95 
96 struct sackblk {
97 	tcp_seq start;		/* start seq no. of sack block */
98 	tcp_seq end;		/* end seq no. */
99 };
100 
101 struct sackhole {
102 	tcp_seq start;		/* start seq no. of hole */
103 	tcp_seq end;		/* end seq no. */
104 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
105 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
106 };
107 
108 struct sackhint {
109 	struct sackhole	*nexthole;
110 	int32_t		sack_bytes_rexmit;
111 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
112 
113 	int32_t		delivered_data; /* Newly acked data from last SACK */
114 
115 	int32_t		sacked_bytes;	/* Total sacked bytes reported by the
116 					 * receiver via sack option
117 					 */
118 	uint32_t	recover_fs;	/* Flight Size at the start of Loss recovery */
119 	uint32_t	prr_delivered;	/* Total bytes delivered using PRR */
120 	uint32_t	prr_out;	/* Bytes sent during IN_RECOVERY */
121 };
122 
123 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
124 
125 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
126 
127 /*
128  * Tcp control block, one per tcp; fields:
129  * Organized for 64 byte cacheline efficiency based
130  * on common tcp_input/tcp_output processing.
131  */
132 struct tcpcb {
133 	/* Cache line 1 */
134 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
135 	struct tcp_function_block *t_fb;/* TCP function call block */
136 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
137 	uint32_t t_maxseg:24,		/* maximum segment size */
138 		t_logstate:8;		/* State of "black box" logging */
139 	uint32_t t_port:16,		/* Tunneling (over udp) port */
140 		t_state:4,		/* state of this connection */
141 		t_idle_reduce : 1,
142 		t_delayed_ack: 7,	/* Delayed ack variable */
143 		t_fin_is_rst: 1,	/* Are fin's treated as resets */
144 		t_log_state_set: 1,
145 		bits_spare : 2;
146 	u_int	t_flags;
147 	tcp_seq	snd_una;		/* sent but unacknowledged */
148 	tcp_seq	snd_max;		/* highest sequence number sent;
149 					 * used to recognize retransmits
150 					 */
151 	tcp_seq	snd_nxt;		/* send next */
152 	tcp_seq	snd_up;			/* send urgent pointer */
153 	uint32_t  snd_wnd;		/* send window */
154 	uint32_t  snd_cwnd;		/* congestion-controlled window */
155 	uint32_t t_peakrate_thr; 	/* pre-calculated peak rate threshold */
156 	/* Cache line 2 */
157 	u_int32_t  ts_offset;		/* our timestamp offset */
158 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
159 	int	rcv_numsacks;		/* # distinct sack blks present */
160 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
161 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
162 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
163 	tcp_seq	rcv_nxt;		/* receive next */
164 	tcp_seq	rcv_adv;		/* advertised window */
165 	uint32_t  rcv_wnd;		/* receive window */
166 	u_int	t_flags2;		/* More tcpcb flags storage */
167 	int	t_srtt;			/* smoothed round-trip time */
168 	int	t_rttvar;		/* variance in round-trip time */
169 	u_int32_t  ts_recent;		/* timestamp echo data */
170 	u_char	snd_scale;		/* window scaling for send window */
171 	u_char	rcv_scale;		/* window scaling for recv window */
172 	u_char	snd_limited;		/* segments limited transmitted */
173 	u_char	request_r_scale;	/* pending window scaling */
174 	tcp_seq	last_ack_sent;
175 	u_int	t_rcvtime;		/* inactivity time */
176 	/* Cache line 3 */
177 	tcp_seq	rcv_up;			/* receive urgent pointer */
178 	int	t_segqlen;		/* segment reassembly queue length */
179 	uint32_t t_segqmbuflen;		/* Count of bytes mbufs on all entries */
180 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
181 	struct mbuf      *t_in_pkt;
182 	struct mbuf	 *t_tail_pkt;
183 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
184 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
185 	uint32_t  snd_ssthresh;		/* snd_cwnd size threshold for
186 					 * for slow start exponential to
187 					 * linear switch
188 					 */
189 	tcp_seq	snd_wl1;		/* window update seg seq number */
190 	/* Cache line 4 */
191 	tcp_seq	snd_wl2;		/* window update seg ack number */
192 
193 	tcp_seq	irs;			/* initial receive sequence number */
194 	tcp_seq	iss;			/* initial send sequence number */
195 	u_int	t_acktime;		/* RACK and BBR incoming new data was acked */
196 	u_int	t_sndtime;		/* time last data was sent */
197 	u_int	ts_recent_age;		/* when last updated */
198 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
199 	uint16_t cl4_spare;		/* Spare to adjust CL 4 */
200 	char	t_oobflags;		/* have some */
201 	char	t_iobc;			/* input character */
202 	int	t_rxtcur;		/* current retransmit value (ticks) */
203 
204 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
205 	u_int	t_rtttime;		/* RTT measurement start time */
206 
207 	tcp_seq	t_rtseq;		/* sequence number being timed */
208 	u_int	t_starttime;		/* time connection was established */
209 	u_int	t_fbyte_in;		/* ticks time when first byte queued in */
210 	u_int	t_fbyte_out;		/* ticks time when first byte queued out */
211 
212 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
213 	int	t_blackhole_enter;	/* when to enter blackhole detection */
214 	int	t_blackhole_exit;	/* when to exit blackhole detection */
215 	u_int	t_rttmin;		/* minimum rtt allowed */
216 
217 	u_int	t_rttbest;		/* best rtt we've seen */
218 
219 	int	t_softerror;		/* possible error not yet reported */
220 	uint32_t  max_sndwnd;		/* largest window peer has offered */
221 	/* Cache line 5 */
222 	uint32_t  snd_cwnd_prev;	/* cwnd prior to retransmit */
223 	uint32_t  snd_ssthresh_prev;	/* ssthresh prior to retransmit */
224 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
225 	int	t_sndzerowin;		/* zero-window updates sent */
226 	u_long	t_rttupdated;		/* number of times rtt sampled */
227 	int	snd_numholes;		/* number of holes seen by sender */
228 	u_int	t_badrxtwin;		/* window for retransmit recovery */
229 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
230 					/* SACK scoreboard (sorted) */
231 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
232 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
233 	struct sackhint	sackhint;	/* SACK scoreboard hint */
234 	int	t_rttlow;		/* smallest observerved RTT */
235 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
236 	struct toedev	*tod;		/* toedev handling this connection */
237 	int	t_sndrexmitpack;	/* retransmit packets sent */
238 	int	t_rcvoopack;		/* out-of-order packets received */
239 	void	*t_toe;			/* TOE pcb pointer */
240 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
241 	struct cc_var	*ccv;		/* congestion control specific vars */
242 	struct osd	*osd;		/* storage for Khelp module data */
243 	int	t_bytes_acked;		/* # bytes acked during current RTT */
244 	u_int   t_maxunacktime;
245 	u_int	t_keepinit;		/* time to establish connection */
246 	u_int	t_keepidle;		/* time before keepalive probes begin */
247 	u_int	t_keepintvl;		/* interval between keepalives */
248 	u_int	t_keepcnt;		/* number of keepalives before close */
249 	int	t_dupacks;		/* consecutive dup acks recd */
250 	int	t_lognum;		/* Number of log entries */
251 	int	t_loglimit;		/* Maximum number of log entries */
252 	int64_t	t_pacing_rate;		/* bytes / sec, -1 => unlimited */
253 	struct tcp_log_stailq t_logs;	/* Log buffer */
254 	struct tcp_log_id_node *t_lin;
255 	struct tcp_log_id_bucket *t_lib;
256 	const char *t_output_caller;	/* Function that called tcp_output */
257 	struct statsblob *t_stats;	/* Per-connection stats */
258 	uint32_t t_logsn;		/* Log "serial number" */
259 	uint32_t gput_ts;		/* Time goodput measurement started */
260 	tcp_seq gput_seq;		/* Outbound measurement seq */
261 	tcp_seq gput_ack;		/* Inbound measurement ack */
262 	int32_t t_stats_gput_prev;	/* XXXLAS: Prev gput measurement */
263 	uint32_t t_maxpeakrate;		/* max peak rate set by user, in bytes/s */
264 	uint32_t t_sndtlppack;		/* tail loss probe packets sent */
265 	uint64_t t_sndtlpbyte;		/* total tail loss probe bytes sent */
266 	uint64_t t_sndbytes;		/* total bytes sent */
267 	uint64_t t_snd_rxt_bytes;	/* total bytes retransmitted */
268 	uint32_t t_dsack_bytes;		/* Total number of dsack bytes we have received */
269 	uint32_t t_dsack_tlp_bytes;	/* Total number of dsack bytes we have received for TLPs sent */
270 	uint32_t t_dsack_pack;		/* Total dsack packets we have recieved */
271 
272 	uint8_t t_tfo_client_cookie_len; /* TCP Fast Open client cookie length */
273 	uint32_t t_end_info_status;	/* Status flag of end info */
274 	unsigned int *t_tfo_pending;	/* TCP Fast Open server pending counter */
275 	union {
276 		uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
277 		uint64_t server;
278 	} t_tfo_cookie;			/* TCP Fast Open cookie to send */
279 	union {
280 		uint8_t t_end_info_bytes[TCP_END_BYTE_INFO];
281 		uint64_t t_end_info;
282 	};
283 #ifdef TCPPCAP
284 	struct mbufq t_inpkts;		/* List of saved input packets. */
285 	struct mbufq t_outpkts;		/* List of saved output packets. */
286 #endif
287 };
288 #endif	/* _KERNEL || _WANT_TCPCB */
289 
290 #ifdef _KERNEL
291 struct tcptemp {
292 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
293 	struct	tcphdr tt_t;
294 };
295 
296 /* Enable TCP/UDP tunneling port */
297 #define TCP_TUNNELING_PORT_MIN		0
298 #define TCP_TUNNELING_PORT_MAX		65535
299 #define TCP_TUNNELING_PORT_DEFAULT	0
300 
301 /* Enable TCP/UDP tunneling port */
302 #define TCP_TUNNELING_OVERHEAD_MIN	sizeof(struct udphdr)
303 #define TCP_TUNNELING_OVERHEAD_MAX	1024
304 #define TCP_TUNNELING_OVERHEAD_DEFAULT	TCP_TUNNELING_OVERHEAD_MIN
305 
306 /* Minimum map entries limit value, if set */
307 #define TCP_MIN_MAP_ENTRIES_LIMIT	128
308 
309 /*
310  * TODO: We yet need to brave plowing in
311  * to tcp_input() and the pru_usrreq() block.
312  * Right now these go to the old standards which
313  * are somewhat ok, but in the long term may
314  * need to be changed. If we do tackle tcp_input()
315  * then we need to get rid of the tcp_do_segment()
316  * function below.
317  */
318 /* Flags for tcp functions */
319 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
320 
321 /*
322  * If defining the optional tcp_timers, in the
323  * tfb_tcp_timer_stop call you must use the
324  * callout_async_drain() function with the
325  * tcp_timer_discard callback. You should check
326  * the return of callout_async_drain() and if 0
327  * increment tt_draincnt. Since the timer sub-system
328  * does not know your callbacks you must provide a
329  * stop_all function that loops through and calls
330  * tcp_timer_stop() with each of your defined timers.
331  * Adding a tfb_tcp_handoff_ok function allows the socket
332  * option to change stacks to query you even if the
333  * connection is in a later stage. You return 0 to
334  * say you can take over and run your stack, you return
335  * non-zero (an error number) to say no you can't.
336  * If the function is undefined you can only change
337  * in the early states (before connect or listen).
338  * tfb_tcp_fb_fini is changed to add a flag to tell
339  * the old stack if the tcb is being destroyed or
340  * not. A one in the flag means the TCB is being
341  * destroyed, a zero indicates its transitioning to
342  * another stack (via socket option).
343  */
344 struct tcp_function_block {
345 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
346 	int	(*tfb_tcp_output)(struct tcpcb *);
347 	int	(*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
348 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
349 			    struct socket *, struct tcpcb *,
350 		        int, int, uint8_t);
351 	int     (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int);
352 	int      (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *,
353 			    struct socket *, struct tcpcb *,
354 			    int, int, uint8_t,
355 			    int, struct timeval *);
356 	void	(*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
357 			    struct socket *, struct tcpcb *,
358 			    int, int, uint8_t,
359 			    int, struct timeval *);
360 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
361 			    struct inpcb *inp, struct tcpcb *tp);
362 	/* Optional memory allocation/free routine */
363 	int	(*tfb_tcp_fb_init)(struct tcpcb *);
364 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
365 	/* Optional timers, must define all if you define one */
366 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
367 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
368 			    uint32_t, u_int);
369 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
370 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
371 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
372 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
373 	void	(*tfb_tcp_mtu_chg)(struct tcpcb *);
374 	int	(*tfb_pru_options)(struct tcpcb *, int);
375 	void	(*tfb_hwtls_change)(struct tcpcb *, int);
376 	volatile uint32_t tfb_refcnt;
377 	uint32_t  tfb_flags;
378 	uint8_t	tfb_id;
379 };
380 
381 struct tcp_function {
382 	TAILQ_ENTRY(tcp_function)	tf_next;
383 	char				tf_name[TCP_FUNCTION_NAME_LEN_MAX];
384 	struct tcp_function_block	*tf_fb;
385 };
386 
387 TAILQ_HEAD(tcp_funchead, tcp_function);
388 #endif	/* _KERNEL */
389 
390 /*
391  * Flags and utility macros for the t_flags field.
392  */
393 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
394 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
395 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
396 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
397 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
398 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
399 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
400 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
401 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
402 #define	TF_SACK_PERMIT	0x00000200	/* other side said I could SACK */
403 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
404 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
405 #define	TF_NOPUSH	0x00001000	/* don't push */
406 #define	TF_PREVVALID	0x00002000	/* saved values for bad rxmit valid */
407 #define	TF_WAKESOR	0x00004000	/* wake up receive socket */
408 #define	TF_GPUTINPROG	0x00008000	/* Goodput measurement in progress */
409 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
410 #define	TF_LQ_OVERFLOW	0x00020000	/* listen queue overflow */
411 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
412 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
413 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
414 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
415 #define	TF_SIGNATURE	0x00400000	/* require MD5 digests (RFC2385) */
416 #define	TF_FORCEDATA	0x00800000	/* force out a byte */
417 #define	TF_TSO		0x01000000	/* TSO enabled on this connection */
418 #define	TF_TOE		0x02000000	/* this connection is offloaded */
419 #define	TF_UNUSED0	0x04000000	/* unused */
420 #define	TF_UNUSED1	0x08000000	/* unused */
421 #define	TF_LRD		0x10000000	/* Lost Retransmission Detection */
422 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
423 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
424 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
425 
426 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
427 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
428 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
429 
430 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
431 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
432 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
433 
434 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
435 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
436 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
437 
438 #if defined(_KERNEL) && !defined(TCP_RFC7413)
439 #define	IS_FASTOPEN(t_flags)		(false)
440 #else
441 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
442 #endif
443 
444 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
445 
446 /*
447  * Flags for the t_oobflags field.
448  */
449 #define	TCPOOB_HAVEDATA	0x01
450 #define	TCPOOB_HADDATA	0x02
451 
452 /*
453  * Flags for the extended TCP flags field, t_flags2
454  */
455 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
456 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
457 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
458 #define	TF2_LOG_AUTO		0x00000008 /* Session is auto-logging. */
459 #define TF2_DROP_AF_DATA 	0x00000010 /* Drop after all data ack'd */
460 #define	TF2_ECN_PERMIT		0x00000020 /* connection ECN-ready */
461 #define	TF2_ECN_SND_CWR		0x00000040 /* ECN CWR in queue */
462 #define	TF2_ECN_SND_ECE		0x00000080 /* ECN ECE in queue */
463 #define	TF2_ACE_PERMIT		0x00000100 /* Accurate ECN mode */
464 #define TF2_FBYTES_COMPLETE	0x00000400 /* We have first bytes in and out */
465 /*
466  * Structure to hold TCP options that are only used during segment
467  * processing (in tcp_input), but not held in the tcpcb.
468  * It's basically used to reduce the number of parameters
469  * to tcp_dooptions and tcp_addoptions.
470  * The binary order of the to_flags is relevant for packing of the
471  * options in tcp_addoptions.
472  */
473 struct tcpopt {
474 	u_int32_t	to_flags;	/* which options are present */
475 #define	TOF_MSS		0x0001		/* maximum segment size */
476 #define	TOF_SCALE	0x0002		/* window scaling */
477 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
478 #define	TOF_TS		0x0010		/* timestamp */
479 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
480 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
481 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
482 #define	TOF_MAXOPT	0x0200
483 	u_int32_t	to_tsval;	/* new timestamp */
484 	u_int32_t	to_tsecr;	/* reflected timestamp */
485 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
486 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
487 	u_int8_t	*to_tfo_cookie; /* pointer to the TFO cookie */
488 	u_int16_t	to_mss;		/* maximum segment size */
489 	u_int8_t	to_wscale;	/* window scaling */
490 	u_int8_t	to_nsacks;	/* number of SACK blocks */
491 	u_int8_t	to_tfo_len;	/* TFO cookie length */
492 	u_int32_t	to_spare;	/* UTO */
493 };
494 
495 /*
496  * Flags for tcp_dooptions.
497  */
498 #define	TO_SYN		0x01		/* parse SYN-only options */
499 
500 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
501 	uint32_t	rmx_mtu;	/* MTU for this path */
502 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
503 	uint32_t	rmx_rtt;	/* estimated round trip time */
504 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
505 	uint32_t	rmx_cwnd;	/* congestion window */
506 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
507 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
508 };
509 
510 /*
511  * Used by tcp_maxmtu() to communicate interface specific features
512  * and limits at the time of connection setup.
513  */
514 struct tcp_ifcap {
515 	int	ifcap;
516 	u_int	tsomax;
517 	u_int	tsomaxsegcount;
518 	u_int	tsomaxsegsize;
519 };
520 
521 #ifndef _NETINET_IN_PCB_H_
522 struct in_conninfo;
523 #endif /* _NETINET_IN_PCB_H_ */
524 
525 struct tcptw {
526 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
527 	uint32_t  t_port:16,		/* UDP port number if TCPoUDP */
528 		t_unused:16;
529 	tcp_seq		snd_nxt;
530 	tcp_seq		rcv_nxt;
531 	tcp_seq		iss;
532 	tcp_seq		irs;
533 	u_short		last_win;	/* cached window value */
534 	short		tw_so_options;	/* copy of so_options */
535 	struct ucred	*tw_cred;	/* user credentials */
536 	u_int32_t	t_recent;
537 	u_int32_t	ts_offset;	/* our timestamp offset */
538 	u_int		t_starttime;
539 	int		tw_time;
540 	TAILQ_ENTRY(tcptw) tw_2msl;
541 	void		*tw_pspare;	/* TCP_SIGNATURE */
542 	u_int		*tw_spare;	/* TCP_SIGNATURE */
543 };
544 
545 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
546 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
547 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
548 
549 /*
550  * The smoothed round-trip time and estimated variance
551  * are stored as fixed point numbers scaled by the values below.
552  * For convenience, these scales are also used in smoothing the average
553  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
554  * With these scales, srtt has 3 bits to the right of the binary point,
555  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
556  * binary point, and is smoothed with an ALPHA of 0.75.
557  */
558 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
559 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
560 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
561 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
562 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
563 
564 /*
565  * The initial retransmission should happen at rtt + 4 * rttvar.
566  * Because of the way we do the smoothing, srtt and rttvar
567  * will each average +1/2 tick of bias.  When we compute
568  * the retransmit timer, we want 1/2 tick of rounding and
569  * 1 extra tick because of +-1/2 tick uncertainty in the
570  * firing of the timer.  The bias will give us exactly the
571  * 1.5 tick we need.  But, because the bias is
572  * statistical, we have to test that we don't drop below
573  * the minimum feasible timer (which is 2 ticks).
574  * This version of the macro adapted from a paper by Lawrence
575  * Brakmo and Larry Peterson which outlines a problem caused
576  * by insufficient precision in the original implementation,
577  * which results in inappropriately large RTO values for very
578  * fast networks.
579  */
580 #define	TCP_REXMTVAL(tp) \
581 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
582 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
583 
584 /*
585  * TCP statistics.
586  * Many of these should be kept per connection,
587  * but that's inconvenient at the moment.
588  */
589 struct	tcpstat {
590 	uint64_t tcps_connattempt;	/* connections initiated */
591 	uint64_t tcps_accepts;		/* connections accepted */
592 	uint64_t tcps_connects;		/* connections established */
593 	uint64_t tcps_drops;		/* connections dropped */
594 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
595 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
596 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
597 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
598 	uint64_t tcps_rttupdated;	/* times we succeeded */
599 	uint64_t tcps_delack;		/* delayed acks sent */
600 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
601 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
602 	uint64_t tcps_persisttimeo;	/* persist timeouts */
603 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
604 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
605 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
606 
607 	uint64_t tcps_sndtotal;		/* total packets sent */
608 	uint64_t tcps_sndpack;		/* data packets sent */
609 	uint64_t tcps_sndbyte;		/* data bytes sent */
610 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
611 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
612 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
613 	uint64_t tcps_sndacks;		/* ack-only packets sent */
614 	uint64_t tcps_sndprobe;		/* window probes sent */
615 	uint64_t tcps_sndurg;		/* packets sent with URG only */
616 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
617 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
618 
619 	uint64_t tcps_rcvtotal;		/* total packets received */
620 	uint64_t tcps_rcvpack;		/* packets received in sequence */
621 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
622 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
623 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
624 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
625 	uint64_t tcps_rcvshort;		/* packets received too short */
626 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
627 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
628 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
629 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
630 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
631 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
632 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
633 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
634 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
635 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
636 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
637 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
638 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
639 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
640 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
641 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
642 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
643 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
644 	uint64_t tcps_pcbcachemiss;
645 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
646 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
647 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
648 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
649 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
650 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
651 	uint64_t tcps_persistdrop;	/* timeout in persist state */
652 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
653 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
654 	uint64_t tcps_listendrop;	/* listen queue overflows */
655 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
656 
657 	uint64_t tcps_sc_added;		/* entry added to syncache */
658 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
659 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
660 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
661 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
662 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
663 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
664 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
665 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
666 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
667 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
668 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
669 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
670 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
671 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
672 
673 	uint64_t tcps_hc_added;		/* entry added to hostcache */
674 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
675 
676 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
677 
678 	/* SACK related stats */
679 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
680 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
681 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
682 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
683 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
684 	uint64_t tcps_sack_lostrexmt;	    /* SACK lost retransmission recovered */
685 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
686 
687 	/* ECN related stats */
688 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
689 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
690 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
691 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
692 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
693 
694 	/* TCP_SIGNATURE related stats */
695 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
696 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
697 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
698 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
699 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
700 
701 	/* Path MTU Discovery Black Hole Detection related stats */
702 	uint64_t tcps_pmtud_blackhole_activated;	 /* Black Hole Count */
703 	uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
704 	uint64_t tcps_pmtud_blackhole_failed;		 /* Black Hole Failure Count */
705 
706 	uint64_t tcps_tunneled_pkts;	/* Packets encap's in UDP received */
707 	uint64_t tcps_tunneled_errs;	/* Packets that had errors that were UDP encaped */
708 
709 	/* Dsack related stats */
710 	uint64_t tcps_dsack_count;	/* Number of ACKs arriving with DSACKs */
711 	uint64_t tcps_dsack_bytes;	/* Number of bytes DSACK'ed no TLP */
712 	uint64_t tcps_dsack_tlp_bytes;	/* Number of bytes DSACK'ed due to TLPs */
713 
714 	uint64_t _pad[6];		/* 3 UTO, 3 TBD */
715 };
716 
717 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
718 
719 #ifdef _KERNEL
720 #define	TI_UNLOCKED	1
721 #define	TI_RLOCKED	2
722 #include <sys/counter.h>
723 
724 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
725 /*
726  * In-kernel consumers can use these accessor macros directly to update
727  * stats.
728  */
729 #define	TCPSTAT_ADD(name, val)	\
730     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
731 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
732 
733 /*
734  * Kernel module consumers must use this accessor macro.
735  */
736 void	kmod_tcpstat_add(int statnum, int val);
737 #define	KMOD_TCPSTAT_ADD(name, val)					\
738     kmod_tcpstat_add(offsetof(struct tcpstat, name) / sizeof(uint64_t), val)
739 #define	KMOD_TCPSTAT_INC(name)	KMOD_TCPSTAT_ADD(name, 1)
740 
741 /*
742  * Running TCP connection count by state.
743  */
744 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
745 #define	V_tcps_states	VNET(tcps_states)
746 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
747 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
748 
749 /*
750  * TCP specific helper hook point identifiers.
751  */
752 #define	HHOOK_TCP_EST_IN		0
753 #define	HHOOK_TCP_EST_OUT		1
754 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
755 
756 struct tcp_hhook_data {
757 	struct tcpcb	*tp;
758 	struct tcphdr	*th;
759 	struct tcpopt	*to;
760 	uint32_t	len;
761 	int		tso;
762 	tcp_seq		curack;
763 };
764 #ifdef TCP_HHOOK
765 void hhook_run_tcp_est_out(struct tcpcb *tp,
766 	struct tcphdr *th, struct tcpopt *to,
767 	uint32_t len, int tso);
768 #endif
769 #endif
770 
771 /*
772  * TCB structure exported to user-land via sysctl(3).
773  *
774  * Fields prefixed with "xt_" are unique to the export structure, and fields
775  * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
776  *
777  * Legend:
778  * (s) - used by userland utilities in src
779  * (p) - used by utilities in ports
780  * (3) - is known to be used by third party software not in ports
781  * (n) - no known usage
782  *
783  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
784  * included.  Not all of our clients do.
785  */
786 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
787 struct xtcpcb {
788 	ksize_t	xt_len;		/* length of this structure */
789 	struct xinpcb	xt_inp;
790 	char		xt_stack[TCP_FUNCTION_NAME_LEN_MAX];	/* (s) */
791 	char		xt_logid[TCP_LOG_ID_LEN];	/* (s) */
792 	char		xt_cc[TCP_CA_NAME_MAX];	/* (s) */
793 	int64_t		spare64[6];
794 	int32_t		t_state;		/* (s,p) */
795 	uint32_t	t_flags;		/* (s,p) */
796 	int32_t		t_sndzerowin;		/* (s) */
797 	int32_t		t_sndrexmitpack;	/* (s) */
798 	int32_t		t_rcvoopack;		/* (s) */
799 	int32_t		t_rcvtime;		/* (s) */
800 	int32_t		tt_rexmt;		/* (s) */
801 	int32_t		tt_persist;		/* (s) */
802 	int32_t		tt_keep;		/* (s) */
803 	int32_t		tt_2msl;		/* (s) */
804 	int32_t		tt_delack;		/* (s) */
805 	int32_t		t_logstate;		/* (3) */
806 	uint32_t	t_snd_cwnd;		/* (s) */
807 	uint32_t	t_snd_ssthresh;		/* (s) */
808 	uint32_t	t_maxseg;		/* (s) */
809 	uint32_t	t_rcv_wnd;		/* (s) */
810 	uint32_t	t_snd_wnd;		/* (s) */
811 	uint32_t	xt_ecn;			/* (s) */
812 	uint32_t	t_dsack_bytes;		/* (n) */
813 	uint32_t	t_dsack_tlp_bytes;	/* (n) */
814 	uint32_t	t_dsack_pack;		/* (n) */
815 	uint16_t	xt_encaps_port;		/* (s) */
816 	int16_t		spare16;
817 	int32_t		spare32[22];
818 } __aligned(8);
819 
820 #ifdef _KERNEL
821 void	tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
822 #endif
823 #endif
824 
825 /*
826  * TCP function information (name-to-id mapping, aliases, and refcnt)
827  * exported to user-land via sysctl(3).
828  */
829 struct tcp_function_info {
830 	uint32_t	tfi_refcnt;
831 	uint8_t		tfi_id;
832 	char		tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
833 	char		tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
834 };
835 
836 /*
837  * Identifiers for TCP sysctl nodes
838  */
839 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
840 #define	TCPCTL_MSSDFLT		3	/* MSS default */
841 #define TCPCTL_STATS		4	/* statistics */
842 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
843 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
844 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
845 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
846 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
847 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
848 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
849 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
850 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
851 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
852 #define	TCPCTL_DROP		15	/* drop tcp connection */
853 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
854 
855 #ifdef _KERNEL
856 #ifdef SYSCTL_DECL
857 SYSCTL_DECL(_net_inet_tcp);
858 SYSCTL_DECL(_net_inet_tcp_sack);
859 MALLOC_DECLARE(M_TCPLOG);
860 #endif
861 
862 VNET_DECLARE(int, tcp_log_in_vain);
863 #define	V_tcp_log_in_vain		VNET(tcp_log_in_vain)
864 
865 /*
866  * Global TCP tunables shared between different stacks.
867  * Please keep the list sorted.
868  */
869 VNET_DECLARE(int, drop_synfin);
870 VNET_DECLARE(int, path_mtu_discovery);
871 VNET_DECLARE(int, tcp_abc_l_var);
872 VNET_DECLARE(int, tcp_autorcvbuf_max);
873 VNET_DECLARE(int, tcp_autosndbuf_inc);
874 VNET_DECLARE(int, tcp_autosndbuf_max);
875 VNET_DECLARE(int, tcp_delack_enabled);
876 VNET_DECLARE(int, tcp_do_autorcvbuf);
877 VNET_DECLARE(int, tcp_do_autosndbuf);
878 VNET_DECLARE(int, tcp_do_ecn);
879 VNET_DECLARE(int, tcp_do_lrd);
880 VNET_DECLARE(int, tcp_do_prr);
881 VNET_DECLARE(int, tcp_do_prr_conservative);
882 VNET_DECLARE(int, tcp_do_newcwv);
883 VNET_DECLARE(int, tcp_do_rfc1323);
884 VNET_DECLARE(int, tcp_tolerate_missing_ts);
885 VNET_DECLARE(int, tcp_do_rfc3042);
886 VNET_DECLARE(int, tcp_do_rfc3390);
887 VNET_DECLARE(int, tcp_do_rfc3465);
888 VNET_DECLARE(int, tcp_do_newsack);
889 VNET_DECLARE(int, tcp_do_sack);
890 VNET_DECLARE(int, tcp_do_tso);
891 VNET_DECLARE(int, tcp_ecn_maxretries);
892 VNET_DECLARE(int, tcp_initcwnd_segments);
893 VNET_DECLARE(int, tcp_insecure_rst);
894 VNET_DECLARE(int, tcp_insecure_syn);
895 VNET_DECLARE(uint32_t, tcp_map_entries_limit);
896 VNET_DECLARE(uint32_t, tcp_map_split_limit);
897 VNET_DECLARE(int, tcp_minmss);
898 VNET_DECLARE(int, tcp_mssdflt);
899 #ifdef STATS
900 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl);
901 VNET_DECLARE(int, tcp_perconn_stats_enable);
902 #endif /* STATS */
903 VNET_DECLARE(int, tcp_recvspace);
904 VNET_DECLARE(int, tcp_sack_globalholes);
905 VNET_DECLARE(int, tcp_sack_globalmaxholes);
906 VNET_DECLARE(int, tcp_sack_maxholes);
907 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
908 VNET_DECLARE(int, tcp_sendspace);
909 VNET_DECLARE(int, tcp_udp_tunneling_overhead);
910 VNET_DECLARE(int, tcp_udp_tunneling_port);
911 VNET_DECLARE(struct inpcbhead, tcb);
912 VNET_DECLARE(struct inpcbinfo, tcbinfo);
913 
914 #define	V_tcp_do_lrd			VNET(tcp_do_lrd)
915 #define	V_tcp_do_prr			VNET(tcp_do_prr)
916 #define	V_tcp_do_prr_conservative	VNET(tcp_do_prr_conservative)
917 #define	V_tcp_do_newcwv			VNET(tcp_do_newcwv)
918 #define	V_drop_synfin			VNET(drop_synfin)
919 #define	V_path_mtu_discovery		VNET(path_mtu_discovery)
920 #define	V_tcb				VNET(tcb)
921 #define	V_tcbinfo			VNET(tcbinfo)
922 #define	V_tcp_abc_l_var			VNET(tcp_abc_l_var)
923 #define	V_tcp_autorcvbuf_max		VNET(tcp_autorcvbuf_max)
924 #define	V_tcp_autosndbuf_inc		VNET(tcp_autosndbuf_inc)
925 #define	V_tcp_autosndbuf_max		VNET(tcp_autosndbuf_max)
926 #define	V_tcp_delack_enabled		VNET(tcp_delack_enabled)
927 #define	V_tcp_do_autorcvbuf		VNET(tcp_do_autorcvbuf)
928 #define	V_tcp_do_autosndbuf		VNET(tcp_do_autosndbuf)
929 #define	V_tcp_do_ecn			VNET(tcp_do_ecn)
930 #define	V_tcp_do_rfc1323		VNET(tcp_do_rfc1323)
931 #define	V_tcp_tolerate_missing_ts	VNET(tcp_tolerate_missing_ts)
932 #define V_tcp_ts_offset_per_conn	VNET(tcp_ts_offset_per_conn)
933 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
934 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)
935 #define	V_tcp_do_rfc3465		VNET(tcp_do_rfc3465)
936 #define	V_tcp_do_newsack		VNET(tcp_do_newsack)
937 #define	V_tcp_do_sack			VNET(tcp_do_sack)
938 #define	V_tcp_do_tso			VNET(tcp_do_tso)
939 #define	V_tcp_ecn_maxretries		VNET(tcp_ecn_maxretries)
940 #define	V_tcp_initcwnd_segments		VNET(tcp_initcwnd_segments)
941 #define	V_tcp_insecure_rst		VNET(tcp_insecure_rst)
942 #define	V_tcp_insecure_syn		VNET(tcp_insecure_syn)
943 #define	V_tcp_map_entries_limit		VNET(tcp_map_entries_limit)
944 #define	V_tcp_map_split_limit		VNET(tcp_map_split_limit)
945 #define	V_tcp_minmss			VNET(tcp_minmss)
946 #define	V_tcp_mssdflt			VNET(tcp_mssdflt)
947 #ifdef STATS
948 #define	V_tcp_perconn_stats_dflt_tpl	VNET(tcp_perconn_stats_dflt_tpl)
949 #define	V_tcp_perconn_stats_enable	VNET(tcp_perconn_stats_enable)
950 #endif /* STATS */
951 #define	V_tcp_recvspace			VNET(tcp_recvspace)
952 #define	V_tcp_sack_globalholes		VNET(tcp_sack_globalholes)
953 #define	V_tcp_sack_globalmaxholes	VNET(tcp_sack_globalmaxholes)
954 #define	V_tcp_sack_maxholes		VNET(tcp_sack_maxholes)
955 #define	V_tcp_sc_rst_sock_fail		VNET(tcp_sc_rst_sock_fail)
956 #define	V_tcp_sendspace			VNET(tcp_sendspace)
957 #define	V_tcp_udp_tunneling_overhead	VNET(tcp_udp_tunneling_overhead)
958 #define	V_tcp_udp_tunneling_port	VNET(tcp_udp_tunneling_port)
959 
960 #ifdef TCP_HHOOK
961 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
962 #define	V_tcp_hhh		VNET(tcp_hhh)
963 #endif
964 
965 int	 tcp_addoptions(struct tcpopt *, u_char *);
966 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
967 struct tcpcb *
968 	 tcp_close(struct tcpcb *);
969 void	 tcp_discardcb(struct tcpcb *);
970 void	 tcp_twstart(struct tcpcb *);
971 void	 tcp_twclose(struct tcptw *, int);
972 void	 tcp_ctlinput(int, struct sockaddr *, void *);
973 int	 tcp_ctloutput(struct socket *, struct sockopt *);
974 void 	 tcp_ctlinput_viaudp(int, struct sockaddr *, void *, void *);
975 struct tcpcb *
976 	 tcp_drop(struct tcpcb *, int);
977 void	 tcp_drain(void);
978 void	 tcp_init(void);
979 void	 tcp_fini(void *);
980 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
981 	    const void *);
982 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
983 	    const void *);
984 int	 tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *,
985 	    struct mbuf *);
986 void	 tcp_reass_global_init(void);
987 void	 tcp_reass_flush(struct tcpcb *);
988 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
989 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
990 		     struct tcpcb *, int, int);
991 void	tcp_pulloutofband(struct socket *,
992 		     struct tcphdr *, struct mbuf *, int);
993 void	tcp_xmit_timer(struct tcpcb *, int);
994 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
995 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
996 			    uint16_t nsegs, uint16_t type);
997 void 	cc_conn_init(struct tcpcb *tp);
998 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
999 void    cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos);
1000 void	cc_ecnpkt_handler_flags(struct tcpcb *tp, uint16_t flags, uint8_t iptos);
1001 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
1002 #ifdef TCP_HHOOK
1003 void	hhook_run_tcp_est_in(struct tcpcb *tp,
1004 			    struct tcphdr *th, struct tcpopt *to);
1005 #endif
1006 
1007 int	 tcp_input(struct mbuf **, int *, int);
1008 int	 tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
1009 	    struct tcpcb *, int);
1010 int	 tcp_input_with_port(struct mbuf **, int *, int, uint16_t);
1011 void	 tcp_handle_wakeup(struct tcpcb *, struct socket *);
1012 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
1013 			struct socket *, struct tcpcb *, int, int, uint8_t);
1014 
1015 int register_tcp_functions(struct tcp_function_block *blk, int wait);
1016 int register_tcp_functions_as_names(struct tcp_function_block *blk,
1017     int wait, const char *names[], int *num_names);
1018 int register_tcp_functions_as_name(struct tcp_function_block *blk,
1019     const char *name, int wait);
1020 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
1021     bool force);
1022 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
1023 int find_tcp_function_alias(struct tcp_function_block *blk, struct tcp_function_set *fs);
1024 void tcp_switch_back_to_default(struct tcpcb *tp);
1025 struct tcp_function_block *
1026 find_and_ref_tcp_fb(struct tcp_function_block *fs);
1027 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
1028 
1029 extern counter_u64_t tcp_inp_lro_direct_queue;
1030 extern counter_u64_t tcp_inp_lro_wokeup_queue;
1031 extern counter_u64_t tcp_inp_lro_compressed;
1032 extern counter_u64_t tcp_inp_lro_locks_taken;
1033 extern counter_u64_t tcp_extra_mbuf;
1034 extern counter_u64_t tcp_would_have_but;
1035 extern counter_u64_t tcp_comp_total;
1036 extern counter_u64_t tcp_uncomp_total;
1037 extern counter_u64_t tcp_bad_csums;
1038 
1039 #ifdef NETFLIX_EXP_DETECTION
1040 /* Various SACK attack thresholds */
1041 extern int32_t tcp_force_detection;
1042 extern int32_t tcp_sack_to_ack_thresh;
1043 extern int32_t tcp_sack_to_move_thresh;
1044 extern int32_t tcp_restoral_thresh;
1045 extern int32_t tcp_sad_decay_val;
1046 extern int32_t tcp_sad_pacing_interval;
1047 extern int32_t tcp_sad_low_pps;
1048 extern int32_t tcp_map_minimum;
1049 extern int32_t tcp_attack_on_turns_on_logging;
1050 #endif
1051 extern uint32_t tcp_ack_war_time_window;
1052 extern uint32_t tcp_ack_war_cnt;
1053 
1054 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
1055 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
1056 void	 tcp6_use_min_mtu(struct tcpcb *);
1057 u_int	 tcp_maxseg(const struct tcpcb *);
1058 u_int	 tcp_fixed_maxseg(const struct tcpcb *);
1059 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
1060 	    struct tcp_ifcap *);
1061 void	 tcp_mss(struct tcpcb *, int);
1062 int	 tcp_mssopt(struct in_conninfo *);
1063 struct inpcb *
1064 	 tcp_drop_syn_sent(struct inpcb *, int);
1065 struct tcpcb *
1066 	 tcp_newtcpcb(struct inpcb *);
1067 int	 tcp_output(struct tcpcb *);
1068 void	 tcp_state_change(struct tcpcb *, int);
1069 void	 tcp_respond(struct tcpcb *, void *,
1070 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
1071 void	 tcp_tw_init(void);
1072 #ifdef VIMAGE
1073 void	 tcp_tw_destroy(void);
1074 #endif
1075 void	 tcp_tw_zone_change(void);
1076 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
1077 	    struct mbuf *, int);
1078 void	 tcp_setpersist(struct tcpcb *);
1079 void	 tcp_slowtimo(void);
1080 void	 tcp_record_dsack(struct tcpcb *tp, tcp_seq start, tcp_seq end, int tlp);
1081 struct tcptemp *
1082 	 tcpip_maketemplate(struct inpcb *);
1083 void	 tcpip_fillheaders(struct inpcb *, uint16_t, void *, void *);
1084 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
1085 int	 tcp_timer_suspend(struct tcpcb *, uint32_t);
1086 void	 tcp_timers_unsuspend(struct tcpcb *, uint32_t);
1087 int	 tcp_timer_active(struct tcpcb *, uint32_t);
1088 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
1089 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
1090 int	 inp_to_cpuid(struct inpcb *inp);
1091 /*
1092  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
1093  */
1094 void	 tcp_hc_init(void);
1095 #ifdef VIMAGE
1096 void	 tcp_hc_destroy(void);
1097 #endif
1098 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
1099 uint32_t tcp_hc_getmtu(struct in_conninfo *);
1100 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
1101 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
1102 
1103 extern	struct pr_usrreqs tcp_usrreqs;
1104 
1105 uint32_t tcp_new_ts_offset(struct in_conninfo *);
1106 tcp_seq	 tcp_new_isn(struct in_conninfo *);
1107 
1108 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
1109 int	 tcp_dsack_block_exists(struct tcpcb *);
1110 void	 tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
1111 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1112 void	 tcp_clean_dsack_blocks(struct tcpcb *tp);
1113 void	 tcp_clean_sackreport(struct tcpcb *tp);
1114 void	 tcp_sack_adjust(struct tcpcb *tp);
1115 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1116 void	 tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *);
1117 void	 tcp_lost_retransmission(struct tcpcb *, struct tcphdr *);
1118 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1119 void	 tcp_free_sackholes(struct tcpcb *tp);
1120 void	 tcp_sack_lost_retransmission(struct tcpcb *, struct tcphdr *);
1121 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
1122 int	 tcp_compute_pipe(struct tcpcb *);
1123 uint32_t tcp_compute_initwnd(uint32_t);
1124 void	 tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
1125 int	 tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes,
1126     size_t seed_len);
1127 int tcp_can_enable_pacing(void);
1128 void tcp_decrement_paced_conn(void);
1129 
1130 struct mbuf *
1131 	 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
1132 	   int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls);
1133 
1134 int	tcp_stats_init(void);
1135 void tcp_log_end_status(struct tcpcb *tp, uint8_t status);
1136 
1137 static inline void
1138 tcp_fields_to_host(struct tcphdr *th)
1139 {
1140 
1141 	th->th_seq = ntohl(th->th_seq);
1142 	th->th_ack = ntohl(th->th_ack);
1143 	th->th_win = ntohs(th->th_win);
1144 	th->th_urp = ntohs(th->th_urp);
1145 }
1146 
1147 static inline void
1148 tcp_fields_to_net(struct tcphdr *th)
1149 {
1150 
1151 	th->th_seq = htonl(th->th_seq);
1152 	th->th_ack = htonl(th->th_ack);
1153 	th->th_win = htons(th->th_win);
1154 	th->th_urp = htons(th->th_urp);
1155 }
1156 
1157 static inline void
1158 tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt,
1159     uint8_t is_tlp, int hw_tls)
1160 {
1161 	if (is_tlp) {
1162 		tp->t_sndtlppack++;
1163 		tp->t_sndtlpbyte += len;
1164 	}
1165 	/* To get total bytes sent you must add t_snd_rxt_bytes to t_sndbytes */
1166 	if (is_rxt)
1167 		tp->t_snd_rxt_bytes += len;
1168 	else
1169 		tp->t_sndbytes += len;
1170 
1171 #ifdef KERN_TLS
1172 	if (hw_tls && is_rxt && len != 0) {
1173 		uint64_t rexmit_percent = (1000ULL * tp->t_snd_rxt_bytes) / (10ULL * (tp->t_snd_rxt_bytes + tp->t_sndbytes));
1174 		if (rexmit_percent > ktls_ifnet_max_rexmit_pct)
1175 			ktls_disable_ifnet(tp);
1176 	}
1177 #endif
1178 
1179 }
1180 #endif /* _KERNEL */
1181 
1182 #endif /* _NETINET_TCP_VAR_H_ */
1183