xref: /freebsd/sys/netinet/tcp_var.h (revision 9e5787d2284e187abb5b654d924394a65772e004)
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 <net/vnet.h>
43 #include <sys/mbuf.h>
44 #endif
45 
46 #define TCP_END_BYTE_INFO 8	/* Bytes that makeup the "end information array" */
47 /* Types of ending byte info */
48 #define TCP_EI_EMPTY_SLOT	0
49 #define TCP_EI_STATUS_CLIENT_FIN	0x1
50 #define TCP_EI_STATUS_CLIENT_RST	0x2
51 #define TCP_EI_STATUS_SERVER_FIN	0x3
52 #define TCP_EI_STATUS_SERVER_RST	0x4
53 #define TCP_EI_STATUS_RETRAN		0x5
54 #define TCP_EI_STATUS_PROGRESS		0x6
55 #define TCP_EI_STATUS_PERSIST_MAX	0x7
56 #define TCP_EI_STATUS_KEEP_MAX		0x8
57 #define TCP_EI_STATUS_DATA_A_CLOSE	0x9
58 #define TCP_EI_STATUS_RST_IN_FRONT	0xa
59 #define TCP_EI_STATUS_2MSL		0xb
60 #define TCP_EI_STATUS_MAX_VALUE		0xb
61 
62 /************************************************/
63 /* Status bits we track to assure no duplicates,
64  * the bits here are not used by the code but
65  * for human representation. To check a bit we
66  * take and shift over by 1 minus the value (1-8).
67  */
68 /************************************************/
69 #define TCP_EI_BITS_CLIENT_FIN	0x001
70 #define TCP_EI_BITS_CLIENT_RST	0x002
71 #define TCP_EI_BITS_SERVER_FIN	0x004
72 #define TCP_EI_BITS_SERVER_RST	0x008
73 #define TCP_EI_BITS_RETRAN	0x010
74 #define TCP_EI_BITS_PROGRESS	0x020
75 #define TCP_EI_BITS_PRESIST_MAX	0x040
76 #define TCP_EI_BITS_KEEP_MAX	0x080
77 #define TCP_EI_BITS_DATA_A_CLO  0x100
78 #define TCP_EI_BITS_RST_IN_FR	0x200	/* a front state reset */
79 #define TCP_EI_BITS_2MS_TIMER	0x400	/* 2 MSL timer expired */
80 
81 #if defined(_KERNEL) || defined(_WANT_TCPCB)
82 /* TCP segment queue entry */
83 struct tseg_qent {
84 	TAILQ_ENTRY(tseg_qent) tqe_q;
85 	struct	mbuf   *tqe_m;		/* mbuf contains packet */
86 	struct  mbuf   *tqe_last;	/* last mbuf in chain */
87 	tcp_seq tqe_start;		/* TCP Sequence number start */
88 	int	tqe_len;		/* TCP segment data length */
89 	uint32_t tqe_flags;		/* The flags from the th->th_flags */
90 	uint32_t tqe_mbuf_cnt;		/* Count of mbuf overhead */
91 };
92 TAILQ_HEAD(tsegqe_head, tseg_qent);
93 
94 struct sackblk {
95 	tcp_seq start;		/* start seq no. of sack block */
96 	tcp_seq end;		/* end seq no. */
97 };
98 
99 struct sackhole {
100 	tcp_seq start;		/* start seq no. of hole */
101 	tcp_seq end;		/* end seq no. */
102 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
103 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
104 };
105 
106 struct sackhint {
107 	struct sackhole	*nexthole;
108 	int32_t		sack_bytes_rexmit;
109 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
110 
111 	int32_t		delivered_data; /* Newly acked data from last SACK */
112 
113 	int32_t		sacked_bytes;	/* Total sacked bytes reported by the
114 					 * receiver via sack option
115 					 */
116 	uint32_t	_pad1[1];	/* TBD */
117 	uint64_t	_pad[1];	/* TBD */
118 };
119 
120 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
121 
122 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
123 
124 /*
125  * Tcp control block, one per tcp; fields:
126  * Organized for 64 byte cacheline efficiency based
127  * on common tcp_input/tcp_output processing.
128  */
129 struct tcpcb {
130 	/* Cache line 1 */
131 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
132 	struct tcp_function_block *t_fb;/* TCP function call block */
133 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
134 	uint32_t t_maxseg:24,		/* maximum segment size */
135 		t_logstate:8;		/* State of "black box" logging */
136 	uint32_t t_port:16,		/* Tunneling (over udp) port */
137 		t_state:4,		/* state of this connection */
138 		t_idle_reduce : 1,
139 		t_delayed_ack: 7,	/* Delayed ack variable */
140 		t_fin_is_rst: 1,	/* Are fin's treated as resets */
141 		t_log_state_set: 1,
142 		bits_spare : 2;
143 	u_int	t_flags;
144 	tcp_seq	snd_una;		/* sent but unacknowledged */
145 	tcp_seq	snd_max;		/* highest sequence number sent;
146 					 * used to recognize retransmits
147 					 */
148 	tcp_seq	snd_nxt;		/* send next */
149 	tcp_seq	snd_up;			/* send urgent pointer */
150 	uint32_t  snd_wnd;		/* send window */
151 	uint32_t  snd_cwnd;		/* congestion-controlled window */
152 	uint32_t t_peakrate_thr; 	/* pre-calculated peak rate threshold */
153 	/* Cache line 2 */
154 	u_int32_t  ts_offset;		/* our timestamp offset */
155 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
156 	int	rcv_numsacks;		/* # distinct sack blks present */
157 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
158 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
159 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
160 	tcp_seq	rcv_nxt;		/* receive next */
161 	tcp_seq	rcv_adv;		/* advertised window */
162 	uint32_t  rcv_wnd;		/* receive window */
163 	u_int	t_flags2;		/* More tcpcb flags storage */
164 	int	t_srtt;			/* smoothed round-trip time */
165 	int	t_rttvar;		/* variance in round-trip time */
166 	u_int32_t  ts_recent;		/* timestamp echo data */
167 	u_char	snd_scale;		/* window scaling for send window */
168 	u_char	rcv_scale;		/* window scaling for recv window */
169 	u_char	snd_limited;		/* segments limited transmitted */
170 	u_char	request_r_scale;	/* pending window scaling */
171 	tcp_seq	last_ack_sent;
172 	u_int	t_rcvtime;		/* inactivity time */
173 	/* Cache line 3 */
174 	tcp_seq	rcv_up;			/* receive urgent pointer */
175 	int	t_segqlen;		/* segment reassembly queue length */
176 	uint32_t t_segqmbuflen;		/* Count of bytes mbufs on all entries */
177 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
178 	struct mbuf      *t_in_pkt;
179 	struct mbuf	 *t_tail_pkt;
180 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
181 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
182 	uint32_t  snd_ssthresh;		/* snd_cwnd size threshold for
183 					 * for slow start exponential to
184 					 * linear switch
185 					 */
186 	tcp_seq	snd_wl1;		/* window update seg seq number */
187 	/* Cache line 4 */
188 	tcp_seq	snd_wl2;		/* window update seg ack number */
189 
190 	tcp_seq	irs;			/* initial receive sequence number */
191 	tcp_seq	iss;			/* initial send sequence number */
192 	u_int	t_acktime;		/* RACK and BBR incoming new data was acked */
193 	u_int	t_sndtime;		/* time last data was sent */
194 	u_int	ts_recent_age;		/* when last updated */
195 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
196 	uint16_t cl4_spare;		/* Spare to adjust CL 4 */
197 	char	t_oobflags;		/* have some */
198 	char	t_iobc;			/* input character */
199 	int	t_rxtcur;		/* current retransmit value (ticks) */
200 
201 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
202 	u_int	t_rtttime;		/* RTT measurement start time */
203 
204 	tcp_seq	t_rtseq;		/* sequence number being timed */
205 	u_int	t_starttime;		/* time connection was established */
206 	u_int	t_fbyte_in;		/* ticks time when first byte queued in */
207 	u_int	t_fbyte_out;		/* ticks time when first byte queued out */
208 
209 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
210 	int	t_blackhole_enter;	/* when to enter blackhole detection */
211 	int	t_blackhole_exit;	/* when to exit blackhole detection */
212 	u_int	t_rttmin;		/* minimum rtt allowed */
213 
214 	u_int	t_rttbest;		/* best rtt we've seen */
215 
216 	int	t_softerror;		/* possible error not yet reported */
217 	uint32_t  max_sndwnd;		/* largest window peer has offered */
218 	/* Cache line 5 */
219 	uint32_t  snd_cwnd_prev;	/* cwnd prior to retransmit */
220 	uint32_t  snd_ssthresh_prev;	/* ssthresh prior to retransmit */
221 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
222 	int	t_sndzerowin;		/* zero-window updates sent */
223 	u_long	t_rttupdated;		/* number of times rtt sampled */
224 	int	snd_numholes;		/* number of holes seen by sender */
225 	u_int	t_badrxtwin;		/* window for retransmit recovery */
226 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
227 					/* SACK scoreboard (sorted) */
228 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
229 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
230 	struct sackhint	sackhint;	/* SACK scoreboard hint */
231 	int	t_rttlow;		/* smallest observerved RTT */
232 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
233 	struct toedev	*tod;		/* toedev handling this connection */
234 	int	t_sndrexmitpack;	/* retransmit packets sent */
235 	int	t_rcvoopack;		/* out-of-order packets received */
236 	void	*t_toe;			/* TOE pcb pointer */
237 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
238 	struct cc_var	*ccv;		/* congestion control specific vars */
239 	struct osd	*osd;		/* storage for Khelp module data */
240 	int	t_bytes_acked;		/* # bytes acked during current RTT */
241 	u_int   t_maxunacktime;
242 	u_int	t_keepinit;		/* time to establish connection */
243 	u_int	t_keepidle;		/* time before keepalive probes begin */
244 	u_int	t_keepintvl;		/* interval between keepalives */
245 	u_int	t_keepcnt;		/* number of keepalives before close */
246 	int	t_dupacks;		/* consecutive dup acks recd */
247 	int	t_lognum;		/* Number of log entries */
248 	int	t_loglimit;		/* Maximum number of log entries */
249 	struct tcp_log_stailq t_logs;	/* Log buffer */
250 	struct tcp_log_id_node *t_lin;
251 	struct tcp_log_id_bucket *t_lib;
252 	const char *t_output_caller;	/* Function that called tcp_output */
253 	struct statsblob *t_stats;	/* Per-connection stats */
254 	uint32_t t_logsn;		/* Log "serial number" */
255 	uint32_t gput_ts;		/* Time goodput measurement started */
256 	tcp_seq gput_seq;		/* Outbound measurement seq */
257 	tcp_seq gput_ack;		/* Inbound measurement ack */
258 	int32_t t_stats_gput_prev;	/* XXXLAS: Prev gput measurement */
259 	uint8_t t_tfo_client_cookie_len; /* TCP Fast Open client cookie length */
260 	uint32_t t_end_info_status;	/* Status flag of end info */
261 	unsigned int *t_tfo_pending;	/* TCP Fast Open server pending counter */
262 	union {
263 		uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
264 		uint64_t server;
265 	} t_tfo_cookie;			/* TCP Fast Open cookie to send */
266 	union {
267 		uint8_t t_end_info_bytes[TCP_END_BYTE_INFO];
268 		uint64_t t_end_info;
269 	};
270 #ifdef TCPPCAP
271 	struct mbufq t_inpkts;		/* List of saved input packets. */
272 	struct mbufq t_outpkts;		/* List of saved output packets. */
273 #endif
274 };
275 #endif	/* _KERNEL || _WANT_TCPCB */
276 
277 #ifdef _KERNEL
278 struct tcptemp {
279 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
280 	struct	tcphdr tt_t;
281 };
282 
283 /* Minimum map entries limit value, if set */
284 #define TCP_MIN_MAP_ENTRIES_LIMIT	128
285 
286 /*
287  * TODO: We yet need to brave plowing in
288  * to tcp_input() and the pru_usrreq() block.
289  * Right now these go to the old standards which
290  * are somewhat ok, but in the long term may
291  * need to be changed. If we do tackle tcp_input()
292  * then we need to get rid of the tcp_do_segment()
293  * function below.
294  */
295 /* Flags for tcp functions */
296 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
297 
298 /*
299  * If defining the optional tcp_timers, in the
300  * tfb_tcp_timer_stop call you must use the
301  * callout_async_drain() function with the
302  * tcp_timer_discard callback. You should check
303  * the return of callout_async_drain() and if 0
304  * increment tt_draincnt. Since the timer sub-system
305  * does not know your callbacks you must provide a
306  * stop_all function that loops through and calls
307  * tcp_timer_stop() with each of your defined timers.
308  * Adding a tfb_tcp_handoff_ok function allows the socket
309  * option to change stacks to query you even if the
310  * connection is in a later stage. You return 0 to
311  * say you can take over and run your stack, you return
312  * non-zero (an error number) to say no you can't.
313  * If the function is undefined you can only change
314  * in the early states (before connect or listen).
315  * tfb_tcp_fb_fini is changed to add a flag to tell
316  * the old stack if the tcb is being destroyed or
317  * not. A one in the flag means the TCB is being
318  * destroyed, a zero indicates its transitioning to
319  * another stack (via socket option).
320  */
321 struct tcp_function_block {
322 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
323 	int	(*tfb_tcp_output)(struct tcpcb *);
324 	int	(*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
325 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
326 			    struct socket *, struct tcpcb *,
327 		        int, int, uint8_t);
328 	int     (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int);
329 	int      (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *,
330 			    struct socket *, struct tcpcb *,
331 			    int, int, uint8_t,
332 			    int, struct timeval *);
333 	void	(*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
334 			    struct socket *, struct tcpcb *,
335 			    int, int, uint8_t,
336 			    int, struct timeval *);
337 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
338 			    struct inpcb *inp, struct tcpcb *tp);
339 	/* Optional memory allocation/free routine */
340 	int	(*tfb_tcp_fb_init)(struct tcpcb *);
341 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
342 	/* Optional timers, must define all if you define one */
343 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
344 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
345 			    uint32_t, u_int);
346 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
347 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
348 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
349 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
350 	void	(*tfb_tcp_mtu_chg)(struct tcpcb *);
351 	int	(*tfb_pru_options)(struct tcpcb *, int);
352 	volatile uint32_t tfb_refcnt;
353 	uint32_t  tfb_flags;
354 	uint8_t	tfb_id;
355 };
356 
357 struct tcp_function {
358 	TAILQ_ENTRY(tcp_function)	tf_next;
359 	char				tf_name[TCP_FUNCTION_NAME_LEN_MAX];
360 	struct tcp_function_block	*tf_fb;
361 };
362 
363 TAILQ_HEAD(tcp_funchead, tcp_function);
364 #endif	/* _KERNEL */
365 
366 /*
367  * Flags and utility macros for the t_flags field.
368  */
369 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
370 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
371 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
372 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
373 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
374 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
375 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
376 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
377 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
378 #define	TF_SACK_PERMIT	0x00000200	/* other side said I could SACK */
379 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
380 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
381 #define	TF_NOPUSH	0x00001000	/* don't push */
382 #define	TF_PREVVALID	0x00002000	/* saved values for bad rxmit valid */
383 #define	TF_UNUSED1	0x00004000	/* unused */
384 #define	TF_GPUTINPROG	0x00008000	/* Goodput measurement in progress */
385 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
386 #define	TF_LQ_OVERFLOW	0x00020000	/* listen queue overflow */
387 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
388 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
389 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
390 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
391 #define	TF_SIGNATURE	0x00400000	/* require MD5 digests (RFC2385) */
392 #define	TF_FORCEDATA	0x00800000	/* force out a byte */
393 #define	TF_TSO		0x01000000	/* TSO enabled on this connection */
394 #define	TF_TOE		0x02000000	/* this connection is offloaded */
395 #define	TF_UNUSED3	0x04000000	/* unused */
396 #define	TF_UNUSED4	0x08000000	/* unused */
397 #define	TF_UNUSED5	0x10000000	/* unused */
398 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
399 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
400 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
401 
402 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
403 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
404 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
405 
406 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
407 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
408 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
409 
410 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
411 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
412 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
413 
414 #if defined(_KERNEL) && !defined(TCP_RFC7413)
415 #define	IS_FASTOPEN(t_flags)		(false)
416 #else
417 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
418 #endif
419 
420 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
421 
422 /*
423  * Flags for the t_oobflags field.
424  */
425 #define	TCPOOB_HAVEDATA	0x01
426 #define	TCPOOB_HADDATA	0x02
427 
428 /*
429  * Flags for the extended TCP flags field, t_flags2
430  */
431 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
432 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
433 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
434 #define	TF2_LOG_AUTO		0x00000008 /* Session is auto-logging. */
435 #define TF2_DROP_AF_DATA 	0x00000010 /* Drop after all data ack'd */
436 #define	TF2_ECN_PERMIT		0x00000020 /* connection ECN-ready */
437 #define	TF2_ECN_SND_CWR		0x00000040 /* ECN CWR in queue */
438 #define	TF2_ECN_SND_ECE		0x00000080 /* ECN ECE in queue */
439 #define	TF2_ACE_PERMIT		0x00000100 /* Accurate ECN mode */
440 #define TF2_FBYTES_COMPLETE	0x00000400 /* We have first bytes in and out */
441 /*
442  * Structure to hold TCP options that are only used during segment
443  * processing (in tcp_input), but not held in the tcpcb.
444  * It's basically used to reduce the number of parameters
445  * to tcp_dooptions and tcp_addoptions.
446  * The binary order of the to_flags is relevant for packing of the
447  * options in tcp_addoptions.
448  */
449 struct tcpopt {
450 	u_int32_t	to_flags;	/* which options are present */
451 #define	TOF_MSS		0x0001		/* maximum segment size */
452 #define	TOF_SCALE	0x0002		/* window scaling */
453 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
454 #define	TOF_TS		0x0010		/* timestamp */
455 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
456 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
457 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
458 #define	TOF_MAXOPT	0x0200
459 	u_int32_t	to_tsval;	/* new timestamp */
460 	u_int32_t	to_tsecr;	/* reflected timestamp */
461 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
462 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
463 	u_int8_t	*to_tfo_cookie; /* pointer to the TFO cookie */
464 	u_int16_t	to_mss;		/* maximum segment size */
465 	u_int8_t	to_wscale;	/* window scaling */
466 	u_int8_t	to_nsacks;	/* number of SACK blocks */
467 	u_int8_t	to_tfo_len;	/* TFO cookie length */
468 	u_int32_t	to_spare;	/* UTO */
469 };
470 
471 /*
472  * Flags for tcp_dooptions.
473  */
474 #define	TO_SYN		0x01		/* parse SYN-only options */
475 
476 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
477 	uint32_t	rmx_mtu;	/* MTU for this path */
478 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
479 	uint32_t	rmx_rtt;	/* estimated round trip time */
480 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
481 	uint32_t	rmx_cwnd;	/* congestion window */
482 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
483 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
484 };
485 
486 /*
487  * Used by tcp_maxmtu() to communicate interface specific features
488  * and limits at the time of connection setup.
489  */
490 struct tcp_ifcap {
491 	int	ifcap;
492 	u_int	tsomax;
493 	u_int	tsomaxsegcount;
494 	u_int	tsomaxsegsize;
495 };
496 
497 #ifndef _NETINET_IN_PCB_H_
498 struct in_conninfo;
499 #endif /* _NETINET_IN_PCB_H_ */
500 
501 struct tcptw {
502 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
503 	tcp_seq		snd_nxt;
504 	tcp_seq		rcv_nxt;
505 	tcp_seq		iss;
506 	tcp_seq		irs;
507 	u_short		last_win;	/* cached window value */
508 	short		tw_so_options;	/* copy of so_options */
509 	struct ucred	*tw_cred;	/* user credentials */
510 	u_int32_t	t_recent;
511 	u_int32_t	ts_offset;	/* our timestamp offset */
512 	u_int		t_starttime;
513 	int		tw_time;
514 	TAILQ_ENTRY(tcptw) tw_2msl;
515 	void		*tw_pspare;	/* TCP_SIGNATURE */
516 	u_int		*tw_spare;	/* TCP_SIGNATURE */
517 };
518 
519 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
520 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
521 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
522 
523 /*
524  * The smoothed round-trip time and estimated variance
525  * are stored as fixed point numbers scaled by the values below.
526  * For convenience, these scales are also used in smoothing the average
527  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
528  * With these scales, srtt has 3 bits to the right of the binary point,
529  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
530  * binary point, and is smoothed with an ALPHA of 0.75.
531  */
532 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
533 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
534 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
535 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
536 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
537 
538 /*
539  * The initial retransmission should happen at rtt + 4 * rttvar.
540  * Because of the way we do the smoothing, srtt and rttvar
541  * will each average +1/2 tick of bias.  When we compute
542  * the retransmit timer, we want 1/2 tick of rounding and
543  * 1 extra tick because of +-1/2 tick uncertainty in the
544  * firing of the timer.  The bias will give us exactly the
545  * 1.5 tick we need.  But, because the bias is
546  * statistical, we have to test that we don't drop below
547  * the minimum feasible timer (which is 2 ticks).
548  * This version of the macro adapted from a paper by Lawrence
549  * Brakmo and Larry Peterson which outlines a problem caused
550  * by insufficient precision in the original implementation,
551  * which results in inappropriately large RTO values for very
552  * fast networks.
553  */
554 #define	TCP_REXMTVAL(tp) \
555 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
556 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
557 
558 /*
559  * TCP statistics.
560  * Many of these should be kept per connection,
561  * but that's inconvenient at the moment.
562  */
563 struct	tcpstat {
564 	uint64_t tcps_connattempt;	/* connections initiated */
565 	uint64_t tcps_accepts;		/* connections accepted */
566 	uint64_t tcps_connects;		/* connections established */
567 	uint64_t tcps_drops;		/* connections dropped */
568 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
569 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
570 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
571 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
572 	uint64_t tcps_rttupdated;	/* times we succeeded */
573 	uint64_t tcps_delack;		/* delayed acks sent */
574 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
575 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
576 	uint64_t tcps_persisttimeo;	/* persist timeouts */
577 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
578 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
579 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
580 
581 	uint64_t tcps_sndtotal;		/* total packets sent */
582 	uint64_t tcps_sndpack;		/* data packets sent */
583 	uint64_t tcps_sndbyte;		/* data bytes sent */
584 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
585 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
586 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
587 	uint64_t tcps_sndacks;		/* ack-only packets sent */
588 	uint64_t tcps_sndprobe;		/* window probes sent */
589 	uint64_t tcps_sndurg;		/* packets sent with URG only */
590 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
591 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
592 
593 	uint64_t tcps_rcvtotal;		/* total packets received */
594 	uint64_t tcps_rcvpack;		/* packets received in sequence */
595 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
596 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
597 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
598 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
599 	uint64_t tcps_rcvshort;		/* packets received too short */
600 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
601 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
602 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
603 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
604 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
605 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
606 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
607 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
608 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
609 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
610 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
611 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
612 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
613 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
614 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
615 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
616 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
617 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
618 	uint64_t tcps_pcbcachemiss;
619 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
620 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
621 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
622 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
623 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
624 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
625 	uint64_t tcps_persistdrop;	/* timeout in persist state */
626 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
627 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
628 	uint64_t tcps_listendrop;	/* listen queue overflows */
629 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
630 
631 	uint64_t tcps_sc_added;		/* entry added to syncache */
632 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
633 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
634 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
635 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
636 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
637 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
638 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
639 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
640 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
641 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
642 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
643 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
644 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
645 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
646 
647 	uint64_t tcps_hc_added;		/* entry added to hostcache */
648 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
649 
650 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
651 
652 	/* SACK related stats */
653 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
654 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
655 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
656 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
657 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
658 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
659 
660 	/* ECN related stats */
661 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
662 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
663 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
664 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
665 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
666 
667 	/* TCP_SIGNATURE related stats */
668 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
669 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
670 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
671 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
672 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
673 
674 	/* Path MTU Discovery Black Hole Detection related stats */
675 	uint64_t tcps_pmtud_blackhole_activated;	 /* Black Hole Count */
676 	uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
677 	uint64_t tcps_pmtud_blackhole_failed;		 /* Black Hole Failure Count */
678 
679 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
680 };
681 
682 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
683 
684 #ifdef _KERNEL
685 #define	TI_UNLOCKED	1
686 #define	TI_RLOCKED	2
687 #include <sys/counter.h>
688 
689 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
690 /*
691  * In-kernel consumers can use these accessor macros directly to update
692  * stats.
693  */
694 #define	TCPSTAT_ADD(name, val)	\
695     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
696 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
697 
698 /*
699  * Kernel module consumers must use this accessor macro.
700  */
701 void	kmod_tcpstat_add(int statnum, int val);
702 #define	KMOD_TCPSTAT_ADD(name, val)					\
703     kmod_tcpstat_add(offsetof(struct tcpstat, name) / sizeof(uint64_t), val)
704 #define	KMOD_TCPSTAT_INC(name)	KMOD_TCPSTAT_ADD(name, 1)
705 
706 /*
707  * Running TCP connection count by state.
708  */
709 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
710 #define	V_tcps_states	VNET(tcps_states)
711 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
712 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
713 
714 /*
715  * TCP specific helper hook point identifiers.
716  */
717 #define	HHOOK_TCP_EST_IN		0
718 #define	HHOOK_TCP_EST_OUT		1
719 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
720 
721 struct tcp_hhook_data {
722 	struct tcpcb	*tp;
723 	struct tcphdr	*th;
724 	struct tcpopt	*to;
725 	uint32_t	len;
726 	int		tso;
727 	tcp_seq		curack;
728 };
729 #ifdef TCP_HHOOK
730 void hhook_run_tcp_est_out(struct tcpcb *tp,
731 	struct tcphdr *th, struct tcpopt *to,
732 	uint32_t len, int tso);
733 #endif
734 #endif
735 
736 /*
737  * TCB structure exported to user-land via sysctl(3).
738  *
739  * Fields prefixed with "xt_" are unique to the export structure, and fields
740  * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
741  *
742  * Legend:
743  * (s) - used by userland utilities in src
744  * (p) - used by utilities in ports
745  * (3) - is known to be used by third party software not in ports
746  * (n) - no known usage
747  *
748  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
749  * included.  Not all of our clients do.
750  */
751 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
752 struct xtcpcb {
753 	ksize_t	xt_len;		/* length of this structure */
754 	struct xinpcb	xt_inp;
755 	char		xt_stack[TCP_FUNCTION_NAME_LEN_MAX];	/* (s) */
756 	char		xt_logid[TCP_LOG_ID_LEN];	/* (s) */
757 	int64_t		spare64[8];
758 	int32_t		t_state;		/* (s,p) */
759 	uint32_t	t_flags;		/* (s,p) */
760 	int32_t		t_sndzerowin;		/* (s) */
761 	int32_t		t_sndrexmitpack;	/* (s) */
762 	int32_t		t_rcvoopack;		/* (s) */
763 	int32_t		t_rcvtime;		/* (s) */
764 	int32_t		tt_rexmt;		/* (s) */
765 	int32_t		tt_persist;		/* (s) */
766 	int32_t		tt_keep;		/* (s) */
767 	int32_t		tt_2msl;		/* (s) */
768 	int32_t		tt_delack;		/* (s) */
769 	int32_t		t_logstate;		/* (3) */
770 	int32_t		spare32[32];
771 } __aligned(8);
772 
773 #ifdef _KERNEL
774 void	tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
775 #endif
776 #endif
777 
778 /*
779  * TCP function information (name-to-id mapping, aliases, and refcnt)
780  * exported to user-land via sysctl(3).
781  */
782 struct tcp_function_info {
783 	uint32_t	tfi_refcnt;
784 	uint8_t		tfi_id;
785 	char		tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
786 	char		tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
787 };
788 
789 /*
790  * Identifiers for TCP sysctl nodes
791  */
792 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
793 #define	TCPCTL_MSSDFLT		3	/* MSS default */
794 #define TCPCTL_STATS		4	/* statistics */
795 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
796 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
797 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
798 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
799 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
800 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
801 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
802 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
803 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
804 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
805 #define	TCPCTL_DROP		15	/* drop tcp connection */
806 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
807 
808 #ifdef _KERNEL
809 #ifdef SYSCTL_DECL
810 SYSCTL_DECL(_net_inet_tcp);
811 SYSCTL_DECL(_net_inet_tcp_sack);
812 MALLOC_DECLARE(M_TCPLOG);
813 #endif
814 
815 VNET_DECLARE(int, tcp_log_in_vain);
816 #define	V_tcp_log_in_vain		VNET(tcp_log_in_vain)
817 
818 /*
819  * Global TCP tunables shared between different stacks.
820  * Please keep the list sorted.
821  */
822 VNET_DECLARE(int, drop_synfin);
823 VNET_DECLARE(int, path_mtu_discovery);
824 VNET_DECLARE(int, tcp_abc_l_var);
825 VNET_DECLARE(int, tcp_autorcvbuf_max);
826 VNET_DECLARE(int, tcp_autosndbuf_inc);
827 VNET_DECLARE(int, tcp_autosndbuf_max);
828 VNET_DECLARE(int, tcp_delack_enabled);
829 VNET_DECLARE(int, tcp_do_autorcvbuf);
830 VNET_DECLARE(int, tcp_do_autosndbuf);
831 VNET_DECLARE(int, tcp_do_ecn);
832 VNET_DECLARE(int, tcp_do_newcwv);
833 VNET_DECLARE(int, tcp_do_rfc1323);
834 VNET_DECLARE(int, tcp_do_rfc3042);
835 VNET_DECLARE(int, tcp_do_rfc3390);
836 VNET_DECLARE(int, tcp_do_rfc3465);
837 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
838 VNET_DECLARE(int, tcp_do_sack);
839 VNET_DECLARE(int, tcp_do_tso);
840 VNET_DECLARE(int, tcp_ecn_maxretries);
841 VNET_DECLARE(int, tcp_initcwnd_segments);
842 VNET_DECLARE(int, tcp_insecure_rst);
843 VNET_DECLARE(int, tcp_insecure_syn);
844 VNET_DECLARE(uint32_t, tcp_map_entries_limit);
845 VNET_DECLARE(uint32_t, tcp_map_split_limit);
846 VNET_DECLARE(int, tcp_minmss);
847 VNET_DECLARE(int, tcp_mssdflt);
848 #ifdef STATS
849 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl);
850 VNET_DECLARE(int, tcp_perconn_stats_enable);
851 #endif /* STATS */
852 VNET_DECLARE(int, tcp_recvspace);
853 VNET_DECLARE(int, tcp_sack_globalholes);
854 VNET_DECLARE(int, tcp_sack_globalmaxholes);
855 VNET_DECLARE(int, tcp_sack_maxholes);
856 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
857 VNET_DECLARE(int, tcp_sendspace);
858 VNET_DECLARE(struct inpcbhead, tcb);
859 VNET_DECLARE(struct inpcbinfo, tcbinfo);
860 
861 #define	V_tcp_do_newcwv			VNET(tcp_do_newcwv)
862 #define	V_drop_synfin			VNET(drop_synfin)
863 #define	V_path_mtu_discovery		VNET(path_mtu_discovery)
864 #define	V_tcb				VNET(tcb)
865 #define	V_tcbinfo			VNET(tcbinfo)
866 #define	V_tcp_abc_l_var			VNET(tcp_abc_l_var)
867 #define	V_tcp_autorcvbuf_max		VNET(tcp_autorcvbuf_max)
868 #define	V_tcp_autosndbuf_inc		VNET(tcp_autosndbuf_inc)
869 #define	V_tcp_autosndbuf_max		VNET(tcp_autosndbuf_max)
870 #define	V_tcp_delack_enabled		VNET(tcp_delack_enabled)
871 #define	V_tcp_do_autorcvbuf		VNET(tcp_do_autorcvbuf)
872 #define	V_tcp_do_autosndbuf		VNET(tcp_do_autosndbuf)
873 #define	V_tcp_do_ecn			VNET(tcp_do_ecn)
874 #define	V_tcp_do_rfc1323		VNET(tcp_do_rfc1323)
875 #define V_tcp_ts_offset_per_conn	VNET(tcp_ts_offset_per_conn)
876 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
877 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)
878 #define	V_tcp_do_rfc3465		VNET(tcp_do_rfc3465)
879 #define	V_tcp_do_rfc6675_pipe		VNET(tcp_do_rfc6675_pipe)
880 #define	V_tcp_do_sack			VNET(tcp_do_sack)
881 #define	V_tcp_do_tso			VNET(tcp_do_tso)
882 #define	V_tcp_ecn_maxretries		VNET(tcp_ecn_maxretries)
883 #define	V_tcp_initcwnd_segments		VNET(tcp_initcwnd_segments)
884 #define	V_tcp_insecure_rst		VNET(tcp_insecure_rst)
885 #define	V_tcp_insecure_syn		VNET(tcp_insecure_syn)
886 #define	V_tcp_map_entries_limit		VNET(tcp_map_entries_limit)
887 #define	V_tcp_map_split_limit		VNET(tcp_map_split_limit)
888 #define	V_tcp_minmss			VNET(tcp_minmss)
889 #define	V_tcp_mssdflt			VNET(tcp_mssdflt)
890 #ifdef STATS
891 #define	V_tcp_perconn_stats_dflt_tpl	VNET(tcp_perconn_stats_dflt_tpl)
892 #define	V_tcp_perconn_stats_enable	VNET(tcp_perconn_stats_enable)
893 #endif /* STATS */
894 #define	V_tcp_recvspace			VNET(tcp_recvspace)
895 #define	V_tcp_sack_globalholes		VNET(tcp_sack_globalholes)
896 #define	V_tcp_sack_globalmaxholes	VNET(tcp_sack_globalmaxholes)
897 #define	V_tcp_sack_maxholes		VNET(tcp_sack_maxholes)
898 #define	V_tcp_sc_rst_sock_fail		VNET(tcp_sc_rst_sock_fail)
899 #define	V_tcp_sendspace			VNET(tcp_sendspace)
900 #define	V_tcp_udp_tunneling_overhead	VNET(tcp_udp_tunneling_overhead)
901 #define	V_tcp_udp_tunneling_port	VNET(tcp_udp_tunneling_port)
902 
903 #ifdef TCP_HHOOK
904 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
905 #define	V_tcp_hhh		VNET(tcp_hhh)
906 #endif
907 
908 int	 tcp_addoptions(struct tcpopt *, u_char *);
909 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
910 struct tcpcb *
911 	 tcp_close(struct tcpcb *);
912 void	 tcp_discardcb(struct tcpcb *);
913 void	 tcp_twstart(struct tcpcb *);
914 void	 tcp_twclose(struct tcptw *, int);
915 void	 tcp_ctlinput(int, struct sockaddr *, void *);
916 int	 tcp_ctloutput(struct socket *, struct sockopt *);
917 struct tcpcb *
918 	 tcp_drop(struct tcpcb *, int);
919 void	 tcp_drain(void);
920 void	 tcp_init(void);
921 void	 tcp_fini(void *);
922 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
923 	    const void *);
924 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
925 	    const void *);
926 int	 tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *, struct mbuf *);
927 void	 tcp_reass_global_init(void);
928 void	 tcp_reass_flush(struct tcpcb *);
929 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
930 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
931 		     struct tcpcb *, int, int);
932 void	tcp_pulloutofband(struct socket *,
933 		     struct tcphdr *, struct mbuf *, int);
934 void	tcp_xmit_timer(struct tcpcb *, int);
935 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
936 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
937 			    uint16_t nsegs, uint16_t type);
938 void 	cc_conn_init(struct tcpcb *tp);
939 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
940 void    cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos);
941 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
942 #ifdef TCP_HHOOK
943 void	hhook_run_tcp_est_in(struct tcpcb *tp,
944 			    struct tcphdr *th, struct tcpopt *to);
945 #endif
946 
947 int	 tcp_input(struct mbuf **, int *, int);
948 int	 tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
949 	    struct tcpcb *, int);
950 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
951 			struct socket *, struct tcpcb *, int, int, uint8_t);
952 
953 int register_tcp_functions(struct tcp_function_block *blk, int wait);
954 int register_tcp_functions_as_names(struct tcp_function_block *blk,
955     int wait, const char *names[], int *num_names);
956 int register_tcp_functions_as_name(struct tcp_function_block *blk,
957     const char *name, int wait);
958 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
959     bool force);
960 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
961 void tcp_switch_back_to_default(struct tcpcb *tp);
962 struct tcp_function_block *
963 find_and_ref_tcp_fb(struct tcp_function_block *fs);
964 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
965 
966 extern counter_u64_t tcp_inp_lro_direct_queue;
967 extern counter_u64_t tcp_inp_lro_wokeup_queue;
968 extern counter_u64_t tcp_inp_lro_compressed;
969 extern counter_u64_t tcp_inp_lro_single_push;
970 extern counter_u64_t tcp_inp_lro_locks_taken;
971 extern counter_u64_t tcp_inp_lro_sack_wake;
972 
973 #ifdef NETFLIX_EXP_DETECTION
974 /* Various SACK attack thresholds */
975 extern int32_t tcp_force_detection;
976 extern int32_t tcp_sack_to_ack_thresh;
977 extern int32_t tcp_sack_to_move_thresh;
978 extern int32_t tcp_restoral_thresh;
979 extern int32_t tcp_sad_decay_val;
980 extern int32_t tcp_sad_pacing_interval;
981 extern int32_t tcp_sad_low_pps;
982 extern int32_t tcp_map_minimum;
983 extern int32_t tcp_attack_on_turns_on_logging;
984 #endif
985 
986 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
987 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
988 u_int	 tcp_maxseg(const struct tcpcb *);
989 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
990 	    struct tcp_ifcap *);
991 void	 tcp_mss(struct tcpcb *, int);
992 int	 tcp_mssopt(struct in_conninfo *);
993 struct inpcb *
994 	 tcp_drop_syn_sent(struct inpcb *, int);
995 struct tcpcb *
996 	 tcp_newtcpcb(struct inpcb *);
997 int	 tcp_output(struct tcpcb *);
998 void	 tcp_state_change(struct tcpcb *, int);
999 void	 tcp_respond(struct tcpcb *, void *,
1000 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
1001 void	 tcp_tw_init(void);
1002 #ifdef VIMAGE
1003 void	 tcp_tw_destroy(void);
1004 #endif
1005 void	 tcp_tw_zone_change(void);
1006 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
1007 	    struct mbuf *, int);
1008 void	 tcp_setpersist(struct tcpcb *);
1009 void	 tcp_slowtimo(void);
1010 struct tcptemp *
1011 	 tcpip_maketemplate(struct inpcb *);
1012 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
1013 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
1014 int	 tcp_timer_suspend(struct tcpcb *, uint32_t);
1015 void	 tcp_timers_unsuspend(struct tcpcb *, uint32_t);
1016 int	 tcp_timer_active(struct tcpcb *, uint32_t);
1017 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
1018 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
1019 int	 inp_to_cpuid(struct inpcb *inp);
1020 /*
1021  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
1022  */
1023 void	 tcp_hc_init(void);
1024 #ifdef VIMAGE
1025 void	 tcp_hc_destroy(void);
1026 #endif
1027 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
1028 uint32_t tcp_hc_getmtu(struct in_conninfo *);
1029 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
1030 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
1031 
1032 extern	struct pr_usrreqs tcp_usrreqs;
1033 
1034 uint32_t tcp_new_ts_offset(struct in_conninfo *);
1035 tcp_seq	 tcp_new_isn(struct in_conninfo *);
1036 
1037 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
1038 void	 tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
1039 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1040 void	 tcp_clean_dsack_blocks(struct tcpcb *tp);
1041 void	 tcp_clean_sackreport(struct tcpcb *tp);
1042 void	 tcp_sack_adjust(struct tcpcb *tp);
1043 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1044 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1045 void	 tcp_free_sackholes(struct tcpcb *tp);
1046 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
1047 int	 tcp_compute_pipe(struct tcpcb *);
1048 uint32_t tcp_compute_initwnd(uint32_t);
1049 void	 tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
1050 int	 tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes,
1051     size_t seed_len);
1052 struct mbuf *
1053 	 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
1054 	   int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls);
1055 
1056 int	tcp_stats_init(void);
1057 void tcp_log_end_status(struct tcpcb *tp, uint8_t status);
1058 
1059 static inline void
1060 tcp_fields_to_host(struct tcphdr *th)
1061 {
1062 
1063 	th->th_seq = ntohl(th->th_seq);
1064 	th->th_ack = ntohl(th->th_ack);
1065 	th->th_win = ntohs(th->th_win);
1066 	th->th_urp = ntohs(th->th_urp);
1067 }
1068 
1069 static inline void
1070 tcp_fields_to_net(struct tcphdr *th)
1071 {
1072 
1073 	th->th_seq = htonl(th->th_seq);
1074 	th->th_ack = htonl(th->th_ack);
1075 	th->th_win = htons(th->th_win);
1076 	th->th_urp = htons(th->th_urp);
1077 }
1078 #endif /* _KERNEL */
1079 
1080 #endif /* _NETINET_TCP_VAR_H_ */
1081