xref: /freebsd/sys/netinet/tcp_var.h (revision 480093f4440d54b30b3025afeac24b48f2ba7a2e)
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 #if defined(_KERNEL) || defined(_WANT_TCPCB)
47 /* TCP segment queue entry */
48 struct tseg_qent {
49 	TAILQ_ENTRY(tseg_qent) tqe_q;
50 	struct	mbuf   *tqe_m;		/* mbuf contains packet */
51 	struct  mbuf   *tqe_last;	/* last mbuf in chain */
52 	tcp_seq tqe_start;		/* TCP Sequence number start */
53 	int	tqe_len;		/* TCP segment data length */
54 	uint32_t tqe_flags;		/* The flags from the th->th_flags */
55 	uint32_t tqe_mbuf_cnt;		/* Count of mbuf overhead */
56 };
57 TAILQ_HEAD(tsegqe_head, tseg_qent);
58 
59 struct sackblk {
60 	tcp_seq start;		/* start seq no. of sack block */
61 	tcp_seq end;		/* end seq no. */
62 };
63 
64 struct sackhole {
65 	tcp_seq start;		/* start seq no. of hole */
66 	tcp_seq end;		/* end seq no. */
67 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
68 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
69 };
70 
71 struct sackhint {
72 	struct sackhole	*nexthole;
73 	int		sack_bytes_rexmit;
74 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
75 
76 	int		ispare;		/* explicit pad for 64bit alignment */
77 	int             sacked_bytes;	/*
78 					 * Total sacked bytes reported by the
79 					 * receiver via sack option
80 					 */
81 	uint32_t	_pad1[1];	/* TBD */
82 	uint64_t	_pad[1];	/* TBD */
83 };
84 
85 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
86 
87 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
88 
89 /*
90  * Tcp control block, one per tcp; fields:
91  * Organized for 64 byte cacheline efficiency based
92  * on common tcp_input/tcp_output processing.
93  */
94 struct tcpcb {
95 	/* Cache line 1 */
96 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
97 	struct tcp_function_block *t_fb;/* TCP function call block */
98 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
99 	uint32_t t_maxseg:24,		/* maximum segment size */
100 		t_logstate:8;		/* State of "black box" logging */
101 	uint32_t t_port:16,		/* Tunneling (over udp) port */
102 		t_state:4,		/* state of this connection */
103 		t_idle_reduce : 1,
104 		t_delayed_ack: 7,	/* Delayed ack variable */
105 		t_fin_is_rst: 1,	/* Are fin's treated as resets */
106 		t_log_state_set: 1,
107 		bits_spare : 2;
108 	u_int	t_flags;
109 	tcp_seq	snd_una;		/* sent but unacknowledged */
110 	tcp_seq	snd_max;		/* highest sequence number sent;
111 					 * used to recognize retransmits
112 					 */
113 	tcp_seq	snd_nxt;		/* send next */
114 	tcp_seq	snd_up;			/* send urgent pointer */
115 	uint32_t  snd_wnd;		/* send window */
116 	uint32_t  snd_cwnd;		/* congestion-controlled window */
117 	uint32_t t_peakrate_thr; 	/* pre-calculated peak rate threshold */
118 	/* Cache line 2 */
119 	u_int32_t  ts_offset;		/* our timestamp offset */
120 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
121 	int	rcv_numsacks;		/* # distinct sack blks present */
122 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
123 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
124 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
125 	tcp_seq	rcv_nxt;		/* receive next */
126 	tcp_seq	rcv_adv;		/* advertised window */
127 	uint32_t  rcv_wnd;		/* receive window */
128 	u_int	t_flags2;		/* More tcpcb flags storage */
129 	int	t_srtt;			/* smoothed round-trip time */
130 	int	t_rttvar;		/* variance in round-trip time */
131 	u_int32_t  ts_recent;		/* timestamp echo data */
132 	u_char	snd_scale;		/* window scaling for send window */
133 	u_char	rcv_scale;		/* window scaling for recv window */
134 	u_char	snd_limited;		/* segments limited transmitted */
135 	u_char	request_r_scale;	/* pending window scaling */
136 	tcp_seq	last_ack_sent;
137 	u_int	t_rcvtime;		/* inactivity time */
138 	/* Cache line 3 */
139 	tcp_seq	rcv_up;			/* receive urgent pointer */
140 	int	t_segqlen;		/* segment reassembly queue length */
141 	uint32_t t_segqmbuflen;		/* Count of bytes mbufs on all entries */
142 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
143 	struct mbuf      *t_in_pkt;
144 	struct mbuf	 *t_tail_pkt;
145 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
146 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
147 	uint32_t  snd_ssthresh;		/* snd_cwnd size threshold for
148 					 * for slow start exponential to
149 					 * linear switch
150 					 */
151 	tcp_seq	snd_wl1;		/* window update seg seq number */
152 	/* Cache line 4 */
153 	tcp_seq	snd_wl2;		/* window update seg ack number */
154 
155 	tcp_seq	irs;			/* initial receive sequence number */
156 	tcp_seq	iss;		        /* initial send sequence number */
157 	u_int   t_acktime;
158 	u_int	ts_recent_age;		/* when last updated */
159 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
160 	uint16_t cl4_spare;		/* Spare to adjust CL 4 */
161 	char	t_oobflags;		/* have some */
162 	char	t_iobc;			/* input character */
163 	int	t_rxtcur;		/* current retransmit value (ticks) */
164 
165 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
166 	u_int	t_rtttime;		/* RTT measurement start time */
167 
168 	tcp_seq	t_rtseq;		/* sequence number being timed */
169 	u_int	t_starttime;		/* time connection was established */
170 
171 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
172 	u_int	t_rttmin;		/* minimum rtt allowed */
173 
174 	u_int	t_rttbest;		/* best rtt we've seen */
175 
176 	int	t_softerror;		/* possible error not yet reported */
177 	uint32_t  max_sndwnd;		/* largest window peer has offered */
178 	/* Cache line 5 */
179 	uint32_t  snd_cwnd_prev;	/* cwnd prior to retransmit */
180 	uint32_t  snd_ssthresh_prev;	/* ssthresh prior to retransmit */
181 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
182 	int	t_sndzerowin;		/* zero-window updates sent */
183 	u_long	t_rttupdated;		/* number of times rtt sampled */
184 	int	snd_numholes;		/* number of holes seen by sender */
185 	u_int	t_badrxtwin;		/* window for retransmit recovery */
186 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
187 					/* SACK scoreboard (sorted) */
188 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
189 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
190 					   episode starts at this seq number */
191 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
192 	struct sackhint	sackhint;	/* SACK scoreboard hint */
193 	int	t_rttlow;		/* smallest observerved RTT */
194 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
195 	struct toedev	*tod;		/* toedev handling this connection */
196 	int	t_sndrexmitpack;	/* retransmit packets sent */
197 	int	t_rcvoopack;		/* out-of-order packets received */
198 	void	*t_toe;			/* TOE pcb pointer */
199 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
200 	struct cc_var	*ccv;		/* congestion control specific vars */
201 	struct osd	*osd;		/* storage for Khelp module data */
202 	int	t_bytes_acked;		/* # bytes acked during current RTT */
203 	u_int   t_maxunacktime;
204 	u_int	t_keepinit;		/* time to establish connection */
205 	u_int	t_keepidle;		/* time before keepalive probes begin */
206 	u_int	t_keepintvl;		/* interval between keepalives */
207 	u_int	t_keepcnt;		/* number of keepalives before close */
208 	int	t_dupacks;		/* consecutive dup acks recd */
209 	int	t_lognum;		/* Number of log entries */
210 	int	t_loglimit;		/* Maximum number of log entries */
211 	struct tcp_log_stailq t_logs;	/* Log buffer */
212 	struct tcp_log_id_node *t_lin;
213 	struct tcp_log_id_bucket *t_lib;
214 	const char *t_output_caller;	/* Function that called tcp_output */
215 	struct statsblob *t_stats;	/* Per-connection stats */
216 	uint32_t t_logsn;		/* Log "serial number" */
217 	uint32_t gput_ts;		/* Time goodput measurement started */
218 	tcp_seq gput_seq;		/* Outbound measurement seq */
219 	tcp_seq gput_ack;		/* Inbound measurement ack */
220 	int32_t t_stats_gput_prev;	/* XXXLAS: Prev gput measurement */
221 	uint8_t t_tfo_client_cookie_len; /* TCP Fast Open client cookie length */
222 	unsigned int *t_tfo_pending;	/* TCP Fast Open server pending counter */
223 	union {
224 		uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
225 		uint64_t server;
226 	} t_tfo_cookie;			/* TCP Fast Open cookie to send */
227 #ifdef TCPPCAP
228 	struct mbufq t_inpkts;		/* List of saved input packets. */
229 	struct mbufq t_outpkts;		/* List of saved output packets. */
230 #endif
231 };
232 #endif	/* _KERNEL || _WANT_TCPCB */
233 
234 #ifdef _KERNEL
235 struct tcptemp {
236 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
237 	struct	tcphdr tt_t;
238 };
239 
240 /* Minimum map entries limit value, if set */
241 #define TCP_MIN_MAP_ENTRIES_LIMIT	128
242 
243 /*
244  * TODO: We yet need to brave plowing in
245  * to tcp_input() and the pru_usrreq() block.
246  * Right now these go to the old standards which
247  * are somewhat ok, but in the long term may
248  * need to be changed. If we do tackle tcp_input()
249  * then we need to get rid of the tcp_do_segment()
250  * function below.
251  */
252 /* Flags for tcp functions */
253 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
254 
255 /*
256  * If defining the optional tcp_timers, in the
257  * tfb_tcp_timer_stop call you must use the
258  * callout_async_drain() function with the
259  * tcp_timer_discard callback. You should check
260  * the return of callout_async_drain() and if 0
261  * increment tt_draincnt. Since the timer sub-system
262  * does not know your callbacks you must provide a
263  * stop_all function that loops through and calls
264  * tcp_timer_stop() with each of your defined timers.
265  * Adding a tfb_tcp_handoff_ok function allows the socket
266  * option to change stacks to query you even if the
267  * connection is in a later stage. You return 0 to
268  * say you can take over and run your stack, you return
269  * non-zero (an error number) to say no you can't.
270  * If the function is undefined you can only change
271  * in the early states (before connect or listen).
272  * tfb_tcp_fb_fini is changed to add a flag to tell
273  * the old stack if the tcb is being destroyed or
274  * not. A one in the flag means the TCB is being
275  * destroyed, a zero indicates its transitioning to
276  * another stack (via socket option).
277  */
278 struct tcp_function_block {
279 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
280 	int	(*tfb_tcp_output)(struct tcpcb *);
281 	int	(*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
282 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
283 			    struct socket *, struct tcpcb *,
284 		        int, int, uint8_t);
285 	int     (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int);
286 	int      (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *,
287 			    struct socket *, struct tcpcb *,
288 			    int, int, uint8_t,
289 			    int, struct timeval *);
290 	void	(*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
291 			    struct socket *, struct tcpcb *,
292 			    int, int, uint8_t,
293 			    int, struct timeval *);
294 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
295 			    struct inpcb *inp, struct tcpcb *tp);
296 	/* Optional memory allocation/free routine */
297 	int	(*tfb_tcp_fb_init)(struct tcpcb *);
298 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
299 	/* Optional timers, must define all if you define one */
300 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
301 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
302 			    uint32_t, u_int);
303 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
304 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
305 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
306 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
307 	void	(*tfb_tcp_mtu_chg)(struct tcpcb *);
308 	volatile uint32_t tfb_refcnt;
309 	uint32_t  tfb_flags;
310 	uint8_t	tfb_id;
311 };
312 
313 struct tcp_function {
314 	TAILQ_ENTRY(tcp_function)	tf_next;
315 	char				tf_name[TCP_FUNCTION_NAME_LEN_MAX];
316 	struct tcp_function_block	*tf_fb;
317 };
318 
319 TAILQ_HEAD(tcp_funchead, tcp_function);
320 #endif	/* _KERNEL */
321 
322 /*
323  * Flags and utility macros for the t_flags field.
324  */
325 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
326 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
327 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
328 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
329 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
330 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
331 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
332 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
333 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
334 #define	TF_SACK_PERMIT	0x00000200	/* other side said I could SACK */
335 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
336 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
337 #define	TF_NOPUSH	0x00001000	/* don't push */
338 #define	TF_PREVVALID	0x00002000	/* saved values for bad rxmit valid */
339 #define	TF_UNUSED1	0x00004000	/* unused */
340 #define	TF_GPUTINPROG	0x00008000	/* Goodput measurement in progress */
341 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
342 #define	TF_LQ_OVERFLOW	0x00020000	/* listen queue overflow */
343 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
344 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
345 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
346 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
347 #define	TF_SIGNATURE	0x00400000	/* require MD5 digests (RFC2385) */
348 #define	TF_FORCEDATA	0x00800000	/* force out a byte */
349 #define	TF_TSO		0x01000000	/* TSO enabled on this connection */
350 #define	TF_TOE		0x02000000	/* this connection is offloaded */
351 #define	TF_UNUSED3	0x04000000	/* unused */
352 #define	TF_UNUSED4	0x08000000	/* unused */
353 #define	TF_UNUSED5	0x10000000	/* unused */
354 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
355 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
356 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
357 
358 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
359 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
360 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
361 
362 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
363 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
364 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
365 
366 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
367 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
368 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
369 
370 #if defined(_KERNEL) && !defined(TCP_RFC7413)
371 #define	IS_FASTOPEN(t_flags)		(false)
372 #else
373 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
374 #endif
375 
376 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
377 
378 /*
379  * Flags for the t_oobflags field.
380  */
381 #define	TCPOOB_HAVEDATA	0x01
382 #define	TCPOOB_HADDATA	0x02
383 
384 /*
385  * Flags for the extended TCP flags field, t_flags2
386  */
387 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
388 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
389 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
390 #define	TF2_LOG_AUTO		0x00000008 /* Session is auto-logging. */
391 #define TF2_DROP_AF_DATA 	0x00000010 /* Drop after all data ack'd */
392 #define	TF2_ECN_PERMIT		0x00000020 /* connection ECN-ready */
393 #define	TF2_ECN_SND_CWR		0x00000040 /* ECN CWR in queue */
394 #define	TF2_ECN_SND_ECE		0x00000080 /* ECN ECE in queue */
395 #define	TF2_ACE_PERMIT		0x00000100 /* Accurate ECN mode */
396 
397 /*
398  * Structure to hold TCP options that are only used during segment
399  * processing (in tcp_input), but not held in the tcpcb.
400  * It's basically used to reduce the number of parameters
401  * to tcp_dooptions and tcp_addoptions.
402  * The binary order of the to_flags is relevant for packing of the
403  * options in tcp_addoptions.
404  */
405 struct tcpopt {
406 	u_int32_t	to_flags;	/* which options are present */
407 #define	TOF_MSS		0x0001		/* maximum segment size */
408 #define	TOF_SCALE	0x0002		/* window scaling */
409 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
410 #define	TOF_TS		0x0010		/* timestamp */
411 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
412 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
413 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
414 #define	TOF_MAXOPT	0x0200
415 	u_int32_t	to_tsval;	/* new timestamp */
416 	u_int32_t	to_tsecr;	/* reflected timestamp */
417 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
418 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
419 	u_int8_t	*to_tfo_cookie; /* pointer to the TFO cookie */
420 	u_int16_t	to_mss;		/* maximum segment size */
421 	u_int8_t	to_wscale;	/* window scaling */
422 	u_int8_t	to_nsacks;	/* number of SACK blocks */
423 	u_int8_t	to_tfo_len;	/* TFO cookie length */
424 	u_int32_t	to_spare;	/* UTO */
425 };
426 
427 /*
428  * Flags for tcp_dooptions.
429  */
430 #define	TO_SYN		0x01		/* parse SYN-only options */
431 
432 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
433 	uint32_t	rmx_mtu;	/* MTU for this path */
434 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
435 	uint32_t	rmx_rtt;	/* estimated round trip time */
436 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
437 	uint32_t	rmx_cwnd;	/* congestion window */
438 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
439 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
440 };
441 
442 /*
443  * Used by tcp_maxmtu() to communicate interface specific features
444  * and limits at the time of connection setup.
445  */
446 struct tcp_ifcap {
447 	int	ifcap;
448 	u_int	tsomax;
449 	u_int	tsomaxsegcount;
450 	u_int	tsomaxsegsize;
451 };
452 
453 #ifndef _NETINET_IN_PCB_H_
454 struct in_conninfo;
455 #endif /* _NETINET_IN_PCB_H_ */
456 
457 struct tcptw {
458 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
459 	tcp_seq		snd_nxt;
460 	tcp_seq		rcv_nxt;
461 	tcp_seq		iss;
462 	tcp_seq		irs;
463 	u_short		last_win;	/* cached window value */
464 	short		tw_so_options;	/* copy of so_options */
465 	struct ucred	*tw_cred;	/* user credentials */
466 	u_int32_t	t_recent;
467 	u_int32_t	ts_offset;	/* our timestamp offset */
468 	u_int		t_starttime;
469 	int		tw_time;
470 	TAILQ_ENTRY(tcptw) tw_2msl;
471 	void		*tw_pspare;	/* TCP_SIGNATURE */
472 	u_int		*tw_spare;	/* TCP_SIGNATURE */
473 };
474 
475 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
476 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
477 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
478 
479 /*
480  * The smoothed round-trip time and estimated variance
481  * are stored as fixed point numbers scaled by the values below.
482  * For convenience, these scales are also used in smoothing the average
483  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
484  * With these scales, srtt has 3 bits to the right of the binary point,
485  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
486  * binary point, and is smoothed with an ALPHA of 0.75.
487  */
488 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
489 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
490 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
491 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
492 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
493 
494 /*
495  * The initial retransmission should happen at rtt + 4 * rttvar.
496  * Because of the way we do the smoothing, srtt and rttvar
497  * will each average +1/2 tick of bias.  When we compute
498  * the retransmit timer, we want 1/2 tick of rounding and
499  * 1 extra tick because of +-1/2 tick uncertainty in the
500  * firing of the timer.  The bias will give us exactly the
501  * 1.5 tick we need.  But, because the bias is
502  * statistical, we have to test that we don't drop below
503  * the minimum feasible timer (which is 2 ticks).
504  * This version of the macro adapted from a paper by Lawrence
505  * Brakmo and Larry Peterson which outlines a problem caused
506  * by insufficient precision in the original implementation,
507  * which results in inappropriately large RTO values for very
508  * fast networks.
509  */
510 #define	TCP_REXMTVAL(tp) \
511 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
512 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
513 
514 /*
515  * TCP statistics.
516  * Many of these should be kept per connection,
517  * but that's inconvenient at the moment.
518  */
519 struct	tcpstat {
520 	uint64_t tcps_connattempt;	/* connections initiated */
521 	uint64_t tcps_accepts;		/* connections accepted */
522 	uint64_t tcps_connects;		/* connections established */
523 	uint64_t tcps_drops;		/* connections dropped */
524 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
525 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
526 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
527 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
528 	uint64_t tcps_rttupdated;	/* times we succeeded */
529 	uint64_t tcps_delack;		/* delayed acks sent */
530 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
531 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
532 	uint64_t tcps_persisttimeo;	/* persist timeouts */
533 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
534 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
535 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
536 
537 	uint64_t tcps_sndtotal;		/* total packets sent */
538 	uint64_t tcps_sndpack;		/* data packets sent */
539 	uint64_t tcps_sndbyte;		/* data bytes sent */
540 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
541 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
542 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
543 	uint64_t tcps_sndacks;		/* ack-only packets sent */
544 	uint64_t tcps_sndprobe;		/* window probes sent */
545 	uint64_t tcps_sndurg;		/* packets sent with URG only */
546 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
547 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
548 
549 	uint64_t tcps_rcvtotal;		/* total packets received */
550 	uint64_t tcps_rcvpack;		/* packets received in sequence */
551 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
552 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
553 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
554 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
555 	uint64_t tcps_rcvshort;		/* packets received too short */
556 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
557 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
558 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
559 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
560 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
561 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
562 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
563 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
564 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
565 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
566 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
567 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
568 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
569 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
570 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
571 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
572 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
573 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
574 	uint64_t tcps_pcbcachemiss;
575 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
576 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
577 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
578 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
579 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
580 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
581 	uint64_t tcps_persistdrop;	/* timeout in persist state */
582 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
583 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
584 	uint64_t tcps_listendrop;	/* listen queue overflows */
585 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
586 
587 	uint64_t tcps_sc_added;		/* entry added to syncache */
588 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
589 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
590 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
591 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
592 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
593 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
594 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
595 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
596 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
597 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
598 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
599 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
600 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
601 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
602 
603 	uint64_t tcps_hc_added;		/* entry added to hostcache */
604 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
605 
606 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
607 
608 	/* SACK related stats */
609 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
610 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
611 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
612 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
613 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
614 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
615 
616 	/* ECN related stats */
617 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
618 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
619 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
620 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
621 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
622 
623 	/* TCP_SIGNATURE related stats */
624 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
625 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
626 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
627 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
628 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
629 
630 	/* Path MTU Discovery Black Hole Detection related stats */
631 	uint64_t tcps_pmtud_blackhole_activated;	 /* Black Hole Count */
632 	uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
633 	uint64_t tcps_pmtud_blackhole_failed;		 /* Black Hole Failure Count */
634 
635 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
636 };
637 
638 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
639 
640 #ifdef _KERNEL
641 #define	TI_UNLOCKED	1
642 #define	TI_RLOCKED	2
643 #include <sys/counter.h>
644 
645 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
646 /*
647  * In-kernel consumers can use these accessor macros directly to update
648  * stats.
649  */
650 #define	TCPSTAT_ADD(name, val)	\
651     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
652 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
653 
654 /*
655  * Kernel module consumers must use this accessor macro.
656  */
657 void	kmod_tcpstat_inc(int statnum);
658 #define	KMOD_TCPSTAT_INC(name)						\
659     kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
660 
661 /*
662  * Running TCP connection count by state.
663  */
664 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
665 #define	V_tcps_states	VNET(tcps_states)
666 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
667 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
668 
669 /*
670  * TCP specific helper hook point identifiers.
671  */
672 #define	HHOOK_TCP_EST_IN		0
673 #define	HHOOK_TCP_EST_OUT		1
674 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
675 
676 struct tcp_hhook_data {
677 	struct tcpcb	*tp;
678 	struct tcphdr	*th;
679 	struct tcpopt	*to;
680 	uint32_t	len;
681 	int		tso;
682 	tcp_seq		curack;
683 };
684 #ifdef TCP_HHOOK
685 void hhook_run_tcp_est_out(struct tcpcb *tp,
686 	struct tcphdr *th, struct tcpopt *to,
687 	uint32_t len, int tso);
688 #endif
689 #endif
690 
691 /*
692  * TCB structure exported to user-land via sysctl(3).
693  *
694  * Fields prefixed with "xt_" are unique to the export structure, and fields
695  * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
696  *
697  * Legend:
698  * (s) - used by userland utilities in src
699  * (p) - used by utilities in ports
700  * (3) - is known to be used by third party software not in ports
701  * (n) - no known usage
702  *
703  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
704  * included.  Not all of our clients do.
705  */
706 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
707 struct xtcpcb {
708 	ksize_t	xt_len;		/* length of this structure */
709 	struct xinpcb	xt_inp;
710 	char		xt_stack[TCP_FUNCTION_NAME_LEN_MAX];	/* (s) */
711 	char		xt_logid[TCP_LOG_ID_LEN];	/* (s) */
712 	int64_t		spare64[8];
713 	int32_t		t_state;		/* (s,p) */
714 	uint32_t	t_flags;		/* (s,p) */
715 	int32_t		t_sndzerowin;		/* (s) */
716 	int32_t		t_sndrexmitpack;	/* (s) */
717 	int32_t		t_rcvoopack;		/* (s) */
718 	int32_t		t_rcvtime;		/* (s) */
719 	int32_t		tt_rexmt;		/* (s) */
720 	int32_t		tt_persist;		/* (s) */
721 	int32_t		tt_keep;		/* (s) */
722 	int32_t		tt_2msl;		/* (s) */
723 	int32_t		tt_delack;		/* (s) */
724 	int32_t		t_logstate;		/* (3) */
725 	int32_t		spare32[32];
726 } __aligned(8);
727 
728 #ifdef _KERNEL
729 void	tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
730 #endif
731 #endif
732 
733 /*
734  * TCP function information (name-to-id mapping, aliases, and refcnt)
735  * exported to user-land via sysctl(3).
736  */
737 struct tcp_function_info {
738 	uint32_t	tfi_refcnt;
739 	uint8_t		tfi_id;
740 	char		tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
741 	char		tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
742 };
743 
744 /*
745  * Identifiers for TCP sysctl nodes
746  */
747 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
748 #define	TCPCTL_MSSDFLT		3	/* MSS default */
749 #define TCPCTL_STATS		4	/* statistics */
750 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
751 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
752 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
753 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
754 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
755 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
756 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
757 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
758 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
759 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
760 #define	TCPCTL_DROP		15	/* drop tcp connection */
761 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
762 
763 #ifdef _KERNEL
764 #ifdef SYSCTL_DECL
765 SYSCTL_DECL(_net_inet_tcp);
766 SYSCTL_DECL(_net_inet_tcp_sack);
767 MALLOC_DECLARE(M_TCPLOG);
768 #endif
769 
770 VNET_DECLARE(int, tcp_log_in_vain);
771 #define	V_tcp_log_in_vain		VNET(tcp_log_in_vain)
772 
773 /*
774  * Global TCP tunables shared between different stacks.
775  * Please keep the list sorted.
776  */
777 VNET_DECLARE(int, drop_synfin);
778 VNET_DECLARE(int, path_mtu_discovery);
779 VNET_DECLARE(int, tcp_abc_l_var);
780 VNET_DECLARE(int, tcp_autorcvbuf_max);
781 VNET_DECLARE(int, tcp_autosndbuf_inc);
782 VNET_DECLARE(int, tcp_autosndbuf_max);
783 VNET_DECLARE(int, tcp_delack_enabled);
784 VNET_DECLARE(int, tcp_do_autorcvbuf);
785 VNET_DECLARE(int, tcp_do_autosndbuf);
786 VNET_DECLARE(int, tcp_do_ecn);
787 VNET_DECLARE(int, tcp_do_newcwv);
788 VNET_DECLARE(int, tcp_do_rfc1323);
789 VNET_DECLARE(int, tcp_do_rfc3042);
790 VNET_DECLARE(int, tcp_do_rfc3390);
791 VNET_DECLARE(int, tcp_do_rfc3465);
792 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
793 VNET_DECLARE(int, tcp_do_sack);
794 VNET_DECLARE(int, tcp_do_tso);
795 VNET_DECLARE(int, tcp_ecn_maxretries);
796 VNET_DECLARE(int, tcp_initcwnd_segments);
797 VNET_DECLARE(int, tcp_insecure_rst);
798 VNET_DECLARE(int, tcp_insecure_syn);
799 VNET_DECLARE(uint32_t, tcp_map_entries_limit);
800 VNET_DECLARE(uint32_t, tcp_map_split_limit);
801 VNET_DECLARE(int, tcp_minmss);
802 VNET_DECLARE(int, tcp_mssdflt);
803 #ifdef STATS
804 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl);
805 VNET_DECLARE(int, tcp_perconn_stats_enable);
806 #endif /* STATS */
807 VNET_DECLARE(int, tcp_recvspace);
808 VNET_DECLARE(int, tcp_sack_globalholes);
809 VNET_DECLARE(int, tcp_sack_globalmaxholes);
810 VNET_DECLARE(int, tcp_sack_maxholes);
811 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
812 VNET_DECLARE(int, tcp_sendspace);
813 VNET_DECLARE(struct inpcbhead, tcb);
814 VNET_DECLARE(struct inpcbinfo, tcbinfo);
815 
816 #define	V_tcp_do_newcwv			VNET(tcp_do_newcwv)
817 #define	V_drop_synfin			VNET(drop_synfin)
818 #define	V_path_mtu_discovery		VNET(path_mtu_discovery)
819 #define	V_tcb				VNET(tcb)
820 #define	V_tcbinfo			VNET(tcbinfo)
821 #define	V_tcp_abc_l_var			VNET(tcp_abc_l_var)
822 #define	V_tcp_autorcvbuf_max		VNET(tcp_autorcvbuf_max)
823 #define	V_tcp_autosndbuf_inc		VNET(tcp_autosndbuf_inc)
824 #define	V_tcp_autosndbuf_max		VNET(tcp_autosndbuf_max)
825 #define	V_tcp_delack_enabled		VNET(tcp_delack_enabled)
826 #define	V_tcp_do_autorcvbuf		VNET(tcp_do_autorcvbuf)
827 #define	V_tcp_do_autosndbuf		VNET(tcp_do_autosndbuf)
828 #define	V_tcp_do_ecn			VNET(tcp_do_ecn)
829 #define	V_tcp_do_rfc1323		VNET(tcp_do_rfc1323)
830 #define V_tcp_ts_offset_per_conn	VNET(tcp_ts_offset_per_conn)
831 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
832 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)
833 #define	V_tcp_do_rfc3465		VNET(tcp_do_rfc3465)
834 #define	V_tcp_do_rfc6675_pipe		VNET(tcp_do_rfc6675_pipe)
835 #define	V_tcp_do_sack			VNET(tcp_do_sack)
836 #define	V_tcp_do_tso			VNET(tcp_do_tso)
837 #define	V_tcp_ecn_maxretries		VNET(tcp_ecn_maxretries)
838 #define	V_tcp_initcwnd_segments		VNET(tcp_initcwnd_segments)
839 #define	V_tcp_insecure_rst		VNET(tcp_insecure_rst)
840 #define	V_tcp_insecure_syn		VNET(tcp_insecure_syn)
841 #define	V_tcp_map_entries_limit		VNET(tcp_map_entries_limit)
842 #define	V_tcp_map_split_limit		VNET(tcp_map_split_limit)
843 #define	V_tcp_minmss			VNET(tcp_minmss)
844 #define	V_tcp_mssdflt			VNET(tcp_mssdflt)
845 #ifdef STATS
846 #define	V_tcp_perconn_stats_dflt_tpl	VNET(tcp_perconn_stats_dflt_tpl)
847 #define	V_tcp_perconn_stats_enable	VNET(tcp_perconn_stats_enable)
848 #endif /* STATS */
849 #define	V_tcp_recvspace			VNET(tcp_recvspace)
850 #define	V_tcp_sack_globalholes		VNET(tcp_sack_globalholes)
851 #define	V_tcp_sack_globalmaxholes	VNET(tcp_sack_globalmaxholes)
852 #define	V_tcp_sack_maxholes		VNET(tcp_sack_maxholes)
853 #define	V_tcp_sc_rst_sock_fail		VNET(tcp_sc_rst_sock_fail)
854 #define	V_tcp_sendspace			VNET(tcp_sendspace)
855 #define	V_tcp_udp_tunneling_overhead	VNET(tcp_udp_tunneling_overhead)
856 #define	V_tcp_udp_tunneling_port	VNET(tcp_udp_tunneling_port)
857 
858 #ifdef TCP_HHOOK
859 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
860 #define	V_tcp_hhh		VNET(tcp_hhh)
861 #endif
862 
863 int	 tcp_addoptions(struct tcpopt *, u_char *);
864 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
865 struct tcpcb *
866 	 tcp_close(struct tcpcb *);
867 void	 tcp_discardcb(struct tcpcb *);
868 void	 tcp_twstart(struct tcpcb *);
869 void	 tcp_twclose(struct tcptw *, int);
870 void	 tcp_ctlinput(int, struct sockaddr *, void *);
871 int	 tcp_ctloutput(struct socket *, struct sockopt *);
872 struct tcpcb *
873 	 tcp_drop(struct tcpcb *, int);
874 void	 tcp_drain(void);
875 void	 tcp_init(void);
876 void	 tcp_fini(void *);
877 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
878 	    const void *);
879 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
880 	    const void *);
881 int	 tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *, struct mbuf *);
882 void	 tcp_reass_global_init(void);
883 void	 tcp_reass_flush(struct tcpcb *);
884 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
885 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
886 		     struct tcpcb *, int, int);
887 void	tcp_pulloutofband(struct socket *,
888 		     struct tcphdr *, struct mbuf *, int);
889 void	tcp_xmit_timer(struct tcpcb *, int);
890 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
891 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
892 			    uint16_t nsegs, uint16_t type);
893 void 	cc_conn_init(struct tcpcb *tp);
894 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
895 void    cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos);
896 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
897 #ifdef TCP_HHOOK
898 void	hhook_run_tcp_est_in(struct tcpcb *tp,
899 			    struct tcphdr *th, struct tcpopt *to);
900 #endif
901 
902 int	 tcp_input(struct mbuf **, int *, int);
903 int	 tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
904 	    struct tcpcb *, int);
905 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
906 			struct socket *, struct tcpcb *, int, int, uint8_t);
907 
908 int register_tcp_functions(struct tcp_function_block *blk, int wait);
909 int register_tcp_functions_as_names(struct tcp_function_block *blk,
910     int wait, const char *names[], int *num_names);
911 int register_tcp_functions_as_name(struct tcp_function_block *blk,
912     const char *name, int wait);
913 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
914     bool force);
915 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
916 void tcp_switch_back_to_default(struct tcpcb *tp);
917 struct tcp_function_block *
918 find_and_ref_tcp_fb(struct tcp_function_block *fs);
919 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
920 
921 extern counter_u64_t tcp_inp_lro_direct_queue;
922 extern counter_u64_t tcp_inp_lro_wokeup_queue;
923 extern counter_u64_t tcp_inp_lro_compressed;
924 extern counter_u64_t tcp_inp_lro_single_push;
925 extern counter_u64_t tcp_inp_lro_locks_taken;
926 extern counter_u64_t tcp_inp_lro_sack_wake;
927 
928 #ifdef NETFLIX_EXP_DETECTION
929 /* Various SACK attack thresholds */
930 extern int32_t tcp_force_detection;
931 extern int32_t tcp_sack_to_ack_thresh;
932 extern int32_t tcp_sack_to_move_thresh;
933 extern int32_t tcp_restoral_thresh;
934 extern int32_t tcp_sad_decay_val;
935 extern int32_t tcp_sad_pacing_interval;
936 extern int32_t tcp_sad_low_pps;
937 extern int32_t tcp_map_minimum;
938 extern int32_t tcp_attack_on_turns_on_logging;
939 #endif
940 
941 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
942 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
943 u_int	 tcp_maxseg(const struct tcpcb *);
944 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
945 	    struct tcp_ifcap *);
946 void	 tcp_mss(struct tcpcb *, int);
947 int	 tcp_mssopt(struct in_conninfo *);
948 struct inpcb *
949 	 tcp_drop_syn_sent(struct inpcb *, int);
950 struct tcpcb *
951 	 tcp_newtcpcb(struct inpcb *);
952 int	 tcp_output(struct tcpcb *);
953 void	 tcp_state_change(struct tcpcb *, int);
954 void	 tcp_respond(struct tcpcb *, void *,
955 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
956 void	 tcp_tw_init(void);
957 #ifdef VIMAGE
958 void	 tcp_tw_destroy(void);
959 #endif
960 void	 tcp_tw_zone_change(void);
961 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
962 	    struct mbuf *, int);
963 void	 tcp_setpersist(struct tcpcb *);
964 void	 tcp_slowtimo(void);
965 struct tcptemp *
966 	 tcpip_maketemplate(struct inpcb *);
967 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
968 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
969 int	 tcp_timer_suspend(struct tcpcb *, uint32_t);
970 void	 tcp_timers_unsuspend(struct tcpcb *, uint32_t);
971 int	 tcp_timer_active(struct tcpcb *, uint32_t);
972 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
973 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
974 int	 inp_to_cpuid(struct inpcb *inp);
975 /*
976  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
977  */
978 void	 tcp_hc_init(void);
979 #ifdef VIMAGE
980 void	 tcp_hc_destroy(void);
981 #endif
982 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
983 uint32_t tcp_hc_getmtu(struct in_conninfo *);
984 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
985 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
986 
987 extern	struct pr_usrreqs tcp_usrreqs;
988 
989 uint32_t tcp_new_ts_offset(struct in_conninfo *);
990 tcp_seq	 tcp_new_isn(struct in_conninfo *);
991 
992 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
993 void	 tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
994 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
995 void	 tcp_clean_dsack_blocks(struct tcpcb *tp);
996 void	 tcp_clean_sackreport(struct tcpcb *tp);
997 void	 tcp_sack_adjust(struct tcpcb *tp);
998 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
999 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1000 void	 tcp_free_sackholes(struct tcpcb *tp);
1001 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
1002 int	 tcp_compute_pipe(struct tcpcb *);
1003 uint32_t tcp_compute_initwnd(uint32_t);
1004 void	 tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
1005 int	 tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes,
1006     size_t seed_len);
1007 struct mbuf *
1008 	 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
1009 	   int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls);
1010 
1011 int	tcp_stats_init(void);
1012 
1013 static inline void
1014 tcp_fields_to_host(struct tcphdr *th)
1015 {
1016 
1017 	th->th_seq = ntohl(th->th_seq);
1018 	th->th_ack = ntohl(th->th_ack);
1019 	th->th_win = ntohs(th->th_win);
1020 	th->th_urp = ntohs(th->th_urp);
1021 }
1022 
1023 static inline void
1024 tcp_fields_to_net(struct tcphdr *th)
1025 {
1026 
1027 	th->th_seq = htonl(th->th_seq);
1028 	th->th_ack = htonl(th->th_ack);
1029 	th->th_win = htons(th->th_win);
1030 	th->th_urp = htons(th->th_urp);
1031 }
1032 #endif /* _KERNEL */
1033 
1034 #endif /* _NETINET_TCP_VAR_H_ */
1035