xref: /freebsd/sys/netinet/tcp_var.h (revision 9768746b)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1993, 1994, 1995
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
32  * $FreeBSD$
33  */
34 
35 #ifndef _NETINET_TCP_VAR_H_
36 #define _NETINET_TCP_VAR_H_
37 
38 #include <netinet/tcp.h>
39 #include <netinet/tcp_fsm.h>
40 
41 #ifdef _KERNEL
42 #include "opt_kern_tls.h"
43 #include <net/vnet.h>
44 #include <sys/mbuf.h>
45 #include <sys/ktls.h>
46 #endif
47 
48 #define TCP_END_BYTE_INFO 8	/* Bytes that makeup the "end information array" */
49 /* Types of ending byte info */
50 #define TCP_EI_EMPTY_SLOT	0
51 #define TCP_EI_STATUS_CLIENT_FIN	0x1
52 #define TCP_EI_STATUS_CLIENT_RST	0x2
53 #define TCP_EI_STATUS_SERVER_FIN	0x3
54 #define TCP_EI_STATUS_SERVER_RST	0x4
55 #define TCP_EI_STATUS_RETRAN		0x5
56 #define TCP_EI_STATUS_PROGRESS		0x6
57 #define TCP_EI_STATUS_PERSIST_MAX	0x7
58 #define TCP_EI_STATUS_KEEP_MAX		0x8
59 #define TCP_EI_STATUS_DATA_A_CLOSE	0x9
60 #define TCP_EI_STATUS_RST_IN_FRONT	0xa
61 #define TCP_EI_STATUS_2MSL		0xb
62 #define TCP_EI_STATUS_MAX_VALUE		0xb
63 
64 /************************************************/
65 /* Status bits we track to assure no duplicates,
66  * the bits here are not used by the code but
67  * for human representation. To check a bit we
68  * take and shift over by 1 minus the value (1-8).
69  */
70 /************************************************/
71 #define TCP_EI_BITS_CLIENT_FIN	0x001
72 #define TCP_EI_BITS_CLIENT_RST	0x002
73 #define TCP_EI_BITS_SERVER_FIN	0x004
74 #define TCP_EI_BITS_SERVER_RST	0x008
75 #define TCP_EI_BITS_RETRAN	0x010
76 #define TCP_EI_BITS_PROGRESS	0x020
77 #define TCP_EI_BITS_PRESIST_MAX	0x040
78 #define TCP_EI_BITS_KEEP_MAX	0x080
79 #define TCP_EI_BITS_DATA_A_CLO  0x100
80 #define TCP_EI_BITS_RST_IN_FR	0x200	/* a front state reset */
81 #define TCP_EI_BITS_2MS_TIMER	0x400	/* 2 MSL timer expired */
82 
83 #if defined(_KERNEL) || defined(_WANT_TCPCB)
84 #include <netinet/cc/cc.h>
85 
86 /* TCP segment queue entry */
87 struct tseg_qent {
88 	TAILQ_ENTRY(tseg_qent) tqe_q;
89 	struct	mbuf   *tqe_m;		/* mbuf contains packet */
90 	struct  mbuf   *tqe_last;	/* last mbuf in chain */
91 	tcp_seq tqe_start;		/* TCP Sequence number start */
92 	int	tqe_len;		/* TCP segment data length */
93 	uint32_t tqe_flags;		/* The flags from tcp_get_flags() */
94 	uint32_t tqe_mbuf_cnt;		/* Count of mbuf overhead */
95 };
96 TAILQ_HEAD(tsegqe_head, tseg_qent);
97 
98 struct sackblk {
99 	tcp_seq start;		/* start seq no. of sack block */
100 	tcp_seq end;		/* end seq no. */
101 };
102 
103 struct sackhole {
104 	tcp_seq start;		/* start seq no. of hole */
105 	tcp_seq end;		/* end seq no. */
106 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
107 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
108 };
109 
110 struct sackhint {
111 	struct sackhole	*nexthole;
112 	int32_t		sack_bytes_rexmit;
113 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
114 
115 	int32_t		delivered_data; /* Newly acked data from last SACK */
116 
117 	int32_t		sacked_bytes;	/* Total sacked bytes reported by the
118 					 * receiver via sack option
119 					 */
120 	uint32_t	recover_fs;	/* Flight Size at the start of Loss recovery */
121 	uint32_t	prr_delivered;	/* Total bytes delivered using PRR */
122 	uint32_t	prr_out;	/* Bytes sent during IN_RECOVERY */
123 };
124 
125 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
126 
127 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
128 
129 typedef enum {
130 	TT_DELACK = 0,
131 	TT_REXMT,
132 	TT_PERSIST,
133 	TT_KEEP,
134 	TT_2MSL,
135 	TT_N,
136 } tt_which;
137 
138 /*
139  * Tcp control block, one per tcp connection.
140  */
141 struct tcpcb {
142 	struct inpcb t_inpcb;		/* embedded protocol independent cb */
143 #define	t_start_zero	t_fb
144 #define	t_zero_size	(sizeof(struct tcpcb) - \
145 			    offsetof(struct tcpcb, t_start_zero))
146 	struct tcp_function_block *t_fb;/* TCP function call block */
147 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
148 
149 	struct callout t_callout;
150 	sbintime_t t_timers[TT_N];
151 	sbintime_t t_precisions[TT_N];
152 
153 	uint32_t t_maxseg:24,		/* maximum segment size */
154 		t_logstate:8;		/* State of "black box" logging */
155 	uint32_t t_port:16,		/* Tunneling (over udp) port */
156 		t_state:4,		/* state of this connection */
157 		t_idle_reduce : 1,
158 		t_delayed_ack: 7,	/* Delayed ack variable */
159 		t_fin_is_rst: 1,	/* Are fin's treated as resets */
160 		t_log_state_set: 1,
161 		bits_spare : 2;
162 	u_int	t_flags;
163 	tcp_seq	snd_una;		/* sent but unacknowledged */
164 	tcp_seq	snd_max;		/* highest sequence number sent;
165 					 * used to recognize retransmits
166 					 */
167 	tcp_seq snd_nxt;		/* send next */
168 	tcp_seq snd_up;			/* send urgent pointer */
169 	uint32_t snd_wnd;		/* send window */
170 	uint32_t snd_cwnd;		/* congestion-controlled window */
171 	uint32_t t_peakrate_thr; 	/* pre-calculated peak rate threshold */
172 	uint32_t ts_offset;		/* our timestamp offset */
173 	uint32_t rfbuf_ts;		/* recv buffer autoscaling timestamp */
174 	int	rcv_numsacks;		/* # distinct sack blks present */
175 	u_int	t_tsomax;		/* TSO total burst length limit */
176 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
177 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
178 	tcp_seq	rcv_nxt;		/* receive next */
179 	tcp_seq	rcv_adv;		/* advertised window */
180 	uint32_t rcv_wnd;		/* receive window */
181 	u_int	t_flags2;		/* More tcpcb flags storage */
182 	int	t_srtt;			/* smoothed round-trip time */
183 	int	t_rttvar;		/* variance in round-trip time */
184 	uint32_t ts_recent;		/* timestamp echo data */
185 	u_char	snd_scale;		/* window scaling for send window */
186 	u_char	rcv_scale;		/* window scaling for recv window */
187 	u_char	snd_limited;		/* segments limited transmitted */
188 	u_char	request_r_scale;	/* pending window scaling */
189 	tcp_seq	last_ack_sent;
190 	u_int	t_rcvtime;		/* inactivity time */
191 	tcp_seq	rcv_up;			/* receive urgent pointer */
192 	int	t_segqlen;		/* segment reassembly queue length */
193 	uint32_t t_segqmbuflen;		/* total reassembly queue byte length */
194 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
195 	struct mbuf *t_in_pkt;
196 	struct mbuf *t_tail_pkt;
197 	uint32_t snd_ssthresh;		/* snd_cwnd size threshold for
198 					 * for slow start exponential to
199 					 * linear switch
200 					 */
201 	tcp_seq	snd_wl1;		/* window update seg seq number */
202 	tcp_seq	snd_wl2;		/* window update seg ack number */
203 
204 	tcp_seq	irs;			/* initial receive sequence number */
205 	tcp_seq	iss;			/* initial send sequence number */
206 	u_int	t_acktime;		/* RACK and BBR incoming new data was acked */
207 	u_int	t_sndtime;		/* time last data was sent */
208 	u_int	ts_recent_age;		/* when last updated */
209 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
210 	char	t_oobflags;		/* have some */
211 	char	t_iobc;			/* input character */
212 	uint8_t t_nic_ktls_xmit:1,	/* active nic ktls xmit sessions */
213 		t_nic_ktls_xmit_dis:1,	/* disabled nic xmit ktls? */
214 		t_nic_ktls_spare:6;	/* spare nic ktls */
215 	int	t_rxtcur;		/* current retransmit value (ticks) */
216 
217 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
218 	u_int	t_rtttime;		/* RTT measurement start time */
219 
220 	tcp_seq	t_rtseq;		/* sequence number being timed */
221 	u_int	t_starttime;		/* time connection was established */
222 	u_int	t_fbyte_in;		/* ticks time first byte queued in */
223 	u_int	t_fbyte_out;		/* ticks time first byte queued out */
224 
225 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
226 	int	t_blackhole_enter;	/* when to enter blackhole detection */
227 	int	t_blackhole_exit;	/* when to exit blackhole detection */
228 	u_int	t_rttmin;		/* minimum rtt allowed */
229 
230 	int	t_softerror;		/* possible error not yet reported */
231 	uint32_t max_sndwnd;		/* largest window peer has offered */
232 	uint32_t snd_cwnd_prev;		/* cwnd prior to retransmit */
233 	uint32_t snd_ssthresh_prev;	/* ssthresh prior to retransmit */
234 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
235 	int	t_sndzerowin;		/* zero-window updates sent */
236 	int	snd_numholes;		/* number of holes seen by sender */
237 	u_int	t_badrxtwin;		/* window for retransmit recovery */
238 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
239 					/* SACK scoreboard (sorted) */
240 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
241 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
242 	struct sackhint	sackhint;	/* SACK scoreboard hint */
243 	int	t_rttlow;		/* smallest observerved RTT */
244 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
245 	struct toedev	*tod;		/* toedev handling this connection */
246 	int	t_sndrexmitpack;	/* retransmit packets sent */
247 	int	t_rcvoopack;		/* out-of-order packets received */
248 	void	*t_toe;			/* TOE pcb pointer */
249 	struct cc_algo	*t_cc;		/* congestion control algorithm */
250 	struct cc_var	t_ccv;		/* congestion control specific vars */
251 	int	t_bytes_acked;		/* # bytes acked during current RTT */
252 	u_int	t_maxunacktime;
253 	u_int	t_keepinit;		/* time to establish connection */
254 	u_int	t_keepidle;		/* time before keepalive probes begin */
255 	u_int	t_keepintvl;		/* interval between keepalives */
256 	u_int	t_keepcnt;		/* number of keepalives before close */
257 	int	t_dupacks;		/* consecutive dup acks recd */
258 	int	t_lognum;		/* Number of log entries */
259 	int	t_loglimit;		/* Maximum number of log entries */
260 	uint32_t t_rcep;		/* Number of received CE marked pkts */
261 	uint32_t t_scep;		/* Synced number of delivered CE pkts */
262 	int64_t	t_pacing_rate;		/* bytes / sec, -1 => unlimited */
263 	struct tcp_log_stailq t_logs;	/* Log buffer */
264 	struct tcp_log_id_node *t_lin;
265 	struct tcp_log_id_bucket *t_lib;
266 	const char *t_output_caller;	/* Function that called tcp_output */
267 	struct statsblob *t_stats;	/* Per-connection stats */
268 	uint32_t t_logsn;		/* Log "serial number" */
269 	uint32_t gput_ts;		/* Time goodput measurement started */
270 	tcp_seq gput_seq;		/* Outbound measurement seq */
271 	tcp_seq gput_ack;		/* Inbound measurement ack */
272 	int32_t t_stats_gput_prev;	/* XXXLAS: Prev gput measurement */
273 	uint32_t t_maxpeakrate;		/* max peak rate set by user, bytes/s */
274 	uint32_t t_sndtlppack;		/* tail loss probe packets sent */
275 	uint64_t t_sndtlpbyte;		/* total tail loss probe bytes sent */
276 	uint64_t t_sndbytes;		/* total bytes sent */
277 	uint64_t t_snd_rxt_bytes;	/* total bytes retransmitted */
278 	uint32_t t_dsack_bytes;		/* dsack bytes received */
279 	uint32_t t_dsack_tlp_bytes;	/* dsack bytes received for TLPs sent */
280 	uint32_t t_dsack_pack;		/* dsack packets we have eceived */
281 	uint8_t t_rttupdated;		/* number of times rtt sampled */
282 	/* TCP Fast Open */
283 	uint8_t t_tfo_client_cookie_len; /* TFO client cookie length */
284 	uint32_t t_end_info_status;	/* Status flag of end info */
285 	unsigned int *t_tfo_pending;	/* TFO server pending counter */
286 	union {
287 		uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
288 		uint64_t server;
289 	} t_tfo_cookie;			/* TCP Fast Open cookie to send */
290 	union {
291 		uint8_t t_end_info_bytes[TCP_END_BYTE_INFO];
292 		uint64_t t_end_info;
293 	};
294 #ifdef TCPPCAP
295 	struct mbufq t_inpkts;		/* List of saved input packets. */
296 	struct mbufq t_outpkts;		/* List of saved output packets. */
297 #endif
298 #ifdef TCP_HHOOK
299 	struct osd	t_osd;		/* storage for Khelp module data */
300 #endif
301 };
302 #endif	/* _KERNEL || _WANT_TCPCB */
303 
304 #ifdef _KERNEL
305 struct tcptemp {
306 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
307 	struct	tcphdr tt_t;
308 };
309 
310 /* Enable TCP/UDP tunneling port */
311 #define TCP_TUNNELING_PORT_MIN		0
312 #define TCP_TUNNELING_PORT_MAX		65535
313 #define TCP_TUNNELING_PORT_DEFAULT	0
314 
315 /* Enable TCP/UDP tunneling port */
316 #define TCP_TUNNELING_OVERHEAD_MIN	sizeof(struct udphdr)
317 #define TCP_TUNNELING_OVERHEAD_MAX	1024
318 #define TCP_TUNNELING_OVERHEAD_DEFAULT	TCP_TUNNELING_OVERHEAD_MIN
319 
320 /* Minimum map entries limit value, if set */
321 #define TCP_MIN_MAP_ENTRIES_LIMIT	128
322 
323 /*
324  * TODO: We yet need to brave plowing in
325  * to tcp_input() and the pru_usrreq() block.
326  * Right now these go to the old standards which
327  * are somewhat ok, but in the long term may
328  * need to be changed. If we do tackle tcp_input()
329  * then we need to get rid of the tcp_do_segment()
330  * function below.
331  */
332 /* Flags for tcp functions */
333 #define	TCP_FUNC_BEING_REMOVED	0x01   	/* Can no longer be referenced */
334 #define	TCP_FUNC_OUTPUT_CANDROP	0x02   	/* tfb_tcp_output may ask tcp_drop */
335 
336 /*
337  * If defining the optional tcp_timers, in the
338  * tfb_tcp_timer_stop call you must use the
339  * callout_async_drain() function with the
340  * tcp_timer_discard callback. You should check
341  * the return of callout_async_drain() and if 0
342  * increment tt_draincnt. Since the timer sub-system
343  * does not know your callbacks you must provide a
344  * stop_all function that loops through and calls
345  * tcp_timer_stop() with each of your defined timers.
346  * Adding a tfb_tcp_handoff_ok function allows the socket
347  * option to change stacks to query you even if the
348  * connection is in a later stage. You return 0 to
349  * say you can take over and run your stack, you return
350  * non-zero (an error number) to say no you can't.
351  * If the function is undefined you can only change
352  * in the early states (before connect or listen).
353  * tfb_tcp_fb_fini is changed to add a flag to tell
354  * the old stack if the tcb is being destroyed or
355  * not. A one in the flag means the TCB is being
356  * destroyed, a zero indicates its transitioning to
357  * another stack (via socket option).
358  */
359 struct tcp_function_block {
360 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
361 	int	(*tfb_tcp_output)(struct tcpcb *);
362 	int	(*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
363 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
364 			    struct socket *, struct tcpcb *,
365 		        int, int, uint8_t);
366 	int     (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int);
367 	int      (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *,
368 			    struct socket *, struct tcpcb *,
369 			    int, int, uint8_t,
370 			    int, struct timeval *);
371 	void	(*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
372 			    struct socket *, struct tcpcb *,
373 			    int, int, uint8_t,
374 			    int, struct timeval *);
375 	int     (*tfb_tcp_ctloutput)(struct inpcb *inp, struct sockopt *sopt);
376 	/* Optional memory allocation/free routine */
377 	int	(*tfb_tcp_fb_init)(struct tcpcb *);
378 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
379 	/* Optional timers, must define all if you define one */
380 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
381 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
382 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
383 	void	(*tfb_tcp_mtu_chg)(struct tcpcb *);
384 	int	(*tfb_pru_options)(struct tcpcb *, int);
385 	void	(*tfb_hwtls_change)(struct tcpcb *, int);
386 	int     (*tfb_compute_pipe)(struct tcpcb *tp);
387 	volatile uint32_t tfb_refcnt;
388 	uint32_t  tfb_flags;
389 	uint8_t	tfb_id;
390 };
391 
392 struct tcp_function {
393 	TAILQ_ENTRY(tcp_function)	tf_next;
394 	char				tf_name[TCP_FUNCTION_NAME_LEN_MAX];
395 	struct tcp_function_block	*tf_fb;
396 };
397 
398 TAILQ_HEAD(tcp_funchead, tcp_function);
399 
400 struct tcpcb * tcp_drop(struct tcpcb *, int);
401 
402 #ifdef _NETINET_IN_PCB_H_
403 #define	intotcpcb(inp)	__containerof((inp), struct tcpcb, t_inpcb)
404 #define	sototcpcb(so)	intotcpcb(sotoinpcb(so))
405 #define	tptoinpcb(tp)	(&(tp)->t_inpcb)
406 #define	tptosocket(tp)	(tp)->t_inpcb.inp_socket
407 
408 /*
409  * tcp_output()
410  * Handles tcp_drop request from advanced stacks and reports that inpcb is
411  * gone with negative return code.
412  * Drop in replacement for the default stack.
413  */
414 static inline int
415 tcp_output(struct tcpcb *tp)
416 {
417 	struct inpcb *inp = tptoinpcb(tp);
418 	int rv;
419 
420 	INP_WLOCK_ASSERT(inp);
421 
422 	rv = tp->t_fb->tfb_tcp_output(tp);
423 	if (rv < 0) {
424 		KASSERT(tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP,
425 		    ("TCP stack %s requested tcp_drop(%p)",
426 		    tp->t_fb->tfb_tcp_block_name, tp));
427 		tp = tcp_drop(tp, -rv);
428 		if (tp)
429 			INP_WUNLOCK(inp);
430 	}
431 
432 	return (rv);
433 }
434 
435 /*
436  * tcp_output_unlock()
437  * Always returns unlocked, handles drop request from advanced stacks.
438  * Always returns positive error code.
439  */
440 static inline int
441 tcp_output_unlock(struct tcpcb *tp)
442 {
443 	struct inpcb *inp = tptoinpcb(tp);
444 	int rv;
445 
446 	INP_WLOCK_ASSERT(inp);
447 
448 	rv = tp->t_fb->tfb_tcp_output(tp);
449 	if (rv < 0) {
450 		KASSERT(tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP,
451 		    ("TCP stack %s requested tcp_drop(%p)",
452 		    tp->t_fb->tfb_tcp_block_name, tp));
453 		rv = -rv;
454 		tp = tcp_drop(tp, rv);
455 		if (tp)
456 			INP_WUNLOCK(inp);
457 	} else
458 		INP_WUNLOCK(inp);
459 
460 	return (rv);
461 }
462 
463 /*
464  * tcp_output_nodrop()
465  * Always returns locked.  It is caller's responsibility to run tcp_drop()!
466  * Useful in syscall implementations, when we want to perform some logging
467  * and/or tracing with tcpcb before calling tcp_drop().  To be used with
468  * tcp_unlock_or_drop() later.
469  *
470  * XXXGL: maybe don't allow stacks to return a drop request at certain
471  * TCP states? Why would it do in connect(2)? In recv(2)?
472  */
473 static inline int
474 tcp_output_nodrop(struct tcpcb *tp)
475 {
476 	int rv;
477 
478 	INP_WLOCK_ASSERT(tptoinpcb(tp));
479 
480 	rv = tp->t_fb->tfb_tcp_output(tp);
481 	KASSERT(rv >= 0 || tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP,
482 	    ("TCP stack %s requested tcp_drop(%p)",
483 	    tp->t_fb->tfb_tcp_block_name, tp));
484 	return (rv);
485 }
486 
487 /*
488  * tcp_unlock_or_drop()
489  * Handle return code from tfb_tcp_output() after we have logged/traced,
490  * to be used with tcp_output_nodrop().
491  */
492 static inline int
493 tcp_unlock_or_drop(struct tcpcb *tp, int tcp_output_retval)
494 {
495 	struct inpcb *inp = tptoinpcb(tp);
496 
497 	INP_WLOCK_ASSERT(inp);
498 
499         if (tcp_output_retval < 0) {
500                 tcp_output_retval = -tcp_output_retval;
501                 if (tcp_drop(tp, tcp_output_retval) != NULL)
502                         INP_WUNLOCK(inp);
503         } else
504 		INP_WUNLOCK(inp);
505 
506 	return (tcp_output_retval);
507 }
508 #endif	/* _NETINET_IN_PCB_H_ */
509 
510 static int inline
511 tcp_packets_this_ack(struct tcpcb *tp, tcp_seq ack)
512 {
513 	return ((ack - tp->snd_una) / tp->t_maxseg +
514 		((((ack - tp->snd_una) % tp->t_maxseg) != 0) ? 1 : 0));
515 }
516 #endif	/* _KERNEL */
517 
518 /*
519  * Flags and utility macros for the t_flags field.
520  */
521 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
522 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
523 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
524 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
525 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
526 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
527 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
528 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
529 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
530 #define	TF_SACK_PERMIT	0x00000200	/* other side said I could SACK */
531 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
532 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
533 #define	TF_NOPUSH	0x00001000	/* don't push */
534 #define	TF_PREVVALID	0x00002000	/* saved values for bad rxmit valid
535 					 * Note: accessing and restoring from
536 					 * these may only be done in the 1st
537 					 * RTO recovery round (t_rxtshift == 1)
538 					 */
539 #define	TF_WAKESOR	0x00004000	/* wake up receive socket */
540 #define	TF_GPUTINPROG	0x00008000	/* Goodput measurement in progress */
541 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
542 #define	TF_SONOTCONN	0x00020000	/* needs soisconnected() on ESTAB */
543 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
544 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
545 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
546 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
547 #define	TF_SIGNATURE	0x00400000	/* require MD5 digests (RFC2385) */
548 #define	TF_FORCEDATA	0x00800000	/* force out a byte */
549 #define	TF_TSO		0x01000000	/* TSO enabled on this connection */
550 #define	TF_TOE		0x02000000	/* this connection is offloaded */
551 #define	TF_CLOSED	0x04000000	/* close(2) called on socket */
552 #define	TF_UNUSED1	0x08000000	/* unused */
553 #define	TF_LRD		0x10000000	/* Lost Retransmission Detection */
554 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
555 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
556 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
557 
558 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
559 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
560 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
561 
562 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
563 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
564 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
565 
566 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
567 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
568 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
569 
570 #if defined(_KERNEL) && !defined(TCP_RFC7413)
571 #define	IS_FASTOPEN(t_flags)		(false)
572 #else
573 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
574 #endif
575 
576 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
577 
578 /*
579  * Flags for the t_oobflags field.
580  */
581 #define	TCPOOB_HAVEDATA	0x01
582 #define	TCPOOB_HADDATA	0x02
583 
584 /*
585  * Flags for the extended TCP flags field, t_flags2
586  */
587 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
588 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
589 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
590 #define	TF2_LOG_AUTO		0x00000008 /* Session is auto-logging. */
591 #define	TF2_DROP_AF_DATA	0x00000010 /* Drop after all data ack'd */
592 #define	TF2_ECN_PERMIT		0x00000020 /* connection ECN-ready */
593 #define	TF2_ECN_SND_CWR		0x00000040 /* ECN CWR in queue */
594 #define	TF2_ECN_SND_ECE		0x00000080 /* ECN ECE in queue */
595 #define	TF2_ACE_PERMIT		0x00000100 /* Accurate ECN mode */
596 #define	TF2_FBYTES_COMPLETE	0x00000400 /* We have first bytes in and out */
597 #define	TF2_ECN_USE_ECT1	0x00000800 /* Use ECT(1) marking on session */
598 
599 /*
600  * Structure to hold TCP options that are only used during segment
601  * processing (in tcp_input), but not held in the tcpcb.
602  * It's basically used to reduce the number of parameters
603  * to tcp_dooptions and tcp_addoptions.
604  * The binary order of the to_flags is relevant for packing of the
605  * options in tcp_addoptions.
606  */
607 struct tcpopt {
608 	u_int32_t	to_flags;	/* which options are present */
609 #define	TOF_MSS		0x0001		/* maximum segment size */
610 #define	TOF_SCALE	0x0002		/* window scaling */
611 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
612 #define	TOF_TS		0x0010		/* timestamp */
613 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
614 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
615 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
616 #define	TOF_MAXOPT	0x0200
617 	u_int32_t	to_tsval;	/* new timestamp */
618 	u_int32_t	to_tsecr;	/* reflected timestamp */
619 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
620 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
621 	u_int8_t	*to_tfo_cookie; /* pointer to the TFO cookie */
622 	u_int16_t	to_mss;		/* maximum segment size */
623 	u_int8_t	to_wscale;	/* window scaling */
624 	u_int8_t	to_nsacks;	/* number of SACK blocks */
625 	u_int8_t	to_tfo_len;	/* TFO cookie length */
626 	u_int32_t	to_spare;	/* UTO */
627 };
628 
629 /*
630  * Flags for tcp_dooptions.
631  */
632 #define	TO_SYN		0x01		/* parse SYN-only options */
633 
634 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
635 	uint32_t	rmx_mtu;	/* MTU for this path */
636 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
637 	uint32_t	rmx_rtt;	/* estimated round trip time */
638 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
639 	uint32_t	rmx_cwnd;	/* congestion window */
640 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
641 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
642 };
643 
644 /*
645  * Used by tcp_maxmtu() to communicate interface specific features
646  * and limits at the time of connection setup.
647  */
648 struct tcp_ifcap {
649 	int	ifcap;
650 	u_int	tsomax;
651 	u_int	tsomaxsegcount;
652 	u_int	tsomaxsegsize;
653 };
654 
655 #ifndef _NETINET_IN_PCB_H_
656 struct in_conninfo;
657 #endif /* _NETINET_IN_PCB_H_ */
658 
659 /*
660  * The smoothed round-trip time and estimated variance
661  * are stored as fixed point numbers scaled by the values below.
662  * For convenience, these scales are also used in smoothing the average
663  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
664  * With these scales, srtt has 3 bits to the right of the binary point,
665  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
666  * binary point, and is smoothed with an ALPHA of 0.75.
667  */
668 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
669 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
670 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
671 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
672 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
673 
674 /*
675  * The initial retransmission should happen at rtt + 4 * rttvar.
676  * Because of the way we do the smoothing, srtt and rttvar
677  * will each average +1/2 tick of bias.  When we compute
678  * the retransmit timer, we want 1/2 tick of rounding and
679  * 1 extra tick because of +-1/2 tick uncertainty in the
680  * firing of the timer.  The bias will give us exactly the
681  * 1.5 tick we need.  But, because the bias is
682  * statistical, we have to test that we don't drop below
683  * the minimum feasible timer (which is 2 ticks).
684  * This version of the macro adapted from a paper by Lawrence
685  * Brakmo and Larry Peterson which outlines a problem caused
686  * by insufficient precision in the original implementation,
687  * which results in inappropriately large RTO values for very
688  * fast networks.
689  */
690 #define	TCP_REXMTVAL(tp) \
691 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
692 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
693 
694 /*
695  * TCP statistics.
696  * Many of these should be kept per connection,
697  * but that's inconvenient at the moment.
698  */
699 struct	tcpstat {
700 	uint64_t tcps_connattempt;	/* connections initiated */
701 	uint64_t tcps_accepts;		/* connections accepted */
702 	uint64_t tcps_connects;		/* connections established */
703 	uint64_t tcps_drops;		/* connections dropped */
704 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
705 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
706 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
707 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
708 	uint64_t tcps_rttupdated;	/* times we succeeded */
709 	uint64_t tcps_delack;		/* delayed acks sent */
710 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
711 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
712 	uint64_t tcps_persisttimeo;	/* persist timeouts */
713 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
714 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
715 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
716 	uint64_t tcps_progdrops;	/* drops due to no progress */
717 
718 	uint64_t tcps_sndtotal;		/* total packets sent */
719 	uint64_t tcps_sndpack;		/* data packets sent */
720 	uint64_t tcps_sndbyte;		/* data bytes sent */
721 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
722 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
723 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
724 	uint64_t tcps_sndacks;		/* ack-only packets sent */
725 	uint64_t tcps_sndprobe;		/* window probes sent */
726 	uint64_t tcps_sndurg;		/* packets sent with URG only */
727 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
728 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
729 
730 	uint64_t tcps_rcvtotal;		/* total packets received */
731 	uint64_t tcps_rcvpack;		/* packets received in sequence */
732 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
733 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
734 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
735 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
736 	uint64_t tcps_rcvshort;		/* packets received too short */
737 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
738 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
739 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
740 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
741 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
742 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
743 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
744 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
745 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
746 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
747 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
748 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
749 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
750 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
751 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
752 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
753 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
754 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
755 	uint64_t tcps_pcbcachemiss;
756 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
757 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
758 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
759 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
760 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
761 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
762 	uint64_t tcps_persistdrop;	/* timeout in persist state */
763 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
764 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
765 	uint64_t tcps_listendrop;	/* listen queue overflows */
766 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
767 
768 	uint64_t tcps_sc_added;		/* entry added to syncache */
769 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
770 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
771 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
772 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
773 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
774 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
775 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
776 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
777 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
778 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
779 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
780 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
781 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
782 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
783 
784 	uint64_t tcps_hc_added;		/* entry added to hostcache */
785 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
786 
787 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
788 
789 	/* SACK related stats */
790 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
791 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
792 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
793 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
794 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
795 	uint64_t tcps_sack_lostrexmt;	    /* SACK lost retransmission recovered */
796 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
797 
798 	/* ECN related stats */
799 	uint64_t tcps_ecn_rcvce;		/* ECN Congestion Experienced */
800 	uint64_t tcps_ecn_rcvect0;		/* ECN Capable Transport */
801 	uint64_t tcps_ecn_rcvect1;		/* ECN Capable Transport */
802 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
803 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
804 
805 	/* TCP_SIGNATURE related stats */
806 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
807 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
808 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
809 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
810 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
811 
812 	/* Path MTU Discovery Black Hole Detection related stats */
813 	uint64_t tcps_pmtud_blackhole_activated;	 /* Black Hole Count */
814 	uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
815 	uint64_t tcps_pmtud_blackhole_failed;		 /* Black Hole Failure Count */
816 
817 	uint64_t tcps_tunneled_pkts;	/* Packets encap's in UDP received */
818 	uint64_t tcps_tunneled_errs;	/* Packets that had errors that were UDP encaped */
819 
820 	/* Dsack related stats */
821 	uint64_t tcps_dsack_count;	/* Number of ACKs arriving with DSACKs */
822 	uint64_t tcps_dsack_bytes;	/* Number of bytes DSACK'ed no TLP */
823 	uint64_t tcps_dsack_tlp_bytes;	/* Number of bytes DSACK'ed due to TLPs */
824 
825 	/* TCPS_TIME_WAIT usage stats */
826 	uint64_t tcps_tw_recycles;	/* Times time-wait was recycled. */
827 	uint64_t tcps_tw_resets;	/* Times time-wait sent a reset. */
828 	uint64_t tcps_tw_responds;	/* Times time-wait sent a valid ack. */
829 
830 	/* Accurate ECN Handshake stats */
831 	uint64_t tcps_ace_nect;		/* ACE SYN packet with Non-ECT */
832 	uint64_t tcps_ace_ect1;		/* ACE SYN packet with ECT1 */
833 	uint64_t tcps_ace_ect0;		/* ACE SYN packet with ECT0 */
834 	uint64_t tcps_ace_ce;		/* ACE SYN packet with CE */
835 
836 	/* ECN related stats */
837 	uint64_t tcps_ecn_sndect0;		/* ECN Capable Transport */
838 	uint64_t tcps_ecn_sndect1;		/* ECN Capable Transport */
839 
840 	uint64_t _pad[4];		/* 4 TBD placeholder for STABLE */
841 };
842 
843 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
844 
845 #ifdef _KERNEL
846 #define	TI_UNLOCKED	1
847 #define	TI_RLOCKED	2
848 #include <sys/counter.h>
849 
850 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
851 /*
852  * In-kernel consumers can use these accessor macros directly to update
853  * stats.
854  */
855 #define	TCPSTAT_ADD(name, val)	\
856     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
857 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
858 
859 /*
860  * Kernel module consumers must use this accessor macro.
861  */
862 void	kmod_tcpstat_add(int statnum, int val);
863 #define	KMOD_TCPSTAT_ADD(name, val)					\
864     kmod_tcpstat_add(offsetof(struct tcpstat, name) / sizeof(uint64_t), val)
865 #define	KMOD_TCPSTAT_INC(name)	KMOD_TCPSTAT_ADD(name, 1)
866 
867 /*
868  * Running TCP connection count by state.
869  */
870 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
871 #define	V_tcps_states	VNET(tcps_states)
872 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
873 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
874 
875 /*
876  * TCP specific helper hook point identifiers.
877  */
878 #define	HHOOK_TCP_EST_IN		0
879 #define	HHOOK_TCP_EST_OUT		1
880 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
881 
882 struct tcp_hhook_data {
883 	struct tcpcb	*tp;
884 	struct tcphdr	*th;
885 	struct tcpopt	*to;
886 	uint32_t	len;
887 	int		tso;
888 	tcp_seq		curack;
889 };
890 #ifdef TCP_HHOOK
891 void hhook_run_tcp_est_out(struct tcpcb *tp,
892 	struct tcphdr *th, struct tcpopt *to,
893 	uint32_t len, int tso);
894 #endif
895 #endif
896 
897 /*
898  * TCB structure exported to user-land via sysctl(3).
899  *
900  * Fields prefixed with "xt_" are unique to the export structure, and fields
901  * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
902  *
903  * Legend:
904  * (s) - used by userland utilities in src
905  * (p) - used by utilities in ports
906  * (3) - is known to be used by third party software not in ports
907  * (n) - no known usage
908  *
909  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
910  * included.  Not all of our clients do.
911  */
912 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
913 struct xtcpcb {
914 	ksize_t	xt_len;		/* length of this structure */
915 	struct xinpcb	xt_inp;
916 	char		xt_stack[TCP_FUNCTION_NAME_LEN_MAX];	/* (s) */
917 	char		xt_logid[TCP_LOG_ID_LEN];	/* (s) */
918 	char		xt_cc[TCP_CA_NAME_MAX];	/* (s) */
919 	int64_t		spare64[6];
920 	int32_t		t_state;		/* (s,p) */
921 	uint32_t	t_flags;		/* (s,p) */
922 	int32_t		t_sndzerowin;		/* (s) */
923 	int32_t		t_sndrexmitpack;	/* (s) */
924 	int32_t		t_rcvoopack;		/* (s) */
925 	int32_t		t_rcvtime;		/* (s) */
926 	int32_t		tt_rexmt;		/* (s) */
927 	int32_t		tt_persist;		/* (s) */
928 	int32_t		tt_keep;		/* (s) */
929 	int32_t		tt_2msl;		/* (s) */
930 	int32_t		tt_delack;		/* (s) */
931 	int32_t		t_logstate;		/* (3) */
932 	uint32_t	t_snd_cwnd;		/* (s) */
933 	uint32_t	t_snd_ssthresh;		/* (s) */
934 	uint32_t	t_maxseg;		/* (s) */
935 	uint32_t	t_rcv_wnd;		/* (s) */
936 	uint32_t	t_snd_wnd;		/* (s) */
937 	uint32_t	xt_ecn;			/* (s) */
938 	uint32_t	t_dsack_bytes;		/* (n) */
939 	uint32_t	t_dsack_tlp_bytes;	/* (n) */
940 	uint32_t	t_dsack_pack;		/* (n) */
941 	uint16_t	xt_encaps_port;		/* (s) */
942 	int16_t		spare16;
943 	int32_t		spare32[22];
944 } __aligned(8);
945 
946 #ifdef _KERNEL
947 void	tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
948 #endif
949 #endif
950 
951 /*
952  * TCP function information (name-to-id mapping, aliases, and refcnt)
953  * exported to user-land via sysctl(3).
954  */
955 struct tcp_function_info {
956 	uint32_t	tfi_refcnt;
957 	uint8_t		tfi_id;
958 	char		tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
959 	char		tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
960 };
961 
962 /*
963  * Identifiers for TCP sysctl nodes
964  */
965 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
966 #define	TCPCTL_MSSDFLT		3	/* MSS default */
967 #define TCPCTL_STATS		4	/* statistics */
968 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
969 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
970 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
971 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
972 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
973 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
974 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
975 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
976 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
977 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
978 #define	TCPCTL_DROP		15	/* drop tcp connection */
979 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
980 
981 #ifdef _KERNEL
982 #ifdef SYSCTL_DECL
983 SYSCTL_DECL(_net_inet_tcp);
984 SYSCTL_DECL(_net_inet_tcp_sack);
985 MALLOC_DECLARE(M_TCPLOG);
986 #endif
987 
988 VNET_DECLARE(int, tcp_log_in_vain);
989 #define	V_tcp_log_in_vain		VNET(tcp_log_in_vain)
990 
991 /*
992  * Global TCP tunables shared between different stacks.
993  * Please keep the list sorted.
994  */
995 VNET_DECLARE(int, drop_synfin);
996 VNET_DECLARE(int, path_mtu_discovery);
997 VNET_DECLARE(int, tcp_abc_l_var);
998 VNET_DECLARE(int, tcp_autorcvbuf_max);
999 VNET_DECLARE(int, tcp_autosndbuf_inc);
1000 VNET_DECLARE(int, tcp_autosndbuf_max);
1001 VNET_DECLARE(int, tcp_delack_enabled);
1002 VNET_DECLARE(int, tcp_do_autorcvbuf);
1003 VNET_DECLARE(int, tcp_do_autosndbuf);
1004 VNET_DECLARE(int, tcp_do_ecn);
1005 VNET_DECLARE(int, tcp_do_lrd);
1006 VNET_DECLARE(int, tcp_do_prr);
1007 VNET_DECLARE(int, tcp_do_prr_conservative);
1008 VNET_DECLARE(int, tcp_do_newcwv);
1009 VNET_DECLARE(int, tcp_do_rfc1323);
1010 VNET_DECLARE(int, tcp_tolerate_missing_ts);
1011 VNET_DECLARE(int, tcp_do_rfc3042);
1012 VNET_DECLARE(int, tcp_do_rfc3390);
1013 VNET_DECLARE(int, tcp_do_rfc3465);
1014 VNET_DECLARE(int, tcp_do_newsack);
1015 VNET_DECLARE(int, tcp_do_sack);
1016 VNET_DECLARE(int, tcp_do_tso);
1017 VNET_DECLARE(int, tcp_ecn_maxretries);
1018 VNET_DECLARE(int, tcp_initcwnd_segments);
1019 VNET_DECLARE(int, tcp_insecure_rst);
1020 VNET_DECLARE(int, tcp_insecure_syn);
1021 VNET_DECLARE(uint32_t, tcp_map_entries_limit);
1022 VNET_DECLARE(uint32_t, tcp_map_split_limit);
1023 VNET_DECLARE(int, tcp_minmss);
1024 VNET_DECLARE(int, tcp_mssdflt);
1025 #ifdef STATS
1026 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl);
1027 VNET_DECLARE(int, tcp_perconn_stats_enable);
1028 #endif /* STATS */
1029 VNET_DECLARE(int, tcp_recvspace);
1030 VNET_DECLARE(int, tcp_sack_globalholes);
1031 VNET_DECLARE(int, tcp_sack_globalmaxholes);
1032 VNET_DECLARE(int, tcp_sack_maxholes);
1033 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
1034 VNET_DECLARE(int, tcp_sendspace);
1035 VNET_DECLARE(int, tcp_udp_tunneling_overhead);
1036 VNET_DECLARE(int, tcp_udp_tunneling_port);
1037 VNET_DECLARE(struct inpcbinfo, tcbinfo);
1038 
1039 #define	V_tcp_do_lrd			VNET(tcp_do_lrd)
1040 #define	V_tcp_do_prr			VNET(tcp_do_prr)
1041 #define	V_tcp_do_prr_conservative	VNET(tcp_do_prr_conservative)
1042 #define	V_tcp_do_newcwv			VNET(tcp_do_newcwv)
1043 #define	V_drop_synfin			VNET(drop_synfin)
1044 #define	V_path_mtu_discovery		VNET(path_mtu_discovery)
1045 #define	V_tcbinfo			VNET(tcbinfo)
1046 #define	V_tcp_abc_l_var			VNET(tcp_abc_l_var)
1047 #define	V_tcp_autorcvbuf_max		VNET(tcp_autorcvbuf_max)
1048 #define	V_tcp_autosndbuf_inc		VNET(tcp_autosndbuf_inc)
1049 #define	V_tcp_autosndbuf_max		VNET(tcp_autosndbuf_max)
1050 #define	V_tcp_delack_enabled		VNET(tcp_delack_enabled)
1051 #define	V_tcp_do_autorcvbuf		VNET(tcp_do_autorcvbuf)
1052 #define	V_tcp_do_autosndbuf		VNET(tcp_do_autosndbuf)
1053 #define	V_tcp_do_ecn			VNET(tcp_do_ecn)
1054 #define	V_tcp_do_rfc1323		VNET(tcp_do_rfc1323)
1055 #define	V_tcp_tolerate_missing_ts	VNET(tcp_tolerate_missing_ts)
1056 #define V_tcp_ts_offset_per_conn	VNET(tcp_ts_offset_per_conn)
1057 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
1058 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)
1059 #define	V_tcp_do_rfc3465		VNET(tcp_do_rfc3465)
1060 #define	V_tcp_do_newsack		VNET(tcp_do_newsack)
1061 #define	V_tcp_do_sack			VNET(tcp_do_sack)
1062 #define	V_tcp_do_tso			VNET(tcp_do_tso)
1063 #define	V_tcp_ecn_maxretries		VNET(tcp_ecn_maxretries)
1064 #define	V_tcp_initcwnd_segments		VNET(tcp_initcwnd_segments)
1065 #define	V_tcp_insecure_rst		VNET(tcp_insecure_rst)
1066 #define	V_tcp_insecure_syn		VNET(tcp_insecure_syn)
1067 #define	V_tcp_map_entries_limit		VNET(tcp_map_entries_limit)
1068 #define	V_tcp_map_split_limit		VNET(tcp_map_split_limit)
1069 #define	V_tcp_minmss			VNET(tcp_minmss)
1070 #define	V_tcp_mssdflt			VNET(tcp_mssdflt)
1071 #ifdef STATS
1072 #define	V_tcp_perconn_stats_dflt_tpl	VNET(tcp_perconn_stats_dflt_tpl)
1073 #define	V_tcp_perconn_stats_enable	VNET(tcp_perconn_stats_enable)
1074 #endif /* STATS */
1075 #define	V_tcp_recvspace			VNET(tcp_recvspace)
1076 #define	V_tcp_sack_globalholes		VNET(tcp_sack_globalholes)
1077 #define	V_tcp_sack_globalmaxholes	VNET(tcp_sack_globalmaxholes)
1078 #define	V_tcp_sack_maxholes		VNET(tcp_sack_maxholes)
1079 #define	V_tcp_sc_rst_sock_fail		VNET(tcp_sc_rst_sock_fail)
1080 #define	V_tcp_sendspace			VNET(tcp_sendspace)
1081 #define	V_tcp_udp_tunneling_overhead	VNET(tcp_udp_tunneling_overhead)
1082 #define	V_tcp_udp_tunneling_port	VNET(tcp_udp_tunneling_port)
1083 
1084 #ifdef TCP_HHOOK
1085 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
1086 #define	V_tcp_hhh		VNET(tcp_hhh)
1087 #endif
1088 
1089 int	 tcp_addoptions(struct tcpopt *, u_char *);
1090 struct tcpcb *
1091 	 tcp_close(struct tcpcb *);
1092 void	 tcp_discardcb(struct tcpcb *);
1093 void	 tcp_twstart(struct tcpcb *);
1094 int	 tcp_ctloutput(struct socket *, struct sockopt *);
1095 void	 tcp_fini(void *);
1096 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, const void *,
1097 	    const void *);
1098 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, const void *,
1099 	    const void *);
1100 int	 tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *,
1101 	    struct mbuf *);
1102 void	 tcp_reass_global_init(void);
1103 void	 tcp_reass_flush(struct tcpcb *);
1104 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
1105 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
1106 		     struct tcpcb *, int, int);
1107 void	tcp_pulloutofband(struct socket *,
1108 		     struct tcphdr *, struct mbuf *, int);
1109 void	tcp_xmit_timer(struct tcpcb *, int);
1110 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
1111 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
1112 			    uint16_t nsegs, uint16_t type);
1113 void 	cc_conn_init(struct tcpcb *tp);
1114 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
1115 void    cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos);
1116 void	cc_ecnpkt_handler_flags(struct tcpcb *tp, uint16_t flags, uint8_t iptos);
1117 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
1118 #ifdef TCP_HHOOK
1119 void	hhook_run_tcp_est_in(struct tcpcb *tp,
1120 			    struct tcphdr *th, struct tcpopt *to);
1121 #endif
1122 
1123 int	 tcp_input(struct mbuf **, int *, int);
1124 int	 tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
1125 	    struct tcpcb *, int);
1126 int	 tcp_input_with_port(struct mbuf **, int *, int, uint16_t);
1127 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
1128 			struct socket *, struct tcpcb *, int, int, uint8_t);
1129 
1130 int register_tcp_functions(struct tcp_function_block *blk, int wait);
1131 int register_tcp_functions_as_names(struct tcp_function_block *blk,
1132     int wait, const char *names[], int *num_names);
1133 int register_tcp_functions_as_name(struct tcp_function_block *blk,
1134     const char *name, int wait);
1135 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
1136     bool force);
1137 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
1138 int find_tcp_function_alias(struct tcp_function_block *blk, struct tcp_function_set *fs);
1139 void tcp_switch_back_to_default(struct tcpcb *tp);
1140 struct tcp_function_block *
1141 find_and_ref_tcp_fb(struct tcp_function_block *fs);
1142 int tcp_default_ctloutput(struct inpcb *inp, struct sockopt *sopt);
1143 int tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt);
1144 
1145 extern counter_u64_t tcp_inp_lro_direct_queue;
1146 extern counter_u64_t tcp_inp_lro_wokeup_queue;
1147 extern counter_u64_t tcp_inp_lro_compressed;
1148 extern counter_u64_t tcp_inp_lro_locks_taken;
1149 extern counter_u64_t tcp_extra_mbuf;
1150 extern counter_u64_t tcp_would_have_but;
1151 extern counter_u64_t tcp_comp_total;
1152 extern counter_u64_t tcp_uncomp_total;
1153 extern counter_u64_t tcp_bad_csums;
1154 
1155 #ifdef NETFLIX_EXP_DETECTION
1156 /* Various SACK attack thresholds */
1157 extern int32_t tcp_force_detection;
1158 extern int32_t tcp_sack_to_ack_thresh;
1159 extern int32_t tcp_sack_to_move_thresh;
1160 extern int32_t tcp_restoral_thresh;
1161 extern int32_t tcp_sad_decay_val;
1162 extern int32_t tcp_sad_pacing_interval;
1163 extern int32_t tcp_sad_low_pps;
1164 extern int32_t tcp_map_minimum;
1165 extern int32_t tcp_attack_on_turns_on_logging;
1166 #endif
1167 extern uint32_t tcp_ack_war_time_window;
1168 extern uint32_t tcp_ack_war_cnt;
1169 
1170 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
1171 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
1172 void	 tcp6_use_min_mtu(struct tcpcb *);
1173 u_int	 tcp_maxseg(const struct tcpcb *);
1174 u_int	 tcp_fixed_maxseg(const struct tcpcb *);
1175 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
1176 	    struct tcp_ifcap *);
1177 void	 tcp_mss(struct tcpcb *, int);
1178 int	 tcp_mssopt(struct in_conninfo *);
1179 struct tcpcb *
1180 	 tcp_newtcpcb(struct inpcb *);
1181 int	 tcp_default_output(struct tcpcb *);
1182 void	 tcp_state_change(struct tcpcb *, int);
1183 void	 tcp_respond(struct tcpcb *, void *,
1184 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, uint16_t);
1185 bool	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
1186 	    struct mbuf *, int);
1187 void	 tcp_setpersist(struct tcpcb *);
1188 void	 tcp_record_dsack(struct tcpcb *tp, tcp_seq start, tcp_seq end, int tlp);
1189 struct tcptemp *
1190 	 tcpip_maketemplate(struct inpcb *);
1191 void	 tcpip_fillheaders(struct inpcb *, uint16_t, void *, void *);
1192 void	 tcp_timer_activate(struct tcpcb *, tt_which, u_int);
1193 bool	 tcp_timer_active(struct tcpcb *, tt_which);
1194 void	 tcp_timer_stop(struct tcpcb *);
1195 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
1196 int	 inp_to_cpuid(struct inpcb *inp);
1197 /*
1198  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
1199  */
1200 void	 tcp_hc_init(void);
1201 #ifdef VIMAGE
1202 void	 tcp_hc_destroy(void);
1203 #endif
1204 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
1205 uint32_t tcp_hc_getmtu(struct in_conninfo *);
1206 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
1207 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
1208 void 	 cc_after_idle(struct tcpcb *tp);
1209 
1210 extern	struct protosw tcp_protosw;		/* shared for TOE */
1211 extern	struct protosw tcp6_protosw;		/* shared for TOE */
1212 
1213 uint32_t tcp_new_ts_offset(struct in_conninfo *);
1214 tcp_seq	 tcp_new_isn(struct in_conninfo *);
1215 
1216 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
1217 int	 tcp_dsack_block_exists(struct tcpcb *);
1218 void	 tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
1219 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1220 void	 tcp_clean_dsack_blocks(struct tcpcb *tp);
1221 void	 tcp_clean_sackreport(struct tcpcb *tp);
1222 void	 tcp_sack_adjust(struct tcpcb *tp);
1223 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1224 void	 tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *);
1225 void	 tcp_lost_retransmission(struct tcpcb *, struct tcphdr *);
1226 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1227 void	 tcp_free_sackholes(struct tcpcb *tp);
1228 void	 tcp_sack_lost_retransmission(struct tcpcb *, struct tcphdr *);
1229 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
1230 int	 tcp_compute_pipe(struct tcpcb *);
1231 uint32_t tcp_compute_initwnd(uint32_t);
1232 void	 tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
1233 int	 tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes,
1234     size_t seed_len);
1235 int tcp_can_enable_pacing(void);
1236 void tcp_decrement_paced_conn(void);
1237 
1238 struct mbuf *
1239 	 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
1240 	   int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls);
1241 
1242 int	tcp_stats_init(void);
1243 void tcp_log_end_status(struct tcpcb *tp, uint8_t status);
1244 
1245 static inline void
1246 tcp_fields_to_host(struct tcphdr *th)
1247 {
1248 
1249 	th->th_seq = ntohl(th->th_seq);
1250 	th->th_ack = ntohl(th->th_ack);
1251 	th->th_win = ntohs(th->th_win);
1252 	th->th_urp = ntohs(th->th_urp);
1253 }
1254 
1255 static inline void
1256 tcp_fields_to_net(struct tcphdr *th)
1257 {
1258 
1259 	th->th_seq = htonl(th->th_seq);
1260 	th->th_ack = htonl(th->th_ack);
1261 	th->th_win = htons(th->th_win);
1262 	th->th_urp = htons(th->th_urp);
1263 }
1264 
1265 static inline uint16_t
1266 tcp_get_flags(const struct tcphdr *th)
1267 {
1268         return (((uint16_t)th->th_x2 << 8) | th->th_flags);
1269 }
1270 
1271 static inline void
1272 tcp_set_flags(struct tcphdr *th, uint16_t flags)
1273 {
1274         th->th_x2    = (flags >> 8) & 0x0f;
1275         th->th_flags = flags & 0xff;
1276 }
1277 
1278 static inline void
1279 tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt,
1280     uint8_t is_tlp, int hw_tls)
1281 {
1282 	if (is_tlp) {
1283 		tp->t_sndtlppack++;
1284 		tp->t_sndtlpbyte += len;
1285 	}
1286 	/* To get total bytes sent you must add t_snd_rxt_bytes to t_sndbytes */
1287 	if (is_rxt)
1288 		tp->t_snd_rxt_bytes += len;
1289 	else
1290 		tp->t_sndbytes += len;
1291 
1292 #ifdef KERN_TLS
1293 	if (hw_tls && is_rxt && len != 0) {
1294 		uint64_t rexmit_percent = (1000ULL * tp->t_snd_rxt_bytes) / (10ULL * (tp->t_snd_rxt_bytes + tp->t_sndbytes));
1295 		if (rexmit_percent > ktls_ifnet_max_rexmit_pct)
1296 			ktls_disable_ifnet(tp);
1297 	}
1298 #endif
1299 
1300 }
1301 #endif /* _KERNEL */
1302 
1303 #endif /* _NETINET_TCP_VAR_H_ */
1304