xref: /freebsd/sys/netinet/tcp_var.h (revision 7cd22ac43418da08448d0bab1009ff3cbda85120)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1993, 1994, 1995
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
32  * $FreeBSD$
33  */
34 
35 #ifndef _NETINET_TCP_VAR_H_
36 #define _NETINET_TCP_VAR_H_
37 
38 #include <netinet/tcp.h>
39 #include <netinet/tcp_fsm.h>
40 
41 #ifdef _KERNEL
42 #include "opt_kern_tls.h"
43 #include <net/vnet.h>
44 #include <sys/mbuf.h>
45 #include <sys/ktls.h>
46 #endif
47 
48 #define TCP_END_BYTE_INFO 8	/* Bytes that makeup the "end information array" */
49 /* Types of ending byte info */
50 #define TCP_EI_EMPTY_SLOT	0
51 #define TCP_EI_STATUS_CLIENT_FIN	0x1
52 #define TCP_EI_STATUS_CLIENT_RST	0x2
53 #define TCP_EI_STATUS_SERVER_FIN	0x3
54 #define TCP_EI_STATUS_SERVER_RST	0x4
55 #define TCP_EI_STATUS_RETRAN		0x5
56 #define TCP_EI_STATUS_PROGRESS		0x6
57 #define TCP_EI_STATUS_PERSIST_MAX	0x7
58 #define TCP_EI_STATUS_KEEP_MAX		0x8
59 #define TCP_EI_STATUS_DATA_A_CLOSE	0x9
60 #define TCP_EI_STATUS_RST_IN_FRONT	0xa
61 #define TCP_EI_STATUS_2MSL		0xb
62 #define TCP_EI_STATUS_MAX_VALUE		0xb
63 
64 /************************************************/
65 /* Status bits we track to assure no duplicates,
66  * the bits here are not used by the code but
67  * for human representation. To check a bit we
68  * take and shift over by 1 minus the value (1-8).
69  */
70 /************************************************/
71 #define TCP_EI_BITS_CLIENT_FIN	0x001
72 #define TCP_EI_BITS_CLIENT_RST	0x002
73 #define TCP_EI_BITS_SERVER_FIN	0x004
74 #define TCP_EI_BITS_SERVER_RST	0x008
75 #define TCP_EI_BITS_RETRAN	0x010
76 #define TCP_EI_BITS_PROGRESS	0x020
77 #define TCP_EI_BITS_PRESIST_MAX	0x040
78 #define TCP_EI_BITS_KEEP_MAX	0x080
79 #define TCP_EI_BITS_DATA_A_CLO  0x100
80 #define TCP_EI_BITS_RST_IN_FR	0x200	/* a front state reset */
81 #define TCP_EI_BITS_2MS_TIMER	0x400	/* 2 MSL timer expired */
82 
83 #if defined(_KERNEL) || defined(_WANT_TCPCB)
84 /* TCP segment queue entry */
85 struct tseg_qent {
86 	TAILQ_ENTRY(tseg_qent) tqe_q;
87 	struct	mbuf   *tqe_m;		/* mbuf contains packet */
88 	struct  mbuf   *tqe_last;	/* last mbuf in chain */
89 	tcp_seq tqe_start;		/* TCP Sequence number start */
90 	int	tqe_len;		/* TCP segment data length */
91 	uint32_t tqe_flags;		/* The flags from the th->th_flags */
92 	uint32_t tqe_mbuf_cnt;		/* Count of mbuf overhead */
93 };
94 TAILQ_HEAD(tsegqe_head, tseg_qent);
95 
96 struct sackblk {
97 	tcp_seq start;		/* start seq no. of sack block */
98 	tcp_seq end;		/* end seq no. */
99 };
100 
101 struct sackhole {
102 	tcp_seq start;		/* start seq no. of hole */
103 	tcp_seq end;		/* end seq no. */
104 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
105 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
106 };
107 
108 struct sackhint {
109 	struct sackhole	*nexthole;
110 	int32_t		sack_bytes_rexmit;
111 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
112 
113 	int32_t		delivered_data; /* Newly acked data from last SACK */
114 
115 	int32_t		sacked_bytes;	/* Total sacked bytes reported by the
116 					 * receiver via sack option
117 					 */
118 	uint32_t	recover_fs;	/* Flight Size at the start of Loss recovery */
119 	uint32_t	prr_delivered;	/* Total bytes delivered using PRR */
120 	uint32_t	prr_out;	/* Bytes sent during IN_RECOVERY */
121 };
122 
123 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
124 
125 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
126 
127 /*
128  * Tcp control block, one per tcp; fields:
129  * Organized for 64 byte cacheline efficiency based
130  * on common tcp_input/tcp_output processing.
131  */
132 struct tcpcb {
133 	/* Cache line 1 */
134 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
135 	struct tcp_function_block *t_fb;/* TCP function call block */
136 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
137 	uint32_t t_maxseg:24,		/* maximum segment size */
138 		t_logstate:8;		/* State of "black box" logging */
139 	uint32_t t_port:16,		/* Tunneling (over udp) port */
140 		t_state:4,		/* state of this connection */
141 		t_idle_reduce : 1,
142 		t_delayed_ack: 7,	/* Delayed ack variable */
143 		t_fin_is_rst: 1,	/* Are fin's treated as resets */
144 		t_log_state_set: 1,
145 		bits_spare : 2;
146 	u_int	t_flags;
147 	tcp_seq	snd_una;		/* sent but unacknowledged */
148 	tcp_seq	snd_max;		/* highest sequence number sent;
149 					 * used to recognize retransmits
150 					 */
151 	tcp_seq	snd_nxt;		/* send next */
152 	tcp_seq	snd_up;			/* send urgent pointer */
153 	uint32_t  snd_wnd;		/* send window */
154 	uint32_t  snd_cwnd;		/* congestion-controlled window */
155 	uint32_t t_peakrate_thr; 	/* pre-calculated peak rate threshold */
156 	/* Cache line 2 */
157 	u_int32_t  ts_offset;		/* our timestamp offset */
158 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
159 	int	rcv_numsacks;		/* # distinct sack blks present */
160 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
161 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
162 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
163 	tcp_seq	rcv_nxt;		/* receive next */
164 	tcp_seq	rcv_adv;		/* advertised window */
165 	uint32_t  rcv_wnd;		/* receive window */
166 	u_int	t_flags2;		/* More tcpcb flags storage */
167 	int	t_srtt;			/* smoothed round-trip time */
168 	int	t_rttvar;		/* variance in round-trip time */
169 	u_int32_t  ts_recent;		/* timestamp echo data */
170 	u_char	snd_scale;		/* window scaling for send window */
171 	u_char	rcv_scale;		/* window scaling for recv window */
172 	u_char	snd_limited;		/* segments limited transmitted */
173 	u_char	request_r_scale;	/* pending window scaling */
174 	tcp_seq	last_ack_sent;
175 	u_int	t_rcvtime;		/* inactivity time */
176 	/* Cache line 3 */
177 	tcp_seq	rcv_up;			/* receive urgent pointer */
178 	int	t_segqlen;		/* segment reassembly queue length */
179 	uint32_t t_segqmbuflen;		/* Count of bytes mbufs on all entries */
180 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
181 	struct mbuf      *t_in_pkt;
182 	struct mbuf	 *t_tail_pkt;
183 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
184 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
185 	uint32_t  snd_ssthresh;		/* snd_cwnd size threshold for
186 					 * for slow start exponential to
187 					 * linear switch
188 					 */
189 	tcp_seq	snd_wl1;		/* window update seg seq number */
190 	/* Cache line 4 */
191 	tcp_seq	snd_wl2;		/* window update seg ack number */
192 
193 	tcp_seq	irs;			/* initial receive sequence number */
194 	tcp_seq	iss;			/* initial send sequence number */
195 	u_int	t_acktime;		/* RACK and BBR incoming new data was acked */
196 	u_int	t_sndtime;		/* time last data was sent */
197 	u_int	ts_recent_age;		/* when last updated */
198 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
199 	uint16_t cl4_spare;		/* Spare to adjust CL 4 */
200 	char	t_oobflags;		/* have some */
201 	char	t_iobc;			/* input character */
202 	int	t_rxtcur;		/* current retransmit value (ticks) */
203 
204 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
205 	u_int	t_rtttime;		/* RTT measurement start time */
206 
207 	tcp_seq	t_rtseq;		/* sequence number being timed */
208 	u_int	t_starttime;		/* time connection was established */
209 	u_int	t_fbyte_in;		/* ticks time when first byte queued in */
210 	u_int	t_fbyte_out;		/* ticks time when first byte queued out */
211 
212 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
213 	int	t_blackhole_enter;	/* when to enter blackhole detection */
214 	int	t_blackhole_exit;	/* when to exit blackhole detection */
215 	u_int	t_rttmin;		/* minimum rtt allowed */
216 
217 	u_int	t_rttbest;		/* best rtt we've seen */
218 
219 	int	t_softerror;		/* possible error not yet reported */
220 	uint32_t  max_sndwnd;		/* largest window peer has offered */
221 	/* Cache line 5 */
222 	uint32_t  snd_cwnd_prev;	/* cwnd prior to retransmit */
223 	uint32_t  snd_ssthresh_prev;	/* ssthresh prior to retransmit */
224 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
225 	int	t_sndzerowin;		/* zero-window updates sent */
226 	u_long	t_rttupdated;		/* number of times rtt sampled */
227 	int	snd_numholes;		/* number of holes seen by sender */
228 	u_int	t_badrxtwin;		/* window for retransmit recovery */
229 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
230 					/* SACK scoreboard (sorted) */
231 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
232 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
233 	struct sackhint	sackhint;	/* SACK scoreboard hint */
234 	int	t_rttlow;		/* smallest observerved RTT */
235 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
236 	struct toedev	*tod;		/* toedev handling this connection */
237 	int	t_sndrexmitpack;	/* retransmit packets sent */
238 	int	t_rcvoopack;		/* out-of-order packets received */
239 	void	*t_toe;			/* TOE pcb pointer */
240 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
241 	struct cc_var	*ccv;		/* congestion control specific vars */
242 	struct osd	*osd;		/* storage for Khelp module data */
243 	int	t_bytes_acked;		/* # bytes acked during current RTT */
244 	u_int   t_maxunacktime;
245 	u_int	t_keepinit;		/* time to establish connection */
246 	u_int	t_keepidle;		/* time before keepalive probes begin */
247 	u_int	t_keepintvl;		/* interval between keepalives */
248 	u_int	t_keepcnt;		/* number of keepalives before close */
249 	int	t_dupacks;		/* consecutive dup acks recd */
250 	int	t_lognum;		/* Number of log entries */
251 	int	t_loglimit;		/* Maximum number of log entries */
252 	int64_t	t_pacing_rate;		/* bytes / sec, -1 => unlimited */
253 	struct tcp_log_stailq t_logs;	/* Log buffer */
254 	struct tcp_log_id_node *t_lin;
255 	struct tcp_log_id_bucket *t_lib;
256 	const char *t_output_caller;	/* Function that called tcp_output */
257 	struct statsblob *t_stats;	/* Per-connection stats */
258 	uint32_t t_logsn;		/* Log "serial number" */
259 	uint32_t gput_ts;		/* Time goodput measurement started */
260 	tcp_seq gput_seq;		/* Outbound measurement seq */
261 	tcp_seq gput_ack;		/* Inbound measurement ack */
262 	int32_t t_stats_gput_prev;	/* XXXLAS: Prev gput measurement */
263 	uint32_t t_maxpeakrate;		/* max peak rate set by user, in bytes/s */
264 	uint32_t t_sndtlppack;		/* tail loss probe packets sent */
265 	uint64_t t_sndtlpbyte;		/* total tail loss probe bytes sent */
266 	uint64_t t_sndbytes;		/* total bytes sent */
267 	uint64_t t_snd_rxt_bytes;	/* total bytes retransmitted */
268 
269 	uint8_t t_tfo_client_cookie_len; /* TCP Fast Open client cookie length */
270 	uint32_t t_end_info_status;	/* Status flag of end info */
271 	unsigned int *t_tfo_pending;	/* TCP Fast Open server pending counter */
272 	union {
273 		uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
274 		uint64_t server;
275 	} t_tfo_cookie;			/* TCP Fast Open cookie to send */
276 	union {
277 		uint8_t t_end_info_bytes[TCP_END_BYTE_INFO];
278 		uint64_t t_end_info;
279 	};
280 #ifdef TCPPCAP
281 	struct mbufq t_inpkts;		/* List of saved input packets. */
282 	struct mbufq t_outpkts;		/* List of saved output packets. */
283 #endif
284 };
285 #endif	/* _KERNEL || _WANT_TCPCB */
286 
287 #ifdef _KERNEL
288 struct tcptemp {
289 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
290 	struct	tcphdr tt_t;
291 };
292 
293 /* Enable TCP/UDP tunneling port */
294 #define TCP_TUNNELING_PORT_MIN		0
295 #define TCP_TUNNELING_PORT_MAX		65535
296 #define TCP_TUNNELING_PORT_DEFAULT	0
297 
298 /* Enable TCP/UDP tunneling port */
299 #define TCP_TUNNELING_OVERHEAD_MIN	sizeof(struct udphdr)
300 #define TCP_TUNNELING_OVERHEAD_MAX	1024
301 #define TCP_TUNNELING_OVERHEAD_DEFAULT	TCP_TUNNELING_OVERHEAD_MIN
302 
303 /* Minimum map entries limit value, if set */
304 #define TCP_MIN_MAP_ENTRIES_LIMIT	128
305 
306 /*
307  * TODO: We yet need to brave plowing in
308  * to tcp_input() and the pru_usrreq() block.
309  * Right now these go to the old standards which
310  * are somewhat ok, but in the long term may
311  * need to be changed. If we do tackle tcp_input()
312  * then we need to get rid of the tcp_do_segment()
313  * function below.
314  */
315 /* Flags for tcp functions */
316 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
317 
318 /*
319  * If defining the optional tcp_timers, in the
320  * tfb_tcp_timer_stop call you must use the
321  * callout_async_drain() function with the
322  * tcp_timer_discard callback. You should check
323  * the return of callout_async_drain() and if 0
324  * increment tt_draincnt. Since the timer sub-system
325  * does not know your callbacks you must provide a
326  * stop_all function that loops through and calls
327  * tcp_timer_stop() with each of your defined timers.
328  * Adding a tfb_tcp_handoff_ok function allows the socket
329  * option to change stacks to query you even if the
330  * connection is in a later stage. You return 0 to
331  * say you can take over and run your stack, you return
332  * non-zero (an error number) to say no you can't.
333  * If the function is undefined you can only change
334  * in the early states (before connect or listen).
335  * tfb_tcp_fb_fini is changed to add a flag to tell
336  * the old stack if the tcb is being destroyed or
337  * not. A one in the flag means the TCB is being
338  * destroyed, a zero indicates its transitioning to
339  * another stack (via socket option).
340  */
341 struct tcp_function_block {
342 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
343 	int	(*tfb_tcp_output)(struct tcpcb *);
344 	int	(*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
345 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
346 			    struct socket *, struct tcpcb *,
347 		        int, int, uint8_t);
348 	int     (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int);
349 	int      (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *,
350 			    struct socket *, struct tcpcb *,
351 			    int, int, uint8_t,
352 			    int, struct timeval *);
353 	void	(*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
354 			    struct socket *, struct tcpcb *,
355 			    int, int, uint8_t,
356 			    int, struct timeval *);
357 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
358 			    struct inpcb *inp, struct tcpcb *tp);
359 	/* Optional memory allocation/free routine */
360 	int	(*tfb_tcp_fb_init)(struct tcpcb *);
361 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
362 	/* Optional timers, must define all if you define one */
363 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
364 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
365 			    uint32_t, u_int);
366 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
367 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
368 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
369 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
370 	void	(*tfb_tcp_mtu_chg)(struct tcpcb *);
371 	int	(*tfb_pru_options)(struct tcpcb *, int);
372 	void	(*tfb_hwtls_change)(struct tcpcb *, int);
373 	volatile uint32_t tfb_refcnt;
374 	uint32_t  tfb_flags;
375 	uint8_t	tfb_id;
376 };
377 
378 struct tcp_function {
379 	TAILQ_ENTRY(tcp_function)	tf_next;
380 	char				tf_name[TCP_FUNCTION_NAME_LEN_MAX];
381 	struct tcp_function_block	*tf_fb;
382 };
383 
384 TAILQ_HEAD(tcp_funchead, tcp_function);
385 #endif	/* _KERNEL */
386 
387 /*
388  * Flags and utility macros for the t_flags field.
389  */
390 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
391 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
392 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
393 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
394 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
395 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
396 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
397 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
398 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
399 #define	TF_SACK_PERMIT	0x00000200	/* other side said I could SACK */
400 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
401 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
402 #define	TF_NOPUSH	0x00001000	/* don't push */
403 #define	TF_PREVVALID	0x00002000	/* saved values for bad rxmit valid */
404 #define	TF_WAKESOR	0x00004000	/* wake up receive socket */
405 #define	TF_GPUTINPROG	0x00008000	/* Goodput measurement in progress */
406 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
407 #define	TF_LQ_OVERFLOW	0x00020000	/* listen queue overflow */
408 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
409 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
410 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
411 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
412 #define	TF_SIGNATURE	0x00400000	/* require MD5 digests (RFC2385) */
413 #define	TF_FORCEDATA	0x00800000	/* force out a byte */
414 #define	TF_TSO		0x01000000	/* TSO enabled on this connection */
415 #define	TF_TOE		0x02000000	/* this connection is offloaded */
416 #define	TF_UNUSED0	0x04000000	/* unused */
417 #define	TF_UNUSED1	0x08000000	/* unused */
418 #define	TF_LRD		0x10000000	/* Lost Retransmission Detection */
419 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
420 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
421 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
422 
423 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
424 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
425 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
426 
427 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
428 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
429 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
430 
431 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
432 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
433 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
434 
435 #if defined(_KERNEL) && !defined(TCP_RFC7413)
436 #define	IS_FASTOPEN(t_flags)		(false)
437 #else
438 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
439 #endif
440 
441 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
442 
443 /*
444  * Flags for the t_oobflags field.
445  */
446 #define	TCPOOB_HAVEDATA	0x01
447 #define	TCPOOB_HADDATA	0x02
448 
449 /*
450  * Flags for the extended TCP flags field, t_flags2
451  */
452 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
453 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
454 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
455 #define	TF2_LOG_AUTO		0x00000008 /* Session is auto-logging. */
456 #define TF2_DROP_AF_DATA 	0x00000010 /* Drop after all data ack'd */
457 #define	TF2_ECN_PERMIT		0x00000020 /* connection ECN-ready */
458 #define	TF2_ECN_SND_CWR		0x00000040 /* ECN CWR in queue */
459 #define	TF2_ECN_SND_ECE		0x00000080 /* ECN ECE in queue */
460 #define	TF2_ACE_PERMIT		0x00000100 /* Accurate ECN mode */
461 #define TF2_FBYTES_COMPLETE	0x00000400 /* We have first bytes in and out */
462 /*
463  * Structure to hold TCP options that are only used during segment
464  * processing (in tcp_input), but not held in the tcpcb.
465  * It's basically used to reduce the number of parameters
466  * to tcp_dooptions and tcp_addoptions.
467  * The binary order of the to_flags is relevant for packing of the
468  * options in tcp_addoptions.
469  */
470 struct tcpopt {
471 	u_int32_t	to_flags;	/* which options are present */
472 #define	TOF_MSS		0x0001		/* maximum segment size */
473 #define	TOF_SCALE	0x0002		/* window scaling */
474 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
475 #define	TOF_TS		0x0010		/* timestamp */
476 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
477 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
478 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
479 #define	TOF_MAXOPT	0x0200
480 	u_int32_t	to_tsval;	/* new timestamp */
481 	u_int32_t	to_tsecr;	/* reflected timestamp */
482 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
483 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
484 	u_int8_t	*to_tfo_cookie; /* pointer to the TFO cookie */
485 	u_int16_t	to_mss;		/* maximum segment size */
486 	u_int8_t	to_wscale;	/* window scaling */
487 	u_int8_t	to_nsacks;	/* number of SACK blocks */
488 	u_int8_t	to_tfo_len;	/* TFO cookie length */
489 	u_int32_t	to_spare;	/* UTO */
490 };
491 
492 /*
493  * Flags for tcp_dooptions.
494  */
495 #define	TO_SYN		0x01		/* parse SYN-only options */
496 
497 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
498 	uint32_t	rmx_mtu;	/* MTU for this path */
499 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
500 	uint32_t	rmx_rtt;	/* estimated round trip time */
501 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
502 	uint32_t	rmx_cwnd;	/* congestion window */
503 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
504 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
505 };
506 
507 /*
508  * Used by tcp_maxmtu() to communicate interface specific features
509  * and limits at the time of connection setup.
510  */
511 struct tcp_ifcap {
512 	int	ifcap;
513 	u_int	tsomax;
514 	u_int	tsomaxsegcount;
515 	u_int	tsomaxsegsize;
516 };
517 
518 #ifndef _NETINET_IN_PCB_H_
519 struct in_conninfo;
520 #endif /* _NETINET_IN_PCB_H_ */
521 
522 struct tcptw {
523 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
524 	uint32_t  t_port:16,		/* UDP port number if TCPoUDP */
525 		t_unused:16;
526 	tcp_seq		snd_nxt;
527 	tcp_seq		rcv_nxt;
528 	tcp_seq		iss;
529 	tcp_seq		irs;
530 	u_short		last_win;	/* cached window value */
531 	short		tw_so_options;	/* copy of so_options */
532 	struct ucred	*tw_cred;	/* user credentials */
533 	u_int32_t	t_recent;
534 	u_int32_t	ts_offset;	/* our timestamp offset */
535 	u_int		t_starttime;
536 	int		tw_time;
537 	TAILQ_ENTRY(tcptw) tw_2msl;
538 	void		*tw_pspare;	/* TCP_SIGNATURE */
539 	u_int		*tw_spare;	/* TCP_SIGNATURE */
540 };
541 
542 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
543 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
544 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
545 
546 /*
547  * The smoothed round-trip time and estimated variance
548  * are stored as fixed point numbers scaled by the values below.
549  * For convenience, these scales are also used in smoothing the average
550  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
551  * With these scales, srtt has 3 bits to the right of the binary point,
552  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
553  * binary point, and is smoothed with an ALPHA of 0.75.
554  */
555 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
556 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
557 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
558 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
559 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
560 
561 /*
562  * The initial retransmission should happen at rtt + 4 * rttvar.
563  * Because of the way we do the smoothing, srtt and rttvar
564  * will each average +1/2 tick of bias.  When we compute
565  * the retransmit timer, we want 1/2 tick of rounding and
566  * 1 extra tick because of +-1/2 tick uncertainty in the
567  * firing of the timer.  The bias will give us exactly the
568  * 1.5 tick we need.  But, because the bias is
569  * statistical, we have to test that we don't drop below
570  * the minimum feasible timer (which is 2 ticks).
571  * This version of the macro adapted from a paper by Lawrence
572  * Brakmo and Larry Peterson which outlines a problem caused
573  * by insufficient precision in the original implementation,
574  * which results in inappropriately large RTO values for very
575  * fast networks.
576  */
577 #define	TCP_REXMTVAL(tp) \
578 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
579 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
580 
581 /*
582  * TCP statistics.
583  * Many of these should be kept per connection,
584  * but that's inconvenient at the moment.
585  */
586 struct	tcpstat {
587 	uint64_t tcps_connattempt;	/* connections initiated */
588 	uint64_t tcps_accepts;		/* connections accepted */
589 	uint64_t tcps_connects;		/* connections established */
590 	uint64_t tcps_drops;		/* connections dropped */
591 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
592 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
593 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
594 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
595 	uint64_t tcps_rttupdated;	/* times we succeeded */
596 	uint64_t tcps_delack;		/* delayed acks sent */
597 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
598 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
599 	uint64_t tcps_persisttimeo;	/* persist timeouts */
600 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
601 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
602 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
603 
604 	uint64_t tcps_sndtotal;		/* total packets sent */
605 	uint64_t tcps_sndpack;		/* data packets sent */
606 	uint64_t tcps_sndbyte;		/* data bytes sent */
607 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
608 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
609 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
610 	uint64_t tcps_sndacks;		/* ack-only packets sent */
611 	uint64_t tcps_sndprobe;		/* window probes sent */
612 	uint64_t tcps_sndurg;		/* packets sent with URG only */
613 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
614 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
615 
616 	uint64_t tcps_rcvtotal;		/* total packets received */
617 	uint64_t tcps_rcvpack;		/* packets received in sequence */
618 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
619 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
620 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
621 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
622 	uint64_t tcps_rcvshort;		/* packets received too short */
623 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
624 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
625 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
626 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
627 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
628 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
629 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
630 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
631 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
632 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
633 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
634 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
635 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
636 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
637 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
638 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
639 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
640 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
641 	uint64_t tcps_pcbcachemiss;
642 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
643 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
644 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
645 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
646 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
647 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
648 	uint64_t tcps_persistdrop;	/* timeout in persist state */
649 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
650 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
651 	uint64_t tcps_listendrop;	/* listen queue overflows */
652 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
653 
654 	uint64_t tcps_sc_added;		/* entry added to syncache */
655 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
656 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
657 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
658 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
659 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
660 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
661 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
662 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
663 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
664 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
665 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
666 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
667 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
668 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
669 
670 	uint64_t tcps_hc_added;		/* entry added to hostcache */
671 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
672 
673 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
674 
675 	/* SACK related stats */
676 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
677 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
678 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
679 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
680 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
681 	uint64_t tcps_sack_lostrexmt;	    /* SACK lost retransmission recovered */
682 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
683 
684 	/* ECN related stats */
685 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
686 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
687 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
688 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
689 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
690 
691 	/* TCP_SIGNATURE related stats */
692 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
693 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
694 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
695 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
696 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
697 
698 	/* Path MTU Discovery Black Hole Detection related stats */
699 	uint64_t tcps_pmtud_blackhole_activated;	 /* Black Hole Count */
700 	uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
701 	uint64_t tcps_pmtud_blackhole_failed;		 /* Black Hole Failure Count */
702 
703 	uint64_t tcps_tunneled_pkts;	/* Packets encap's in UDP received */
704 	uint64_t tcps_tunneled_errs;	/* Packets that had errors that were UDP encaped */
705 
706 	uint64_t _pad[9];		/* 6 UTO, 3 TBD */
707 };
708 
709 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
710 
711 #ifdef _KERNEL
712 #define	TI_UNLOCKED	1
713 #define	TI_RLOCKED	2
714 #include <sys/counter.h>
715 
716 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
717 /*
718  * In-kernel consumers can use these accessor macros directly to update
719  * stats.
720  */
721 #define	TCPSTAT_ADD(name, val)	\
722     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
723 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
724 
725 /*
726  * Kernel module consumers must use this accessor macro.
727  */
728 void	kmod_tcpstat_add(int statnum, int val);
729 #define	KMOD_TCPSTAT_ADD(name, val)					\
730     kmod_tcpstat_add(offsetof(struct tcpstat, name) / sizeof(uint64_t), val)
731 #define	KMOD_TCPSTAT_INC(name)	KMOD_TCPSTAT_ADD(name, 1)
732 
733 /*
734  * Running TCP connection count by state.
735  */
736 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
737 #define	V_tcps_states	VNET(tcps_states)
738 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
739 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
740 
741 /*
742  * TCP specific helper hook point identifiers.
743  */
744 #define	HHOOK_TCP_EST_IN		0
745 #define	HHOOK_TCP_EST_OUT		1
746 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
747 
748 struct tcp_hhook_data {
749 	struct tcpcb	*tp;
750 	struct tcphdr	*th;
751 	struct tcpopt	*to;
752 	uint32_t	len;
753 	int		tso;
754 	tcp_seq		curack;
755 };
756 #ifdef TCP_HHOOK
757 void hhook_run_tcp_est_out(struct tcpcb *tp,
758 	struct tcphdr *th, struct tcpopt *to,
759 	uint32_t len, int tso);
760 #endif
761 #endif
762 
763 /*
764  * TCB structure exported to user-land via sysctl(3).
765  *
766  * Fields prefixed with "xt_" are unique to the export structure, and fields
767  * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
768  *
769  * Legend:
770  * (s) - used by userland utilities in src
771  * (p) - used by utilities in ports
772  * (3) - is known to be used by third party software not in ports
773  * (n) - no known usage
774  *
775  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
776  * included.  Not all of our clients do.
777  */
778 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
779 struct xtcpcb {
780 	ksize_t	xt_len;		/* length of this structure */
781 	struct xinpcb	xt_inp;
782 	char		xt_stack[TCP_FUNCTION_NAME_LEN_MAX];	/* (s) */
783 	char		xt_logid[TCP_LOG_ID_LEN];	/* (s) */
784 	char		xt_cc[TCP_CA_NAME_MAX];	/* (s) */
785 	int64_t		spare64[6];
786 	int32_t		t_state;		/* (s,p) */
787 	uint32_t	t_flags;		/* (s,p) */
788 	int32_t		t_sndzerowin;		/* (s) */
789 	int32_t		t_sndrexmitpack;	/* (s) */
790 	int32_t		t_rcvoopack;		/* (s) */
791 	int32_t		t_rcvtime;		/* (s) */
792 	int32_t		tt_rexmt;		/* (s) */
793 	int32_t		tt_persist;		/* (s) */
794 	int32_t		tt_keep;		/* (s) */
795 	int32_t		tt_2msl;		/* (s) */
796 	int32_t		tt_delack;		/* (s) */
797 	int32_t		t_logstate;		/* (3) */
798 	uint32_t	t_snd_cwnd;		/* (s) */
799 	uint32_t	t_snd_ssthresh;		/* (s) */
800 	uint32_t	t_maxseg;		/* (s) */
801 	uint32_t	t_rcv_wnd;		/* (s) */
802 	uint32_t	t_snd_wnd;		/* (s) */
803 	uint32_t	xt_ecn;			/* (s) */
804 	uint16_t	xt_encaps_port;		/* (s) */
805 	int16_t		spare16;
806 	int32_t		spare32[25];
807 } __aligned(8);
808 
809 #ifdef _KERNEL
810 void	tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
811 #endif
812 #endif
813 
814 /*
815  * TCP function information (name-to-id mapping, aliases, and refcnt)
816  * exported to user-land via sysctl(3).
817  */
818 struct tcp_function_info {
819 	uint32_t	tfi_refcnt;
820 	uint8_t		tfi_id;
821 	char		tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
822 	char		tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
823 };
824 
825 /*
826  * Identifiers for TCP sysctl nodes
827  */
828 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
829 #define	TCPCTL_MSSDFLT		3	/* MSS default */
830 #define TCPCTL_STATS		4	/* statistics */
831 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
832 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
833 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
834 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
835 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
836 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
837 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
838 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
839 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
840 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
841 #define	TCPCTL_DROP		15	/* drop tcp connection */
842 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
843 
844 #ifdef _KERNEL
845 #ifdef SYSCTL_DECL
846 SYSCTL_DECL(_net_inet_tcp);
847 SYSCTL_DECL(_net_inet_tcp_sack);
848 MALLOC_DECLARE(M_TCPLOG);
849 #endif
850 
851 VNET_DECLARE(int, tcp_log_in_vain);
852 #define	V_tcp_log_in_vain		VNET(tcp_log_in_vain)
853 
854 /*
855  * Global TCP tunables shared between different stacks.
856  * Please keep the list sorted.
857  */
858 VNET_DECLARE(int, drop_synfin);
859 VNET_DECLARE(int, path_mtu_discovery);
860 VNET_DECLARE(int, tcp_abc_l_var);
861 VNET_DECLARE(int, tcp_autorcvbuf_max);
862 VNET_DECLARE(int, tcp_autosndbuf_inc);
863 VNET_DECLARE(int, tcp_autosndbuf_max);
864 VNET_DECLARE(int, tcp_delack_enabled);
865 VNET_DECLARE(int, tcp_do_autorcvbuf);
866 VNET_DECLARE(int, tcp_do_autosndbuf);
867 VNET_DECLARE(int, tcp_do_ecn);
868 VNET_DECLARE(int, tcp_do_lrd);
869 VNET_DECLARE(int, tcp_do_prr);
870 VNET_DECLARE(int, tcp_do_prr_conservative);
871 VNET_DECLARE(int, tcp_do_newcwv);
872 VNET_DECLARE(int, tcp_do_rfc1323);
873 VNET_DECLARE(int, tcp_tolerate_missing_ts);
874 VNET_DECLARE(int, tcp_do_rfc3042);
875 VNET_DECLARE(int, tcp_do_rfc3390);
876 VNET_DECLARE(int, tcp_do_rfc3465);
877 VNET_DECLARE(int, tcp_do_newsack);
878 VNET_DECLARE(int, tcp_do_sack);
879 VNET_DECLARE(int, tcp_do_tso);
880 VNET_DECLARE(int, tcp_ecn_maxretries);
881 VNET_DECLARE(int, tcp_initcwnd_segments);
882 VNET_DECLARE(int, tcp_insecure_rst);
883 VNET_DECLARE(int, tcp_insecure_syn);
884 VNET_DECLARE(uint32_t, tcp_map_entries_limit);
885 VNET_DECLARE(uint32_t, tcp_map_split_limit);
886 VNET_DECLARE(int, tcp_minmss);
887 VNET_DECLARE(int, tcp_mssdflt);
888 #ifdef STATS
889 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl);
890 VNET_DECLARE(int, tcp_perconn_stats_enable);
891 #endif /* STATS */
892 VNET_DECLARE(int, tcp_recvspace);
893 VNET_DECLARE(int, tcp_sack_globalholes);
894 VNET_DECLARE(int, tcp_sack_globalmaxholes);
895 VNET_DECLARE(int, tcp_sack_maxholes);
896 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
897 VNET_DECLARE(int, tcp_sendspace);
898 VNET_DECLARE(int, tcp_udp_tunneling_overhead);
899 VNET_DECLARE(int, tcp_udp_tunneling_port);
900 VNET_DECLARE(struct inpcbhead, tcb);
901 VNET_DECLARE(struct inpcbinfo, tcbinfo);
902 
903 #define	V_tcp_do_lrd			VNET(tcp_do_lrd)
904 #define	V_tcp_do_prr			VNET(tcp_do_prr)
905 #define	V_tcp_do_prr_conservative	VNET(tcp_do_prr_conservative)
906 #define	V_tcp_do_newcwv			VNET(tcp_do_newcwv)
907 #define	V_drop_synfin			VNET(drop_synfin)
908 #define	V_path_mtu_discovery		VNET(path_mtu_discovery)
909 #define	V_tcb				VNET(tcb)
910 #define	V_tcbinfo			VNET(tcbinfo)
911 #define	V_tcp_abc_l_var			VNET(tcp_abc_l_var)
912 #define	V_tcp_autorcvbuf_max		VNET(tcp_autorcvbuf_max)
913 #define	V_tcp_autosndbuf_inc		VNET(tcp_autosndbuf_inc)
914 #define	V_tcp_autosndbuf_max		VNET(tcp_autosndbuf_max)
915 #define	V_tcp_delack_enabled		VNET(tcp_delack_enabled)
916 #define	V_tcp_do_autorcvbuf		VNET(tcp_do_autorcvbuf)
917 #define	V_tcp_do_autosndbuf		VNET(tcp_do_autosndbuf)
918 #define	V_tcp_do_ecn			VNET(tcp_do_ecn)
919 #define	V_tcp_do_rfc1323		VNET(tcp_do_rfc1323)
920 #define	V_tcp_tolerate_missing_ts	VNET(tcp_tolerate_missing_ts)
921 #define V_tcp_ts_offset_per_conn	VNET(tcp_ts_offset_per_conn)
922 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
923 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)
924 #define	V_tcp_do_rfc3465		VNET(tcp_do_rfc3465)
925 #define	V_tcp_do_newsack		VNET(tcp_do_newsack)
926 #define	V_tcp_do_sack			VNET(tcp_do_sack)
927 #define	V_tcp_do_tso			VNET(tcp_do_tso)
928 #define	V_tcp_ecn_maxretries		VNET(tcp_ecn_maxretries)
929 #define	V_tcp_initcwnd_segments		VNET(tcp_initcwnd_segments)
930 #define	V_tcp_insecure_rst		VNET(tcp_insecure_rst)
931 #define	V_tcp_insecure_syn		VNET(tcp_insecure_syn)
932 #define	V_tcp_map_entries_limit		VNET(tcp_map_entries_limit)
933 #define	V_tcp_map_split_limit		VNET(tcp_map_split_limit)
934 #define	V_tcp_minmss			VNET(tcp_minmss)
935 #define	V_tcp_mssdflt			VNET(tcp_mssdflt)
936 #ifdef STATS
937 #define	V_tcp_perconn_stats_dflt_tpl	VNET(tcp_perconn_stats_dflt_tpl)
938 #define	V_tcp_perconn_stats_enable	VNET(tcp_perconn_stats_enable)
939 #endif /* STATS */
940 #define	V_tcp_recvspace			VNET(tcp_recvspace)
941 #define	V_tcp_sack_globalholes		VNET(tcp_sack_globalholes)
942 #define	V_tcp_sack_globalmaxholes	VNET(tcp_sack_globalmaxholes)
943 #define	V_tcp_sack_maxholes		VNET(tcp_sack_maxholes)
944 #define	V_tcp_sc_rst_sock_fail		VNET(tcp_sc_rst_sock_fail)
945 #define	V_tcp_sendspace			VNET(tcp_sendspace)
946 #define	V_tcp_udp_tunneling_overhead	VNET(tcp_udp_tunneling_overhead)
947 #define	V_tcp_udp_tunneling_port	VNET(tcp_udp_tunneling_port)
948 
949 #ifdef TCP_HHOOK
950 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
951 #define	V_tcp_hhh		VNET(tcp_hhh)
952 #endif
953 
954 int	 tcp_addoptions(struct tcpopt *, u_char *);
955 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
956 struct tcpcb *
957 	 tcp_close(struct tcpcb *);
958 void	 tcp_discardcb(struct tcpcb *);
959 void	 tcp_twstart(struct tcpcb *);
960 void	 tcp_twclose(struct tcptw *, int);
961 void	 tcp_ctlinput(int, struct sockaddr *, void *);
962 int	 tcp_ctloutput(struct socket *, struct sockopt *);
963 void 	 tcp_ctlinput_viaudp(int, struct sockaddr *, void *, void *);
964 struct tcpcb *
965 	 tcp_drop(struct tcpcb *, int);
966 void	 tcp_drain(void);
967 void	 tcp_init(void);
968 void	 tcp_fini(void *);
969 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
970 	    const void *);
971 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
972 	    const void *);
973 int	 tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *,
974 	    struct mbuf *);
975 void	 tcp_reass_global_init(void);
976 void	 tcp_reass_flush(struct tcpcb *);
977 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
978 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
979 		     struct tcpcb *, int, int);
980 void	tcp_pulloutofband(struct socket *,
981 		     struct tcphdr *, struct mbuf *, int);
982 void	tcp_xmit_timer(struct tcpcb *, int);
983 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
984 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
985 			    uint16_t nsegs, uint16_t type);
986 void 	cc_conn_init(struct tcpcb *tp);
987 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
988 void    cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos);
989 void	cc_ecnpkt_handler_flags(struct tcpcb *tp, uint16_t flags, uint8_t iptos);
990 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
991 #ifdef TCP_HHOOK
992 void	hhook_run_tcp_est_in(struct tcpcb *tp,
993 			    struct tcphdr *th, struct tcpopt *to);
994 #endif
995 
996 int	 tcp_input(struct mbuf **, int *, int);
997 int	 tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
998 	    struct tcpcb *, int);
999 int	 tcp_input_with_port(struct mbuf **, int *, int, uint16_t);
1000 void	 tcp_handle_wakeup(struct tcpcb *, struct socket *);
1001 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
1002 			struct socket *, struct tcpcb *, int, int, uint8_t);
1003 
1004 int register_tcp_functions(struct tcp_function_block *blk, int wait);
1005 int register_tcp_functions_as_names(struct tcp_function_block *blk,
1006     int wait, const char *names[], int *num_names);
1007 int register_tcp_functions_as_name(struct tcp_function_block *blk,
1008     const char *name, int wait);
1009 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
1010     bool force);
1011 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
1012 void tcp_switch_back_to_default(struct tcpcb *tp);
1013 struct tcp_function_block *
1014 find_and_ref_tcp_fb(struct tcp_function_block *fs);
1015 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
1016 
1017 extern counter_u64_t tcp_inp_lro_direct_queue;
1018 extern counter_u64_t tcp_inp_lro_wokeup_queue;
1019 extern counter_u64_t tcp_inp_lro_compressed;
1020 extern counter_u64_t tcp_inp_lro_locks_taken;
1021 extern counter_u64_t tcp_extra_mbuf;
1022 extern counter_u64_t tcp_would_have_but;
1023 extern counter_u64_t tcp_comp_total;
1024 extern counter_u64_t tcp_uncomp_total;
1025 
1026 #ifdef NETFLIX_EXP_DETECTION
1027 /* Various SACK attack thresholds */
1028 extern int32_t tcp_force_detection;
1029 extern int32_t tcp_sack_to_ack_thresh;
1030 extern int32_t tcp_sack_to_move_thresh;
1031 extern int32_t tcp_restoral_thresh;
1032 extern int32_t tcp_sad_decay_val;
1033 extern int32_t tcp_sad_pacing_interval;
1034 extern int32_t tcp_sad_low_pps;
1035 extern int32_t tcp_map_minimum;
1036 extern int32_t tcp_attack_on_turns_on_logging;
1037 #endif
1038 extern uint32_t tcp_ack_war_time_window;
1039 extern uint32_t tcp_ack_war_cnt;
1040 
1041 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
1042 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
1043 u_int	 tcp_maxseg(const struct tcpcb *);
1044 u_int	 tcp_fixed_maxseg(const struct tcpcb *);
1045 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
1046 	    struct tcp_ifcap *);
1047 void	 tcp_mss(struct tcpcb *, int);
1048 int	 tcp_mssopt(struct in_conninfo *);
1049 struct inpcb *
1050 	 tcp_drop_syn_sent(struct inpcb *, int);
1051 struct tcpcb *
1052 	 tcp_newtcpcb(struct inpcb *);
1053 int	 tcp_output(struct tcpcb *);
1054 void	 tcp_state_change(struct tcpcb *, int);
1055 void	 tcp_respond(struct tcpcb *, void *,
1056 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
1057 void	 tcp_tw_init(void);
1058 #ifdef VIMAGE
1059 void	 tcp_tw_destroy(void);
1060 #endif
1061 void	 tcp_tw_zone_change(void);
1062 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
1063 	    struct mbuf *, int);
1064 void	 tcp_setpersist(struct tcpcb *);
1065 void	 tcp_slowtimo(void);
1066 struct tcptemp *
1067 	 tcpip_maketemplate(struct inpcb *);
1068 void	 tcpip_fillheaders(struct inpcb *, uint16_t, void *, void *);
1069 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
1070 int	 tcp_timer_suspend(struct tcpcb *, uint32_t);
1071 void	 tcp_timers_unsuspend(struct tcpcb *, uint32_t);
1072 int	 tcp_timer_active(struct tcpcb *, uint32_t);
1073 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
1074 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
1075 int	 inp_to_cpuid(struct inpcb *inp);
1076 /*
1077  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
1078  */
1079 void	 tcp_hc_init(void);
1080 #ifdef VIMAGE
1081 void	 tcp_hc_destroy(void);
1082 #endif
1083 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
1084 uint32_t tcp_hc_getmtu(struct in_conninfo *);
1085 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
1086 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
1087 
1088 extern	struct pr_usrreqs tcp_usrreqs;
1089 
1090 uint32_t tcp_new_ts_offset(struct in_conninfo *);
1091 tcp_seq	 tcp_new_isn(struct in_conninfo *);
1092 
1093 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
1094 int	 tcp_dsack_block_exists(struct tcpcb *);
1095 void	 tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
1096 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1097 void	 tcp_clean_dsack_blocks(struct tcpcb *tp);
1098 void	 tcp_clean_sackreport(struct tcpcb *tp);
1099 void	 tcp_sack_adjust(struct tcpcb *tp);
1100 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1101 void	 tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *);
1102 void	 tcp_lost_retransmission(struct tcpcb *, struct tcphdr *);
1103 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1104 void	 tcp_free_sackholes(struct tcpcb *tp);
1105 void	 tcp_sack_lost_retransmission(struct tcpcb *, struct tcphdr *);
1106 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
1107 int	 tcp_compute_pipe(struct tcpcb *);
1108 uint32_t tcp_compute_initwnd(uint32_t);
1109 void	 tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
1110 int	 tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes,
1111     size_t seed_len);
1112 int tcp_can_enable_pacing(void);
1113 void tcp_decrement_paced_conn(void);
1114 
1115 struct mbuf *
1116 	 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
1117 	   int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls);
1118 
1119 int	tcp_stats_init(void);
1120 void tcp_log_end_status(struct tcpcb *tp, uint8_t status);
1121 
1122 static inline void
1123 tcp_fields_to_host(struct tcphdr *th)
1124 {
1125 
1126 	th->th_seq = ntohl(th->th_seq);
1127 	th->th_ack = ntohl(th->th_ack);
1128 	th->th_win = ntohs(th->th_win);
1129 	th->th_urp = ntohs(th->th_urp);
1130 }
1131 
1132 static inline void
1133 tcp_fields_to_net(struct tcphdr *th)
1134 {
1135 
1136 	th->th_seq = htonl(th->th_seq);
1137 	th->th_ack = htonl(th->th_ack);
1138 	th->th_win = htons(th->th_win);
1139 	th->th_urp = htons(th->th_urp);
1140 }
1141 
1142 static inline void
1143 tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt,
1144     uint8_t is_tlp, int hw_tls)
1145 {
1146 	uint64_t rexmit_percent;
1147 
1148 	if (is_tlp) {
1149 		tp->t_sndtlppack++;
1150 		tp->t_sndtlpbyte += len;
1151 	}
1152 	/* To get total bytes sent you must add t_snd_rxt_bytes to t_sndbytes */
1153 	if (is_rxt)
1154 		tp->t_snd_rxt_bytes += len;
1155 	else
1156 		tp->t_sndbytes += len;
1157 
1158 	if (hw_tls && is_rxt) {
1159 		rexmit_percent = (1000ULL * tp->t_snd_rxt_bytes) / (10ULL * (tp->t_snd_rxt_bytes + tp->t_sndbytes));
1160 		if (rexmit_percent > ktls_ifnet_max_rexmit_pct)
1161 			ktls_disable_ifnet(tp);
1162 	}
1163 
1164 }
1165 #endif /* _KERNEL */
1166 
1167 #endif /* _NETINET_TCP_VAR_H_ */
1168