xref: /netbsd-src/sys/netinet/tcp_subr.c (revision 267197ec1eebfcb9810ea27a89625b6ddf68e3e7)
1 /*	$NetBSD: tcp_subr.c,v 1.222 2008/02/05 09:38:48 yamt Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * 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 project 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 PROJECT 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 PROJECT 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 
32 /*-
33  * Copyright (c) 1997, 1998, 2000, 2001 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation
38  * Facility, NASA Ames Research Center.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. All advertising materials mentioning features or use of this software
49  *    must display the following acknowledgement:
50  *	This product includes software developed by the NetBSD
51  *	Foundation, Inc. and its contributors.
52  * 4. Neither the name of The NetBSD Foundation nor the names of its
53  *    contributors may be used to endorse or promote products derived
54  *    from this software without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66  * POSSIBILITY OF SUCH DAMAGE.
67  */
68 
69 /*
70  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
71  *	The Regents of the University of California.  All rights reserved.
72  *
73  * Redistribution and use in source and binary forms, with or without
74  * modification, are permitted provided that the following conditions
75  * are met:
76  * 1. Redistributions of source code must retain the above copyright
77  *    notice, this list of conditions and the following disclaimer.
78  * 2. Redistributions in binary form must reproduce the above copyright
79  *    notice, this list of conditions and the following disclaimer in the
80  *    documentation and/or other materials provided with the distribution.
81  * 3. Neither the name of the University nor the names of its contributors
82  *    may be used to endorse or promote products derived from this software
83  *    without specific prior written permission.
84  *
85  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
86  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
87  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
88  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
89  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
90  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
91  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
92  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
93  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
94  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
95  * SUCH DAMAGE.
96  *
97  *	@(#)tcp_subr.c	8.2 (Berkeley) 5/24/95
98  */
99 
100 #include <sys/cdefs.h>
101 __KERNEL_RCSID(0, "$NetBSD: tcp_subr.c,v 1.222 2008/02/05 09:38:48 yamt Exp $");
102 
103 #include "opt_inet.h"
104 #include "opt_ipsec.h"
105 #include "opt_tcp_compat_42.h"
106 #include "opt_inet_csum.h"
107 #include "opt_mbuftrace.h"
108 #include "rnd.h"
109 
110 #include <sys/param.h>
111 #include <sys/proc.h>
112 #include <sys/systm.h>
113 #include <sys/malloc.h>
114 #include <sys/mbuf.h>
115 #include <sys/socket.h>
116 #include <sys/socketvar.h>
117 #include <sys/protosw.h>
118 #include <sys/errno.h>
119 #include <sys/kernel.h>
120 #include <sys/pool.h>
121 #if NRND > 0
122 #include <sys/md5.h>
123 #include <sys/rnd.h>
124 #endif
125 
126 #include <net/route.h>
127 #include <net/if.h>
128 
129 #include <netinet/in.h>
130 #include <netinet/in_systm.h>
131 #include <netinet/ip.h>
132 #include <netinet/in_pcb.h>
133 #include <netinet/ip_var.h>
134 #include <netinet/ip_icmp.h>
135 
136 #ifdef INET6
137 #ifndef INET
138 #include <netinet/in.h>
139 #endif
140 #include <netinet/ip6.h>
141 #include <netinet6/in6_pcb.h>
142 #include <netinet6/ip6_var.h>
143 #include <netinet6/in6_var.h>
144 #include <netinet6/ip6protosw.h>
145 #include <netinet/icmp6.h>
146 #include <netinet6/nd6.h>
147 #endif
148 
149 #include <netinet/tcp.h>
150 #include <netinet/tcp_fsm.h>
151 #include <netinet/tcp_seq.h>
152 #include <netinet/tcp_timer.h>
153 #include <netinet/tcp_var.h>
154 #include <netinet/tcp_congctl.h>
155 #include <netinet/tcpip.h>
156 
157 #ifdef IPSEC
158 #include <netinet6/ipsec.h>
159 #include <netkey/key.h>
160 #endif /*IPSEC*/
161 
162 #ifdef FAST_IPSEC
163 #include <netipsec/ipsec.h>
164 #include <netipsec/xform.h>
165 #ifdef INET6
166 #include <netipsec/ipsec6.h>
167 #endif
168  #include <netipsec/key.h>
169 #endif	/* FAST_IPSEC*/
170 
171 
172 struct	inpcbtable tcbtable;	/* head of queue of active tcpcb's */
173 struct	tcpstat tcpstat;	/* tcp statistics */
174 u_int32_t tcp_now;		/* for RFC 1323 timestamps */
175 
176 /* patchable/settable parameters for tcp */
177 int 	tcp_mssdflt = TCP_MSS;
178 int	tcp_minmss = TCP_MINMSS;
179 int 	tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ;
180 int	tcp_do_rfc1323 = 1;	/* window scaling / timestamps (obsolete) */
181 #if NRND > 0
182 int	tcp_do_rfc1948 = 0;	/* ISS by cryptographic hash */
183 #endif
184 int	tcp_do_sack = 1;	/* selective acknowledgement */
185 int	tcp_do_win_scale = 1;	/* RFC1323 window scaling */
186 int	tcp_do_timestamps = 1;	/* RFC1323 timestamps */
187 int	tcp_ack_on_push = 0;	/* set to enable immediate ACK-on-PUSH */
188 int	tcp_do_ecn = 0;		/* Explicit Congestion Notification */
189 #ifndef TCP_INIT_WIN
190 #define	TCP_INIT_WIN	0	/* initial slow start window */
191 #endif
192 #ifndef TCP_INIT_WIN_LOCAL
193 #define	TCP_INIT_WIN_LOCAL 4	/* initial slow start window for local nets */
194 #endif
195 int	tcp_init_win = TCP_INIT_WIN;
196 int	tcp_init_win_local = TCP_INIT_WIN_LOCAL;
197 int	tcp_mss_ifmtu = 0;
198 #ifdef TCP_COMPAT_42
199 int	tcp_compat_42 = 1;
200 #else
201 int	tcp_compat_42 = 0;
202 #endif
203 int	tcp_rst_ppslim = 100;	/* 100pps */
204 int	tcp_ackdrop_ppslim = 100;	/* 100pps */
205 int	tcp_do_loopback_cksum = 0;
206 int	tcp_do_abc = 1;		/* RFC3465 Appropriate byte counting. */
207 int	tcp_abc_aggressive = 1;	/* 1: L=2*SMSS  0: L=1*SMSS */
208 int	tcp_sack_tp_maxholes = 32;
209 int	tcp_sack_globalmaxholes = 1024;
210 int	tcp_sack_globalholes = 0;
211 int	tcp_ecn_maxretries = 1;
212 
213 /* tcb hash */
214 #ifndef TCBHASHSIZE
215 #define	TCBHASHSIZE	128
216 #endif
217 int	tcbhashsize = TCBHASHSIZE;
218 
219 /* syn hash parameters */
220 #define	TCP_SYN_HASH_SIZE	293
221 #define	TCP_SYN_BUCKET_SIZE	35
222 int	tcp_syn_cache_size = TCP_SYN_HASH_SIZE;
223 int	tcp_syn_cache_limit = TCP_SYN_HASH_SIZE*TCP_SYN_BUCKET_SIZE;
224 int	tcp_syn_bucket_limit = 3*TCP_SYN_BUCKET_SIZE;
225 struct	syn_cache_head tcp_syn_cache[TCP_SYN_HASH_SIZE];
226 
227 int	tcp_freeq(struct tcpcb *);
228 
229 #ifdef INET
230 void	tcp_mtudisc_callback(struct in_addr);
231 #endif
232 #ifdef INET6
233 void	tcp6_mtudisc_callback(struct in6_addr *);
234 #endif
235 
236 #ifdef INET6
237 void	tcp6_mtudisc(struct in6pcb *, int);
238 #endif
239 
240 POOL_INIT(tcpcb_pool, sizeof(struct tcpcb), 0, 0, 0, "tcpcbpl", NULL,
241     IPL_SOFTNET);
242 
243 #ifdef TCP_CSUM_COUNTERS
244 #include <sys/device.h>
245 
246 #if defined(INET)
247 struct evcnt tcp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
248     NULL, "tcp", "hwcsum bad");
249 struct evcnt tcp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
250     NULL, "tcp", "hwcsum ok");
251 struct evcnt tcp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
252     NULL, "tcp", "hwcsum data");
253 struct evcnt tcp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
254     NULL, "tcp", "swcsum");
255 
256 EVCNT_ATTACH_STATIC(tcp_hwcsum_bad);
257 EVCNT_ATTACH_STATIC(tcp_hwcsum_ok);
258 EVCNT_ATTACH_STATIC(tcp_hwcsum_data);
259 EVCNT_ATTACH_STATIC(tcp_swcsum);
260 #endif /* defined(INET) */
261 
262 #if defined(INET6)
263 struct evcnt tcp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
264     NULL, "tcp6", "hwcsum bad");
265 struct evcnt tcp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
266     NULL, "tcp6", "hwcsum ok");
267 struct evcnt tcp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
268     NULL, "tcp6", "hwcsum data");
269 struct evcnt tcp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
270     NULL, "tcp6", "swcsum");
271 
272 EVCNT_ATTACH_STATIC(tcp6_hwcsum_bad);
273 EVCNT_ATTACH_STATIC(tcp6_hwcsum_ok);
274 EVCNT_ATTACH_STATIC(tcp6_hwcsum_data);
275 EVCNT_ATTACH_STATIC(tcp6_swcsum);
276 #endif /* defined(INET6) */
277 #endif /* TCP_CSUM_COUNTERS */
278 
279 
280 #ifdef TCP_OUTPUT_COUNTERS
281 #include <sys/device.h>
282 
283 struct evcnt tcp_output_bigheader = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
284     NULL, "tcp", "output big header");
285 struct evcnt tcp_output_predict_hit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
286     NULL, "tcp", "output predict hit");
287 struct evcnt tcp_output_predict_miss = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
288     NULL, "tcp", "output predict miss");
289 struct evcnt tcp_output_copysmall = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
290     NULL, "tcp", "output copy small");
291 struct evcnt tcp_output_copybig = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
292     NULL, "tcp", "output copy big");
293 struct evcnt tcp_output_refbig = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
294     NULL, "tcp", "output reference big");
295 
296 EVCNT_ATTACH_STATIC(tcp_output_bigheader);
297 EVCNT_ATTACH_STATIC(tcp_output_predict_hit);
298 EVCNT_ATTACH_STATIC(tcp_output_predict_miss);
299 EVCNT_ATTACH_STATIC(tcp_output_copysmall);
300 EVCNT_ATTACH_STATIC(tcp_output_copybig);
301 EVCNT_ATTACH_STATIC(tcp_output_refbig);
302 
303 #endif /* TCP_OUTPUT_COUNTERS */
304 
305 #ifdef TCP_REASS_COUNTERS
306 #include <sys/device.h>
307 
308 struct evcnt tcp_reass_ = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
309     NULL, "tcp_reass", "calls");
310 struct evcnt tcp_reass_empty = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
311     &tcp_reass_, "tcp_reass", "insert into empty queue");
312 struct evcnt tcp_reass_iteration[8] = {
313     EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", ">7 iterations"),
314     EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "1 iteration"),
315     EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "2 iterations"),
316     EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "3 iterations"),
317     EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "4 iterations"),
318     EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "5 iterations"),
319     EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "6 iterations"),
320     EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "7 iterations"),
321 };
322 struct evcnt tcp_reass_prependfirst = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
323     &tcp_reass_, "tcp_reass", "prepend to first");
324 struct evcnt tcp_reass_prepend = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
325     &tcp_reass_, "tcp_reass", "prepend");
326 struct evcnt tcp_reass_insert = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
327     &tcp_reass_, "tcp_reass", "insert");
328 struct evcnt tcp_reass_inserttail = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
329     &tcp_reass_, "tcp_reass", "insert at tail");
330 struct evcnt tcp_reass_append = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
331     &tcp_reass_, "tcp_reass", "append");
332 struct evcnt tcp_reass_appendtail = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
333     &tcp_reass_, "tcp_reass", "append to tail fragment");
334 struct evcnt tcp_reass_overlaptail = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
335     &tcp_reass_, "tcp_reass", "overlap at end");
336 struct evcnt tcp_reass_overlapfront = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
337     &tcp_reass_, "tcp_reass", "overlap at start");
338 struct evcnt tcp_reass_segdup = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
339     &tcp_reass_, "tcp_reass", "duplicate segment");
340 struct evcnt tcp_reass_fragdup = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
341     &tcp_reass_, "tcp_reass", "duplicate fragment");
342 
343 EVCNT_ATTACH_STATIC(tcp_reass_);
344 EVCNT_ATTACH_STATIC(tcp_reass_empty);
345 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 0);
346 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 1);
347 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 2);
348 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 3);
349 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 4);
350 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 5);
351 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 6);
352 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 7);
353 EVCNT_ATTACH_STATIC(tcp_reass_prependfirst);
354 EVCNT_ATTACH_STATIC(tcp_reass_prepend);
355 EVCNT_ATTACH_STATIC(tcp_reass_insert);
356 EVCNT_ATTACH_STATIC(tcp_reass_inserttail);
357 EVCNT_ATTACH_STATIC(tcp_reass_append);
358 EVCNT_ATTACH_STATIC(tcp_reass_appendtail);
359 EVCNT_ATTACH_STATIC(tcp_reass_overlaptail);
360 EVCNT_ATTACH_STATIC(tcp_reass_overlapfront);
361 EVCNT_ATTACH_STATIC(tcp_reass_segdup);
362 EVCNT_ATTACH_STATIC(tcp_reass_fragdup);
363 
364 #endif /* TCP_REASS_COUNTERS */
365 
366 #ifdef MBUFTRACE
367 struct mowner tcp_mowner = MOWNER_INIT("tcp", "");
368 struct mowner tcp_rx_mowner = MOWNER_INIT("tcp", "rx");
369 struct mowner tcp_tx_mowner = MOWNER_INIT("tcp", "tx");
370 struct mowner tcp_sock_mowner = MOWNER_INIT("tcp", "sock");
371 struct mowner tcp_sock_rx_mowner = MOWNER_INIT("tcp", "sock rx");
372 struct mowner tcp_sock_tx_mowner = MOWNER_INIT("tcp", "sock tx");
373 #endif
374 
375 /*
376  * Tcp initialization
377  */
378 void
379 tcp_init(void)
380 {
381 	int hlen;
382 
383 	in_pcbinit(&tcbtable, tcbhashsize, tcbhashsize);
384 
385 	hlen = sizeof(struct ip) + sizeof(struct tcphdr);
386 #ifdef INET6
387 	if (sizeof(struct ip) < sizeof(struct ip6_hdr))
388 		hlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
389 #endif
390 	if (max_protohdr < hlen)
391 		max_protohdr = hlen;
392 	if (max_linkhdr + hlen > MHLEN)
393 		panic("tcp_init");
394 
395 #ifdef INET
396 	icmp_mtudisc_callback_register(tcp_mtudisc_callback);
397 #endif
398 #ifdef INET6
399 	icmp6_mtudisc_callback_register(tcp6_mtudisc_callback);
400 #endif
401 
402 	/* Initialize timer state. */
403 	tcp_timer_init();
404 
405 	/* Initialize the compressed state engine. */
406 	syn_cache_init();
407 
408 	/* Initialize the congestion control algorithms. */
409 	tcp_congctl_init();
410 
411 	/* Initialize the TCPCB template. */
412 	tcp_tcpcb_template();
413 
414 	MOWNER_ATTACH(&tcp_tx_mowner);
415 	MOWNER_ATTACH(&tcp_rx_mowner);
416 	MOWNER_ATTACH(&tcp_reass_mowner);
417 	MOWNER_ATTACH(&tcp_sock_mowner);
418 	MOWNER_ATTACH(&tcp_sock_tx_mowner);
419 	MOWNER_ATTACH(&tcp_sock_rx_mowner);
420 	MOWNER_ATTACH(&tcp_mowner);
421 }
422 
423 /*
424  * Create template to be used to send tcp packets on a connection.
425  * Call after host entry created, allocates an mbuf and fills
426  * in a skeletal tcp/ip header, minimizing the amount of work
427  * necessary when the connection is used.
428  */
429 struct mbuf *
430 tcp_template(struct tcpcb *tp)
431 {
432 	struct inpcb *inp = tp->t_inpcb;
433 #ifdef INET6
434 	struct in6pcb *in6p = tp->t_in6pcb;
435 #endif
436 	struct tcphdr *n;
437 	struct mbuf *m;
438 	int hlen;
439 
440 	switch (tp->t_family) {
441 	case AF_INET:
442 		hlen = sizeof(struct ip);
443 		if (inp)
444 			break;
445 #ifdef INET6
446 		if (in6p) {
447 			/* mapped addr case */
448 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
449 			 && IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
450 				break;
451 		}
452 #endif
453 		return NULL;	/*EINVAL*/
454 #ifdef INET6
455 	case AF_INET6:
456 		hlen = sizeof(struct ip6_hdr);
457 		if (in6p) {
458 			/* more sainty check? */
459 			break;
460 		}
461 		return NULL;	/*EINVAL*/
462 #endif
463 	default:
464 		hlen = 0;	/*pacify gcc*/
465 		return NULL;	/*EAFNOSUPPORT*/
466 	}
467 #ifdef DIAGNOSTIC
468 	if (hlen + sizeof(struct tcphdr) > MCLBYTES)
469 		panic("mclbytes too small for t_template");
470 #endif
471 	m = tp->t_template;
472 	if (m && m->m_len == hlen + sizeof(struct tcphdr))
473 		;
474 	else {
475 		if (m)
476 			m_freem(m);
477 		m = tp->t_template = NULL;
478 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
479 		if (m && hlen + sizeof(struct tcphdr) > MHLEN) {
480 			MCLGET(m, M_DONTWAIT);
481 			if ((m->m_flags & M_EXT) == 0) {
482 				m_free(m);
483 				m = NULL;
484 			}
485 		}
486 		if (m == NULL)
487 			return NULL;
488 		MCLAIM(m, &tcp_mowner);
489 		m->m_pkthdr.len = m->m_len = hlen + sizeof(struct tcphdr);
490 	}
491 
492 	bzero(mtod(m, void *), m->m_len);
493 
494 	n = (struct tcphdr *)(mtod(m, char *) + hlen);
495 
496 	switch (tp->t_family) {
497 	case AF_INET:
498 	    {
499 		struct ipovly *ipov;
500 		mtod(m, struct ip *)->ip_v = 4;
501 		mtod(m, struct ip *)->ip_hl = hlen >> 2;
502 		ipov = mtod(m, struct ipovly *);
503 		ipov->ih_pr = IPPROTO_TCP;
504 		ipov->ih_len = htons(sizeof(struct tcphdr));
505 		if (inp) {
506 			ipov->ih_src = inp->inp_laddr;
507 			ipov->ih_dst = inp->inp_faddr;
508 		}
509 #ifdef INET6
510 		else if (in6p) {
511 			/* mapped addr case */
512 			bcopy(&in6p->in6p_laddr.s6_addr32[3], &ipov->ih_src,
513 				sizeof(ipov->ih_src));
514 			bcopy(&in6p->in6p_faddr.s6_addr32[3], &ipov->ih_dst,
515 				sizeof(ipov->ih_dst));
516 		}
517 #endif
518 		/*
519 		 * Compute the pseudo-header portion of the checksum
520 		 * now.  We incrementally add in the TCP option and
521 		 * payload lengths later, and then compute the TCP
522 		 * checksum right before the packet is sent off onto
523 		 * the wire.
524 		 */
525 		n->th_sum = in_cksum_phdr(ipov->ih_src.s_addr,
526 		    ipov->ih_dst.s_addr,
527 		    htons(sizeof(struct tcphdr) + IPPROTO_TCP));
528 		break;
529 	    }
530 #ifdef INET6
531 	case AF_INET6:
532 	    {
533 		struct ip6_hdr *ip6;
534 		mtod(m, struct ip *)->ip_v = 6;
535 		ip6 = mtod(m, struct ip6_hdr *);
536 		ip6->ip6_nxt = IPPROTO_TCP;
537 		ip6->ip6_plen = htons(sizeof(struct tcphdr));
538 		ip6->ip6_src = in6p->in6p_laddr;
539 		ip6->ip6_dst = in6p->in6p_faddr;
540 		ip6->ip6_flow = in6p->in6p_flowinfo & IPV6_FLOWINFO_MASK;
541 		if (ip6_auto_flowlabel) {
542 			ip6->ip6_flow &= ~IPV6_FLOWLABEL_MASK;
543 			ip6->ip6_flow |=
544 			    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
545 		}
546 		ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
547 		ip6->ip6_vfc |= IPV6_VERSION;
548 
549 		/*
550 		 * Compute the pseudo-header portion of the checksum
551 		 * now.  We incrementally add in the TCP option and
552 		 * payload lengths later, and then compute the TCP
553 		 * checksum right before the packet is sent off onto
554 		 * the wire.
555 		 */
556 		n->th_sum = in6_cksum_phdr(&in6p->in6p_laddr,
557 		    &in6p->in6p_faddr, htonl(sizeof(struct tcphdr)),
558 		    htonl(IPPROTO_TCP));
559 		break;
560 	    }
561 #endif
562 	}
563 	if (inp) {
564 		n->th_sport = inp->inp_lport;
565 		n->th_dport = inp->inp_fport;
566 	}
567 #ifdef INET6
568 	else if (in6p) {
569 		n->th_sport = in6p->in6p_lport;
570 		n->th_dport = in6p->in6p_fport;
571 	}
572 #endif
573 	n->th_seq = 0;
574 	n->th_ack = 0;
575 	n->th_x2 = 0;
576 	n->th_off = 5;
577 	n->th_flags = 0;
578 	n->th_win = 0;
579 	n->th_urp = 0;
580 	return (m);
581 }
582 
583 /*
584  * Send a single message to the TCP at address specified by
585  * the given TCP/IP header.  If m == 0, then we make a copy
586  * of the tcpiphdr at ti and send directly to the addressed host.
587  * This is used to force keep alive messages out using the TCP
588  * template for a connection tp->t_template.  If flags are given
589  * then we send a message back to the TCP which originated the
590  * segment ti, and discard the mbuf containing it and any other
591  * attached mbufs.
592  *
593  * In any case the ack and sequence number of the transmitted
594  * segment are as specified by the parameters.
595  */
596 int
597 tcp_respond(struct tcpcb *tp, struct mbuf *template, struct mbuf *m,
598     struct tcphdr *th0, tcp_seq ack, tcp_seq seq, int flags)
599 {
600 #ifdef INET6
601 	struct rtentry *rt;
602 #endif
603 	struct route *ro;
604 	int error, tlen, win = 0;
605 	int hlen;
606 	struct ip *ip;
607 #ifdef INET6
608 	struct ip6_hdr *ip6;
609 #endif
610 	int family;	/* family on packet, not inpcb/in6pcb! */
611 	struct tcphdr *th;
612 	struct socket *so;
613 
614 	if (tp != NULL && (flags & TH_RST) == 0) {
615 #ifdef DIAGNOSTIC
616 		if (tp->t_inpcb && tp->t_in6pcb)
617 			panic("tcp_respond: both t_inpcb and t_in6pcb are set");
618 #endif
619 #ifdef INET
620 		if (tp->t_inpcb)
621 			win = sbspace(&tp->t_inpcb->inp_socket->so_rcv);
622 #endif
623 #ifdef INET6
624 		if (tp->t_in6pcb)
625 			win = sbspace(&tp->t_in6pcb->in6p_socket->so_rcv);
626 #endif
627 	}
628 
629 	th = NULL;	/* Quell uninitialized warning */
630 	ip = NULL;
631 #ifdef INET6
632 	ip6 = NULL;
633 #endif
634 	if (m == 0) {
635 		if (!template)
636 			return EINVAL;
637 
638 		/* get family information from template */
639 		switch (mtod(template, struct ip *)->ip_v) {
640 		case 4:
641 			family = AF_INET;
642 			hlen = sizeof(struct ip);
643 			break;
644 #ifdef INET6
645 		case 6:
646 			family = AF_INET6;
647 			hlen = sizeof(struct ip6_hdr);
648 			break;
649 #endif
650 		default:
651 			return EAFNOSUPPORT;
652 		}
653 
654 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
655 		if (m) {
656 			MCLAIM(m, &tcp_tx_mowner);
657 			MCLGET(m, M_DONTWAIT);
658 			if ((m->m_flags & M_EXT) == 0) {
659 				m_free(m);
660 				m = NULL;
661 			}
662 		}
663 		if (m == NULL)
664 			return (ENOBUFS);
665 
666 		if (tcp_compat_42)
667 			tlen = 1;
668 		else
669 			tlen = 0;
670 
671 		m->m_data += max_linkhdr;
672 		bcopy(mtod(template, void *), mtod(m, void *),
673 			template->m_len);
674 		switch (family) {
675 		case AF_INET:
676 			ip = mtod(m, struct ip *);
677 			th = (struct tcphdr *)(ip + 1);
678 			break;
679 #ifdef INET6
680 		case AF_INET6:
681 			ip6 = mtod(m, struct ip6_hdr *);
682 			th = (struct tcphdr *)(ip6 + 1);
683 			break;
684 #endif
685 #if 0
686 		default:
687 			/* noone will visit here */
688 			m_freem(m);
689 			return EAFNOSUPPORT;
690 #endif
691 		}
692 		flags = TH_ACK;
693 	} else {
694 
695 		if ((m->m_flags & M_PKTHDR) == 0) {
696 #if 0
697 			printf("non PKTHDR to tcp_respond\n");
698 #endif
699 			m_freem(m);
700 			return EINVAL;
701 		}
702 #ifdef DIAGNOSTIC
703 		if (!th0)
704 			panic("th0 == NULL in tcp_respond");
705 #endif
706 
707 		/* get family information from m */
708 		switch (mtod(m, struct ip *)->ip_v) {
709 		case 4:
710 			family = AF_INET;
711 			hlen = sizeof(struct ip);
712 			ip = mtod(m, struct ip *);
713 			break;
714 #ifdef INET6
715 		case 6:
716 			family = AF_INET6;
717 			hlen = sizeof(struct ip6_hdr);
718 			ip6 = mtod(m, struct ip6_hdr *);
719 			break;
720 #endif
721 		default:
722 			m_freem(m);
723 			return EAFNOSUPPORT;
724 		}
725 		/* clear h/w csum flags inherited from rx packet */
726 		m->m_pkthdr.csum_flags = 0;
727 
728 		if ((flags & TH_SYN) == 0 || sizeof(*th0) > (th0->th_off << 2))
729 			tlen = sizeof(*th0);
730 		else
731 			tlen = th0->th_off << 2;
732 
733 		if (m->m_len > hlen + tlen && (m->m_flags & M_EXT) == 0 &&
734 		    mtod(m, char *) + hlen == (char *)th0) {
735 			m->m_len = hlen + tlen;
736 			m_freem(m->m_next);
737 			m->m_next = NULL;
738 		} else {
739 			struct mbuf *n;
740 
741 #ifdef DIAGNOSTIC
742 			if (max_linkhdr + hlen + tlen > MCLBYTES) {
743 				m_freem(m);
744 				return EMSGSIZE;
745 			}
746 #endif
747 			MGETHDR(n, M_DONTWAIT, MT_HEADER);
748 			if (n && max_linkhdr + hlen + tlen > MHLEN) {
749 				MCLGET(n, M_DONTWAIT);
750 				if ((n->m_flags & M_EXT) == 0) {
751 					m_freem(n);
752 					n = NULL;
753 				}
754 			}
755 			if (!n) {
756 				m_freem(m);
757 				return ENOBUFS;
758 			}
759 
760 			MCLAIM(n, &tcp_tx_mowner);
761 			n->m_data += max_linkhdr;
762 			n->m_len = hlen + tlen;
763 			m_copyback(n, 0, hlen, mtod(m, void *));
764 			m_copyback(n, hlen, tlen, (void *)th0);
765 
766 			m_freem(m);
767 			m = n;
768 			n = NULL;
769 		}
770 
771 #define xchg(a,b,type) { type t; t=a; a=b; b=t; }
772 		switch (family) {
773 		case AF_INET:
774 			ip = mtod(m, struct ip *);
775 			th = (struct tcphdr *)(ip + 1);
776 			ip->ip_p = IPPROTO_TCP;
777 			xchg(ip->ip_dst, ip->ip_src, struct in_addr);
778 			ip->ip_p = IPPROTO_TCP;
779 			break;
780 #ifdef INET6
781 		case AF_INET6:
782 			ip6 = mtod(m, struct ip6_hdr *);
783 			th = (struct tcphdr *)(ip6 + 1);
784 			ip6->ip6_nxt = IPPROTO_TCP;
785 			xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
786 			ip6->ip6_nxt = IPPROTO_TCP;
787 			break;
788 #endif
789 #if 0
790 		default:
791 			/* noone will visit here */
792 			m_freem(m);
793 			return EAFNOSUPPORT;
794 #endif
795 		}
796 		xchg(th->th_dport, th->th_sport, u_int16_t);
797 #undef xchg
798 		tlen = 0;	/*be friendly with the following code*/
799 	}
800 	th->th_seq = htonl(seq);
801 	th->th_ack = htonl(ack);
802 	th->th_x2 = 0;
803 	if ((flags & TH_SYN) == 0) {
804 		if (tp)
805 			win >>= tp->rcv_scale;
806 		if (win > TCP_MAXWIN)
807 			win = TCP_MAXWIN;
808 		th->th_win = htons((u_int16_t)win);
809 		th->th_off = sizeof (struct tcphdr) >> 2;
810 		tlen += sizeof(*th);
811 	} else
812 		tlen += th->th_off << 2;
813 	m->m_len = hlen + tlen;
814 	m->m_pkthdr.len = hlen + tlen;
815 	m->m_pkthdr.rcvif = (struct ifnet *) 0;
816 	th->th_flags = flags;
817 	th->th_urp = 0;
818 
819 	switch (family) {
820 #ifdef INET
821 	case AF_INET:
822 	    {
823 		struct ipovly *ipov = (struct ipovly *)ip;
824 		bzero(ipov->ih_x1, sizeof ipov->ih_x1);
825 		ipov->ih_len = htons((u_int16_t)tlen);
826 
827 		th->th_sum = 0;
828 		th->th_sum = in_cksum(m, hlen + tlen);
829 		ip->ip_len = htons(hlen + tlen);
830 		ip->ip_ttl = ip_defttl;
831 		break;
832 	    }
833 #endif
834 #ifdef INET6
835 	case AF_INET6:
836 	    {
837 		th->th_sum = 0;
838 		th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr),
839 				tlen);
840 		ip6->ip6_plen = htons(tlen);
841 		if (tp && tp->t_in6pcb) {
842 			struct ifnet *oifp;
843 			ro = &tp->t_in6pcb->in6p_route;
844 			oifp = (rt = rtcache_validate(ro)) != NULL ? rt->rt_ifp
845 			                                           : NULL;
846 			ip6->ip6_hlim = in6_selecthlim(tp->t_in6pcb, oifp);
847 		} else
848 			ip6->ip6_hlim = ip6_defhlim;
849 		ip6->ip6_flow &= ~IPV6_FLOWINFO_MASK;
850 		if (ip6_auto_flowlabel) {
851 			ip6->ip6_flow |=
852 			    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
853 		}
854 		break;
855 	    }
856 #endif
857 	}
858 
859 	if (tp && tp->t_inpcb)
860 		so = tp->t_inpcb->inp_socket;
861 #ifdef INET6
862 	else if (tp && tp->t_in6pcb)
863 		so = tp->t_in6pcb->in6p_socket;
864 #endif
865 	else
866 		so = NULL;
867 
868 	if (tp != NULL && tp->t_inpcb != NULL) {
869 		ro = &tp->t_inpcb->inp_route;
870 #ifdef DIAGNOSTIC
871 		if (family != AF_INET)
872 			panic("tcp_respond: address family mismatch");
873 		if (!in_hosteq(ip->ip_dst, tp->t_inpcb->inp_faddr)) {
874 			panic("tcp_respond: ip_dst %x != inp_faddr %x",
875 			    ntohl(ip->ip_dst.s_addr),
876 			    ntohl(tp->t_inpcb->inp_faddr.s_addr));
877 		}
878 #endif
879 	}
880 #ifdef INET6
881 	else if (tp != NULL && tp->t_in6pcb != NULL) {
882 		ro = (struct route *)&tp->t_in6pcb->in6p_route;
883 #ifdef DIAGNOSTIC
884 		if (family == AF_INET) {
885 			if (!IN6_IS_ADDR_V4MAPPED(&tp->t_in6pcb->in6p_faddr))
886 				panic("tcp_respond: not mapped addr");
887 			if (bcmp(&ip->ip_dst,
888 			    &tp->t_in6pcb->in6p_faddr.s6_addr32[3],
889 			    sizeof(ip->ip_dst)) != 0) {
890 				panic("tcp_respond: ip_dst != in6p_faddr");
891 			}
892 		} else if (family == AF_INET6) {
893 			if (!IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
894 			    &tp->t_in6pcb->in6p_faddr))
895 				panic("tcp_respond: ip6_dst != in6p_faddr");
896 		} else
897 			panic("tcp_respond: address family mismatch");
898 #endif
899 	}
900 #endif
901 	else
902 		ro = NULL;
903 
904 	switch (family) {
905 #ifdef INET
906 	case AF_INET:
907 		error = ip_output(m, NULL, ro,
908 		    (tp && tp->t_mtudisc ? IP_MTUDISC : 0),
909 		    (struct ip_moptions *)0, so);
910 		break;
911 #endif
912 #ifdef INET6
913 	case AF_INET6:
914 		error = ip6_output(m, NULL, ro, 0, NULL, so, NULL);
915 		break;
916 #endif
917 	default:
918 		error = EAFNOSUPPORT;
919 		break;
920 	}
921 
922 	return (error);
923 }
924 
925 /*
926  * Template TCPCB.  Rather than zeroing a new TCPCB and initializing
927  * a bunch of members individually, we maintain this template for the
928  * static and mostly-static components of the TCPCB, and copy it into
929  * the new TCPCB instead.
930  */
931 static struct tcpcb tcpcb_template = {
932 	.t_srtt = TCPTV_SRTTBASE,
933 	.t_rttmin = TCPTV_MIN,
934 
935 	.snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT,
936 	.snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT,
937 	.snd_numholes = 0,
938 
939 	.t_partialacks = -1,
940 	.t_bytes_acked = 0,
941 };
942 
943 /*
944  * Updates the TCPCB template whenever a parameter that would affect
945  * the template is changed.
946  */
947 void
948 tcp_tcpcb_template(void)
949 {
950 	struct tcpcb *tp = &tcpcb_template;
951 	int flags;
952 
953 	tp->t_peermss = tcp_mssdflt;
954 	tp->t_ourmss = tcp_mssdflt;
955 	tp->t_segsz = tcp_mssdflt;
956 
957 	flags = 0;
958 	if (tcp_do_rfc1323 && tcp_do_win_scale)
959 		flags |= TF_REQ_SCALE;
960 	if (tcp_do_rfc1323 && tcp_do_timestamps)
961 		flags |= TF_REQ_TSTMP;
962 	tp->t_flags = flags;
963 
964 	/*
965 	 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no
966 	 * rtt estimate.  Set rttvar so that srtt + 2 * rttvar gives
967 	 * reasonable initial retransmit time.
968 	 */
969 	tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ << (TCP_RTTVAR_SHIFT + 2 - 1);
970 	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
971 	    TCPTV_MIN, TCPTV_REXMTMAX);
972 
973 	/* Keep Alive */
974 	tp->t_keepinit = tcp_keepinit;
975 	tp->t_keepidle = tcp_keepidle;
976 	tp->t_keepintvl = tcp_keepintvl;
977 	tp->t_keepcnt = tcp_keepcnt;
978 	tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl;
979 }
980 
981 /*
982  * Create a new TCP control block, making an
983  * empty reassembly queue and hooking it to the argument
984  * protocol control block.
985  */
986 /* family selects inpcb, or in6pcb */
987 struct tcpcb *
988 tcp_newtcpcb(int family, void *aux)
989 {
990 #ifdef INET6
991 	struct rtentry *rt;
992 #endif
993 	struct tcpcb *tp;
994 	int i;
995 
996 	/* XXX Consider using a pool_cache for speed. */
997 	tp = pool_get(&tcpcb_pool, PR_NOWAIT);	/* splsoftnet via tcp_usrreq */
998 	if (tp == NULL)
999 		return (NULL);
1000 	memcpy(tp, &tcpcb_template, sizeof(*tp));
1001 	TAILQ_INIT(&tp->segq);
1002 	TAILQ_INIT(&tp->timeq);
1003 	tp->t_family = family;		/* may be overridden later on */
1004 	TAILQ_INIT(&tp->snd_holes);
1005 	LIST_INIT(&tp->t_sc);		/* XXX can template this */
1006 
1007 	/* Don't sweat this loop; hopefully the compiler will unroll it. */
1008 	for (i = 0; i < TCPT_NTIMERS; i++) {
1009 		callout_init(&tp->t_timer[i], 0);
1010 		TCP_TIMER_INIT(tp, i);
1011 	}
1012 	callout_init(&tp->t_delack_ch, 0);
1013 
1014 	switch (family) {
1015 	case AF_INET:
1016 	    {
1017 		struct inpcb *inp = (struct inpcb *)aux;
1018 
1019 		inp->inp_ip.ip_ttl = ip_defttl;
1020 		inp->inp_ppcb = (void *)tp;
1021 
1022 		tp->t_inpcb = inp;
1023 		tp->t_mtudisc = ip_mtudisc;
1024 		break;
1025 	    }
1026 #ifdef INET6
1027 	case AF_INET6:
1028 	    {
1029 		struct in6pcb *in6p = (struct in6pcb *)aux;
1030 
1031 		in6p->in6p_ip6.ip6_hlim = in6_selecthlim(in6p,
1032 			(rt = rtcache_validate(&in6p->in6p_route)) != NULL
1033 			    ? rt->rt_ifp
1034 			    : NULL);
1035 		in6p->in6p_ppcb = (void *)tp;
1036 
1037 		tp->t_in6pcb = in6p;
1038 		/* for IPv6, always try to run path MTU discovery */
1039 		tp->t_mtudisc = 1;
1040 		break;
1041 	    }
1042 #endif /* INET6 */
1043 	default:
1044 		for (i = 0; i < TCPT_NTIMERS; i++)
1045 			callout_destroy(&tp->t_timer[i]);
1046 		callout_destroy(&tp->t_delack_ch);
1047 		pool_put(&tcpcb_pool, tp);	/* splsoftnet via tcp_usrreq */
1048 		return (NULL);
1049 	}
1050 
1051 	/*
1052 	 * Initialize our timebase.  When we send timestamps, we take
1053 	 * the delta from tcp_now -- this means each connection always
1054 	 * gets a timebase of 1, which makes it, among other things,
1055 	 * more difficult to determine how long a system has been up,
1056 	 * and thus how many TCP sequence increments have occurred.
1057 	 *
1058 	 * We start with 1, because 0 doesn't work with linux, which
1059 	 * considers timestamp 0 in a SYN packet as a bug and disables
1060 	 * timestamps.
1061 	 */
1062 	tp->ts_timebase = tcp_now - 1;
1063 
1064 	tp->t_congctl = tcp_congctl_global;
1065 	tp->t_congctl->refcnt++;
1066 
1067 	return (tp);
1068 }
1069 
1070 /*
1071  * Drop a TCP connection, reporting
1072  * the specified error.  If connection is synchronized,
1073  * then send a RST to peer.
1074  */
1075 struct tcpcb *
1076 tcp_drop(struct tcpcb *tp, int errno)
1077 {
1078 	struct socket *so = NULL;
1079 
1080 #ifdef DIAGNOSTIC
1081 	if (tp->t_inpcb && tp->t_in6pcb)
1082 		panic("tcp_drop: both t_inpcb and t_in6pcb are set");
1083 #endif
1084 #ifdef INET
1085 	if (tp->t_inpcb)
1086 		so = tp->t_inpcb->inp_socket;
1087 #endif
1088 #ifdef INET6
1089 	if (tp->t_in6pcb)
1090 		so = tp->t_in6pcb->in6p_socket;
1091 #endif
1092 	if (!so)
1093 		return NULL;
1094 
1095 	if (TCPS_HAVERCVDSYN(tp->t_state)) {
1096 		tp->t_state = TCPS_CLOSED;
1097 		(void) tcp_output(tp);
1098 		tcpstat.tcps_drops++;
1099 	} else
1100 		tcpstat.tcps_conndrops++;
1101 	if (errno == ETIMEDOUT && tp->t_softerror)
1102 		errno = tp->t_softerror;
1103 	so->so_error = errno;
1104 	return (tcp_close(tp));
1105 }
1106 
1107 /*
1108  * Return whether this tcpcb is marked as dead, indicating
1109  * to the calling timer function that no further action should
1110  * be taken, as we are about to release this tcpcb.  The release
1111  * of the storage will be done if this is the last timer running.
1112  *
1113  * This should be called from the callout handler function after
1114  * callout_ack() is done, so that the number of invoking timer
1115  * functions is 0.
1116  */
1117 int
1118 tcp_isdead(struct tcpcb *tp)
1119 {
1120 	int i, dead = (tp->t_flags & TF_DEAD);
1121 
1122 	if (__predict_false(dead)) {
1123 		if (tcp_timers_invoking(tp) > 0)
1124 				/* not quite there yet -- count separately? */
1125 			return dead;
1126 		tcpstat.tcps_delayed_free++;
1127 		for (i = 0; i < TCPT_NTIMERS; i++)
1128 			callout_destroy(&tp->t_timer[i]);
1129 		callout_destroy(&tp->t_delack_ch);
1130 		pool_put(&tcpcb_pool, tp);	/* splsoftnet via tcp_timer.c */
1131 	}
1132 	return dead;
1133 }
1134 
1135 /*
1136  * Close a TCP control block:
1137  *	discard all space held by the tcp
1138  *	discard internet protocol block
1139  *	wake up any sleepers
1140  */
1141 struct tcpcb *
1142 tcp_close(struct tcpcb *tp)
1143 {
1144 	struct inpcb *inp;
1145 #ifdef INET6
1146 	struct in6pcb *in6p;
1147 #endif
1148 	struct socket *so;
1149 #ifdef RTV_RTT
1150 	struct rtentry *rt;
1151 #endif
1152 	struct route *ro;
1153 	int j;
1154 
1155 	inp = tp->t_inpcb;
1156 #ifdef INET6
1157 	in6p = tp->t_in6pcb;
1158 #endif
1159 	so = NULL;
1160 	ro = NULL;
1161 	if (inp) {
1162 		so = inp->inp_socket;
1163 		ro = &inp->inp_route;
1164 	}
1165 #ifdef INET6
1166 	else if (in6p) {
1167 		so = in6p->in6p_socket;
1168 		ro = (struct route *)&in6p->in6p_route;
1169 	}
1170 #endif
1171 
1172 #ifdef RTV_RTT
1173 	/*
1174 	 * If we sent enough data to get some meaningful characteristics,
1175 	 * save them in the routing entry.  'Enough' is arbitrarily
1176 	 * defined as the sendpipesize (default 4K) * 16.  This would
1177 	 * give us 16 rtt samples assuming we only get one sample per
1178 	 * window (the usual case on a long haul net).  16 samples is
1179 	 * enough for the srtt filter to converge to within 5% of the correct
1180 	 * value; fewer samples and we could save a very bogus rtt.
1181 	 *
1182 	 * Don't update the default route's characteristics and don't
1183 	 * update anything that the user "locked".
1184 	 */
1185 	if (SEQ_LT(tp->iss + so->so_snd.sb_hiwat * 16, tp->snd_max) &&
1186 	    ro && (rt = rtcache_validate(ro)) != NULL &&
1187 	    !in_nullhost(satocsin(rt_getkey(rt))->sin_addr)) {
1188 		u_long i = 0;
1189 
1190 		if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) {
1191 			i = tp->t_srtt *
1192 			    ((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTT_SHIFT + 2));
1193 			if (rt->rt_rmx.rmx_rtt && i)
1194 				/*
1195 				 * filter this update to half the old & half
1196 				 * the new values, converting scale.
1197 				 * See route.h and tcp_var.h for a
1198 				 * description of the scaling constants.
1199 				 */
1200 				rt->rt_rmx.rmx_rtt =
1201 				    (rt->rt_rmx.rmx_rtt + i) / 2;
1202 			else
1203 				rt->rt_rmx.rmx_rtt = i;
1204 		}
1205 		if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) {
1206 			i = tp->t_rttvar *
1207 			    ((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTTVAR_SHIFT + 2));
1208 			if (rt->rt_rmx.rmx_rttvar && i)
1209 				rt->rt_rmx.rmx_rttvar =
1210 				    (rt->rt_rmx.rmx_rttvar + i) / 2;
1211 			else
1212 				rt->rt_rmx.rmx_rttvar = i;
1213 		}
1214 		/*
1215 		 * update the pipelimit (ssthresh) if it has been updated
1216 		 * already or if a pipesize was specified & the threshhold
1217 		 * got below half the pipesize.  I.e., wait for bad news
1218 		 * before we start updating, then update on both good
1219 		 * and bad news.
1220 		 */
1221 		if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 &&
1222 		    (i = tp->snd_ssthresh) && rt->rt_rmx.rmx_ssthresh) ||
1223 		    i < (rt->rt_rmx.rmx_sendpipe / 2)) {
1224 			/*
1225 			 * convert the limit from user data bytes to
1226 			 * packets then to packet data bytes.
1227 			 */
1228 			i = (i + tp->t_segsz / 2) / tp->t_segsz;
1229 			if (i < 2)
1230 				i = 2;
1231 			i *= (u_long)(tp->t_segsz + sizeof (struct tcpiphdr));
1232 			if (rt->rt_rmx.rmx_ssthresh)
1233 				rt->rt_rmx.rmx_ssthresh =
1234 				    (rt->rt_rmx.rmx_ssthresh + i) / 2;
1235 			else
1236 				rt->rt_rmx.rmx_ssthresh = i;
1237 		}
1238 	}
1239 #endif /* RTV_RTT */
1240 	/* free the reassembly queue, if any */
1241 	TCP_REASS_LOCK(tp);
1242 	(void) tcp_freeq(tp);
1243 	TCP_REASS_UNLOCK(tp);
1244 
1245 	/* free the SACK holes list. */
1246 	tcp_free_sackholes(tp);
1247 
1248 	tp->t_congctl->refcnt--;
1249 
1250 	tcp_canceltimers(tp);
1251 	TCP_CLEAR_DELACK(tp);
1252 	syn_cache_cleanup(tp);
1253 
1254 	if (tp->t_template) {
1255 		m_free(tp->t_template);
1256 		tp->t_template = NULL;
1257 	}
1258 	if (tcp_timers_invoking(tp))
1259 		tp->t_flags |= TF_DEAD;
1260 	else {
1261 		for (j = 0; j < TCPT_NTIMERS; j++)
1262 			callout_destroy(&tp->t_timer[j]);
1263 		callout_destroy(&tp->t_delack_ch);
1264 		pool_put(&tcpcb_pool, tp);
1265 	}
1266 
1267 	if (inp) {
1268 		inp->inp_ppcb = 0;
1269 		soisdisconnected(so);
1270 		in_pcbdetach(inp);
1271 	}
1272 #ifdef INET6
1273 	else if (in6p) {
1274 		in6p->in6p_ppcb = 0;
1275 		soisdisconnected(so);
1276 		in6_pcbdetach(in6p);
1277 	}
1278 #endif
1279 	tcpstat.tcps_closed++;
1280 	return ((struct tcpcb *)0);
1281 }
1282 
1283 int
1284 tcp_freeq(tp)
1285 	struct tcpcb *tp;
1286 {
1287 	struct ipqent *qe;
1288 	int rv = 0;
1289 #ifdef TCPREASS_DEBUG
1290 	int i = 0;
1291 #endif
1292 
1293 	TCP_REASS_LOCK_CHECK(tp);
1294 
1295 	while ((qe = TAILQ_FIRST(&tp->segq)) != NULL) {
1296 #ifdef TCPREASS_DEBUG
1297 		printf("tcp_freeq[%p,%d]: %u:%u(%u) 0x%02x\n",
1298 			tp, i++, qe->ipqe_seq, qe->ipqe_seq + qe->ipqe_len,
1299 			qe->ipqe_len, qe->ipqe_flags & (TH_SYN|TH_FIN|TH_RST));
1300 #endif
1301 		TAILQ_REMOVE(&tp->segq, qe, ipqe_q);
1302 		TAILQ_REMOVE(&tp->timeq, qe, ipqe_timeq);
1303 		m_freem(qe->ipqe_m);
1304 		tcpipqent_free(qe);
1305 		rv = 1;
1306 	}
1307 	tp->t_segqlen = 0;
1308 	KASSERT(TAILQ_EMPTY(&tp->timeq));
1309 	return (rv);
1310 }
1311 
1312 /*
1313  * Protocol drain routine.  Called when memory is in short supply.
1314  */
1315 void
1316 tcp_drain(void)
1317 {
1318 	struct inpcb_hdr *inph;
1319 	struct tcpcb *tp;
1320 
1321 	/*
1322 	 * Free the sequence queue of all TCP connections.
1323 	 */
1324 	CIRCLEQ_FOREACH(inph, &tcbtable.inpt_queue, inph_queue) {
1325 		switch (inph->inph_af) {
1326 		case AF_INET:
1327 			tp = intotcpcb((struct inpcb *)inph);
1328 			break;
1329 #ifdef INET6
1330 		case AF_INET6:
1331 			tp = in6totcpcb((struct in6pcb *)inph);
1332 			break;
1333 #endif
1334 		default:
1335 			tp = NULL;
1336 			break;
1337 		}
1338 		if (tp != NULL) {
1339 			/*
1340 			 * We may be called from a device's interrupt
1341 			 * context.  If the tcpcb is already busy,
1342 			 * just bail out now.
1343 			 */
1344 			if (tcp_reass_lock_try(tp) == 0)
1345 				continue;
1346 			if (tcp_freeq(tp))
1347 				tcpstat.tcps_connsdrained++;
1348 			TCP_REASS_UNLOCK(tp);
1349 		}
1350 	}
1351 }
1352 
1353 /*
1354  * Notify a tcp user of an asynchronous error;
1355  * store error as soft error, but wake up user
1356  * (for now, won't do anything until can select for soft error).
1357  */
1358 void
1359 tcp_notify(struct inpcb *inp, int error)
1360 {
1361 	struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb;
1362 	struct socket *so = inp->inp_socket;
1363 
1364 	/*
1365 	 * Ignore some errors if we are hooked up.
1366 	 * If connection hasn't completed, has retransmitted several times,
1367 	 * and receives a second error, give up now.  This is better
1368 	 * than waiting a long time to establish a connection that
1369 	 * can never complete.
1370 	 */
1371 	if (tp->t_state == TCPS_ESTABLISHED &&
1372 	     (error == EHOSTUNREACH || error == ENETUNREACH ||
1373 	      error == EHOSTDOWN)) {
1374 		return;
1375 	} else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 &&
1376 	    tp->t_rxtshift > 3 && tp->t_softerror)
1377 		so->so_error = error;
1378 	else
1379 		tp->t_softerror = error;
1380 	wakeup((void *) &so->so_timeo);
1381 	sorwakeup(so);
1382 	sowwakeup(so);
1383 }
1384 
1385 #ifdef INET6
1386 void
1387 tcp6_notify(struct in6pcb *in6p, int error)
1388 {
1389 	struct tcpcb *tp = (struct tcpcb *)in6p->in6p_ppcb;
1390 	struct socket *so = in6p->in6p_socket;
1391 
1392 	/*
1393 	 * Ignore some errors if we are hooked up.
1394 	 * If connection hasn't completed, has retransmitted several times,
1395 	 * and receives a second error, give up now.  This is better
1396 	 * than waiting a long time to establish a connection that
1397 	 * can never complete.
1398 	 */
1399 	if (tp->t_state == TCPS_ESTABLISHED &&
1400 	     (error == EHOSTUNREACH || error == ENETUNREACH ||
1401 	      error == EHOSTDOWN)) {
1402 		return;
1403 	} else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 &&
1404 	    tp->t_rxtshift > 3 && tp->t_softerror)
1405 		so->so_error = error;
1406 	else
1407 		tp->t_softerror = error;
1408 	wakeup((void *) &so->so_timeo);
1409 	sorwakeup(so);
1410 	sowwakeup(so);
1411 }
1412 #endif
1413 
1414 #ifdef INET6
1415 void
1416 tcp6_ctlinput(int cmd, const struct sockaddr *sa, void *d)
1417 {
1418 	struct tcphdr th;
1419 	void (*notify)(struct in6pcb *, int) = tcp6_notify;
1420 	int nmatch;
1421 	struct ip6_hdr *ip6;
1422 	const struct sockaddr_in6 *sa6_src = NULL;
1423 	const struct sockaddr_in6 *sa6 = (const struct sockaddr_in6 *)sa;
1424 	struct mbuf *m;
1425 	int off;
1426 
1427 	if (sa->sa_family != AF_INET6 ||
1428 	    sa->sa_len != sizeof(struct sockaddr_in6))
1429 		return;
1430 	if ((unsigned)cmd >= PRC_NCMDS)
1431 		return;
1432 	else if (cmd == PRC_QUENCH) {
1433 		/*
1434 		 * Don't honor ICMP Source Quench messages meant for
1435 		 * TCP connections.
1436 		 */
1437 		return;
1438 	} else if (PRC_IS_REDIRECT(cmd))
1439 		notify = in6_rtchange, d = NULL;
1440 	else if (cmd == PRC_MSGSIZE)
1441 		; /* special code is present, see below */
1442 	else if (cmd == PRC_HOSTDEAD)
1443 		d = NULL;
1444 	else if (inet6ctlerrmap[cmd] == 0)
1445 		return;
1446 
1447 	/* if the parameter is from icmp6, decode it. */
1448 	if (d != NULL) {
1449 		struct ip6ctlparam *ip6cp = (struct ip6ctlparam *)d;
1450 		m = ip6cp->ip6c_m;
1451 		ip6 = ip6cp->ip6c_ip6;
1452 		off = ip6cp->ip6c_off;
1453 		sa6_src = ip6cp->ip6c_src;
1454 	} else {
1455 		m = NULL;
1456 		ip6 = NULL;
1457 		sa6_src = &sa6_any;
1458 		off = 0;
1459 	}
1460 
1461 	if (ip6) {
1462 		/*
1463 		 * XXX: We assume that when ip6 is non NULL,
1464 		 * M and OFF are valid.
1465 		 */
1466 
1467 		/* check if we can safely examine src and dst ports */
1468 		if (m->m_pkthdr.len < off + sizeof(th)) {
1469 			if (cmd == PRC_MSGSIZE)
1470 				icmp6_mtudisc_update((struct ip6ctlparam *)d, 0);
1471 			return;
1472 		}
1473 
1474 		bzero(&th, sizeof(th));
1475 		m_copydata(m, off, sizeof(th), (void *)&th);
1476 
1477 		if (cmd == PRC_MSGSIZE) {
1478 			int valid = 0;
1479 
1480 			/*
1481 			 * Check to see if we have a valid TCP connection
1482 			 * corresponding to the address in the ICMPv6 message
1483 			 * payload.
1484 			 */
1485 			if (in6_pcblookup_connect(&tcbtable, &sa6->sin6_addr,
1486 			    th.th_dport,
1487 			    (const struct in6_addr *)&sa6_src->sin6_addr,
1488 			    th.th_sport, 0))
1489 				valid++;
1490 
1491 			/*
1492 			 * Depending on the value of "valid" and routing table
1493 			 * size (mtudisc_{hi,lo}wat), we will:
1494 			 * - recalcurate the new MTU and create the
1495 			 *   corresponding routing entry, or
1496 			 * - ignore the MTU change notification.
1497 			 */
1498 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
1499 
1500 			/*
1501 			 * no need to call in6_pcbnotify, it should have been
1502 			 * called via callback if necessary
1503 			 */
1504 			return;
1505 		}
1506 
1507 		nmatch = in6_pcbnotify(&tcbtable, sa, th.th_dport,
1508 		    (const struct sockaddr *)sa6_src, th.th_sport, cmd, NULL, notify);
1509 		if (nmatch == 0 && syn_cache_count &&
1510 		    (inet6ctlerrmap[cmd] == EHOSTUNREACH ||
1511 		     inet6ctlerrmap[cmd] == ENETUNREACH ||
1512 		     inet6ctlerrmap[cmd] == EHOSTDOWN))
1513 			syn_cache_unreach((const struct sockaddr *)sa6_src,
1514 					  sa, &th);
1515 	} else {
1516 		(void) in6_pcbnotify(&tcbtable, sa, 0,
1517 		    (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify);
1518 	}
1519 }
1520 #endif
1521 
1522 #ifdef INET
1523 /* assumes that ip header and tcp header are contiguous on mbuf */
1524 void *
1525 tcp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
1526 {
1527 	struct ip *ip = v;
1528 	struct tcphdr *th;
1529 	struct icmp *icp;
1530 	extern const int inetctlerrmap[];
1531 	void (*notify)(struct inpcb *, int) = tcp_notify;
1532 	int errno;
1533 	int nmatch;
1534 	struct tcpcb *tp;
1535 	u_int mtu;
1536 	tcp_seq seq;
1537 	struct inpcb *inp;
1538 #ifdef INET6
1539 	struct in6pcb *in6p;
1540 	struct in6_addr src6, dst6;
1541 #endif
1542 
1543 	if (sa->sa_family != AF_INET ||
1544 	    sa->sa_len != sizeof(struct sockaddr_in))
1545 		return NULL;
1546 	if ((unsigned)cmd >= PRC_NCMDS)
1547 		return NULL;
1548 	errno = inetctlerrmap[cmd];
1549 	if (cmd == PRC_QUENCH)
1550 		/*
1551 		 * Don't honor ICMP Source Quench messages meant for
1552 		 * TCP connections.
1553 		 */
1554 		return NULL;
1555 	else if (PRC_IS_REDIRECT(cmd))
1556 		notify = in_rtchange, ip = 0;
1557 	else if (cmd == PRC_MSGSIZE && ip && ip->ip_v == 4) {
1558 		/*
1559 		 * Check to see if we have a valid TCP connection
1560 		 * corresponding to the address in the ICMP message
1561 		 * payload.
1562 		 *
1563 		 * Boundary check is made in icmp_input(), with ICMP_ADVLENMIN.
1564 		 */
1565 		th = (struct tcphdr *)((char *)ip + (ip->ip_hl << 2));
1566 #ifdef INET6
1567 		memset(&src6, 0, sizeof(src6));
1568 		memset(&dst6, 0, sizeof(dst6));
1569 		src6.s6_addr16[5] = dst6.s6_addr16[5] = 0xffff;
1570 		memcpy(&src6.s6_addr32[3], &ip->ip_src, sizeof(struct in_addr));
1571 		memcpy(&dst6.s6_addr32[3], &ip->ip_dst, sizeof(struct in_addr));
1572 #endif
1573 		if ((inp = in_pcblookup_connect(&tcbtable, ip->ip_dst,
1574 		    th->th_dport, ip->ip_src, th->th_sport)) != NULL)
1575 #ifdef INET6
1576 			in6p = NULL;
1577 #else
1578 			;
1579 #endif
1580 #ifdef INET6
1581 		else if ((in6p = in6_pcblookup_connect(&tcbtable, &dst6,
1582 		    th->th_dport, &src6, th->th_sport, 0)) != NULL)
1583 			;
1584 #endif
1585 		else
1586 			return NULL;
1587 
1588 		/*
1589 		 * Now that we've validated that we are actually communicating
1590 		 * with the host indicated in the ICMP message, locate the
1591 		 * ICMP header, recalculate the new MTU, and create the
1592 		 * corresponding routing entry.
1593 		 */
1594 		icp = (struct icmp *)((char *)ip -
1595 		    offsetof(struct icmp, icmp_ip));
1596 		if (inp) {
1597 			if ((tp = intotcpcb(inp)) == NULL)
1598 				return NULL;
1599 		}
1600 #ifdef INET6
1601 		else if (in6p) {
1602 			if ((tp = in6totcpcb(in6p)) == NULL)
1603 				return NULL;
1604 		}
1605 #endif
1606 		else
1607 			return NULL;
1608 		seq = ntohl(th->th_seq);
1609 		if (SEQ_LT(seq, tp->snd_una) || SEQ_GT(seq, tp->snd_max))
1610 			return NULL;
1611 		/*
1612 		 * If the ICMP message advertises a Next-Hop MTU
1613 		 * equal or larger than the maximum packet size we have
1614 		 * ever sent, drop the message.
1615 		 */
1616 		mtu = (u_int)ntohs(icp->icmp_nextmtu);
1617 		if (mtu >= tp->t_pmtud_mtu_sent)
1618 			return NULL;
1619 		if (mtu >= tcp_hdrsz(tp) + tp->t_pmtud_mss_acked) {
1620 			/*
1621 			 * Calculate new MTU, and create corresponding
1622 			 * route (traditional PMTUD).
1623 			 */
1624 			tp->t_flags &= ~TF_PMTUD_PEND;
1625 			icmp_mtudisc(icp, ip->ip_dst);
1626 		} else {
1627 			/*
1628 			 * Record the information got in the ICMP
1629 			 * message; act on it later.
1630 			 * If we had already recorded an ICMP message,
1631 			 * replace the old one only if the new message
1632 			 * refers to an older TCP segment
1633 			 */
1634 			if (tp->t_flags & TF_PMTUD_PEND) {
1635 				if (SEQ_LT(tp->t_pmtud_th_seq, seq))
1636 					return NULL;
1637 			} else
1638 				tp->t_flags |= TF_PMTUD_PEND;
1639 			tp->t_pmtud_th_seq = seq;
1640 			tp->t_pmtud_nextmtu = icp->icmp_nextmtu;
1641 			tp->t_pmtud_ip_len = icp->icmp_ip.ip_len;
1642 			tp->t_pmtud_ip_hl = icp->icmp_ip.ip_hl;
1643 		}
1644 		return NULL;
1645 	} else if (cmd == PRC_HOSTDEAD)
1646 		ip = 0;
1647 	else if (errno == 0)
1648 		return NULL;
1649 	if (ip && ip->ip_v == 4 && sa->sa_family == AF_INET) {
1650 		th = (struct tcphdr *)((char *)ip + (ip->ip_hl << 2));
1651 		nmatch = in_pcbnotify(&tcbtable, satocsin(sa)->sin_addr,
1652 		    th->th_dport, ip->ip_src, th->th_sport, errno, notify);
1653 		if (nmatch == 0 && syn_cache_count &&
1654 		    (inetctlerrmap[cmd] == EHOSTUNREACH ||
1655 		    inetctlerrmap[cmd] == ENETUNREACH ||
1656 		    inetctlerrmap[cmd] == EHOSTDOWN)) {
1657 			struct sockaddr_in sin;
1658 			bzero(&sin, sizeof(sin));
1659 			sin.sin_len = sizeof(sin);
1660 			sin.sin_family = AF_INET;
1661 			sin.sin_port = th->th_sport;
1662 			sin.sin_addr = ip->ip_src;
1663 			syn_cache_unreach((struct sockaddr *)&sin, sa, th);
1664 		}
1665 
1666 		/* XXX mapped address case */
1667 	} else
1668 		in_pcbnotifyall(&tcbtable, satocsin(sa)->sin_addr, errno,
1669 		    notify);
1670 	return NULL;
1671 }
1672 
1673 /*
1674  * When a source quench is received, we are being notified of congestion.
1675  * Close the congestion window down to the Loss Window (one segment).
1676  * We will gradually open it again as we proceed.
1677  */
1678 void
1679 tcp_quench(struct inpcb *inp, int errno)
1680 {
1681 	struct tcpcb *tp = intotcpcb(inp);
1682 
1683 	if (tp) {
1684 		tp->snd_cwnd = tp->t_segsz;
1685 		tp->t_bytes_acked = 0;
1686 	}
1687 }
1688 #endif
1689 
1690 #ifdef INET6
1691 void
1692 tcp6_quench(struct in6pcb *in6p, int errno)
1693 {
1694 	struct tcpcb *tp = in6totcpcb(in6p);
1695 
1696 	if (tp) {
1697 		tp->snd_cwnd = tp->t_segsz;
1698 		tp->t_bytes_acked = 0;
1699 	}
1700 }
1701 #endif
1702 
1703 #ifdef INET
1704 /*
1705  * Path MTU Discovery handlers.
1706  */
1707 void
1708 tcp_mtudisc_callback(struct in_addr faddr)
1709 {
1710 #ifdef INET6
1711 	struct in6_addr in6;
1712 #endif
1713 
1714 	in_pcbnotifyall(&tcbtable, faddr, EMSGSIZE, tcp_mtudisc);
1715 #ifdef INET6
1716 	memset(&in6, 0, sizeof(in6));
1717 	in6.s6_addr16[5] = 0xffff;
1718 	memcpy(&in6.s6_addr32[3], &faddr, sizeof(struct in_addr));
1719 	tcp6_mtudisc_callback(&in6);
1720 #endif
1721 }
1722 
1723 /*
1724  * On receipt of path MTU corrections, flush old route and replace it
1725  * with the new one.  Retransmit all unacknowledged packets, to ensure
1726  * that all packets will be received.
1727  */
1728 void
1729 tcp_mtudisc(struct inpcb *inp, int errno)
1730 {
1731 	struct tcpcb *tp = intotcpcb(inp);
1732 	struct rtentry *rt = in_pcbrtentry(inp);
1733 
1734 	if (tp != 0) {
1735 		if (rt != 0) {
1736 			/*
1737 			 * If this was not a host route, remove and realloc.
1738 			 */
1739 			if ((rt->rt_flags & RTF_HOST) == 0) {
1740 				in_rtchange(inp, errno);
1741 				if ((rt = in_pcbrtentry(inp)) == 0)
1742 					return;
1743 			}
1744 
1745 			/*
1746 			 * Slow start out of the error condition.  We
1747 			 * use the MTU because we know it's smaller
1748 			 * than the previously transmitted segment.
1749 			 *
1750 			 * Note: This is more conservative than the
1751 			 * suggestion in draft-floyd-incr-init-win-03.
1752 			 */
1753 			if (rt->rt_rmx.rmx_mtu != 0)
1754 				tp->snd_cwnd =
1755 				    TCP_INITIAL_WINDOW(tcp_init_win,
1756 				    rt->rt_rmx.rmx_mtu);
1757 		}
1758 
1759 		/*
1760 		 * Resend unacknowledged packets.
1761 		 */
1762 		tp->snd_nxt = tp->snd_una;
1763 		tcp_output(tp);
1764 	}
1765 }
1766 #endif
1767 
1768 #ifdef INET6
1769 /*
1770  * Path MTU Discovery handlers.
1771  */
1772 void
1773 tcp6_mtudisc_callback(struct in6_addr *faddr)
1774 {
1775 	struct sockaddr_in6 sin6;
1776 
1777 	bzero(&sin6, sizeof(sin6));
1778 	sin6.sin6_family = AF_INET6;
1779 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1780 	sin6.sin6_addr = *faddr;
1781 	(void) in6_pcbnotify(&tcbtable, (struct sockaddr *)&sin6, 0,
1782 	    (const struct sockaddr *)&sa6_any, 0, PRC_MSGSIZE, NULL, tcp6_mtudisc);
1783 }
1784 
1785 void
1786 tcp6_mtudisc(struct in6pcb *in6p, int errno)
1787 {
1788 	struct tcpcb *tp = in6totcpcb(in6p);
1789 	struct rtentry *rt = in6_pcbrtentry(in6p);
1790 
1791 	if (tp != 0) {
1792 		if (rt != 0) {
1793 			/*
1794 			 * If this was not a host route, remove and realloc.
1795 			 */
1796 			if ((rt->rt_flags & RTF_HOST) == 0) {
1797 				in6_rtchange(in6p, errno);
1798 				if ((rt = in6_pcbrtentry(in6p)) == 0)
1799 					return;
1800 			}
1801 
1802 			/*
1803 			 * Slow start out of the error condition.  We
1804 			 * use the MTU because we know it's smaller
1805 			 * than the previously transmitted segment.
1806 			 *
1807 			 * Note: This is more conservative than the
1808 			 * suggestion in draft-floyd-incr-init-win-03.
1809 			 */
1810 			if (rt->rt_rmx.rmx_mtu != 0)
1811 				tp->snd_cwnd =
1812 				    TCP_INITIAL_WINDOW(tcp_init_win,
1813 				    rt->rt_rmx.rmx_mtu);
1814 		}
1815 
1816 		/*
1817 		 * Resend unacknowledged packets.
1818 		 */
1819 		tp->snd_nxt = tp->snd_una;
1820 		tcp_output(tp);
1821 	}
1822 }
1823 #endif /* INET6 */
1824 
1825 /*
1826  * Compute the MSS to advertise to the peer.  Called only during
1827  * the 3-way handshake.  If we are the server (peer initiated
1828  * connection), we are called with a pointer to the interface
1829  * on which the SYN packet arrived.  If we are the client (we
1830  * initiated connection), we are called with a pointer to the
1831  * interface out which this connection should go.
1832  *
1833  * NOTE: Do not subtract IP option/extension header size nor IPsec
1834  * header size from MSS advertisement.  MSS option must hold the maximum
1835  * segment size we can accept, so it must always be:
1836  *	 max(if mtu) - ip header - tcp header
1837  */
1838 u_long
1839 tcp_mss_to_advertise(const struct ifnet *ifp, int af)
1840 {
1841 	extern u_long in_maxmtu;
1842 	u_long mss = 0;
1843 	u_long hdrsiz;
1844 
1845 	/*
1846 	 * In order to avoid defeating path MTU discovery on the peer,
1847 	 * we advertise the max MTU of all attached networks as our MSS,
1848 	 * per RFC 1191, section 3.1.
1849 	 *
1850 	 * We provide the option to advertise just the MTU of
1851 	 * the interface on which we hope this connection will
1852 	 * be receiving.  If we are responding to a SYN, we
1853 	 * will have a pretty good idea about this, but when
1854 	 * initiating a connection there is a bit more doubt.
1855 	 *
1856 	 * We also need to ensure that loopback has a large enough
1857 	 * MSS, as the loopback MTU is never included in in_maxmtu.
1858 	 */
1859 
1860 	if (ifp != NULL)
1861 		switch (af) {
1862 		case AF_INET:
1863 			mss = ifp->if_mtu;
1864 			break;
1865 #ifdef INET6
1866 		case AF_INET6:
1867 			mss = IN6_LINKMTU(ifp);
1868 			break;
1869 #endif
1870 		}
1871 
1872 	if (tcp_mss_ifmtu == 0)
1873 		switch (af) {
1874 		case AF_INET:
1875 			mss = max(in_maxmtu, mss);
1876 			break;
1877 #ifdef INET6
1878 		case AF_INET6:
1879 			mss = max(in6_maxmtu, mss);
1880 			break;
1881 #endif
1882 		}
1883 
1884 	switch (af) {
1885 	case AF_INET:
1886 		hdrsiz = sizeof(struct ip);
1887 		break;
1888 #ifdef INET6
1889 	case AF_INET6:
1890 		hdrsiz = sizeof(struct ip6_hdr);
1891 		break;
1892 #endif
1893 	default:
1894 		hdrsiz = 0;
1895 		break;
1896 	}
1897 	hdrsiz += sizeof(struct tcphdr);
1898 	if (mss > hdrsiz)
1899 		mss -= hdrsiz;
1900 
1901 	mss = max(tcp_mssdflt, mss);
1902 	return (mss);
1903 }
1904 
1905 /*
1906  * Set connection variables based on the peer's advertised MSS.
1907  * We are passed the TCPCB for the actual connection.  If we
1908  * are the server, we are called by the compressed state engine
1909  * when the 3-way handshake is complete.  If we are the client,
1910  * we are called when we receive the SYN,ACK from the server.
1911  *
1912  * NOTE: Our advertised MSS value must be initialized in the TCPCB
1913  * before this routine is called!
1914  */
1915 void
1916 tcp_mss_from_peer(struct tcpcb *tp, int offer)
1917 {
1918 	struct socket *so;
1919 #if defined(RTV_SPIPE) || defined(RTV_SSTHRESH)
1920 	struct rtentry *rt;
1921 #endif
1922 	u_long bufsize;
1923 	int mss;
1924 
1925 #ifdef DIAGNOSTIC
1926 	if (tp->t_inpcb && tp->t_in6pcb)
1927 		panic("tcp_mss_from_peer: both t_inpcb and t_in6pcb are set");
1928 #endif
1929 	so = NULL;
1930 	rt = NULL;
1931 #ifdef INET
1932 	if (tp->t_inpcb) {
1933 		so = tp->t_inpcb->inp_socket;
1934 #if defined(RTV_SPIPE) || defined(RTV_SSTHRESH)
1935 		rt = in_pcbrtentry(tp->t_inpcb);
1936 #endif
1937 	}
1938 #endif
1939 #ifdef INET6
1940 	if (tp->t_in6pcb) {
1941 		so = tp->t_in6pcb->in6p_socket;
1942 #if defined(RTV_SPIPE) || defined(RTV_SSTHRESH)
1943 		rt = in6_pcbrtentry(tp->t_in6pcb);
1944 #endif
1945 	}
1946 #endif
1947 
1948 	/*
1949 	 * As per RFC1122, use the default MSS value, unless they
1950 	 * sent us an offer.  Do not accept offers less than 256 bytes.
1951 	 */
1952 	mss = tcp_mssdflt;
1953 	if (offer)
1954 		mss = offer;
1955 	mss = max(mss, 256);		/* sanity */
1956 	tp->t_peermss = mss;
1957 	mss -= tcp_optlen(tp);
1958 #ifdef INET
1959 	if (tp->t_inpcb)
1960 		mss -= ip_optlen(tp->t_inpcb);
1961 #endif
1962 #ifdef INET6
1963 	if (tp->t_in6pcb)
1964 		mss -= ip6_optlen(tp->t_in6pcb);
1965 #endif
1966 
1967 	/*
1968 	 * If there's a pipesize, change the socket buffer to that size.
1969 	 * Make the socket buffer an integral number of MSS units.  If
1970 	 * the MSS is larger than the socket buffer, artificially decrease
1971 	 * the MSS.
1972 	 */
1973 #ifdef RTV_SPIPE
1974 	if (rt != NULL && rt->rt_rmx.rmx_sendpipe != 0)
1975 		bufsize = rt->rt_rmx.rmx_sendpipe;
1976 	else
1977 #endif
1978 	{
1979 		KASSERT(so != NULL);
1980 		bufsize = so->so_snd.sb_hiwat;
1981 	}
1982 	if (bufsize < mss)
1983 		mss = bufsize;
1984 	else {
1985 		bufsize = roundup(bufsize, mss);
1986 		if (bufsize > sb_max)
1987 			bufsize = sb_max;
1988 		(void) sbreserve(&so->so_snd, bufsize, so);
1989 	}
1990 	tp->t_segsz = mss;
1991 
1992 #ifdef RTV_SSTHRESH
1993 	if (rt != NULL && rt->rt_rmx.rmx_ssthresh) {
1994 		/*
1995 		 * There's some sort of gateway or interface buffer
1996 		 * limit on the path.  Use this to set the slow
1997 		 * start threshold, but set the threshold to no less
1998 		 * than 2 * MSS.
1999 		 */
2000 		tp->snd_ssthresh = max(2 * mss, rt->rt_rmx.rmx_ssthresh);
2001 	}
2002 #endif
2003 }
2004 
2005 /*
2006  * Processing necessary when a TCP connection is established.
2007  */
2008 void
2009 tcp_established(struct tcpcb *tp)
2010 {
2011 	struct socket *so;
2012 #ifdef RTV_RPIPE
2013 	struct rtentry *rt;
2014 #endif
2015 	u_long bufsize;
2016 
2017 #ifdef DIAGNOSTIC
2018 	if (tp->t_inpcb && tp->t_in6pcb)
2019 		panic("tcp_established: both t_inpcb and t_in6pcb are set");
2020 #endif
2021 	so = NULL;
2022 	rt = NULL;
2023 #ifdef INET
2024 	if (tp->t_inpcb) {
2025 		so = tp->t_inpcb->inp_socket;
2026 #if defined(RTV_RPIPE)
2027 		rt = in_pcbrtentry(tp->t_inpcb);
2028 #endif
2029 	}
2030 #endif
2031 #ifdef INET6
2032 	if (tp->t_in6pcb) {
2033 		so = tp->t_in6pcb->in6p_socket;
2034 #if defined(RTV_RPIPE)
2035 		rt = in6_pcbrtentry(tp->t_in6pcb);
2036 #endif
2037 	}
2038 #endif
2039 
2040 	tp->t_state = TCPS_ESTABLISHED;
2041 	TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle);
2042 
2043 #ifdef RTV_RPIPE
2044 	if (rt != NULL && rt->rt_rmx.rmx_recvpipe != 0)
2045 		bufsize = rt->rt_rmx.rmx_recvpipe;
2046 	else
2047 #endif
2048 	{
2049 		KASSERT(so != NULL);
2050 		bufsize = so->so_rcv.sb_hiwat;
2051 	}
2052 	if (bufsize > tp->t_ourmss) {
2053 		bufsize = roundup(bufsize, tp->t_ourmss);
2054 		if (bufsize > sb_max)
2055 			bufsize = sb_max;
2056 		(void) sbreserve(&so->so_rcv, bufsize, so);
2057 	}
2058 }
2059 
2060 /*
2061  * Check if there's an initial rtt or rttvar.  Convert from the
2062  * route-table units to scaled multiples of the slow timeout timer.
2063  * Called only during the 3-way handshake.
2064  */
2065 void
2066 tcp_rmx_rtt(struct tcpcb *tp)
2067 {
2068 #ifdef RTV_RTT
2069 	struct rtentry *rt = NULL;
2070 	int rtt;
2071 
2072 #ifdef DIAGNOSTIC
2073 	if (tp->t_inpcb && tp->t_in6pcb)
2074 		panic("tcp_rmx_rtt: both t_inpcb and t_in6pcb are set");
2075 #endif
2076 #ifdef INET
2077 	if (tp->t_inpcb)
2078 		rt = in_pcbrtentry(tp->t_inpcb);
2079 #endif
2080 #ifdef INET6
2081 	if (tp->t_in6pcb)
2082 		rt = in6_pcbrtentry(tp->t_in6pcb);
2083 #endif
2084 	if (rt == NULL)
2085 		return;
2086 
2087 	if (tp->t_srtt == 0 && (rtt = rt->rt_rmx.rmx_rtt)) {
2088 		/*
2089 		 * XXX The lock bit for MTU indicates that the value
2090 		 * is also a minimum value; this is subject to time.
2091 		 */
2092 		if (rt->rt_rmx.rmx_locks & RTV_RTT)
2093 			TCPT_RANGESET(tp->t_rttmin,
2094 			    rtt / (RTM_RTTUNIT / PR_SLOWHZ),
2095 			    TCPTV_MIN, TCPTV_REXMTMAX);
2096 		tp->t_srtt = rtt /
2097 		    ((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTT_SHIFT + 2));
2098 		if (rt->rt_rmx.rmx_rttvar) {
2099 			tp->t_rttvar = rt->rt_rmx.rmx_rttvar /
2100 			    ((RTM_RTTUNIT / PR_SLOWHZ) >>
2101 				(TCP_RTTVAR_SHIFT + 2));
2102 		} else {
2103 			/* Default variation is +- 1 rtt */
2104 			tp->t_rttvar =
2105 			    tp->t_srtt >> (TCP_RTT_SHIFT - TCP_RTTVAR_SHIFT);
2106 		}
2107 		TCPT_RANGESET(tp->t_rxtcur,
2108 		    ((tp->t_srtt >> 2) + tp->t_rttvar) >> (1 + 2),
2109 		    tp->t_rttmin, TCPTV_REXMTMAX);
2110 	}
2111 #endif
2112 }
2113 
2114 tcp_seq	 tcp_iss_seq = 0;	/* tcp initial seq # */
2115 #if NRND > 0
2116 u_int8_t tcp_iss_secret[16];	/* 128 bits; should be plenty */
2117 #endif
2118 
2119 /*
2120  * Get a new sequence value given a tcp control block
2121  */
2122 tcp_seq
2123 tcp_new_iss(struct tcpcb *tp, tcp_seq addin)
2124 {
2125 
2126 #ifdef INET
2127 	if (tp->t_inpcb != NULL) {
2128 		return (tcp_new_iss1(&tp->t_inpcb->inp_laddr,
2129 		    &tp->t_inpcb->inp_faddr, tp->t_inpcb->inp_lport,
2130 		    tp->t_inpcb->inp_fport, sizeof(tp->t_inpcb->inp_laddr),
2131 		    addin));
2132 	}
2133 #endif
2134 #ifdef INET6
2135 	if (tp->t_in6pcb != NULL) {
2136 		return (tcp_new_iss1(&tp->t_in6pcb->in6p_laddr,
2137 		    &tp->t_in6pcb->in6p_faddr, tp->t_in6pcb->in6p_lport,
2138 		    tp->t_in6pcb->in6p_fport, sizeof(tp->t_in6pcb->in6p_laddr),
2139 		    addin));
2140 	}
2141 #endif
2142 	/* Not possible. */
2143 	panic("tcp_new_iss");
2144 }
2145 
2146 /*
2147  * This routine actually generates a new TCP initial sequence number.
2148  */
2149 tcp_seq
2150 tcp_new_iss1(void *laddr, void *faddr, u_int16_t lport, u_int16_t fport,
2151     size_t addrsz, tcp_seq addin)
2152 {
2153 	tcp_seq tcp_iss;
2154 
2155 #if NRND > 0
2156 	static int beenhere;
2157 
2158 	/*
2159 	 * If we haven't been here before, initialize our cryptographic
2160 	 * hash secret.
2161 	 */
2162 	if (beenhere == 0) {
2163 		rnd_extract_data(tcp_iss_secret, sizeof(tcp_iss_secret),
2164 		    RND_EXTRACT_ANY);
2165 		beenhere = 1;
2166 	}
2167 
2168 	if (tcp_do_rfc1948) {
2169 		MD5_CTX ctx;
2170 		u_int8_t hash[16];	/* XXX MD5 knowledge */
2171 
2172 		/*
2173 		 * Compute the base value of the ISS.  It is a hash
2174 		 * of (saddr, sport, daddr, dport, secret).
2175 		 */
2176 		MD5Init(&ctx);
2177 
2178 		MD5Update(&ctx, (u_char *) laddr, addrsz);
2179 		MD5Update(&ctx, (u_char *) &lport, sizeof(lport));
2180 
2181 		MD5Update(&ctx, (u_char *) faddr, addrsz);
2182 		MD5Update(&ctx, (u_char *) &fport, sizeof(fport));
2183 
2184 		MD5Update(&ctx, tcp_iss_secret, sizeof(tcp_iss_secret));
2185 
2186 		MD5Final(hash, &ctx);
2187 
2188 		memcpy(&tcp_iss, hash, sizeof(tcp_iss));
2189 
2190 		/*
2191 		 * Now increment our "timer", and add it in to
2192 		 * the computed value.
2193 		 *
2194 		 * XXX Use `addin'?
2195 		 * XXX TCP_ISSINCR too large to use?
2196 		 */
2197 		tcp_iss_seq += TCP_ISSINCR;
2198 #ifdef TCPISS_DEBUG
2199 		printf("ISS hash 0x%08x, ", tcp_iss);
2200 #endif
2201 		tcp_iss += tcp_iss_seq + addin;
2202 #ifdef TCPISS_DEBUG
2203 		printf("new ISS 0x%08x\n", tcp_iss);
2204 #endif
2205 	} else
2206 #endif /* NRND > 0 */
2207 	{
2208 		/*
2209 		 * Randomize.
2210 		 */
2211 #if NRND > 0
2212 		rnd_extract_data(&tcp_iss, sizeof(tcp_iss), RND_EXTRACT_ANY);
2213 #else
2214 		tcp_iss = arc4random();
2215 #endif
2216 
2217 		/*
2218 		 * If we were asked to add some amount to a known value,
2219 		 * we will take a random value obtained above, mask off
2220 		 * the upper bits, and add in the known value.  We also
2221 		 * add in a constant to ensure that we are at least a
2222 		 * certain distance from the original value.
2223 		 *
2224 		 * This is used when an old connection is in timed wait
2225 		 * and we have a new one coming in, for instance.
2226 		 */
2227 		if (addin != 0) {
2228 #ifdef TCPISS_DEBUG
2229 			printf("Random %08x, ", tcp_iss);
2230 #endif
2231 			tcp_iss &= TCP_ISS_RANDOM_MASK;
2232 			tcp_iss += addin + TCP_ISSINCR;
2233 #ifdef TCPISS_DEBUG
2234 			printf("Old ISS %08x, ISS %08x\n", addin, tcp_iss);
2235 #endif
2236 		} else {
2237 			tcp_iss &= TCP_ISS_RANDOM_MASK;
2238 			tcp_iss += tcp_iss_seq;
2239 			tcp_iss_seq += TCP_ISSINCR;
2240 #ifdef TCPISS_DEBUG
2241 			printf("ISS %08x\n", tcp_iss);
2242 #endif
2243 		}
2244 	}
2245 
2246 	if (tcp_compat_42) {
2247 		/*
2248 		 * Limit it to the positive range for really old TCP
2249 		 * implementations.
2250 		 * Just AND off the top bit instead of checking if
2251 		 * is set first - saves a branch 50% of the time.
2252 		 */
2253 		tcp_iss &= 0x7fffffff;		/* XXX */
2254 	}
2255 
2256 	return (tcp_iss);
2257 }
2258 
2259 #if defined(IPSEC) || defined(FAST_IPSEC)
2260 /* compute ESP/AH header size for TCP, including outer IP header. */
2261 size_t
2262 ipsec4_hdrsiz_tcp(struct tcpcb *tp)
2263 {
2264 	struct inpcb *inp;
2265 	size_t hdrsiz;
2266 
2267 	/* XXX mapped addr case (tp->t_in6pcb) */
2268 	if (!tp || !tp->t_template || !(inp = tp->t_inpcb))
2269 		return 0;
2270 	switch (tp->t_family) {
2271 	case AF_INET:
2272 		/* XXX: should use currect direction. */
2273 		hdrsiz = ipsec4_hdrsiz(tp->t_template, IPSEC_DIR_OUTBOUND, inp);
2274 		break;
2275 	default:
2276 		hdrsiz = 0;
2277 		break;
2278 	}
2279 
2280 	return hdrsiz;
2281 }
2282 
2283 #ifdef INET6
2284 size_t
2285 ipsec6_hdrsiz_tcp(struct tcpcb *tp)
2286 {
2287 	struct in6pcb *in6p;
2288 	size_t hdrsiz;
2289 
2290 	if (!tp || !tp->t_template || !(in6p = tp->t_in6pcb))
2291 		return 0;
2292 	switch (tp->t_family) {
2293 	case AF_INET6:
2294 		/* XXX: should use currect direction. */
2295 		hdrsiz = ipsec6_hdrsiz(tp->t_template, IPSEC_DIR_OUTBOUND, in6p);
2296 		break;
2297 	case AF_INET:
2298 		/* mapped address case - tricky */
2299 	default:
2300 		hdrsiz = 0;
2301 		break;
2302 	}
2303 
2304 	return hdrsiz;
2305 }
2306 #endif
2307 #endif /*IPSEC*/
2308 
2309 /*
2310  * Determine the length of the TCP options for this connection.
2311  *
2312  * XXX:  What do we do for SACK, when we add that?  Just reserve
2313  *       all of the space?  Otherwise we can't exactly be incrementing
2314  *       cwnd by an amount that varies depending on the amount we last
2315  *       had to SACK!
2316  */
2317 
2318 u_int
2319 tcp_optlen(struct tcpcb *tp)
2320 {
2321 	u_int optlen;
2322 
2323 	optlen = 0;
2324 	if ((tp->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP|TF_NOOPT)) ==
2325 	    (TF_REQ_TSTMP | TF_RCVD_TSTMP))
2326 		optlen += TCPOLEN_TSTAMP_APPA;
2327 
2328 #ifdef TCP_SIGNATURE
2329 	if (tp->t_flags & TF_SIGNATURE)
2330 		optlen += TCPOLEN_SIGNATURE + 2;
2331 #endif /* TCP_SIGNATURE */
2332 
2333 	return optlen;
2334 }
2335 
2336 u_int
2337 tcp_hdrsz(struct tcpcb *tp)
2338 {
2339 	u_int hlen;
2340 
2341 	switch (tp->t_family) {
2342 #ifdef INET6
2343 	case AF_INET6:
2344 		hlen = sizeof(struct ip6_hdr);
2345 		break;
2346 #endif
2347 	case AF_INET:
2348 		hlen = sizeof(struct ip);
2349 		break;
2350 	default:
2351 		hlen = 0;
2352 		break;
2353 	}
2354 	hlen += sizeof(struct tcphdr);
2355 
2356 	if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
2357 	    (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)
2358 		hlen += TCPOLEN_TSTAMP_APPA;
2359 #ifdef TCP_SIGNATURE
2360 	if (tp->t_flags & TF_SIGNATURE)
2361 		hlen += TCPOLEN_SIGLEN;
2362 #endif
2363 	return hlen;
2364 }
2365