xref: /openbsd-src/sys/netinet/tcp_usrreq.c (revision a5429850edcc9dd5646cc8ddb251ed22eba08b09)
1 /*	$OpenBSD: tcp_usrreq.c,v 1.208 2022/09/13 09:05:47 mvs Exp $	*/
2 /*	$NetBSD: tcp_usrreq.c,v 1.20 1996/02/13 23:44:16 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
33  *
34  * NRL grants permission for redistribution and use in source and binary
35  * forms, with or without modification, of the software and documentation
36  * created at NRL provided that the following conditions are met:
37  *
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 3. All advertising materials mentioning features or use of this software
44  *    must display the following acknowledgements:
45  *	This product includes software developed by the University of
46  *	California, Berkeley and its contributors.
47  *	This product includes software developed at the Information
48  *	Technology Division, US Naval Research Laboratory.
49  * 4. Neither the name of the NRL nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
54  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
56  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
57  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
58  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
59  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
60  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
61  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
62  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
63  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
64  *
65  * The views and conclusions contained in the software and documentation
66  * are those of the authors and should not be interpreted as representing
67  * official policies, either expressed or implied, of the US Naval
68  * Research Laboratory (NRL).
69  */
70 
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/mbuf.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/protosw.h>
77 #include <sys/stat.h>
78 #include <sys/sysctl.h>
79 #include <sys/domain.h>
80 #include <sys/kernel.h>
81 #include <sys/pool.h>
82 #include <sys/proc.h>
83 
84 #include <net/if.h>
85 #include <net/if_var.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip.h>
91 #include <netinet/in_pcb.h>
92 #include <netinet/ip_var.h>
93 #include <netinet/tcp.h>
94 #include <netinet/tcp_fsm.h>
95 #include <netinet/tcp_seq.h>
96 #include <netinet/tcp_timer.h>
97 #include <netinet/tcp_var.h>
98 #include <netinet/tcp_debug.h>
99 
100 #ifdef INET6
101 #include <netinet6/in6_var.h>
102 #endif
103 
104 #ifndef TCP_SENDSPACE
105 #define	TCP_SENDSPACE	1024*16
106 #endif
107 u_int	tcp_sendspace = TCP_SENDSPACE;
108 #ifndef TCP_RECVSPACE
109 #define	TCP_RECVSPACE	1024*16
110 #endif
111 u_int	tcp_recvspace = TCP_RECVSPACE;
112 u_int	tcp_autorcvbuf_inc = 16 * 1024;
113 
114 const struct pr_usrreqs tcp_usrreqs = {
115 	.pru_attach	= tcp_attach,
116 	.pru_detach	= tcp_detach,
117 	.pru_bind	= tcp_bind,
118 	.pru_listen	= tcp_listen,
119 	.pru_connect	= tcp_connect,
120 	.pru_accept	= tcp_accept,
121 	.pru_disconnect	= tcp_disconnect,
122 	.pru_shutdown	= tcp_shutdown,
123 	.pru_rcvd	= tcp_rcvd,
124 	.pru_send	= tcp_send,
125 	.pru_abort	= tcp_abort,
126 	.pru_sense	= tcp_sense,
127 	.pru_rcvoob	= tcp_rcvoob,
128 	.pru_sendoob	= tcp_sendoob,
129 	.pru_control	= in_control,
130 	.pru_sockaddr	= tcp_sockaddr,
131 	.pru_peeraddr	= tcp_peeraddr,
132 };
133 
134 #ifdef INET6
135 const struct pr_usrreqs tcp6_usrreqs = {
136 	.pru_attach	= tcp_attach,
137 	.pru_detach	= tcp_detach,
138 	.pru_bind	= tcp_bind,
139 	.pru_listen	= tcp_listen,
140 	.pru_connect	= tcp_connect,
141 	.pru_accept	= tcp_accept,
142 	.pru_disconnect	= tcp_disconnect,
143 	.pru_shutdown	= tcp_shutdown,
144 	.pru_rcvd	= tcp_rcvd,
145 	.pru_send	= tcp_send,
146 	.pru_abort	= tcp_abort,
147 	.pru_sense	= tcp_sense,
148 	.pru_rcvoob	= tcp_rcvoob,
149 	.pru_sendoob	= tcp_sendoob,
150 	.pru_control	= in6_control,
151 	.pru_sockaddr	= tcp_sockaddr,
152 	.pru_peeraddr	= tcp_peeraddr,
153 };
154 #endif
155 
156 static int pr_slowhz = PR_SLOWHZ;
157 const struct sysctl_bounded_args tcpctl_vars[] = {
158 	{ TCPCTL_SLOWHZ, &pr_slowhz, SYSCTL_INT_READONLY },
159 	{ TCPCTL_RFC1323, &tcp_do_rfc1323, 0, 1 },
160 	{ TCPCTL_KEEPINITTIME, &tcptv_keep_init, 1, 3 * TCPTV_KEEP_INIT },
161 	{ TCPCTL_KEEPIDLE, &tcp_keepidle, 1, 5 * TCPTV_KEEP_IDLE },
162 	{ TCPCTL_KEEPINTVL, &tcp_keepintvl, 1, 3 * TCPTV_KEEPINTVL },
163 	{ TCPCTL_SACK, &tcp_do_sack, 0, 1 },
164 	{ TCPCTL_MSSDFLT, &tcp_mssdflt, TCP_MSS, 65535 },
165 	{ TCPCTL_RSTPPSLIMIT, &tcp_rst_ppslim, 1, 1000 * 1000 },
166 	{ TCPCTL_ACK_ON_PUSH, &tcp_ack_on_push, 0, 1 },
167 #ifdef TCP_ECN
168 	{ TCPCTL_ECN, &tcp_do_ecn, 0, 1 },
169 #endif
170 	{ TCPCTL_SYN_CACHE_LIMIT, &tcp_syn_cache_limit, 1, 1000 * 1000 },
171 	{ TCPCTL_SYN_BUCKET_LIMIT, &tcp_syn_bucket_limit, 1, INT_MAX },
172 	{ TCPCTL_RFC3390, &tcp_do_rfc3390, 0, 2 },
173 	{ TCPCTL_ALWAYS_KEEPALIVE, &tcp_always_keepalive, 0, 1 },
174 };
175 
176 struct	inpcbtable tcbtable;
177 
178 int	tcp_fill_info(struct tcpcb *, struct socket *, struct mbuf *);
179 int	tcp_ident(void *, size_t *, void *, size_t, int);
180 
181 static inline int tcp_sogetpcb(struct socket *, struct inpcb **,
182                       struct tcpcb **);
183 
184 static inline int
185 tcp_sogetpcb(struct socket *so, struct inpcb **rinp, struct tcpcb **rtp)
186 {
187 	struct inpcb *inp;
188 	struct tcpcb *tp;
189 
190 	/*
191 	 * When a TCP is attached to a socket, then there will be
192 	 * a (struct inpcb) pointed at by the socket, and this
193 	 * structure will point at a subsidiary (struct tcpcb).
194 	 */
195 	if ((inp = sotoinpcb(so)) == NULL || (tp = intotcpcb(inp)) == NULL) {
196 		if (so->so_error)
197 			return so->so_error;
198 		return EINVAL;
199 	}
200 
201 	*rinp = inp;
202 	*rtp = tp;
203 
204 	return 0;
205 }
206 
207 /*
208  * Export internal TCP state information via a struct tcp_info without
209  * leaking any sensitive information. Sequence numbers are reported
210  * relative to the initial sequence number.
211  */
212 int
213 tcp_fill_info(struct tcpcb *tp, struct socket *so, struct mbuf *m)
214 {
215 	struct proc *p = curproc;
216 	struct tcp_info *ti;
217 	u_int t = 1000000 / PR_SLOWHZ;
218 	uint32_t now;
219 
220 	if (sizeof(*ti) > MLEN) {
221 		MCLGETL(m, M_WAITOK, sizeof(*ti));
222 		if (!ISSET(m->m_flags, M_EXT))
223 			return ENOMEM;
224 	}
225 	ti = mtod(m, struct tcp_info *);
226 	m->m_len = sizeof(*ti);
227 	memset(ti, 0, sizeof(*ti));
228 	now = READ_ONCE(tcp_now);
229 
230 	ti->tcpi_state = tp->t_state;
231 	if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
232 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
233 	if (tp->t_flags & TF_SACK_PERMIT)
234 		ti->tcpi_options |= TCPI_OPT_SACK;
235 	if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
236 		ti->tcpi_options |= TCPI_OPT_WSCALE;
237 		ti->tcpi_snd_wscale = tp->snd_scale;
238 		ti->tcpi_rcv_wscale = tp->rcv_scale;
239 	}
240 #ifdef TCP_ECN
241 	if (tp->t_flags & TF_ECN_PERMIT)
242 		ti->tcpi_options |= TCPI_OPT_ECN;
243 #endif
244 
245 	ti->tcpi_rto = tp->t_rxtcur * t;
246 	ti->tcpi_snd_mss = tp->t_maxseg;
247 	ti->tcpi_rcv_mss = tp->t_peermss;
248 
249 	ti->tcpi_last_data_sent = (now - tp->t_sndtime) * t;
250 	ti->tcpi_last_ack_sent = (now - tp->t_sndacktime) * t;
251 	ti->tcpi_last_data_recv = (now - tp->t_rcvtime) * t;
252 	ti->tcpi_last_ack_recv = (now - tp->t_rcvacktime) * t;
253 
254 	ti->tcpi_rtt = ((uint64_t)tp->t_srtt * t) >>
255 	    (TCP_RTT_SHIFT + TCP_RTT_BASE_SHIFT);
256 	ti->tcpi_rttvar = ((uint64_t)tp->t_rttvar * t) >>
257 	    (TCP_RTTVAR_SHIFT + TCP_RTT_BASE_SHIFT);
258 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
259 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
260 
261 	ti->tcpi_rcv_space = tp->rcv_wnd;
262 
263 	/*
264 	 * Provide only minimal information for unprivileged processes.
265 	 */
266 	if (suser(p) != 0)
267 		return 0;
268 
269 	/* FreeBSD-specific extension fields for tcp_info.  */
270 	ti->tcpi_snd_wnd = tp->snd_wnd;
271 	ti->tcpi_snd_nxt = tp->snd_nxt - tp->iss;
272 	ti->tcpi_rcv_nxt = tp->rcv_nxt - tp->irs;
273 	/* missing tcpi_toe_tid */
274 	ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
275 	ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
276 	ti->tcpi_snd_zerowin = tp->t_sndzerowin;
277 
278 	/* OpenBSD extensions */
279 	ti->tcpi_rttmin = tp->t_rttmin * t;
280 	ti->tcpi_max_sndwnd = tp->max_sndwnd;
281 	ti->tcpi_rcv_adv = tp->rcv_adv - tp->irs;
282 	ti->tcpi_rcv_up = tp->rcv_up - tp->irs;
283 	ti->tcpi_snd_una = tp->snd_una - tp->iss;
284 	ti->tcpi_snd_up = tp->snd_up - tp->iss;
285 	ti->tcpi_snd_wl1 = tp->snd_wl1 - tp->iss;
286 	ti->tcpi_snd_wl2 = tp->snd_wl2 - tp->iss;
287 	ti->tcpi_snd_max = tp->snd_max - tp->iss;
288 
289 	ti->tcpi_ts_recent = tp->ts_recent; /* XXX value from the wire */
290 	ti->tcpi_ts_recent_age = (now - tp->ts_recent_age) * t;
291 	ti->tcpi_rfbuf_cnt = tp->rfbuf_cnt;
292 	ti->tcpi_rfbuf_ts = (now - tp->rfbuf_ts) * t;
293 
294 	ti->tcpi_so_rcv_sb_cc = so->so_rcv.sb_cc;
295 	ti->tcpi_so_rcv_sb_hiwat = so->so_rcv.sb_hiwat;
296 	ti->tcpi_so_rcv_sb_lowat = so->so_rcv.sb_lowat;
297 	ti->tcpi_so_rcv_sb_wat = so->so_rcv.sb_wat;
298 	ti->tcpi_so_snd_sb_cc = so->so_snd.sb_cc;
299 	ti->tcpi_so_snd_sb_hiwat = so->so_snd.sb_hiwat;
300 	ti->tcpi_so_snd_sb_lowat = so->so_snd.sb_lowat;
301 	ti->tcpi_so_snd_sb_wat = so->so_snd.sb_wat;
302 
303 	return 0;
304 }
305 
306 int
307 tcp_ctloutput(int op, struct socket *so, int level, int optname,
308     struct mbuf *m)
309 {
310 	int error = 0;
311 	struct inpcb *inp;
312 	struct tcpcb *tp;
313 	int i;
314 
315 	inp = sotoinpcb(so);
316 	if (inp == NULL)
317 		return (ECONNRESET);
318 	if (level != IPPROTO_TCP) {
319 		switch (so->so_proto->pr_domain->dom_family) {
320 #ifdef INET6
321 		case PF_INET6:
322 			error = ip6_ctloutput(op, so, level, optname, m);
323 			break;
324 #endif /* INET6 */
325 		case PF_INET:
326 			error = ip_ctloutput(op, so, level, optname, m);
327 			break;
328 		default:
329 			error = EAFNOSUPPORT;	/*?*/
330 			break;
331 		}
332 		return (error);
333 	}
334 	tp = intotcpcb(inp);
335 
336 	switch (op) {
337 
338 	case PRCO_SETOPT:
339 		switch (optname) {
340 
341 		case TCP_NODELAY:
342 			if (m == NULL || m->m_len < sizeof (int))
343 				error = EINVAL;
344 			else if (*mtod(m, int *))
345 				tp->t_flags |= TF_NODELAY;
346 			else
347 				tp->t_flags &= ~TF_NODELAY;
348 			break;
349 
350 		case TCP_NOPUSH:
351 			if (m == NULL || m->m_len < sizeof (int))
352 				error = EINVAL;
353 			else if (*mtod(m, int *))
354 				tp->t_flags |= TF_NOPUSH;
355 			else if (tp->t_flags & TF_NOPUSH) {
356 				tp->t_flags &= ~TF_NOPUSH;
357 				if (TCPS_HAVEESTABLISHED(tp->t_state))
358 					error = tcp_output(tp);
359 			}
360 			break;
361 
362 		case TCP_MAXSEG:
363 			if (m == NULL || m->m_len < sizeof (int)) {
364 				error = EINVAL;
365 				break;
366 			}
367 
368 			i = *mtod(m, int *);
369 			if (i > 0 && i <= tp->t_maxseg)
370 				tp->t_maxseg = i;
371 			else
372 				error = EINVAL;
373 			break;
374 
375 		case TCP_SACK_ENABLE:
376 			if (m == NULL || m->m_len < sizeof (int)) {
377 				error = EINVAL;
378 				break;
379 			}
380 
381 			if (TCPS_HAVEESTABLISHED(tp->t_state)) {
382 				error = EPERM;
383 				break;
384 			}
385 
386 			if (tp->t_flags & TF_SIGNATURE) {
387 				error = EPERM;
388 				break;
389 			}
390 
391 			if (*mtod(m, int *))
392 				tp->sack_enable = 1;
393 			else
394 				tp->sack_enable = 0;
395 			break;
396 #ifdef TCP_SIGNATURE
397 		case TCP_MD5SIG:
398 			if (m == NULL || m->m_len < sizeof (int)) {
399 				error = EINVAL;
400 				break;
401 			}
402 
403 			if (TCPS_HAVEESTABLISHED(tp->t_state)) {
404 				error = EPERM;
405 				break;
406 			}
407 
408 			if (*mtod(m, int *)) {
409 				tp->t_flags |= TF_SIGNATURE;
410 				tp->sack_enable = 0;
411 			} else
412 				tp->t_flags &= ~TF_SIGNATURE;
413 			break;
414 #endif /* TCP_SIGNATURE */
415 		default:
416 			error = ENOPROTOOPT;
417 			break;
418 		}
419 		break;
420 
421 	case PRCO_GETOPT:
422 		switch (optname) {
423 		case TCP_NODELAY:
424 			m->m_len = sizeof(int);
425 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
426 			break;
427 		case TCP_NOPUSH:
428 			m->m_len = sizeof(int);
429 			*mtod(m, int *) = tp->t_flags & TF_NOPUSH;
430 			break;
431 		case TCP_MAXSEG:
432 			m->m_len = sizeof(int);
433 			*mtod(m, int *) = tp->t_maxseg;
434 			break;
435 		case TCP_SACK_ENABLE:
436 			m->m_len = sizeof(int);
437 			*mtod(m, int *) = tp->sack_enable;
438 			break;
439 		case TCP_INFO:
440 			error = tcp_fill_info(tp, so, m);
441 			break;
442 #ifdef TCP_SIGNATURE
443 		case TCP_MD5SIG:
444 			m->m_len = sizeof(int);
445 			*mtod(m, int *) = tp->t_flags & TF_SIGNATURE;
446 			break;
447 #endif
448 		default:
449 			error = ENOPROTOOPT;
450 			break;
451 		}
452 		break;
453 	}
454 	return (error);
455 }
456 
457 /*
458  * Attach TCP protocol to socket, allocating
459  * internet protocol control block, tcp control block,
460  * buffer space, and entering LISTEN state to accept connections.
461  */
462 int
463 tcp_attach(struct socket *so, int proto)
464 {
465 	struct tcpcb *tp;
466 	struct inpcb *inp;
467 	int error;
468 
469 	if (so->so_pcb)
470 		return EISCONN;
471 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0 ||
472 	    sbcheckreserve(so->so_snd.sb_wat, tcp_sendspace) ||
473 	    sbcheckreserve(so->so_rcv.sb_wat, tcp_recvspace)) {
474 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
475 		if (error)
476 			return (error);
477 	}
478 
479 	NET_ASSERT_LOCKED();
480 	error = in_pcballoc(so, &tcbtable);
481 	if (error)
482 		return (error);
483 	inp = sotoinpcb(so);
484 	tp = tcp_newtcpcb(inp);
485 	if (tp == NULL) {
486 		unsigned int nofd = so->so_state & SS_NOFDREF;	/* XXX */
487 
488 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
489 		in_pcbdetach(inp);
490 		so->so_state |= nofd;
491 		return (ENOBUFS);
492 	}
493 	tp->t_state = TCPS_CLOSED;
494 #ifdef INET6
495 	/* we disallow IPv4 mapped address completely. */
496 	if (inp->inp_flags & INP_IPV6)
497 		tp->pf = PF_INET6;
498 	else
499 		tp->pf = PF_INET;
500 #else
501 	tp->pf = PF_INET;
502 #endif
503 	if ((so->so_options & SO_LINGER) && so->so_linger == 0)
504 		so->so_linger = TCP_LINGERTIME;
505 
506 	if (so->so_options & SO_DEBUG)
507 		tcp_trace(TA_USER, TCPS_CLOSED, tp, tp, NULL, PRU_ATTACH, 0);
508 	return (0);
509 }
510 
511 int
512 tcp_detach(struct socket *so)
513 {
514 	struct inpcb *inp;
515 	struct tcpcb *otp = NULL, *tp;
516 	int error = 0;
517 	short ostate;
518 
519 	soassertlocked(so);
520 
521 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
522 		return (error);
523 
524 	if (so->so_options & SO_DEBUG) {
525 		otp = tp;
526 		ostate = tp->t_state;
527 	}
528 
529 	/*
530 	 * Detach the TCP protocol from the socket.
531 	 * If the protocol state is non-embryonic, then can't
532 	 * do this directly: have to initiate a PRU_DISCONNECT,
533 	 * which may finish later; embryonic TCB's can just
534 	 * be discarded here.
535 	 */
536 	tp = tcp_dodisconnect(tp);
537 
538 	if (otp)
539 		tcp_trace(TA_USER, ostate, tp, otp, NULL, PRU_DETACH, 0);
540 	return (error);
541 }
542 
543 /*
544  * Give the socket an address.
545  */
546 int
547 tcp_bind(struct socket *so, struct mbuf *nam, struct proc *p)
548 {
549 	struct inpcb *inp;
550 	struct tcpcb *tp;
551 	int error;
552 	short ostate;
553 
554 	soassertlocked(so);
555 
556 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
557 		return (error);
558 
559 	if (so->so_options & SO_DEBUG)
560 		ostate = tp->t_state;
561 
562 	error = in_pcbbind(inp, nam, p);
563 
564 	if (so->so_options & SO_DEBUG)
565 		tcp_trace(TA_USER, ostate, tp, tp, NULL, PRU_BIND, 0);
566 	return (error);
567 }
568 
569 /*
570  * Prepare to accept connections.
571  */
572 int
573 tcp_listen(struct socket *so)
574 {
575 	struct inpcb *inp;
576 	struct tcpcb *tp, *otp = NULL;
577 	int error;
578 	short ostate;
579 
580 	soassertlocked(so);
581 
582 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
583 		return (error);
584 
585 	if (so->so_options & SO_DEBUG) {
586 		otp = tp;
587 		ostate = tp->t_state;
588 	}
589 
590 	if (inp->inp_lport == 0)
591 		if ((error = in_pcbbind(inp, NULL, curproc)))
592 			goto out;
593 
594 	/*
595 	 * If the in_pcbbind() above is called, the tp->pf
596 	 * should still be whatever it was before.
597 	 */
598 	tp->t_state = TCPS_LISTEN;
599 
600 out:
601 	if (otp)
602 		tcp_trace(TA_USER, ostate, tp, otp, NULL, PRU_LISTEN, 0);
603 	return (error);
604 }
605 
606 /*
607  * Initiate connection to peer.
608  * Create a template for use in transmissions on this connection.
609  * Enter SYN_SENT state, and mark socket as connecting.
610  * Start keep-alive timer, and seed output sequence space.
611  * Send initial segment on connection.
612  */
613 int
614 tcp_connect(struct socket *so, struct mbuf *nam)
615 {
616 	struct inpcb *inp;
617 	struct tcpcb *tp, *otp = NULL;
618 	int error;
619 	short ostate;
620 
621 	soassertlocked(so);
622 
623 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
624 		return (error);
625 
626 	if (so->so_options & SO_DEBUG) {
627 		otp = tp;
628 		ostate = tp->t_state;
629 	}
630 
631 #ifdef INET6
632 	if (inp->inp_flags & INP_IPV6) {
633 		struct sockaddr_in6 *sin6;
634 
635 		if ((error = in6_nam2sin6(nam, &sin6)))
636 			goto out;
637 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
638 		    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
639 			error = EINVAL;
640 			goto out;
641 		}
642 		error = in6_pcbconnect(inp, nam);
643 	} else
644 #endif /* INET6 */
645 	{
646 		struct sockaddr_in *sin;
647 
648 		if ((error = in_nam2sin(nam, &sin)))
649 			goto out;
650 		if ((sin->sin_addr.s_addr == INADDR_ANY) ||
651 		    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
652 		    IN_MULTICAST(sin->sin_addr.s_addr) ||
653 		    in_broadcast(sin->sin_addr, inp->inp_rtableid)) {
654 			error = EINVAL;
655 			goto out;
656 		}
657 		error = in_pcbconnect(inp, nam);
658 	}
659 	if (error)
660 		goto out;
661 
662 	tp->t_template = tcp_template(tp);
663 	if (tp->t_template == 0) {
664 		in_pcbdisconnect(inp);
665 		error = ENOBUFS;
666 		goto out;
667 	}
668 
669 	so->so_state |= SS_CONNECTOUT;
670 
671 	/* Compute window scaling to request.  */
672 	tcp_rscale(tp, sb_max);
673 
674 	soisconnecting(so);
675 	tcpstat_inc(tcps_connattempt);
676 	tp->t_state = TCPS_SYN_SENT;
677 	TCP_TIMER_ARM(tp, TCPT_KEEP, tcptv_keep_init);
678 	tcp_set_iss_tsm(tp);
679 	tcp_sendseqinit(tp);
680 	tp->snd_last = tp->snd_una;
681 	error = tcp_output(tp);
682 
683 out:
684 	if (otp)
685 		tcp_trace(TA_USER, ostate, tp, otp, NULL, PRU_CONNECT, 0);
686 	return (error);
687 }
688 
689 /*
690  * Accept a connection.  Essentially all the work is done at higher
691  * levels; just return the address of the peer, storing through addr.
692  */
693 int
694 tcp_accept(struct socket *so, struct mbuf *nam)
695 {
696 	struct inpcb *inp;
697 	struct tcpcb *tp;
698 	int error;
699 	short ostate;
700 
701 	soassertlocked(so);
702 
703 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
704 		return (error);
705 
706 	if (so->so_options & SO_DEBUG)
707 		ostate = tp->t_state;
708 
709 #ifdef INET6
710 	if (inp->inp_flags & INP_IPV6)
711 		in6_setpeeraddr(inp, nam);
712 	else
713 #endif
714 		in_setpeeraddr(inp, nam);
715 
716 	if (so->so_options & SO_DEBUG)
717 		tcp_trace(TA_USER, ostate, tp, tp, NULL, PRU_ACCEPT, 0);
718 	return (error);
719 }
720 
721 /*
722  * Initiate disconnect from peer.
723  * If connection never passed embryonic stage, just drop;
724  * else if don't need to let data drain, then can just drop anyways,
725  * else have to begin TCP shutdown process: mark socket disconnecting,
726  * drain unread data, state switch to reflect user close, and
727  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
728  * when peer sends FIN and acks ours.
729  *
730  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
731  */
732 int
733 tcp_disconnect(struct socket *so)
734 {
735 	struct inpcb *inp;
736 	struct tcpcb *tp, *otp = NULL;
737 	int error;
738 	short ostate;
739 
740 	soassertlocked(so);
741 
742 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
743 		return (error);
744 
745 	if (so->so_options & SO_DEBUG) {
746 		otp = tp;
747 		ostate = tp->t_state;
748 	}
749 
750 	tp = tcp_dodisconnect(tp);
751 
752 	if (otp)
753 		tcp_trace(TA_USER, ostate, tp, otp, NULL, PRU_DISCONNECT, 0);
754 	return (0);
755 }
756 
757 /*
758  * Mark the connection as being incapable of further output.
759  */
760 int
761 tcp_shutdown(struct socket *so)
762 {
763 	struct inpcb *inp;
764 	struct tcpcb *tp, *otp = NULL;
765 	int error;
766 	short ostate;
767 
768 	soassertlocked(so);
769 
770 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
771 		return (error);
772 
773 	if (so->so_options & SO_DEBUG) {
774 		otp = tp;
775 		ostate = tp->t_state;
776 	}
777 
778 	if (so->so_state & SS_CANTSENDMORE)
779 		goto out;
780 
781 	socantsendmore(so);
782 	tp = tcp_usrclosed(tp);
783 	if (tp)
784 		error = tcp_output(tp);
785 
786 out:
787 	if (otp)
788 		tcp_trace(TA_USER, ostate, tp, otp, NULL, PRU_SHUTDOWN, 0);
789 	return (error);
790 }
791 
792 /*
793  * After a receive, possibly send window update to peer.
794  */
795 void
796 tcp_rcvd(struct socket *so)
797 {
798 	struct inpcb *inp;
799 	struct tcpcb *tp;
800 	short ostate;
801 
802 	soassertlocked(so);
803 
804 	if (tcp_sogetpcb(so, &inp, &tp))
805 		return;
806 
807 	if (so->so_options & SO_DEBUG)
808 		ostate = tp->t_state;
809 
810 	/*
811 	 * soreceive() calls this function when a user receives
812 	 * ancillary data on a listening socket. We don't call
813 	 * tcp_output in such a case, since there is no header
814 	 * template for a listening socket and hence the kernel
815 	 * will panic.
816 	 */
817 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
818 		(void) tcp_output(tp);
819 
820 	if (so->so_options & SO_DEBUG)
821 		tcp_trace(TA_USER, ostate, tp, tp, NULL, PRU_RCVD, 0);
822 }
823 
824 /*
825  * Do a send by putting data in output queue and updating urgent
826  * marker if URG set.  Possibly send more data.
827  */
828 int
829 tcp_send(struct socket *so, struct mbuf *m, struct mbuf *nam,
830     struct mbuf *control)
831 {
832 	struct inpcb *inp;
833 	struct tcpcb *tp;
834 	int error;
835 	short ostate;
836 
837 	soassertlocked(so);
838 
839 	if (control && control->m_len) {
840 		error = EINVAL;
841 		goto out;
842 	}
843 
844 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
845 		goto out;
846 
847 	if (so->so_options & SO_DEBUG)
848 		ostate = tp->t_state;
849 
850 	sbappendstream(so, &so->so_snd, m);
851 	m = NULL;
852 
853 	error = tcp_output(tp);
854 
855 	if (so->so_options & SO_DEBUG)
856 		tcp_trace(TA_USER, ostate, tp, tp, NULL, PRU_SEND, 0);
857 
858 out:
859 	m_freem(control);
860 	m_freem(m);
861 
862 	return (error);
863 }
864 
865 /*
866  * Abort the TCP.
867  */
868 int
869 tcp_abort(struct socket *so)
870 {
871 	struct inpcb *inp;
872 	struct tcpcb *tp, *otp = NULL;
873 	int error;
874 	short ostate;
875 
876 	soassertlocked(so);
877 
878 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
879 		return (error);
880 
881 	if (so->so_options & SO_DEBUG) {
882 		otp = tp;
883 		ostate = tp->t_state;
884 	}
885 
886 	tp = tcp_drop(tp, ECONNABORTED);
887 
888 	if (otp)
889 		tcp_trace(TA_USER, ostate, tp, otp, NULL, PRU_ABORT, 0);
890 	return (0);
891 }
892 
893 int
894 tcp_sense(struct socket *so, struct stat *ub)
895 {
896 	struct inpcb *inp;
897 	struct tcpcb *tp;
898 	int error;
899 
900 	soassertlocked(so);
901 
902 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
903 		return (error);
904 
905 	ub->st_blksize = so->so_snd.sb_hiwat;
906 
907 	if (so->so_options & SO_DEBUG)
908 		tcp_trace(TA_USER, tp->t_state, tp, tp, NULL, PRU_SENSE, 0);
909 	return (0);
910 }
911 
912 int
913 tcp_rcvoob(struct socket *so, struct mbuf *m, int flags)
914 {
915 	struct inpcb *inp;
916 	struct tcpcb *tp;
917 	int error;
918 
919 	soassertlocked(so);
920 
921 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
922 		return (error);
923 
924 	if ((so->so_oobmark == 0 &&
925 	    (so->so_state & SS_RCVATMARK) == 0) ||
926 	    so->so_options & SO_OOBINLINE ||
927 	    tp->t_oobflags & TCPOOB_HADDATA) {
928 		error = EINVAL;
929 		goto out;
930 	}
931 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
932 		error = EWOULDBLOCK;
933 		goto out;
934 	}
935 	m->m_len = 1;
936 	*mtod(m, caddr_t) = tp->t_iobc;
937 	if ((flags & MSG_PEEK) == 0)
938 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
939 out:
940 	if (so->so_options & SO_DEBUG)
941 		tcp_trace(TA_USER, tp->t_state, tp, tp, NULL, PRU_RCVOOB, 0);
942 	return (error);
943 }
944 
945 int
946 tcp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *nam,
947     struct mbuf *control)
948 {
949 	struct inpcb *inp;
950 	struct tcpcb *tp;
951 	int error;
952 	short ostate;
953 
954 	soassertlocked(so);
955 
956 	if (control && control->m_len) {
957 		error = EINVAL;
958 		goto release;
959 	}
960 
961 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
962 		goto release;
963 
964 	if (so->so_options & SO_DEBUG)
965 		ostate = tp->t_state;
966 
967 	if (sbspace(so, &so->so_snd) < -512) {
968 		error = ENOBUFS;
969 		goto out;
970 	}
971 
972 	/*
973 	 * According to RFC961 (Assigned Protocols),
974 	 * the urgent pointer points to the last octet
975 	 * of urgent data.  We continue, however,
976 	 * to consider it to indicate the first octet
977 	 * of data past the urgent section.
978 	 * Otherwise, snd_up should be one lower.
979 	 */
980 	sbappendstream(so, &so->so_snd, m);
981 	m = NULL;
982 	tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
983 	tp->t_force = 1;
984 	error = tcp_output(tp);
985 	tp->t_force = 0;
986 
987 out:
988 	if (so->so_options & SO_DEBUG)
989 		tcp_trace(TA_USER, ostate, tp, tp, NULL, PRU_SENDOOB, 0);
990 
991 release:
992 	m_freem(control);
993 	m_freem(m);
994 
995 	return (error);
996 }
997 
998 int
999 tcp_sockaddr(struct socket *so, struct mbuf *nam)
1000 {
1001 	struct inpcb *inp;
1002 	struct tcpcb *tp;
1003 	int error;
1004 
1005 	soassertlocked(so);
1006 
1007 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
1008 		return (error);
1009 
1010 #ifdef INET6
1011 	if (inp->inp_flags & INP_IPV6)
1012 		in6_setsockaddr(inp, nam);
1013 	else
1014 #endif
1015 		in_setsockaddr(inp, nam);
1016 
1017 	if (so->so_options & SO_DEBUG)
1018 		tcp_trace(TA_USER, tp->t_state, tp, tp, NULL,
1019 		    PRU_SOCKADDR, 0);
1020 	return (0);
1021 }
1022 
1023 int
1024 tcp_peeraddr(struct socket *so, struct mbuf *nam)
1025 {
1026 	struct inpcb *inp;
1027 	struct tcpcb *tp;
1028 	int error;
1029 
1030 	soassertlocked(so);
1031 
1032 	if ((error = tcp_sogetpcb(so, &inp, &tp)))
1033 		return (error);
1034 
1035 #ifdef INET6
1036 	if (inp->inp_flags & INP_IPV6)
1037 		in6_setpeeraddr(inp, nam);
1038 	else
1039 #endif
1040 		in_setpeeraddr(inp, nam);
1041 
1042 	if (so->so_options & SO_DEBUG)
1043 		tcp_trace(TA_USER, tp->t_state, tp, tp, NULL,
1044 		    PRU_PEERADDR, 0);
1045 	return (0);
1046 }
1047 
1048 /*
1049  * Initiate (or continue) disconnect.
1050  * If embryonic state, just send reset (once).
1051  * If in ``let data drain'' option and linger null, just drop.
1052  * Otherwise (hard), mark socket disconnecting and drop
1053  * current input data; switch states based on user close, and
1054  * send segment to peer (with FIN).
1055  */
1056 struct tcpcb *
1057 tcp_dodisconnect(struct tcpcb *tp)
1058 {
1059 	struct socket *so = tp->t_inpcb->inp_socket;
1060 
1061 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
1062 		tp = tcp_close(tp);
1063 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1064 		tp = tcp_drop(tp, 0);
1065 	else {
1066 		soisdisconnecting(so);
1067 		sbflush(so, &so->so_rcv);
1068 		tp = tcp_usrclosed(tp);
1069 		if (tp)
1070 			(void) tcp_output(tp);
1071 	}
1072 	return (tp);
1073 }
1074 
1075 /*
1076  * User issued close, and wish to trail through shutdown states:
1077  * if never received SYN, just forget it.  If got a SYN from peer,
1078  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1079  * If already got a FIN from peer, then almost done; go to LAST_ACK
1080  * state.  In all other cases, have already sent FIN to peer (e.g.
1081  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1082  * for peer to send FIN or not respond to keep-alives, etc.
1083  * We can let the user exit from the close as soon as the FIN is acked.
1084  */
1085 struct tcpcb *
1086 tcp_usrclosed(struct tcpcb *tp)
1087 {
1088 
1089 	switch (tp->t_state) {
1090 
1091 	case TCPS_CLOSED:
1092 	case TCPS_LISTEN:
1093 	case TCPS_SYN_SENT:
1094 		tp->t_state = TCPS_CLOSED;
1095 		tp = tcp_close(tp);
1096 		break;
1097 
1098 	case TCPS_SYN_RECEIVED:
1099 	case TCPS_ESTABLISHED:
1100 		tp->t_state = TCPS_FIN_WAIT_1;
1101 		break;
1102 
1103 	case TCPS_CLOSE_WAIT:
1104 		tp->t_state = TCPS_LAST_ACK;
1105 		break;
1106 	}
1107 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1108 		soisdisconnected(tp->t_inpcb->inp_socket);
1109 		/*
1110 		 * If we are in FIN_WAIT_2, we arrived here because the
1111 		 * application did a shutdown of the send side.  Like the
1112 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
1113 		 * a full close, we start a timer to make sure sockets are
1114 		 * not left in FIN_WAIT_2 forever.
1115 		 */
1116 		if (tp->t_state == TCPS_FIN_WAIT_2)
1117 			TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle);
1118 	}
1119 	return (tp);
1120 }
1121 
1122 /*
1123  * Look up a socket for ident or tcpdrop, ...
1124  */
1125 int
1126 tcp_ident(void *oldp, size_t *oldlenp, void *newp, size_t newlen, int dodrop)
1127 {
1128 	int error = 0;
1129 	struct tcp_ident_mapping tir;
1130 	struct inpcb *inp;
1131 	struct tcpcb *tp = NULL;
1132 	struct sockaddr_in *fin, *lin;
1133 #ifdef INET6
1134 	struct sockaddr_in6 *fin6, *lin6;
1135 	struct in6_addr f6, l6;
1136 #endif
1137 
1138 	NET_ASSERT_LOCKED();
1139 
1140 	if (dodrop) {
1141 		if (oldp != NULL || *oldlenp != 0)
1142 			return (EINVAL);
1143 		if (newp == NULL)
1144 			return (EPERM);
1145 		if (newlen < sizeof(tir))
1146 			return (ENOMEM);
1147 		if ((error = copyin(newp, &tir, sizeof (tir))) != 0 )
1148 			return (error);
1149 	} else {
1150 		if (oldp == NULL)
1151 			return (EINVAL);
1152 		if (*oldlenp < sizeof(tir))
1153 			return (ENOMEM);
1154 		if (newp != NULL || newlen != 0)
1155 			return (EINVAL);
1156 		if ((error = copyin(oldp, &tir, sizeof (tir))) != 0 )
1157 			return (error);
1158 	}
1159 	switch (tir.faddr.ss_family) {
1160 #ifdef INET6
1161 	case AF_INET6:
1162 		fin6 = (struct sockaddr_in6 *)&tir.faddr;
1163 		error = in6_embedscope(&f6, fin6, NULL);
1164 		if (error)
1165 			return EINVAL;	/*?*/
1166 		lin6 = (struct sockaddr_in6 *)&tir.laddr;
1167 		error = in6_embedscope(&l6, lin6, NULL);
1168 		if (error)
1169 			return EINVAL;	/*?*/
1170 		break;
1171 #endif
1172 	case AF_INET:
1173 		fin = (struct sockaddr_in *)&tir.faddr;
1174 		lin = (struct sockaddr_in *)&tir.laddr;
1175 		break;
1176 	default:
1177 		return (EINVAL);
1178 	}
1179 
1180 	switch (tir.faddr.ss_family) {
1181 #ifdef INET6
1182 	case AF_INET6:
1183 		inp = in6_pcblookup(&tcbtable, &f6,
1184 		    fin6->sin6_port, &l6, lin6->sin6_port, tir.rdomain);
1185 		break;
1186 #endif
1187 	case AF_INET:
1188 		inp = in_pcblookup(&tcbtable, fin->sin_addr,
1189 		    fin->sin_port, lin->sin_addr, lin->sin_port, tir.rdomain);
1190 		break;
1191 	default:
1192 		unhandled_af(tir.faddr.ss_family);
1193 	}
1194 
1195 	if (dodrop) {
1196 		if (inp && (tp = intotcpcb(inp)) &&
1197 		    ((inp->inp_socket->so_options & SO_ACCEPTCONN) == 0))
1198 			tp = tcp_drop(tp, ECONNABORTED);
1199 		else
1200 			error = ESRCH;
1201 		in_pcbunref(inp);
1202 		return (error);
1203 	}
1204 
1205 	if (inp == NULL) {
1206 		tcpstat_inc(tcps_pcbhashmiss);
1207 		switch (tir.faddr.ss_family) {
1208 #ifdef INET6
1209 		case AF_INET6:
1210 			inp = in6_pcblookup_listen(&tcbtable,
1211 			    &l6, lin6->sin6_port, NULL, tir.rdomain);
1212 			break;
1213 #endif
1214 		case AF_INET:
1215 			inp = in_pcblookup_listen(&tcbtable,
1216 			    lin->sin_addr, lin->sin_port, NULL, tir.rdomain);
1217 			break;
1218 		}
1219 	}
1220 
1221 	if (inp != NULL && (inp->inp_socket->so_state & SS_CONNECTOUT)) {
1222 		tir.ruid = inp->inp_socket->so_ruid;
1223 		tir.euid = inp->inp_socket->so_euid;
1224 	} else {
1225 		tir.ruid = -1;
1226 		tir.euid = -1;
1227 	}
1228 
1229 	*oldlenp = sizeof (tir);
1230 	error = copyout((void *)&tir, oldp, sizeof (tir));
1231 	in_pcbunref(inp);
1232 	return (error);
1233 }
1234 
1235 int
1236 tcp_sysctl_tcpstat(void *oldp, size_t *oldlenp, void *newp)
1237 {
1238 	uint64_t counters[tcps_ncounters];
1239 	struct tcpstat tcpstat;
1240 	struct syn_cache_set *set;
1241 	int i = 0;
1242 
1243 #define ASSIGN(field)	do { tcpstat.field = counters[i++]; } while (0)
1244 
1245 	memset(&tcpstat, 0, sizeof tcpstat);
1246 	counters_read(tcpcounters, counters, nitems(counters));
1247 	ASSIGN(tcps_connattempt);
1248 	ASSIGN(tcps_accepts);
1249 	ASSIGN(tcps_connects);
1250 	ASSIGN(tcps_drops);
1251 	ASSIGN(tcps_conndrops);
1252 	ASSIGN(tcps_closed);
1253 	ASSIGN(tcps_segstimed);
1254 	ASSIGN(tcps_rttupdated);
1255 	ASSIGN(tcps_delack);
1256 	ASSIGN(tcps_timeoutdrop);
1257 	ASSIGN(tcps_rexmttimeo);
1258 	ASSIGN(tcps_persisttimeo);
1259 	ASSIGN(tcps_persistdrop);
1260 	ASSIGN(tcps_keeptimeo);
1261 	ASSIGN(tcps_keepprobe);
1262 	ASSIGN(tcps_keepdrops);
1263 	ASSIGN(tcps_sndtotal);
1264 	ASSIGN(tcps_sndpack);
1265 	ASSIGN(tcps_sndbyte);
1266 	ASSIGN(tcps_sndrexmitpack);
1267 	ASSIGN(tcps_sndrexmitbyte);
1268 	ASSIGN(tcps_sndrexmitfast);
1269 	ASSIGN(tcps_sndacks);
1270 	ASSIGN(tcps_sndprobe);
1271 	ASSIGN(tcps_sndurg);
1272 	ASSIGN(tcps_sndwinup);
1273 	ASSIGN(tcps_sndctrl);
1274 	ASSIGN(tcps_rcvtotal);
1275 	ASSIGN(tcps_rcvpack);
1276 	ASSIGN(tcps_rcvbyte);
1277 	ASSIGN(tcps_rcvbadsum);
1278 	ASSIGN(tcps_rcvbadoff);
1279 	ASSIGN(tcps_rcvmemdrop);
1280 	ASSIGN(tcps_rcvnosec);
1281 	ASSIGN(tcps_rcvshort);
1282 	ASSIGN(tcps_rcvduppack);
1283 	ASSIGN(tcps_rcvdupbyte);
1284 	ASSIGN(tcps_rcvpartduppack);
1285 	ASSIGN(tcps_rcvpartdupbyte);
1286 	ASSIGN(tcps_rcvoopack);
1287 	ASSIGN(tcps_rcvoobyte);
1288 	ASSIGN(tcps_rcvpackafterwin);
1289 	ASSIGN(tcps_rcvbyteafterwin);
1290 	ASSIGN(tcps_rcvafterclose);
1291 	ASSIGN(tcps_rcvwinprobe);
1292 	ASSIGN(tcps_rcvdupack);
1293 	ASSIGN(tcps_rcvacktoomuch);
1294 	ASSIGN(tcps_rcvacktooold);
1295 	ASSIGN(tcps_rcvackpack);
1296 	ASSIGN(tcps_rcvackbyte);
1297 	ASSIGN(tcps_rcvwinupd);
1298 	ASSIGN(tcps_pawsdrop);
1299 	ASSIGN(tcps_predack);
1300 	ASSIGN(tcps_preddat);
1301 	ASSIGN(tcps_pcbhashmiss);
1302 	ASSIGN(tcps_noport);
1303 	ASSIGN(tcps_badsyn);
1304 	ASSIGN(tcps_dropsyn);
1305 	ASSIGN(tcps_rcvbadsig);
1306 	ASSIGN(tcps_rcvgoodsig);
1307 	ASSIGN(tcps_inswcsum);
1308 	ASSIGN(tcps_outswcsum);
1309 	ASSIGN(tcps_ecn_accepts);
1310 	ASSIGN(tcps_ecn_rcvece);
1311 	ASSIGN(tcps_ecn_rcvcwr);
1312 	ASSIGN(tcps_ecn_rcvce);
1313 	ASSIGN(tcps_ecn_sndect);
1314 	ASSIGN(tcps_ecn_sndece);
1315 	ASSIGN(tcps_ecn_sndcwr);
1316 	ASSIGN(tcps_cwr_ecn);
1317 	ASSIGN(tcps_cwr_frecovery);
1318 	ASSIGN(tcps_cwr_timeout);
1319 	ASSIGN(tcps_sc_added);
1320 	ASSIGN(tcps_sc_completed);
1321 	ASSIGN(tcps_sc_timed_out);
1322 	ASSIGN(tcps_sc_overflowed);
1323 	ASSIGN(tcps_sc_reset);
1324 	ASSIGN(tcps_sc_unreach);
1325 	ASSIGN(tcps_sc_bucketoverflow);
1326 	ASSIGN(tcps_sc_aborted);
1327 	ASSIGN(tcps_sc_dupesyn);
1328 	ASSIGN(tcps_sc_dropped);
1329 	ASSIGN(tcps_sc_collisions);
1330 	ASSIGN(tcps_sc_retransmitted);
1331 	ASSIGN(tcps_sc_seedrandom);
1332 	ASSIGN(tcps_sc_hash_size);
1333 	ASSIGN(tcps_sc_entry_count);
1334 	ASSIGN(tcps_sc_entry_limit);
1335 	ASSIGN(tcps_sc_bucket_maxlen);
1336 	ASSIGN(tcps_sc_bucket_limit);
1337 	ASSIGN(tcps_sc_uses_left);
1338 	ASSIGN(tcps_conndrained);
1339 	ASSIGN(tcps_sack_recovery_episode);
1340 	ASSIGN(tcps_sack_rexmits);
1341 	ASSIGN(tcps_sack_rexmit_bytes);
1342 	ASSIGN(tcps_sack_rcv_opts);
1343 	ASSIGN(tcps_sack_snd_opts);
1344 	ASSIGN(tcps_sack_drop_opts);
1345 
1346 #undef ASSIGN
1347 
1348 	set = &tcp_syn_cache[tcp_syn_cache_active];
1349 	tcpstat.tcps_sc_hash_size = set->scs_size;
1350 	tcpstat.tcps_sc_entry_count = set->scs_count;
1351 	tcpstat.tcps_sc_entry_limit = tcp_syn_cache_limit;
1352 	tcpstat.tcps_sc_bucket_maxlen = 0;
1353 	for (i = 0; i < set->scs_size; i++) {
1354 		if (tcpstat.tcps_sc_bucket_maxlen <
1355 		    set->scs_buckethead[i].sch_length)
1356 			tcpstat.tcps_sc_bucket_maxlen =
1357 				set->scs_buckethead[i].sch_length;
1358 	}
1359 	tcpstat.tcps_sc_bucket_limit = tcp_syn_bucket_limit;
1360 	tcpstat.tcps_sc_uses_left = set->scs_use;
1361 
1362 	return (sysctl_rdstruct(oldp, oldlenp, newp,
1363 	    &tcpstat, sizeof(tcpstat)));
1364 }
1365 
1366 /*
1367  * Sysctl for tcp variables.
1368  */
1369 int
1370 tcp_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
1371     size_t newlen)
1372 {
1373 	int error, nval;
1374 
1375 	/* All sysctl names at this level are terminal. */
1376 	if (namelen != 1)
1377 		return (ENOTDIR);
1378 
1379 	switch (name[0]) {
1380 	case TCPCTL_BADDYNAMIC:
1381 		NET_LOCK();
1382 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
1383 		    baddynamicports.tcp, sizeof(baddynamicports.tcp));
1384 		NET_UNLOCK();
1385 		return (error);
1386 
1387 	case TCPCTL_ROOTONLY:
1388 		if (newp && securelevel > 0)
1389 			return (EPERM);
1390 		NET_LOCK();
1391 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
1392 		    rootonlyports.tcp, sizeof(rootonlyports.tcp));
1393 		NET_UNLOCK();
1394 		return (error);
1395 
1396 	case TCPCTL_IDENT:
1397 		NET_LOCK();
1398 		error = tcp_ident(oldp, oldlenp, newp, newlen, 0);
1399 		NET_UNLOCK();
1400 		return (error);
1401 
1402 	case TCPCTL_DROP:
1403 		NET_LOCK();
1404 		error = tcp_ident(oldp, oldlenp, newp, newlen, 1);
1405 		NET_UNLOCK();
1406 		return (error);
1407 
1408 	case TCPCTL_REASS_LIMIT:
1409 		NET_LOCK();
1410 		nval = tcp_reass_limit;
1411 		error = sysctl_int(oldp, oldlenp, newp, newlen, &nval);
1412 		if (!error && nval != tcp_reass_limit) {
1413 			error = pool_sethardlimit(&tcpqe_pool, nval, NULL, 0);
1414 			if (!error)
1415 				tcp_reass_limit = nval;
1416 		}
1417 		NET_UNLOCK();
1418 		return (error);
1419 
1420 	case TCPCTL_SACKHOLE_LIMIT:
1421 		NET_LOCK();
1422 		nval = tcp_sackhole_limit;
1423 		error = sysctl_int(oldp, oldlenp, newp, newlen, &nval);
1424 		if (!error && nval != tcp_sackhole_limit) {
1425 			error = pool_sethardlimit(&sackhl_pool, nval, NULL, 0);
1426 			if (!error)
1427 				tcp_sackhole_limit = nval;
1428 		}
1429 		NET_UNLOCK();
1430 		return (error);
1431 
1432 	case TCPCTL_STATS:
1433 		return (tcp_sysctl_tcpstat(oldp, oldlenp, newp));
1434 
1435 	case TCPCTL_SYN_USE_LIMIT:
1436 		NET_LOCK();
1437 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1438 		    &tcp_syn_use_limit, 0, INT_MAX);
1439 		if (!error && newp != NULL) {
1440 			/*
1441 			 * Global tcp_syn_use_limit is used when reseeding a
1442 			 * new cache.  Also update the value in active cache.
1443 			 */
1444 			if (tcp_syn_cache[0].scs_use > tcp_syn_use_limit)
1445 				tcp_syn_cache[0].scs_use = tcp_syn_use_limit;
1446 			if (tcp_syn_cache[1].scs_use > tcp_syn_use_limit)
1447 				tcp_syn_cache[1].scs_use = tcp_syn_use_limit;
1448 		}
1449 		NET_UNLOCK();
1450 		return (error);
1451 
1452 	case TCPCTL_SYN_HASH_SIZE:
1453 		NET_LOCK();
1454 		nval = tcp_syn_hash_size;
1455 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1456 		    &nval, 1, 100000);
1457 		if (!error && nval != tcp_syn_hash_size) {
1458 			/*
1459 			 * If global hash size has been changed,
1460 			 * switch sets as soon as possible.  Then
1461 			 * the actual hash array will be reallocated.
1462 			 */
1463 			if (tcp_syn_cache[0].scs_size != nval)
1464 				tcp_syn_cache[0].scs_use = 0;
1465 			if (tcp_syn_cache[1].scs_size != nval)
1466 				tcp_syn_cache[1].scs_use = 0;
1467 			tcp_syn_hash_size = nval;
1468 		}
1469 		NET_UNLOCK();
1470 		return (error);
1471 
1472 	default:
1473 		NET_LOCK();
1474 		error = sysctl_bounded_arr(tcpctl_vars, nitems(tcpctl_vars), name,
1475 		     namelen, oldp, oldlenp, newp, newlen);
1476 		NET_UNLOCK();
1477 		return (error);
1478 	}
1479 	/* NOTREACHED */
1480 }
1481 
1482 /*
1483  * Scale the send buffer so that inflight data is not accounted against
1484  * the limit. The buffer will scale with the congestion window, if the
1485  * the receiver stops acking data the window will shrink and therefore
1486  * the buffer size will shrink as well.
1487  * In low memory situation try to shrink the buffer to the initial size
1488  * disabling the send buffer scaling as long as the situation persists.
1489  */
1490 void
1491 tcp_update_sndspace(struct tcpcb *tp)
1492 {
1493 	struct socket *so = tp->t_inpcb->inp_socket;
1494 	u_long nmax = so->so_snd.sb_hiwat;
1495 
1496 	if (sbchecklowmem()) {
1497 		/* low on memory try to get rid of some */
1498 		if (tcp_sendspace < nmax)
1499 			nmax = tcp_sendspace;
1500 	} else if (so->so_snd.sb_wat != tcp_sendspace)
1501 		/* user requested buffer size, auto-scaling disabled */
1502 		nmax = so->so_snd.sb_wat;
1503 	else
1504 		/* automatic buffer scaling */
1505 		nmax = MIN(sb_max, so->so_snd.sb_wat + tp->snd_max -
1506 		    tp->snd_una);
1507 
1508 	/* a writable socket must be preserved because of poll(2) semantics */
1509 	if (sbspace(so, &so->so_snd) >= so->so_snd.sb_lowat) {
1510 		if (nmax < so->so_snd.sb_cc + so->so_snd.sb_lowat)
1511 			nmax = so->so_snd.sb_cc + so->so_snd.sb_lowat;
1512 		/* keep in sync with sbreserve() calculation */
1513 		if (nmax * 8 < so->so_snd.sb_mbcnt + so->so_snd.sb_lowat)
1514 			nmax = (so->so_snd.sb_mbcnt+so->so_snd.sb_lowat+7) / 8;
1515 	}
1516 
1517 	/* round to MSS boundary */
1518 	nmax = roundup(nmax, tp->t_maxseg);
1519 
1520 	if (nmax != so->so_snd.sb_hiwat)
1521 		sbreserve(so, &so->so_snd, nmax);
1522 }
1523 
1524 /*
1525  * Scale the recv buffer by looking at how much data was transferred in
1526  * on approximated RTT. If more than a big part of the recv buffer was
1527  * transferred during that time we increase the buffer by a constant.
1528  * In low memory situation try to shrink the buffer to the initial size.
1529  */
1530 void
1531 tcp_update_rcvspace(struct tcpcb *tp)
1532 {
1533 	struct socket *so = tp->t_inpcb->inp_socket;
1534 	u_long nmax = so->so_rcv.sb_hiwat;
1535 
1536 	if (sbchecklowmem()) {
1537 		/* low on memory try to get rid of some */
1538 		if (tcp_recvspace < nmax)
1539 			nmax = tcp_recvspace;
1540 	} else if (so->so_rcv.sb_wat != tcp_recvspace)
1541 		/* user requested buffer size, auto-scaling disabled */
1542 		nmax = so->so_rcv.sb_wat;
1543 	else {
1544 		/* automatic buffer scaling */
1545 		if (tp->rfbuf_cnt > so->so_rcv.sb_hiwat / 8 * 7)
1546 			nmax = MIN(sb_max, so->so_rcv.sb_hiwat +
1547 			    tcp_autorcvbuf_inc);
1548 	}
1549 
1550 	/* a readable socket must be preserved because of poll(2) semantics */
1551 	if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat &&
1552 	    nmax < so->so_snd.sb_lowat)
1553 		nmax = so->so_snd.sb_lowat;
1554 
1555 	if (nmax == so->so_rcv.sb_hiwat)
1556 		return;
1557 
1558 	/* round to MSS boundary */
1559 	nmax = roundup(nmax, tp->t_maxseg);
1560 	sbreserve(so, &so->so_rcv, nmax);
1561 }
1562