Home | History | Annotate | Line # | Download | only in netinet
tcp_usrreq.c revision 1.171.2.1
      1 /*	$NetBSD: tcp_usrreq.c,v 1.171.2.1 2014/08/10 06:56:25 tls 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, 2005, 2006 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  * This code is derived from software contributed to The NetBSD Foundation
     40  * by Charles M. Hannum.
     41  * This code is derived from software contributed to The NetBSD Foundation
     42  * by Rui Paulo.
     43  *
     44  * Redistribution and use in source and binary forms, with or without
     45  * modification, are permitted provided that the following conditions
     46  * are met:
     47  * 1. Redistributions of source code must retain the above copyright
     48  *    notice, this list of conditions and the following disclaimer.
     49  * 2. Redistributions in binary form must reproduce the above copyright
     50  *    notice, this list of conditions and the following disclaimer in the
     51  *    documentation and/or other materials provided with the distribution.
     52  *
     53  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     54  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     56  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     57  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     58  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     59  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     60  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     61  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     62  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     63  * POSSIBILITY OF SUCH DAMAGE.
     64  */
     65 
     66 /*
     67  * Copyright (c) 1982, 1986, 1988, 1993, 1995
     68  *	The Regents of the University of California.  All rights reserved.
     69  *
     70  * Redistribution and use in source and binary forms, with or without
     71  * modification, are permitted provided that the following conditions
     72  * are met:
     73  * 1. Redistributions of source code must retain the above copyright
     74  *    notice, this list of conditions and the following disclaimer.
     75  * 2. Redistributions in binary form must reproduce the above copyright
     76  *    notice, this list of conditions and the following disclaimer in the
     77  *    documentation and/or other materials provided with the distribution.
     78  * 3. Neither the name of the University nor the names of its contributors
     79  *    may be used to endorse or promote products derived from this software
     80  *    without specific prior written permission.
     81  *
     82  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     83  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     84  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     85  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     86  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     87  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     88  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     89  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     90  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     91  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     92  * SUCH DAMAGE.
     93  *
     94  *	@(#)tcp_usrreq.c	8.5 (Berkeley) 6/21/95
     95  */
     96 
     97 /*
     98  * TCP protocol interface to socket abstraction.
     99  */
    100 
    101 #include <sys/cdefs.h>
    102 __KERNEL_RCSID(0, "$NetBSD: tcp_usrreq.c,v 1.171.2.1 2014/08/10 06:56:25 tls Exp $");
    103 
    104 #include "opt_inet.h"
    105 #include "opt_ipsec.h"
    106 #include "opt_tcp_debug.h"
    107 #include "opt_mbuftrace.h"
    108 
    109 #include <sys/param.h>
    110 #include <sys/systm.h>
    111 #include <sys/kernel.h>
    112 #include <sys/malloc.h>
    113 #include <sys/mbuf.h>
    114 #include <sys/socket.h>
    115 #include <sys/socketvar.h>
    116 #include <sys/protosw.h>
    117 #include <sys/errno.h>
    118 #include <sys/stat.h>
    119 #include <sys/proc.h>
    120 #include <sys/domain.h>
    121 #include <sys/sysctl.h>
    122 #include <sys/kauth.h>
    123 #include <sys/uidinfo.h>
    124 
    125 #include <net/if.h>
    126 #include <net/route.h>
    127 
    128 #include <netinet/in.h>
    129 #include <netinet/in_systm.h>
    130 #include <netinet/in_var.h>
    131 #include <netinet/ip.h>
    132 #include <netinet/in_pcb.h>
    133 #include <netinet/ip_var.h>
    134 #include <netinet/in_offload.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/scope6_var.h>
    144 #endif
    145 
    146 #include <netinet/tcp.h>
    147 #include <netinet/tcp_fsm.h>
    148 #include <netinet/tcp_seq.h>
    149 #include <netinet/tcp_timer.h>
    150 #include <netinet/tcp_var.h>
    151 #include <netinet/tcp_private.h>
    152 #include <netinet/tcp_congctl.h>
    153 #include <netinet/tcpip.h>
    154 #include <netinet/tcp_debug.h>
    155 #include <netinet/tcp_vtw.h>
    156 
    157 #include "opt_tcp_space.h"
    158 
    159 static int
    160 tcp_debug_capture(struct tcpcb *tp, int req)
    161 {
    162 #ifdef KPROF
    163 	tcp_acounts[tp->t_state][req]++;
    164 #endif
    165 #ifdef TCP_DEBUG
    166 	return tp->t_state;
    167 #endif
    168 	return 0;
    169 }
    170 
    171 static inline void
    172 tcp_debug_trace(struct socket *so, struct tcpcb *tp, int ostate, int req)
    173 {
    174 #ifdef TCP_DEBUG
    175 	if (tp && (so->so_options & SO_DEBUG))
    176 		tcp_trace(TA_USER, ostate, tp, NULL, req);
    177 #endif
    178 }
    179 
    180 static int
    181 tcp_getpcb(struct socket *so, struct inpcb **inp,
    182     struct in6pcb **in6p, struct tcpcb **tp)
    183 {
    184 
    185 	KASSERT(solocked(so));
    186 
    187 	/*
    188 	 * When a TCP is attached to a socket, then there will be
    189 	 * a (struct inpcb) pointed at by the socket, and this
    190 	 * structure will point at a subsidary (struct tcpcb).
    191 	 */
    192 	switch (so->so_proto->pr_domain->dom_family) {
    193 #ifdef INET
    194 	case PF_INET:
    195 		*inp = sotoinpcb(so);
    196 		if (*inp == NULL)
    197 			return EINVAL;
    198 		*tp = intotcpcb(*inp);
    199 		break;
    200 #endif
    201 #ifdef INET6
    202 	case PF_INET6:
    203 		*in6p = sotoin6pcb(so);
    204 		if (*in6p == NULL)
    205 			return EINVAL;
    206 		*tp = in6totcpcb(*in6p);
    207 		break;
    208 #endif
    209 	default:
    210 		return EAFNOSUPPORT;
    211 	}
    212 
    213 	KASSERT(tp != NULL);
    214 
    215 	return 0;
    216 }
    217 
    218 /*
    219  * Process a TCP user request for TCP tb.  If this is a send request
    220  * then m is the mbuf chain of send data.  If this is a timer expiration
    221  * (called from the software clock routine), then timertype tells which timer.
    222  */
    223 static int
    224 tcp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
    225     struct mbuf *control, struct lwp *l)
    226 {
    227 	KASSERT(req != PRU_ATTACH);
    228 	KASSERT(req != PRU_DETACH);
    229 	KASSERT(req != PRU_ACCEPT);
    230 	KASSERT(req != PRU_BIND);
    231 	KASSERT(req != PRU_LISTEN);
    232 	KASSERT(req != PRU_CONNECT);
    233 	KASSERT(req != PRU_CONNECT2);
    234 	KASSERT(req != PRU_DISCONNECT);
    235 	KASSERT(req != PRU_SHUTDOWN);
    236 	KASSERT(req != PRU_ABORT);
    237 	KASSERT(req != PRU_CONTROL);
    238 	KASSERT(req != PRU_SENSE);
    239 	KASSERT(req != PRU_PEERADDR);
    240 	KASSERT(req != PRU_SOCKADDR);
    241 	KASSERT(req != PRU_RCVD);
    242 	KASSERT(req != PRU_RCVOOB);
    243 	KASSERT(req != PRU_SEND);
    244 	KASSERT(req != PRU_SENDOOB);
    245 	KASSERT(req != PRU_PURGEIF);
    246 
    247 	KASSERT(solocked(so));
    248 
    249 	panic("tcp_usrreq");
    250 
    251 	return 0;
    252 }
    253 
    254 static void
    255 change_keepalive(struct socket *so, struct tcpcb *tp)
    256 {
    257 	tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl;
    258 	TCP_TIMER_DISARM(tp, TCPT_KEEP);
    259 	TCP_TIMER_DISARM(tp, TCPT_2MSL);
    260 
    261 	if (tp->t_state == TCPS_SYN_RECEIVED ||
    262 	    tp->t_state == TCPS_SYN_SENT) {
    263 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
    264 	} else if (so->so_options & SO_KEEPALIVE &&
    265 	    tp->t_state <= TCPS_CLOSE_WAIT) {
    266 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl);
    267 	} else {
    268 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle);
    269 	}
    270 
    271 	if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
    272 		TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
    273 }
    274 
    275 int
    276 tcp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
    277 {
    278 	int error = 0, s;
    279 	struct inpcb *inp;
    280 #ifdef INET6
    281 	struct in6pcb *in6p;
    282 #endif
    283 	struct tcpcb *tp;
    284 	u_int ui;
    285 	int family;	/* family of the socket */
    286 	int level, optname, optval;
    287 
    288 	level = sopt->sopt_level;
    289 	optname = sopt->sopt_name;
    290 
    291 	family = so->so_proto->pr_domain->dom_family;
    292 
    293 	s = splsoftnet();
    294 	switch (family) {
    295 #ifdef INET
    296 	case PF_INET:
    297 		inp = sotoinpcb(so);
    298 #ifdef INET6
    299 		in6p = NULL;
    300 #endif
    301 		break;
    302 #endif
    303 #ifdef INET6
    304 	case PF_INET6:
    305 		inp = NULL;
    306 		in6p = sotoin6pcb(so);
    307 		break;
    308 #endif
    309 	default:
    310 		splx(s);
    311 		panic("%s: af %d", __func__, family);
    312 	}
    313 #ifndef INET6
    314 	if (inp == NULL)
    315 #else
    316 	if (inp == NULL && in6p == NULL)
    317 #endif
    318 	{
    319 		splx(s);
    320 		return (ECONNRESET);
    321 	}
    322 	if (level != IPPROTO_TCP) {
    323 		switch (family) {
    324 #ifdef INET
    325 		case PF_INET:
    326 			error = ip_ctloutput(op, so, sopt);
    327 			break;
    328 #endif
    329 #ifdef INET6
    330 		case PF_INET6:
    331 			error = ip6_ctloutput(op, so, sopt);
    332 			break;
    333 #endif
    334 		}
    335 		splx(s);
    336 		return (error);
    337 	}
    338 	if (inp)
    339 		tp = intotcpcb(inp);
    340 #ifdef INET6
    341 	else if (in6p)
    342 		tp = in6totcpcb(in6p);
    343 #endif
    344 	else
    345 		tp = NULL;
    346 
    347 	switch (op) {
    348 	case PRCO_SETOPT:
    349 		switch (optname) {
    350 #ifdef TCP_SIGNATURE
    351 		case TCP_MD5SIG:
    352 			error = sockopt_getint(sopt, &optval);
    353 			if (error)
    354 				break;
    355 			if (optval > 0)
    356 				tp->t_flags |= TF_SIGNATURE;
    357 			else
    358 				tp->t_flags &= ~TF_SIGNATURE;
    359 			break;
    360 #endif /* TCP_SIGNATURE */
    361 
    362 		case TCP_NODELAY:
    363 			error = sockopt_getint(sopt, &optval);
    364 			if (error)
    365 				break;
    366 			if (optval)
    367 				tp->t_flags |= TF_NODELAY;
    368 			else
    369 				tp->t_flags &= ~TF_NODELAY;
    370 			break;
    371 
    372 		case TCP_MAXSEG:
    373 			error = sockopt_getint(sopt, &optval);
    374 			if (error)
    375 				break;
    376 			if (optval > 0 && optval <= tp->t_peermss)
    377 				tp->t_peermss = optval; /* limit on send size */
    378 			else
    379 				error = EINVAL;
    380 			break;
    381 #ifdef notyet
    382 		case TCP_CONGCTL:
    383 			/* XXX string overflow XXX */
    384 			error = tcp_congctl_select(tp, sopt->sopt_data);
    385 			break;
    386 #endif
    387 
    388 		case TCP_KEEPIDLE:
    389 			error = sockopt_get(sopt, &ui, sizeof(ui));
    390 			if (error)
    391 				break;
    392 			if (ui > 0) {
    393 				tp->t_keepidle = ui;
    394 				change_keepalive(so, tp);
    395 			} else
    396 				error = EINVAL;
    397 			break;
    398 
    399 		case TCP_KEEPINTVL:
    400 			error = sockopt_get(sopt, &ui, sizeof(ui));
    401 			if (error)
    402 				break;
    403 			if (ui > 0) {
    404 				tp->t_keepintvl = ui;
    405 				change_keepalive(so, tp);
    406 			} else
    407 				error = EINVAL;
    408 			break;
    409 
    410 		case TCP_KEEPCNT:
    411 			error = sockopt_get(sopt, &ui, sizeof(ui));
    412 			if (error)
    413 				break;
    414 			if (ui > 0) {
    415 				tp->t_keepcnt = ui;
    416 				change_keepalive(so, tp);
    417 			} else
    418 				error = EINVAL;
    419 			break;
    420 
    421 		case TCP_KEEPINIT:
    422 			error = sockopt_get(sopt, &ui, sizeof(ui));
    423 			if (error)
    424 				break;
    425 			if (ui > 0) {
    426 				tp->t_keepinit = ui;
    427 				change_keepalive(so, tp);
    428 			} else
    429 				error = EINVAL;
    430 			break;
    431 
    432 		default:
    433 			error = ENOPROTOOPT;
    434 			break;
    435 		}
    436 		break;
    437 
    438 	case PRCO_GETOPT:
    439 		switch (optname) {
    440 #ifdef TCP_SIGNATURE
    441 		case TCP_MD5SIG:
    442 			optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
    443 			error = sockopt_set(sopt, &optval, sizeof(optval));
    444 			break;
    445 #endif
    446 		case TCP_NODELAY:
    447 			optval = tp->t_flags & TF_NODELAY;
    448 			error = sockopt_set(sopt, &optval, sizeof(optval));
    449 			break;
    450 		case TCP_MAXSEG:
    451 			optval = tp->t_peermss;
    452 			error = sockopt_set(sopt, &optval, sizeof(optval));
    453 			break;
    454 #ifdef notyet
    455 		case TCP_CONGCTL:
    456 			break;
    457 #endif
    458 		default:
    459 			error = ENOPROTOOPT;
    460 			break;
    461 		}
    462 		break;
    463 	}
    464 	splx(s);
    465 	return (error);
    466 }
    467 
    468 #ifndef TCP_SENDSPACE
    469 #define	TCP_SENDSPACE	1024*32
    470 #endif
    471 int	tcp_sendspace = TCP_SENDSPACE;
    472 #ifndef TCP_RECVSPACE
    473 #define	TCP_RECVSPACE	1024*32
    474 #endif
    475 int	tcp_recvspace = TCP_RECVSPACE;
    476 
    477 /*
    478  * tcp_attach: attach TCP protocol to socket, allocating internet protocol
    479  * control block, TCP control block, buffer space and entering LISTEN state
    480  * if to accept connections.
    481  */
    482 static int
    483 tcp_attach(struct socket *so, int proto)
    484 {
    485 	struct tcpcb *tp;
    486 	struct inpcb *inp;
    487 #ifdef INET6
    488 	struct in6pcb *in6p;
    489 #endif
    490 	int s, error, family;
    491 
    492 	/* Assign the lock (must happen even if we will error out). */
    493 	s = splsoftnet();
    494 	sosetlock(so);
    495 	KASSERT(solocked(so));
    496 
    497 	family = so->so_proto->pr_domain->dom_family;
    498 	switch (family) {
    499 #ifdef INET
    500 	case PF_INET:
    501 		inp = sotoinpcb(so);
    502 #ifdef INET6
    503 		in6p = NULL;
    504 #endif
    505 		break;
    506 #endif
    507 #ifdef INET6
    508 	case PF_INET6:
    509 		inp = NULL;
    510 		in6p = sotoin6pcb(so);
    511 		break;
    512 #endif
    513 	default:
    514 		error = EAFNOSUPPORT;
    515 		goto out;
    516 	}
    517 
    518 	KASSERT(inp == NULL);
    519 #ifdef INET6
    520 	KASSERT(in6p == NULL);
    521 #endif
    522 
    523 #ifdef MBUFTRACE
    524 	so->so_mowner = &tcp_sock_mowner;
    525 	so->so_rcv.sb_mowner = &tcp_sock_rx_mowner;
    526 	so->so_snd.sb_mowner = &tcp_sock_tx_mowner;
    527 #endif
    528 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
    529 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
    530 		if (error)
    531 			goto out;
    532 	}
    533 
    534 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
    535 	so->so_snd.sb_flags |= SB_AUTOSIZE;
    536 
    537 	switch (family) {
    538 #ifdef INET
    539 	case PF_INET:
    540 		error = in_pcballoc(so, &tcbtable);
    541 		if (error)
    542 			goto out;
    543 		inp = sotoinpcb(so);
    544 #ifdef INET6
    545 		in6p = NULL;
    546 #endif
    547 		break;
    548 #endif
    549 #ifdef INET6
    550 	case PF_INET6:
    551 		error = in6_pcballoc(so, &tcbtable);
    552 		if (error)
    553 			goto out;
    554 		inp = NULL;
    555 		in6p = sotoin6pcb(so);
    556 		break;
    557 #endif
    558 	default:
    559 		error = EAFNOSUPPORT;
    560 		goto out;
    561 	}
    562 	if (inp)
    563 		tp = tcp_newtcpcb(family, (void *)inp);
    564 #ifdef INET6
    565 	else if (in6p)
    566 		tp = tcp_newtcpcb(family, (void *)in6p);
    567 #endif
    568 	else
    569 		tp = NULL;
    570 
    571 	if (tp == NULL) {
    572 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
    573 
    574 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
    575 #ifdef INET
    576 		if (inp)
    577 			in_pcbdetach(inp);
    578 #endif
    579 #ifdef INET6
    580 		if (in6p)
    581 			in6_pcbdetach(in6p);
    582 #endif
    583 		so->so_state |= nofd;
    584 		error = ENOBUFS;
    585 		goto out;
    586 	}
    587 	tp->t_state = TCPS_CLOSED;
    588 	if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
    589 		so->so_linger = TCP_LINGERTIME;
    590 	}
    591 out:
    592 	KASSERT(solocked(so));
    593 	splx(s);
    594 	return error;
    595 }
    596 
    597 static void
    598 tcp_detach(struct socket *so)
    599 {
    600 	struct inpcb *inp = NULL;
    601 	struct in6pcb *in6p = NULL;
    602 	struct tcpcb *tp = NULL;
    603 	int s;
    604 
    605 	if (tcp_getpcb(so, &inp, &in6p, &tp) != 0)
    606 		return;
    607 
    608 	s = splsoftnet();
    609 	(void)tcp_disconnect1(tp);
    610 	splx(s);
    611 }
    612 
    613 static int
    614 tcp_accept(struct socket *so, struct mbuf *nam)
    615 {
    616 	struct inpcb *inp = NULL;
    617 	struct in6pcb *in6p = NULL;
    618 	struct tcpcb *tp = NULL;
    619 	int ostate = 0;
    620 	int error = 0;
    621 	int s;
    622 
    623 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    624 		return error;
    625 
    626 	ostate = tcp_debug_capture(tp, PRU_ACCEPT);
    627 
    628 	/*
    629 	 * Accept a connection.  Essentially all the work is
    630 	 * done at higher levels; just return the address
    631 	 * of the peer, storing through addr.
    632 	 */
    633 	s = splsoftnet();
    634 #ifdef INET
    635 	if (inp) {
    636 		in_setpeeraddr(inp, nam);
    637 	}
    638 #endif
    639 #ifdef INET6
    640 	if (in6p) {
    641 		in6_setpeeraddr(in6p, nam);
    642 	}
    643 #endif
    644 	tcp_debug_trace(so, tp, ostate, PRU_ACCEPT);
    645 	splx(s);
    646 
    647 	return 0;
    648 }
    649 
    650 static int
    651 tcp_bind(struct socket *so, struct mbuf *nam, struct lwp *l)
    652 {
    653 	struct inpcb *inp = NULL;
    654 	struct in6pcb *in6p = NULL;
    655 	struct tcpcb *tp = NULL;
    656 	int s;
    657 	int error = 0;
    658 	int ostate = 0;
    659 
    660 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    661 		return error;
    662 
    663 	ostate = tcp_debug_capture(tp, PRU_BIND);
    664 
    665 	/*
    666 	 * Give the socket an address.
    667 	 */
    668 	s = splsoftnet();
    669 	switch (so->so_proto->pr_domain->dom_family) {
    670 #ifdef INET
    671 	case PF_INET:
    672 		error = in_pcbbind(inp, nam, l);
    673 		break;
    674 #endif
    675 #ifdef INET6
    676 	case PF_INET6:
    677 		error = in6_pcbbind(in6p, nam, l);
    678 		if (!error) {
    679 			/* mapped addr case */
    680 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
    681 				tp->t_family = AF_INET;
    682 			else
    683 				tp->t_family = AF_INET6;
    684 		}
    685 		break;
    686 #endif
    687 	}
    688 	tcp_debug_trace(so, tp, ostate, PRU_BIND);
    689 	splx(s);
    690 
    691 	return error;
    692 }
    693 
    694 static int
    695 tcp_listen(struct socket *so, struct lwp *l)
    696 {
    697 	struct inpcb *inp = NULL;
    698 	struct in6pcb *in6p = NULL;
    699 	struct tcpcb *tp = NULL;
    700 	int error = 0;
    701 	int ostate = 0;
    702 	int s;
    703 
    704 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    705 		return error;
    706 
    707 	ostate = tcp_debug_capture(tp, PRU_LISTEN);
    708 
    709 	/*
    710 	 * Prepare to accept connections.
    711 	 */
    712 	s = splsoftnet();
    713 #ifdef INET
    714 	if (inp && inp->inp_lport == 0) {
    715 		error = in_pcbbind(inp, NULL, l);
    716 		if (error)
    717 			goto release;
    718 	}
    719 #endif
    720 #ifdef INET6
    721 	if (in6p && in6p->in6p_lport == 0) {
    722 		error = in6_pcbbind(in6p, NULL, l);
    723 		if (error)
    724 			goto release;
    725 	}
    726 #endif
    727 	tp->t_state = TCPS_LISTEN;
    728 
    729 release:
    730 	tcp_debug_trace(so, tp, ostate, PRU_LISTEN);
    731 	splx(s);
    732 
    733 	return error;
    734 }
    735 
    736 static int
    737 tcp_connect(struct socket *so, struct mbuf *nam, struct lwp *l)
    738 {
    739 	struct inpcb *inp = NULL;
    740 	struct in6pcb *in6p = NULL;
    741 	struct tcpcb *tp = NULL;
    742 	int s;
    743 	int error = 0;
    744 	int ostate = 0;
    745 
    746 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    747 		return error;
    748 
    749 	ostate = tcp_debug_capture(tp, PRU_CONNECT);
    750 
    751 	/*
    752 	 * Initiate connection to peer.
    753 	 * Create a template for use in transmissions on this connection.
    754 	 * Enter SYN_SENT state, and mark socket as connecting.
    755 	 * Start keep-alive timer, and seed output sequence space.
    756 	 * Send initial segment on connection.
    757 	 */
    758 	s = splsoftnet();
    759 #ifdef INET
    760 	if (inp) {
    761 		if (inp->inp_lport == 0) {
    762 			error = in_pcbbind(inp, NULL, l);
    763 			if (error)
    764 				goto release;
    765 		}
    766 		error = in_pcbconnect(inp, nam, l);
    767 	}
    768 #endif
    769 #ifdef INET6
    770 	if (in6p) {
    771 		if (in6p->in6p_lport == 0) {
    772 			error = in6_pcbbind(in6p, NULL, l);
    773 			if (error)
    774 				goto release;
    775 		}
    776 		error = in6_pcbconnect(in6p, nam, l);
    777 		if (!error) {
    778 			/* mapped addr case */
    779 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
    780 				tp->t_family = AF_INET;
    781 			else
    782 				tp->t_family = AF_INET6;
    783 		}
    784 	}
    785 #endif
    786 	if (error)
    787 		goto release;
    788 	tp->t_template = tcp_template(tp);
    789 	if (tp->t_template == 0) {
    790 #ifdef INET
    791 		if (inp)
    792 			in_pcbdisconnect(inp);
    793 #endif
    794 #ifdef INET6
    795 		if (in6p)
    796 			in6_pcbdisconnect(in6p);
    797 #endif
    798 		error = ENOBUFS;
    799 		goto release;
    800 	}
    801 	/*
    802 	 * Compute window scaling to request.
    803 	 * XXX: This should be moved to tcp_output().
    804 	 */
    805 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
    806 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
    807 		tp->request_r_scale++;
    808 	soisconnecting(so);
    809 	TCP_STATINC(TCP_STAT_CONNATTEMPT);
    810 	tp->t_state = TCPS_SYN_SENT;
    811 	TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
    812 	tp->iss = tcp_new_iss(tp, 0);
    813 	tcp_sendseqinit(tp);
    814 	error = tcp_output(tp);
    815 
    816 release:
    817 	tcp_debug_trace(so, tp, ostate, PRU_CONNECT);
    818 	splx(s);
    819 
    820 	return error;
    821 }
    822 
    823 static int
    824 tcp_connect2(struct socket *so, struct socket *so2)
    825 {
    826 	struct inpcb *inp = NULL;
    827 	struct in6pcb *in6p = NULL;
    828 	struct tcpcb *tp = NULL;
    829 	int error = 0;
    830 	int ostate = 0;
    831 
    832 	KASSERT(solocked(so));
    833 
    834 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    835 		return error;
    836 
    837 	ostate = tcp_debug_capture(tp, PRU_CONNECT2);
    838 
    839 	tcp_debug_trace(so, tp, ostate, PRU_CONNECT2);
    840 
    841 	return EOPNOTSUPP;
    842 }
    843 
    844 static int
    845 tcp_disconnect(struct socket *so)
    846 {
    847 	struct inpcb *inp = NULL;
    848 	struct in6pcb *in6p = NULL;
    849 	struct tcpcb *tp = NULL;
    850 	int error = 0;
    851 	int ostate = 0;
    852 	int s;
    853 
    854 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    855 		return error;
    856 
    857 	ostate = tcp_debug_capture(tp, PRU_DISCONNECT);
    858 
    859 	/*
    860 	 * Initiate disconnect from peer.
    861 	 * If connection never passed embryonic stage, just drop;
    862 	 * else if don't need to let data drain, then can just drop anyways,
    863 	 * else have to begin TCP shutdown process: mark socket disconnecting,
    864 	 * drain unread data, state switch to reflect user close, and
    865 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
    866 	 * when peer sends FIN and acks ours.
    867 	 *
    868 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
    869 	 */
    870 	s = splsoftnet();
    871 	tp = tcp_disconnect1(tp);
    872 	tcp_debug_trace(so, tp, ostate, PRU_DISCONNECT);
    873 	splx(s);
    874 
    875 	return error;
    876 }
    877 
    878 static int
    879 tcp_shutdown(struct socket *so)
    880 {
    881 	struct inpcb *inp = NULL;
    882 	struct in6pcb *in6p = NULL;
    883 	struct tcpcb *tp = NULL;
    884 	int error = 0;
    885 	int ostate = 0;
    886 	int s;
    887 
    888 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    889 		return error;
    890 
    891 	ostate = tcp_debug_capture(tp, PRU_SHUTDOWN);
    892 	/*
    893 	 * Mark the connection as being incapable of further output.
    894 	 */
    895 	s = splsoftnet();
    896 	socantsendmore(so);
    897 	tp = tcp_usrclosed(tp);
    898 	if (tp)
    899 		error = tcp_output(tp);
    900 	tcp_debug_trace(so, tp, ostate, PRU_SHUTDOWN);
    901 	splx(s);
    902 
    903 	return error;
    904 }
    905 
    906 static int
    907 tcp_abort(struct socket *so)
    908 {
    909 	struct inpcb *inp = NULL;
    910 	struct in6pcb *in6p = NULL;
    911 	struct tcpcb *tp = NULL;
    912 	int error = 0;
    913 	int ostate = 0;
    914 	int s;
    915 
    916 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    917 		return error;
    918 
    919 	ostate = tcp_debug_capture(tp, PRU_ABORT);
    920 
    921 	/*
    922 	 * Abort the TCP.
    923 	 */
    924 	s = splsoftnet();
    925 	tp = tcp_drop(tp, ECONNABORTED);
    926 	tcp_debug_trace(so, tp, ostate, PRU_ABORT);
    927 	splx(s);
    928 
    929 	return error;
    930 }
    931 
    932 static int
    933 tcp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
    934 {
    935 	switch (so->so_proto->pr_domain->dom_family) {
    936 #ifdef INET
    937 	case PF_INET:
    938 		return in_control(so, cmd, nam, ifp);
    939 #endif
    940 #ifdef INET6
    941 	case PF_INET6:
    942 		return in6_control(so, cmd, nam, ifp);
    943 #endif
    944 	default:
    945 		return EAFNOSUPPORT;
    946 	}
    947 }
    948 
    949 static int
    950 tcp_stat(struct socket *so, struct stat *ub)
    951 {
    952 	KASSERT(solocked(so));
    953 
    954 	/* stat: don't bother with a blocksize.  */
    955 	return 0;
    956 }
    957 
    958 static int
    959 tcp_peeraddr(struct socket *so, struct mbuf *nam)
    960 {
    961 	struct inpcb *inp = NULL;
    962 	struct in6pcb *in6p = NULL;
    963 	struct tcpcb *tp = NULL;
    964 	int ostate = 0;
    965 	int error = 0;
    966 	int s;
    967 
    968 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    969 		return error;
    970 
    971 	ostate = tcp_debug_capture(tp, PRU_PEERADDR);
    972 
    973 	s = splsoftnet();
    974 #ifdef INET
    975 	if (inp)
    976 		in_setpeeraddr(inp, nam);
    977 #endif
    978 #ifdef INET6
    979 	if (in6p)
    980 		in6_setpeeraddr(in6p, nam);
    981 #endif
    982 	tcp_debug_trace(so, tp, ostate, PRU_PEERADDR);
    983 	splx(s);
    984 
    985 	return 0;
    986 }
    987 
    988 static int
    989 tcp_sockaddr(struct socket *so, struct mbuf *nam)
    990 {
    991 	struct inpcb *inp = NULL;
    992 	struct in6pcb *in6p = NULL;
    993 	struct tcpcb *tp = NULL;
    994 	int ostate = 0;
    995 	int error = 0;
    996 	int s;
    997 
    998 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
    999 		return error;
   1000 
   1001 	ostate = tcp_debug_capture(tp, PRU_SOCKADDR);
   1002 
   1003 	s = splsoftnet();
   1004 #ifdef INET
   1005 	if (inp)
   1006 		in_setsockaddr(inp, nam);
   1007 #endif
   1008 #ifdef INET6
   1009 	if (in6p)
   1010 		in6_setsockaddr(in6p, nam);
   1011 #endif
   1012 	tcp_debug_trace(so, tp, ostate, PRU_SOCKADDR);
   1013 	splx(s);
   1014 
   1015 	return 0;
   1016 }
   1017 
   1018 static int
   1019 tcp_rcvd(struct socket *so, int flags, struct lwp *l)
   1020 {
   1021 	struct inpcb *inp = NULL;
   1022 	struct in6pcb *in6p = NULL;
   1023 	struct tcpcb *tp = NULL;
   1024 	int ostate = 0;
   1025 	int error = 0;
   1026 	int s;
   1027 
   1028 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
   1029 		return error;
   1030 
   1031 	ostate = tcp_debug_capture(tp, PRU_RCVD);
   1032 
   1033 	/*
   1034 	 * After a receive, possibly send window update to peer.
   1035 	 *
   1036 	 * soreceive() calls this function when a user receives
   1037 	 * ancillary data on a listening socket. We don't call
   1038 	 * tcp_output in such a case, since there is no header
   1039 	 * template for a listening socket and hence the kernel
   1040 	 * will panic.
   1041 	 */
   1042 	s = splsoftnet();
   1043 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
   1044 		(void) tcp_output(tp);
   1045 	splx(s);
   1046 
   1047 	tcp_debug_trace(so, tp, ostate, PRU_RCVD);
   1048 
   1049 	return 0;
   1050 }
   1051 
   1052 static int
   1053 tcp_recvoob(struct socket *so, struct mbuf *m, int flags)
   1054 {
   1055 	struct inpcb *inp = NULL;
   1056 	struct in6pcb *in6p = NULL;
   1057 	struct tcpcb *tp = NULL;
   1058 	int ostate = 0;
   1059 	int error = 0;
   1060 	int s;
   1061 
   1062 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
   1063 		return error;
   1064 
   1065 	ostate = tcp_debug_capture(tp, PRU_RCVOOB);
   1066 
   1067 	s = splsoftnet();
   1068 	if ((so->so_oobmark == 0 &&
   1069 	    (so->so_state & SS_RCVATMARK) == 0) ||
   1070 	    so->so_options & SO_OOBINLINE ||
   1071 	    tp->t_oobflags & TCPOOB_HADDATA) {
   1072 		splx(s);
   1073 		return EINVAL;
   1074 	}
   1075 
   1076 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
   1077 		splx(s);
   1078 		return EWOULDBLOCK;
   1079 	}
   1080 
   1081 	m->m_len = 1;
   1082 	*mtod(m, char *) = tp->t_iobc;
   1083 	if ((flags & MSG_PEEK) == 0)
   1084 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
   1085 
   1086 	tcp_debug_trace(so, tp, ostate, PRU_RCVOOB);
   1087 	splx(s);
   1088 
   1089 	return 0;
   1090 }
   1091 
   1092 static int
   1093 tcp_send(struct socket *so, struct mbuf *m, struct mbuf *nam,
   1094     struct mbuf *control, struct lwp *l)
   1095 {
   1096 	struct inpcb *inp = NULL;
   1097 	struct in6pcb *in6p = NULL;
   1098 	struct tcpcb *tp = NULL;
   1099 	int ostate = 0;
   1100 	int error = 0;
   1101 	int s;
   1102 
   1103 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
   1104 		return error;
   1105 
   1106 	ostate = tcp_debug_capture(tp, PRU_SEND);
   1107 
   1108 	/*
   1109 	 * Do a send by putting data in output queue and updating urgent
   1110 	 * marker if URG set.  Possibly send more data.
   1111 	 */
   1112 	s = splsoftnet();
   1113 	if (control && control->m_len) {
   1114 		m_freem(control);
   1115 		m_freem(m);
   1116 		tcp_debug_trace(so, tp, ostate, PRU_SEND);
   1117 		splx(s);
   1118 		return EINVAL;
   1119 	}
   1120 
   1121 	sbappendstream(&so->so_snd, m);
   1122 	error = tcp_output(tp);
   1123 	tcp_debug_trace(so, tp, ostate, PRU_SEND);
   1124 	splx(s);
   1125 
   1126 	return error;
   1127 }
   1128 
   1129 static int
   1130 tcp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
   1131 {
   1132 	struct inpcb *inp = NULL;
   1133 	struct in6pcb *in6p = NULL;
   1134 	struct tcpcb *tp = NULL;
   1135 	int ostate = 0;
   1136 	int error = 0;
   1137 	int s;
   1138 
   1139 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
   1140 		return error;
   1141 
   1142 	ostate = tcp_debug_capture(tp, PRU_SENDOOB);
   1143 
   1144 	s = splsoftnet();
   1145 	if (sbspace(&so->so_snd) < -512) {
   1146 		m_freem(m);
   1147 		splx(s);
   1148 		return ENOBUFS;
   1149 	}
   1150 	/*
   1151 	 * According to RFC961 (Assigned Protocols),
   1152 	 * the urgent pointer points to the last octet
   1153 	 * of urgent data.  We continue, however,
   1154 	 * to consider it to indicate the first octet
   1155 	 * of data past the urgent section.
   1156 	 * Otherwise, snd_up should be one lower.
   1157 	 */
   1158 	sbappendstream(&so->so_snd, m);
   1159 	tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
   1160 	tp->t_force = 1;
   1161 	error = tcp_output(tp);
   1162 	tp->t_force = 0;
   1163 	tcp_debug_trace(so, tp, ostate, PRU_SENDOOB);
   1164 	splx(s);
   1165 
   1166 	return error;
   1167 }
   1168 
   1169 static int
   1170 tcp_purgeif(struct socket *so, struct ifnet *ifp)
   1171 {
   1172 	int s;
   1173 
   1174 	s = splsoftnet();
   1175 	mutex_enter(softnet_lock);
   1176 	switch (so->so_proto->pr_domain->dom_family) {
   1177 #ifdef INET
   1178 	case PF_INET:
   1179 		in_pcbpurgeif0(&tcbtable, ifp);
   1180 		in_purgeif(ifp);
   1181 		in_pcbpurgeif(&tcbtable, ifp);
   1182 		break;
   1183 #endif
   1184 #ifdef INET6
   1185 	case PF_INET6:
   1186 		in6_pcbpurgeif0(&tcbtable, ifp);
   1187 		in6_purgeif(ifp);
   1188 		in6_pcbpurgeif(&tcbtable, ifp);
   1189 		break;
   1190 #endif
   1191 	default:
   1192 		mutex_exit(softnet_lock);
   1193 		splx(s);
   1194 		return EAFNOSUPPORT;
   1195 	}
   1196 	mutex_exit(softnet_lock);
   1197 	splx(s);
   1198 
   1199 	return 0;
   1200 }
   1201 
   1202 /*
   1203  * Initiate (or continue) disconnect.
   1204  * If embryonic state, just send reset (once).
   1205  * If in ``let data drain'' option and linger null, just drop.
   1206  * Otherwise (hard), mark socket disconnecting and drop
   1207  * current input data; switch states based on user close, and
   1208  * send segment to peer (with FIN).
   1209  */
   1210 struct tcpcb *
   1211 tcp_disconnect1(struct tcpcb *tp)
   1212 {
   1213 	struct socket *so;
   1214 
   1215 	if (tp->t_inpcb)
   1216 		so = tp->t_inpcb->inp_socket;
   1217 #ifdef INET6
   1218 	else if (tp->t_in6pcb)
   1219 		so = tp->t_in6pcb->in6p_socket;
   1220 #endif
   1221 	else
   1222 		so = NULL;
   1223 
   1224 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
   1225 		tp = tcp_close(tp);
   1226 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
   1227 		tp = tcp_drop(tp, 0);
   1228 	else {
   1229 		soisdisconnecting(so);
   1230 		sbflush(&so->so_rcv);
   1231 		tp = tcp_usrclosed(tp);
   1232 		if (tp)
   1233 			(void) tcp_output(tp);
   1234 	}
   1235 	return (tp);
   1236 }
   1237 
   1238 /*
   1239  * User issued close, and wish to trail through shutdown states:
   1240  * if never received SYN, just forget it.  If got a SYN from peer,
   1241  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
   1242  * If already got a FIN from peer, then almost done; go to LAST_ACK
   1243  * state.  In all other cases, have already sent FIN to peer (e.g.
   1244  * after PRU_SHUTDOWN), and just have to play tedious game waiting
   1245  * for peer to send FIN or not respond to keep-alives, etc.
   1246  * We can let the user exit from the close as soon as the FIN is acked.
   1247  */
   1248 struct tcpcb *
   1249 tcp_usrclosed(struct tcpcb *tp)
   1250 {
   1251 
   1252 	switch (tp->t_state) {
   1253 
   1254 	case TCPS_CLOSED:
   1255 	case TCPS_LISTEN:
   1256 	case TCPS_SYN_SENT:
   1257 		tp->t_state = TCPS_CLOSED;
   1258 		tp = tcp_close(tp);
   1259 		break;
   1260 
   1261 	case TCPS_SYN_RECEIVED:
   1262 	case TCPS_ESTABLISHED:
   1263 		tp->t_state = TCPS_FIN_WAIT_1;
   1264 		break;
   1265 
   1266 	case TCPS_CLOSE_WAIT:
   1267 		tp->t_state = TCPS_LAST_ACK;
   1268 		break;
   1269 	}
   1270 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
   1271 		struct socket *so;
   1272 		if (tp->t_inpcb)
   1273 			so = tp->t_inpcb->inp_socket;
   1274 #ifdef INET6
   1275 		else if (tp->t_in6pcb)
   1276 			so = tp->t_in6pcb->in6p_socket;
   1277 #endif
   1278 		else
   1279 			so = NULL;
   1280 		if (so)
   1281 			soisdisconnected(so);
   1282 		/*
   1283 		 * If we are in FIN_WAIT_2, we arrived here because the
   1284 		 * application did a shutdown of the send side.  Like the
   1285 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
   1286 		 * a full close, we start a timer to make sure sockets are
   1287 		 * not left in FIN_WAIT_2 forever.
   1288 		 */
   1289 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
   1290 			TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
   1291 		else if (tp->t_state == TCPS_TIME_WAIT
   1292 			 && ((tp->t_inpcb
   1293 			      && (tcp4_vtw_enable & 1)
   1294 			      && vtw_add(AF_INET, tp))
   1295 			     ||
   1296 			     (tp->t_in6pcb
   1297 			      && (tcp6_vtw_enable & 1)
   1298 			      && vtw_add(AF_INET6, tp)))) {
   1299 			tp = 0;
   1300 		}
   1301 	}
   1302 	return (tp);
   1303 }
   1304 
   1305 /*
   1306  * sysctl helper routine for net.inet.ip.mssdflt.  it can't be less
   1307  * than 32.
   1308  */
   1309 static int
   1310 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS)
   1311 {
   1312 	int error, mssdflt;
   1313 	struct sysctlnode node;
   1314 
   1315 	mssdflt = tcp_mssdflt;
   1316 	node = *rnode;
   1317 	node.sysctl_data = &mssdflt;
   1318 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1319 	if (error || newp == NULL)
   1320 		return (error);
   1321 
   1322 	if (mssdflt < 32)
   1323 		return (EINVAL);
   1324 	tcp_mssdflt = mssdflt;
   1325 
   1326 	mutex_enter(softnet_lock);
   1327 	tcp_tcpcb_template();
   1328 	mutex_exit(softnet_lock);
   1329 
   1330 	return (0);
   1331 }
   1332 
   1333 /*
   1334  * sysctl helper for TCP CB template update
   1335  */
   1336 static int
   1337 sysctl_update_tcpcb_template(SYSCTLFN_ARGS)
   1338 {
   1339 	int t, error;
   1340 	struct sysctlnode node;
   1341 
   1342 	/* follow procedures in sysctl(9) manpage */
   1343 	t = *(int *)rnode->sysctl_data;
   1344 	node = *rnode;
   1345 	node.sysctl_data = &t;
   1346 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1347 	if (error || newp == NULL)
   1348 		return error;
   1349 
   1350 	if (t < 0)
   1351 		return EINVAL;
   1352 
   1353 	*(int *)rnode->sysctl_data = t;
   1354 
   1355 	mutex_enter(softnet_lock);
   1356 	tcp_tcpcb_template();
   1357 	mutex_exit(softnet_lock);
   1358 
   1359 	return 0;
   1360 }
   1361 
   1362 /*
   1363  * sysctl helper routine for setting port related values under
   1364  * net.inet.ip and net.inet6.ip6.  does basic range checking and does
   1365  * additional checks for each type.  this code has placed in
   1366  * tcp_input.c since INET and INET6 both use the same tcp code.
   1367  *
   1368  * this helper is not static so that both inet and inet6 can use it.
   1369  */
   1370 int
   1371 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS)
   1372 {
   1373 	int error, tmp;
   1374 	int apmin, apmax;
   1375 #ifndef IPNOPRIVPORTS
   1376 	int lpmin, lpmax;
   1377 #endif /* IPNOPRIVPORTS */
   1378 	struct sysctlnode node;
   1379 
   1380 	if (namelen != 0)
   1381 		return (EINVAL);
   1382 
   1383 	switch (name[-3]) {
   1384 #ifdef INET
   1385 	    case PF_INET:
   1386 		apmin = anonportmin;
   1387 		apmax = anonportmax;
   1388 #ifndef IPNOPRIVPORTS
   1389 		lpmin = lowportmin;
   1390 		lpmax = lowportmax;
   1391 #endif /* IPNOPRIVPORTS */
   1392 		break;
   1393 #endif /* INET */
   1394 #ifdef INET6
   1395 	    case PF_INET6:
   1396 		apmin = ip6_anonportmin;
   1397 		apmax = ip6_anonportmax;
   1398 #ifndef IPNOPRIVPORTS
   1399 		lpmin = ip6_lowportmin;
   1400 		lpmax = ip6_lowportmax;
   1401 #endif /* IPNOPRIVPORTS */
   1402 		break;
   1403 #endif /* INET6 */
   1404 	    default:
   1405 		return (EINVAL);
   1406 	}
   1407 
   1408 	/*
   1409 	 * insert temporary copy into node, perform lookup on
   1410 	 * temporary, then restore pointer
   1411 	 */
   1412 	node = *rnode;
   1413 	tmp = *(int*)rnode->sysctl_data;
   1414 	node.sysctl_data = &tmp;
   1415 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1416 	if (error || newp == NULL)
   1417 		return (error);
   1418 
   1419 	/*
   1420 	 * simple port range check
   1421 	 */
   1422 	if (tmp < 0 || tmp > 65535)
   1423 		return (EINVAL);
   1424 
   1425 	/*
   1426 	 * per-node range checks
   1427 	 */
   1428 	switch (rnode->sysctl_num) {
   1429 	case IPCTL_ANONPORTMIN:
   1430 	case IPV6CTL_ANONPORTMIN:
   1431 		if (tmp >= apmax)
   1432 			return (EINVAL);
   1433 #ifndef IPNOPRIVPORTS
   1434 		if (tmp < IPPORT_RESERVED)
   1435                         return (EINVAL);
   1436 #endif /* IPNOPRIVPORTS */
   1437 		break;
   1438 
   1439 	case IPCTL_ANONPORTMAX:
   1440 	case IPV6CTL_ANONPORTMAX:
   1441                 if (apmin >= tmp)
   1442 			return (EINVAL);
   1443 #ifndef IPNOPRIVPORTS
   1444 		if (tmp < IPPORT_RESERVED)
   1445                         return (EINVAL);
   1446 #endif /* IPNOPRIVPORTS */
   1447 		break;
   1448 
   1449 #ifndef IPNOPRIVPORTS
   1450 	case IPCTL_LOWPORTMIN:
   1451 	case IPV6CTL_LOWPORTMIN:
   1452 		if (tmp >= lpmax ||
   1453 		    tmp > IPPORT_RESERVEDMAX ||
   1454 		    tmp < IPPORT_RESERVEDMIN)
   1455 			return (EINVAL);
   1456 		break;
   1457 
   1458 	case IPCTL_LOWPORTMAX:
   1459 	case IPV6CTL_LOWPORTMAX:
   1460 		if (lpmin >= tmp ||
   1461 		    tmp > IPPORT_RESERVEDMAX ||
   1462 		    tmp < IPPORT_RESERVEDMIN)
   1463 			return (EINVAL);
   1464 		break;
   1465 #endif /* IPNOPRIVPORTS */
   1466 
   1467 	default:
   1468 		return (EINVAL);
   1469 	}
   1470 
   1471 	*(int*)rnode->sysctl_data = tmp;
   1472 
   1473 	return (0);
   1474 }
   1475 
   1476 static inline int
   1477 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp)
   1478 {
   1479 	if (oldp) {
   1480 		size_t sz;
   1481 		uid_t uid;
   1482 		int error;
   1483 
   1484 		if (sockp->so_cred == NULL)
   1485 			return EPERM;
   1486 
   1487 		uid = kauth_cred_geteuid(sockp->so_cred);
   1488 		sz = MIN(sizeof(uid), *oldlenp);
   1489 		if ((error = copyout(&uid, oldp, sz)) != 0)
   1490 			return error;
   1491 	}
   1492 	*oldlenp = sizeof(uid_t);
   1493 	return 0;
   1494 }
   1495 
   1496 static inline int
   1497 inet4_ident_core(struct in_addr raddr, u_int rport,
   1498     struct in_addr laddr, u_int lport,
   1499     void *oldp, size_t *oldlenp,
   1500     struct lwp *l, int dodrop)
   1501 {
   1502 	struct inpcb *inp;
   1503 	struct socket *sockp;
   1504 
   1505 	inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0);
   1506 
   1507 	if (inp == NULL || (sockp = inp->inp_socket) == NULL)
   1508 		return ESRCH;
   1509 
   1510 	if (dodrop) {
   1511 		struct tcpcb *tp;
   1512 		int error;
   1513 
   1514 		if (inp == NULL || (tp = intotcpcb(inp)) == NULL ||
   1515 		    (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0)
   1516 			return ESRCH;
   1517 
   1518 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
   1519 		    KAUTH_REQ_NETWORK_SOCKET_DROP, inp->inp_socket, tp, NULL);
   1520 		if (error)
   1521 			return (error);
   1522 
   1523 		(void)tcp_drop(tp, ECONNABORTED);
   1524 		return 0;
   1525 	}
   1526 	else
   1527 		return copyout_uid(sockp, oldp, oldlenp);
   1528 }
   1529 
   1530 #ifdef INET6
   1531 static inline int
   1532 inet6_ident_core(struct in6_addr *raddr, u_int rport,
   1533     struct in6_addr *laddr, u_int lport,
   1534     void *oldp, size_t *oldlenp,
   1535     struct lwp *l, int dodrop)
   1536 {
   1537 	struct in6pcb *in6p;
   1538 	struct socket *sockp;
   1539 
   1540 	in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0, 0);
   1541 
   1542 	if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL)
   1543 		return ESRCH;
   1544 
   1545 	if (dodrop) {
   1546 		struct tcpcb *tp;
   1547 		int error;
   1548 
   1549 		if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL ||
   1550 		    (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0)
   1551 			return ESRCH;
   1552 
   1553 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
   1554 		    KAUTH_REQ_NETWORK_SOCKET_DROP, in6p->in6p_socket, tp, NULL);
   1555 		if (error)
   1556 			return (error);
   1557 
   1558 		(void)tcp_drop(tp, ECONNABORTED);
   1559 		return 0;
   1560 	}
   1561 	else
   1562 		return copyout_uid(sockp, oldp, oldlenp);
   1563 }
   1564 #endif
   1565 
   1566 /*
   1567  * sysctl helper routine for the net.inet.tcp.drop and
   1568  * net.inet6.tcp6.drop nodes.
   1569  */
   1570 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident
   1571 
   1572 /*
   1573  * sysctl helper routine for the net.inet.tcp.ident and
   1574  * net.inet6.tcp6.ident nodes.  contains backwards compat code for the
   1575  * old way of looking up the ident information for ipv4 which involves
   1576  * stuffing the port/addr pairs into the mib lookup.
   1577  */
   1578 static int
   1579 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS)
   1580 {
   1581 #ifdef INET
   1582 	struct sockaddr_in *si4[2];
   1583 #endif /* INET */
   1584 #ifdef INET6
   1585 	struct sockaddr_in6 *si6[2];
   1586 #endif /* INET6 */
   1587 	struct sockaddr_storage sa[2];
   1588 	int error, pf, dodrop;
   1589 
   1590 	dodrop = name[-1] == TCPCTL_DROP;
   1591 	if (dodrop) {
   1592 		if (oldp != NULL || *oldlenp != 0)
   1593 			return EINVAL;
   1594 		if (newp == NULL)
   1595 			return EPERM;
   1596 		if (newlen < sizeof(sa))
   1597 			return ENOMEM;
   1598 	}
   1599 	if (namelen != 4 && namelen != 0)
   1600 		return EINVAL;
   1601 	if (name[-2] != IPPROTO_TCP)
   1602 		return EINVAL;
   1603 	pf = name[-3];
   1604 
   1605 	/* old style lookup, ipv4 only */
   1606 	if (namelen == 4) {
   1607 #ifdef INET
   1608 		struct in_addr laddr, raddr;
   1609 		u_int lport, rport;
   1610 
   1611 		if (pf != PF_INET)
   1612 			return EPROTONOSUPPORT;
   1613 		raddr.s_addr = (uint32_t)name[0];
   1614 		rport = (u_int)name[1];
   1615 		laddr.s_addr = (uint32_t)name[2];
   1616 		lport = (u_int)name[3];
   1617 
   1618 		mutex_enter(softnet_lock);
   1619 		error = inet4_ident_core(raddr, rport, laddr, lport,
   1620 		    oldp, oldlenp, l, dodrop);
   1621 		mutex_exit(softnet_lock);
   1622 		return error;
   1623 #else /* INET */
   1624 		return EINVAL;
   1625 #endif /* INET */
   1626 	}
   1627 
   1628 	if (newp == NULL || newlen != sizeof(sa))
   1629 		return EINVAL;
   1630 	error = copyin(newp, &sa, newlen);
   1631 	if (error)
   1632 		return error;
   1633 
   1634 	/*
   1635 	 * requested families must match
   1636 	 */
   1637 	if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family)
   1638 		return EINVAL;
   1639 
   1640 	switch (pf) {
   1641 #ifdef INET6
   1642 	case PF_INET6:
   1643 		si6[0] = (struct sockaddr_in6*)&sa[0];
   1644 		si6[1] = (struct sockaddr_in6*)&sa[1];
   1645 		if (si6[0]->sin6_len != sizeof(*si6[0]) ||
   1646 		    si6[1]->sin6_len != sizeof(*si6[1]))
   1647 			return EINVAL;
   1648 
   1649 		if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) &&
   1650 		    !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) {
   1651 			error = sa6_embedscope(si6[0], ip6_use_defzone);
   1652 			if (error)
   1653 				return error;
   1654 			error = sa6_embedscope(si6[1], ip6_use_defzone);
   1655 			if (error)
   1656 				return error;
   1657 
   1658 			mutex_enter(softnet_lock);
   1659 			error = inet6_ident_core(&si6[0]->sin6_addr,
   1660 			    si6[0]->sin6_port, &si6[1]->sin6_addr,
   1661 			    si6[1]->sin6_port, oldp, oldlenp, l, dodrop);
   1662 			mutex_exit(softnet_lock);
   1663 			return error;
   1664 		}
   1665 
   1666 		if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) !=
   1667 		    IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr))
   1668 			return EINVAL;
   1669 
   1670 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]);
   1671 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]);
   1672 		/*FALLTHROUGH*/
   1673 #endif /* INET6 */
   1674 #ifdef INET
   1675 	case PF_INET:
   1676 		si4[0] = (struct sockaddr_in*)&sa[0];
   1677 		si4[1] = (struct sockaddr_in*)&sa[1];
   1678 		if (si4[0]->sin_len != sizeof(*si4[0]) ||
   1679 		    si4[0]->sin_len != sizeof(*si4[1]))
   1680 			return EINVAL;
   1681 
   1682 		mutex_enter(softnet_lock);
   1683 		error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port,
   1684 		    si4[1]->sin_addr, si4[1]->sin_port,
   1685 		    oldp, oldlenp, l, dodrop);
   1686 		mutex_exit(softnet_lock);
   1687 		return error;
   1688 #endif /* INET */
   1689 	default:
   1690 		return EPROTONOSUPPORT;
   1691 	}
   1692 }
   1693 
   1694 /*
   1695  * sysctl helper for the inet and inet6 pcblists.  handles tcp/udp and
   1696  * inet/inet6, as well as raw pcbs for each.  specifically not
   1697  * declared static so that raw sockets and udp/udp6 can use it as
   1698  * well.
   1699  */
   1700 int
   1701 sysctl_inpcblist(SYSCTLFN_ARGS)
   1702 {
   1703 #ifdef INET
   1704 	struct sockaddr_in *in;
   1705 	const struct inpcb *inp;
   1706 #endif
   1707 #ifdef INET6
   1708 	struct sockaddr_in6 *in6;
   1709 	const struct in6pcb *in6p;
   1710 #endif
   1711 	struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data);
   1712 	const struct inpcb_hdr *inph;
   1713 	struct tcpcb *tp;
   1714 	struct kinfo_pcb pcb;
   1715 	char *dp;
   1716 	size_t len, needed, elem_size, out_size;
   1717 	int error, elem_count, pf, proto, pf2;
   1718 
   1719 	if (namelen != 4)
   1720 		return (EINVAL);
   1721 
   1722 	if (oldp != NULL) {
   1723 		    len = *oldlenp;
   1724 		    elem_size = name[2];
   1725 		    elem_count = name[3];
   1726 		    if (elem_size != sizeof(pcb))
   1727 			    return EINVAL;
   1728 	} else {
   1729 		    len = 0;
   1730 		    elem_count = INT_MAX;
   1731 		    elem_size = sizeof(pcb);
   1732 	}
   1733 	error = 0;
   1734 	dp = oldp;
   1735 	out_size = elem_size;
   1736 	needed = 0;
   1737 
   1738 	if (namelen == 1 && name[0] == CTL_QUERY)
   1739 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
   1740 
   1741 	if (name - oname != 4)
   1742 		return (EINVAL);
   1743 
   1744 	pf = oname[1];
   1745 	proto = oname[2];
   1746 	pf2 = (oldp != NULL) ? pf : 0;
   1747 
   1748 	mutex_enter(softnet_lock);
   1749 
   1750 	TAILQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) {
   1751 #ifdef INET
   1752 		inp = (const struct inpcb *)inph;
   1753 #endif
   1754 #ifdef INET6
   1755 		in6p = (const struct in6pcb *)inph;
   1756 #endif
   1757 
   1758 		if (inph->inph_af != pf)
   1759 			continue;
   1760 
   1761 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
   1762 		    KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL,
   1763 		    NULL) != 0)
   1764 			continue;
   1765 
   1766 		memset(&pcb, 0, sizeof(pcb));
   1767 
   1768 		pcb.ki_family = pf;
   1769 		pcb.ki_type = proto;
   1770 
   1771 		switch (pf2) {
   1772 		case 0:
   1773 			/* just probing for size */
   1774 			break;
   1775 #ifdef INET
   1776 		case PF_INET:
   1777 			pcb.ki_family = inp->inp_socket->so_proto->
   1778 			    pr_domain->dom_family;
   1779 			pcb.ki_type = inp->inp_socket->so_proto->
   1780 			    pr_type;
   1781 			pcb.ki_protocol = inp->inp_socket->so_proto->
   1782 			    pr_protocol;
   1783 			pcb.ki_pflags = inp->inp_flags;
   1784 
   1785 			pcb.ki_sostate = inp->inp_socket->so_state;
   1786 			pcb.ki_prstate = inp->inp_state;
   1787 			if (proto == IPPROTO_TCP) {
   1788 				tp = intotcpcb(inp);
   1789 				pcb.ki_tstate = tp->t_state;
   1790 				pcb.ki_tflags = tp->t_flags;
   1791 			}
   1792 
   1793 			pcb.ki_pcbaddr = PTRTOUINT64(inp);
   1794 			pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb);
   1795 			pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket);
   1796 
   1797 			pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc;
   1798 			pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc;
   1799 
   1800 			in = satosin(&pcb.ki_src);
   1801 			in->sin_len = sizeof(*in);
   1802 			in->sin_family = pf;
   1803 			in->sin_port = inp->inp_lport;
   1804 			in->sin_addr = inp->inp_laddr;
   1805 			if (pcb.ki_prstate >= INP_CONNECTED) {
   1806 				in = satosin(&pcb.ki_dst);
   1807 				in->sin_len = sizeof(*in);
   1808 				in->sin_family = pf;
   1809 				in->sin_port = inp->inp_fport;
   1810 				in->sin_addr = inp->inp_faddr;
   1811 			}
   1812 			break;
   1813 #endif
   1814 #ifdef INET6
   1815 		case PF_INET6:
   1816 			pcb.ki_family = in6p->in6p_socket->so_proto->
   1817 			    pr_domain->dom_family;
   1818 			pcb.ki_type = in6p->in6p_socket->so_proto->pr_type;
   1819 			pcb.ki_protocol = in6p->in6p_socket->so_proto->
   1820 			    pr_protocol;
   1821 			pcb.ki_pflags = in6p->in6p_flags;
   1822 
   1823 			pcb.ki_sostate = in6p->in6p_socket->so_state;
   1824 			pcb.ki_prstate = in6p->in6p_state;
   1825 			if (proto == IPPROTO_TCP) {
   1826 				tp = in6totcpcb(in6p);
   1827 				pcb.ki_tstate = tp->t_state;
   1828 				pcb.ki_tflags = tp->t_flags;
   1829 			}
   1830 
   1831 			pcb.ki_pcbaddr = PTRTOUINT64(in6p);
   1832 			pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb);
   1833 			pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket);
   1834 
   1835 			pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc;
   1836 			pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc;
   1837 
   1838 			in6 = satosin6(&pcb.ki_src);
   1839 			in6->sin6_len = sizeof(*in6);
   1840 			in6->sin6_family = pf;
   1841 			in6->sin6_port = in6p->in6p_lport;
   1842 			in6->sin6_flowinfo = in6p->in6p_flowinfo;
   1843 			in6->sin6_addr = in6p->in6p_laddr;
   1844 			in6->sin6_scope_id = 0; /* XXX? */
   1845 
   1846 			if (pcb.ki_prstate >= IN6P_CONNECTED) {
   1847 				in6 = satosin6(&pcb.ki_dst);
   1848 				in6->sin6_len = sizeof(*in6);
   1849 				in6->sin6_family = pf;
   1850 				in6->sin6_port = in6p->in6p_fport;
   1851 				in6->sin6_flowinfo = in6p->in6p_flowinfo;
   1852 				in6->sin6_addr = in6p->in6p_faddr;
   1853 				in6->sin6_scope_id = 0; /* XXX? */
   1854 			}
   1855 			break;
   1856 #endif
   1857 		}
   1858 
   1859 		if (len >= elem_size && elem_count > 0) {
   1860 			error = copyout(&pcb, dp, out_size);
   1861 			if (error) {
   1862 				mutex_exit(softnet_lock);
   1863 				return (error);
   1864 			}
   1865 			dp += elem_size;
   1866 			len -= elem_size;
   1867 		}
   1868 		needed += elem_size;
   1869 		if (elem_count > 0 && elem_count != INT_MAX)
   1870 			elem_count--;
   1871 	}
   1872 
   1873 	*oldlenp = needed;
   1874 	if (oldp == NULL)
   1875 		*oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb);
   1876 
   1877 	mutex_exit(softnet_lock);
   1878 
   1879 	return (error);
   1880 }
   1881 
   1882 static int
   1883 sysctl_tcp_congctl(SYSCTLFN_ARGS)
   1884 {
   1885 	struct sysctlnode node;
   1886 	int error;
   1887 	char newname[TCPCC_MAXLEN];
   1888 
   1889 	strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1);
   1890 
   1891 	node = *rnode;
   1892 	node.sysctl_data = newname;
   1893 	node.sysctl_size = sizeof(newname);
   1894 
   1895 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1896 
   1897 	if (error ||
   1898 	    newp == NULL ||
   1899 	    strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0)
   1900 		return error;
   1901 
   1902 	mutex_enter(softnet_lock);
   1903 	error = tcp_congctl_select(NULL, newname);
   1904 	mutex_exit(softnet_lock);
   1905 
   1906 	return error;
   1907 }
   1908 
   1909 static int
   1910 sysctl_tcp_init_win(SYSCTLFN_ARGS)
   1911 {
   1912 	int error;
   1913 	u_int iw;
   1914 	struct sysctlnode node;
   1915 
   1916 	iw = *(u_int *)rnode->sysctl_data;
   1917 	node = *rnode;
   1918 	node.sysctl_data = &iw;
   1919 	node.sysctl_size = sizeof(iw);
   1920 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1921 	if (error || newp == NULL)
   1922 		return error;
   1923 
   1924 	if (iw >= __arraycount(tcp_init_win_max))
   1925 		return EINVAL;
   1926 	*(u_int *)rnode->sysctl_data = iw;
   1927 	return 0;
   1928 }
   1929 
   1930 static int
   1931 sysctl_tcp_keep(SYSCTLFN_ARGS)
   1932 {
   1933 	int error;
   1934 	u_int tmp;
   1935 	struct sysctlnode node;
   1936 
   1937 	node = *rnode;
   1938 	tmp = *(u_int *)rnode->sysctl_data;
   1939 	node.sysctl_data = &tmp;
   1940 
   1941 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1942 	if (error || newp == NULL)
   1943 		return error;
   1944 
   1945 	mutex_enter(softnet_lock);
   1946 
   1947 	*(u_int *)rnode->sysctl_data = tmp;
   1948 	tcp_tcpcb_template();	/* update the template */
   1949 
   1950 	mutex_exit(softnet_lock);
   1951 	return 0;
   1952 }
   1953 
   1954 static int
   1955 sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS)
   1956 {
   1957 
   1958 	return (NETSTAT_SYSCTL(tcpstat_percpu, TCP_NSTATS));
   1959 }
   1960 
   1961 /*
   1962  * this (second stage) setup routine is a replacement for tcp_sysctl()
   1963  * (which is currently used for ipv4 and ipv6)
   1964  */
   1965 static void
   1966 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname,
   1967 			   const char *tcpname)
   1968 {
   1969 	const struct sysctlnode *sack_node;
   1970 	const struct sysctlnode *abc_node;
   1971 	const struct sysctlnode *ecn_node;
   1972 	const struct sysctlnode *congctl_node;
   1973 	const struct sysctlnode *mslt_node;
   1974 	const struct sysctlnode *vtw_node;
   1975 #ifdef TCP_DEBUG
   1976 	extern struct tcp_debug tcp_debug[TCP_NDEBUG];
   1977 	extern int tcp_debx;
   1978 #endif
   1979 
   1980 	sysctl_createv(clog, 0, NULL, NULL,
   1981 		       CTLFLAG_PERMANENT,
   1982 		       CTLTYPE_NODE, pfname, NULL,
   1983 		       NULL, 0, NULL, 0,
   1984 		       CTL_NET, pf, CTL_EOL);
   1985 	sysctl_createv(clog, 0, NULL, NULL,
   1986 		       CTLFLAG_PERMANENT,
   1987 		       CTLTYPE_NODE, tcpname,
   1988 		       SYSCTL_DESCR("TCP related settings"),
   1989 		       NULL, 0, NULL, 0,
   1990 		       CTL_NET, pf, IPPROTO_TCP, CTL_EOL);
   1991 
   1992 	sysctl_createv(clog, 0, NULL, NULL,
   1993 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1994 		       CTLTYPE_INT, "rfc1323",
   1995 		       SYSCTL_DESCR("Enable RFC1323 TCP extensions"),
   1996 		       sysctl_update_tcpcb_template, 0, &tcp_do_rfc1323, 0,
   1997 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL);
   1998 	sysctl_createv(clog, 0, NULL, NULL,
   1999 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2000 		       CTLTYPE_INT, "sendspace",
   2001 		       SYSCTL_DESCR("Default TCP send buffer size"),
   2002 		       NULL, 0, &tcp_sendspace, 0,
   2003 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL);
   2004 	sysctl_createv(clog, 0, NULL, NULL,
   2005 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2006 		       CTLTYPE_INT, "recvspace",
   2007 		       SYSCTL_DESCR("Default TCP receive buffer size"),
   2008 		       NULL, 0, &tcp_recvspace, 0,
   2009 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL);
   2010 	sysctl_createv(clog, 0, NULL, NULL,
   2011 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2012 		       CTLTYPE_INT, "mssdflt",
   2013 		       SYSCTL_DESCR("Default maximum segment size"),
   2014 		       sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0,
   2015 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL);
   2016 	sysctl_createv(clog, 0, NULL, NULL,
   2017 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2018 		       CTLTYPE_INT, "minmss",
   2019 		       SYSCTL_DESCR("Lower limit for TCP maximum segment size"),
   2020 		       NULL, 0, &tcp_minmss, 0,
   2021 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2022 	sysctl_createv(clog, 0, NULL, NULL,
   2023 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2024 		       CTLTYPE_INT, "msl",
   2025 		       SYSCTL_DESCR("Maximum Segment Life"),
   2026 		       NULL, 0, &tcp_msl, 0,
   2027 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSL, CTL_EOL);
   2028 	sysctl_createv(clog, 0, NULL, NULL,
   2029 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2030 		       CTLTYPE_INT, "syn_cache_limit",
   2031 		       SYSCTL_DESCR("Maximum number of entries in the TCP "
   2032 				    "compressed state engine"),
   2033 		       NULL, 0, &tcp_syn_cache_limit, 0,
   2034 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT,
   2035 		       CTL_EOL);
   2036 	sysctl_createv(clog, 0, NULL, NULL,
   2037 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2038 		       CTLTYPE_INT, "syn_bucket_limit",
   2039 		       SYSCTL_DESCR("Maximum number of entries per hash "
   2040 				    "bucket in the TCP compressed state "
   2041 				    "engine"),
   2042 		       NULL, 0, &tcp_syn_bucket_limit, 0,
   2043 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT,
   2044 		       CTL_EOL);
   2045 #if 0 /* obsoleted */
   2046 	sysctl_createv(clog, 0, NULL, NULL,
   2047 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2048 		       CTLTYPE_INT, "syn_cache_interval",
   2049 		       SYSCTL_DESCR("TCP compressed state engine's timer interval"),
   2050 		       NULL, 0, &tcp_syn_cache_interval, 0,
   2051 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER,
   2052 		       CTL_EOL);
   2053 #endif
   2054 	sysctl_createv(clog, 0, NULL, NULL,
   2055 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2056 		       CTLTYPE_INT, "init_win",
   2057 		       SYSCTL_DESCR("Initial TCP congestion window"),
   2058 		       sysctl_tcp_init_win, 0, &tcp_init_win, 0,
   2059 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL);
   2060 	sysctl_createv(clog, 0, NULL, NULL,
   2061 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2062 		       CTLTYPE_INT, "mss_ifmtu",
   2063 		       SYSCTL_DESCR("Use interface MTU for calculating MSS"),
   2064 		       NULL, 0, &tcp_mss_ifmtu, 0,
   2065 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL);
   2066 	sysctl_createv(clog, 0, NULL, &sack_node,
   2067 		       CTLFLAG_PERMANENT,
   2068 		       CTLTYPE_NODE, "sack",
   2069 		       SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"),
   2070 		       NULL, 0, NULL, 0,
   2071 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL);
   2072 
   2073 	/* Congctl subtree */
   2074 	sysctl_createv(clog, 0, NULL, &congctl_node,
   2075 		       CTLFLAG_PERMANENT,
   2076 		       CTLTYPE_NODE, "congctl",
   2077 		       SYSCTL_DESCR("TCP Congestion Control"),
   2078 	    	       NULL, 0, NULL, 0,
   2079 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2080 	sysctl_createv(clog, 0, &congctl_node, NULL,
   2081 		       CTLFLAG_PERMANENT,
   2082 		       CTLTYPE_STRING, "available",
   2083 		       SYSCTL_DESCR("Available Congestion Control Mechanisms"),
   2084 		       NULL, 0, tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL);
   2085 	sysctl_createv(clog, 0, &congctl_node, NULL,
   2086 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2087 		       CTLTYPE_STRING, "selected",
   2088 		       SYSCTL_DESCR("Selected Congestion Control Mechanism"),
   2089 		       sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN,
   2090 		       CTL_CREATE, CTL_EOL);
   2091 
   2092 	sysctl_createv(clog, 0, NULL, NULL,
   2093 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2094 		       CTLTYPE_INT, "win_scale",
   2095 		       SYSCTL_DESCR("Use RFC1323 window scale options"),
   2096 		       sysctl_update_tcpcb_template, 0, &tcp_do_win_scale, 0,
   2097 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL);
   2098 	sysctl_createv(clog, 0, NULL, NULL,
   2099 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2100 		       CTLTYPE_INT, "timestamps",
   2101 		       SYSCTL_DESCR("Use RFC1323 time stamp options"),
   2102 		       sysctl_update_tcpcb_template, 0, &tcp_do_timestamps, 0,
   2103 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL);
   2104 	sysctl_createv(clog, 0, NULL, NULL,
   2105 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2106 		       CTLTYPE_INT, "compat_42",
   2107 		       SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"),
   2108 		       NULL, 0, &tcp_compat_42, 0,
   2109 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL);
   2110 	sysctl_createv(clog, 0, NULL, NULL,
   2111 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2112 		       CTLTYPE_INT, "cwm",
   2113 		       SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window "
   2114 				    "Monitoring"),
   2115 		       NULL, 0, &tcp_cwm, 0,
   2116 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL);
   2117 	sysctl_createv(clog, 0, NULL, NULL,
   2118 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2119 		       CTLTYPE_INT, "cwm_burstsize",
   2120 		       SYSCTL_DESCR("Congestion Window Monitoring allowed "
   2121 				    "burst count in packets"),
   2122 		       NULL, 0, &tcp_cwm_burstsize, 0,
   2123 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE,
   2124 		       CTL_EOL);
   2125 	sysctl_createv(clog, 0, NULL, NULL,
   2126 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2127 		       CTLTYPE_INT, "ack_on_push",
   2128 		       SYSCTL_DESCR("Immediately return ACK when PSH is "
   2129 				    "received"),
   2130 		       NULL, 0, &tcp_ack_on_push, 0,
   2131 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL);
   2132 	sysctl_createv(clog, 0, NULL, NULL,
   2133 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2134 		       CTLTYPE_INT, "keepidle",
   2135 		       SYSCTL_DESCR("Allowed connection idle ticks before a "
   2136 				    "keepalive probe is sent"),
   2137 		       sysctl_tcp_keep, 0, &tcp_keepidle, 0,
   2138 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL);
   2139 	sysctl_createv(clog, 0, NULL, NULL,
   2140 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2141 		       CTLTYPE_INT, "keepintvl",
   2142 		       SYSCTL_DESCR("Ticks before next keepalive probe is sent"),
   2143 		       sysctl_tcp_keep, 0, &tcp_keepintvl, 0,
   2144 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL);
   2145 	sysctl_createv(clog, 0, NULL, NULL,
   2146 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2147 		       CTLTYPE_INT, "keepcnt",
   2148 		       SYSCTL_DESCR("Number of keepalive probes to send"),
   2149 		       sysctl_tcp_keep, 0, &tcp_keepcnt, 0,
   2150 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL);
   2151 	sysctl_createv(clog, 0, NULL, NULL,
   2152 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
   2153 		       CTLTYPE_INT, "slowhz",
   2154 		       SYSCTL_DESCR("Keepalive ticks per second"),
   2155 		       NULL, PR_SLOWHZ, NULL, 0,
   2156 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL);
   2157 	sysctl_createv(clog, 0, NULL, NULL,
   2158 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2159 		       CTLTYPE_INT, "log_refused",
   2160 		       SYSCTL_DESCR("Log refused TCP connections"),
   2161 		       NULL, 0, &tcp_log_refused, 0,
   2162 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL);
   2163 #if 0 /* obsoleted */
   2164 	sysctl_createv(clog, 0, NULL, NULL,
   2165 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2166 		       CTLTYPE_INT, "rstratelimit", NULL,
   2167 		       NULL, 0, &tcp_rst_ratelim, 0,
   2168 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL);
   2169 #endif
   2170 	sysctl_createv(clog, 0, NULL, NULL,
   2171 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2172 		       CTLTYPE_INT, "rstppslimit",
   2173 		       SYSCTL_DESCR("Maximum number of RST packets to send "
   2174 				    "per second"),
   2175 		       NULL, 0, &tcp_rst_ppslim, 0,
   2176 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL);
   2177 	sysctl_createv(clog, 0, NULL, NULL,
   2178 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2179 		       CTLTYPE_INT, "delack_ticks",
   2180 		       SYSCTL_DESCR("Number of ticks to delay sending an ACK"),
   2181 		       NULL, 0, &tcp_delack_ticks, 0,
   2182 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL);
   2183 	sysctl_createv(clog, 0, NULL, NULL,
   2184 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2185 		       CTLTYPE_INT, "init_win_local",
   2186 		       SYSCTL_DESCR("Initial TCP window size (in segments)"),
   2187 		       sysctl_tcp_init_win, 0, &tcp_init_win_local, 0,
   2188 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL,
   2189 		       CTL_EOL);
   2190 	sysctl_createv(clog, 0, NULL, NULL,
   2191 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2192 		       CTLTYPE_STRUCT, "ident",
   2193 		       SYSCTL_DESCR("RFC1413 Identification Protocol lookups"),
   2194 		       sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t),
   2195 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL);
   2196 	sysctl_createv(clog, 0, NULL, NULL,
   2197 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2198 		       CTLTYPE_INT, "do_loopback_cksum",
   2199 		       SYSCTL_DESCR("Perform TCP checksum on loopback"),
   2200 		       NULL, 0, &tcp_do_loopback_cksum, 0,
   2201 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM,
   2202 		       CTL_EOL);
   2203 	sysctl_createv(clog, 0, NULL, NULL,
   2204 		       CTLFLAG_PERMANENT,
   2205 		       CTLTYPE_STRUCT, "pcblist",
   2206 		       SYSCTL_DESCR("TCP protocol control block list"),
   2207 		       sysctl_inpcblist, 0, &tcbtable, 0,
   2208 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
   2209 		       CTL_EOL);
   2210 	sysctl_createv(clog, 0, NULL, NULL,
   2211 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2212 		       CTLTYPE_INT, "keepinit",
   2213 		       SYSCTL_DESCR("Ticks before initial tcp connection times out"),
   2214 		       sysctl_tcp_keep, 0, &tcp_keepinit, 0,
   2215 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2216 
   2217 	/* TCP socket buffers auto-sizing nodes */
   2218 	sysctl_createv(clog, 0, NULL, NULL,
   2219 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2220 		       CTLTYPE_INT, "recvbuf_auto",
   2221 		       SYSCTL_DESCR("Enable automatic receive "
   2222 		           "buffer sizing (experimental)"),
   2223 		       NULL, 0, &tcp_do_autorcvbuf, 0,
   2224 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2225 	sysctl_createv(clog, 0, NULL, NULL,
   2226 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2227 		       CTLTYPE_INT, "recvbuf_inc",
   2228 		       SYSCTL_DESCR("Incrementor step size of "
   2229 		           "automatic receive buffer"),
   2230 		       NULL, 0, &tcp_autorcvbuf_inc, 0,
   2231 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2232 	sysctl_createv(clog, 0, NULL, NULL,
   2233 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2234 		       CTLTYPE_INT, "recvbuf_max",
   2235 		       SYSCTL_DESCR("Max size of automatic receive buffer"),
   2236 		       NULL, 0, &tcp_autorcvbuf_max, 0,
   2237 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2238 
   2239 	sysctl_createv(clog, 0, NULL, NULL,
   2240 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2241 		       CTLTYPE_INT, "sendbuf_auto",
   2242 		       SYSCTL_DESCR("Enable automatic send "
   2243 		           "buffer sizing (experimental)"),
   2244 		       NULL, 0, &tcp_do_autosndbuf, 0,
   2245 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2246 	sysctl_createv(clog, 0, NULL, NULL,
   2247 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2248 		       CTLTYPE_INT, "sendbuf_inc",
   2249 		       SYSCTL_DESCR("Incrementor step size of "
   2250 		           "automatic send buffer"),
   2251 		       NULL, 0, &tcp_autosndbuf_inc, 0,
   2252 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2253 	sysctl_createv(clog, 0, NULL, NULL,
   2254 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2255 		       CTLTYPE_INT, "sendbuf_max",
   2256 		       SYSCTL_DESCR("Max size of automatic send buffer"),
   2257 		       NULL, 0, &tcp_autosndbuf_max, 0,
   2258 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2259 
   2260 	/* ECN subtree */
   2261 	sysctl_createv(clog, 0, NULL, &ecn_node,
   2262 	    	       CTLFLAG_PERMANENT,
   2263 		       CTLTYPE_NODE, "ecn",
   2264 	    	       SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"),
   2265 	    	       NULL, 0, NULL, 0,
   2266 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2267 	sysctl_createv(clog, 0, &ecn_node, NULL,
   2268 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2269 		       CTLTYPE_INT, "enable",
   2270 		       SYSCTL_DESCR("Enable TCP Explicit Congestion "
   2271 			   "Notification"),
   2272 	    	       NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL);
   2273 	sysctl_createv(clog, 0, &ecn_node, NULL,
   2274 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2275 		       CTLTYPE_INT, "maxretries",
   2276 		       SYSCTL_DESCR("Number of times to retry ECN setup "
   2277 			       "before disabling ECN on the connection"),
   2278 	    	       NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL);
   2279 
   2280 	/* SACK gets it's own little subtree. */
   2281 	sysctl_createv(clog, 0, NULL, &sack_node,
   2282 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2283 		       CTLTYPE_INT, "enable",
   2284 		       SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"),
   2285 		       NULL, 0, &tcp_do_sack, 0,
   2286 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
   2287 	sysctl_createv(clog, 0, NULL, &sack_node,
   2288 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2289 		       CTLTYPE_INT, "maxholes",
   2290 		       SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"),
   2291 		       NULL, 0, &tcp_sack_tp_maxholes, 0,
   2292 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
   2293 	sysctl_createv(clog, 0, NULL, &sack_node,
   2294 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2295 		       CTLTYPE_INT, "globalmaxholes",
   2296 		       SYSCTL_DESCR("Global maximum number of TCP SACK holes"),
   2297 		       NULL, 0, &tcp_sack_globalmaxholes, 0,
   2298 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
   2299 	sysctl_createv(clog, 0, NULL, &sack_node,
   2300 		       CTLFLAG_PERMANENT,
   2301 		       CTLTYPE_INT, "globalholes",
   2302 		       SYSCTL_DESCR("Global number of TCP SACK holes"),
   2303 		       NULL, 0, &tcp_sack_globalholes, 0,
   2304 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
   2305 
   2306 	sysctl_createv(clog, 0, NULL, NULL,
   2307 		       CTLFLAG_PERMANENT,
   2308 		       CTLTYPE_STRUCT, "stats",
   2309 		       SYSCTL_DESCR("TCP statistics"),
   2310 		       sysctl_net_inet_tcp_stats, 0, NULL, 0,
   2311 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS,
   2312 		       CTL_EOL);
   2313         sysctl_createv(clog, 0, NULL, NULL,
   2314                        CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2315                        CTLTYPE_INT, "local_by_rtt",
   2316                        SYSCTL_DESCR("Use RTT estimator to decide which hosts "
   2317 				    "are local"),
   2318 		       NULL, 0, &tcp_rttlocal, 0,
   2319 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2320 #ifdef TCP_DEBUG
   2321 	sysctl_createv(clog, 0, NULL, NULL,
   2322 		       CTLFLAG_PERMANENT,
   2323 		       CTLTYPE_STRUCT, "debug",
   2324 		       SYSCTL_DESCR("TCP sockets debug information"),
   2325 		       NULL, 0, &tcp_debug, sizeof(tcp_debug),
   2326 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG,
   2327 		       CTL_EOL);
   2328 	sysctl_createv(clog, 0, NULL, NULL,
   2329 		       CTLFLAG_PERMANENT,
   2330 		       CTLTYPE_INT, "debx",
   2331 		       SYSCTL_DESCR("Number of TCP debug sockets messages"),
   2332 		       NULL, 0, &tcp_debx, sizeof(tcp_debx),
   2333 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX,
   2334 		       CTL_EOL);
   2335 #endif
   2336 	sysctl_createv(clog, 0, NULL, NULL,
   2337 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2338 		       CTLTYPE_STRUCT, "drop",
   2339 		       SYSCTL_DESCR("TCP drop connection"),
   2340 		       sysctl_net_inet_tcp_drop, 0, NULL, 0,
   2341 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL);
   2342 	sysctl_createv(clog, 0, NULL, NULL,
   2343 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2344 		       CTLTYPE_INT, "iss_hash",
   2345 		       SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic "
   2346 				    "hash computation"),
   2347 		       NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948),
   2348 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
   2349 		       CTL_EOL);
   2350 
   2351 	/* ABC subtree */
   2352 
   2353 	sysctl_createv(clog, 0, NULL, &abc_node,
   2354 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc",
   2355 		       SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"),
   2356 		       NULL, 0, NULL, 0,
   2357 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2358 	sysctl_createv(clog, 0, &abc_node, NULL,
   2359 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2360 		       CTLTYPE_INT, "enable",
   2361 		       SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"),
   2362 		       NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL);
   2363 	sysctl_createv(clog, 0, &abc_node, NULL,
   2364 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2365 		       CTLTYPE_INT, "aggressive",
   2366 		       SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"),
   2367 		       NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL);
   2368 
   2369 	/* MSL tuning subtree */
   2370 
   2371 	sysctl_createv(clog, 0, NULL, &mslt_node,
   2372 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "mslt",
   2373 		       SYSCTL_DESCR("MSL Tuning for TIME_WAIT truncation"),
   2374 		       NULL, 0, NULL, 0,
   2375 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2376 	sysctl_createv(clog, 0, &mslt_node, NULL,
   2377 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2378 		       CTLTYPE_INT, "enable",
   2379 		       SYSCTL_DESCR("Enable TIME_WAIT truncation"),
   2380 		       NULL, 0, &tcp_msl_enable, 0, CTL_CREATE, CTL_EOL);
   2381 	sysctl_createv(clog, 0, &mslt_node, NULL,
   2382 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2383 		       CTLTYPE_INT, "loopback",
   2384 		       SYSCTL_DESCR("MSL value to use for loopback connections"),
   2385 		       NULL, 0, &tcp_msl_loop, 0, CTL_CREATE, CTL_EOL);
   2386 	sysctl_createv(clog, 0, &mslt_node, NULL,
   2387 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2388 		       CTLTYPE_INT, "local",
   2389 		       SYSCTL_DESCR("MSL value to use for local connections"),
   2390 		       NULL, 0, &tcp_msl_local, 0, CTL_CREATE, CTL_EOL);
   2391 	sysctl_createv(clog, 0, &mslt_node, NULL,
   2392 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2393 		       CTLTYPE_INT, "remote",
   2394 		       SYSCTL_DESCR("MSL value to use for remote connections"),
   2395 		       NULL, 0, &tcp_msl_remote, 0, CTL_CREATE, CTL_EOL);
   2396 	sysctl_createv(clog, 0, &mslt_node, NULL,
   2397 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2398 		       CTLTYPE_INT, "remote_threshold",
   2399 		       SYSCTL_DESCR("RTT estimate value to promote local to remote"),
   2400 		       NULL, 0, &tcp_msl_remote_threshold, 0, CTL_CREATE, CTL_EOL);
   2401 
   2402 	/* vestigial TIME_WAIT tuning subtree */
   2403 
   2404 	sysctl_createv(clog, 0, NULL, &vtw_node,
   2405 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "vtw",
   2406 		       SYSCTL_DESCR("Tuning for Vestigial TIME_WAIT"),
   2407 		       NULL, 0, NULL, 0,
   2408 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
   2409 	sysctl_createv(clog, 0, &vtw_node, NULL,
   2410 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2411 		       CTLTYPE_INT, "enable",
   2412 		       SYSCTL_DESCR("Enable Vestigial TIME_WAIT"),
   2413 		       sysctl_tcp_vtw_enable, 0,
   2414 	               (pf == AF_INET) ? &tcp4_vtw_enable : &tcp6_vtw_enable,
   2415 		       0, CTL_CREATE, CTL_EOL);
   2416 	sysctl_createv(clog, 0, &vtw_node, NULL,
   2417 		       CTLFLAG_PERMANENT|CTLFLAG_READONLY,
   2418 		       CTLTYPE_INT, "entries",
   2419 		       SYSCTL_DESCR("Maximum number of vestigial TIME_WAIT entries"),
   2420 		       NULL, 0, &tcp_vtw_entries, 0, CTL_CREATE, CTL_EOL);
   2421 }
   2422 
   2423 void
   2424 tcp_usrreq_init(void)
   2425 {
   2426 
   2427 #ifdef INET
   2428 	sysctl_net_inet_tcp_setup2(NULL, PF_INET, "inet", "tcp");
   2429 #endif
   2430 #ifdef INET6
   2431 	sysctl_net_inet_tcp_setup2(NULL, PF_INET6, "inet6", "tcp6");
   2432 #endif
   2433 }
   2434 
   2435 PR_WRAP_USRREQS(tcp)
   2436 #define	tcp_attach	tcp_attach_wrapper
   2437 #define	tcp_detach	tcp_detach_wrapper
   2438 #define	tcp_accept	tcp_accept_wrapper
   2439 #define	tcp_bind	tcp_bind_wrapper
   2440 #define	tcp_listen	tcp_listen_wrapper
   2441 #define	tcp_connect	tcp_connect_wrapper
   2442 #define	tcp_connect2	tcp_connect2_wrapper
   2443 #define	tcp_disconnect	tcp_disconnect_wrapper
   2444 #define	tcp_shutdown	tcp_shutdown_wrapper
   2445 #define	tcp_abort	tcp_abort_wrapper
   2446 #define	tcp_ioctl	tcp_ioctl_wrapper
   2447 #define	tcp_stat	tcp_stat_wrapper
   2448 #define	tcp_peeraddr	tcp_peeraddr_wrapper
   2449 #define	tcp_sockaddr	tcp_sockaddr_wrapper
   2450 #define	tcp_rcvd	tcp_rcvd_wrapper
   2451 #define	tcp_recvoob	tcp_recvoob_wrapper
   2452 #define	tcp_send	tcp_send_wrapper
   2453 #define	tcp_sendoob	tcp_sendoob_wrapper
   2454 #define	tcp_purgeif	tcp_purgeif_wrapper
   2455 #define	tcp_usrreq	tcp_usrreq_wrapper
   2456 
   2457 const struct pr_usrreqs tcp_usrreqs = {
   2458 	.pr_attach	= tcp_attach,
   2459 	.pr_detach	= tcp_detach,
   2460 	.pr_accept	= tcp_accept,
   2461 	.pr_bind	= tcp_bind,
   2462 	.pr_listen	= tcp_listen,
   2463 	.pr_connect	= tcp_connect,
   2464 	.pr_connect2	= tcp_connect2,
   2465 	.pr_disconnect	= tcp_disconnect,
   2466 	.pr_shutdown	= tcp_shutdown,
   2467 	.pr_abort	= tcp_abort,
   2468 	.pr_ioctl	= tcp_ioctl,
   2469 	.pr_stat	= tcp_stat,
   2470 	.pr_peeraddr	= tcp_peeraddr,
   2471 	.pr_sockaddr	= tcp_sockaddr,
   2472 	.pr_rcvd	= tcp_rcvd,
   2473 	.pr_recvoob	= tcp_recvoob,
   2474 	.pr_send	= tcp_send,
   2475 	.pr_sendoob	= tcp_sendoob,
   2476 	.pr_purgeif	= tcp_purgeif,
   2477 	.pr_generic	= tcp_usrreq,
   2478 };
   2479