xref: /dflybsd-src/sys/kern/uipc_socket.c (revision e6f30c11b835a7878a0ca02133e6bbb9abfad4ab)
1 /*
2  * Copyright (c) 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 2004 The DragonFly Project.  All rights reserved.
4  *
5  * This code is derived from software contributed to The DragonFly Project
6  * by Jeffrey M. Hsu.
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 DragonFly Project nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific, prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 2004 Jeffrey M. Hsu.  All rights reserved.
36  *
37  * License terms: all terms for the DragonFly license above plus the following:
38  *
39  * 4. All advertising materials mentioning features or use of this software
40  *    must display the following acknowledgement:
41  *
42  *	This product includes software developed by Jeffrey M. Hsu
43  *	for the DragonFly Project.
44  *
45  *    This requirement may be waived with permission from Jeffrey Hsu.
46  *    This requirement will sunset and may be removed on July 8 2005,
47  *    after which the standard DragonFly license (as shown above) will
48  *    apply.
49  */
50 
51 /*
52  * Copyright (c) 1982, 1986, 1988, 1990, 1993
53  *	The Regents of the University of California.  All rights reserved.
54  *
55  * Redistribution and use in source and binary forms, with or without
56  * modification, are permitted provided that the following conditions
57  * are met:
58  * 1. Redistributions of source code must retain the above copyright
59  *    notice, this list of conditions and the following disclaimer.
60  * 2. Redistributions in binary form must reproduce the above copyright
61  *    notice, this list of conditions and the following disclaimer in the
62  *    documentation and/or other materials provided with the distribution.
63  * 3. All advertising materials mentioning features or use of this software
64  *    must display the following acknowledgement:
65  *	This product includes software developed by the University of
66  *	California, Berkeley and its contributors.
67  * 4. Neither the name of the University nor the names of its contributors
68  *    may be used to endorse or promote products derived from this software
69  *    without specific prior written permission.
70  *
71  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
72  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
73  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
74  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
75  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
76  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
77  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
78  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
79  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
80  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
81  * SUCH DAMAGE.
82  *
83  *	@(#)uipc_socket.c	8.3 (Berkeley) 4/15/94
84  * $FreeBSD: src/sys/kern/uipc_socket.c,v 1.68.2.24 2003/11/11 17:18:18 silby Exp $
85  * $DragonFly: src/sys/kern/uipc_socket.c,v 1.34 2005/06/07 19:08:55 hsu Exp $
86  */
87 
88 #include "opt_inet.h"
89 
90 #include <sys/param.h>
91 #include <sys/systm.h>
92 #include <sys/fcntl.h>
93 #include <sys/malloc.h>
94 #include <sys/mbuf.h>
95 #include <sys/domain.h>
96 #include <sys/file.h>			/* for struct knote */
97 #include <sys/kernel.h>
98 #include <sys/malloc.h>
99 #include <sys/event.h>
100 #include <sys/poll.h>
101 #include <sys/proc.h>
102 #include <sys/protosw.h>
103 #include <sys/socket.h>
104 #include <sys/socketvar.h>
105 #include <sys/socketops.h>
106 #include <sys/resourcevar.h>
107 #include <sys/signalvar.h>
108 #include <sys/sysctl.h>
109 #include <sys/uio.h>
110 #include <sys/jail.h>
111 #include <vm/vm_zone.h>
112 
113 #include <sys/thread2.h>
114 
115 #include <machine/limits.h>
116 
117 #ifdef INET
118 static int	 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
119 #endif /* INET */
120 
121 static void 	filt_sordetach(struct knote *kn);
122 static int 	filt_soread(struct knote *kn, long hint);
123 static void 	filt_sowdetach(struct knote *kn);
124 static int	filt_sowrite(struct knote *kn, long hint);
125 static int	filt_solisten(struct knote *kn, long hint);
126 
127 static struct filterops solisten_filtops =
128 	{ 1, NULL, filt_sordetach, filt_solisten };
129 static struct filterops soread_filtops =
130 	{ 1, NULL, filt_sordetach, filt_soread };
131 static struct filterops sowrite_filtops =
132 	{ 1, NULL, filt_sowdetach, filt_sowrite };
133 
134 struct	vm_zone *socket_zone;
135 so_gen_t	so_gencnt;	/* generation count for sockets */
136 
137 MALLOC_DEFINE(M_SONAME, "soname", "socket name");
138 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");
139 
140 
141 static int somaxconn = SOMAXCONN;
142 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW,
143     &somaxconn, 0, "Maximum pending socket connection queue size");
144 
145 /*
146  * Socket operation routines.
147  * These routines are called by the routines in
148  * sys_socket.c or from a system process, and
149  * implement the semantics of socket operations by
150  * switching out to the protocol specific routines.
151  */
152 
153 /*
154  * Get a socket structure from our zone, and initialize it.
155  * We don't implement `waitok' yet (see comments in uipc_domain.c).
156  * Note that it would probably be better to allocate socket
157  * and PCB at the same time, but I'm not convinced that all
158  * the protocols can be easily modified to do this.
159  */
160 struct socket *
161 soalloc(waitok)
162 	int waitok;
163 {
164 	struct socket *so;
165 
166 	so = zalloc(socket_zone);
167 	if (so) {
168 		/* XXX race condition for reentrant kernel */
169 		bzero(so, sizeof *so);
170 		so->so_gencnt = ++so_gencnt;
171 		TAILQ_INIT(&so->so_aiojobq);
172 		TAILQ_INIT(&so->so_rcv.sb_sel.si_mlist);
173 		TAILQ_INIT(&so->so_snd.sb_sel.si_mlist);
174 	}
175 	return so;
176 }
177 
178 int
179 socreate(int dom, struct socket **aso, int type,
180 	int proto, struct thread *td)
181 {
182 	struct proc *p = td->td_proc;
183 	struct protosw *prp;
184 	struct socket *so;
185 	struct pru_attach_info ai;
186 	int error;
187 
188 	if (proto)
189 		prp = pffindproto(dom, proto, type);
190 	else
191 		prp = pffindtype(dom, type);
192 
193 	if (prp == 0 || prp->pr_usrreqs->pru_attach == 0)
194 		return (EPROTONOSUPPORT);
195 
196 	if (p->p_ucred->cr_prison && jail_socket_unixiproute_only &&
197 	    prp->pr_domain->dom_family != PF_LOCAL &&
198 	    prp->pr_domain->dom_family != PF_INET &&
199 	    prp->pr_domain->dom_family != PF_ROUTE) {
200 		return (EPROTONOSUPPORT);
201 	}
202 
203 	if (prp->pr_type != type)
204 		return (EPROTOTYPE);
205 	so = soalloc(p != 0);
206 	if (so == 0)
207 		return (ENOBUFS);
208 
209 	TAILQ_INIT(&so->so_incomp);
210 	TAILQ_INIT(&so->so_comp);
211 	so->so_type = type;
212 	so->so_cred = crhold(p->p_ucred);
213 	so->so_proto = prp;
214 	ai.sb_rlimit = &p->p_rlimit[RLIMIT_SBSIZE];
215 	ai.p_ucred = p->p_ucred;
216 	ai.fd_rdir = p->p_fd->fd_rdir;
217 	error = so_pru_attach(so, proto, &ai);
218 	if (error) {
219 		so->so_state |= SS_NOFDREF;
220 		sofree(so);
221 		return (error);
222 	}
223 	*aso = so;
224 	return (0);
225 }
226 
227 int
228 sobind(struct socket *so, struct sockaddr *nam, struct thread *td)
229 {
230 	int error;
231 
232 	crit_enter();
233 	error = so_pru_bind(so, nam, td);
234 	crit_exit();
235 	return (error);
236 }
237 
238 void
239 sodealloc(struct socket *so)
240 {
241 
242 	so->so_gencnt = ++so_gencnt;
243 	if (so->so_rcv.sb_hiwat)
244 		(void)chgsbsize(so->so_cred->cr_uidinfo,
245 		    &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
246 	if (so->so_snd.sb_hiwat)
247 		(void)chgsbsize(so->so_cred->cr_uidinfo,
248 		    &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
249 #ifdef INET
250 	/* remove accept filter if present */
251 	if (so->so_accf != NULL)
252 		do_setopt_accept_filter(so, NULL);
253 #endif /* INET */
254 	crfree(so->so_cred);
255 	zfree(socket_zone, so);
256 }
257 
258 int
259 solisten(struct socket *so, int backlog, struct thread *td)
260 {
261 	int error;
262 
263 	crit_enter();
264  	if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) {
265 		crit_exit();
266  		return (EINVAL);
267  	}
268 
269 	error = so_pru_listen(so, td);
270 	if (error) {
271 		crit_exit();
272 		return (error);
273 	}
274 	if (TAILQ_EMPTY(&so->so_comp))
275 		so->so_options |= SO_ACCEPTCONN;
276 	if (backlog < 0 || backlog > somaxconn)
277 		backlog = somaxconn;
278 	so->so_qlimit = backlog;
279 	crit_exit();
280 	return (0);
281 }
282 
283 void
284 sofree(struct socket *so)
285 {
286 	struct socket *head = so->so_head;
287 
288 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
289 		return;
290 	if (head != NULL) {
291 		if (so->so_state & SS_INCOMP) {
292 			TAILQ_REMOVE(&head->so_incomp, so, so_list);
293 			head->so_incqlen--;
294 		} else if (so->so_state & SS_COMP) {
295 			/*
296 			 * We must not decommission a socket that's
297 			 * on the accept(2) queue.  If we do, then
298 			 * accept(2) may hang after select(2) indicated
299 			 * that the listening socket was ready.
300 			 */
301 			return;
302 		} else {
303 			panic("sofree: not queued");
304 		}
305 		so->so_state &= ~SS_INCOMP;
306 		so->so_head = NULL;
307 	}
308 	sbrelease(&so->so_snd, so);
309 	sorflush(so);
310 	sodealloc(so);
311 }
312 
313 /*
314  * Close a socket on last file table reference removal.
315  * Initiate disconnect if connected.
316  * Free socket when disconnect complete.
317  */
318 int
319 soclose(struct socket *so)
320 {
321 	int error = 0;
322 
323 	crit_enter();
324 	funsetown(so->so_sigio);
325 	if (so->so_pcb == NULL)
326 		goto discard;
327 	if (so->so_state & SS_ISCONNECTED) {
328 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
329 			error = sodisconnect(so);
330 			if (error)
331 				goto drop;
332 		}
333 		if (so->so_options & SO_LINGER) {
334 			if ((so->so_state & SS_ISDISCONNECTING) &&
335 			    (so->so_state & SS_NBIO))
336 				goto drop;
337 			while (so->so_state & SS_ISCONNECTED) {
338 				error = tsleep((caddr_t)&so->so_timeo,
339 				    PCATCH, "soclos", so->so_linger * hz);
340 				if (error)
341 					break;
342 			}
343 		}
344 	}
345 drop:
346 	if (so->so_pcb) {
347 		int error2;
348 
349 		error2 = so_pru_detach(so);
350 		if (error == 0)
351 			error = error2;
352 	}
353 discard:
354 	if (so->so_options & SO_ACCEPTCONN) {
355 		struct socket *sp, *sonext;
356 
357 		sp = TAILQ_FIRST(&so->so_incomp);
358 		for (; sp != NULL; sp = sonext) {
359 			sonext = TAILQ_NEXT(sp, so_list);
360 			(void) soabort(sp);
361 		}
362 		for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
363 			sonext = TAILQ_NEXT(sp, so_list);
364 			/* Dequeue from so_comp since sofree() won't do it */
365 			TAILQ_REMOVE(&so->so_comp, sp, so_list);
366 			so->so_qlen--;
367 			sp->so_state &= ~SS_COMP;
368 			sp->so_head = NULL;
369 			(void) soabort(sp);
370 		}
371 	}
372 	if (so->so_state & SS_NOFDREF)
373 		panic("soclose: NOFDREF");
374 	so->so_state |= SS_NOFDREF;
375 	sofree(so);
376 	crit_exit();
377 	return (error);
378 }
379 
380 /*
381  * Must be called from a critical section.
382  */
383 int
384 soabort(so)
385 	struct socket *so;
386 {
387 	int error;
388 
389 	error = so_pru_abort(so);
390 	if (error) {
391 		sofree(so);
392 		return error;
393 	}
394 	return (0);
395 }
396 
397 int
398 soaccept(struct socket *so, struct sockaddr **nam)
399 {
400 	int error;
401 
402 	crit_enter();
403 	if ((so->so_state & SS_NOFDREF) == 0)
404 		panic("soaccept: !NOFDREF");
405 	so->so_state &= ~SS_NOFDREF;
406 	error = so_pru_accept(so, nam);
407 	crit_exit();
408 	return (error);
409 }
410 
411 int
412 soconnect(struct socket *so, struct sockaddr *nam, struct thread *td)
413 {
414 	int error;
415 
416 	if (so->so_options & SO_ACCEPTCONN)
417 		return (EOPNOTSUPP);
418 	crit_enter();
419 	/*
420 	 * If protocol is connection-based, can only connect once.
421 	 * Otherwise, if connected, try to disconnect first.
422 	 * This allows user to disconnect by connecting to, e.g.,
423 	 * a null address.
424 	 */
425 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
426 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
427 	    (error = sodisconnect(so)))) {
428 		error = EISCONN;
429 	} else {
430 		/*
431 		 * Prevent accumulated error from previous connection
432 		 * from biting us.
433 		 */
434 		so->so_error = 0;
435 		error = so_pru_connect(so, nam, td);
436 	}
437 	crit_exit();
438 	return (error);
439 }
440 
441 int
442 soconnect2(struct socket *so1, struct socket *so2)
443 {
444 	int error;
445 
446 	crit_enter();
447 	error = so_pru_connect2(so1, so2);
448 	crit_exit();
449 	return (error);
450 }
451 
452 int
453 sodisconnect(struct socket *so)
454 {
455 	int error;
456 
457 	crit_enter();
458 	if ((so->so_state & SS_ISCONNECTED) == 0) {
459 		error = ENOTCONN;
460 		goto bad;
461 	}
462 	if (so->so_state & SS_ISDISCONNECTING) {
463 		error = EALREADY;
464 		goto bad;
465 	}
466 	error = so_pru_disconnect(so);
467 bad:
468 	crit_exit();
469 	return (error);
470 }
471 
472 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
473 /*
474  * Send on a socket.
475  * If send must go all at once and message is larger than
476  * send buffering, then hard error.
477  * Lock against other senders.
478  * If must go all at once and not enough room now, then
479  * inform user that this would block and do nothing.
480  * Otherwise, if nonblocking, send as much as possible.
481  * The data to be sent is described by "uio" if nonzero,
482  * otherwise by the mbuf chain "top" (which must be null
483  * if uio is not).  Data provided in mbuf chain must be small
484  * enough to send all at once.
485  *
486  * Returns nonzero on error, timeout or signal; callers
487  * must check for short counts if EINTR/ERESTART are returned.
488  * Data and control buffers are freed on return.
489  */
490 int
491 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
492 	struct mbuf *top, struct mbuf *control, int flags,
493 	struct thread *td)
494 {
495 	struct mbuf **mp;
496 	struct mbuf *m;
497 	long space, len, resid;
498 	int clen = 0, error, dontroute, mlen;
499 	int atomic = sosendallatonce(so) || top;
500 	int pru_flags;
501 
502 	if (uio)
503 		resid = uio->uio_resid;
504 	else
505 		resid = top->m_pkthdr.len;
506 	/*
507 	 * In theory resid should be unsigned.
508 	 * However, space must be signed, as it might be less than 0
509 	 * if we over-committed, and we must use a signed comparison
510 	 * of space and resid.  On the other hand, a negative resid
511 	 * causes us to loop sending 0-length segments to the protocol.
512 	 *
513 	 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
514 	 * type sockets since that's an error.
515 	 */
516 	if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
517 		error = EINVAL;
518 		goto out;
519 	}
520 
521 	dontroute =
522 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
523 	    (so->so_proto->pr_flags & PR_ATOMIC);
524 	if (td->td_proc && td->td_proc->p_stats)
525 		td->td_proc->p_stats->p_ru.ru_msgsnd++;
526 	if (control)
527 		clen = control->m_len;
528 #define	gotoerr(errno)	{ error = errno; crit_exit(); goto release; }
529 
530 restart:
531 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
532 	if (error)
533 		goto out;
534 	do {
535 		crit_enter();
536 		if (so->so_state & SS_CANTSENDMORE)
537 			gotoerr(EPIPE);
538 		if (so->so_error) {
539 			error = so->so_error;
540 			so->so_error = 0;
541 			crit_exit();
542 			goto release;
543 		}
544 		if ((so->so_state & SS_ISCONNECTED) == 0) {
545 			/*
546 			 * `sendto' and `sendmsg' is allowed on a connection-
547 			 * based socket if it supports implied connect.
548 			 * Return ENOTCONN if not connected and no address is
549 			 * supplied.
550 			 */
551 			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
552 			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
553 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
554 				    !(resid == 0 && clen != 0))
555 					gotoerr(ENOTCONN);
556 			} else if (addr == 0)
557 			    gotoerr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
558 				   ENOTCONN : EDESTADDRREQ);
559 		}
560 		space = sbspace(&so->so_snd);
561 		if (flags & MSG_OOB)
562 			space += 1024;
563 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
564 		    clen > so->so_snd.sb_hiwat)
565 			gotoerr(EMSGSIZE);
566 		if (space < resid + clen && uio &&
567 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
568 			if (so->so_state & SS_NBIO)
569 				gotoerr(EWOULDBLOCK);
570 			sbunlock(&so->so_snd);
571 			error = sbwait(&so->so_snd);
572 			crit_exit();
573 			if (error)
574 				goto out;
575 			goto restart;
576 		}
577 		crit_exit();
578 		mp = &top;
579 		space -= clen;
580 		do {
581 		    if (uio == NULL) {
582 			/*
583 			 * Data is prepackaged in "top".
584 			 */
585 			resid = 0;
586 			if (flags & MSG_EOR)
587 				top->m_flags |= M_EOR;
588 		    } else do {
589 			m = m_getl(resid, MB_WAIT, MT_DATA,
590 				   top == NULL ? M_PKTHDR : 0, &mlen);
591 			if (top == NULL) {
592 				m->m_pkthdr.len = 0;
593 				m->m_pkthdr.rcvif = (struct ifnet *)0;
594 			}
595 			len = min(min(mlen, resid), space);
596 			if (resid < MINCLSIZE) {
597 				/*
598 				 * For datagram protocols, leave room
599 				 * for protocol headers in first mbuf.
600 				 */
601 				if (atomic && top == 0 && len < mlen)
602 					MH_ALIGN(m, len);
603 			}
604 			space -= len;
605 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
606 			resid = uio->uio_resid;
607 			m->m_len = len;
608 			*mp = m;
609 			top->m_pkthdr.len += len;
610 			if (error)
611 				goto release;
612 			mp = &m->m_next;
613 			if (resid <= 0) {
614 				if (flags & MSG_EOR)
615 					top->m_flags |= M_EOR;
616 				break;
617 			}
618 		    } while (space > 0 && atomic);
619 		    if (dontroute)
620 			    so->so_options |= SO_DONTROUTE;
621 		    if (flags & MSG_OOB) {
622 		    	    pru_flags = PRUS_OOB;
623 		    } else if ((flags & MSG_EOF) &&
624 		    	       (so->so_proto->pr_flags & PR_IMPLOPCL) &&
625 		    	       (resid <= 0)) {
626 			    /*
627 			     * If the user set MSG_EOF, the protocol
628 			     * understands this flag and nothing left to
629 			     * send then use PRU_SEND_EOF instead of PRU_SEND.
630 			     */
631 		    	    pru_flags = PRUS_EOF;
632 		    } else if (resid > 0 && space > 0) {
633 			    /* If there is more to send, set PRUS_MORETOCOME */
634 		    	    pru_flags = PRUS_MORETOCOME;
635 		    } else {
636 		    	    pru_flags = 0;
637 		    }
638 		    crit_enter();
639 		    /*
640 		     * XXX all the SS_CANTSENDMORE checks previously
641 		     * done could be out of date.  We could have recieved
642 		     * a reset packet in an interrupt or maybe we slept
643 		     * while doing page faults in uiomove() etc. We could
644 		     * probably recheck again inside the splnet() protection
645 		     * here, but there are probably other places that this
646 		     * also happens.  We must rethink this.
647 		     */
648 		    error = so_pru_send(so, pru_flags, top, addr, control, td);
649 		    crit_exit();
650 		    if (dontroute)
651 			    so->so_options &= ~SO_DONTROUTE;
652 		    clen = 0;
653 		    control = 0;
654 		    top = 0;
655 		    mp = &top;
656 		    if (error)
657 			    goto release;
658 		} while (resid && space > 0);
659 	} while (resid);
660 
661 release:
662 	sbunlock(&so->so_snd);
663 out:
664 	if (top)
665 		m_freem(top);
666 	if (control)
667 		m_freem(control);
668 	return (error);
669 }
670 
671 /*
672  * A specialization of sosend() for UDP based on protocol-specific knowledge:
673  *   so->so_proto->pr_flags has the PR_ATOMIC field set.  This means that
674  *	sosendallatonce() returns true,
675  *	the "atomic" variable is true,
676  *	and sosendudp() blocks until space is available for the entire send.
677  *   so->so_proto->pr_flags does not have the PR_CONNREQUIRED or
678  *	PR_IMPLOPCL flags set.
679  *   UDP has no out-of-band data.
680  *   UDP has no control data.
681  *   UDP does not support MSG_EOR.
682  */
683 int
684 sosendudp(struct socket *so, struct sockaddr *addr, struct uio *uio,
685 	  struct mbuf *top, struct mbuf *control, int flags, struct thread *td)
686 {
687 	int resid, error;
688 	boolean_t dontroute;		/* temporary SO_DONTROUTE setting */
689 
690 	if (td->td_proc && td->td_proc->p_stats)
691 		td->td_proc->p_stats->p_ru.ru_msgsnd++;
692 	if (control)
693 		m_freem(control);
694 
695 	KASSERT((uio && !top) || (top && !uio), ("bad arguments to sosendudp"));
696 	resid = uio ? uio->uio_resid : top->m_pkthdr.len;
697 
698 restart:
699 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
700 	if (error)
701 		goto out;
702 
703 	crit_enter();
704 	if (so->so_state & SS_CANTSENDMORE)
705 		gotoerr(EPIPE);
706 	if (so->so_error) {
707 		error = so->so_error;
708 		so->so_error = 0;
709 		crit_exit();
710 		goto release;
711 	}
712 	if (!(so->so_state & SS_ISCONNECTED) && addr == NULL)
713 		gotoerr(EDESTADDRREQ);
714 	if (resid > so->so_snd.sb_hiwat)
715 		gotoerr(EMSGSIZE);
716 	if (uio && sbspace(&so->so_snd) < resid) {
717 		if (so->so_state & SS_NBIO)
718 			gotoerr(EWOULDBLOCK);
719 		sbunlock(&so->so_snd);
720 		error = sbwait(&so->so_snd);
721 		crit_exit();
722 		if (error)
723 			goto out;
724 		goto restart;
725 	}
726 	crit_exit();
727 
728 	if (uio) {
729 		top = m_uiomove(uio);
730 		if (top == NULL)
731 			goto release;
732 	}
733 
734 	dontroute = (flags & MSG_DONTROUTE) && !(so->so_options & SO_DONTROUTE);
735 	if (dontroute)
736 		so->so_options |= SO_DONTROUTE;
737 
738 	error = so_pru_send(so, 0, top, addr, NULL, td);
739 	top = NULL;		/* sent or freed in lower layer */
740 
741 	if (dontroute)
742 		so->so_options &= ~SO_DONTROUTE;
743 
744 release:
745 	sbunlock(&so->so_snd);
746 out:
747 	if (top)
748 		m_freem(top);
749 	return (error);
750 }
751 
752 /*
753  * Implement receive operations on a socket.
754  * We depend on the way that records are added to the sockbuf
755  * by sbappend*.  In particular, each record (mbufs linked through m_next)
756  * must begin with an address if the protocol so specifies,
757  * followed by an optional mbuf or mbufs containing ancillary data,
758  * and then zero or more mbufs of data.
759  * In order to avoid blocking network interrupts for the entire time here,
760  * we exit the critical section while doing the actual copy to user space.
761  * Although the sockbuf is locked, new data may still be appended,
762  * and thus we must maintain consistency of the sockbuf during that time.
763  *
764  * The caller may receive the data as a single mbuf chain by supplying
765  * an mbuf **mp0 for use in returning the chain.  The uio is then used
766  * only for the count in uio_resid.
767  */
768 int
769 soreceive(so, psa, uio, mp0, controlp, flagsp)
770 	struct socket *so;
771 	struct sockaddr **psa;
772 	struct uio *uio;
773 	struct mbuf **mp0;
774 	struct mbuf **controlp;
775 	int *flagsp;
776 {
777 	struct mbuf *m, **mp;
778 	int flags, len, error, offset;
779 	struct protosw *pr = so->so_proto;
780 	struct mbuf *nextrecord;
781 	int moff, type = 0;
782 	int orig_resid = uio->uio_resid;
783 
784 	mp = mp0;
785 	if (psa)
786 		*psa = 0;
787 	if (controlp)
788 		*controlp = 0;
789 	if (flagsp)
790 		flags = *flagsp &~ MSG_EOR;
791 	else
792 		flags = 0;
793 	if (flags & MSG_OOB) {
794 		m = m_get(MB_WAIT, MT_DATA);
795 		if (m == NULL)
796 			return (ENOBUFS);
797 		error = so_pru_rcvoob(so, m, flags & MSG_PEEK);
798 		if (error)
799 			goto bad;
800 		do {
801 			error = uiomove(mtod(m, caddr_t),
802 			    (int) min(uio->uio_resid, m->m_len), uio);
803 			m = m_free(m);
804 		} while (uio->uio_resid && error == 0 && m);
805 bad:
806 		if (m)
807 			m_freem(m);
808 		return (error);
809 	}
810 	if (mp)
811 		*mp = (struct mbuf *)0;
812 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
813 		so_pru_rcvd(so, 0);
814 
815 restart:
816 	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
817 	if (error)
818 		return (error);
819 	crit_enter();
820 
821 	m = so->so_rcv.sb_mb;
822 	/*
823 	 * If we have less data than requested, block awaiting more
824 	 * (subject to any timeout) if:
825 	 *   1. the current count is less than the low water mark, or
826 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
827 	 *	receive operation at once if we block (resid <= hiwat).
828 	 *   3. MSG_DONTWAIT is not set
829 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
830 	 * we have to do the receive in sections, and thus risk returning
831 	 * a short count if a timeout or signal occurs after we start.
832 	 */
833 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
834 	    so->so_rcv.sb_cc < uio->uio_resid) &&
835 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
836 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
837 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
838 		KASSERT(m != 0 || !so->so_rcv.sb_cc, ("receive 1"));
839 		if (so->so_error) {
840 			if (m)
841 				goto dontblock;
842 			error = so->so_error;
843 			if ((flags & MSG_PEEK) == 0)
844 				so->so_error = 0;
845 			goto release;
846 		}
847 		if (so->so_state & SS_CANTRCVMORE) {
848 			if (m)
849 				goto dontblock;
850 			else
851 				goto release;
852 		}
853 		for (; m; m = m->m_next)
854 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
855 				m = so->so_rcv.sb_mb;
856 				goto dontblock;
857 			}
858 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
859 		    (pr->pr_flags & PR_CONNREQUIRED)) {
860 			error = ENOTCONN;
861 			goto release;
862 		}
863 		if (uio->uio_resid == 0)
864 			goto release;
865 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
866 			error = EWOULDBLOCK;
867 			goto release;
868 		}
869 		sbunlock(&so->so_rcv);
870 		error = sbwait(&so->so_rcv);
871 		crit_exit();
872 		if (error)
873 			return (error);
874 		goto restart;
875 	}
876 dontblock:
877 	if (uio->uio_td && uio->uio_td->td_proc)
878 		uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++;
879 	nextrecord = m->m_nextpkt;
880 	if (pr->pr_flags & PR_ADDR) {
881 		KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
882 		orig_resid = 0;
883 		if (psa)
884 			*psa = dup_sockaddr(mtod(m, struct sockaddr *));
885 		if (flags & MSG_PEEK) {
886 			m = m->m_next;
887 		} else {
888 			sbfree(&so->so_rcv, m);
889 			m->m_nextpkt = NULL;
890 			so->so_rcv.sb_mb = m_free(m);
891 			m = so->so_rcv.sb_mb;
892 		}
893 	}
894 	while (m && m->m_type == MT_CONTROL && error == 0) {
895 		if (flags & MSG_PEEK) {
896 			if (controlp)
897 				*controlp = m_copy(m, 0, m->m_len);
898 			m = m->m_next;
899 		} else {
900 			sbfree(&so->so_rcv, m);
901 			m->m_nextpkt = NULL;
902 			if (controlp) {
903 				if (pr->pr_domain->dom_externalize &&
904 				    mtod(m, struct cmsghdr *)->cmsg_type ==
905 				    SCM_RIGHTS)
906 				   error = (*pr->pr_domain->dom_externalize)(m);
907 				*controlp = m;
908 				so->so_rcv.sb_mb = m->m_next;
909 				m->m_next = NULL;
910 				m = so->so_rcv.sb_mb;
911 			} else {
912 				so->so_rcv.sb_mb = m_free(m);
913 				m = so->so_rcv.sb_mb;
914 			}
915 		}
916 		if (controlp) {
917 			orig_resid = 0;
918 			controlp = &(*controlp)->m_next;
919 		}
920 	}
921 	if (m) {
922 		if ((flags & MSG_PEEK) == 0)
923 			m->m_nextpkt = nextrecord;
924 		type = m->m_type;
925 		if (type == MT_OOBDATA)
926 			flags |= MSG_OOB;
927 	}
928 	moff = 0;
929 	offset = 0;
930 	while (m && uio->uio_resid > 0 && error == 0) {
931 		if (m->m_type == MT_OOBDATA) {
932 			if (type != MT_OOBDATA)
933 				break;
934 		} else if (type == MT_OOBDATA)
935 			break;
936 		else
937 		    KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
938 			("receive 3"));
939 		so->so_state &= ~SS_RCVATMARK;
940 		len = uio->uio_resid;
941 		if (so->so_oobmark && len > so->so_oobmark - offset)
942 			len = so->so_oobmark - offset;
943 		if (len > m->m_len - moff)
944 			len = m->m_len - moff;
945 		/*
946 		 * If mp is set, just pass back the mbufs.
947 		 * Otherwise copy them out via the uio, then free.
948 		 * Sockbuf must be consistent here (points to current mbuf,
949 		 * it points to next record) when we drop priority;
950 		 * we must note any additions to the sockbuf when we
951 		 * block interrupts again.
952 		 */
953 		if (mp == 0) {
954 			crit_exit();
955 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
956 			crit_enter();
957 			if (error)
958 				goto release;
959 		} else
960 			uio->uio_resid -= len;
961 		if (len == m->m_len - moff) {
962 			if (m->m_flags & M_EOR)
963 				flags |= MSG_EOR;
964 			if (flags & MSG_PEEK) {
965 				m = m->m_next;
966 				moff = 0;
967 			} else {
968 				nextrecord = m->m_nextpkt;
969 				m->m_nextpkt = NULL;
970 				sbfree(&so->so_rcv, m);
971 				if (mp) {
972 					*mp = m;
973 					mp = &m->m_next;
974 					so->so_rcv.sb_mb = m = m->m_next;
975 					*mp = (struct mbuf *)0;
976 				} else {
977 					so->so_rcv.sb_mb = m = m_free(m);
978 				}
979 				if (m)
980 					m->m_nextpkt = nextrecord;
981 				else
982 					so->so_rcv.sb_lastmbuf = NULL;
983 			}
984 		} else {
985 			if (flags & MSG_PEEK)
986 				moff += len;
987 			else {
988 				if (mp)
989 					*mp = m_copym(m, 0, len, MB_WAIT);
990 				m->m_data += len;
991 				m->m_len -= len;
992 				so->so_rcv.sb_cc -= len;
993 			}
994 		}
995 		if (so->so_oobmark) {
996 			if ((flags & MSG_PEEK) == 0) {
997 				so->so_oobmark -= len;
998 				if (so->so_oobmark == 0) {
999 					so->so_state |= SS_RCVATMARK;
1000 					break;
1001 				}
1002 			} else {
1003 				offset += len;
1004 				if (offset == so->so_oobmark)
1005 					break;
1006 			}
1007 		}
1008 		if (flags & MSG_EOR)
1009 			break;
1010 		/*
1011 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
1012 		 * we must not quit until "uio->uio_resid == 0" or an error
1013 		 * termination.  If a signal/timeout occurs, return
1014 		 * with a short count but without error.
1015 		 * Keep sockbuf locked against other readers.
1016 		 */
1017 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
1018 		    !sosendallatonce(so) && !nextrecord) {
1019 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
1020 				break;
1021 			/*
1022 			 * The window might have closed to zero, make
1023 			 * sure we send an ack now that we've drained
1024 			 * the buffer or we might end up blocking until
1025 			 * the idle takes over (5 seconds).
1026 			 */
1027 			if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1028 				so_pru_rcvd(so, flags);
1029 			error = sbwait(&so->so_rcv);
1030 			if (error) {
1031 				sbunlock(&so->so_rcv);
1032 				crit_exit();
1033 				return (0);
1034 			}
1035 			m = so->so_rcv.sb_mb;
1036 			if (m)
1037 				nextrecord = m->m_nextpkt;
1038 		}
1039 	}
1040 
1041 	if (m && pr->pr_flags & PR_ATOMIC)
1042 		flags |= MSG_TRUNC;
1043 	if (!(flags & MSG_PEEK)) {
1044 		if (m == NULL) {
1045 			so->so_rcv.sb_mb = nextrecord;
1046 			so->so_rcv.sb_lastmbuf = NULL;
1047 		} else {
1048 			if (pr->pr_flags & PR_ATOMIC)
1049 				sbdroprecord(&so->so_rcv);
1050 			else if (m->m_nextpkt == NULL) {
1051 				KASSERT(so->so_rcv.sb_mb == m,
1052 				    ("sb_mb %p != m %p", so->so_rcv.sb_mb, m));
1053 				so->so_rcv.sb_lastrecord = m;
1054 			}
1055 		}
1056 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1057 			so_pru_rcvd(so, flags);
1058 	}
1059 
1060 	if (orig_resid == uio->uio_resid && orig_resid &&
1061 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1062 		sbunlock(&so->so_rcv);
1063 		crit_exit();
1064 		goto restart;
1065 	}
1066 
1067 	if (flagsp)
1068 		*flagsp |= flags;
1069 release:
1070 	sbunlock(&so->so_rcv);
1071 	crit_exit();
1072 	return (error);
1073 }
1074 
1075 int
1076 soshutdown(so, how)
1077 	struct socket *so;
1078 	int how;
1079 {
1080 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1081 		return (EINVAL);
1082 
1083 	if (how != SHUT_WR)
1084 		sorflush(so);
1085 	if (how != SHUT_RD)
1086 		return (so_pru_shutdown(so));
1087 	return (0);
1088 }
1089 
1090 void
1091 sorflush(so)
1092 	struct socket *so;
1093 {
1094 	struct sockbuf *sb = &so->so_rcv;
1095 	struct protosw *pr = so->so_proto;
1096 	struct sockbuf asb;
1097 
1098 	sb->sb_flags |= SB_NOINTR;
1099 	(void) sblock(sb, M_WAITOK);
1100 
1101 	crit_enter();
1102 	socantrcvmore(so);
1103 	sbunlock(sb);
1104 	asb = *sb;
1105 	bzero((caddr_t)sb, sizeof (*sb));
1106 	if (asb.sb_flags & SB_KNOTE) {
1107 		sb->sb_sel.si_note = asb.sb_sel.si_note;
1108 		sb->sb_flags = SB_KNOTE;
1109 	}
1110 	crit_exit();
1111 
1112 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1113 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
1114 	sbrelease(&asb, so);
1115 }
1116 
1117 #ifdef INET
1118 static int
1119 do_setopt_accept_filter(so, sopt)
1120 	struct	socket *so;
1121 	struct	sockopt *sopt;
1122 {
1123 	struct accept_filter_arg	*afap = NULL;
1124 	struct accept_filter	*afp;
1125 	struct so_accf	*af = so->so_accf;
1126 	int	error = 0;
1127 
1128 	/* do not set/remove accept filters on non listen sockets */
1129 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
1130 		error = EINVAL;
1131 		goto out;
1132 	}
1133 
1134 	/* removing the filter */
1135 	if (sopt == NULL) {
1136 		if (af != NULL) {
1137 			if (af->so_accept_filter != NULL &&
1138 				af->so_accept_filter->accf_destroy != NULL) {
1139 				af->so_accept_filter->accf_destroy(so);
1140 			}
1141 			if (af->so_accept_filter_str != NULL) {
1142 				FREE(af->so_accept_filter_str, M_ACCF);
1143 			}
1144 			FREE(af, M_ACCF);
1145 			so->so_accf = NULL;
1146 		}
1147 		so->so_options &= ~SO_ACCEPTFILTER;
1148 		return (0);
1149 	}
1150 	/* adding a filter */
1151 	/* must remove previous filter first */
1152 	if (af != NULL) {
1153 		error = EINVAL;
1154 		goto out;
1155 	}
1156 	/* don't put large objects on the kernel stack */
1157 	MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1158 	error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1159 	afap->af_name[sizeof(afap->af_name)-1] = '\0';
1160 	afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1161 	if (error)
1162 		goto out;
1163 	afp = accept_filt_get(afap->af_name);
1164 	if (afp == NULL) {
1165 		error = ENOENT;
1166 		goto out;
1167 	}
1168 	MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK);
1169 	bzero(af, sizeof(*af));
1170 	if (afp->accf_create != NULL) {
1171 		if (afap->af_name[0] != '\0') {
1172 			int len = strlen(afap->af_name) + 1;
1173 
1174 			MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1175 			strcpy(af->so_accept_filter_str, afap->af_name);
1176 		}
1177 		af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1178 		if (af->so_accept_filter_arg == NULL) {
1179 			FREE(af->so_accept_filter_str, M_ACCF);
1180 			FREE(af, M_ACCF);
1181 			so->so_accf = NULL;
1182 			error = EINVAL;
1183 			goto out;
1184 		}
1185 	}
1186 	af->so_accept_filter = afp;
1187 	so->so_accf = af;
1188 	so->so_options |= SO_ACCEPTFILTER;
1189 out:
1190 	if (afap != NULL)
1191 		FREE(afap, M_TEMP);
1192 	return (error);
1193 }
1194 #endif /* INET */
1195 
1196 /*
1197  * Perhaps this routine, and sooptcopyout(), below, ought to come in
1198  * an additional variant to handle the case where the option value needs
1199  * to be some kind of integer, but not a specific size.
1200  * In addition to their use here, these functions are also called by the
1201  * protocol-level pr_ctloutput() routines.
1202  */
1203 int
1204 sooptcopyin(sopt, buf, len, minlen)
1205 	struct	sockopt *sopt;
1206 	void	*buf;
1207 	size_t	len;
1208 	size_t	minlen;
1209 {
1210 	size_t	valsize;
1211 
1212 	/*
1213 	 * If the user gives us more than we wanted, we ignore it,
1214 	 * but if we don't get the minimum length the caller
1215 	 * wants, we return EINVAL.  On success, sopt->sopt_valsize
1216 	 * is set to however much we actually retrieved.
1217 	 */
1218 	if ((valsize = sopt->sopt_valsize) < minlen)
1219 		return EINVAL;
1220 	if (valsize > len)
1221 		sopt->sopt_valsize = valsize = len;
1222 
1223 	if (sopt->sopt_td != NULL)
1224 		return (copyin(sopt->sopt_val, buf, valsize));
1225 
1226 	bcopy(sopt->sopt_val, buf, valsize);
1227 	return 0;
1228 }
1229 
1230 int
1231 sosetopt(so, sopt)
1232 	struct socket *so;
1233 	struct sockopt *sopt;
1234 {
1235 	int	error, optval;
1236 	struct	linger l;
1237 	struct	timeval tv;
1238 	u_long  val;
1239 
1240 	error = 0;
1241 	sopt->sopt_dir = SOPT_SET;
1242 	if (sopt->sopt_level != SOL_SOCKET) {
1243 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1244 			return (so_pr_ctloutput(so, sopt));
1245 		}
1246 		error = ENOPROTOOPT;
1247 	} else {
1248 		switch (sopt->sopt_name) {
1249 #ifdef INET
1250 		case SO_ACCEPTFILTER:
1251 			error = do_setopt_accept_filter(so, sopt);
1252 			if (error)
1253 				goto bad;
1254 			break;
1255 #endif /* INET */
1256 		case SO_LINGER:
1257 			error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1258 			if (error)
1259 				goto bad;
1260 
1261 			so->so_linger = l.l_linger;
1262 			if (l.l_onoff)
1263 				so->so_options |= SO_LINGER;
1264 			else
1265 				so->so_options &= ~SO_LINGER;
1266 			break;
1267 
1268 		case SO_DEBUG:
1269 		case SO_KEEPALIVE:
1270 		case SO_DONTROUTE:
1271 		case SO_USELOOPBACK:
1272 		case SO_BROADCAST:
1273 		case SO_REUSEADDR:
1274 		case SO_REUSEPORT:
1275 		case SO_OOBINLINE:
1276 		case SO_TIMESTAMP:
1277 			error = sooptcopyin(sopt, &optval, sizeof optval,
1278 					    sizeof optval);
1279 			if (error)
1280 				goto bad;
1281 			if (optval)
1282 				so->so_options |= sopt->sopt_name;
1283 			else
1284 				so->so_options &= ~sopt->sopt_name;
1285 			break;
1286 
1287 		case SO_SNDBUF:
1288 		case SO_RCVBUF:
1289 		case SO_SNDLOWAT:
1290 		case SO_RCVLOWAT:
1291 			error = sooptcopyin(sopt, &optval, sizeof optval,
1292 					    sizeof optval);
1293 			if (error)
1294 				goto bad;
1295 
1296 			/*
1297 			 * Values < 1 make no sense for any of these
1298 			 * options, so disallow them.
1299 			 */
1300 			if (optval < 1) {
1301 				error = EINVAL;
1302 				goto bad;
1303 			}
1304 
1305 			switch (sopt->sopt_name) {
1306 			case SO_SNDBUF:
1307 			case SO_RCVBUF:
1308 				if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1309 				    &so->so_snd : &so->so_rcv, (u_long)optval,
1310 				    so,
1311 				    &curproc->p_rlimit[RLIMIT_SBSIZE]) == 0) {
1312 					error = ENOBUFS;
1313 					goto bad;
1314 				}
1315 				break;
1316 
1317 			/*
1318 			 * Make sure the low-water is never greater than
1319 			 * the high-water.
1320 			 */
1321 			case SO_SNDLOWAT:
1322 				so->so_snd.sb_lowat =
1323 				    (optval > so->so_snd.sb_hiwat) ?
1324 				    so->so_snd.sb_hiwat : optval;
1325 				break;
1326 			case SO_RCVLOWAT:
1327 				so->so_rcv.sb_lowat =
1328 				    (optval > so->so_rcv.sb_hiwat) ?
1329 				    so->so_rcv.sb_hiwat : optval;
1330 				break;
1331 			}
1332 			break;
1333 
1334 		case SO_SNDTIMEO:
1335 		case SO_RCVTIMEO:
1336 			error = sooptcopyin(sopt, &tv, sizeof tv,
1337 					    sizeof tv);
1338 			if (error)
1339 				goto bad;
1340 
1341 			/* assert(hz > 0); */
1342 			if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1343 			    tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1344 				error = EDOM;
1345 				goto bad;
1346 			}
1347 			/* assert(tick > 0); */
1348 			/* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1349 			val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1350 			if (val > SHRT_MAX) {
1351 				error = EDOM;
1352 				goto bad;
1353 			}
1354 			if (val == 0 && tv.tv_usec != 0)
1355 				val = 1;
1356 
1357 			switch (sopt->sopt_name) {
1358 			case SO_SNDTIMEO:
1359 				so->so_snd.sb_timeo = val;
1360 				break;
1361 			case SO_RCVTIMEO:
1362 				so->so_rcv.sb_timeo = val;
1363 				break;
1364 			}
1365 			break;
1366 		default:
1367 			error = ENOPROTOOPT;
1368 			break;
1369 		}
1370 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1371 			(void) so_pr_ctloutput(so, sopt);
1372 		}
1373 	}
1374 bad:
1375 	return (error);
1376 }
1377 
1378 /* Helper routine for getsockopt */
1379 int
1380 sooptcopyout(struct sockopt *sopt, const void *buf, size_t len)
1381 {
1382 	int	error;
1383 	size_t	valsize;
1384 
1385 	error = 0;
1386 
1387 	/*
1388 	 * Documented get behavior is that we always return a value,
1389 	 * possibly truncated to fit in the user's buffer.
1390 	 * Traditional behavior is that we always tell the user
1391 	 * precisely how much we copied, rather than something useful
1392 	 * like the total amount we had available for her.
1393 	 * Note that this interface is not idempotent; the entire answer must
1394 	 * generated ahead of time.
1395 	 */
1396 	valsize = min(len, sopt->sopt_valsize);
1397 	sopt->sopt_valsize = valsize;
1398 	if (sopt->sopt_val != 0) {
1399 		if (sopt->sopt_td != NULL)
1400 			error = copyout(buf, sopt->sopt_val, valsize);
1401 		else
1402 			bcopy(buf, sopt->sopt_val, valsize);
1403 	}
1404 	return error;
1405 }
1406 
1407 int
1408 sogetopt(so, sopt)
1409 	struct socket *so;
1410 	struct sockopt *sopt;
1411 {
1412 	int	error, optval;
1413 	struct	linger l;
1414 	struct	timeval tv;
1415 #ifdef INET
1416 	struct accept_filter_arg *afap;
1417 #endif
1418 
1419 	error = 0;
1420 	sopt->sopt_dir = SOPT_GET;
1421 	if (sopt->sopt_level != SOL_SOCKET) {
1422 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1423 			return (so_pr_ctloutput(so, sopt));
1424 		} else
1425 			return (ENOPROTOOPT);
1426 	} else {
1427 		switch (sopt->sopt_name) {
1428 #ifdef INET
1429 		case SO_ACCEPTFILTER:
1430 			if ((so->so_options & SO_ACCEPTCONN) == 0)
1431 				return (EINVAL);
1432 			MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1433 				M_TEMP, M_WAITOK);
1434 			bzero(afap, sizeof(*afap));
1435 			if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1436 				strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1437 				if (so->so_accf->so_accept_filter_str != NULL)
1438 					strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1439 			}
1440 			error = sooptcopyout(sopt, afap, sizeof(*afap));
1441 			FREE(afap, M_TEMP);
1442 			break;
1443 #endif /* INET */
1444 
1445 		case SO_LINGER:
1446 			l.l_onoff = so->so_options & SO_LINGER;
1447 			l.l_linger = so->so_linger;
1448 			error = sooptcopyout(sopt, &l, sizeof l);
1449 			break;
1450 
1451 		case SO_USELOOPBACK:
1452 		case SO_DONTROUTE:
1453 		case SO_DEBUG:
1454 		case SO_KEEPALIVE:
1455 		case SO_REUSEADDR:
1456 		case SO_REUSEPORT:
1457 		case SO_BROADCAST:
1458 		case SO_OOBINLINE:
1459 		case SO_TIMESTAMP:
1460 			optval = so->so_options & sopt->sopt_name;
1461 integer:
1462 			error = sooptcopyout(sopt, &optval, sizeof optval);
1463 			break;
1464 
1465 		case SO_TYPE:
1466 			optval = so->so_type;
1467 			goto integer;
1468 
1469 		case SO_ERROR:
1470 			optval = so->so_error;
1471 			so->so_error = 0;
1472 			goto integer;
1473 
1474 		case SO_SNDBUF:
1475 			optval = so->so_snd.sb_hiwat;
1476 			goto integer;
1477 
1478 		case SO_RCVBUF:
1479 			optval = so->so_rcv.sb_hiwat;
1480 			goto integer;
1481 
1482 		case SO_SNDLOWAT:
1483 			optval = so->so_snd.sb_lowat;
1484 			goto integer;
1485 
1486 		case SO_RCVLOWAT:
1487 			optval = so->so_rcv.sb_lowat;
1488 			goto integer;
1489 
1490 		case SO_SNDTIMEO:
1491 		case SO_RCVTIMEO:
1492 			optval = (sopt->sopt_name == SO_SNDTIMEO ?
1493 				  so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1494 
1495 			tv.tv_sec = optval / hz;
1496 			tv.tv_usec = (optval % hz) * tick;
1497 			error = sooptcopyout(sopt, &tv, sizeof tv);
1498 			break;
1499 
1500 		default:
1501 			error = ENOPROTOOPT;
1502 			break;
1503 		}
1504 		return (error);
1505 	}
1506 }
1507 
1508 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1509 int
1510 soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1511 {
1512 	struct mbuf *m, *m_prev;
1513 	int sopt_size = sopt->sopt_valsize, msize;
1514 
1515 	m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_DATA,
1516 		   0, &msize);
1517 	if (m == NULL)
1518 		return (ENOBUFS);
1519 	m->m_len = min(msize, sopt_size);
1520 	sopt_size -= m->m_len;
1521 	*mp = m;
1522 	m_prev = m;
1523 
1524 	while (sopt_size > 0) {
1525 		m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT,
1526 			   MT_DATA, 0, &msize);
1527 		if (m == NULL) {
1528 			m_freem(*mp);
1529 			return (ENOBUFS);
1530 		}
1531 		m->m_len = min(msize, sopt_size);
1532 		sopt_size -= m->m_len;
1533 		m_prev->m_next = m;
1534 		m_prev = m;
1535 	}
1536 	return (0);
1537 }
1538 
1539 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1540 int
1541 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1542 {
1543 	struct mbuf *m0 = m;
1544 
1545 	if (sopt->sopt_val == NULL)
1546 		return 0;
1547 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1548 		if (sopt->sopt_td != NULL) {
1549 			int error;
1550 
1551 			error = copyin(sopt->sopt_val, mtod(m, char *),
1552 				       m->m_len);
1553 			if (error != 0) {
1554 				m_freem(m0);
1555 				return (error);
1556 			}
1557 		} else
1558 			bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1559 		sopt->sopt_valsize -= m->m_len;
1560 		sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len;
1561 		m = m->m_next;
1562 	}
1563 	if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1564 		panic("ip6_sooptmcopyin");
1565 	return 0;
1566 }
1567 
1568 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1569 int
1570 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1571 {
1572 	struct mbuf *m0 = m;
1573 	size_t valsize = 0;
1574 
1575 	if (sopt->sopt_val == NULL)
1576 		return 0;
1577 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1578 		if (sopt->sopt_td != NULL) {
1579 			int error;
1580 
1581 			error = copyout(mtod(m, char *), sopt->sopt_val,
1582 				       m->m_len);
1583 			if (error != 0) {
1584 				m_freem(m0);
1585 				return (error);
1586 			}
1587 		} else
1588 			bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1589 	       sopt->sopt_valsize -= m->m_len;
1590 	       sopt->sopt_val = (caddr_t)sopt->sopt_val + m->m_len;
1591 	       valsize += m->m_len;
1592 	       m = m->m_next;
1593 	}
1594 	if (m != NULL) {
1595 		/* enough soopt buffer should be given from user-land */
1596 		m_freem(m0);
1597 		return (EINVAL);
1598 	}
1599 	sopt->sopt_valsize = valsize;
1600 	return 0;
1601 }
1602 
1603 void
1604 sohasoutofband(so)
1605 	struct socket *so;
1606 {
1607 	if (so->so_sigio != NULL)
1608 		pgsigio(so->so_sigio, SIGURG, 0);
1609 	selwakeup(&so->so_rcv.sb_sel);
1610 }
1611 
1612 int
1613 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td)
1614 {
1615 	int revents = 0;
1616 
1617 	crit_enter();
1618 
1619 	if (events & (POLLIN | POLLRDNORM))
1620 		if (soreadable(so))
1621 			revents |= events & (POLLIN | POLLRDNORM);
1622 
1623 	if (events & POLLINIGNEOF)
1624 		if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat ||
1625 			!TAILQ_EMPTY(&so->so_comp) || so->so_error)
1626 			revents |= POLLINIGNEOF;
1627 
1628 	if (events & (POLLOUT | POLLWRNORM))
1629 		if (sowriteable(so))
1630 			revents |= events & (POLLOUT | POLLWRNORM);
1631 
1632 	if (events & (POLLPRI | POLLRDBAND))
1633 		if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1634 			revents |= events & (POLLPRI | POLLRDBAND);
1635 
1636 	if (revents == 0) {
1637 		if (events &
1638 			(POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM |
1639 			 POLLRDBAND)) {
1640 			selrecord(td, &so->so_rcv.sb_sel);
1641 			so->so_rcv.sb_flags |= SB_SEL;
1642 		}
1643 
1644 		if (events & (POLLOUT | POLLWRNORM)) {
1645 			selrecord(td, &so->so_snd.sb_sel);
1646 			so->so_snd.sb_flags |= SB_SEL;
1647 		}
1648 	}
1649 
1650 	crit_exit();
1651 	return (revents);
1652 }
1653 
1654 int
1655 sokqfilter(struct file *fp, struct knote *kn)
1656 {
1657 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1658 	struct sockbuf *sb;
1659 
1660 	switch (kn->kn_filter) {
1661 	case EVFILT_READ:
1662 		if (so->so_options & SO_ACCEPTCONN)
1663 			kn->kn_fop = &solisten_filtops;
1664 		else
1665 			kn->kn_fop = &soread_filtops;
1666 		sb = &so->so_rcv;
1667 		break;
1668 	case EVFILT_WRITE:
1669 		kn->kn_fop = &sowrite_filtops;
1670 		sb = &so->so_snd;
1671 		break;
1672 	default:
1673 		return (1);
1674 	}
1675 
1676 	crit_enter();
1677 	SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1678 	sb->sb_flags |= SB_KNOTE;
1679 	crit_exit();
1680 	return (0);
1681 }
1682 
1683 static void
1684 filt_sordetach(struct knote *kn)
1685 {
1686 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1687 
1688 	crit_enter();
1689 	SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1690 	if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1691 		so->so_rcv.sb_flags &= ~SB_KNOTE;
1692 	crit_exit();
1693 }
1694 
1695 /*ARGSUSED*/
1696 static int
1697 filt_soread(struct knote *kn, long hint)
1698 {
1699 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1700 
1701 	kn->kn_data = so->so_rcv.sb_cc;
1702 	if (so->so_state & SS_CANTRCVMORE) {
1703 		kn->kn_flags |= EV_EOF;
1704 		kn->kn_fflags = so->so_error;
1705 		return (1);
1706 	}
1707 	if (so->so_error)	/* temporary udp error */
1708 		return (1);
1709 	if (kn->kn_sfflags & NOTE_LOWAT)
1710 		return (kn->kn_data >= kn->kn_sdata);
1711 	return (kn->kn_data >= so->so_rcv.sb_lowat);
1712 }
1713 
1714 static void
1715 filt_sowdetach(struct knote *kn)
1716 {
1717 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1718 
1719 	crit_enter();
1720 	SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1721 	if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1722 		so->so_snd.sb_flags &= ~SB_KNOTE;
1723 	crit_exit();
1724 }
1725 
1726 /*ARGSUSED*/
1727 static int
1728 filt_sowrite(struct knote *kn, long hint)
1729 {
1730 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1731 
1732 	kn->kn_data = sbspace(&so->so_snd);
1733 	if (so->so_state & SS_CANTSENDMORE) {
1734 		kn->kn_flags |= EV_EOF;
1735 		kn->kn_fflags = so->so_error;
1736 		return (1);
1737 	}
1738 	if (so->so_error)	/* temporary udp error */
1739 		return (1);
1740 	if (((so->so_state & SS_ISCONNECTED) == 0) &&
1741 	    (so->so_proto->pr_flags & PR_CONNREQUIRED))
1742 		return (0);
1743 	if (kn->kn_sfflags & NOTE_LOWAT)
1744 		return (kn->kn_data >= kn->kn_sdata);
1745 	return (kn->kn_data >= so->so_snd.sb_lowat);
1746 }
1747 
1748 /*ARGSUSED*/
1749 static int
1750 filt_solisten(struct knote *kn, long hint)
1751 {
1752 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1753 
1754 	kn->kn_data = so->so_qlen;
1755 	return (! TAILQ_EMPTY(&so->so_comp));
1756 }
1757