xref: /freebsd/sys/kern/uipc_socket.c (revision 77a0943ded95b9e6438f7db70c4a28e4d93946d4)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)uipc_socket.c	8.3 (Berkeley) 4/15/94
34  * $FreeBSD$
35  */
36 
37 #include "opt_inet.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/fcntl.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/domain.h>
45 #include <sys/file.h>			/* for struct knote */
46 #include <sys/kernel.h>
47 #include <sys/malloc.h>
48 #include <sys/event.h>
49 #include <sys/poll.h>
50 #include <sys/proc.h>
51 #include <sys/protosw.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/resourcevar.h>
55 #include <sys/signalvar.h>
56 #include <sys/sysctl.h>
57 #include <sys/uio.h>
58 #include <sys/jail.h>
59 #include <vm/vm_zone.h>
60 
61 #include <machine/limits.h>
62 
63 #ifdef INET
64 static int	 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
65 #endif
66 
67 static int 	filt_sorattach(struct knote *kn);
68 static void 	filt_sordetach(struct knote *kn);
69 static int 	filt_soread(struct knote *kn, long hint);
70 static int 	filt_sowattach(struct knote *kn);
71 static void 	filt_sowdetach(struct knote *kn);
72 static int	filt_sowrite(struct knote *kn, long hint);
73 static int	filt_solisten(struct knote *kn, long hint);
74 
75 static struct filterops solisten_filtops =
76 	{ 1, filt_sorattach, filt_sordetach, filt_solisten };
77 
78 struct filterops so_rwfiltops[] = {
79 	{ 1, filt_sorattach, filt_sordetach, filt_soread },
80 	{ 1, filt_sowattach, filt_sowdetach, filt_sowrite },
81 };
82 
83 struct	vm_zone *socket_zone;
84 so_gen_t	so_gencnt;	/* generation count for sockets */
85 
86 MALLOC_DEFINE(M_SONAME, "soname", "socket name");
87 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");
88 
89 SYSCTL_DECL(_kern_ipc);
90 
91 static int somaxconn = SOMAXCONN;
92 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW,
93     &somaxconn, 0, "Maximum pending socket connection queue size");
94 
95 /*
96  * Socket operation routines.
97  * These routines are called by the routines in
98  * sys_socket.c or from a system process, and
99  * implement the semantics of socket operations by
100  * switching out to the protocol specific routines.
101  */
102 
103 /*
104  * Get a socket structure from our zone, and initialize it.
105  * We don't implement `waitok' yet (see comments in uipc_domain.c).
106  * Note that it would probably be better to allocate socket
107  * and PCB at the same time, but I'm not convinced that all
108  * the protocols can be easily modified to do this.
109  */
110 struct socket *
111 soalloc(waitok)
112 	int waitok;
113 {
114 	struct socket *so;
115 
116 	so = zalloci(socket_zone);
117 	if (so) {
118 		/* XXX race condition for reentrant kernel */
119 		bzero(so, sizeof *so);
120 		so->so_gencnt = ++so_gencnt;
121 		so->so_zone = socket_zone;
122 		TAILQ_INIT(&so->so_aiojobq);
123 	}
124 	return so;
125 }
126 
127 int
128 socreate(dom, aso, type, proto, p)
129 	int dom;
130 	struct socket **aso;
131 	register int type;
132 	int proto;
133 	struct proc *p;
134 {
135 	register struct protosw *prp;
136 	register struct socket *so;
137 	register int error;
138 
139 	if (proto)
140 		prp = pffindproto(dom, proto, type);
141 	else
142 		prp = pffindtype(dom, type);
143 
144 	if (prp == 0 || prp->pr_usrreqs->pru_attach == 0)
145 		return (EPROTONOSUPPORT);
146 
147 	if (p->p_prison && jail_socket_unixiproute_only &&
148 	    prp->pr_domain->dom_family != PF_LOCAL &&
149 	    prp->pr_domain->dom_family != PF_INET &&
150 	    prp->pr_domain->dom_family != PF_ROUTE) {
151 		return (EPROTONOSUPPORT);
152 	}
153 
154 	if (prp->pr_type != type)
155 		return (EPROTOTYPE);
156 	so = soalloc(p != 0);
157 	if (so == 0)
158 		return (ENOBUFS);
159 
160 	TAILQ_INIT(&so->so_incomp);
161 	TAILQ_INIT(&so->so_comp);
162 	so->so_type = type;
163 	so->so_cred = p->p_ucred;
164 	crhold(so->so_cred);
165 	so->so_proto = prp;
166 	error = (*prp->pr_usrreqs->pru_attach)(so, proto, p);
167 	if (error) {
168 		so->so_state |= SS_NOFDREF;
169 		sofree(so);
170 		return (error);
171 	}
172 	*aso = so;
173 	return (0);
174 }
175 
176 int
177 sobind(so, nam, p)
178 	struct socket *so;
179 	struct sockaddr *nam;
180 	struct proc *p;
181 {
182 	int s = splnet();
183 	int error;
184 
185 	error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, p);
186 	splx(s);
187 	return (error);
188 }
189 
190 void
191 sodealloc(so)
192 	struct socket *so;
193 {
194 
195 	so->so_gencnt = ++so_gencnt;
196 	if (so->so_rcv.sb_hiwat)
197 		(void)chgsbsize(so->so_cred->cr_uidinfo,
198 		    &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
199 	if (so->so_snd.sb_hiwat)
200 		(void)chgsbsize(so->so_cred->cr_uidinfo,
201 		    &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
202 #ifdef INET
203 	if (so->so_accf != NULL) {
204 		if (so->so_accf->so_accept_filter != NULL &&
205 			so->so_accf->so_accept_filter->accf_destroy != NULL) {
206 			so->so_accf->so_accept_filter->accf_destroy(so);
207 		}
208 		if (so->so_accf->so_accept_filter_str != NULL)
209 			FREE(so->so_accf->so_accept_filter_str, M_ACCF);
210 		FREE(so->so_accf, M_ACCF);
211 	}
212 #endif
213 	crfree(so->so_cred);
214 	zfreei(so->so_zone, so);
215 }
216 
217 int
218 solisten(so, backlog, p)
219 	register struct socket *so;
220 	int backlog;
221 	struct proc *p;
222 {
223 	int s, error;
224 
225 	s = splnet();
226 	error = (*so->so_proto->pr_usrreqs->pru_listen)(so, p);
227 	if (error) {
228 		splx(s);
229 		return (error);
230 	}
231 	if (TAILQ_EMPTY(&so->so_comp))
232 		so->so_options |= SO_ACCEPTCONN;
233 	if (backlog < 0 || backlog > somaxconn)
234 		backlog = somaxconn;
235 	so->so_qlimit = backlog;
236 	splx(s);
237 	return (0);
238 }
239 
240 void
241 sofree(so)
242 	register struct socket *so;
243 {
244 	struct socket *head = so->so_head;
245 
246 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
247 		return;
248 	if (head != NULL) {
249 		if (so->so_state & SS_INCOMP) {
250 			TAILQ_REMOVE(&head->so_incomp, so, so_list);
251 			head->so_incqlen--;
252 		} else if (so->so_state & SS_COMP) {
253 			/*
254 			 * We must not decommission a socket that's
255 			 * on the accept(2) queue.  If we do, then
256 			 * accept(2) may hang after select(2) indicated
257 			 * that the listening socket was ready.
258 			 */
259 			return;
260 		} else {
261 			panic("sofree: not queued");
262 		}
263 		head->so_qlen--;
264 		so->so_state &= ~SS_INCOMP;
265 		so->so_head = NULL;
266 	}
267 	sbrelease(&so->so_snd, so);
268 	sorflush(so);
269 	sodealloc(so);
270 }
271 
272 /*
273  * Close a socket on last file table reference removal.
274  * Initiate disconnect if connected.
275  * Free socket when disconnect complete.
276  */
277 int
278 soclose(so)
279 	register struct socket *so;
280 {
281 	int s = splnet();		/* conservative */
282 	int error = 0;
283 
284 	funsetown(so->so_sigio);
285 	if (so->so_options & SO_ACCEPTCONN) {
286 		struct socket *sp, *sonext;
287 
288 		sp = TAILQ_FIRST(&so->so_incomp);
289 		for (; sp != NULL; sp = sonext) {
290 			sonext = TAILQ_NEXT(sp, so_list);
291 			(void) soabort(sp);
292 		}
293 		for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
294 			sonext = TAILQ_NEXT(sp, so_list);
295 			/* Dequeue from so_comp since sofree() won't do it */
296 			TAILQ_REMOVE(&so->so_comp, sp, so_list);
297 			so->so_qlen--;
298 			sp->so_state &= ~SS_COMP;
299 			sp->so_head = NULL;
300 			(void) soabort(sp);
301 		}
302 	}
303 	if (so->so_pcb == 0)
304 		goto discard;
305 	if (so->so_state & SS_ISCONNECTED) {
306 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
307 			error = sodisconnect(so);
308 			if (error)
309 				goto drop;
310 		}
311 		if (so->so_options & SO_LINGER) {
312 			if ((so->so_state & SS_ISDISCONNECTING) &&
313 			    (so->so_state & SS_NBIO))
314 				goto drop;
315 			while (so->so_state & SS_ISCONNECTED) {
316 				error = tsleep((caddr_t)&so->so_timeo,
317 				    PSOCK | PCATCH, "soclos", so->so_linger * hz);
318 				if (error)
319 					break;
320 			}
321 		}
322 	}
323 drop:
324 	if (so->so_pcb) {
325 		int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so);
326 		if (error == 0)
327 			error = error2;
328 	}
329 discard:
330 	if (so->so_state & SS_NOFDREF)
331 		panic("soclose: NOFDREF");
332 	so->so_state |= SS_NOFDREF;
333 	sofree(so);
334 	splx(s);
335 	return (error);
336 }
337 
338 /*
339  * Must be called at splnet...
340  */
341 int
342 soabort(so)
343 	struct socket *so;
344 {
345 	int error;
346 
347 	error = (*so->so_proto->pr_usrreqs->pru_abort)(so);
348 	if (error) {
349 		sofree(so);
350 		return error;
351 	}
352 	return (0);
353 }
354 
355 int
356 soaccept(so, nam)
357 	register struct socket *so;
358 	struct sockaddr **nam;
359 {
360 	int s = splnet();
361 	int error;
362 
363 	if ((so->so_state & SS_NOFDREF) == 0)
364 		panic("soaccept: !NOFDREF");
365 	so->so_state &= ~SS_NOFDREF;
366  	if ((so->so_state & SS_ISDISCONNECTED) == 0)
367 		error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam);
368 	else {
369 		if (nam)
370 			*nam = 0;
371 		error = 0;
372 	}
373 	splx(s);
374 	return (error);
375 }
376 
377 int
378 soconnect(so, nam, p)
379 	register struct socket *so;
380 	struct sockaddr *nam;
381 	struct proc *p;
382 {
383 	int s;
384 	int error;
385 
386 	if (so->so_options & SO_ACCEPTCONN)
387 		return (EOPNOTSUPP);
388 	s = splnet();
389 	/*
390 	 * If protocol is connection-based, can only connect once.
391 	 * Otherwise, if connected, try to disconnect first.
392 	 * This allows user to disconnect by connecting to, e.g.,
393 	 * a null address.
394 	 */
395 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
396 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
397 	    (error = sodisconnect(so))))
398 		error = EISCONN;
399 	else
400 		error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, p);
401 	splx(s);
402 	return (error);
403 }
404 
405 int
406 soconnect2(so1, so2)
407 	register struct socket *so1;
408 	struct socket *so2;
409 {
410 	int s = splnet();
411 	int error;
412 
413 	error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2);
414 	splx(s);
415 	return (error);
416 }
417 
418 int
419 sodisconnect(so)
420 	register struct socket *so;
421 {
422 	int s = splnet();
423 	int error;
424 
425 	if ((so->so_state & SS_ISCONNECTED) == 0) {
426 		error = ENOTCONN;
427 		goto bad;
428 	}
429 	if (so->so_state & SS_ISDISCONNECTING) {
430 		error = EALREADY;
431 		goto bad;
432 	}
433 	error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so);
434 bad:
435 	splx(s);
436 	return (error);
437 }
438 
439 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
440 /*
441  * Send on a socket.
442  * If send must go all at once and message is larger than
443  * send buffering, then hard error.
444  * Lock against other senders.
445  * If must go all at once and not enough room now, then
446  * inform user that this would block and do nothing.
447  * Otherwise, if nonblocking, send as much as possible.
448  * The data to be sent is described by "uio" if nonzero,
449  * otherwise by the mbuf chain "top" (which must be null
450  * if uio is not).  Data provided in mbuf chain must be small
451  * enough to send all at once.
452  *
453  * Returns nonzero on error, timeout or signal; callers
454  * must check for short counts if EINTR/ERESTART are returned.
455  * Data and control buffers are freed on return.
456  */
457 int
458 sosend(so, addr, uio, top, control, flags, p)
459 	register struct socket *so;
460 	struct sockaddr *addr;
461 	struct uio *uio;
462 	struct mbuf *top;
463 	struct mbuf *control;
464 	int flags;
465 	struct proc *p;
466 {
467 	struct mbuf **mp;
468 	register struct mbuf *m;
469 	register long space, len, resid;
470 	int clen = 0, error, s, dontroute, mlen;
471 	int atomic = sosendallatonce(so) || top;
472 
473 	if (uio)
474 		resid = uio->uio_resid;
475 	else
476 		resid = top->m_pkthdr.len;
477 	/*
478 	 * In theory resid should be unsigned.
479 	 * However, space must be signed, as it might be less than 0
480 	 * if we over-committed, and we must use a signed comparison
481 	 * of space and resid.  On the other hand, a negative resid
482 	 * causes us to loop sending 0-length segments to the protocol.
483 	 *
484 	 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
485 	 * type sockets since that's an error.
486 	 */
487 	if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
488 		error = EINVAL;
489 		goto out;
490 	}
491 
492 	dontroute =
493 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
494 	    (so->so_proto->pr_flags & PR_ATOMIC);
495 	if (p)
496 		p->p_stats->p_ru.ru_msgsnd++;
497 	if (control)
498 		clen = control->m_len;
499 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
500 
501 restart:
502 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
503 	if (error)
504 		goto out;
505 	do {
506 		s = splnet();
507 		if (so->so_state & SS_CANTSENDMORE)
508 			snderr(EPIPE);
509 		if (so->so_error) {
510 			error = so->so_error;
511 			so->so_error = 0;
512 			splx(s);
513 			goto release;
514 		}
515 		if ((so->so_state & SS_ISCONNECTED) == 0) {
516 			/*
517 			 * `sendto' and `sendmsg' is allowed on a connection-
518 			 * based socket if it supports implied connect.
519 			 * Return ENOTCONN if not connected and no address is
520 			 * supplied.
521 			 */
522 			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
523 			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
524 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
525 				    !(resid == 0 && clen != 0))
526 					snderr(ENOTCONN);
527 			} else if (addr == 0)
528 			    snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
529 				   ENOTCONN : EDESTADDRREQ);
530 		}
531 		space = sbspace(&so->so_snd);
532 		if (flags & MSG_OOB)
533 			space += 1024;
534 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
535 		    clen > so->so_snd.sb_hiwat)
536 			snderr(EMSGSIZE);
537 		if (space < resid + clen && uio &&
538 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
539 			if (so->so_state & SS_NBIO)
540 				snderr(EWOULDBLOCK);
541 			sbunlock(&so->so_snd);
542 			error = sbwait(&so->so_snd);
543 			splx(s);
544 			if (error)
545 				goto out;
546 			goto restart;
547 		}
548 		splx(s);
549 		mp = &top;
550 		space -= clen;
551 		do {
552 		    if (uio == NULL) {
553 			/*
554 			 * Data is prepackaged in "top".
555 			 */
556 			resid = 0;
557 			if (flags & MSG_EOR)
558 				top->m_flags |= M_EOR;
559 		    } else do {
560 			if (top == 0) {
561 				MGETHDR(m, M_WAIT, MT_DATA);
562 				if (m == NULL) {
563 					error = ENOBUFS;
564 					goto release;
565 				}
566 				mlen = MHLEN;
567 				m->m_pkthdr.len = 0;
568 				m->m_pkthdr.rcvif = (struct ifnet *)0;
569 			} else {
570 				MGET(m, M_WAIT, MT_DATA);
571 				if (m == NULL) {
572 					error = ENOBUFS;
573 					goto release;
574 				}
575 				mlen = MLEN;
576 			}
577 			if (resid >= MINCLSIZE) {
578 				MCLGET(m, M_WAIT);
579 				if ((m->m_flags & M_EXT) == 0)
580 					goto nopages;
581 				mlen = MCLBYTES;
582 				len = min(min(mlen, resid), space);
583 			} else {
584 nopages:
585 				len = min(min(mlen, resid), space);
586 				/*
587 				 * For datagram protocols, leave room
588 				 * for protocol headers in first mbuf.
589 				 */
590 				if (atomic && top == 0 && len < mlen)
591 					MH_ALIGN(m, len);
592 			}
593 			space -= len;
594 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
595 			resid = uio->uio_resid;
596 			m->m_len = len;
597 			*mp = m;
598 			top->m_pkthdr.len += len;
599 			if (error)
600 				goto release;
601 			mp = &m->m_next;
602 			if (resid <= 0) {
603 				if (flags & MSG_EOR)
604 					top->m_flags |= M_EOR;
605 				break;
606 			}
607 		    } while (space > 0 && atomic);
608 		    if (dontroute)
609 			    so->so_options |= SO_DONTROUTE;
610 		    s = splnet();				/* XXX */
611 		    /*
612 		     * XXX all the SS_CANTSENDMORE checks previously
613 		     * done could be out of date.  We could have recieved
614 		     * a reset packet in an interrupt or maybe we slept
615 		     * while doing page faults in uiomove() etc. We could
616 		     * probably recheck again inside the splnet() protection
617 		     * here, but there are probably other places that this
618 		     * also happens.  We must rethink this.
619 		     */
620 		    error = (*so->so_proto->pr_usrreqs->pru_send)(so,
621 			(flags & MSG_OOB) ? PRUS_OOB :
622 			/*
623 			 * If the user set MSG_EOF, the protocol
624 			 * understands this flag and nothing left to
625 			 * send then use PRU_SEND_EOF instead of PRU_SEND.
626 			 */
627 			((flags & MSG_EOF) &&
628 			 (so->so_proto->pr_flags & PR_IMPLOPCL) &&
629 			 (resid <= 0)) ?
630 				PRUS_EOF :
631 			/* If there is more to send set PRUS_MORETOCOME */
632 			(resid > 0 && space > 0) ? PRUS_MORETOCOME : 0,
633 			top, addr, control, p);
634 		    splx(s);
635 		    if (dontroute)
636 			    so->so_options &= ~SO_DONTROUTE;
637 		    clen = 0;
638 		    control = 0;
639 		    top = 0;
640 		    mp = &top;
641 		    if (error)
642 			goto release;
643 		} while (resid && space > 0);
644 	} while (resid);
645 
646 release:
647 	sbunlock(&so->so_snd);
648 out:
649 	if (top)
650 		m_freem(top);
651 	if (control)
652 		m_freem(control);
653 	return (error);
654 }
655 
656 /*
657  * Implement receive operations on a socket.
658  * We depend on the way that records are added to the sockbuf
659  * by sbappend*.  In particular, each record (mbufs linked through m_next)
660  * must begin with an address if the protocol so specifies,
661  * followed by an optional mbuf or mbufs containing ancillary data,
662  * and then zero or more mbufs of data.
663  * In order to avoid blocking network interrupts for the entire time here,
664  * we splx() while doing the actual copy to user space.
665  * Although the sockbuf is locked, new data may still be appended,
666  * and thus we must maintain consistency of the sockbuf during that time.
667  *
668  * The caller may receive the data as a single mbuf chain by supplying
669  * an mbuf **mp0 for use in returning the chain.  The uio is then used
670  * only for the count in uio_resid.
671  */
672 int
673 soreceive(so, psa, uio, mp0, controlp, flagsp)
674 	register struct socket *so;
675 	struct sockaddr **psa;
676 	struct uio *uio;
677 	struct mbuf **mp0;
678 	struct mbuf **controlp;
679 	int *flagsp;
680 {
681 	register struct mbuf *m, **mp;
682 	register int flags, len, error, s, offset;
683 	struct protosw *pr = so->so_proto;
684 	struct mbuf *nextrecord;
685 	int moff, type = 0;
686 	int orig_resid = uio->uio_resid;
687 
688 	mp = mp0;
689 	if (psa)
690 		*psa = 0;
691 	if (controlp)
692 		*controlp = 0;
693 	if (flagsp)
694 		flags = *flagsp &~ MSG_EOR;
695 	else
696 		flags = 0;
697 	if (flags & MSG_OOB) {
698 		m = m_get(M_WAIT, MT_DATA);
699 		if (m == NULL)
700 			return (ENOBUFS);
701 		error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK);
702 		if (error)
703 			goto bad;
704 		do {
705 			error = uiomove(mtod(m, caddr_t),
706 			    (int) min(uio->uio_resid, m->m_len), uio);
707 			m = m_free(m);
708 		} while (uio->uio_resid && error == 0 && m);
709 bad:
710 		if (m)
711 			m_freem(m);
712 		return (error);
713 	}
714 	if (mp)
715 		*mp = (struct mbuf *)0;
716 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
717 		(*pr->pr_usrreqs->pru_rcvd)(so, 0);
718 
719 restart:
720 	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
721 	if (error)
722 		return (error);
723 	s = splnet();
724 
725 	m = so->so_rcv.sb_mb;
726 	/*
727 	 * If we have less data than requested, block awaiting more
728 	 * (subject to any timeout) if:
729 	 *   1. the current count is less than the low water mark, or
730 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
731 	 *	receive operation at once if we block (resid <= hiwat).
732 	 *   3. MSG_DONTWAIT is not set
733 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
734 	 * we have to do the receive in sections, and thus risk returning
735 	 * a short count if a timeout or signal occurs after we start.
736 	 */
737 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
738 	    so->so_rcv.sb_cc < uio->uio_resid) &&
739 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
740 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
741 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
742 		KASSERT(m != 0 || !so->so_rcv.sb_cc, ("receive 1"));
743 		if (so->so_error) {
744 			if (m)
745 				goto dontblock;
746 			error = so->so_error;
747 			if ((flags & MSG_PEEK) == 0)
748 				so->so_error = 0;
749 			goto release;
750 		}
751 		if (so->so_state & SS_CANTRCVMORE) {
752 			if (m)
753 				goto dontblock;
754 			else
755 				goto release;
756 		}
757 		for (; m; m = m->m_next)
758 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
759 				m = so->so_rcv.sb_mb;
760 				goto dontblock;
761 			}
762 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
763 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
764 			error = ENOTCONN;
765 			goto release;
766 		}
767 		if (uio->uio_resid == 0)
768 			goto release;
769 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
770 			error = EWOULDBLOCK;
771 			goto release;
772 		}
773 		sbunlock(&so->so_rcv);
774 		error = sbwait(&so->so_rcv);
775 		splx(s);
776 		if (error)
777 			return (error);
778 		goto restart;
779 	}
780 dontblock:
781 	if (uio->uio_procp)
782 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
783 	nextrecord = m->m_nextpkt;
784 	if (pr->pr_flags & PR_ADDR) {
785 		KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
786 		orig_resid = 0;
787 		if (psa)
788 			*psa = dup_sockaddr(mtod(m, struct sockaddr *),
789 					    mp0 == 0);
790 		if (flags & MSG_PEEK) {
791 			m = m->m_next;
792 		} else {
793 			sbfree(&so->so_rcv, m);
794 			MFREE(m, so->so_rcv.sb_mb);
795 			m = so->so_rcv.sb_mb;
796 		}
797 	}
798 	while (m && m->m_type == MT_CONTROL && error == 0) {
799 		if (flags & MSG_PEEK) {
800 			if (controlp)
801 				*controlp = m_copy(m, 0, m->m_len);
802 			m = m->m_next;
803 		} else {
804 			sbfree(&so->so_rcv, m);
805 			if (controlp) {
806 				if (pr->pr_domain->dom_externalize &&
807 				    mtod(m, struct cmsghdr *)->cmsg_type ==
808 				    SCM_RIGHTS)
809 				   error = (*pr->pr_domain->dom_externalize)(m);
810 				*controlp = m;
811 				so->so_rcv.sb_mb = m->m_next;
812 				m->m_next = 0;
813 				m = so->so_rcv.sb_mb;
814 			} else {
815 				MFREE(m, so->so_rcv.sb_mb);
816 				m = so->so_rcv.sb_mb;
817 			}
818 		}
819 		if (controlp) {
820 			orig_resid = 0;
821 			controlp = &(*controlp)->m_next;
822 		}
823 	}
824 	if (m) {
825 		if ((flags & MSG_PEEK) == 0)
826 			m->m_nextpkt = nextrecord;
827 		type = m->m_type;
828 		if (type == MT_OOBDATA)
829 			flags |= MSG_OOB;
830 	}
831 	moff = 0;
832 	offset = 0;
833 	while (m && uio->uio_resid > 0 && error == 0) {
834 		if (m->m_type == MT_OOBDATA) {
835 			if (type != MT_OOBDATA)
836 				break;
837 		} else if (type == MT_OOBDATA)
838 			break;
839 		else
840 		    KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
841 			("receive 3"));
842 		so->so_state &= ~SS_RCVATMARK;
843 		len = uio->uio_resid;
844 		if (so->so_oobmark && len > so->so_oobmark - offset)
845 			len = so->so_oobmark - offset;
846 		if (len > m->m_len - moff)
847 			len = m->m_len - moff;
848 		/*
849 		 * If mp is set, just pass back the mbufs.
850 		 * Otherwise copy them out via the uio, then free.
851 		 * Sockbuf must be consistent here (points to current mbuf,
852 		 * it points to next record) when we drop priority;
853 		 * we must note any additions to the sockbuf when we
854 		 * block interrupts again.
855 		 */
856 		if (mp == 0) {
857 			splx(s);
858 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
859 			s = splnet();
860 			if (error)
861 				goto release;
862 		} else
863 			uio->uio_resid -= len;
864 		if (len == m->m_len - moff) {
865 			if (m->m_flags & M_EOR)
866 				flags |= MSG_EOR;
867 			if (flags & MSG_PEEK) {
868 				m = m->m_next;
869 				moff = 0;
870 			} else {
871 				nextrecord = m->m_nextpkt;
872 				sbfree(&so->so_rcv, m);
873 				if (mp) {
874 					*mp = m;
875 					mp = &m->m_next;
876 					so->so_rcv.sb_mb = m = m->m_next;
877 					*mp = (struct mbuf *)0;
878 				} else {
879 					MFREE(m, so->so_rcv.sb_mb);
880 					m = so->so_rcv.sb_mb;
881 				}
882 				if (m)
883 					m->m_nextpkt = nextrecord;
884 			}
885 		} else {
886 			if (flags & MSG_PEEK)
887 				moff += len;
888 			else {
889 				if (mp)
890 					*mp = m_copym(m, 0, len, M_WAIT);
891 				m->m_data += len;
892 				m->m_len -= len;
893 				so->so_rcv.sb_cc -= len;
894 			}
895 		}
896 		if (so->so_oobmark) {
897 			if ((flags & MSG_PEEK) == 0) {
898 				so->so_oobmark -= len;
899 				if (so->so_oobmark == 0) {
900 					so->so_state |= SS_RCVATMARK;
901 					break;
902 				}
903 			} else {
904 				offset += len;
905 				if (offset == so->so_oobmark)
906 					break;
907 			}
908 		}
909 		if (flags & MSG_EOR)
910 			break;
911 		/*
912 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
913 		 * we must not quit until "uio->uio_resid == 0" or an error
914 		 * termination.  If a signal/timeout occurs, return
915 		 * with a short count but without error.
916 		 * Keep sockbuf locked against other readers.
917 		 */
918 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
919 		    !sosendallatonce(so) && !nextrecord) {
920 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
921 				break;
922 			error = sbwait(&so->so_rcv);
923 			if (error) {
924 				sbunlock(&so->so_rcv);
925 				splx(s);
926 				return (0);
927 			}
928 			m = so->so_rcv.sb_mb;
929 			if (m)
930 				nextrecord = m->m_nextpkt;
931 		}
932 	}
933 
934 	if (m && pr->pr_flags & PR_ATOMIC) {
935 		flags |= MSG_TRUNC;
936 		if ((flags & MSG_PEEK) == 0)
937 			(void) sbdroprecord(&so->so_rcv);
938 	}
939 	if ((flags & MSG_PEEK) == 0) {
940 		if (m == 0)
941 			so->so_rcv.sb_mb = nextrecord;
942 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
943 			(*pr->pr_usrreqs->pru_rcvd)(so, flags);
944 	}
945 	if (orig_resid == uio->uio_resid && orig_resid &&
946 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
947 		sbunlock(&so->so_rcv);
948 		splx(s);
949 		goto restart;
950 	}
951 
952 	if (flagsp)
953 		*flagsp |= flags;
954 release:
955 	sbunlock(&so->so_rcv);
956 	splx(s);
957 	return (error);
958 }
959 
960 int
961 soshutdown(so, how)
962 	register struct socket *so;
963 	register int how;
964 {
965 	register struct protosw *pr = so->so_proto;
966 
967 	how++;
968 	if (how & FREAD)
969 		sorflush(so);
970 	if (how & FWRITE)
971 		return ((*pr->pr_usrreqs->pru_shutdown)(so));
972 	return (0);
973 }
974 
975 void
976 sorflush(so)
977 	register struct socket *so;
978 {
979 	register struct sockbuf *sb = &so->so_rcv;
980 	register struct protosw *pr = so->so_proto;
981 	register int s;
982 	struct sockbuf asb;
983 
984 	sb->sb_flags |= SB_NOINTR;
985 	(void) sblock(sb, M_WAITOK);
986 	s = splimp();
987 	socantrcvmore(so);
988 	sbunlock(sb);
989 	asb = *sb;
990 	bzero((caddr_t)sb, sizeof (*sb));
991 	splx(s);
992 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
993 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
994 	sbrelease(&asb, so);
995 }
996 
997 #ifdef INET
998 static int
999 do_setopt_accept_filter(so, sopt)
1000 	struct	socket *so;
1001 	struct	sockopt *sopt;
1002 {
1003 	struct accept_filter_arg	*afap = NULL;
1004 	struct accept_filter	*afp;
1005 	struct so_accf	*af = so->so_accf;
1006 	int	error = 0;
1007 
1008 	/* do not set/remove accept filters on non listen sockets */
1009 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
1010 		error = EINVAL;
1011 		goto out;
1012 	}
1013 
1014 	/* removing the filter */
1015 	if (sopt == NULL) {
1016 		if (af != NULL) {
1017 			if (af->so_accept_filter != NULL &&
1018 				af->so_accept_filter->accf_destroy != NULL) {
1019 				af->so_accept_filter->accf_destroy(so);
1020 			}
1021 			if (af->so_accept_filter_str != NULL) {
1022 				FREE(af->so_accept_filter_str, M_ACCF);
1023 			}
1024 			FREE(af, M_ACCF);
1025 			so->so_accf = NULL;
1026 		}
1027 		so->so_options &= ~SO_ACCEPTFILTER;
1028 		return (0);
1029 	}
1030 	/* adding a filter */
1031 	/* must remove previous filter first */
1032 	if (af != NULL) {
1033 		error = EINVAL;
1034 		goto out;
1035 	}
1036 	/* don't put large objects on the kernel stack */
1037 	MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1038 	error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1039 	afap->af_name[sizeof(afap->af_name)-1] = '\0';
1040 	afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1041 	if (error)
1042 		goto out;
1043 	afp = accept_filt_get(afap->af_name);
1044 	if (afp == NULL) {
1045 		error = ENOENT;
1046 		goto out;
1047 	}
1048 	MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK);
1049 	bzero(af, sizeof(*af));
1050 	if (afp->accf_create != NULL) {
1051 		if (afap->af_name[0] != '\0') {
1052 			int len = strlen(afap->af_name) + 1;
1053 
1054 			MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1055 			strcpy(af->so_accept_filter_str, afap->af_name);
1056 		}
1057 		af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1058 		if (af->so_accept_filter_arg == NULL) {
1059 			FREE(af->so_accept_filter_str, M_ACCF);
1060 			FREE(af, M_ACCF);
1061 			so->so_accf = NULL;
1062 			error = EINVAL;
1063 			goto out;
1064 		}
1065 	}
1066 	af->so_accept_filter = afp;
1067 	so->so_accf = af;
1068 	so->so_options |= SO_ACCEPTFILTER;
1069 out:
1070 	if (afap != NULL)
1071 		FREE(afap, M_TEMP);
1072 	return (error);
1073 }
1074 #endif /* INET */
1075 
1076 /*
1077  * Perhaps this routine, and sooptcopyout(), below, ought to come in
1078  * an additional variant to handle the case where the option value needs
1079  * to be some kind of integer, but not a specific size.
1080  * In addition to their use here, these functions are also called by the
1081  * protocol-level pr_ctloutput() routines.
1082  */
1083 int
1084 sooptcopyin(sopt, buf, len, minlen)
1085 	struct	sockopt *sopt;
1086 	void	*buf;
1087 	size_t	len;
1088 	size_t	minlen;
1089 {
1090 	size_t	valsize;
1091 
1092 	/*
1093 	 * If the user gives us more than we wanted, we ignore it,
1094 	 * but if we don't get the minimum length the caller
1095 	 * wants, we return EINVAL.  On success, sopt->sopt_valsize
1096 	 * is set to however much we actually retrieved.
1097 	 */
1098 	if ((valsize = sopt->sopt_valsize) < minlen)
1099 		return EINVAL;
1100 	if (valsize > len)
1101 		sopt->sopt_valsize = valsize = len;
1102 
1103 	if (sopt->sopt_p != 0)
1104 		return (copyin(sopt->sopt_val, buf, valsize));
1105 
1106 	bcopy(sopt->sopt_val, buf, valsize);
1107 	return 0;
1108 }
1109 
1110 int
1111 sosetopt(so, sopt)
1112 	struct socket *so;
1113 	struct sockopt *sopt;
1114 {
1115 	int	error, optval;
1116 	struct	linger l;
1117 	struct	timeval tv;
1118 	u_long  val;
1119 
1120 	error = 0;
1121 	if (sopt->sopt_level != SOL_SOCKET) {
1122 		if (so->so_proto && so->so_proto->pr_ctloutput)
1123 			return ((*so->so_proto->pr_ctloutput)
1124 				  (so, sopt));
1125 		error = ENOPROTOOPT;
1126 	} else {
1127 		switch (sopt->sopt_name) {
1128 #ifdef INET
1129 		case SO_ACCEPTFILTER:
1130 			error = do_setopt_accept_filter(so, sopt);
1131 			if (error)
1132 				goto bad;
1133 			break;
1134 #endif
1135 		case SO_LINGER:
1136 			error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1137 			if (error)
1138 				goto bad;
1139 
1140 			so->so_linger = l.l_linger;
1141 			if (l.l_onoff)
1142 				so->so_options |= SO_LINGER;
1143 			else
1144 				so->so_options &= ~SO_LINGER;
1145 			break;
1146 
1147 		case SO_DEBUG:
1148 		case SO_KEEPALIVE:
1149 		case SO_DONTROUTE:
1150 		case SO_USELOOPBACK:
1151 		case SO_BROADCAST:
1152 		case SO_REUSEADDR:
1153 		case SO_REUSEPORT:
1154 		case SO_OOBINLINE:
1155 		case SO_TIMESTAMP:
1156 			error = sooptcopyin(sopt, &optval, sizeof optval,
1157 					    sizeof optval);
1158 			if (error)
1159 				goto bad;
1160 			if (optval)
1161 				so->so_options |= sopt->sopt_name;
1162 			else
1163 				so->so_options &= ~sopt->sopt_name;
1164 			break;
1165 
1166 		case SO_SNDBUF:
1167 		case SO_RCVBUF:
1168 		case SO_SNDLOWAT:
1169 		case SO_RCVLOWAT:
1170 			error = sooptcopyin(sopt, &optval, sizeof optval,
1171 					    sizeof optval);
1172 			if (error)
1173 				goto bad;
1174 
1175 			/*
1176 			 * Values < 1 make no sense for any of these
1177 			 * options, so disallow them.
1178 			 */
1179 			if (optval < 1) {
1180 				error = EINVAL;
1181 				goto bad;
1182 			}
1183 
1184 			switch (sopt->sopt_name) {
1185 			case SO_SNDBUF:
1186 			case SO_RCVBUF:
1187 				if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1188 				    &so->so_snd : &so->so_rcv, (u_long)optval,
1189 				    so, curproc) == 0) {
1190 					error = ENOBUFS;
1191 					goto bad;
1192 				}
1193 				break;
1194 
1195 			/*
1196 			 * Make sure the low-water is never greater than
1197 			 * the high-water.
1198 			 */
1199 			case SO_SNDLOWAT:
1200 				so->so_snd.sb_lowat =
1201 				    (optval > so->so_snd.sb_hiwat) ?
1202 				    so->so_snd.sb_hiwat : optval;
1203 				break;
1204 			case SO_RCVLOWAT:
1205 				so->so_rcv.sb_lowat =
1206 				    (optval > so->so_rcv.sb_hiwat) ?
1207 				    so->so_rcv.sb_hiwat : optval;
1208 				break;
1209 			}
1210 			break;
1211 
1212 		case SO_SNDTIMEO:
1213 		case SO_RCVTIMEO:
1214 			error = sooptcopyin(sopt, &tv, sizeof tv,
1215 					    sizeof tv);
1216 			if (error)
1217 				goto bad;
1218 
1219 			/* assert(hz > 0); */
1220 			if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1221 			    tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1222 				error = EDOM;
1223 				goto bad;
1224 			}
1225 			/* assert(tick > 0); */
1226 			/* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1227 			val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1228 			if (val > SHRT_MAX) {
1229 				error = EDOM;
1230 				goto bad;
1231 			}
1232 
1233 			switch (sopt->sopt_name) {
1234 			case SO_SNDTIMEO:
1235 				so->so_snd.sb_timeo = val;
1236 				break;
1237 			case SO_RCVTIMEO:
1238 				so->so_rcv.sb_timeo = val;
1239 				break;
1240 			}
1241 			break;
1242 		default:
1243 			error = ENOPROTOOPT;
1244 			break;
1245 		}
1246 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1247 			(void) ((*so->so_proto->pr_ctloutput)
1248 				  (so, sopt));
1249 		}
1250 	}
1251 bad:
1252 	return (error);
1253 }
1254 
1255 /* Helper routine for getsockopt */
1256 int
1257 sooptcopyout(sopt, buf, len)
1258 	struct	sockopt *sopt;
1259 	void	*buf;
1260 	size_t	len;
1261 {
1262 	int	error;
1263 	size_t	valsize;
1264 
1265 	error = 0;
1266 
1267 	/*
1268 	 * Documented get behavior is that we always return a value,
1269 	 * possibly truncated to fit in the user's buffer.
1270 	 * Traditional behavior is that we always tell the user
1271 	 * precisely how much we copied, rather than something useful
1272 	 * like the total amount we had available for her.
1273 	 * Note that this interface is not idempotent; the entire answer must
1274 	 * generated ahead of time.
1275 	 */
1276 	valsize = min(len, sopt->sopt_valsize);
1277 	sopt->sopt_valsize = valsize;
1278 	if (sopt->sopt_val != 0) {
1279 		if (sopt->sopt_p != 0)
1280 			error = copyout(buf, sopt->sopt_val, valsize);
1281 		else
1282 			bcopy(buf, sopt->sopt_val, valsize);
1283 	}
1284 	return error;
1285 }
1286 
1287 int
1288 sogetopt(so, sopt)
1289 	struct socket *so;
1290 	struct sockopt *sopt;
1291 {
1292 	int	error, optval;
1293 	struct	linger l;
1294 	struct	timeval tv;
1295 #ifdef INET
1296 	struct accept_filter_arg *afap;
1297 #endif
1298 
1299 	error = 0;
1300 	if (sopt->sopt_level != SOL_SOCKET) {
1301 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1302 			return ((*so->so_proto->pr_ctloutput)
1303 				  (so, sopt));
1304 		} else
1305 			return (ENOPROTOOPT);
1306 	} else {
1307 		switch (sopt->sopt_name) {
1308 #ifdef INET
1309 		case SO_ACCEPTFILTER:
1310 			if ((so->so_options & SO_ACCEPTCONN) == 0)
1311 				return (EINVAL);
1312 			MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1313 				M_TEMP, M_WAITOK);
1314 			bzero(afap, sizeof(*afap));
1315 			if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1316 				strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1317 				if (so->so_accf->so_accept_filter_str != NULL)
1318 					strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1319 			}
1320 			error = sooptcopyout(sopt, afap, sizeof(*afap));
1321 			FREE(afap, M_TEMP);
1322 			break;
1323 #endif
1324 
1325 		case SO_LINGER:
1326 			l.l_onoff = so->so_options & SO_LINGER;
1327 			l.l_linger = so->so_linger;
1328 			error = sooptcopyout(sopt, &l, sizeof l);
1329 			break;
1330 
1331 		case SO_USELOOPBACK:
1332 		case SO_DONTROUTE:
1333 		case SO_DEBUG:
1334 		case SO_KEEPALIVE:
1335 		case SO_REUSEADDR:
1336 		case SO_REUSEPORT:
1337 		case SO_BROADCAST:
1338 		case SO_OOBINLINE:
1339 		case SO_TIMESTAMP:
1340 			optval = so->so_options & sopt->sopt_name;
1341 integer:
1342 			error = sooptcopyout(sopt, &optval, sizeof optval);
1343 			break;
1344 
1345 		case SO_TYPE:
1346 			optval = so->so_type;
1347 			goto integer;
1348 
1349 		case SO_ERROR:
1350 			optval = so->so_error;
1351 			so->so_error = 0;
1352 			goto integer;
1353 
1354 		case SO_SNDBUF:
1355 			optval = so->so_snd.sb_hiwat;
1356 			goto integer;
1357 
1358 		case SO_RCVBUF:
1359 			optval = so->so_rcv.sb_hiwat;
1360 			goto integer;
1361 
1362 		case SO_SNDLOWAT:
1363 			optval = so->so_snd.sb_lowat;
1364 			goto integer;
1365 
1366 		case SO_RCVLOWAT:
1367 			optval = so->so_rcv.sb_lowat;
1368 			goto integer;
1369 
1370 		case SO_SNDTIMEO:
1371 		case SO_RCVTIMEO:
1372 			optval = (sopt->sopt_name == SO_SNDTIMEO ?
1373 				  so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1374 
1375 			tv.tv_sec = optval / hz;
1376 			tv.tv_usec = (optval % hz) * tick;
1377 			error = sooptcopyout(sopt, &tv, sizeof tv);
1378 			break;
1379 
1380 		default:
1381 			error = ENOPROTOOPT;
1382 			break;
1383 		}
1384 		return (error);
1385 	}
1386 }
1387 
1388 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1389 int
1390 soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1391 {
1392 	struct mbuf *m, *m_prev;
1393 	int sopt_size = sopt->sopt_valsize;
1394 
1395 	MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_DATA);
1396 	if (m == 0)
1397 		return ENOBUFS;
1398 	if (sopt_size > MLEN) {
1399 		MCLGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT);
1400 		if ((m->m_flags & M_EXT) == 0) {
1401 			m_free(m);
1402 			return ENOBUFS;
1403 		}
1404 		m->m_len = min(MCLBYTES, sopt_size);
1405 	} else {
1406 		m->m_len = min(MLEN, sopt_size);
1407 	}
1408 	sopt_size -= m->m_len;
1409 	*mp = m;
1410 	m_prev = m;
1411 
1412 	while (sopt_size) {
1413 		MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_DATA);
1414 		if (m == 0) {
1415 			m_freem(*mp);
1416 			return ENOBUFS;
1417 		}
1418 		if (sopt_size > MLEN) {
1419 			MCLGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT);
1420 			if ((m->m_flags & M_EXT) == 0) {
1421 				m_freem(*mp);
1422 				return ENOBUFS;
1423 			}
1424 			m->m_len = min(MCLBYTES, sopt_size);
1425 		} else {
1426 			m->m_len = min(MLEN, sopt_size);
1427 		}
1428 		sopt_size -= m->m_len;
1429 		m_prev->m_next = m;
1430 		m_prev = m;
1431 	}
1432 	return 0;
1433 }
1434 
1435 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1436 int
1437 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1438 {
1439 	struct mbuf *m0 = m;
1440 
1441 	if (sopt->sopt_val == NULL)
1442 		return 0;
1443 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1444 		if (sopt->sopt_p != NULL) {
1445 			int error;
1446 
1447 			error = copyin(sopt->sopt_val, mtod(m, char *),
1448 				       m->m_len);
1449 			if (error != 0) {
1450 				m_freem(m0);
1451 				return(error);
1452 			}
1453 		} else
1454 			bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1455 		sopt->sopt_valsize -= m->m_len;
1456 		(caddr_t)sopt->sopt_val += m->m_len;
1457 		m = m->m_next;
1458 	}
1459 	if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1460 		panic("ip6_sooptmcopyin");
1461 	return 0;
1462 }
1463 
1464 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1465 int
1466 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1467 {
1468 	struct mbuf *m0 = m;
1469 	size_t valsize = 0;
1470 
1471 	if (sopt->sopt_val == NULL)
1472 		return 0;
1473 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1474 		if (sopt->sopt_p != NULL) {
1475 			int error;
1476 
1477 			error = copyout(mtod(m, char *), sopt->sopt_val,
1478 				       m->m_len);
1479 			if (error != 0) {
1480 				m_freem(m0);
1481 				return(error);
1482 			}
1483 		} else
1484 			bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1485 	       sopt->sopt_valsize -= m->m_len;
1486 	       (caddr_t)sopt->sopt_val += m->m_len;
1487 	       valsize += m->m_len;
1488 	       m = m->m_next;
1489 	}
1490 	if (m != NULL) {
1491 		/* enough soopt buffer should be given from user-land */
1492 		m_freem(m0);
1493 		return(EINVAL);
1494 	}
1495 	sopt->sopt_valsize = valsize;
1496 	return 0;
1497 }
1498 
1499 void
1500 sohasoutofband(so)
1501 	register struct socket *so;
1502 {
1503 	if (so->so_sigio != NULL)
1504 		pgsigio(so->so_sigio, SIGURG, 0);
1505 	selwakeup(&so->so_rcv.sb_sel);
1506 }
1507 
1508 int
1509 sopoll(struct socket *so, int events, struct ucred *cred, struct proc *p)
1510 {
1511 	int revents = 0;
1512 	int s = splnet();
1513 
1514 	if (events & (POLLIN | POLLRDNORM))
1515 		if (soreadable(so))
1516 			revents |= events & (POLLIN | POLLRDNORM);
1517 
1518 	if (events & (POLLOUT | POLLWRNORM))
1519 		if (sowriteable(so))
1520 			revents |= events & (POLLOUT | POLLWRNORM);
1521 
1522 	if (events & (POLLPRI | POLLRDBAND))
1523 		if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1524 			revents |= events & (POLLPRI | POLLRDBAND);
1525 
1526 	if (revents == 0) {
1527 		if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
1528 			selrecord(p, &so->so_rcv.sb_sel);
1529 			so->so_rcv.sb_flags |= SB_SEL;
1530 		}
1531 
1532 		if (events & (POLLOUT | POLLWRNORM)) {
1533 			selrecord(p, &so->so_snd.sb_sel);
1534 			so->so_snd.sb_flags |= SB_SEL;
1535 		}
1536 	}
1537 
1538 	splx(s);
1539 	return (revents);
1540 }
1541 
1542 static int
1543 filt_sorattach(struct knote *kn)
1544 {
1545 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1546 	int s = splnet();
1547 
1548 	if (so->so_options & SO_ACCEPTCONN)
1549 		kn->kn_fop = &solisten_filtops;
1550 	SLIST_INSERT_HEAD(&so->so_rcv.sb_sel.si_note, kn, kn_selnext);
1551 	so->so_rcv.sb_flags |= SB_KNOTE;
1552 	splx(s);
1553 	return (0);
1554 }
1555 
1556 static void
1557 filt_sordetach(struct knote *kn)
1558 {
1559 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1560 	int s = splnet();
1561 
1562 	SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1563 	if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1564 		so->so_rcv.sb_flags &= ~SB_KNOTE;
1565 	splx(s);
1566 }
1567 
1568 /*ARGSUSED*/
1569 static int
1570 filt_soread(struct knote *kn, long hint)
1571 {
1572 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1573 
1574 	kn->kn_data = so->so_rcv.sb_cc;
1575 	if (so->so_state & SS_CANTRCVMORE) {
1576 		kn->kn_flags |= EV_EOF;
1577 		return (1);
1578 	}
1579 	if (so->so_error)	/* temporary udp error */
1580 		return (1);
1581 	return (kn->kn_data >= so->so_rcv.sb_lowat);
1582 }
1583 
1584 static int
1585 filt_sowattach(struct knote *kn)
1586 {
1587 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1588 	int s = splnet();
1589 
1590 	SLIST_INSERT_HEAD(&so->so_snd.sb_sel.si_note, kn, kn_selnext);
1591 	so->so_snd.sb_flags |= SB_KNOTE;
1592 	splx(s);
1593 	return (0);
1594 }
1595 
1596 static void
1597 filt_sowdetach(struct knote *kn)
1598 {
1599 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1600 	int s = splnet();
1601 
1602 	SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1603 	if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1604 		so->so_snd.sb_flags &= ~SB_KNOTE;
1605 	splx(s);
1606 }
1607 
1608 /*ARGSUSED*/
1609 static int
1610 filt_sowrite(struct knote *kn, long hint)
1611 {
1612 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1613 
1614 	kn->kn_data = sbspace(&so->so_snd);
1615 	if (so->so_state & SS_CANTSENDMORE) {
1616 		kn->kn_flags |= EV_EOF;
1617 		return (1);
1618 	}
1619 	if (so->so_error)	/* temporary udp error */
1620 		return (1);
1621 	if (((so->so_state & SS_ISCONNECTED) == 0) &&
1622 	    (so->so_proto->pr_flags & PR_CONNREQUIRED))
1623 		return (0);
1624 	return (kn->kn_data >= so->so_snd.sb_lowat);
1625 }
1626 
1627 /*ARGSUSED*/
1628 static int
1629 filt_solisten(struct knote *kn, long hint)
1630 {
1631 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1632 
1633 	kn->kn_data = so->so_qlen - so->so_incqlen;
1634 	return (! TAILQ_EMPTY(&so->so_comp));
1635 }
1636