xref: /freebsd/sys/kern/uipc_socket.c (revision 3ff369fed2a08f32dda232c10470b949bef9489f)
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/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/mutex.h>
46 #include <sys/domain.h>
47 #include <sys/file.h>			/* for struct knote */
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
50 #include <sys/event.h>
51 #include <sys/poll.h>
52 #include <sys/proc.h>
53 #include <sys/protosw.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/resourcevar.h>
57 #include <sys/signalvar.h>
58 #include <sys/sysctl.h>
59 #include <sys/uio.h>
60 #include <sys/jail.h>
61 
62 #include <vm/uma.h>
63 
64 #include <machine/limits.h>
65 
66 #ifdef INET
67 static int	 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
68 #endif
69 
70 static void 	filt_sordetach(struct knote *kn);
71 static int 	filt_soread(struct knote *kn, long hint);
72 static void 	filt_sowdetach(struct knote *kn);
73 static int	filt_sowrite(struct knote *kn, long hint);
74 static int	filt_solisten(struct knote *kn, long hint);
75 
76 static struct filterops solisten_filtops =
77 	{ 1, NULL, filt_sordetach, filt_solisten };
78 static struct filterops soread_filtops =
79 	{ 1, NULL, filt_sordetach, filt_soread };
80 static struct filterops sowrite_filtops =
81 	{ 1, NULL, filt_sowdetach, filt_sowrite };
82 
83 uma_zone_t 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 static int numopensockets;
95 SYSCTL_INT(_kern_ipc, OID_AUTO, numopensockets, CTLFLAG_RD,
96     &numopensockets, 0, "Number of open sockets");
97 
98 
99 /*
100  * Socket operation routines.
101  * These routines are called by the routines in
102  * sys_socket.c or from a system process, and
103  * implement the semantics of socket operations by
104  * switching out to the protocol specific routines.
105  */
106 
107 /*
108  * Get a socket structure from our zone, and initialize it.
109  * Note that it would probably be better to allocate socket
110  * and PCB at the same time, but I'm not convinced that all
111  * the protocols can be easily modified to do this.
112  *
113  * soalloc() returns a socket with a ref count of 0.
114  */
115 struct socket *
116 soalloc(waitok)
117 	int waitok;
118 {
119 	struct socket *so;
120 	int flag;
121 
122 	if (waitok == 1)
123 		flag = M_WAITOK;
124 	else
125 		flag = M_NOWAIT;
126 	flag |= M_ZERO;
127 	so = uma_zalloc(socket_zone, flag);
128 	if (so) {
129 		/* XXX race condition for reentrant kernel */
130 		so->so_gencnt = ++so_gencnt;
131 		/* sx_init(&so->so_sxlock, "socket sxlock"); */
132 		TAILQ_INIT(&so->so_aiojobq);
133 		++numopensockets;
134 	}
135 	return so;
136 }
137 
138 /*
139  * socreate returns a socket with a ref count of 1.  The socket should be
140  * closed with soclose().
141  */
142 int
143 socreate(dom, aso, type, proto, cred, td)
144 	int dom;
145 	struct socket **aso;
146 	register int type;
147 	int proto;
148 	struct ucred *cred;
149 	struct thread *td;
150 {
151 	register struct protosw *prp;
152 	register struct socket *so;
153 	register int error;
154 
155 	if (proto)
156 		prp = pffindproto(dom, proto, type);
157 	else
158 		prp = pffindtype(dom, type);
159 
160 	if (prp == 0 || prp->pr_usrreqs->pru_attach == 0)
161 		return (EPROTONOSUPPORT);
162 
163 	if (jailed(td->td_ucred) && jail_socket_unixiproute_only &&
164 	    prp->pr_domain->dom_family != PF_LOCAL &&
165 	    prp->pr_domain->dom_family != PF_INET &&
166 	    prp->pr_domain->dom_family != PF_ROUTE) {
167 		return (EPROTONOSUPPORT);
168 	}
169 
170 	if (prp->pr_type != type)
171 		return (EPROTOTYPE);
172 	so = soalloc(M_NOWAIT);
173 	if (so == NULL)
174 		return (ENOBUFS);
175 
176 	TAILQ_INIT(&so->so_incomp);
177 	TAILQ_INIT(&so->so_comp);
178 	so->so_type = type;
179 	so->so_cred = crhold(cred);
180 	so->so_proto = prp;
181 	soref(so);
182 	error = (*prp->pr_usrreqs->pru_attach)(so, proto, td);
183 	if (error) {
184 		so->so_state |= SS_NOFDREF;
185 		sorele(so);
186 		return (error);
187 	}
188 	*aso = so;
189 	return (0);
190 }
191 
192 int
193 sobind(so, nam, td)
194 	struct socket *so;
195 	struct sockaddr *nam;
196 	struct thread *td;
197 {
198 	int s = splnet();
199 	int error;
200 
201 	error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, td);
202 	splx(s);
203 	return (error);
204 }
205 
206 static void
207 sodealloc(struct socket *so)
208 {
209 
210 	KASSERT(so->so_count == 0, ("sodealloc(): so_count %d", so->so_count));
211 	so->so_gencnt = ++so_gencnt;
212 	if (so->so_rcv.sb_hiwat)
213 		(void)chgsbsize(so->so_cred->cr_uidinfo,
214 		    &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
215 	if (so->so_snd.sb_hiwat)
216 		(void)chgsbsize(so->so_cred->cr_uidinfo,
217 		    &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
218 #ifdef INET
219 	if (so->so_accf != NULL) {
220 		if (so->so_accf->so_accept_filter != NULL &&
221 			so->so_accf->so_accept_filter->accf_destroy != NULL) {
222 			so->so_accf->so_accept_filter->accf_destroy(so);
223 		}
224 		if (so->so_accf->so_accept_filter_str != NULL)
225 			FREE(so->so_accf->so_accept_filter_str, M_ACCF);
226 		FREE(so->so_accf, M_ACCF);
227 	}
228 #endif
229 	crfree(so->so_cred);
230 	/* sx_destroy(&so->so_sxlock); */
231 	uma_zfree(socket_zone, so);
232 	--numopensockets;
233 }
234 
235 int
236 solisten(so, backlog, td)
237 	register struct socket *so;
238 	int backlog;
239 	struct thread *td;
240 {
241 	int s, error;
242 
243 	s = splnet();
244 	error = (*so->so_proto->pr_usrreqs->pru_listen)(so, td);
245 	if (error) {
246 		splx(s);
247 		return (error);
248 	}
249 	if (TAILQ_EMPTY(&so->so_comp))
250 		so->so_options |= SO_ACCEPTCONN;
251 	if (backlog < 0 || backlog > somaxconn)
252 		backlog = somaxconn;
253 	so->so_qlimit = backlog;
254 	splx(s);
255 	return (0);
256 }
257 
258 void
259 sofree(so)
260 	register struct socket *so;
261 {
262 	struct socket *head = so->so_head;
263 
264 	KASSERT(so->so_count == 0, ("socket %p so_count not 0", so));
265 
266 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
267 		return;
268 	if (head != NULL) {
269 		if (so->so_state & SS_INCOMP) {
270 			TAILQ_REMOVE(&head->so_incomp, so, so_list);
271 			head->so_incqlen--;
272 		} else if (so->so_state & SS_COMP) {
273 			/*
274 			 * We must not decommission a socket that's
275 			 * on the accept(2) queue.  If we do, then
276 			 * accept(2) may hang after select(2) indicated
277 			 * that the listening socket was ready.
278 			 */
279 			return;
280 		} else {
281 			panic("sofree: not queued");
282 		}
283 		so->so_state &= ~SS_INCOMP;
284 		so->so_head = NULL;
285 	}
286 	sbrelease(&so->so_snd, so);
287 	sorflush(so);
288 	sodealloc(so);
289 }
290 
291 /*
292  * Close a socket on last file table reference removal.
293  * Initiate disconnect if connected.
294  * Free socket when disconnect complete.
295  *
296  * This function will sorele() the socket.  Note that soclose() may be
297  * called prior to the ref count reaching zero.  The actual socket
298  * structure will not be freed until the ref count reaches zero.
299  */
300 int
301 soclose(so)
302 	register struct socket *so;
303 {
304 	int s = splnet();		/* conservative */
305 	int error = 0;
306 
307 	funsetown(&so->so_sigio);
308 	if (so->so_options & SO_ACCEPTCONN) {
309 		struct socket *sp, *sonext;
310 
311 		sp = TAILQ_FIRST(&so->so_incomp);
312 		for (; sp != NULL; sp = sonext) {
313 			sonext = TAILQ_NEXT(sp, so_list);
314 			(void) soabort(sp);
315 		}
316 		for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
317 			sonext = TAILQ_NEXT(sp, so_list);
318 			/* Dequeue from so_comp since sofree() won't do it */
319 			TAILQ_REMOVE(&so->so_comp, sp, so_list);
320 			so->so_qlen--;
321 			sp->so_state &= ~SS_COMP;
322 			sp->so_head = NULL;
323 			(void) soabort(sp);
324 		}
325 	}
326 	if (so->so_pcb == 0)
327 		goto discard;
328 	if (so->so_state & SS_ISCONNECTED) {
329 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
330 			error = sodisconnect(so);
331 			if (error)
332 				goto drop;
333 		}
334 		if (so->so_options & SO_LINGER) {
335 			if ((so->so_state & SS_ISDISCONNECTING) &&
336 			    (so->so_state & SS_NBIO))
337 				goto drop;
338 			while (so->so_state & SS_ISCONNECTED) {
339 				error = tsleep((caddr_t)&so->so_timeo,
340 				    PSOCK | PCATCH, "soclos", so->so_linger * hz);
341 				if (error)
342 					break;
343 			}
344 		}
345 	}
346 drop:
347 	if (so->so_pcb) {
348 		int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so);
349 		if (error == 0)
350 			error = error2;
351 	}
352 discard:
353 	if (so->so_state & SS_NOFDREF)
354 		panic("soclose: NOFDREF");
355 	so->so_state |= SS_NOFDREF;
356 	sorele(so);
357 	splx(s);
358 	return (error);
359 }
360 
361 /*
362  * Must be called at splnet...
363  */
364 int
365 soabort(so)
366 	struct socket *so;
367 {
368 	int error;
369 
370 	error = (*so->so_proto->pr_usrreqs->pru_abort)(so);
371 	if (error) {
372 		sotryfree(so);	/* note: does not decrement the ref count */
373 		return error;
374 	}
375 	return (0);
376 }
377 
378 int
379 soaccept(so, nam)
380 	register struct socket *so;
381 	struct sockaddr **nam;
382 {
383 	int s = splnet();
384 	int error;
385 
386 	if ((so->so_state & SS_NOFDREF) == 0)
387 		panic("soaccept: !NOFDREF");
388 	so->so_state &= ~SS_NOFDREF;
389 	error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam);
390 	splx(s);
391 	return (error);
392 }
393 
394 int
395 soconnect(so, nam, td)
396 	register struct socket *so;
397 	struct sockaddr *nam;
398 	struct thread *td;
399 {
400 	int s;
401 	int error;
402 
403 	if (so->so_options & SO_ACCEPTCONN)
404 		return (EOPNOTSUPP);
405 	s = splnet();
406 	/*
407 	 * If protocol is connection-based, can only connect once.
408 	 * Otherwise, if connected, try to disconnect first.
409 	 * This allows user to disconnect by connecting to, e.g.,
410 	 * a null address.
411 	 */
412 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
413 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
414 	    (error = sodisconnect(so))))
415 		error = EISCONN;
416 	else
417 		error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, td);
418 	splx(s);
419 	return (error);
420 }
421 
422 int
423 soconnect2(so1, so2)
424 	register struct socket *so1;
425 	struct socket *so2;
426 {
427 	int s = splnet();
428 	int error;
429 
430 	error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2);
431 	splx(s);
432 	return (error);
433 }
434 
435 int
436 sodisconnect(so)
437 	register struct socket *so;
438 {
439 	int s = splnet();
440 	int error;
441 
442 	if ((so->so_state & SS_ISCONNECTED) == 0) {
443 		error = ENOTCONN;
444 		goto bad;
445 	}
446 	if (so->so_state & SS_ISDISCONNECTING) {
447 		error = EALREADY;
448 		goto bad;
449 	}
450 	error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so);
451 bad:
452 	splx(s);
453 	return (error);
454 }
455 
456 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
457 /*
458  * Send on a socket.
459  * If send must go all at once and message is larger than
460  * send buffering, then hard error.
461  * Lock against other senders.
462  * If must go all at once and not enough room now, then
463  * inform user that this would block and do nothing.
464  * Otherwise, if nonblocking, send as much as possible.
465  * The data to be sent is described by "uio" if nonzero,
466  * otherwise by the mbuf chain "top" (which must be null
467  * if uio is not).  Data provided in mbuf chain must be small
468  * enough to send all at once.
469  *
470  * Returns nonzero on error, timeout or signal; callers
471  * must check for short counts if EINTR/ERESTART are returned.
472  * Data and control buffers are freed on return.
473  */
474 int
475 sosend(so, addr, uio, top, control, flags, td)
476 	register struct socket *so;
477 	struct sockaddr *addr;
478 	struct uio *uio;
479 	struct mbuf *top;
480 	struct mbuf *control;
481 	int flags;
482 	struct thread *td;
483 {
484 	struct mbuf **mp;
485 	register struct mbuf *m;
486 	register long space, len, resid;
487 	int clen = 0, error, s, dontroute, mlen;
488 	int atomic = sosendallatonce(so) || top;
489 
490 	if (uio)
491 		resid = uio->uio_resid;
492 	else
493 		resid = top->m_pkthdr.len;
494 	/*
495 	 * In theory resid should be unsigned.
496 	 * However, space must be signed, as it might be less than 0
497 	 * if we over-committed, and we must use a signed comparison
498 	 * of space and resid.  On the other hand, a negative resid
499 	 * causes us to loop sending 0-length segments to the protocol.
500 	 *
501 	 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
502 	 * type sockets since that's an error.
503 	 */
504 	if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
505 		error = EINVAL;
506 		goto out;
507 	}
508 
509 	dontroute =
510 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
511 	    (so->so_proto->pr_flags & PR_ATOMIC);
512 	if (td)
513 		td->td_proc->p_stats->p_ru.ru_msgsnd++;
514 	if (control)
515 		clen = control->m_len;
516 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
517 
518 restart:
519 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
520 	if (error)
521 		goto out;
522 	do {
523 		s = splnet();
524 		if (so->so_state & SS_CANTSENDMORE)
525 			snderr(EPIPE);
526 		if (so->so_error) {
527 			error = so->so_error;
528 			so->so_error = 0;
529 			splx(s);
530 			goto release;
531 		}
532 		if ((so->so_state & SS_ISCONNECTED) == 0) {
533 			/*
534 			 * `sendto' and `sendmsg' is allowed on a connection-
535 			 * based socket if it supports implied connect.
536 			 * Return ENOTCONN if not connected and no address is
537 			 * supplied.
538 			 */
539 			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
540 			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
541 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
542 				    !(resid == 0 && clen != 0))
543 					snderr(ENOTCONN);
544 			} else if (addr == 0)
545 			    snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
546 				   ENOTCONN : EDESTADDRREQ);
547 		}
548 		space = sbspace(&so->so_snd);
549 		if (flags & MSG_OOB)
550 			space += 1024;
551 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
552 		    clen > so->so_snd.sb_hiwat)
553 			snderr(EMSGSIZE);
554 		if (space < resid + clen &&
555 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
556 			if (so->so_state & SS_NBIO)
557 				snderr(EWOULDBLOCK);
558 			sbunlock(&so->so_snd);
559 			error = sbwait(&so->so_snd);
560 			splx(s);
561 			if (error)
562 				goto out;
563 			goto restart;
564 		}
565 		splx(s);
566 		mp = &top;
567 		space -= clen;
568 		do {
569 		    if (uio == NULL) {
570 			/*
571 			 * Data is prepackaged in "top".
572 			 */
573 			resid = 0;
574 			if (flags & MSG_EOR)
575 				top->m_flags |= M_EOR;
576 		    } else do {
577 			if (top == 0) {
578 				MGETHDR(m, M_TRYWAIT, MT_DATA);
579 				if (m == NULL) {
580 					error = ENOBUFS;
581 					goto release;
582 				}
583 				mlen = MHLEN;
584 				m->m_pkthdr.len = 0;
585 				m->m_pkthdr.rcvif = (struct ifnet *)0;
586 			} else {
587 				MGET(m, M_TRYWAIT, MT_DATA);
588 				if (m == NULL) {
589 					error = ENOBUFS;
590 					goto release;
591 				}
592 				mlen = MLEN;
593 			}
594 			if (resid >= MINCLSIZE) {
595 				MCLGET(m, M_TRYWAIT);
596 				if ((m->m_flags & M_EXT) == 0)
597 					goto nopages;
598 				mlen = MCLBYTES;
599 				len = min(min(mlen, resid), space);
600 			} else {
601 nopages:
602 				len = min(min(mlen, resid), space);
603 				/*
604 				 * For datagram protocols, leave room
605 				 * for protocol headers in first mbuf.
606 				 */
607 				if (atomic && top == 0 && len < mlen)
608 					MH_ALIGN(m, len);
609 			}
610 			space -= len;
611 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
612 			resid = uio->uio_resid;
613 			m->m_len = len;
614 			*mp = m;
615 			top->m_pkthdr.len += len;
616 			if (error)
617 				goto release;
618 			mp = &m->m_next;
619 			if (resid <= 0) {
620 				if (flags & MSG_EOR)
621 					top->m_flags |= M_EOR;
622 				break;
623 			}
624 		    } while (space > 0 && atomic);
625 		    if (dontroute)
626 			    so->so_options |= SO_DONTROUTE;
627 		    s = splnet();				/* XXX */
628 		    /*
629 		     * XXX all the SS_CANTSENDMORE checks previously
630 		     * done could be out of date.  We could have recieved
631 		     * a reset packet in an interrupt or maybe we slept
632 		     * while doing page faults in uiomove() etc. We could
633 		     * probably recheck again inside the splnet() protection
634 		     * here, but there are probably other places that this
635 		     * also happens.  We must rethink this.
636 		     */
637 		    error = (*so->so_proto->pr_usrreqs->pru_send)(so,
638 			(flags & MSG_OOB) ? PRUS_OOB :
639 			/*
640 			 * If the user set MSG_EOF, the protocol
641 			 * understands this flag and nothing left to
642 			 * send then use PRU_SEND_EOF instead of PRU_SEND.
643 			 */
644 			((flags & MSG_EOF) &&
645 			 (so->so_proto->pr_flags & PR_IMPLOPCL) &&
646 			 (resid <= 0)) ?
647 				PRUS_EOF :
648 			/* If there is more to send set PRUS_MORETOCOME */
649 			(resid > 0 && space > 0) ? PRUS_MORETOCOME : 0,
650 			top, addr, control, td);
651 		    splx(s);
652 		    if (dontroute)
653 			    so->so_options &= ~SO_DONTROUTE;
654 		    clen = 0;
655 		    control = 0;
656 		    top = 0;
657 		    mp = &top;
658 		    if (error)
659 			goto release;
660 		} while (resid && space > 0);
661 	} while (resid);
662 
663 release:
664 	sbunlock(&so->so_snd);
665 out:
666 	if (top)
667 		m_freem(top);
668 	if (control)
669 		m_freem(control);
670 	return (error);
671 }
672 
673 /*
674  * Implement receive operations on a socket.
675  * We depend on the way that records are added to the sockbuf
676  * by sbappend*.  In particular, each record (mbufs linked through m_next)
677  * must begin with an address if the protocol so specifies,
678  * followed by an optional mbuf or mbufs containing ancillary data,
679  * and then zero or more mbufs of data.
680  * In order to avoid blocking network interrupts for the entire time here,
681  * we splx() while doing the actual copy to user space.
682  * Although the sockbuf is locked, new data may still be appended,
683  * and thus we must maintain consistency of the sockbuf during that time.
684  *
685  * The caller may receive the data as a single mbuf chain by supplying
686  * an mbuf **mp0 for use in returning the chain.  The uio is then used
687  * only for the count in uio_resid.
688  */
689 int
690 soreceive(so, psa, uio, mp0, controlp, flagsp)
691 	register struct socket *so;
692 	struct sockaddr **psa;
693 	struct uio *uio;
694 	struct mbuf **mp0;
695 	struct mbuf **controlp;
696 	int *flagsp;
697 {
698 	struct mbuf *m, **mp;
699 	register int flags, len, error, s, offset;
700 	struct protosw *pr = so->so_proto;
701 	struct mbuf *nextrecord;
702 	int moff, type = 0;
703 	int orig_resid = uio->uio_resid;
704 
705 	mp = mp0;
706 	if (psa)
707 		*psa = 0;
708 	if (controlp)
709 		*controlp = 0;
710 	if (flagsp)
711 		flags = *flagsp &~ MSG_EOR;
712 	else
713 		flags = 0;
714 	if (flags & MSG_OOB) {
715 		m = m_get(M_TRYWAIT, MT_DATA);
716 		if (m == NULL)
717 			return (ENOBUFS);
718 		error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK);
719 		if (error)
720 			goto bad;
721 		do {
722 			error = uiomove(mtod(m, caddr_t),
723 			    (int) min(uio->uio_resid, m->m_len), uio);
724 			m = m_free(m);
725 		} while (uio->uio_resid && error == 0 && m);
726 bad:
727 		if (m)
728 			m_freem(m);
729 		return (error);
730 	}
731 	if (mp)
732 		*mp = (struct mbuf *)0;
733 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
734 		(*pr->pr_usrreqs->pru_rcvd)(so, 0);
735 
736 restart:
737 	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
738 	if (error)
739 		return (error);
740 	s = splnet();
741 
742 	m = so->so_rcv.sb_mb;
743 	/*
744 	 * If we have less data than requested, block awaiting more
745 	 * (subject to any timeout) if:
746 	 *   1. the current count is less than the low water mark, or
747 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
748 	 *	receive operation at once if we block (resid <= hiwat).
749 	 *   3. MSG_DONTWAIT is not set
750 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
751 	 * we have to do the receive in sections, and thus risk returning
752 	 * a short count if a timeout or signal occurs after we start.
753 	 */
754 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
755 	    so->so_rcv.sb_cc < uio->uio_resid) &&
756 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
757 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
758 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
759 		KASSERT(m != 0 || !so->so_rcv.sb_cc,
760 		    ("receive: m == %p so->so_rcv.sb_cc == %lu",
761 		    m, so->so_rcv.sb_cc));
762 		if (so->so_error) {
763 			if (m)
764 				goto dontblock;
765 			error = so->so_error;
766 			if ((flags & MSG_PEEK) == 0)
767 				so->so_error = 0;
768 			goto release;
769 		}
770 		if (so->so_state & SS_CANTRCVMORE) {
771 			if (m)
772 				goto dontblock;
773 			else
774 				goto release;
775 		}
776 		for (; m; m = m->m_next)
777 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
778 				m = so->so_rcv.sb_mb;
779 				goto dontblock;
780 			}
781 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
782 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
783 			error = ENOTCONN;
784 			goto release;
785 		}
786 		if (uio->uio_resid == 0)
787 			goto release;
788 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
789 			error = EWOULDBLOCK;
790 			goto release;
791 		}
792 		sbunlock(&so->so_rcv);
793 		error = sbwait(&so->so_rcv);
794 		splx(s);
795 		if (error)
796 			return (error);
797 		goto restart;
798 	}
799 dontblock:
800 	if (uio->uio_td)
801 		uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++;
802 	nextrecord = m->m_nextpkt;
803 	if (pr->pr_flags & PR_ADDR) {
804 		KASSERT(m->m_type == MT_SONAME,
805 		    ("m->m_type == %d", m->m_type));
806 		orig_resid = 0;
807 		if (psa)
808 			*psa = dup_sockaddr(mtod(m, struct sockaddr *),
809 					    mp0 == 0);
810 		if (flags & MSG_PEEK) {
811 			m = m->m_next;
812 		} else {
813 			sbfree(&so->so_rcv, m);
814 			so->so_rcv.sb_mb = m_free(m);
815 			m = so->so_rcv.sb_mb;
816 		}
817 	}
818 	while (m && m->m_type == MT_CONTROL && error == 0) {
819 		if (flags & MSG_PEEK) {
820 			if (controlp)
821 				*controlp = m_copy(m, 0, m->m_len);
822 			m = m->m_next;
823 		} else {
824 			sbfree(&so->so_rcv, m);
825 			so->so_rcv.sb_mb = m->m_next;
826 			m->m_next = NULL;
827 			if (pr->pr_domain->dom_externalize)
828 				error =
829 				(*pr->pr_domain->dom_externalize)(m, controlp);
830 			else if (controlp)
831 				*controlp = m;
832 			else
833 				m_freem(m);
834 			m = so->so_rcv.sb_mb;
835 		}
836 		if (controlp) {
837 			orig_resid = 0;
838 			do
839 				controlp = &(*controlp)->m_next;
840 			while (*controlp != NULL);
841 		}
842 	}
843 	if (m) {
844 		if ((flags & MSG_PEEK) == 0)
845 			m->m_nextpkt = nextrecord;
846 		type = m->m_type;
847 		if (type == MT_OOBDATA)
848 			flags |= MSG_OOB;
849 	}
850 	moff = 0;
851 	offset = 0;
852 	while (m && uio->uio_resid > 0 && error == 0) {
853 		if (m->m_type == MT_OOBDATA) {
854 			if (type != MT_OOBDATA)
855 				break;
856 		} else if (type == MT_OOBDATA)
857 			break;
858 		else
859 		    KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
860 			("m->m_type == %d", m->m_type));
861 		so->so_state &= ~SS_RCVATMARK;
862 		len = uio->uio_resid;
863 		if (so->so_oobmark && len > so->so_oobmark - offset)
864 			len = so->so_oobmark - offset;
865 		if (len > m->m_len - moff)
866 			len = m->m_len - moff;
867 		/*
868 		 * If mp is set, just pass back the mbufs.
869 		 * Otherwise copy them out via the uio, then free.
870 		 * Sockbuf must be consistent here (points to current mbuf,
871 		 * it points to next record) when we drop priority;
872 		 * we must note any additions to the sockbuf when we
873 		 * block interrupts again.
874 		 */
875 		if (mp == 0) {
876 			splx(s);
877 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
878 			s = splnet();
879 			if (error)
880 				goto release;
881 		} else
882 			uio->uio_resid -= len;
883 		if (len == m->m_len - moff) {
884 			if (m->m_flags & M_EOR)
885 				flags |= MSG_EOR;
886 			if (flags & MSG_PEEK) {
887 				m = m->m_next;
888 				moff = 0;
889 			} else {
890 				nextrecord = m->m_nextpkt;
891 				sbfree(&so->so_rcv, m);
892 				if (mp) {
893 					*mp = m;
894 					mp = &m->m_next;
895 					so->so_rcv.sb_mb = m = m->m_next;
896 					*mp = (struct mbuf *)0;
897 				} else {
898 					so->so_rcv.sb_mb = m_free(m);
899 					m = so->so_rcv.sb_mb;
900 				}
901 				if (m)
902 					m->m_nextpkt = nextrecord;
903 			}
904 		} else {
905 			if (flags & MSG_PEEK)
906 				moff += len;
907 			else {
908 				if (mp)
909 					*mp = m_copym(m, 0, len, M_TRYWAIT);
910 				m->m_data += len;
911 				m->m_len -= len;
912 				so->so_rcv.sb_cc -= len;
913 			}
914 		}
915 		if (so->so_oobmark) {
916 			if ((flags & MSG_PEEK) == 0) {
917 				so->so_oobmark -= len;
918 				if (so->so_oobmark == 0) {
919 					so->so_state |= SS_RCVATMARK;
920 					break;
921 				}
922 			} else {
923 				offset += len;
924 				if (offset == so->so_oobmark)
925 					break;
926 			}
927 		}
928 		if (flags & MSG_EOR)
929 			break;
930 		/*
931 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
932 		 * we must not quit until "uio->uio_resid == 0" or an error
933 		 * termination.  If a signal/timeout occurs, return
934 		 * with a short count but without error.
935 		 * Keep sockbuf locked against other readers.
936 		 */
937 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
938 		    !sosendallatonce(so) && !nextrecord) {
939 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
940 				break;
941 			/*
942 			 * Notify the protocol that some data has been
943 			 * drained before blocking.
944 			 */
945 			if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
946 				(*pr->pr_usrreqs->pru_rcvd)(so, flags);
947 			error = sbwait(&so->so_rcv);
948 			if (error) {
949 				sbunlock(&so->so_rcv);
950 				splx(s);
951 				return (0);
952 			}
953 			m = so->so_rcv.sb_mb;
954 			if (m)
955 				nextrecord = m->m_nextpkt;
956 		}
957 	}
958 
959 	if (m && pr->pr_flags & PR_ATOMIC) {
960 		flags |= MSG_TRUNC;
961 		if ((flags & MSG_PEEK) == 0)
962 			(void) sbdroprecord(&so->so_rcv);
963 	}
964 	if ((flags & MSG_PEEK) == 0) {
965 		if (m == 0)
966 			so->so_rcv.sb_mb = nextrecord;
967 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
968 			(*pr->pr_usrreqs->pru_rcvd)(so, flags);
969 	}
970 	if (orig_resid == uio->uio_resid && orig_resid &&
971 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
972 		sbunlock(&so->so_rcv);
973 		splx(s);
974 		goto restart;
975 	}
976 
977 	if (flagsp)
978 		*flagsp |= flags;
979 release:
980 	sbunlock(&so->so_rcv);
981 	splx(s);
982 	return (error);
983 }
984 
985 int
986 soshutdown(so, how)
987 	register struct socket *so;
988 	register int how;
989 {
990 	register struct protosw *pr = so->so_proto;
991 
992 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
993 		return (EINVAL);
994 
995 	if (how != SHUT_WR)
996 		sorflush(so);
997 	if (how != SHUT_RD)
998 		return ((*pr->pr_usrreqs->pru_shutdown)(so));
999 	return (0);
1000 }
1001 
1002 void
1003 sorflush(so)
1004 	register struct socket *so;
1005 {
1006 	register struct sockbuf *sb = &so->so_rcv;
1007 	register struct protosw *pr = so->so_proto;
1008 	register int s;
1009 	struct sockbuf asb;
1010 
1011 	sb->sb_flags |= SB_NOINTR;
1012 	(void) sblock(sb, M_WAITOK);
1013 	s = splimp();
1014 	socantrcvmore(so);
1015 	sbunlock(sb);
1016 	asb = *sb;
1017 	bzero((caddr_t)sb, sizeof (*sb));
1018 	splx(s);
1019 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1020 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
1021 	sbrelease(&asb, so);
1022 }
1023 
1024 #ifdef INET
1025 static int
1026 do_setopt_accept_filter(so, sopt)
1027 	struct	socket *so;
1028 	struct	sockopt *sopt;
1029 {
1030 	struct accept_filter_arg	*afap = NULL;
1031 	struct accept_filter	*afp;
1032 	struct so_accf	*af = so->so_accf;
1033 	int	error = 0;
1034 
1035 	/* do not set/remove accept filters on non listen sockets */
1036 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
1037 		error = EINVAL;
1038 		goto out;
1039 	}
1040 
1041 	/* removing the filter */
1042 	if (sopt == NULL) {
1043 		if (af != NULL) {
1044 			if (af->so_accept_filter != NULL &&
1045 				af->so_accept_filter->accf_destroy != NULL) {
1046 				af->so_accept_filter->accf_destroy(so);
1047 			}
1048 			if (af->so_accept_filter_str != NULL) {
1049 				FREE(af->so_accept_filter_str, M_ACCF);
1050 			}
1051 			FREE(af, M_ACCF);
1052 			so->so_accf = NULL;
1053 		}
1054 		so->so_options &= ~SO_ACCEPTFILTER;
1055 		return (0);
1056 	}
1057 	/* adding a filter */
1058 	/* must remove previous filter first */
1059 	if (af != NULL) {
1060 		error = EINVAL;
1061 		goto out;
1062 	}
1063 	/* don't put large objects on the kernel stack */
1064 	MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1065 	error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1066 	afap->af_name[sizeof(afap->af_name)-1] = '\0';
1067 	afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1068 	if (error)
1069 		goto out;
1070 	afp = accept_filt_get(afap->af_name);
1071 	if (afp == NULL) {
1072 		error = ENOENT;
1073 		goto out;
1074 	}
1075 	MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO);
1076 	if (afp->accf_create != NULL) {
1077 		if (afap->af_name[0] != '\0') {
1078 			int len = strlen(afap->af_name) + 1;
1079 
1080 			MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1081 			strcpy(af->so_accept_filter_str, afap->af_name);
1082 		}
1083 		af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1084 		if (af->so_accept_filter_arg == NULL) {
1085 			FREE(af->so_accept_filter_str, M_ACCF);
1086 			FREE(af, M_ACCF);
1087 			so->so_accf = NULL;
1088 			error = EINVAL;
1089 			goto out;
1090 		}
1091 	}
1092 	af->so_accept_filter = afp;
1093 	so->so_accf = af;
1094 	so->so_options |= SO_ACCEPTFILTER;
1095 out:
1096 	if (afap != NULL)
1097 		FREE(afap, M_TEMP);
1098 	return (error);
1099 }
1100 #endif /* INET */
1101 
1102 /*
1103  * Perhaps this routine, and sooptcopyout(), below, ought to come in
1104  * an additional variant to handle the case where the option value needs
1105  * to be some kind of integer, but not a specific size.
1106  * In addition to their use here, these functions are also called by the
1107  * protocol-level pr_ctloutput() routines.
1108  */
1109 int
1110 sooptcopyin(sopt, buf, len, minlen)
1111 	struct	sockopt *sopt;
1112 	void	*buf;
1113 	size_t	len;
1114 	size_t	minlen;
1115 {
1116 	size_t	valsize;
1117 
1118 	/*
1119 	 * If the user gives us more than we wanted, we ignore it,
1120 	 * but if we don't get the minimum length the caller
1121 	 * wants, we return EINVAL.  On success, sopt->sopt_valsize
1122 	 * is set to however much we actually retrieved.
1123 	 */
1124 	if ((valsize = sopt->sopt_valsize) < minlen)
1125 		return EINVAL;
1126 	if (valsize > len)
1127 		sopt->sopt_valsize = valsize = len;
1128 
1129 	if (sopt->sopt_td != 0)
1130 		return (copyin(sopt->sopt_val, buf, valsize));
1131 
1132 	bcopy(sopt->sopt_val, buf, valsize);
1133 	return 0;
1134 }
1135 
1136 int
1137 sosetopt(so, sopt)
1138 	struct socket *so;
1139 	struct sockopt *sopt;
1140 {
1141 	int	error, optval;
1142 	struct	linger l;
1143 	struct	timeval tv;
1144 	u_long  val;
1145 
1146 	error = 0;
1147 	if (sopt->sopt_level != SOL_SOCKET) {
1148 		if (so->so_proto && so->so_proto->pr_ctloutput)
1149 			return ((*so->so_proto->pr_ctloutput)
1150 				  (so, sopt));
1151 		error = ENOPROTOOPT;
1152 	} else {
1153 		switch (sopt->sopt_name) {
1154 #ifdef INET
1155 		case SO_ACCEPTFILTER:
1156 			error = do_setopt_accept_filter(so, sopt);
1157 			if (error)
1158 				goto bad;
1159 			break;
1160 #endif
1161 		case SO_LINGER:
1162 			error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1163 			if (error)
1164 				goto bad;
1165 
1166 			so->so_linger = l.l_linger;
1167 			if (l.l_onoff)
1168 				so->so_options |= SO_LINGER;
1169 			else
1170 				so->so_options &= ~SO_LINGER;
1171 			break;
1172 
1173 		case SO_DEBUG:
1174 		case SO_KEEPALIVE:
1175 		case SO_DONTROUTE:
1176 		case SO_USELOOPBACK:
1177 		case SO_BROADCAST:
1178 		case SO_REUSEADDR:
1179 		case SO_REUSEPORT:
1180 		case SO_OOBINLINE:
1181 		case SO_TIMESTAMP:
1182 			error = sooptcopyin(sopt, &optval, sizeof optval,
1183 					    sizeof optval);
1184 			if (error)
1185 				goto bad;
1186 			if (optval)
1187 				so->so_options |= sopt->sopt_name;
1188 			else
1189 				so->so_options &= ~sopt->sopt_name;
1190 			break;
1191 
1192 		case SO_SNDBUF:
1193 		case SO_RCVBUF:
1194 		case SO_SNDLOWAT:
1195 		case SO_RCVLOWAT:
1196 			error = sooptcopyin(sopt, &optval, sizeof optval,
1197 					    sizeof optval);
1198 			if (error)
1199 				goto bad;
1200 
1201 			/*
1202 			 * Values < 1 make no sense for any of these
1203 			 * options, so disallow them.
1204 			 */
1205 			if (optval < 1) {
1206 				error = EINVAL;
1207 				goto bad;
1208 			}
1209 
1210 			switch (sopt->sopt_name) {
1211 			case SO_SNDBUF:
1212 			case SO_RCVBUF:
1213 				if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1214 				    &so->so_snd : &so->so_rcv, (u_long)optval,
1215 				    so, curthread) == 0) {
1216 					error = ENOBUFS;
1217 					goto bad;
1218 				}
1219 				break;
1220 
1221 			/*
1222 			 * Make sure the low-water is never greater than
1223 			 * the high-water.
1224 			 */
1225 			case SO_SNDLOWAT:
1226 				so->so_snd.sb_lowat =
1227 				    (optval > so->so_snd.sb_hiwat) ?
1228 				    so->so_snd.sb_hiwat : optval;
1229 				break;
1230 			case SO_RCVLOWAT:
1231 				so->so_rcv.sb_lowat =
1232 				    (optval > so->so_rcv.sb_hiwat) ?
1233 				    so->so_rcv.sb_hiwat : optval;
1234 				break;
1235 			}
1236 			break;
1237 
1238 		case SO_SNDTIMEO:
1239 		case SO_RCVTIMEO:
1240 			error = sooptcopyin(sopt, &tv, sizeof tv,
1241 					    sizeof tv);
1242 			if (error)
1243 				goto bad;
1244 
1245 			/* assert(hz > 0); */
1246 			if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1247 			    tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1248 				error = EDOM;
1249 				goto bad;
1250 			}
1251 			/* assert(tick > 0); */
1252 			/* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1253 			val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1254 			if (val > SHRT_MAX) {
1255 				error = EDOM;
1256 				goto bad;
1257 			}
1258 
1259 			switch (sopt->sopt_name) {
1260 			case SO_SNDTIMEO:
1261 				so->so_snd.sb_timeo = val;
1262 				break;
1263 			case SO_RCVTIMEO:
1264 				so->so_rcv.sb_timeo = val;
1265 				break;
1266 			}
1267 			break;
1268 		default:
1269 			error = ENOPROTOOPT;
1270 			break;
1271 		}
1272 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1273 			(void) ((*so->so_proto->pr_ctloutput)
1274 				  (so, sopt));
1275 		}
1276 	}
1277 bad:
1278 	return (error);
1279 }
1280 
1281 /* Helper routine for getsockopt */
1282 int
1283 sooptcopyout(sopt, buf, len)
1284 	struct	sockopt *sopt;
1285 	void	*buf;
1286 	size_t	len;
1287 {
1288 	int	error;
1289 	size_t	valsize;
1290 
1291 	error = 0;
1292 
1293 	/*
1294 	 * Documented get behavior is that we always return a value,
1295 	 * possibly truncated to fit in the user's buffer.
1296 	 * Traditional behavior is that we always tell the user
1297 	 * precisely how much we copied, rather than something useful
1298 	 * like the total amount we had available for her.
1299 	 * Note that this interface is not idempotent; the entire answer must
1300 	 * generated ahead of time.
1301 	 */
1302 	valsize = min(len, sopt->sopt_valsize);
1303 	sopt->sopt_valsize = valsize;
1304 	if (sopt->sopt_val != 0) {
1305 		if (sopt->sopt_td != 0)
1306 			error = copyout(buf, sopt->sopt_val, valsize);
1307 		else
1308 			bcopy(buf, sopt->sopt_val, valsize);
1309 	}
1310 	return error;
1311 }
1312 
1313 int
1314 sogetopt(so, sopt)
1315 	struct socket *so;
1316 	struct sockopt *sopt;
1317 {
1318 	int	error, optval;
1319 	struct	linger l;
1320 	struct	timeval tv;
1321 #ifdef INET
1322 	struct accept_filter_arg *afap;
1323 #endif
1324 
1325 	error = 0;
1326 	if (sopt->sopt_level != SOL_SOCKET) {
1327 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1328 			return ((*so->so_proto->pr_ctloutput)
1329 				  (so, sopt));
1330 		} else
1331 			return (ENOPROTOOPT);
1332 	} else {
1333 		switch (sopt->sopt_name) {
1334 #ifdef INET
1335 		case SO_ACCEPTFILTER:
1336 			if ((so->so_options & SO_ACCEPTCONN) == 0)
1337 				return (EINVAL);
1338 			MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1339 				M_TEMP, M_WAITOK | M_ZERO);
1340 			if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1341 				strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1342 				if (so->so_accf->so_accept_filter_str != NULL)
1343 					strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1344 			}
1345 			error = sooptcopyout(sopt, afap, sizeof(*afap));
1346 			FREE(afap, M_TEMP);
1347 			break;
1348 #endif
1349 
1350 		case SO_LINGER:
1351 			l.l_onoff = so->so_options & SO_LINGER;
1352 			l.l_linger = so->so_linger;
1353 			error = sooptcopyout(sopt, &l, sizeof l);
1354 			break;
1355 
1356 		case SO_USELOOPBACK:
1357 		case SO_DONTROUTE:
1358 		case SO_DEBUG:
1359 		case SO_KEEPALIVE:
1360 		case SO_REUSEADDR:
1361 		case SO_REUSEPORT:
1362 		case SO_BROADCAST:
1363 		case SO_OOBINLINE:
1364 		case SO_TIMESTAMP:
1365 			optval = so->so_options & sopt->sopt_name;
1366 integer:
1367 			error = sooptcopyout(sopt, &optval, sizeof optval);
1368 			break;
1369 
1370 		case SO_TYPE:
1371 			optval = so->so_type;
1372 			goto integer;
1373 
1374 		case SO_ERROR:
1375 			optval = so->so_error;
1376 			so->so_error = 0;
1377 			goto integer;
1378 
1379 		case SO_SNDBUF:
1380 			optval = so->so_snd.sb_hiwat;
1381 			goto integer;
1382 
1383 		case SO_RCVBUF:
1384 			optval = so->so_rcv.sb_hiwat;
1385 			goto integer;
1386 
1387 		case SO_SNDLOWAT:
1388 			optval = so->so_snd.sb_lowat;
1389 			goto integer;
1390 
1391 		case SO_RCVLOWAT:
1392 			optval = so->so_rcv.sb_lowat;
1393 			goto integer;
1394 
1395 		case SO_SNDTIMEO:
1396 		case SO_RCVTIMEO:
1397 			optval = (sopt->sopt_name == SO_SNDTIMEO ?
1398 				  so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1399 
1400 			tv.tv_sec = optval / hz;
1401 			tv.tv_usec = (optval % hz) * tick;
1402 			error = sooptcopyout(sopt, &tv, sizeof tv);
1403 			break;
1404 
1405 		default:
1406 			error = ENOPROTOOPT;
1407 			break;
1408 		}
1409 		return (error);
1410 	}
1411 }
1412 
1413 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1414 int
1415 soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1416 {
1417 	struct mbuf *m, *m_prev;
1418 	int sopt_size = sopt->sopt_valsize;
1419 
1420 	MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1421 	if (m == 0)
1422 		return ENOBUFS;
1423 	if (sopt_size > MLEN) {
1424 		MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT);
1425 		if ((m->m_flags & M_EXT) == 0) {
1426 			m_free(m);
1427 			return ENOBUFS;
1428 		}
1429 		m->m_len = min(MCLBYTES, sopt_size);
1430 	} else {
1431 		m->m_len = min(MLEN, sopt_size);
1432 	}
1433 	sopt_size -= m->m_len;
1434 	*mp = m;
1435 	m_prev = m;
1436 
1437 	while (sopt_size) {
1438 		MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1439 		if (m == 0) {
1440 			m_freem(*mp);
1441 			return ENOBUFS;
1442 		}
1443 		if (sopt_size > MLEN) {
1444 			MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT);
1445 			if ((m->m_flags & M_EXT) == 0) {
1446 				m_freem(*mp);
1447 				return ENOBUFS;
1448 			}
1449 			m->m_len = min(MCLBYTES, sopt_size);
1450 		} else {
1451 			m->m_len = min(MLEN, sopt_size);
1452 		}
1453 		sopt_size -= m->m_len;
1454 		m_prev->m_next = m;
1455 		m_prev = m;
1456 	}
1457 	return 0;
1458 }
1459 
1460 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1461 int
1462 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1463 {
1464 	struct mbuf *m0 = m;
1465 
1466 	if (sopt->sopt_val == NULL)
1467 		return 0;
1468 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1469 		if (sopt->sopt_td != NULL) {
1470 			int error;
1471 
1472 			error = copyin(sopt->sopt_val, mtod(m, char *),
1473 				       m->m_len);
1474 			if (error != 0) {
1475 				m_freem(m0);
1476 				return(error);
1477 			}
1478 		} else
1479 			bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1480 		sopt->sopt_valsize -= m->m_len;
1481 		(caddr_t)sopt->sopt_val += m->m_len;
1482 		m = m->m_next;
1483 	}
1484 	if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1485 		panic("ip6_sooptmcopyin");
1486 	return 0;
1487 }
1488 
1489 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1490 int
1491 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1492 {
1493 	struct mbuf *m0 = m;
1494 	size_t valsize = 0;
1495 
1496 	if (sopt->sopt_val == NULL)
1497 		return 0;
1498 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1499 		if (sopt->sopt_td != NULL) {
1500 			int error;
1501 
1502 			error = copyout(mtod(m, char *), sopt->sopt_val,
1503 				       m->m_len);
1504 			if (error != 0) {
1505 				m_freem(m0);
1506 				return(error);
1507 			}
1508 		} else
1509 			bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1510 	       sopt->sopt_valsize -= m->m_len;
1511 	       (caddr_t)sopt->sopt_val += m->m_len;
1512 	       valsize += m->m_len;
1513 	       m = m->m_next;
1514 	}
1515 	if (m != NULL) {
1516 		/* enough soopt buffer should be given from user-land */
1517 		m_freem(m0);
1518 		return(EINVAL);
1519 	}
1520 	sopt->sopt_valsize = valsize;
1521 	return 0;
1522 }
1523 
1524 void
1525 sohasoutofband(so)
1526 	register struct socket *so;
1527 {
1528 	if (so->so_sigio != NULL)
1529 		pgsigio(&so->so_sigio, SIGURG, 0);
1530 	selwakeup(&so->so_rcv.sb_sel);
1531 }
1532 
1533 int
1534 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td)
1535 {
1536 	int revents = 0;
1537 	int s = splnet();
1538 
1539 	if (events & (POLLIN | POLLRDNORM))
1540 		if (soreadable(so))
1541 			revents |= events & (POLLIN | POLLRDNORM);
1542 
1543 	if (events & POLLINIGNEOF)
1544 		if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat ||
1545 		    !TAILQ_EMPTY(&so->so_comp) || so->so_error)
1546 			revents |= POLLINIGNEOF;
1547 
1548 	if (events & (POLLOUT | POLLWRNORM))
1549 		if (sowriteable(so))
1550 			revents |= events & (POLLOUT | POLLWRNORM);
1551 
1552 	if (events & (POLLPRI | POLLRDBAND))
1553 		if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1554 			revents |= events & (POLLPRI | POLLRDBAND);
1555 
1556 	if (revents == 0) {
1557 		if (events &
1558 		    (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM |
1559 		     POLLRDBAND)) {
1560 			selrecord(td, &so->so_rcv.sb_sel);
1561 			so->so_rcv.sb_flags |= SB_SEL;
1562 		}
1563 
1564 		if (events & (POLLOUT | POLLWRNORM)) {
1565 			selrecord(td, &so->so_snd.sb_sel);
1566 			so->so_snd.sb_flags |= SB_SEL;
1567 		}
1568 	}
1569 
1570 	splx(s);
1571 	return (revents);
1572 }
1573 
1574 int
1575 sokqfilter(struct file *fp, struct knote *kn)
1576 {
1577 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1578 	struct sockbuf *sb;
1579 	int s;
1580 
1581 	switch (kn->kn_filter) {
1582 	case EVFILT_READ:
1583 		if (so->so_options & SO_ACCEPTCONN)
1584 			kn->kn_fop = &solisten_filtops;
1585 		else
1586 			kn->kn_fop = &soread_filtops;
1587 		sb = &so->so_rcv;
1588 		break;
1589 	case EVFILT_WRITE:
1590 		kn->kn_fop = &sowrite_filtops;
1591 		sb = &so->so_snd;
1592 		break;
1593 	default:
1594 		return (1);
1595 	}
1596 
1597 	s = splnet();
1598 	SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1599 	sb->sb_flags |= SB_KNOTE;
1600 	splx(s);
1601 	return (0);
1602 }
1603 
1604 static void
1605 filt_sordetach(struct knote *kn)
1606 {
1607 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1608 	int s = splnet();
1609 
1610 	SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1611 	if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1612 		so->so_rcv.sb_flags &= ~SB_KNOTE;
1613 	splx(s);
1614 }
1615 
1616 /*ARGSUSED*/
1617 static int
1618 filt_soread(struct knote *kn, long hint)
1619 {
1620 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1621 
1622 	kn->kn_data = so->so_rcv.sb_cc;
1623 	if (so->so_state & SS_CANTRCVMORE) {
1624 		kn->kn_flags |= EV_EOF;
1625 		kn->kn_fflags = so->so_error;
1626 		return (1);
1627 	}
1628 	if (so->so_error)	/* temporary udp error */
1629 		return (1);
1630 	if (kn->kn_sfflags & NOTE_LOWAT)
1631 		return (kn->kn_data >= kn->kn_sdata);
1632 	return (kn->kn_data >= so->so_rcv.sb_lowat);
1633 }
1634 
1635 static void
1636 filt_sowdetach(struct knote *kn)
1637 {
1638 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1639 	int s = splnet();
1640 
1641 	SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1642 	if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1643 		so->so_snd.sb_flags &= ~SB_KNOTE;
1644 	splx(s);
1645 }
1646 
1647 /*ARGSUSED*/
1648 static int
1649 filt_sowrite(struct knote *kn, long hint)
1650 {
1651 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1652 
1653 	kn->kn_data = sbspace(&so->so_snd);
1654 	if (so->so_state & SS_CANTSENDMORE) {
1655 		kn->kn_flags |= EV_EOF;
1656 		kn->kn_fflags = so->so_error;
1657 		return (1);
1658 	}
1659 	if (so->so_error)	/* temporary udp error */
1660 		return (1);
1661 	if (((so->so_state & SS_ISCONNECTED) == 0) &&
1662 	    (so->so_proto->pr_flags & PR_CONNREQUIRED))
1663 		return (0);
1664 	if (kn->kn_sfflags & NOTE_LOWAT)
1665 		return (kn->kn_data >= kn->kn_sdata);
1666 	return (kn->kn_data >= so->so_snd.sb_lowat);
1667 }
1668 
1669 /*ARGSUSED*/
1670 static int
1671 filt_solisten(struct knote *kn, long hint)
1672 {
1673 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1674 
1675 	kn->kn_data = so->so_qlen;
1676 	return (! TAILQ_EMPTY(&so->so_comp));
1677 }
1678 
1679 int
1680 socheckuid(struct socket *so, uid_t uid)
1681 {
1682 
1683 	if (so == NULL)
1684 		return (EPERM);
1685 	if (so->so_cred->cr_uid == uid)
1686 		return (0);
1687 	return (EPERM);
1688 }
1689