xref: /dflybsd-src/sys/kern/uipc_socket.c (revision c5541aee854b0d32586182b733a9ea4d4c92168b)
1 /*
2  * Copyright (c) 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 1982, 1986, 1988, 1990, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. All advertising materials mentioning features or use of this software
15  *    must display the following acknowledgement:
16  *	This product includes software developed by the University of
17  *	California, Berkeley and its contributors.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)uipc_socket.c	8.3 (Berkeley) 4/15/94
35  * $FreeBSD: src/sys/kern/uipc_socket.c,v 1.68.2.24 2003/11/11 17:18:18 silby Exp $
36  * $DragonFly: src/sys/kern/uipc_socket.c,v 1.16 2004/03/27 11:48:48 hsu Exp $
37  */
38 
39 #include "opt_inet.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/fcntl.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.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/socketops.h>
57 #include <sys/resourcevar.h>
58 #include <sys/signalvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/uio.h>
61 #include <sys/jail.h>
62 #include <vm/vm_zone.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 /* INET */
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 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 = zalloc(socket_zone);
117 	if (so) {
118 		/* XXX race condition for reentrant kernel */
119 		bzero(so, sizeof *so);
120 		so->so_gencnt = ++so_gencnt;
121 		TAILQ_INIT(&so->so_aiojobq);
122 	}
123 	return so;
124 }
125 
126 int
127 socreate(int dom, struct socket **aso, int type,
128 	int proto, struct thread *td)
129 {
130 	struct proc *p = td->td_proc;
131 	struct protosw *prp;
132 	struct socket *so;
133 	struct pru_attach_info ai;
134 	int error;
135 
136 	if (proto)
137 		prp = pffindproto(dom, proto, type);
138 	else
139 		prp = pffindtype(dom, type);
140 
141 	if (prp == 0 || prp->pr_usrreqs->pru_attach == 0)
142 		return (EPROTONOSUPPORT);
143 
144 	if (p->p_ucred->cr_prison && jail_socket_unixiproute_only &&
145 	    prp->pr_domain->dom_family != PF_LOCAL &&
146 	    prp->pr_domain->dom_family != PF_INET &&
147 	    prp->pr_domain->dom_family != PF_ROUTE) {
148 		return (EPROTONOSUPPORT);
149 	}
150 
151 	if (prp->pr_type != type)
152 		return (EPROTOTYPE);
153 	so = soalloc(p != 0);
154 	if (so == 0)
155 		return (ENOBUFS);
156 
157 	TAILQ_INIT(&so->so_incomp);
158 	TAILQ_INIT(&so->so_comp);
159 	so->so_type = type;
160 	so->so_cred = crhold(p->p_ucred);
161 	so->so_proto = prp;
162 	ai.sb_rlimit = &p->p_rlimit[RLIMIT_SBSIZE];
163 	ai.p_ucred = p->p_ucred;
164 	ai.fd_rdir = p->p_fd->fd_rdir;
165 	error = so_pru_attach(so, proto, &ai);
166 	if (error) {
167 		so->so_state |= SS_NOFDREF;
168 		sofree(so);
169 		return (error);
170 	}
171 	*aso = so;
172 	return (0);
173 }
174 
175 int
176 sobind(struct socket *so, struct sockaddr *nam, struct thread *td)
177 {
178 	int s = splnet();
179 	int error;
180 
181 	error = so_pru_bind(so, nam, td);
182 	splx(s);
183 	return (error);
184 }
185 
186 void
187 sodealloc(struct socket *so)
188 {
189 
190 	so->so_gencnt = ++so_gencnt;
191 	if (so->so_rcv.sb_hiwat)
192 		(void)chgsbsize(so->so_cred->cr_uidinfo,
193 		    &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
194 	if (so->so_snd.sb_hiwat)
195 		(void)chgsbsize(so->so_cred->cr_uidinfo,
196 		    &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
197 #ifdef INET
198 	if (so->so_accf != NULL) {
199 		if (so->so_accf->so_accept_filter != NULL &&
200 			so->so_accf->so_accept_filter->accf_destroy != NULL) {
201 			so->so_accf->so_accept_filter->accf_destroy(so);
202 		}
203 		if (so->so_accf->so_accept_filter_str != NULL)
204 			FREE(so->so_accf->so_accept_filter_str, M_ACCF);
205 		FREE(so->so_accf, M_ACCF);
206 	}
207 #endif /* INET */
208 	crfree(so->so_cred);
209 	zfree(socket_zone, so);
210 }
211 
212 int
213 solisten(struct socket *so, int backlog, struct thread *td)
214 {
215 	int s, error;
216 
217 	s = splnet();
218  	if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) {
219  		splx(s);
220  		return (EINVAL);
221  	}
222 
223 	error = so_pru_listen(so, td);
224 	if (error) {
225 		splx(s);
226 		return (error);
227 	}
228 	if (TAILQ_EMPTY(&so->so_comp))
229 		so->so_options |= SO_ACCEPTCONN;
230 	if (backlog < 0 || backlog > somaxconn)
231 		backlog = somaxconn;
232 	so->so_qlimit = backlog;
233 	splx(s);
234 	return (0);
235 }
236 
237 void
238 sofree(struct socket *so)
239 {
240 	struct socket *head = so->so_head;
241 
242 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
243 		return;
244 	if (head != NULL) {
245 		if (so->so_state & SS_INCOMP) {
246 			TAILQ_REMOVE(&head->so_incomp, so, so_list);
247 			head->so_incqlen--;
248 		} else if (so->so_state & SS_COMP) {
249 			/*
250 			 * We must not decommission a socket that's
251 			 * on the accept(2) queue.  If we do, then
252 			 * accept(2) may hang after select(2) indicated
253 			 * that the listening socket was ready.
254 			 */
255 			return;
256 		} else {
257 			panic("sofree: not queued");
258 		}
259 		so->so_state &= ~SS_INCOMP;
260 		so->so_head = NULL;
261 	}
262 	sbrelease(&so->so_snd, so);
263 	sorflush(so);
264 	sodealloc(so);
265 }
266 
267 /*
268  * Close a socket on last file table reference removal.
269  * Initiate disconnect if connected.
270  * Free socket when disconnect complete.
271  */
272 int
273 soclose(struct socket *so)
274 {
275 	int s = splnet();		/* conservative */
276 	int error = 0;
277 
278 	funsetown(so->so_sigio);
279 	if (so->so_options & SO_ACCEPTCONN) {
280 		struct socket *sp, *sonext;
281 
282 		sp = TAILQ_FIRST(&so->so_incomp);
283 		for (; sp != NULL; sp = sonext) {
284 			sonext = TAILQ_NEXT(sp, so_list);
285 			(void) soabort(sp);
286 		}
287 		for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
288 			sonext = TAILQ_NEXT(sp, so_list);
289 			/* Dequeue from so_comp since sofree() won't do it */
290 			TAILQ_REMOVE(&so->so_comp, sp, so_list);
291 			so->so_qlen--;
292 			sp->so_state &= ~SS_COMP;
293 			sp->so_head = NULL;
294 			(void) soabort(sp);
295 		}
296 	}
297 	if (so->so_pcb == 0)
298 		goto discard;
299 	if (so->so_state & SS_ISCONNECTED) {
300 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
301 			error = sodisconnect(so);
302 			if (error)
303 				goto drop;
304 		}
305 		if (so->so_options & SO_LINGER) {
306 			if ((so->so_state & SS_ISDISCONNECTING) &&
307 			    (so->so_state & SS_NBIO))
308 				goto drop;
309 			while (so->so_state & SS_ISCONNECTED) {
310 				error = tsleep((caddr_t)&so->so_timeo,
311 				    PCATCH, "soclos", so->so_linger * hz);
312 				if (error)
313 					break;
314 			}
315 		}
316 	}
317 drop:
318 	if (so->so_pcb) {
319 		int error2;
320 
321 		error2 = so_pru_detach(so);
322 		if (error == 0)
323 			error = error2;
324 	}
325 discard:
326 	if (so->so_state & SS_NOFDREF)
327 		panic("soclose: NOFDREF");
328 	so->so_state |= SS_NOFDREF;
329 	sofree(so);
330 	splx(s);
331 	return (error);
332 }
333 
334 /*
335  * Must be called at splnet...
336  */
337 int
338 soabort(so)
339 	struct socket *so;
340 {
341 	int error;
342 
343 	error = so_pru_abort(so);
344 	if (error) {
345 		sofree(so);
346 		return error;
347 	}
348 	return (0);
349 }
350 
351 int
352 soaccept(struct socket *so, struct sockaddr **nam)
353 {
354 	int s = splnet();
355 	int error;
356 
357 	if ((so->so_state & SS_NOFDREF) == 0)
358 		panic("soaccept: !NOFDREF");
359 	so->so_state &= ~SS_NOFDREF;
360 	error = so_pru_accept(so, nam);
361 	splx(s);
362 	return (error);
363 }
364 
365 int
366 soconnect(struct socket *so, struct sockaddr *nam, struct thread *td)
367 {
368 	int s;
369 	int error;
370 
371 	if (so->so_options & SO_ACCEPTCONN)
372 		return (EOPNOTSUPP);
373 	s = splnet();
374 	/*
375 	 * If protocol is connection-based, can only connect once.
376 	 * Otherwise, if connected, try to disconnect first.
377 	 * This allows user to disconnect by connecting to, e.g.,
378 	 * a null address.
379 	 */
380 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
381 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
382 	    (error = sodisconnect(so))))
383 		error = EISCONN;
384 	else
385 		error = so_pru_connect(so, nam, td);
386 	splx(s);
387 	return (error);
388 }
389 
390 int
391 soconnect2(struct socket *so1, struct socket *so2)
392 {
393 	int s = splnet();
394 	int error;
395 
396 	error = so_pru_connect2(so1, so2);
397 	splx(s);
398 	return (error);
399 }
400 
401 int
402 sodisconnect(struct socket *so)
403 {
404 	int s = splnet();
405 	int error;
406 
407 	if ((so->so_state & SS_ISCONNECTED) == 0) {
408 		error = ENOTCONN;
409 		goto bad;
410 	}
411 	if (so->so_state & SS_ISDISCONNECTING) {
412 		error = EALREADY;
413 		goto bad;
414 	}
415 	error = so_pru_disconnect(so);
416 bad:
417 	splx(s);
418 	return (error);
419 }
420 
421 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
422 /*
423  * Send on a socket.
424  * If send must go all at once and message is larger than
425  * send buffering, then hard error.
426  * Lock against other senders.
427  * If must go all at once and not enough room now, then
428  * inform user that this would block and do nothing.
429  * Otherwise, if nonblocking, send as much as possible.
430  * The data to be sent is described by "uio" if nonzero,
431  * otherwise by the mbuf chain "top" (which must be null
432  * if uio is not).  Data provided in mbuf chain must be small
433  * enough to send all at once.
434  *
435  * Returns nonzero on error, timeout or signal; callers
436  * must check for short counts if EINTR/ERESTART are returned.
437  * Data and control buffers are freed on return.
438  */
439 int
440 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
441 	struct mbuf *top, struct mbuf *control, int flags,
442 	struct thread *td)
443 {
444 	struct mbuf **mp;
445 	struct mbuf *m;
446 	long space, len, resid;
447 	int clen = 0, error, s, dontroute, mlen;
448 	int atomic = sosendallatonce(so) || top;
449 	int pru_flags;
450 
451 	if (uio)
452 		resid = uio->uio_resid;
453 	else
454 		resid = top->m_pkthdr.len;
455 	/*
456 	 * In theory resid should be unsigned.
457 	 * However, space must be signed, as it might be less than 0
458 	 * if we over-committed, and we must use a signed comparison
459 	 * of space and resid.  On the other hand, a negative resid
460 	 * causes us to loop sending 0-length segments to the protocol.
461 	 *
462 	 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
463 	 * type sockets since that's an error.
464 	 */
465 	if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
466 		error = EINVAL;
467 		goto out;
468 	}
469 
470 	dontroute =
471 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
472 	    (so->so_proto->pr_flags & PR_ATOMIC);
473 	if (td->td_proc && td->td_proc->p_stats)
474 		td->td_proc->p_stats->p_ru.ru_msgsnd++;
475 	if (control)
476 		clen = control->m_len;
477 #define	gotoerr(errno)	{ error = errno; splx(s); goto release; }
478 
479 restart:
480 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
481 	if (error)
482 		goto out;
483 	do {
484 		s = splnet();
485 		if (so->so_state & SS_CANTSENDMORE)
486 			gotoerr(EPIPE);
487 		if (so->so_error) {
488 			error = so->so_error;
489 			so->so_error = 0;
490 			splx(s);
491 			goto release;
492 		}
493 		if ((so->so_state & SS_ISCONNECTED) == 0) {
494 			/*
495 			 * `sendto' and `sendmsg' is allowed on a connection-
496 			 * based socket if it supports implied connect.
497 			 * Return ENOTCONN if not connected and no address is
498 			 * supplied.
499 			 */
500 			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
501 			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
502 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
503 				    !(resid == 0 && clen != 0))
504 					gotoerr(ENOTCONN);
505 			} else if (addr == 0)
506 			    gotoerr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
507 				   ENOTCONN : EDESTADDRREQ);
508 		}
509 		space = sbspace(&so->so_snd);
510 		if (flags & MSG_OOB)
511 			space += 1024;
512 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
513 		    clen > so->so_snd.sb_hiwat)
514 			gotoerr(EMSGSIZE);
515 		if (space < resid + clen && uio &&
516 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
517 			if (so->so_state & SS_NBIO)
518 				gotoerr(EWOULDBLOCK);
519 			sbunlock(&so->so_snd);
520 			error = sbwait(&so->so_snd);
521 			splx(s);
522 			if (error)
523 				goto out;
524 			goto restart;
525 		}
526 		splx(s);
527 		mp = &top;
528 		space -= clen;
529 		do {
530 		    if (uio == NULL) {
531 			/*
532 			 * Data is prepackaged in "top".
533 			 */
534 			resid = 0;
535 			if (flags & MSG_EOR)
536 				top->m_flags |= M_EOR;
537 		    } else do {
538 			if (top == 0) {
539 				MGETHDR(m, M_WAIT, MT_DATA);
540 				if (m == NULL) {
541 					error = ENOBUFS;
542 					goto release;
543 				}
544 				mlen = MHLEN;
545 				m->m_pkthdr.len = 0;
546 				m->m_pkthdr.rcvif = (struct ifnet *)0;
547 			} else {
548 				MGET(m, M_WAIT, MT_DATA);
549 				if (m == NULL) {
550 					error = ENOBUFS;
551 					goto release;
552 				}
553 				mlen = MLEN;
554 			}
555 			if (resid >= MINCLSIZE) {
556 				MCLGET(m, M_WAIT);
557 				if ((m->m_flags & M_EXT) == 0)
558 					goto nopages;
559 				mlen = MCLBYTES;
560 				len = min(min(mlen, resid), space);
561 			} else {
562 nopages:
563 				len = min(min(mlen, resid), space);
564 				/*
565 				 * For datagram protocols, leave room
566 				 * for protocol headers in first mbuf.
567 				 */
568 				if (atomic && top == 0 && len < mlen)
569 					MH_ALIGN(m, len);
570 			}
571 			space -= len;
572 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
573 			resid = uio->uio_resid;
574 			m->m_len = len;
575 			*mp = m;
576 			top->m_pkthdr.len += len;
577 			if (error)
578 				goto release;
579 			mp = &m->m_next;
580 			if (resid <= 0) {
581 				if (flags & MSG_EOR)
582 					top->m_flags |= M_EOR;
583 				break;
584 			}
585 		    } while (space > 0 && atomic);
586 		    if (dontroute)
587 			    so->so_options |= SO_DONTROUTE;
588 		    if (flags & MSG_OOB) {
589 		    	    pru_flags = PRUS_OOB;
590 		    } else if ((flags & MSG_EOF) &&
591 		    	       (so->so_proto->pr_flags & PR_IMPLOPCL) &&
592 		    	       (resid <= 0)) {
593 			    /*
594 			     * If the user set MSG_EOF, the protocol
595 			     * understands this flag and nothing left to
596 			     * send then use PRU_SEND_EOF instead of PRU_SEND.
597 			     */
598 		    	    pru_flags = PRUS_EOF;
599 		    } else if (resid > 0 && space > 0) {
600 			    /* If there is more to send, set PRUS_MORETOCOME */
601 		    	    pru_flags = PRUS_MORETOCOME;
602 		    } else {
603 		    	    pru_flags = 0;
604 		    }
605 		    s = splnet();				/* XXX */
606 		    /*
607 		     * XXX all the SS_CANTSENDMORE checks previously
608 		     * done could be out of date.  We could have recieved
609 		     * a reset packet in an interrupt or maybe we slept
610 		     * while doing page faults in uiomove() etc. We could
611 		     * probably recheck again inside the splnet() protection
612 		     * here, but there are probably other places that this
613 		     * also happens.  We must rethink this.
614 		     */
615 		    error = so_pru_send(so, pru_flags, top, addr, control, td);
616 		    splx(s);
617 		    if (dontroute)
618 			    so->so_options &= ~SO_DONTROUTE;
619 		    clen = 0;
620 		    control = 0;
621 		    top = 0;
622 		    mp = &top;
623 		    if (error)
624 			    goto release;
625 		} while (resid && space > 0);
626 	} while (resid);
627 
628 release:
629 	sbunlock(&so->so_snd);
630 out:
631 	if (top)
632 		m_freem(top);
633 	if (control)
634 		m_freem(control);
635 	return (error);
636 }
637 
638 /*
639  * A specialization of sosend() for UDP based on protocol-specific knowledge:
640  *   so->so_proto->pr_flags has the PR_ATOMIC field set.  This means that
641  *	sosendallatonce() returns true,
642  *	the "atomic" variable is true,
643  *	and sosendudp() blocks until space is available for the entire send.
644  *   so->so_proto->pr_flags does not have the PR_CONNREQUIRED or
645  *	PR_IMPLOPCL flags set.
646  *   UDP has no out-of-band data.
647  *   UDP has no control data.
648  *   UDP does not support MSG_EOR.
649  */
650 int
651 sosendudp(struct socket *so, struct sockaddr *addr, struct uio *uio,
652 	  struct mbuf *top, struct mbuf *control, int flags, struct thread *td)
653 {
654 	int resid, error, s;
655 	boolean_t dontroute;		/* temporary SO_DONTROUTE setting */
656 
657 	if (td->td_proc && td->td_proc->p_stats)
658 		td->td_proc->p_stats->p_ru.ru_msgsnd++;
659 	if (control)
660 		m_freem(control);
661 
662 	KASSERT((uio && !top) || (top && !uio), ("bad arguments to sosendudp"));
663 	resid = uio ? uio->uio_resid : top->m_pkthdr.len;
664 
665 restart:
666 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
667 	if (error)
668 		goto out;
669 
670 	s = splnet();
671 	if (so->so_state & SS_CANTSENDMORE)
672 		gotoerr(EPIPE);
673 	if (so->so_error) {
674 		error = so->so_error;
675 		so->so_error = 0;
676 		splx(s);
677 		goto release;
678 	}
679 	if (!(so->so_state & SS_ISCONNECTED) && addr == NULL)
680 		gotoerr(EDESTADDRREQ);
681 	if (resid > so->so_snd.sb_hiwat)
682 		gotoerr(EMSGSIZE);
683 	if (uio && sbspace(&so->so_snd) < resid) {
684 		if (so->so_state & SS_NBIO)
685 			gotoerr(EWOULDBLOCK);
686 		sbunlock(&so->so_snd);
687 		error = sbwait(&so->so_snd);
688 		splx(s);
689 		if (error)
690 			goto out;
691 		goto restart;
692 	}
693 	splx(s);
694 
695 	if (uio) {
696 		top = m_uiomove(uio, M_WAIT, 0);
697 		if (top == NULL)
698 			goto release;
699 	}
700 
701 	dontroute = (flags & MSG_DONTROUTE) && !(so->so_options & SO_DONTROUTE);
702 	if (dontroute)
703 		so->so_options |= SO_DONTROUTE;
704 
705 	error = so_pru_send(so, 0, top, addr, NULL, td);
706 	top = NULL;		/* sent or freed in lower layer */
707 
708 	if (dontroute)
709 		so->so_options &= ~SO_DONTROUTE;
710 
711 release:
712 	sbunlock(&so->so_snd);
713 out:
714 	if (top)
715 		m_freem(top);
716 	return (error);
717 }
718 
719 /*
720  * Implement receive operations on a socket.
721  * We depend on the way that records are added to the sockbuf
722  * by sbappend*.  In particular, each record (mbufs linked through m_next)
723  * must begin with an address if the protocol so specifies,
724  * followed by an optional mbuf or mbufs containing ancillary data,
725  * and then zero or more mbufs of data.
726  * In order to avoid blocking network interrupts for the entire time here,
727  * we splx() while doing the actual copy to user space.
728  * Although the sockbuf is locked, new data may still be appended,
729  * and thus we must maintain consistency of the sockbuf during that time.
730  *
731  * The caller may receive the data as a single mbuf chain by supplying
732  * an mbuf **mp0 for use in returning the chain.  The uio is then used
733  * only for the count in uio_resid.
734  */
735 int
736 soreceive(so, psa, uio, mp0, controlp, flagsp)
737 	struct socket *so;
738 	struct sockaddr **psa;
739 	struct uio *uio;
740 	struct mbuf **mp0;
741 	struct mbuf **controlp;
742 	int *flagsp;
743 {
744 	struct mbuf *m, **mp;
745 	int flags, len, error, s, offset;
746 	struct protosw *pr = so->so_proto;
747 	struct mbuf *nextrecord;
748 	int moff, type = 0;
749 	int orig_resid = uio->uio_resid;
750 
751 	mp = mp0;
752 	if (psa)
753 		*psa = 0;
754 	if (controlp)
755 		*controlp = 0;
756 	if (flagsp)
757 		flags = *flagsp &~ MSG_EOR;
758 	else
759 		flags = 0;
760 	if (flags & MSG_OOB) {
761 		m = m_get(M_WAIT, MT_DATA);
762 		if (m == NULL)
763 			return (ENOBUFS);
764 		error = so_pru_rcvoob(so, m, flags & MSG_PEEK);
765 		if (error)
766 			goto bad;
767 		do {
768 			error = uiomove(mtod(m, caddr_t),
769 			    (int) min(uio->uio_resid, m->m_len), uio);
770 			m = m_free(m);
771 		} while (uio->uio_resid && error == 0 && m);
772 bad:
773 		if (m)
774 			m_freem(m);
775 		return (error);
776 	}
777 	if (mp)
778 		*mp = (struct mbuf *)0;
779 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
780 		so_pru_rcvd(so, 0);
781 
782 restart:
783 	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
784 	if (error)
785 		return (error);
786 	s = splnet();
787 
788 	m = so->so_rcv.sb_mb;
789 	/*
790 	 * If we have less data than requested, block awaiting more
791 	 * (subject to any timeout) if:
792 	 *   1. the current count is less than the low water mark, or
793 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
794 	 *	receive operation at once if we block (resid <= hiwat).
795 	 *   3. MSG_DONTWAIT is not set
796 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
797 	 * we have to do the receive in sections, and thus risk returning
798 	 * a short count if a timeout or signal occurs after we start.
799 	 */
800 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
801 	    so->so_rcv.sb_cc < uio->uio_resid) &&
802 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
803 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
804 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
805 		KASSERT(m != 0 || !so->so_rcv.sb_cc, ("receive 1"));
806 		if (so->so_error) {
807 			if (m)
808 				goto dontblock;
809 			error = so->so_error;
810 			if ((flags & MSG_PEEK) == 0)
811 				so->so_error = 0;
812 			goto release;
813 		}
814 		if (so->so_state & SS_CANTRCVMORE) {
815 			if (m)
816 				goto dontblock;
817 			else
818 				goto release;
819 		}
820 		for (; m; m = m->m_next)
821 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
822 				m = so->so_rcv.sb_mb;
823 				goto dontblock;
824 			}
825 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
826 		    (pr->pr_flags & PR_CONNREQUIRED)) {
827 			error = ENOTCONN;
828 			goto release;
829 		}
830 		if (uio->uio_resid == 0)
831 			goto release;
832 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
833 			error = EWOULDBLOCK;
834 			goto release;
835 		}
836 		sbunlock(&so->so_rcv);
837 		error = sbwait(&so->so_rcv);
838 		splx(s);
839 		if (error)
840 			return (error);
841 		goto restart;
842 	}
843 dontblock:
844 	if (uio->uio_td && uio->uio_td->td_proc)
845 		uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++;
846 	nextrecord = m->m_nextpkt;
847 	if (pr->pr_flags & PR_ADDR) {
848 		KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
849 		orig_resid = 0;
850 		if (psa)
851 			*psa = dup_sockaddr(mtod(m, struct sockaddr *),
852 					    mp0 == 0);
853 		if (flags & MSG_PEEK) {
854 			m = m->m_next;
855 		} else {
856 			sbfree(&so->so_rcv, m);
857 			so->so_rcv.sb_mb = m_free(m);
858 			m = so->so_rcv.sb_mb;
859 		}
860 	}
861 	while (m && m->m_type == MT_CONTROL && error == 0) {
862 		if (flags & MSG_PEEK) {
863 			if (controlp)
864 				*controlp = m_copy(m, 0, m->m_len);
865 			m = m->m_next;
866 		} else {
867 			sbfree(&so->so_rcv, m);
868 			if (controlp) {
869 				if (pr->pr_domain->dom_externalize &&
870 				    mtod(m, struct cmsghdr *)->cmsg_type ==
871 				    SCM_RIGHTS)
872 				   error = (*pr->pr_domain->dom_externalize)(m);
873 				*controlp = m;
874 				so->so_rcv.sb_mb = m->m_next;
875 				m->m_next = 0;
876 				m = so->so_rcv.sb_mb;
877 			} else {
878 				so->so_rcv.sb_mb = m_free(m);
879 				m = so->so_rcv.sb_mb;
880 			}
881 		}
882 		if (controlp) {
883 			orig_resid = 0;
884 			controlp = &(*controlp)->m_next;
885 		}
886 	}
887 	if (m) {
888 		if ((flags & MSG_PEEK) == 0)
889 			m->m_nextpkt = nextrecord;
890 		type = m->m_type;
891 		if (type == MT_OOBDATA)
892 			flags |= MSG_OOB;
893 	}
894 	moff = 0;
895 	offset = 0;
896 	while (m && uio->uio_resid > 0 && error == 0) {
897 		if (m->m_type == MT_OOBDATA) {
898 			if (type != MT_OOBDATA)
899 				break;
900 		} else if (type == MT_OOBDATA)
901 			break;
902 		else
903 		    KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
904 			("receive 3"));
905 		so->so_state &= ~SS_RCVATMARK;
906 		len = uio->uio_resid;
907 		if (so->so_oobmark && len > so->so_oobmark - offset)
908 			len = so->so_oobmark - offset;
909 		if (len > m->m_len - moff)
910 			len = m->m_len - moff;
911 		/*
912 		 * If mp is set, just pass back the mbufs.
913 		 * Otherwise copy them out via the uio, then free.
914 		 * Sockbuf must be consistent here (points to current mbuf,
915 		 * it points to next record) when we drop priority;
916 		 * we must note any additions to the sockbuf when we
917 		 * block interrupts again.
918 		 */
919 		if (mp == 0) {
920 			splx(s);
921 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
922 			s = splnet();
923 			if (error)
924 				goto release;
925 		} else
926 			uio->uio_resid -= len;
927 		if (len == m->m_len - moff) {
928 			if (m->m_flags & M_EOR)
929 				flags |= MSG_EOR;
930 			if (flags & MSG_PEEK) {
931 				m = m->m_next;
932 				moff = 0;
933 			} else {
934 				nextrecord = m->m_nextpkt;
935 				sbfree(&so->so_rcv, m);
936 				if (mp) {
937 					*mp = m;
938 					mp = &m->m_next;
939 					so->so_rcv.sb_mb = m = m->m_next;
940 					*mp = (struct mbuf *)0;
941 				} else {
942 					so->so_rcv.sb_mb = m = m_free(m);
943 				}
944 				if (m)
945 					m->m_nextpkt = nextrecord;
946 			}
947 		} else {
948 			if (flags & MSG_PEEK)
949 				moff += len;
950 			else {
951 				if (mp)
952 					*mp = m_copym(m, 0, len, M_WAIT);
953 				m->m_data += len;
954 				m->m_len -= len;
955 				so->so_rcv.sb_cc -= len;
956 			}
957 		}
958 		if (so->so_oobmark) {
959 			if ((flags & MSG_PEEK) == 0) {
960 				so->so_oobmark -= len;
961 				if (so->so_oobmark == 0) {
962 					so->so_state |= SS_RCVATMARK;
963 					break;
964 				}
965 			} else {
966 				offset += len;
967 				if (offset == so->so_oobmark)
968 					break;
969 			}
970 		}
971 		if (flags & MSG_EOR)
972 			break;
973 		/*
974 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
975 		 * we must not quit until "uio->uio_resid == 0" or an error
976 		 * termination.  If a signal/timeout occurs, return
977 		 * with a short count but without error.
978 		 * Keep sockbuf locked against other readers.
979 		 */
980 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
981 		    !sosendallatonce(so) && !nextrecord) {
982 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
983 				break;
984 			/*
985 			 * The window might have closed to zero, make
986 			 * sure we send an ack now that we've drained
987 			 * the buffer or we might end up blocking until
988 			 * the idle takes over (5 seconds).
989 			 */
990 			if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
991 				so_pru_rcvd(so, flags);
992 			error = sbwait(&so->so_rcv);
993 			if (error) {
994 				sbunlock(&so->so_rcv);
995 				splx(s);
996 				return (0);
997 			}
998 			m = so->so_rcv.sb_mb;
999 			if (m)
1000 				nextrecord = m->m_nextpkt;
1001 		}
1002 	}
1003 
1004 	if (m && pr->pr_flags & PR_ATOMIC) {
1005 		flags |= MSG_TRUNC;
1006 		if ((flags & MSG_PEEK) == 0)
1007 			(void) sbdroprecord(&so->so_rcv);
1008 	}
1009 	if ((flags & MSG_PEEK) == 0) {
1010 		if (m == 0)
1011 			so->so_rcv.sb_mb = nextrecord;
1012 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1013 			so_pru_rcvd(so, flags);
1014 	}
1015 	if (orig_resid == uio->uio_resid && orig_resid &&
1016 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1017 		sbunlock(&so->so_rcv);
1018 		splx(s);
1019 		goto restart;
1020 	}
1021 
1022 	if (flagsp)
1023 		*flagsp |= flags;
1024 release:
1025 	sbunlock(&so->so_rcv);
1026 	splx(s);
1027 	return (error);
1028 }
1029 
1030 int
1031 soshutdown(so, how)
1032 	struct socket *so;
1033 	int how;
1034 {
1035 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1036 		return (EINVAL);
1037 
1038 	if (how != SHUT_WR)
1039 		sorflush(so);
1040 	if (how != SHUT_RD)
1041 		return (so_pru_shutdown(so));
1042 	return (0);
1043 }
1044 
1045 void
1046 sorflush(so)
1047 	struct socket *so;
1048 {
1049 	struct sockbuf *sb = &so->so_rcv;
1050 	struct protosw *pr = so->so_proto;
1051 	int s;
1052 	struct sockbuf asb;
1053 
1054 	sb->sb_flags |= SB_NOINTR;
1055 	(void) sblock(sb, M_WAITOK);
1056 	s = splimp();
1057 	socantrcvmore(so);
1058 	sbunlock(sb);
1059 	asb = *sb;
1060 	bzero((caddr_t)sb, sizeof (*sb));
1061 	if (asb.sb_flags & SB_KNOTE) {
1062 		sb->sb_sel.si_note = asb.sb_sel.si_note;
1063 		sb->sb_flags = SB_KNOTE;
1064 	}
1065 	splx(s);
1066 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1067 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
1068 	sbrelease(&asb, so);
1069 }
1070 
1071 #ifdef INET
1072 static int
1073 do_setopt_accept_filter(so, sopt)
1074 	struct	socket *so;
1075 	struct	sockopt *sopt;
1076 {
1077 	struct accept_filter_arg	*afap = NULL;
1078 	struct accept_filter	*afp;
1079 	struct so_accf	*af = so->so_accf;
1080 	int	error = 0;
1081 
1082 	/* do not set/remove accept filters on non listen sockets */
1083 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
1084 		error = EINVAL;
1085 		goto out;
1086 	}
1087 
1088 	/* removing the filter */
1089 	if (sopt == NULL) {
1090 		if (af != NULL) {
1091 			if (af->so_accept_filter != NULL &&
1092 				af->so_accept_filter->accf_destroy != NULL) {
1093 				af->so_accept_filter->accf_destroy(so);
1094 			}
1095 			if (af->so_accept_filter_str != NULL) {
1096 				FREE(af->so_accept_filter_str, M_ACCF);
1097 			}
1098 			FREE(af, M_ACCF);
1099 			so->so_accf = NULL;
1100 		}
1101 		so->so_options &= ~SO_ACCEPTFILTER;
1102 		return (0);
1103 	}
1104 	/* adding a filter */
1105 	/* must remove previous filter first */
1106 	if (af != NULL) {
1107 		error = EINVAL;
1108 		goto out;
1109 	}
1110 	/* don't put large objects on the kernel stack */
1111 	MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1112 	error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1113 	afap->af_name[sizeof(afap->af_name)-1] = '\0';
1114 	afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1115 	if (error)
1116 		goto out;
1117 	afp = accept_filt_get(afap->af_name);
1118 	if (afp == NULL) {
1119 		error = ENOENT;
1120 		goto out;
1121 	}
1122 	MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK);
1123 	bzero(af, sizeof(*af));
1124 	if (afp->accf_create != NULL) {
1125 		if (afap->af_name[0] != '\0') {
1126 			int len = strlen(afap->af_name) + 1;
1127 
1128 			MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1129 			strcpy(af->so_accept_filter_str, afap->af_name);
1130 		}
1131 		af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1132 		if (af->so_accept_filter_arg == NULL) {
1133 			FREE(af->so_accept_filter_str, M_ACCF);
1134 			FREE(af, M_ACCF);
1135 			so->so_accf = NULL;
1136 			error = EINVAL;
1137 			goto out;
1138 		}
1139 	}
1140 	af->so_accept_filter = afp;
1141 	so->so_accf = af;
1142 	so->so_options |= SO_ACCEPTFILTER;
1143 out:
1144 	if (afap != NULL)
1145 		FREE(afap, M_TEMP);
1146 	return (error);
1147 }
1148 #endif /* INET */
1149 
1150 /*
1151  * Perhaps this routine, and sooptcopyout(), below, ought to come in
1152  * an additional variant to handle the case where the option value needs
1153  * to be some kind of integer, but not a specific size.
1154  * In addition to their use here, these functions are also called by the
1155  * protocol-level pr_ctloutput() routines.
1156  */
1157 int
1158 sooptcopyin(sopt, buf, len, minlen)
1159 	struct	sockopt *sopt;
1160 	void	*buf;
1161 	size_t	len;
1162 	size_t	minlen;
1163 {
1164 	size_t	valsize;
1165 
1166 	/*
1167 	 * If the user gives us more than we wanted, we ignore it,
1168 	 * but if we don't get the minimum length the caller
1169 	 * wants, we return EINVAL.  On success, sopt->sopt_valsize
1170 	 * is set to however much we actually retrieved.
1171 	 */
1172 	if ((valsize = sopt->sopt_valsize) < minlen)
1173 		return EINVAL;
1174 	if (valsize > len)
1175 		sopt->sopt_valsize = valsize = len;
1176 
1177 	if (sopt->sopt_td != NULL)
1178 		return (copyin(sopt->sopt_val, buf, valsize));
1179 
1180 	bcopy(sopt->sopt_val, buf, valsize);
1181 	return 0;
1182 }
1183 
1184 int
1185 sosetopt(so, sopt)
1186 	struct socket *so;
1187 	struct sockopt *sopt;
1188 {
1189 	int	error, optval;
1190 	struct	linger l;
1191 	struct	timeval tv;
1192 	u_long  val;
1193 
1194 	error = 0;
1195 	if (sopt->sopt_level != SOL_SOCKET) {
1196 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1197 			return (so_pr_ctloutput(so, sopt));
1198 		}
1199 		error = ENOPROTOOPT;
1200 	} else {
1201 		switch (sopt->sopt_name) {
1202 #ifdef INET
1203 		case SO_ACCEPTFILTER:
1204 			error = do_setopt_accept_filter(so, sopt);
1205 			if (error)
1206 				goto bad;
1207 			break;
1208 #endif /* INET */
1209 		case SO_LINGER:
1210 			error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1211 			if (error)
1212 				goto bad;
1213 
1214 			so->so_linger = l.l_linger;
1215 			if (l.l_onoff)
1216 				so->so_options |= SO_LINGER;
1217 			else
1218 				so->so_options &= ~SO_LINGER;
1219 			break;
1220 
1221 		case SO_DEBUG:
1222 		case SO_KEEPALIVE:
1223 		case SO_DONTROUTE:
1224 		case SO_USELOOPBACK:
1225 		case SO_BROADCAST:
1226 		case SO_REUSEADDR:
1227 		case SO_REUSEPORT:
1228 		case SO_OOBINLINE:
1229 		case SO_TIMESTAMP:
1230 			error = sooptcopyin(sopt, &optval, sizeof optval,
1231 					    sizeof optval);
1232 			if (error)
1233 				goto bad;
1234 			if (optval)
1235 				so->so_options |= sopt->sopt_name;
1236 			else
1237 				so->so_options &= ~sopt->sopt_name;
1238 			break;
1239 
1240 		case SO_SNDBUF:
1241 		case SO_RCVBUF:
1242 		case SO_SNDLOWAT:
1243 		case SO_RCVLOWAT:
1244 			error = sooptcopyin(sopt, &optval, sizeof optval,
1245 					    sizeof optval);
1246 			if (error)
1247 				goto bad;
1248 
1249 			/*
1250 			 * Values < 1 make no sense for any of these
1251 			 * options, so disallow them.
1252 			 */
1253 			if (optval < 1) {
1254 				error = EINVAL;
1255 				goto bad;
1256 			}
1257 
1258 			switch (sopt->sopt_name) {
1259 			case SO_SNDBUF:
1260 			case SO_RCVBUF:
1261 				if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1262 				    &so->so_snd : &so->so_rcv, (u_long)optval,
1263 				    so,
1264 				    &curproc->p_rlimit[RLIMIT_SBSIZE]) == 0) {
1265 					error = ENOBUFS;
1266 					goto bad;
1267 				}
1268 				break;
1269 
1270 			/*
1271 			 * Make sure the low-water is never greater than
1272 			 * the high-water.
1273 			 */
1274 			case SO_SNDLOWAT:
1275 				so->so_snd.sb_lowat =
1276 				    (optval > so->so_snd.sb_hiwat) ?
1277 				    so->so_snd.sb_hiwat : optval;
1278 				break;
1279 			case SO_RCVLOWAT:
1280 				so->so_rcv.sb_lowat =
1281 				    (optval > so->so_rcv.sb_hiwat) ?
1282 				    so->so_rcv.sb_hiwat : optval;
1283 				break;
1284 			}
1285 			break;
1286 
1287 		case SO_SNDTIMEO:
1288 		case SO_RCVTIMEO:
1289 			error = sooptcopyin(sopt, &tv, sizeof tv,
1290 					    sizeof tv);
1291 			if (error)
1292 				goto bad;
1293 
1294 			/* assert(hz > 0); */
1295 			if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1296 			    tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1297 				error = EDOM;
1298 				goto bad;
1299 			}
1300 			/* assert(tick > 0); */
1301 			/* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1302 			val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1303 			if (val > SHRT_MAX) {
1304 				error = EDOM;
1305 				goto bad;
1306 			}
1307 			if (val == 0 && tv.tv_usec != 0)
1308 				val = 1;
1309 
1310 			switch (sopt->sopt_name) {
1311 			case SO_SNDTIMEO:
1312 				so->so_snd.sb_timeo = val;
1313 				break;
1314 			case SO_RCVTIMEO:
1315 				so->so_rcv.sb_timeo = val;
1316 				break;
1317 			}
1318 			break;
1319 		default:
1320 			error = ENOPROTOOPT;
1321 			break;
1322 		}
1323 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1324 			(void) so_pr_ctloutput(so, sopt);
1325 		}
1326 	}
1327 bad:
1328 	return (error);
1329 }
1330 
1331 /* Helper routine for getsockopt */
1332 int
1333 sooptcopyout(struct sockopt *sopt, const void *buf, size_t len)
1334 {
1335 	int	error;
1336 	size_t	valsize;
1337 
1338 	error = 0;
1339 
1340 	/*
1341 	 * Documented get behavior is that we always return a value,
1342 	 * possibly truncated to fit in the user's buffer.
1343 	 * Traditional behavior is that we always tell the user
1344 	 * precisely how much we copied, rather than something useful
1345 	 * like the total amount we had available for her.
1346 	 * Note that this interface is not idempotent; the entire answer must
1347 	 * generated ahead of time.
1348 	 */
1349 	valsize = min(len, sopt->sopt_valsize);
1350 	sopt->sopt_valsize = valsize;
1351 	if (sopt->sopt_val != 0) {
1352 		if (sopt->sopt_td != NULL)
1353 			error = copyout(buf, sopt->sopt_val, valsize);
1354 		else
1355 			bcopy(buf, sopt->sopt_val, valsize);
1356 	}
1357 	return error;
1358 }
1359 
1360 int
1361 sogetopt(so, sopt)
1362 	struct socket *so;
1363 	struct sockopt *sopt;
1364 {
1365 	int	error, optval;
1366 	struct	linger l;
1367 	struct	timeval tv;
1368 	struct accept_filter_arg *afap;
1369 
1370 	error = 0;
1371 	if (sopt->sopt_level != SOL_SOCKET) {
1372 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1373 			return (so_pr_ctloutput(so, sopt));
1374 		} else
1375 			return (ENOPROTOOPT);
1376 	} else {
1377 		switch (sopt->sopt_name) {
1378 #ifdef INET
1379 		case SO_ACCEPTFILTER:
1380 			if ((so->so_options & SO_ACCEPTCONN) == 0)
1381 				return (EINVAL);
1382 			MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1383 				M_TEMP, M_WAITOK);
1384 			bzero(afap, sizeof(*afap));
1385 			if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1386 				strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1387 				if (so->so_accf->so_accept_filter_str != NULL)
1388 					strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1389 			}
1390 			error = sooptcopyout(sopt, afap, sizeof(*afap));
1391 			FREE(afap, M_TEMP);
1392 			break;
1393 #endif /* INET */
1394 
1395 		case SO_LINGER:
1396 			l.l_onoff = so->so_options & SO_LINGER;
1397 			l.l_linger = so->so_linger;
1398 			error = sooptcopyout(sopt, &l, sizeof l);
1399 			break;
1400 
1401 		case SO_USELOOPBACK:
1402 		case SO_DONTROUTE:
1403 		case SO_DEBUG:
1404 		case SO_KEEPALIVE:
1405 		case SO_REUSEADDR:
1406 		case SO_REUSEPORT:
1407 		case SO_BROADCAST:
1408 		case SO_OOBINLINE:
1409 		case SO_TIMESTAMP:
1410 			optval = so->so_options & sopt->sopt_name;
1411 integer:
1412 			error = sooptcopyout(sopt, &optval, sizeof optval);
1413 			break;
1414 
1415 		case SO_TYPE:
1416 			optval = so->so_type;
1417 			goto integer;
1418 
1419 		case SO_ERROR:
1420 			optval = so->so_error;
1421 			so->so_error = 0;
1422 			goto integer;
1423 
1424 		case SO_SNDBUF:
1425 			optval = so->so_snd.sb_hiwat;
1426 			goto integer;
1427 
1428 		case SO_RCVBUF:
1429 			optval = so->so_rcv.sb_hiwat;
1430 			goto integer;
1431 
1432 		case SO_SNDLOWAT:
1433 			optval = so->so_snd.sb_lowat;
1434 			goto integer;
1435 
1436 		case SO_RCVLOWAT:
1437 			optval = so->so_rcv.sb_lowat;
1438 			goto integer;
1439 
1440 		case SO_SNDTIMEO:
1441 		case SO_RCVTIMEO:
1442 			optval = (sopt->sopt_name == SO_SNDTIMEO ?
1443 				  so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1444 
1445 			tv.tv_sec = optval / hz;
1446 			tv.tv_usec = (optval % hz) * tick;
1447 			error = sooptcopyout(sopt, &tv, sizeof tv);
1448 			break;
1449 
1450 		default:
1451 			error = ENOPROTOOPT;
1452 			break;
1453 		}
1454 		return (error);
1455 	}
1456 }
1457 
1458 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1459 int
1460 soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1461 {
1462 	struct mbuf *m, *m_prev;
1463 	int sopt_size = sopt->sopt_valsize;
1464 
1465 	MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1466 	if (m == 0)
1467 		return ENOBUFS;
1468 	if (sopt_size > MLEN) {
1469 		MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1470 		if ((m->m_flags & M_EXT) == 0) {
1471 			m_free(m);
1472 			return ENOBUFS;
1473 		}
1474 		m->m_len = min(MCLBYTES, sopt_size);
1475 	} else {
1476 		m->m_len = min(MLEN, sopt_size);
1477 	}
1478 	sopt_size -= m->m_len;
1479 	*mp = m;
1480 	m_prev = m;
1481 
1482 	while (sopt_size) {
1483 		MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1484 		if (m == 0) {
1485 			m_freem(*mp);
1486 			return ENOBUFS;
1487 		}
1488 		if (sopt_size > MLEN) {
1489 			MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1490 			if ((m->m_flags & M_EXT) == 0) {
1491 				m_freem(*mp);
1492 				return ENOBUFS;
1493 			}
1494 			m->m_len = min(MCLBYTES, sopt_size);
1495 		} else {
1496 			m->m_len = min(MLEN, sopt_size);
1497 		}
1498 		sopt_size -= m->m_len;
1499 		m_prev->m_next = m;
1500 		m_prev = m;
1501 	}
1502 	return 0;
1503 }
1504 
1505 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1506 int
1507 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1508 {
1509 	struct mbuf *m0 = m;
1510 
1511 	if (sopt->sopt_val == NULL)
1512 		return 0;
1513 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1514 		if (sopt->sopt_td != NULL) {
1515 			int error;
1516 
1517 			error = copyin(sopt->sopt_val, mtod(m, char *),
1518 				       m->m_len);
1519 			if (error != 0) {
1520 				m_freem(m0);
1521 				return(error);
1522 			}
1523 		} else
1524 			bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1525 		sopt->sopt_valsize -= m->m_len;
1526 		(caddr_t)sopt->sopt_val += m->m_len;
1527 		m = m->m_next;
1528 	}
1529 	if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1530 		panic("ip6_sooptmcopyin");
1531 	return 0;
1532 }
1533 
1534 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1535 int
1536 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1537 {
1538 	struct mbuf *m0 = m;
1539 	size_t valsize = 0;
1540 
1541 	if (sopt->sopt_val == NULL)
1542 		return 0;
1543 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1544 		if (sopt->sopt_td != NULL) {
1545 			int error;
1546 
1547 			error = copyout(mtod(m, char *), sopt->sopt_val,
1548 				       m->m_len);
1549 			if (error != 0) {
1550 				m_freem(m0);
1551 				return(error);
1552 			}
1553 		} else
1554 			bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1555 	       sopt->sopt_valsize -= m->m_len;
1556 	       (caddr_t)sopt->sopt_val += m->m_len;
1557 	       valsize += m->m_len;
1558 	       m = m->m_next;
1559 	}
1560 	if (m != NULL) {
1561 		/* enough soopt buffer should be given from user-land */
1562 		m_freem(m0);
1563 		return(EINVAL);
1564 	}
1565 	sopt->sopt_valsize = valsize;
1566 	return 0;
1567 }
1568 
1569 void
1570 sohasoutofband(so)
1571 	struct socket *so;
1572 {
1573 	if (so->so_sigio != NULL)
1574 		pgsigio(so->so_sigio, SIGURG, 0);
1575 	selwakeup(&so->so_rcv.sb_sel);
1576 }
1577 
1578 int
1579 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td)
1580 {
1581 	int revents = 0;
1582 	int s = splnet();
1583 
1584 	if (events & (POLLIN | POLLRDNORM))
1585 		if (soreadable(so))
1586 			revents |= events & (POLLIN | POLLRDNORM);
1587 
1588 	if (events & POLLINIGNEOF)
1589 		if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat ||
1590 			!TAILQ_EMPTY(&so->so_comp) || so->so_error)
1591 			revents |= POLLINIGNEOF;
1592 
1593 	if (events & (POLLOUT | POLLWRNORM))
1594 		if (sowriteable(so))
1595 			revents |= events & (POLLOUT | POLLWRNORM);
1596 
1597 	if (events & (POLLPRI | POLLRDBAND))
1598 		if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1599 			revents |= events & (POLLPRI | POLLRDBAND);
1600 
1601 	if (revents == 0) {
1602 		if (events &
1603 			(POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM |
1604 			 POLLRDBAND)) {
1605 			selrecord(td, &so->so_rcv.sb_sel);
1606 			so->so_rcv.sb_flags |= SB_SEL;
1607 		}
1608 
1609 		if (events & (POLLOUT | POLLWRNORM)) {
1610 			selrecord(td, &so->so_snd.sb_sel);
1611 			so->so_snd.sb_flags |= SB_SEL;
1612 		}
1613 	}
1614 
1615 	splx(s);
1616 	return (revents);
1617 }
1618 
1619 int
1620 sokqfilter(struct file *fp, struct knote *kn)
1621 {
1622 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1623 	struct sockbuf *sb;
1624 	int s;
1625 
1626 	switch (kn->kn_filter) {
1627 	case EVFILT_READ:
1628 		if (so->so_options & SO_ACCEPTCONN)
1629 			kn->kn_fop = &solisten_filtops;
1630 		else
1631 			kn->kn_fop = &soread_filtops;
1632 		sb = &so->so_rcv;
1633 		break;
1634 	case EVFILT_WRITE:
1635 		kn->kn_fop = &sowrite_filtops;
1636 		sb = &so->so_snd;
1637 		break;
1638 	default:
1639 		return (1);
1640 	}
1641 
1642 	s = splnet();
1643 	SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1644 	sb->sb_flags |= SB_KNOTE;
1645 	splx(s);
1646 	return (0);
1647 }
1648 
1649 static void
1650 filt_sordetach(struct knote *kn)
1651 {
1652 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1653 	int s = splnet();
1654 
1655 	SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1656 	if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1657 		so->so_rcv.sb_flags &= ~SB_KNOTE;
1658 	splx(s);
1659 }
1660 
1661 /*ARGSUSED*/
1662 static int
1663 filt_soread(struct knote *kn, long hint)
1664 {
1665 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1666 
1667 	kn->kn_data = so->so_rcv.sb_cc;
1668 	if (so->so_state & SS_CANTRCVMORE) {
1669 		kn->kn_flags |= EV_EOF;
1670 		kn->kn_fflags = so->so_error;
1671 		return (1);
1672 	}
1673 	if (so->so_error)	/* temporary udp error */
1674 		return (1);
1675 	if (kn->kn_sfflags & NOTE_LOWAT)
1676 		return (kn->kn_data >= kn->kn_sdata);
1677 	return (kn->kn_data >= so->so_rcv.sb_lowat);
1678 }
1679 
1680 static void
1681 filt_sowdetach(struct knote *kn)
1682 {
1683 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1684 	int s = splnet();
1685 
1686 	SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1687 	if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1688 		so->so_snd.sb_flags &= ~SB_KNOTE;
1689 	splx(s);
1690 }
1691 
1692 /*ARGSUSED*/
1693 static int
1694 filt_sowrite(struct knote *kn, long hint)
1695 {
1696 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1697 
1698 	kn->kn_data = sbspace(&so->so_snd);
1699 	if (so->so_state & SS_CANTSENDMORE) {
1700 		kn->kn_flags |= EV_EOF;
1701 		kn->kn_fflags = so->so_error;
1702 		return (1);
1703 	}
1704 	if (so->so_error)	/* temporary udp error */
1705 		return (1);
1706 	if (((so->so_state & SS_ISCONNECTED) == 0) &&
1707 	    (so->so_proto->pr_flags & PR_CONNREQUIRED))
1708 		return (0);
1709 	if (kn->kn_sfflags & NOTE_LOWAT)
1710 		return (kn->kn_data >= kn->kn_sdata);
1711 	return (kn->kn_data >= so->so_snd.sb_lowat);
1712 }
1713 
1714 /*ARGSUSED*/
1715 static int
1716 filt_solisten(struct knote *kn, long hint)
1717 {
1718 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1719 
1720 	kn->kn_data = so->so_qlen;
1721 	return (! TAILQ_EMPTY(&so->so_comp));
1722 }
1723