xref: /freebsd/sys/kern/uipc_socket.c (revision 1a2cdef4962b47be5057809ce730a733b7f3c27c)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)uipc_socket.c	8.3 (Berkeley) 4/15/94
34  * $FreeBSD$
35  */
36 
37 #include "opt_inet.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/fcntl.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/domain.h>
45 #include <sys/file.h>			/* for struct knote */
46 #include <sys/kernel.h>
47 #include <sys/malloc.h>
48 #include <sys/event.h>
49 #include <sys/poll.h>
50 #include <sys/proc.h>
51 #include <sys/protosw.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/resourcevar.h>
55 #include <sys/signalvar.h>
56 #include <sys/sysctl.h>
57 #include <sys/uio.h>
58 #include <sys/jail.h>
59 #include <vm/vm_zone.h>
60 
61 #include <machine/limits.h>
62 
63 #ifdef INET
64 static int	 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
65 #endif
66 
67 static void 	filt_sordetach(struct knote *kn);
68 static int 	filt_soread(struct knote *kn, long hint);
69 static void 	filt_sowdetach(struct knote *kn);
70 static int	filt_sowrite(struct knote *kn, long hint);
71 static int	filt_solisten(struct knote *kn, long hint);
72 
73 static struct filterops solisten_filtops =
74 	{ 1, NULL, filt_sordetach, filt_solisten };
75 static struct filterops soread_filtops =
76 	{ 1, NULL, filt_sordetach, filt_soread };
77 static struct filterops sowrite_filtops =
78 	{ 1, NULL, filt_sowdetach, filt_sowrite };
79 
80 struct	vm_zone *socket_zone;
81 so_gen_t	so_gencnt;	/* generation count for sockets */
82 
83 MALLOC_DEFINE(M_SONAME, "soname", "socket name");
84 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");
85 
86 SYSCTL_DECL(_kern_ipc);
87 
88 static int somaxconn = SOMAXCONN;
89 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW,
90     &somaxconn, 0, "Maximum pending socket connection queue size");
91 
92 /*
93  * Socket operation routines.
94  * These routines are called by the routines in
95  * sys_socket.c or from a system process, and
96  * implement the semantics of socket operations by
97  * switching out to the protocol specific routines.
98  */
99 
100 /*
101  * Get a socket structure from our zone, and initialize it.
102  * We don't implement `waitok' yet (see comments in uipc_domain.c).
103  * Note that it would probably be better to allocate socket
104  * and PCB at the same time, but I'm not convinced that all
105  * the protocols can be easily modified to do this.
106  */
107 struct socket *
108 soalloc(waitok)
109 	int waitok;
110 {
111 	struct socket *so;
112 
113 	so = zalloc(socket_zone);
114 	if (so) {
115 		/* XXX race condition for reentrant kernel */
116 		bzero(so, sizeof *so);
117 		so->so_gencnt = ++so_gencnt;
118 		so->so_zone = socket_zone;
119 		TAILQ_INIT(&so->so_aiojobq);
120 	}
121 	return so;
122 }
123 
124 int
125 socreate(dom, aso, type, proto, p)
126 	int dom;
127 	struct socket **aso;
128 	register int type;
129 	int proto;
130 	struct proc *p;
131 {
132 	register struct protosw *prp;
133 	register struct socket *so;
134 	register 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 (jailed(p->p_ucred) && 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 = p->p_ucred;
161 	crhold(so->so_cred);
162 	so->so_proto = prp;
163 	error = (*prp->pr_usrreqs->pru_attach)(so, proto, p);
164 	if (error) {
165 		so->so_state |= SS_NOFDREF;
166 		sofree(so);
167 		return (error);
168 	}
169 	*aso = so;
170 	return (0);
171 }
172 
173 int
174 sobind(so, nam, p)
175 	struct socket *so;
176 	struct sockaddr *nam;
177 	struct proc *p;
178 {
179 	int s = splnet();
180 	int error;
181 
182 	error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, p);
183 	splx(s);
184 	return (error);
185 }
186 
187 void
188 sodealloc(so)
189 	struct socket *so;
190 {
191 
192 	so->so_gencnt = ++so_gencnt;
193 	if (so->so_rcv.sb_hiwat)
194 		(void)chgsbsize(so->so_cred->cr_uidinfo,
195 		    &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
196 	if (so->so_snd.sb_hiwat)
197 		(void)chgsbsize(so->so_cred->cr_uidinfo,
198 		    &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
199 #ifdef INET
200 	if (so->so_accf != NULL) {
201 		if (so->so_accf->so_accept_filter != NULL &&
202 			so->so_accf->so_accept_filter->accf_destroy != NULL) {
203 			so->so_accf->so_accept_filter->accf_destroy(so);
204 		}
205 		if (so->so_accf->so_accept_filter_str != NULL)
206 			FREE(so->so_accf->so_accept_filter_str, M_ACCF);
207 		FREE(so->so_accf, M_ACCF);
208 	}
209 #endif
210 	crfree(so->so_cred);
211 	zfree(so->so_zone, so);
212 }
213 
214 int
215 solisten(so, backlog, p)
216 	register struct socket *so;
217 	int backlog;
218 	struct proc *p;
219 {
220 	int s, error;
221 
222 	s = splnet();
223 	error = (*so->so_proto->pr_usrreqs->pru_listen)(so, p);
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(so)
239 	register struct socket *so;
240 {
241 	struct socket *head = so->so_head;
242 
243 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
244 		return;
245 	if (head != NULL) {
246 		if (so->so_state & SS_INCOMP) {
247 			TAILQ_REMOVE(&head->so_incomp, so, so_list);
248 			head->so_incqlen--;
249 		} else if (so->so_state & SS_COMP) {
250 			/*
251 			 * We must not decommission a socket that's
252 			 * on the accept(2) queue.  If we do, then
253 			 * accept(2) may hang after select(2) indicated
254 			 * that the listening socket was ready.
255 			 */
256 			return;
257 		} else {
258 			panic("sofree: not queued");
259 		}
260 		head->so_qlen--;
261 		so->so_state &= ~SS_INCOMP;
262 		so->so_head = NULL;
263 	}
264 	sbrelease(&so->so_snd, so);
265 	sorflush(so);
266 	sodealloc(so);
267 }
268 
269 /*
270  * Close a socket on last file table reference removal.
271  * Initiate disconnect if connected.
272  * Free socket when disconnect complete.
273  */
274 int
275 soclose(so)
276 	register struct socket *so;
277 {
278 	int s = splnet();		/* conservative */
279 	int error = 0;
280 
281 	funsetown(so->so_sigio);
282 	if (so->so_options & SO_ACCEPTCONN) {
283 		struct socket *sp, *sonext;
284 
285 		sp = TAILQ_FIRST(&so->so_incomp);
286 		for (; sp != NULL; sp = sonext) {
287 			sonext = TAILQ_NEXT(sp, so_list);
288 			(void) soabort(sp);
289 		}
290 		for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
291 			sonext = TAILQ_NEXT(sp, so_list);
292 			/* Dequeue from so_comp since sofree() won't do it */
293 			TAILQ_REMOVE(&so->so_comp, sp, so_list);
294 			so->so_qlen--;
295 			sp->so_state &= ~SS_COMP;
296 			sp->so_head = NULL;
297 			(void) soabort(sp);
298 		}
299 	}
300 	if (so->so_pcb == 0)
301 		goto discard;
302 	if (so->so_state & SS_ISCONNECTED) {
303 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
304 			error = sodisconnect(so);
305 			if (error)
306 				goto drop;
307 		}
308 		if (so->so_options & SO_LINGER) {
309 			if ((so->so_state & SS_ISDISCONNECTING) &&
310 			    (so->so_state & SS_NBIO))
311 				goto drop;
312 			while (so->so_state & SS_ISCONNECTED) {
313 				error = tsleep((caddr_t)&so->so_timeo,
314 				    PSOCK | PCATCH, "soclos", so->so_linger * hz);
315 				if (error)
316 					break;
317 			}
318 		}
319 	}
320 drop:
321 	if (so->so_pcb) {
322 		int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so);
323 		if (error == 0)
324 			error = error2;
325 	}
326 discard:
327 	if (so->so_state & SS_NOFDREF)
328 		panic("soclose: NOFDREF");
329 	so->so_state |= SS_NOFDREF;
330 	sofree(so);
331 	splx(s);
332 	return (error);
333 }
334 
335 /*
336  * Must be called at splnet...
337  */
338 int
339 soabort(so)
340 	struct socket *so;
341 {
342 	int error;
343 
344 	error = (*so->so_proto->pr_usrreqs->pru_abort)(so);
345 	if (error) {
346 		sofree(so);
347 		return error;
348 	}
349 	return (0);
350 }
351 
352 int
353 soaccept(so, nam)
354 	register struct socket *so;
355 	struct sockaddr **nam;
356 {
357 	int s = splnet();
358 	int error;
359 
360 	if ((so->so_state & SS_NOFDREF) == 0)
361 		panic("soaccept: !NOFDREF");
362 	so->so_state &= ~SS_NOFDREF;
363 	error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam);
364 	splx(s);
365 	return (error);
366 }
367 
368 int
369 soconnect(so, nam, p)
370 	register struct socket *so;
371 	struct sockaddr *nam;
372 	struct proc *p;
373 {
374 	int s;
375 	int error;
376 
377 	if (so->so_options & SO_ACCEPTCONN)
378 		return (EOPNOTSUPP);
379 	s = splnet();
380 	/*
381 	 * If protocol is connection-based, can only connect once.
382 	 * Otherwise, if connected, try to disconnect first.
383 	 * This allows user to disconnect by connecting to, e.g.,
384 	 * a null address.
385 	 */
386 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
387 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
388 	    (error = sodisconnect(so))))
389 		error = EISCONN;
390 	else
391 		error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, p);
392 	splx(s);
393 	return (error);
394 }
395 
396 int
397 soconnect2(so1, so2)
398 	register struct socket *so1;
399 	struct socket *so2;
400 {
401 	int s = splnet();
402 	int error;
403 
404 	error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2);
405 	splx(s);
406 	return (error);
407 }
408 
409 int
410 sodisconnect(so)
411 	register struct socket *so;
412 {
413 	int s = splnet();
414 	int error;
415 
416 	if ((so->so_state & SS_ISCONNECTED) == 0) {
417 		error = ENOTCONN;
418 		goto bad;
419 	}
420 	if (so->so_state & SS_ISDISCONNECTING) {
421 		error = EALREADY;
422 		goto bad;
423 	}
424 	error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so);
425 bad:
426 	splx(s);
427 	return (error);
428 }
429 
430 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
431 /*
432  * Send on a socket.
433  * If send must go all at once and message is larger than
434  * send buffering, then hard error.
435  * Lock against other senders.
436  * If must go all at once and not enough room now, then
437  * inform user that this would block and do nothing.
438  * Otherwise, if nonblocking, send as much as possible.
439  * The data to be sent is described by "uio" if nonzero,
440  * otherwise by the mbuf chain "top" (which must be null
441  * if uio is not).  Data provided in mbuf chain must be small
442  * enough to send all at once.
443  *
444  * Returns nonzero on error, timeout or signal; callers
445  * must check for short counts if EINTR/ERESTART are returned.
446  * Data and control buffers are freed on return.
447  */
448 int
449 sosend(so, addr, uio, top, control, flags, p)
450 	register struct socket *so;
451 	struct sockaddr *addr;
452 	struct uio *uio;
453 	struct mbuf *top;
454 	struct mbuf *control;
455 	int flags;
456 	struct proc *p;
457 {
458 	struct mbuf **mp;
459 	register struct mbuf *m;
460 	register long space, len, resid;
461 	int clen = 0, error, s, dontroute, mlen;
462 	int atomic = sosendallatonce(so) || top;
463 
464 	if (uio)
465 		resid = uio->uio_resid;
466 	else
467 		resid = top->m_pkthdr.len;
468 	/*
469 	 * In theory resid should be unsigned.
470 	 * However, space must be signed, as it might be less than 0
471 	 * if we over-committed, and we must use a signed comparison
472 	 * of space and resid.  On the other hand, a negative resid
473 	 * causes us to loop sending 0-length segments to the protocol.
474 	 *
475 	 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
476 	 * type sockets since that's an error.
477 	 */
478 	if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
479 		error = EINVAL;
480 		goto out;
481 	}
482 
483 	dontroute =
484 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
485 	    (so->so_proto->pr_flags & PR_ATOMIC);
486 	if (p)
487 		p->p_stats->p_ru.ru_msgsnd++;
488 	if (control)
489 		clen = control->m_len;
490 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
491 
492 restart:
493 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
494 	if (error)
495 		goto out;
496 	do {
497 		s = splnet();
498 		if (so->so_state & SS_CANTSENDMORE)
499 			snderr(EPIPE);
500 		if (so->so_error) {
501 			error = so->so_error;
502 			so->so_error = 0;
503 			splx(s);
504 			goto release;
505 		}
506 		if ((so->so_state & SS_ISCONNECTED) == 0) {
507 			/*
508 			 * `sendto' and `sendmsg' is allowed on a connection-
509 			 * based socket if it supports implied connect.
510 			 * Return ENOTCONN if not connected and no address is
511 			 * supplied.
512 			 */
513 			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
514 			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
515 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
516 				    !(resid == 0 && clen != 0))
517 					snderr(ENOTCONN);
518 			} else if (addr == 0)
519 			    snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
520 				   ENOTCONN : EDESTADDRREQ);
521 		}
522 		space = sbspace(&so->so_snd);
523 		if (flags & MSG_OOB)
524 			space += 1024;
525 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
526 		    clen > so->so_snd.sb_hiwat)
527 			snderr(EMSGSIZE);
528 		if (space < resid + clen && uio &&
529 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
530 			if (so->so_state & SS_NBIO)
531 				snderr(EWOULDBLOCK);
532 			sbunlock(&so->so_snd);
533 			error = sbwait(&so->so_snd);
534 			splx(s);
535 			if (error)
536 				goto out;
537 			goto restart;
538 		}
539 		splx(s);
540 		mp = &top;
541 		space -= clen;
542 		do {
543 		    if (uio == NULL) {
544 			/*
545 			 * Data is prepackaged in "top".
546 			 */
547 			resid = 0;
548 			if (flags & MSG_EOR)
549 				top->m_flags |= M_EOR;
550 		    } else do {
551 			if (top == 0) {
552 				MGETHDR(m, M_TRYWAIT, MT_DATA);
553 				if (m == NULL) {
554 					error = ENOBUFS;
555 					goto release;
556 				}
557 				mlen = MHLEN;
558 				m->m_pkthdr.len = 0;
559 				m->m_pkthdr.rcvif = (struct ifnet *)0;
560 			} else {
561 				MGET(m, M_TRYWAIT, MT_DATA);
562 				if (m == NULL) {
563 					error = ENOBUFS;
564 					goto release;
565 				}
566 				mlen = MLEN;
567 			}
568 			if (resid >= MINCLSIZE) {
569 				MCLGET(m, M_TRYWAIT);
570 				if ((m->m_flags & M_EXT) == 0)
571 					goto nopages;
572 				mlen = MCLBYTES;
573 				len = min(min(mlen, resid), space);
574 			} else {
575 nopages:
576 				len = min(min(mlen, resid), space);
577 				/*
578 				 * For datagram protocols, leave room
579 				 * for protocol headers in first mbuf.
580 				 */
581 				if (atomic && top == 0 && len < mlen)
582 					MH_ALIGN(m, len);
583 			}
584 			space -= len;
585 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
586 			resid = uio->uio_resid;
587 			m->m_len = len;
588 			*mp = m;
589 			top->m_pkthdr.len += len;
590 			if (error)
591 				goto release;
592 			mp = &m->m_next;
593 			if (resid <= 0) {
594 				if (flags & MSG_EOR)
595 					top->m_flags |= M_EOR;
596 				break;
597 			}
598 		    } while (space > 0 && atomic);
599 		    if (dontroute)
600 			    so->so_options |= SO_DONTROUTE;
601 		    s = splnet();				/* XXX */
602 		    /*
603 		     * XXX all the SS_CANTSENDMORE checks previously
604 		     * done could be out of date.  We could have recieved
605 		     * a reset packet in an interrupt or maybe we slept
606 		     * while doing page faults in uiomove() etc. We could
607 		     * probably recheck again inside the splnet() protection
608 		     * here, but there are probably other places that this
609 		     * also happens.  We must rethink this.
610 		     */
611 		    error = (*so->so_proto->pr_usrreqs->pru_send)(so,
612 			(flags & MSG_OOB) ? PRUS_OOB :
613 			/*
614 			 * If the user set MSG_EOF, the protocol
615 			 * understands this flag and nothing left to
616 			 * send then use PRU_SEND_EOF instead of PRU_SEND.
617 			 */
618 			((flags & MSG_EOF) &&
619 			 (so->so_proto->pr_flags & PR_IMPLOPCL) &&
620 			 (resid <= 0)) ?
621 				PRUS_EOF :
622 			/* If there is more to send set PRUS_MORETOCOME */
623 			(resid > 0 && space > 0) ? PRUS_MORETOCOME : 0,
624 			top, addr, control, p);
625 		    splx(s);
626 		    if (dontroute)
627 			    so->so_options &= ~SO_DONTROUTE;
628 		    clen = 0;
629 		    control = 0;
630 		    top = 0;
631 		    mp = &top;
632 		    if (error)
633 			goto release;
634 		} while (resid && space > 0);
635 	} while (resid);
636 
637 release:
638 	sbunlock(&so->so_snd);
639 out:
640 	if (top)
641 		m_freem(top);
642 	if (control)
643 		m_freem(control);
644 	return (error);
645 }
646 
647 /*
648  * Implement receive operations on a socket.
649  * We depend on the way that records are added to the sockbuf
650  * by sbappend*.  In particular, each record (mbufs linked through m_next)
651  * must begin with an address if the protocol so specifies,
652  * followed by an optional mbuf or mbufs containing ancillary data,
653  * and then zero or more mbufs of data.
654  * In order to avoid blocking network interrupts for the entire time here,
655  * we splx() while doing the actual copy to user space.
656  * Although the sockbuf is locked, new data may still be appended,
657  * and thus we must maintain consistency of the sockbuf during that time.
658  *
659  * The caller may receive the data as a single mbuf chain by supplying
660  * an mbuf **mp0 for use in returning the chain.  The uio is then used
661  * only for the count in uio_resid.
662  */
663 int
664 soreceive(so, psa, uio, mp0, controlp, flagsp)
665 	register struct socket *so;
666 	struct sockaddr **psa;
667 	struct uio *uio;
668 	struct mbuf **mp0;
669 	struct mbuf **controlp;
670 	int *flagsp;
671 {
672 	register struct mbuf *m, **mp;
673 	register int flags, len, error, s, offset;
674 	struct protosw *pr = so->so_proto;
675 	struct mbuf *nextrecord;
676 	int moff, type = 0;
677 	int orig_resid = uio->uio_resid;
678 
679 	mp = mp0;
680 	if (psa)
681 		*psa = 0;
682 	if (controlp)
683 		*controlp = 0;
684 	if (flagsp)
685 		flags = *flagsp &~ MSG_EOR;
686 	else
687 		flags = 0;
688 	if (flags & MSG_OOB) {
689 		m = m_get(M_TRYWAIT, MT_DATA);
690 		if (m == NULL)
691 			return (ENOBUFS);
692 		error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK);
693 		if (error)
694 			goto bad;
695 		do {
696 			error = uiomove(mtod(m, caddr_t),
697 			    (int) min(uio->uio_resid, m->m_len), uio);
698 			m = m_free(m);
699 		} while (uio->uio_resid && error == 0 && m);
700 bad:
701 		if (m)
702 			m_freem(m);
703 		return (error);
704 	}
705 	if (mp)
706 		*mp = (struct mbuf *)0;
707 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
708 		(*pr->pr_usrreqs->pru_rcvd)(so, 0);
709 
710 restart:
711 	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
712 	if (error)
713 		return (error);
714 	s = splnet();
715 
716 	m = so->so_rcv.sb_mb;
717 	/*
718 	 * If we have less data than requested, block awaiting more
719 	 * (subject to any timeout) if:
720 	 *   1. the current count is less than the low water mark, or
721 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
722 	 *	receive operation at once if we block (resid <= hiwat).
723 	 *   3. MSG_DONTWAIT is not set
724 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
725 	 * we have to do the receive in sections, and thus risk returning
726 	 * a short count if a timeout or signal occurs after we start.
727 	 */
728 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
729 	    so->so_rcv.sb_cc < uio->uio_resid) &&
730 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
731 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
732 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
733 		KASSERT(m != 0 || !so->so_rcv.sb_cc, ("receive 1"));
734 		if (so->so_error) {
735 			if (m)
736 				goto dontblock;
737 			error = so->so_error;
738 			if ((flags & MSG_PEEK) == 0)
739 				so->so_error = 0;
740 			goto release;
741 		}
742 		if (so->so_state & SS_CANTRCVMORE) {
743 			if (m)
744 				goto dontblock;
745 			else
746 				goto release;
747 		}
748 		for (; m; m = m->m_next)
749 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
750 				m = so->so_rcv.sb_mb;
751 				goto dontblock;
752 			}
753 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
754 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
755 			error = ENOTCONN;
756 			goto release;
757 		}
758 		if (uio->uio_resid == 0)
759 			goto release;
760 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
761 			error = EWOULDBLOCK;
762 			goto release;
763 		}
764 		sbunlock(&so->so_rcv);
765 		error = sbwait(&so->so_rcv);
766 		splx(s);
767 		if (error)
768 			return (error);
769 		goto restart;
770 	}
771 dontblock:
772 	if (uio->uio_procp)
773 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
774 	nextrecord = m->m_nextpkt;
775 	if (pr->pr_flags & PR_ADDR) {
776 		KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
777 		orig_resid = 0;
778 		if (psa)
779 			*psa = dup_sockaddr(mtod(m, struct sockaddr *),
780 					    mp0 == 0);
781 		if (flags & MSG_PEEK) {
782 			m = m->m_next;
783 		} else {
784 			sbfree(&so->so_rcv, m);
785 			MFREE(m, so->so_rcv.sb_mb);
786 			m = so->so_rcv.sb_mb;
787 		}
788 	}
789 	while (m && m->m_type == MT_CONTROL && error == 0) {
790 		if (flags & MSG_PEEK) {
791 			if (controlp)
792 				*controlp = m_copy(m, 0, m->m_len);
793 			m = m->m_next;
794 		} else {
795 			sbfree(&so->so_rcv, m);
796 			if (controlp) {
797 				if (pr->pr_domain->dom_externalize &&
798 				    mtod(m, struct cmsghdr *)->cmsg_type ==
799 				    SCM_RIGHTS)
800 				   error = (*pr->pr_domain->dom_externalize)(m);
801 				*controlp = m;
802 				so->so_rcv.sb_mb = m->m_next;
803 				m->m_next = 0;
804 				m = so->so_rcv.sb_mb;
805 			} else {
806 				MFREE(m, so->so_rcv.sb_mb);
807 				m = so->so_rcv.sb_mb;
808 			}
809 		}
810 		if (controlp) {
811 			orig_resid = 0;
812 			controlp = &(*controlp)->m_next;
813 		}
814 	}
815 	if (m) {
816 		if ((flags & MSG_PEEK) == 0)
817 			m->m_nextpkt = nextrecord;
818 		type = m->m_type;
819 		if (type == MT_OOBDATA)
820 			flags |= MSG_OOB;
821 	}
822 	moff = 0;
823 	offset = 0;
824 	while (m && uio->uio_resid > 0 && error == 0) {
825 		if (m->m_type == MT_OOBDATA) {
826 			if (type != MT_OOBDATA)
827 				break;
828 		} else if (type == MT_OOBDATA)
829 			break;
830 		else
831 		    KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
832 			("receive 3"));
833 		so->so_state &= ~SS_RCVATMARK;
834 		len = uio->uio_resid;
835 		if (so->so_oobmark && len > so->so_oobmark - offset)
836 			len = so->so_oobmark - offset;
837 		if (len > m->m_len - moff)
838 			len = m->m_len - moff;
839 		/*
840 		 * If mp is set, just pass back the mbufs.
841 		 * Otherwise copy them out via the uio, then free.
842 		 * Sockbuf must be consistent here (points to current mbuf,
843 		 * it points to next record) when we drop priority;
844 		 * we must note any additions to the sockbuf when we
845 		 * block interrupts again.
846 		 */
847 		if (mp == 0) {
848 			splx(s);
849 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
850 			s = splnet();
851 			if (error)
852 				goto release;
853 		} else
854 			uio->uio_resid -= len;
855 		if (len == m->m_len - moff) {
856 			if (m->m_flags & M_EOR)
857 				flags |= MSG_EOR;
858 			if (flags & MSG_PEEK) {
859 				m = m->m_next;
860 				moff = 0;
861 			} else {
862 				nextrecord = m->m_nextpkt;
863 				sbfree(&so->so_rcv, m);
864 				if (mp) {
865 					*mp = m;
866 					mp = &m->m_next;
867 					so->so_rcv.sb_mb = m = m->m_next;
868 					*mp = (struct mbuf *)0;
869 				} else {
870 					MFREE(m, so->so_rcv.sb_mb);
871 					m = so->so_rcv.sb_mb;
872 				}
873 				if (m)
874 					m->m_nextpkt = nextrecord;
875 			}
876 		} else {
877 			if (flags & MSG_PEEK)
878 				moff += len;
879 			else {
880 				if (mp)
881 					*mp = m_copym(m, 0, len, M_TRYWAIT);
882 				m->m_data += len;
883 				m->m_len -= len;
884 				so->so_rcv.sb_cc -= len;
885 			}
886 		}
887 		if (so->so_oobmark) {
888 			if ((flags & MSG_PEEK) == 0) {
889 				so->so_oobmark -= len;
890 				if (so->so_oobmark == 0) {
891 					so->so_state |= SS_RCVATMARK;
892 					break;
893 				}
894 			} else {
895 				offset += len;
896 				if (offset == so->so_oobmark)
897 					break;
898 			}
899 		}
900 		if (flags & MSG_EOR)
901 			break;
902 		/*
903 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
904 		 * we must not quit until "uio->uio_resid == 0" or an error
905 		 * termination.  If a signal/timeout occurs, return
906 		 * with a short count but without error.
907 		 * Keep sockbuf locked against other readers.
908 		 */
909 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
910 		    !sosendallatonce(so) && !nextrecord) {
911 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
912 				break;
913 			/*
914 			 * Notify the protocol that some data has been
915 			 * drained before blocking.
916 			 */
917 			if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
918 				(*pr->pr_usrreqs->pru_rcvd)(so, flags);
919 			error = sbwait(&so->so_rcv);
920 			if (error) {
921 				sbunlock(&so->so_rcv);
922 				splx(s);
923 				return (0);
924 			}
925 			m = so->so_rcv.sb_mb;
926 			if (m)
927 				nextrecord = m->m_nextpkt;
928 		}
929 	}
930 
931 	if (m && pr->pr_flags & PR_ATOMIC) {
932 		flags |= MSG_TRUNC;
933 		if ((flags & MSG_PEEK) == 0)
934 			(void) sbdroprecord(&so->so_rcv);
935 	}
936 	if ((flags & MSG_PEEK) == 0) {
937 		if (m == 0)
938 			so->so_rcv.sb_mb = nextrecord;
939 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
940 			(*pr->pr_usrreqs->pru_rcvd)(so, flags);
941 	}
942 	if (orig_resid == uio->uio_resid && orig_resid &&
943 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
944 		sbunlock(&so->so_rcv);
945 		splx(s);
946 		goto restart;
947 	}
948 
949 	if (flagsp)
950 		*flagsp |= flags;
951 release:
952 	sbunlock(&so->so_rcv);
953 	splx(s);
954 	return (error);
955 }
956 
957 int
958 soshutdown(so, how)
959 	register struct socket *so;
960 	register int how;
961 {
962 	register struct protosw *pr = so->so_proto;
963 
964 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
965 		return (EINVAL);
966 
967 	if (how != SHUT_WR)
968 		sorflush(so);
969 	if (how != SHUT_RD)
970 		return ((*pr->pr_usrreqs->pru_shutdown)(so));
971 	return (0);
972 }
973 
974 void
975 sorflush(so)
976 	register struct socket *so;
977 {
978 	register struct sockbuf *sb = &so->so_rcv;
979 	register struct protosw *pr = so->so_proto;
980 	register int s;
981 	struct sockbuf asb;
982 
983 	sb->sb_flags |= SB_NOINTR;
984 	(void) sblock(sb, M_WAITOK);
985 	s = splimp();
986 	socantrcvmore(so);
987 	sbunlock(sb);
988 	asb = *sb;
989 	bzero((caddr_t)sb, sizeof (*sb));
990 	splx(s);
991 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
992 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
993 	sbrelease(&asb, so);
994 }
995 
996 #ifdef INET
997 static int
998 do_setopt_accept_filter(so, sopt)
999 	struct	socket *so;
1000 	struct	sockopt *sopt;
1001 {
1002 	struct accept_filter_arg	*afap = NULL;
1003 	struct accept_filter	*afp;
1004 	struct so_accf	*af = so->so_accf;
1005 	int	error = 0;
1006 
1007 	/* do not set/remove accept filters on non listen sockets */
1008 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
1009 		error = EINVAL;
1010 		goto out;
1011 	}
1012 
1013 	/* removing the filter */
1014 	if (sopt == NULL) {
1015 		if (af != NULL) {
1016 			if (af->so_accept_filter != NULL &&
1017 				af->so_accept_filter->accf_destroy != NULL) {
1018 				af->so_accept_filter->accf_destroy(so);
1019 			}
1020 			if (af->so_accept_filter_str != NULL) {
1021 				FREE(af->so_accept_filter_str, M_ACCF);
1022 			}
1023 			FREE(af, M_ACCF);
1024 			so->so_accf = NULL;
1025 		}
1026 		so->so_options &= ~SO_ACCEPTFILTER;
1027 		return (0);
1028 	}
1029 	/* adding a filter */
1030 	/* must remove previous filter first */
1031 	if (af != NULL) {
1032 		error = EINVAL;
1033 		goto out;
1034 	}
1035 	/* don't put large objects on the kernel stack */
1036 	MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1037 	error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1038 	afap->af_name[sizeof(afap->af_name)-1] = '\0';
1039 	afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1040 	if (error)
1041 		goto out;
1042 	afp = accept_filt_get(afap->af_name);
1043 	if (afp == NULL) {
1044 		error = ENOENT;
1045 		goto out;
1046 	}
1047 	MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO);
1048 	if (afp->accf_create != NULL) {
1049 		if (afap->af_name[0] != '\0') {
1050 			int len = strlen(afap->af_name) + 1;
1051 
1052 			MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1053 			strcpy(af->so_accept_filter_str, afap->af_name);
1054 		}
1055 		af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1056 		if (af->so_accept_filter_arg == NULL) {
1057 			FREE(af->so_accept_filter_str, M_ACCF);
1058 			FREE(af, M_ACCF);
1059 			so->so_accf = NULL;
1060 			error = EINVAL;
1061 			goto out;
1062 		}
1063 	}
1064 	af->so_accept_filter = afp;
1065 	so->so_accf = af;
1066 	so->so_options |= SO_ACCEPTFILTER;
1067 out:
1068 	if (afap != NULL)
1069 		FREE(afap, M_TEMP);
1070 	return (error);
1071 }
1072 #endif /* INET */
1073 
1074 /*
1075  * Perhaps this routine, and sooptcopyout(), below, ought to come in
1076  * an additional variant to handle the case where the option value needs
1077  * to be some kind of integer, but not a specific size.
1078  * In addition to their use here, these functions are also called by the
1079  * protocol-level pr_ctloutput() routines.
1080  */
1081 int
1082 sooptcopyin(sopt, buf, len, minlen)
1083 	struct	sockopt *sopt;
1084 	void	*buf;
1085 	size_t	len;
1086 	size_t	minlen;
1087 {
1088 	size_t	valsize;
1089 
1090 	/*
1091 	 * If the user gives us more than we wanted, we ignore it,
1092 	 * but if we don't get the minimum length the caller
1093 	 * wants, we return EINVAL.  On success, sopt->sopt_valsize
1094 	 * is set to however much we actually retrieved.
1095 	 */
1096 	if ((valsize = sopt->sopt_valsize) < minlen)
1097 		return EINVAL;
1098 	if (valsize > len)
1099 		sopt->sopt_valsize = valsize = len;
1100 
1101 	if (sopt->sopt_p != 0)
1102 		return (copyin(sopt->sopt_val, buf, valsize));
1103 
1104 	bcopy(sopt->sopt_val, buf, valsize);
1105 	return 0;
1106 }
1107 
1108 int
1109 sosetopt(so, sopt)
1110 	struct socket *so;
1111 	struct sockopt *sopt;
1112 {
1113 	int	error, optval;
1114 	struct	linger l;
1115 	struct	timeval tv;
1116 	u_long  val;
1117 
1118 	error = 0;
1119 	if (sopt->sopt_level != SOL_SOCKET) {
1120 		if (so->so_proto && so->so_proto->pr_ctloutput)
1121 			return ((*so->so_proto->pr_ctloutput)
1122 				  (so, sopt));
1123 		error = ENOPROTOOPT;
1124 	} else {
1125 		switch (sopt->sopt_name) {
1126 #ifdef INET
1127 		case SO_ACCEPTFILTER:
1128 			error = do_setopt_accept_filter(so, sopt);
1129 			if (error)
1130 				goto bad;
1131 			break;
1132 #endif
1133 		case SO_LINGER:
1134 			error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1135 			if (error)
1136 				goto bad;
1137 
1138 			so->so_linger = l.l_linger;
1139 			if (l.l_onoff)
1140 				so->so_options |= SO_LINGER;
1141 			else
1142 				so->so_options &= ~SO_LINGER;
1143 			break;
1144 
1145 		case SO_DEBUG:
1146 		case SO_KEEPALIVE:
1147 		case SO_DONTROUTE:
1148 		case SO_USELOOPBACK:
1149 		case SO_BROADCAST:
1150 		case SO_REUSEADDR:
1151 		case SO_REUSEPORT:
1152 		case SO_OOBINLINE:
1153 		case SO_TIMESTAMP:
1154 			error = sooptcopyin(sopt, &optval, sizeof optval,
1155 					    sizeof optval);
1156 			if (error)
1157 				goto bad;
1158 			if (optval)
1159 				so->so_options |= sopt->sopt_name;
1160 			else
1161 				so->so_options &= ~sopt->sopt_name;
1162 			break;
1163 
1164 		case SO_SNDBUF:
1165 		case SO_RCVBUF:
1166 		case SO_SNDLOWAT:
1167 		case SO_RCVLOWAT:
1168 			error = sooptcopyin(sopt, &optval, sizeof optval,
1169 					    sizeof optval);
1170 			if (error)
1171 				goto bad;
1172 
1173 			/*
1174 			 * Values < 1 make no sense for any of these
1175 			 * options, so disallow them.
1176 			 */
1177 			if (optval < 1) {
1178 				error = EINVAL;
1179 				goto bad;
1180 			}
1181 
1182 			switch (sopt->sopt_name) {
1183 			case SO_SNDBUF:
1184 			case SO_RCVBUF:
1185 				if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1186 				    &so->so_snd : &so->so_rcv, (u_long)optval,
1187 				    so, curproc) == 0) {
1188 					error = ENOBUFS;
1189 					goto bad;
1190 				}
1191 				break;
1192 
1193 			/*
1194 			 * Make sure the low-water is never greater than
1195 			 * the high-water.
1196 			 */
1197 			case SO_SNDLOWAT:
1198 				so->so_snd.sb_lowat =
1199 				    (optval > so->so_snd.sb_hiwat) ?
1200 				    so->so_snd.sb_hiwat : optval;
1201 				break;
1202 			case SO_RCVLOWAT:
1203 				so->so_rcv.sb_lowat =
1204 				    (optval > so->so_rcv.sb_hiwat) ?
1205 				    so->so_rcv.sb_hiwat : optval;
1206 				break;
1207 			}
1208 			break;
1209 
1210 		case SO_SNDTIMEO:
1211 		case SO_RCVTIMEO:
1212 			error = sooptcopyin(sopt, &tv, sizeof tv,
1213 					    sizeof tv);
1214 			if (error)
1215 				goto bad;
1216 
1217 			/* assert(hz > 0); */
1218 			if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1219 			    tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1220 				error = EDOM;
1221 				goto bad;
1222 			}
1223 			/* assert(tick > 0); */
1224 			/* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1225 			val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1226 			if (val > SHRT_MAX) {
1227 				error = EDOM;
1228 				goto bad;
1229 			}
1230 
1231 			switch (sopt->sopt_name) {
1232 			case SO_SNDTIMEO:
1233 				so->so_snd.sb_timeo = val;
1234 				break;
1235 			case SO_RCVTIMEO:
1236 				so->so_rcv.sb_timeo = val;
1237 				break;
1238 			}
1239 			break;
1240 		default:
1241 			error = ENOPROTOOPT;
1242 			break;
1243 		}
1244 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1245 			(void) ((*so->so_proto->pr_ctloutput)
1246 				  (so, sopt));
1247 		}
1248 	}
1249 bad:
1250 	return (error);
1251 }
1252 
1253 /* Helper routine for getsockopt */
1254 int
1255 sooptcopyout(sopt, buf, len)
1256 	struct	sockopt *sopt;
1257 	void	*buf;
1258 	size_t	len;
1259 {
1260 	int	error;
1261 	size_t	valsize;
1262 
1263 	error = 0;
1264 
1265 	/*
1266 	 * Documented get behavior is that we always return a value,
1267 	 * possibly truncated to fit in the user's buffer.
1268 	 * Traditional behavior is that we always tell the user
1269 	 * precisely how much we copied, rather than something useful
1270 	 * like the total amount we had available for her.
1271 	 * Note that this interface is not idempotent; the entire answer must
1272 	 * generated ahead of time.
1273 	 */
1274 	valsize = min(len, sopt->sopt_valsize);
1275 	sopt->sopt_valsize = valsize;
1276 	if (sopt->sopt_val != 0) {
1277 		if (sopt->sopt_p != 0)
1278 			error = copyout(buf, sopt->sopt_val, valsize);
1279 		else
1280 			bcopy(buf, sopt->sopt_val, valsize);
1281 	}
1282 	return error;
1283 }
1284 
1285 int
1286 sogetopt(so, sopt)
1287 	struct socket *so;
1288 	struct sockopt *sopt;
1289 {
1290 	int	error, optval;
1291 	struct	linger l;
1292 	struct	timeval tv;
1293 #ifdef INET
1294 	struct accept_filter_arg *afap;
1295 #endif
1296 
1297 	error = 0;
1298 	if (sopt->sopt_level != SOL_SOCKET) {
1299 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1300 			return ((*so->so_proto->pr_ctloutput)
1301 				  (so, sopt));
1302 		} else
1303 			return (ENOPROTOOPT);
1304 	} else {
1305 		switch (sopt->sopt_name) {
1306 #ifdef INET
1307 		case SO_ACCEPTFILTER:
1308 			if ((so->so_options & SO_ACCEPTCONN) == 0)
1309 				return (EINVAL);
1310 			MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1311 				M_TEMP, M_WAITOK | M_ZERO);
1312 			if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1313 				strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1314 				if (so->so_accf->so_accept_filter_str != NULL)
1315 					strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1316 			}
1317 			error = sooptcopyout(sopt, afap, sizeof(*afap));
1318 			FREE(afap, M_TEMP);
1319 			break;
1320 #endif
1321 
1322 		case SO_LINGER:
1323 			l.l_onoff = so->so_options & SO_LINGER;
1324 			l.l_linger = so->so_linger;
1325 			error = sooptcopyout(sopt, &l, sizeof l);
1326 			break;
1327 
1328 		case SO_USELOOPBACK:
1329 		case SO_DONTROUTE:
1330 		case SO_DEBUG:
1331 		case SO_KEEPALIVE:
1332 		case SO_REUSEADDR:
1333 		case SO_REUSEPORT:
1334 		case SO_BROADCAST:
1335 		case SO_OOBINLINE:
1336 		case SO_TIMESTAMP:
1337 			optval = so->so_options & sopt->sopt_name;
1338 integer:
1339 			error = sooptcopyout(sopt, &optval, sizeof optval);
1340 			break;
1341 
1342 		case SO_TYPE:
1343 			optval = so->so_type;
1344 			goto integer;
1345 
1346 		case SO_ERROR:
1347 			optval = so->so_error;
1348 			so->so_error = 0;
1349 			goto integer;
1350 
1351 		case SO_SNDBUF:
1352 			optval = so->so_snd.sb_hiwat;
1353 			goto integer;
1354 
1355 		case SO_RCVBUF:
1356 			optval = so->so_rcv.sb_hiwat;
1357 			goto integer;
1358 
1359 		case SO_SNDLOWAT:
1360 			optval = so->so_snd.sb_lowat;
1361 			goto integer;
1362 
1363 		case SO_RCVLOWAT:
1364 			optval = so->so_rcv.sb_lowat;
1365 			goto integer;
1366 
1367 		case SO_SNDTIMEO:
1368 		case SO_RCVTIMEO:
1369 			optval = (sopt->sopt_name == SO_SNDTIMEO ?
1370 				  so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1371 
1372 			tv.tv_sec = optval / hz;
1373 			tv.tv_usec = (optval % hz) * tick;
1374 			error = sooptcopyout(sopt, &tv, sizeof tv);
1375 			break;
1376 
1377 		default:
1378 			error = ENOPROTOOPT;
1379 			break;
1380 		}
1381 		return (error);
1382 	}
1383 }
1384 
1385 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1386 int
1387 soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1388 {
1389 	struct mbuf *m, *m_prev;
1390 	int sopt_size = sopt->sopt_valsize;
1391 
1392 	MGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1393 	if (m == 0)
1394 		return ENOBUFS;
1395 	if (sopt_size > MLEN) {
1396 		MCLGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT);
1397 		if ((m->m_flags & M_EXT) == 0) {
1398 			m_free(m);
1399 			return ENOBUFS;
1400 		}
1401 		m->m_len = min(MCLBYTES, sopt_size);
1402 	} else {
1403 		m->m_len = min(MLEN, sopt_size);
1404 	}
1405 	sopt_size -= m->m_len;
1406 	*mp = m;
1407 	m_prev = m;
1408 
1409 	while (sopt_size) {
1410 		MGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1411 		if (m == 0) {
1412 			m_freem(*mp);
1413 			return ENOBUFS;
1414 		}
1415 		if (sopt_size > MLEN) {
1416 			MCLGET(m, sopt->sopt_p ? M_TRYWAIT : M_DONTWAIT);
1417 			if ((m->m_flags & M_EXT) == 0) {
1418 				m_freem(*mp);
1419 				return ENOBUFS;
1420 			}
1421 			m->m_len = min(MCLBYTES, sopt_size);
1422 		} else {
1423 			m->m_len = min(MLEN, sopt_size);
1424 		}
1425 		sopt_size -= m->m_len;
1426 		m_prev->m_next = m;
1427 		m_prev = m;
1428 	}
1429 	return 0;
1430 }
1431 
1432 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1433 int
1434 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1435 {
1436 	struct mbuf *m0 = m;
1437 
1438 	if (sopt->sopt_val == NULL)
1439 		return 0;
1440 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1441 		if (sopt->sopt_p != NULL) {
1442 			int error;
1443 
1444 			error = copyin(sopt->sopt_val, mtod(m, char *),
1445 				       m->m_len);
1446 			if (error != 0) {
1447 				m_freem(m0);
1448 				return(error);
1449 			}
1450 		} else
1451 			bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1452 		sopt->sopt_valsize -= m->m_len;
1453 		(caddr_t)sopt->sopt_val += m->m_len;
1454 		m = m->m_next;
1455 	}
1456 	if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1457 		panic("ip6_sooptmcopyin");
1458 	return 0;
1459 }
1460 
1461 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1462 int
1463 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1464 {
1465 	struct mbuf *m0 = m;
1466 	size_t valsize = 0;
1467 
1468 	if (sopt->sopt_val == NULL)
1469 		return 0;
1470 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1471 		if (sopt->sopt_p != NULL) {
1472 			int error;
1473 
1474 			error = copyout(mtod(m, char *), sopt->sopt_val,
1475 				       m->m_len);
1476 			if (error != 0) {
1477 				m_freem(m0);
1478 				return(error);
1479 			}
1480 		} else
1481 			bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1482 	       sopt->sopt_valsize -= m->m_len;
1483 	       (caddr_t)sopt->sopt_val += m->m_len;
1484 	       valsize += m->m_len;
1485 	       m = m->m_next;
1486 	}
1487 	if (m != NULL) {
1488 		/* enough soopt buffer should be given from user-land */
1489 		m_freem(m0);
1490 		return(EINVAL);
1491 	}
1492 	sopt->sopt_valsize = valsize;
1493 	return 0;
1494 }
1495 
1496 void
1497 sohasoutofband(so)
1498 	register struct socket *so;
1499 {
1500 	if (so->so_sigio != NULL)
1501 		pgsigio(so->so_sigio, SIGURG, 0);
1502 	selwakeup(&so->so_rcv.sb_sel);
1503 }
1504 
1505 int
1506 sopoll(struct socket *so, int events, struct ucred *cred, struct proc *p)
1507 {
1508 	int revents = 0;
1509 	int s = splnet();
1510 
1511 	if (events & (POLLIN | POLLRDNORM))
1512 		if (soreadable(so))
1513 			revents |= events & (POLLIN | POLLRDNORM);
1514 
1515 	if (events & (POLLOUT | POLLWRNORM))
1516 		if (sowriteable(so))
1517 			revents |= events & (POLLOUT | POLLWRNORM);
1518 
1519 	if (events & (POLLPRI | POLLRDBAND))
1520 		if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1521 			revents |= events & (POLLPRI | POLLRDBAND);
1522 
1523 	if (revents == 0) {
1524 		if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
1525 			selrecord(p, &so->so_rcv.sb_sel);
1526 			so->so_rcv.sb_flags |= SB_SEL;
1527 		}
1528 
1529 		if (events & (POLLOUT | POLLWRNORM)) {
1530 			selrecord(p, &so->so_snd.sb_sel);
1531 			so->so_snd.sb_flags |= SB_SEL;
1532 		}
1533 	}
1534 
1535 	splx(s);
1536 	return (revents);
1537 }
1538 
1539 int
1540 sokqfilter(struct file *fp, struct knote *kn)
1541 {
1542 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1543 	struct sockbuf *sb;
1544 	int s;
1545 
1546 	switch (kn->kn_filter) {
1547 	case EVFILT_READ:
1548 		if (so->so_options & SO_ACCEPTCONN)
1549 			kn->kn_fop = &solisten_filtops;
1550 		else
1551 			kn->kn_fop = &soread_filtops;
1552 		sb = &so->so_rcv;
1553 		break;
1554 	case EVFILT_WRITE:
1555 		kn->kn_fop = &sowrite_filtops;
1556 		sb = &so->so_snd;
1557 		break;
1558 	default:
1559 		return (1);
1560 	}
1561 
1562 	s = splnet();
1563 	SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1564 	sb->sb_flags |= SB_KNOTE;
1565 	splx(s);
1566 	return (0);
1567 }
1568 
1569 static void
1570 filt_sordetach(struct knote *kn)
1571 {
1572 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1573 	int s = splnet();
1574 
1575 	SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1576 	if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1577 		so->so_rcv.sb_flags &= ~SB_KNOTE;
1578 	splx(s);
1579 }
1580 
1581 /*ARGSUSED*/
1582 static int
1583 filt_soread(struct knote *kn, long hint)
1584 {
1585 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1586 
1587 	kn->kn_data = so->so_rcv.sb_cc;
1588 	if (so->so_state & SS_CANTRCVMORE) {
1589 		kn->kn_flags |= EV_EOF;
1590 		kn->kn_fflags = so->so_error;
1591 		return (1);
1592 	}
1593 	if (so->so_error)	/* temporary udp error */
1594 		return (1);
1595 	if (kn->kn_sfflags & NOTE_LOWAT)
1596 		return (kn->kn_data >= kn->kn_sdata);
1597 	return (kn->kn_data >= so->so_rcv.sb_lowat);
1598 }
1599 
1600 static void
1601 filt_sowdetach(struct knote *kn)
1602 {
1603 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1604 	int s = splnet();
1605 
1606 	SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1607 	if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1608 		so->so_snd.sb_flags &= ~SB_KNOTE;
1609 	splx(s);
1610 }
1611 
1612 /*ARGSUSED*/
1613 static int
1614 filt_sowrite(struct knote *kn, long hint)
1615 {
1616 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1617 
1618 	kn->kn_data = sbspace(&so->so_snd);
1619 	if (so->so_state & SS_CANTSENDMORE) {
1620 		kn->kn_flags |= EV_EOF;
1621 		kn->kn_fflags = so->so_error;
1622 		return (1);
1623 	}
1624 	if (so->so_error)	/* temporary udp error */
1625 		return (1);
1626 	if (((so->so_state & SS_ISCONNECTED) == 0) &&
1627 	    (so->so_proto->pr_flags & PR_CONNREQUIRED))
1628 		return (0);
1629 	if (kn->kn_sfflags & NOTE_LOWAT)
1630 		return (kn->kn_data >= kn->kn_sdata);
1631 	return (kn->kn_data >= so->so_snd.sb_lowat);
1632 }
1633 
1634 /*ARGSUSED*/
1635 static int
1636 filt_solisten(struct knote *kn, long hint)
1637 {
1638 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1639 
1640 	kn->kn_data = so->so_qlen - so->so_incqlen;
1641 	return (! TAILQ_EMPTY(&so->so_comp));
1642 }
1643