xref: /dflybsd-src/sys/kern/uipc_socket.c (revision ac2e3f5effc58aa364c7e5c199f35ebbae7cda81)
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: src/sys/kern/uipc_socket.c,v 1.68.2.22 2002/12/15 09:24:23 maxim Exp $
35  * $DragonFly: src/sys/kern/uipc_socket.c,v 1.8 2003/07/26 19:42:11 rob Exp $
36  */
37 
38 #include "opt_inet.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/fcntl.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/domain.h>
46 #include <sys/file.h>			/* for struct knote */
47 #include <sys/kernel.h>
48 #include <sys/malloc.h>
49 #include <sys/event.h>
50 #include <sys/poll.h>
51 #include <sys/proc.h>
52 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/resourcevar.h>
56 #include <sys/signalvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/uio.h>
59 #include <sys/jail.h>
60 #include <vm/vm_zone.h>
61 
62 #include <machine/limits.h>
63 
64 #ifdef INET
65 static int	 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
66 #endif /* INET */
67 
68 static void 	filt_sordetach(struct knote *kn);
69 static int 	filt_soread(struct knote *kn, long hint);
70 static void 	filt_sowdetach(struct knote *kn);
71 static int	filt_sowrite(struct knote *kn, long hint);
72 static int	filt_solisten(struct knote *kn, long hint);
73 
74 static struct filterops solisten_filtops =
75 	{ 1, NULL, filt_sordetach, filt_solisten };
76 static struct filterops soread_filtops =
77 	{ 1, NULL, filt_sordetach, filt_soread };
78 static struct filterops sowrite_filtops =
79 	{ 1, NULL, filt_sowdetach, filt_sowrite };
80 
81 struct	vm_zone *socket_zone;
82 so_gen_t	so_gencnt;	/* generation count for sockets */
83 
84 MALLOC_DEFINE(M_SONAME, "soname", "socket name");
85 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");
86 
87 SYSCTL_DECL(_kern_ipc);
88 
89 static int somaxconn = SOMAXCONN;
90 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW,
91     &somaxconn, 0, "Maximum pending socket connection queue size");
92 
93 /*
94  * Socket operation routines.
95  * These routines are called by the routines in
96  * sys_socket.c or from a system process, and
97  * implement the semantics of socket operations by
98  * switching out to the protocol specific routines.
99  */
100 
101 /*
102  * Get a socket structure from our zone, and initialize it.
103  * We don't implement `waitok' yet (see comments in uipc_domain.c).
104  * Note that it would probably be better to allocate socket
105  * and PCB at the same time, but I'm not convinced that all
106  * the protocols can be easily modified to do this.
107  */
108 struct socket *
109 soalloc(waitok)
110 	int waitok;
111 {
112 	struct socket *so;
113 
114 	so = zalloc(socket_zone);
115 	if (so) {
116 		/* XXX race condition for reentrant kernel */
117 		bzero(so, sizeof *so);
118 		so->so_gencnt = ++so_gencnt;
119 		TAILQ_INIT(&so->so_aiojobq);
120 	}
121 	return so;
122 }
123 
124 int
125 socreate(int dom, struct socket **aso, int type,
126 	int proto, struct thread *td)
127 {
128 	struct proc *p = td->td_proc;
129 	struct protosw *prp;
130 	struct socket *so;
131 	int error;
132 
133 	if (proto)
134 		prp = pffindproto(dom, proto, type);
135 	else
136 		prp = pffindtype(dom, type);
137 
138 	if (prp == 0 || prp->pr_usrreqs->pru_attach == 0)
139 		return (EPROTONOSUPPORT);
140 
141 	if (p->p_ucred->cr_prison && jail_socket_unixiproute_only &&
142 	    prp->pr_domain->dom_family != PF_LOCAL &&
143 	    prp->pr_domain->dom_family != PF_INET &&
144 	    prp->pr_domain->dom_family != PF_ROUTE) {
145 		return (EPROTONOSUPPORT);
146 	}
147 
148 	if (prp->pr_type != type)
149 		return (EPROTOTYPE);
150 	so = soalloc(p != 0);
151 	if (so == 0)
152 		return (ENOBUFS);
153 
154 	TAILQ_INIT(&so->so_incomp);
155 	TAILQ_INIT(&so->so_comp);
156 	so->so_type = type;
157 	so->so_cred = crhold(p->p_ucred);
158 	so->so_proto = prp;
159 	error = (*prp->pr_usrreqs->pru_attach)(so, proto, td);
160 	if (error) {
161 		so->so_state |= SS_NOFDREF;
162 		sofree(so);
163 		return (error);
164 	}
165 	*aso = so;
166 	return (0);
167 }
168 
169 int
170 sobind(struct socket *so, struct sockaddr *nam, struct thread *td)
171 {
172 	int s = splnet();
173 	int error;
174 
175 	error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, td);
176 	splx(s);
177 	return (error);
178 }
179 
180 void
181 sodealloc(struct socket *so)
182 {
183 
184 	so->so_gencnt = ++so_gencnt;
185 	if (so->so_rcv.sb_hiwat)
186 		(void)chgsbsize(so->so_cred->cr_uidinfo,
187 		    &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
188 	if (so->so_snd.sb_hiwat)
189 		(void)chgsbsize(so->so_cred->cr_uidinfo,
190 		    &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
191 #ifdef INET
192 	if (so->so_accf != NULL) {
193 		if (so->so_accf->so_accept_filter != NULL &&
194 			so->so_accf->so_accept_filter->accf_destroy != NULL) {
195 			so->so_accf->so_accept_filter->accf_destroy(so);
196 		}
197 		if (so->so_accf->so_accept_filter_str != NULL)
198 			FREE(so->so_accf->so_accept_filter_str, M_ACCF);
199 		FREE(so->so_accf, M_ACCF);
200 	}
201 #endif /* INET */
202 	crfree(so->so_cred);
203 	zfree(socket_zone, so);
204 }
205 
206 int
207 solisten(struct socket *so, int backlog, struct thread *td)
208 {
209 	int s, error;
210 
211 	s = splnet();
212 	error = (*so->so_proto->pr_usrreqs->pru_listen)(so, td);
213 	if (error) {
214 		splx(s);
215 		return (error);
216 	}
217 	if (TAILQ_EMPTY(&so->so_comp))
218 		so->so_options |= SO_ACCEPTCONN;
219 	if (backlog < 0 || backlog > somaxconn)
220 		backlog = somaxconn;
221 	so->so_qlimit = backlog;
222 	splx(s);
223 	return (0);
224 }
225 
226 void
227 sofree(struct socket *so)
228 {
229 	struct socket *head = so->so_head;
230 
231 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
232 		return;
233 	if (head != NULL) {
234 		if (so->so_state & SS_INCOMP) {
235 			TAILQ_REMOVE(&head->so_incomp, so, so_list);
236 			head->so_incqlen--;
237 		} else if (so->so_state & SS_COMP) {
238 			/*
239 			 * We must not decommission a socket that's
240 			 * on the accept(2) queue.  If we do, then
241 			 * accept(2) may hang after select(2) indicated
242 			 * that the listening socket was ready.
243 			 */
244 			return;
245 		} else {
246 			panic("sofree: not queued");
247 		}
248 		so->so_state &= ~SS_INCOMP;
249 		so->so_head = NULL;
250 	}
251 	sbrelease(&so->so_snd, so);
252 	sorflush(so);
253 	sodealloc(so);
254 }
255 
256 /*
257  * Close a socket on last file table reference removal.
258  * Initiate disconnect if connected.
259  * Free socket when disconnect complete.
260  */
261 int
262 soclose(struct socket *so)
263 {
264 	int s = splnet();		/* conservative */
265 	int error = 0;
266 
267 	funsetown(so->so_sigio);
268 	if (so->so_options & SO_ACCEPTCONN) {
269 		struct socket *sp, *sonext;
270 
271 		sp = TAILQ_FIRST(&so->so_incomp);
272 		for (; sp != NULL; sp = sonext) {
273 			sonext = TAILQ_NEXT(sp, so_list);
274 			(void) soabort(sp);
275 		}
276 		for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
277 			sonext = TAILQ_NEXT(sp, so_list);
278 			/* Dequeue from so_comp since sofree() won't do it */
279 			TAILQ_REMOVE(&so->so_comp, sp, so_list);
280 			so->so_qlen--;
281 			sp->so_state &= ~SS_COMP;
282 			sp->so_head = NULL;
283 			(void) soabort(sp);
284 		}
285 	}
286 	if (so->so_pcb == 0)
287 		goto discard;
288 	if (so->so_state & SS_ISCONNECTED) {
289 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
290 			error = sodisconnect(so);
291 			if (error)
292 				goto drop;
293 		}
294 		if (so->so_options & SO_LINGER) {
295 			if ((so->so_state & SS_ISDISCONNECTING) &&
296 			    (so->so_state & SS_NBIO))
297 				goto drop;
298 			while (so->so_state & SS_ISCONNECTED) {
299 				error = tsleep((caddr_t)&so->so_timeo,
300 				    PCATCH, "soclos", so->so_linger * hz);
301 				if (error)
302 					break;
303 			}
304 		}
305 	}
306 drop:
307 	if (so->so_pcb) {
308 		int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so);
309 		if (error == 0)
310 			error = error2;
311 	}
312 discard:
313 	if (so->so_state & SS_NOFDREF)
314 		panic("soclose: NOFDREF");
315 	so->so_state |= SS_NOFDREF;
316 	sofree(so);
317 	splx(s);
318 	return (error);
319 }
320 
321 /*
322  * Must be called at splnet...
323  */
324 int
325 soabort(so)
326 	struct socket *so;
327 {
328 	int error;
329 
330 	error = (*so->so_proto->pr_usrreqs->pru_abort)(so);
331 	if (error) {
332 		sofree(so);
333 		return error;
334 	}
335 	return (0);
336 }
337 
338 int
339 soaccept(struct socket *so, struct sockaddr **nam)
340 {
341 	int s = splnet();
342 	int error;
343 
344 	if ((so->so_state & SS_NOFDREF) == 0)
345 		panic("soaccept: !NOFDREF");
346 	so->so_state &= ~SS_NOFDREF;
347 	error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam);
348 	splx(s);
349 	return (error);
350 }
351 
352 int
353 soconnect(struct socket *so, struct sockaddr *nam, struct thread *td)
354 {
355 	int s;
356 	int error;
357 
358 	if (so->so_options & SO_ACCEPTCONN)
359 		return (EOPNOTSUPP);
360 	s = splnet();
361 	/*
362 	 * If protocol is connection-based, can only connect once.
363 	 * Otherwise, if connected, try to disconnect first.
364 	 * This allows user to disconnect by connecting to, e.g.,
365 	 * a null address.
366 	 */
367 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
368 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
369 	    (error = sodisconnect(so))))
370 		error = EISCONN;
371 	else
372 		error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, td);
373 	splx(s);
374 	return (error);
375 }
376 
377 int
378 soconnect2(struct socket *so1, struct socket *so2)
379 {
380 	int s = splnet();
381 	int error;
382 
383 	error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2);
384 	splx(s);
385 	return (error);
386 }
387 
388 int
389 sodisconnect(struct socket *so)
390 {
391 	int s = splnet();
392 	int error;
393 
394 	if ((so->so_state & SS_ISCONNECTED) == 0) {
395 		error = ENOTCONN;
396 		goto bad;
397 	}
398 	if (so->so_state & SS_ISDISCONNECTING) {
399 		error = EALREADY;
400 		goto bad;
401 	}
402 	error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so);
403 bad:
404 	splx(s);
405 	return (error);
406 }
407 
408 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
409 /*
410  * Send on a socket.
411  * If send must go all at once and message is larger than
412  * send buffering, then hard error.
413  * Lock against other senders.
414  * If must go all at once and not enough room now, then
415  * inform user that this would block and do nothing.
416  * Otherwise, if nonblocking, send as much as possible.
417  * The data to be sent is described by "uio" if nonzero,
418  * otherwise by the mbuf chain "top" (which must be null
419  * if uio is not).  Data provided in mbuf chain must be small
420  * enough to send all at once.
421  *
422  * Returns nonzero on error, timeout or signal; callers
423  * must check for short counts if EINTR/ERESTART are returned.
424  * Data and control buffers are freed on return.
425  */
426 int
427 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
428 	struct mbuf *top, struct mbuf *control, int flags,
429 	struct thread *td)
430 {
431 	struct mbuf **mp;
432 	struct mbuf *m;
433 	long space, len, resid;
434 	int clen = 0, error, s, dontroute, mlen;
435 	int atomic = sosendallatonce(so) || top;
436 
437 	if (uio)
438 		resid = uio->uio_resid;
439 	else
440 		resid = top->m_pkthdr.len;
441 	/*
442 	 * In theory resid should be unsigned.
443 	 * However, space must be signed, as it might be less than 0
444 	 * if we over-committed, and we must use a signed comparison
445 	 * of space and resid.  On the other hand, a negative resid
446 	 * causes us to loop sending 0-length segments to the protocol.
447 	 *
448 	 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
449 	 * type sockets since that's an error.
450 	 */
451 	if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
452 		error = EINVAL;
453 		goto out;
454 	}
455 
456 	dontroute =
457 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
458 	    (so->so_proto->pr_flags & PR_ATOMIC);
459 	if (td->td_proc && td->td_proc->p_stats)
460 		td->td_proc->p_stats->p_ru.ru_msgsnd++;
461 	if (control)
462 		clen = control->m_len;
463 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
464 
465 restart:
466 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
467 	if (error)
468 		goto out;
469 	do {
470 		s = splnet();
471 		if (so->so_state & SS_CANTSENDMORE)
472 			snderr(EPIPE);
473 		if (so->so_error) {
474 			error = so->so_error;
475 			so->so_error = 0;
476 			splx(s);
477 			goto release;
478 		}
479 		if ((so->so_state & SS_ISCONNECTED) == 0) {
480 			/*
481 			 * `sendto' and `sendmsg' is allowed on a connection-
482 			 * based socket if it supports implied connect.
483 			 * Return ENOTCONN if not connected and no address is
484 			 * supplied.
485 			 */
486 			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
487 			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
488 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
489 				    !(resid == 0 && clen != 0))
490 					snderr(ENOTCONN);
491 			} else if (addr == 0)
492 			    snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
493 				   ENOTCONN : EDESTADDRREQ);
494 		}
495 		space = sbspace(&so->so_snd);
496 		if (flags & MSG_OOB)
497 			space += 1024;
498 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
499 		    clen > so->so_snd.sb_hiwat)
500 			snderr(EMSGSIZE);
501 		if (space < resid + clen &&
502 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
503 			if (so->so_state & SS_NBIO)
504 				snderr(EWOULDBLOCK);
505 			sbunlock(&so->so_snd);
506 			error = sbwait(&so->so_snd);
507 			splx(s);
508 			if (error)
509 				goto out;
510 			goto restart;
511 		}
512 		splx(s);
513 		mp = &top;
514 		space -= clen;
515 		do {
516 		    if (uio == NULL) {
517 			/*
518 			 * Data is prepackaged in "top".
519 			 */
520 			resid = 0;
521 			if (flags & MSG_EOR)
522 				top->m_flags |= M_EOR;
523 		    } else do {
524 			if (top == 0) {
525 				MGETHDR(m, M_WAIT, MT_DATA);
526 				if (m == NULL) {
527 					error = ENOBUFS;
528 					goto release;
529 				}
530 				mlen = MHLEN;
531 				m->m_pkthdr.len = 0;
532 				m->m_pkthdr.rcvif = (struct ifnet *)0;
533 			} else {
534 				MGET(m, M_WAIT, MT_DATA);
535 				if (m == NULL) {
536 					error = ENOBUFS;
537 					goto release;
538 				}
539 				mlen = MLEN;
540 			}
541 			if (resid >= MINCLSIZE) {
542 				MCLGET(m, M_WAIT);
543 				if ((m->m_flags & M_EXT) == 0)
544 					goto nopages;
545 				mlen = MCLBYTES;
546 				len = min(min(mlen, resid), space);
547 			} else {
548 nopages:
549 				len = min(min(mlen, resid), space);
550 				/*
551 				 * For datagram protocols, leave room
552 				 * for protocol headers in first mbuf.
553 				 */
554 				if (atomic && top == 0 && len < mlen)
555 					MH_ALIGN(m, len);
556 			}
557 			space -= len;
558 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
559 			resid = uio->uio_resid;
560 			m->m_len = len;
561 			*mp = m;
562 			top->m_pkthdr.len += len;
563 			if (error)
564 				goto release;
565 			mp = &m->m_next;
566 			if (resid <= 0) {
567 				if (flags & MSG_EOR)
568 					top->m_flags |= M_EOR;
569 				break;
570 			}
571 		    } while (space > 0 && atomic);
572 		    if (dontroute)
573 			    so->so_options |= SO_DONTROUTE;
574 		    s = splnet();				/* XXX */
575 		    /*
576 		     * XXX all the SS_CANTSENDMORE checks previously
577 		     * done could be out of date.  We could have recieved
578 		     * a reset packet in an interrupt or maybe we slept
579 		     * while doing page faults in uiomove() etc. We could
580 		     * probably recheck again inside the splnet() protection
581 		     * here, but there are probably other places that this
582 		     * also happens.  We must rethink this.
583 		     */
584 		    error = (*so->so_proto->pr_usrreqs->pru_send)(so,
585 			(flags & MSG_OOB) ? PRUS_OOB :
586 			/*
587 			 * If the user set MSG_EOF, the protocol
588 			 * understands this flag and nothing left to
589 			 * send then use PRU_SEND_EOF instead of PRU_SEND.
590 			 */
591 			((flags & MSG_EOF) &&
592 			 (so->so_proto->pr_flags & PR_IMPLOPCL) &&
593 			 (resid <= 0)) ?
594 				PRUS_EOF :
595 			/* If there is more to send set PRUS_MORETOCOME */
596 			(resid > 0 && space > 0) ? PRUS_MORETOCOME : 0,
597 			top, addr, control, td);
598 		    splx(s);
599 		    if (dontroute)
600 			    so->so_options &= ~SO_DONTROUTE;
601 		    clen = 0;
602 		    control = 0;
603 		    top = 0;
604 		    mp = &top;
605 		    if (error)
606 			goto release;
607 		} while (resid && space > 0);
608 	} while (resid);
609 
610 release:
611 	sbunlock(&so->so_snd);
612 out:
613 	if (top)
614 		m_freem(top);
615 	if (control)
616 		m_freem(control);
617 	return (error);
618 }
619 
620 /*
621  * Implement receive operations on a socket.
622  * We depend on the way that records are added to the sockbuf
623  * by sbappend*.  In particular, each record (mbufs linked through m_next)
624  * must begin with an address if the protocol so specifies,
625  * followed by an optional mbuf or mbufs containing ancillary data,
626  * and then zero or more mbufs of data.
627  * In order to avoid blocking network interrupts for the entire time here,
628  * we splx() while doing the actual copy to user space.
629  * Although the sockbuf is locked, new data may still be appended,
630  * and thus we must maintain consistency of the sockbuf during that time.
631  *
632  * The caller may receive the data as a single mbuf chain by supplying
633  * an mbuf **mp0 for use in returning the chain.  The uio is then used
634  * only for the count in uio_resid.
635  */
636 int
637 soreceive(so, psa, uio, mp0, controlp, flagsp)
638 	struct socket *so;
639 	struct sockaddr **psa;
640 	struct uio *uio;
641 	struct mbuf **mp0;
642 	struct mbuf **controlp;
643 	int *flagsp;
644 {
645 	struct mbuf *m, **mp;
646 	int flags, len, error, s, offset;
647 	struct protosw *pr = so->so_proto;
648 	struct mbuf *nextrecord;
649 	int moff, type = 0;
650 	int orig_resid = uio->uio_resid;
651 
652 	mp = mp0;
653 	if (psa)
654 		*psa = 0;
655 	if (controlp)
656 		*controlp = 0;
657 	if (flagsp)
658 		flags = *flagsp &~ MSG_EOR;
659 	else
660 		flags = 0;
661 	if (flags & MSG_OOB) {
662 		m = m_get(M_WAIT, MT_DATA);
663 		if (m == NULL)
664 			return (ENOBUFS);
665 		error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK);
666 		if (error)
667 			goto bad;
668 		do {
669 			error = uiomove(mtod(m, caddr_t),
670 			    (int) min(uio->uio_resid, m->m_len), uio);
671 			m = m_free(m);
672 		} while (uio->uio_resid && error == 0 && m);
673 bad:
674 		if (m)
675 			m_freem(m);
676 		return (error);
677 	}
678 	if (mp)
679 		*mp = (struct mbuf *)0;
680 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
681 		(*pr->pr_usrreqs->pru_rcvd)(so, 0);
682 
683 restart:
684 	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
685 	if (error)
686 		return (error);
687 	s = splnet();
688 
689 	m = so->so_rcv.sb_mb;
690 	/*
691 	 * If we have less data than requested, block awaiting more
692 	 * (subject to any timeout) if:
693 	 *   1. the current count is less than the low water mark, or
694 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
695 	 *	receive operation at once if we block (resid <= hiwat).
696 	 *   3. MSG_DONTWAIT is not set
697 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
698 	 * we have to do the receive in sections, and thus risk returning
699 	 * a short count if a timeout or signal occurs after we start.
700 	 */
701 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
702 	    so->so_rcv.sb_cc < uio->uio_resid) &&
703 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
704 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
705 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
706 		KASSERT(m != 0 || !so->so_rcv.sb_cc, ("receive 1"));
707 		if (so->so_error) {
708 			if (m)
709 				goto dontblock;
710 			error = so->so_error;
711 			if ((flags & MSG_PEEK) == 0)
712 				so->so_error = 0;
713 			goto release;
714 		}
715 		if (so->so_state & SS_CANTRCVMORE) {
716 			if (m)
717 				goto dontblock;
718 			else
719 				goto release;
720 		}
721 		for (; m; m = m->m_next)
722 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
723 				m = so->so_rcv.sb_mb;
724 				goto dontblock;
725 			}
726 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
727 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
728 			error = ENOTCONN;
729 			goto release;
730 		}
731 		if (uio->uio_resid == 0)
732 			goto release;
733 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
734 			error = EWOULDBLOCK;
735 			goto release;
736 		}
737 		sbunlock(&so->so_rcv);
738 		error = sbwait(&so->so_rcv);
739 		splx(s);
740 		if (error)
741 			return (error);
742 		goto restart;
743 	}
744 dontblock:
745 	if (uio->uio_td && uio->uio_td->td_proc)
746 		uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++;
747 	nextrecord = m->m_nextpkt;
748 	if (pr->pr_flags & PR_ADDR) {
749 		KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
750 		orig_resid = 0;
751 		if (psa)
752 			*psa = dup_sockaddr(mtod(m, struct sockaddr *),
753 					    mp0 == 0);
754 		if (flags & MSG_PEEK) {
755 			m = m->m_next;
756 		} else {
757 			sbfree(&so->so_rcv, m);
758 			so->so_rcv.sb_mb = m_free(m);
759 			m = so->so_rcv.sb_mb;
760 		}
761 	}
762 	while (m && m->m_type == MT_CONTROL && error == 0) {
763 		if (flags & MSG_PEEK) {
764 			if (controlp)
765 				*controlp = m_copy(m, 0, m->m_len);
766 			m = m->m_next;
767 		} else {
768 			sbfree(&so->so_rcv, m);
769 			if (controlp) {
770 				if (pr->pr_domain->dom_externalize &&
771 				    mtod(m, struct cmsghdr *)->cmsg_type ==
772 				    SCM_RIGHTS)
773 				   error = (*pr->pr_domain->dom_externalize)(m);
774 				*controlp = m;
775 				so->so_rcv.sb_mb = m->m_next;
776 				m->m_next = 0;
777 				m = so->so_rcv.sb_mb;
778 			} else {
779 				so->so_rcv.sb_mb = m_free(m);
780 				m = so->so_rcv.sb_mb;
781 			}
782 		}
783 		if (controlp) {
784 			orig_resid = 0;
785 			controlp = &(*controlp)->m_next;
786 		}
787 	}
788 	if (m) {
789 		if ((flags & MSG_PEEK) == 0)
790 			m->m_nextpkt = nextrecord;
791 		type = m->m_type;
792 		if (type == MT_OOBDATA)
793 			flags |= MSG_OOB;
794 	}
795 	moff = 0;
796 	offset = 0;
797 	while (m && uio->uio_resid > 0 && error == 0) {
798 		if (m->m_type == MT_OOBDATA) {
799 			if (type != MT_OOBDATA)
800 				break;
801 		} else if (type == MT_OOBDATA)
802 			break;
803 		else
804 		    KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
805 			("receive 3"));
806 		so->so_state &= ~SS_RCVATMARK;
807 		len = uio->uio_resid;
808 		if (so->so_oobmark && len > so->so_oobmark - offset)
809 			len = so->so_oobmark - offset;
810 		if (len > m->m_len - moff)
811 			len = m->m_len - moff;
812 		/*
813 		 * If mp is set, just pass back the mbufs.
814 		 * Otherwise copy them out via the uio, then free.
815 		 * Sockbuf must be consistent here (points to current mbuf,
816 		 * it points to next record) when we drop priority;
817 		 * we must note any additions to the sockbuf when we
818 		 * block interrupts again.
819 		 */
820 		if (mp == 0) {
821 			splx(s);
822 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
823 			s = splnet();
824 			if (error)
825 				goto release;
826 		} else
827 			uio->uio_resid -= len;
828 		if (len == m->m_len - moff) {
829 			if (m->m_flags & M_EOR)
830 				flags |= MSG_EOR;
831 			if (flags & MSG_PEEK) {
832 				m = m->m_next;
833 				moff = 0;
834 			} else {
835 				nextrecord = m->m_nextpkt;
836 				sbfree(&so->so_rcv, m);
837 				if (mp) {
838 					*mp = m;
839 					mp = &m->m_next;
840 					so->so_rcv.sb_mb = m = m->m_next;
841 					*mp = (struct mbuf *)0;
842 				} else {
843 					so->so_rcv.sb_mb = m = m_free(m);
844 				}
845 				if (m)
846 					m->m_nextpkt = nextrecord;
847 			}
848 		} else {
849 			if (flags & MSG_PEEK)
850 				moff += len;
851 			else {
852 				if (mp)
853 					*mp = m_copym(m, 0, len, M_WAIT);
854 				m->m_data += len;
855 				m->m_len -= len;
856 				so->so_rcv.sb_cc -= len;
857 			}
858 		}
859 		if (so->so_oobmark) {
860 			if ((flags & MSG_PEEK) == 0) {
861 				so->so_oobmark -= len;
862 				if (so->so_oobmark == 0) {
863 					so->so_state |= SS_RCVATMARK;
864 					break;
865 				}
866 			} else {
867 				offset += len;
868 				if (offset == so->so_oobmark)
869 					break;
870 			}
871 		}
872 		if (flags & MSG_EOR)
873 			break;
874 		/*
875 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
876 		 * we must not quit until "uio->uio_resid == 0" or an error
877 		 * termination.  If a signal/timeout occurs, return
878 		 * with a short count but without error.
879 		 * Keep sockbuf locked against other readers.
880 		 */
881 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
882 		    !sosendallatonce(so) && !nextrecord) {
883 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
884 				break;
885 			/*
886 			 * The window might have closed to zero, make
887 			 * sure we send an ack now that we've drained
888 			 * the buffer or we might end up blocking until
889 			 * the idle takes over (5 seconds).
890 			 */
891 			if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
892 				(*pr->pr_usrreqs->pru_rcvd)(so, flags);
893 			error = sbwait(&so->so_rcv);
894 			if (error) {
895 				sbunlock(&so->so_rcv);
896 				splx(s);
897 				return (0);
898 			}
899 			m = so->so_rcv.sb_mb;
900 			if (m)
901 				nextrecord = m->m_nextpkt;
902 		}
903 	}
904 
905 	if (m && pr->pr_flags & PR_ATOMIC) {
906 		flags |= MSG_TRUNC;
907 		if ((flags & MSG_PEEK) == 0)
908 			(void) sbdroprecord(&so->so_rcv);
909 	}
910 	if ((flags & MSG_PEEK) == 0) {
911 		if (m == 0)
912 			so->so_rcv.sb_mb = nextrecord;
913 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
914 			(*pr->pr_usrreqs->pru_rcvd)(so, flags);
915 	}
916 	if (orig_resid == uio->uio_resid && orig_resid &&
917 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
918 		sbunlock(&so->so_rcv);
919 		splx(s);
920 		goto restart;
921 	}
922 
923 	if (flagsp)
924 		*flagsp |= flags;
925 release:
926 	sbunlock(&so->so_rcv);
927 	splx(s);
928 	return (error);
929 }
930 
931 int
932 soshutdown(so, how)
933 	struct socket *so;
934 	int how;
935 {
936 	struct protosw *pr = so->so_proto;
937 
938 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
939 		return (EINVAL);
940 
941 	if (how != SHUT_WR)
942 		sorflush(so);
943 	if (how != SHUT_RD)
944 		return ((*pr->pr_usrreqs->pru_shutdown)(so));
945 	return (0);
946 }
947 
948 void
949 sorflush(so)
950 	struct socket *so;
951 {
952 	struct sockbuf *sb = &so->so_rcv;
953 	struct protosw *pr = so->so_proto;
954 	int s;
955 	struct sockbuf asb;
956 
957 	sb->sb_flags |= SB_NOINTR;
958 	(void) sblock(sb, M_WAITOK);
959 	s = splimp();
960 	socantrcvmore(so);
961 	sbunlock(sb);
962 	asb = *sb;
963 	bzero((caddr_t)sb, sizeof (*sb));
964 	if (asb.sb_flags & SB_KNOTE) {
965 		sb->sb_sel.si_note = asb.sb_sel.si_note;
966 		sb->sb_flags = SB_KNOTE;
967 	}
968 	splx(s);
969 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
970 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
971 	sbrelease(&asb, so);
972 }
973 
974 #ifdef INET
975 static int
976 do_setopt_accept_filter(so, sopt)
977 	struct	socket *so;
978 	struct	sockopt *sopt;
979 {
980 	struct accept_filter_arg	*afap = NULL;
981 	struct accept_filter	*afp;
982 	struct so_accf	*af = so->so_accf;
983 	int	error = 0;
984 
985 	/* do not set/remove accept filters on non listen sockets */
986 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
987 		error = EINVAL;
988 		goto out;
989 	}
990 
991 	/* removing the filter */
992 	if (sopt == NULL) {
993 		if (af != NULL) {
994 			if (af->so_accept_filter != NULL &&
995 				af->so_accept_filter->accf_destroy != NULL) {
996 				af->so_accept_filter->accf_destroy(so);
997 			}
998 			if (af->so_accept_filter_str != NULL) {
999 				FREE(af->so_accept_filter_str, M_ACCF);
1000 			}
1001 			FREE(af, M_ACCF);
1002 			so->so_accf = NULL;
1003 		}
1004 		so->so_options &= ~SO_ACCEPTFILTER;
1005 		return (0);
1006 	}
1007 	/* adding a filter */
1008 	/* must remove previous filter first */
1009 	if (af != NULL) {
1010 		error = EINVAL;
1011 		goto out;
1012 	}
1013 	/* don't put large objects on the kernel stack */
1014 	MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1015 	error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1016 	afap->af_name[sizeof(afap->af_name)-1] = '\0';
1017 	afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1018 	if (error)
1019 		goto out;
1020 	afp = accept_filt_get(afap->af_name);
1021 	if (afp == NULL) {
1022 		error = ENOENT;
1023 		goto out;
1024 	}
1025 	MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK);
1026 	bzero(af, sizeof(*af));
1027 	if (afp->accf_create != NULL) {
1028 		if (afap->af_name[0] != '\0') {
1029 			int len = strlen(afap->af_name) + 1;
1030 
1031 			MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1032 			strcpy(af->so_accept_filter_str, afap->af_name);
1033 		}
1034 		af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1035 		if (af->so_accept_filter_arg == NULL) {
1036 			FREE(af->so_accept_filter_str, M_ACCF);
1037 			FREE(af, M_ACCF);
1038 			so->so_accf = NULL;
1039 			error = EINVAL;
1040 			goto out;
1041 		}
1042 	}
1043 	af->so_accept_filter = afp;
1044 	so->so_accf = af;
1045 	so->so_options |= SO_ACCEPTFILTER;
1046 out:
1047 	if (afap != NULL)
1048 		FREE(afap, M_TEMP);
1049 	return (error);
1050 }
1051 #endif /* INET */
1052 
1053 /*
1054  * Perhaps this routine, and sooptcopyout(), below, ought to come in
1055  * an additional variant to handle the case where the option value needs
1056  * to be some kind of integer, but not a specific size.
1057  * In addition to their use here, these functions are also called by the
1058  * protocol-level pr_ctloutput() routines.
1059  */
1060 int
1061 sooptcopyin(sopt, buf, len, minlen)
1062 	struct	sockopt *sopt;
1063 	void	*buf;
1064 	size_t	len;
1065 	size_t	minlen;
1066 {
1067 	size_t	valsize;
1068 
1069 	/*
1070 	 * If the user gives us more than we wanted, we ignore it,
1071 	 * but if we don't get the minimum length the caller
1072 	 * wants, we return EINVAL.  On success, sopt->sopt_valsize
1073 	 * is set to however much we actually retrieved.
1074 	 */
1075 	if ((valsize = sopt->sopt_valsize) < minlen)
1076 		return EINVAL;
1077 	if (valsize > len)
1078 		sopt->sopt_valsize = valsize = len;
1079 
1080 	if (sopt->sopt_td != NULL)
1081 		return (copyin(sopt->sopt_val, buf, valsize));
1082 
1083 	bcopy(sopt->sopt_val, buf, valsize);
1084 	return 0;
1085 }
1086 
1087 int
1088 sosetopt(so, sopt)
1089 	struct socket *so;
1090 	struct sockopt *sopt;
1091 {
1092 	int	error, optval;
1093 	struct	linger l;
1094 	struct	timeval tv;
1095 	u_long  val;
1096 
1097 	error = 0;
1098 	if (sopt->sopt_level != SOL_SOCKET) {
1099 		if (so->so_proto && so->so_proto->pr_ctloutput)
1100 			return ((*so->so_proto->pr_ctloutput)
1101 				  (so, sopt));
1102 		error = ENOPROTOOPT;
1103 	} else {
1104 		switch (sopt->sopt_name) {
1105 #ifdef INET
1106 		case SO_ACCEPTFILTER:
1107 			error = do_setopt_accept_filter(so, sopt);
1108 			if (error)
1109 				goto bad;
1110 			break;
1111 #endif /* INET */
1112 		case SO_LINGER:
1113 			error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1114 			if (error)
1115 				goto bad;
1116 
1117 			so->so_linger = l.l_linger;
1118 			if (l.l_onoff)
1119 				so->so_options |= SO_LINGER;
1120 			else
1121 				so->so_options &= ~SO_LINGER;
1122 			break;
1123 
1124 		case SO_DEBUG:
1125 		case SO_KEEPALIVE:
1126 		case SO_DONTROUTE:
1127 		case SO_USELOOPBACK:
1128 		case SO_BROADCAST:
1129 		case SO_REUSEADDR:
1130 		case SO_REUSEPORT:
1131 		case SO_OOBINLINE:
1132 		case SO_TIMESTAMP:
1133 			error = sooptcopyin(sopt, &optval, sizeof optval,
1134 					    sizeof optval);
1135 			if (error)
1136 				goto bad;
1137 			if (optval)
1138 				so->so_options |= sopt->sopt_name;
1139 			else
1140 				so->so_options &= ~sopt->sopt_name;
1141 			break;
1142 
1143 		case SO_SNDBUF:
1144 		case SO_RCVBUF:
1145 		case SO_SNDLOWAT:
1146 		case SO_RCVLOWAT:
1147 			error = sooptcopyin(sopt, &optval, sizeof optval,
1148 					    sizeof optval);
1149 			if (error)
1150 				goto bad;
1151 
1152 			/*
1153 			 * Values < 1 make no sense for any of these
1154 			 * options, so disallow them.
1155 			 */
1156 			if (optval < 1) {
1157 				error = EINVAL;
1158 				goto bad;
1159 			}
1160 
1161 			switch (sopt->sopt_name) {
1162 			case SO_SNDBUF:
1163 			case SO_RCVBUF:
1164 				if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1165 				    &so->so_snd : &so->so_rcv, (u_long)optval,
1166 				    so, curproc) == 0) {
1167 					error = ENOBUFS;
1168 					goto bad;
1169 				}
1170 				break;
1171 
1172 			/*
1173 			 * Make sure the low-water is never greater than
1174 			 * the high-water.
1175 			 */
1176 			case SO_SNDLOWAT:
1177 				so->so_snd.sb_lowat =
1178 				    (optval > so->so_snd.sb_hiwat) ?
1179 				    so->so_snd.sb_hiwat : optval;
1180 				break;
1181 			case SO_RCVLOWAT:
1182 				so->so_rcv.sb_lowat =
1183 				    (optval > so->so_rcv.sb_hiwat) ?
1184 				    so->so_rcv.sb_hiwat : optval;
1185 				break;
1186 			}
1187 			break;
1188 
1189 		case SO_SNDTIMEO:
1190 		case SO_RCVTIMEO:
1191 			error = sooptcopyin(sopt, &tv, sizeof tv,
1192 					    sizeof tv);
1193 			if (error)
1194 				goto bad;
1195 
1196 			/* assert(hz > 0); */
1197 			if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1198 			    tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1199 				error = EDOM;
1200 				goto bad;
1201 			}
1202 			/* assert(tick > 0); */
1203 			/* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1204 			val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1205 			if (val > SHRT_MAX) {
1206 				error = EDOM;
1207 				goto bad;
1208 			}
1209 			if (val == 0 && tv.tv_usec != 0)
1210 				val = 1;
1211 
1212 			switch (sopt->sopt_name) {
1213 			case SO_SNDTIMEO:
1214 				so->so_snd.sb_timeo = val;
1215 				break;
1216 			case SO_RCVTIMEO:
1217 				so->so_rcv.sb_timeo = val;
1218 				break;
1219 			}
1220 			break;
1221 		default:
1222 			error = ENOPROTOOPT;
1223 			break;
1224 		}
1225 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1226 			(void) ((*so->so_proto->pr_ctloutput)
1227 				  (so, sopt));
1228 		}
1229 	}
1230 bad:
1231 	return (error);
1232 }
1233 
1234 /* Helper routine for getsockopt */
1235 int
1236 sooptcopyout(sopt, buf, len)
1237 	struct	sockopt *sopt;
1238 	void	*buf;
1239 	size_t	len;
1240 {
1241 	int	error;
1242 	size_t	valsize;
1243 
1244 	error = 0;
1245 
1246 	/*
1247 	 * Documented get behavior is that we always return a value,
1248 	 * possibly truncated to fit in the user's buffer.
1249 	 * Traditional behavior is that we always tell the user
1250 	 * precisely how much we copied, rather than something useful
1251 	 * like the total amount we had available for her.
1252 	 * Note that this interface is not idempotent; the entire answer must
1253 	 * generated ahead of time.
1254 	 */
1255 	valsize = min(len, sopt->sopt_valsize);
1256 	sopt->sopt_valsize = valsize;
1257 	if (sopt->sopt_val != 0) {
1258 		if (sopt->sopt_td != NULL)
1259 			error = copyout(buf, sopt->sopt_val, valsize);
1260 		else
1261 			bcopy(buf, sopt->sopt_val, valsize);
1262 	}
1263 	return error;
1264 }
1265 
1266 int
1267 sogetopt(so, sopt)
1268 	struct socket *so;
1269 	struct sockopt *sopt;
1270 {
1271 	int	error, optval;
1272 	struct	linger l;
1273 	struct	timeval tv;
1274 	struct accept_filter_arg *afap;
1275 
1276 	error = 0;
1277 	if (sopt->sopt_level != SOL_SOCKET) {
1278 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1279 			return ((*so->so_proto->pr_ctloutput)
1280 				  (so, sopt));
1281 		} else
1282 			return (ENOPROTOOPT);
1283 	} else {
1284 		switch (sopt->sopt_name) {
1285 #ifdef INET
1286 		case SO_ACCEPTFILTER:
1287 			if ((so->so_options & SO_ACCEPTCONN) == 0)
1288 				return (EINVAL);
1289 			MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1290 				M_TEMP, M_WAITOK);
1291 			bzero(afap, sizeof(*afap));
1292 			if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1293 				strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1294 				if (so->so_accf->so_accept_filter_str != NULL)
1295 					strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1296 			}
1297 			error = sooptcopyout(sopt, afap, sizeof(*afap));
1298 			FREE(afap, M_TEMP);
1299 			break;
1300 #endif /* INET */
1301 
1302 		case SO_LINGER:
1303 			l.l_onoff = so->so_options & SO_LINGER;
1304 			l.l_linger = so->so_linger;
1305 			error = sooptcopyout(sopt, &l, sizeof l);
1306 			break;
1307 
1308 		case SO_USELOOPBACK:
1309 		case SO_DONTROUTE:
1310 		case SO_DEBUG:
1311 		case SO_KEEPALIVE:
1312 		case SO_REUSEADDR:
1313 		case SO_REUSEPORT:
1314 		case SO_BROADCAST:
1315 		case SO_OOBINLINE:
1316 		case SO_TIMESTAMP:
1317 			optval = so->so_options & sopt->sopt_name;
1318 integer:
1319 			error = sooptcopyout(sopt, &optval, sizeof optval);
1320 			break;
1321 
1322 		case SO_TYPE:
1323 			optval = so->so_type;
1324 			goto integer;
1325 
1326 		case SO_ERROR:
1327 			optval = so->so_error;
1328 			so->so_error = 0;
1329 			goto integer;
1330 
1331 		case SO_SNDBUF:
1332 			optval = so->so_snd.sb_hiwat;
1333 			goto integer;
1334 
1335 		case SO_RCVBUF:
1336 			optval = so->so_rcv.sb_hiwat;
1337 			goto integer;
1338 
1339 		case SO_SNDLOWAT:
1340 			optval = so->so_snd.sb_lowat;
1341 			goto integer;
1342 
1343 		case SO_RCVLOWAT:
1344 			optval = so->so_rcv.sb_lowat;
1345 			goto integer;
1346 
1347 		case SO_SNDTIMEO:
1348 		case SO_RCVTIMEO:
1349 			optval = (sopt->sopt_name == SO_SNDTIMEO ?
1350 				  so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1351 
1352 			tv.tv_sec = optval / hz;
1353 			tv.tv_usec = (optval % hz) * tick;
1354 			error = sooptcopyout(sopt, &tv, sizeof tv);
1355 			break;
1356 
1357 		default:
1358 			error = ENOPROTOOPT;
1359 			break;
1360 		}
1361 		return (error);
1362 	}
1363 }
1364 
1365 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1366 int
1367 soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1368 {
1369 	struct mbuf *m, *m_prev;
1370 	int sopt_size = sopt->sopt_valsize;
1371 
1372 	MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1373 	if (m == 0)
1374 		return ENOBUFS;
1375 	if (sopt_size > MLEN) {
1376 		MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1377 		if ((m->m_flags & M_EXT) == 0) {
1378 			m_free(m);
1379 			return ENOBUFS;
1380 		}
1381 		m->m_len = min(MCLBYTES, sopt_size);
1382 	} else {
1383 		m->m_len = min(MLEN, sopt_size);
1384 	}
1385 	sopt_size -= m->m_len;
1386 	*mp = m;
1387 	m_prev = m;
1388 
1389 	while (sopt_size) {
1390 		MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1391 		if (m == 0) {
1392 			m_freem(*mp);
1393 			return ENOBUFS;
1394 		}
1395 		if (sopt_size > MLEN) {
1396 			MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1397 			if ((m->m_flags & M_EXT) == 0) {
1398 				m_freem(*mp);
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 		m_prev->m_next = m;
1407 		m_prev = m;
1408 	}
1409 	return 0;
1410 }
1411 
1412 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1413 int
1414 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1415 {
1416 	struct mbuf *m0 = m;
1417 
1418 	if (sopt->sopt_val == NULL)
1419 		return 0;
1420 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1421 		if (sopt->sopt_td != NULL) {
1422 			int error;
1423 
1424 			error = copyin(sopt->sopt_val, mtod(m, char *),
1425 				       m->m_len);
1426 			if (error != 0) {
1427 				m_freem(m0);
1428 				return(error);
1429 			}
1430 		} else
1431 			bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1432 		sopt->sopt_valsize -= m->m_len;
1433 		(caddr_t)sopt->sopt_val += m->m_len;
1434 		m = m->m_next;
1435 	}
1436 	if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1437 		panic("ip6_sooptmcopyin");
1438 	return 0;
1439 }
1440 
1441 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1442 int
1443 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1444 {
1445 	struct mbuf *m0 = m;
1446 	size_t valsize = 0;
1447 
1448 	if (sopt->sopt_val == NULL)
1449 		return 0;
1450 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1451 		if (sopt->sopt_td != NULL) {
1452 			int error;
1453 
1454 			error = copyout(mtod(m, char *), sopt->sopt_val,
1455 				       m->m_len);
1456 			if (error != 0) {
1457 				m_freem(m0);
1458 				return(error);
1459 			}
1460 		} else
1461 			bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1462 	       sopt->sopt_valsize -= m->m_len;
1463 	       (caddr_t)sopt->sopt_val += m->m_len;
1464 	       valsize += m->m_len;
1465 	       m = m->m_next;
1466 	}
1467 	if (m != NULL) {
1468 		/* enough soopt buffer should be given from user-land */
1469 		m_freem(m0);
1470 		return(EINVAL);
1471 	}
1472 	sopt->sopt_valsize = valsize;
1473 	return 0;
1474 }
1475 
1476 void
1477 sohasoutofband(so)
1478 	struct socket *so;
1479 {
1480 	if (so->so_sigio != NULL)
1481 		pgsigio(so->so_sigio, SIGURG, 0);
1482 	selwakeup(&so->so_rcv.sb_sel);
1483 }
1484 
1485 int
1486 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td)
1487 {
1488 	int revents = 0;
1489 	int s = splnet();
1490 
1491 	if (events & (POLLIN | POLLRDNORM))
1492 		if (soreadable(so))
1493 			revents |= events & (POLLIN | POLLRDNORM);
1494 
1495 	if (events & (POLLOUT | POLLWRNORM))
1496 		if (sowriteable(so))
1497 			revents |= events & (POLLOUT | POLLWRNORM);
1498 
1499 	if (events & (POLLPRI | POLLRDBAND))
1500 		if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1501 			revents |= events & (POLLPRI | POLLRDBAND);
1502 
1503 	if (revents == 0) {
1504 		if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
1505 			selrecord(td, &so->so_rcv.sb_sel);
1506 			so->so_rcv.sb_flags |= SB_SEL;
1507 		}
1508 
1509 		if (events & (POLLOUT | POLLWRNORM)) {
1510 			selrecord(td, &so->so_snd.sb_sel);
1511 			so->so_snd.sb_flags |= SB_SEL;
1512 		}
1513 	}
1514 
1515 	splx(s);
1516 	return (revents);
1517 }
1518 
1519 int
1520 sokqfilter(struct file *fp, struct knote *kn)
1521 {
1522 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1523 	struct sockbuf *sb;
1524 	int s;
1525 
1526 	switch (kn->kn_filter) {
1527 	case EVFILT_READ:
1528 		if (so->so_options & SO_ACCEPTCONN)
1529 			kn->kn_fop = &solisten_filtops;
1530 		else
1531 			kn->kn_fop = &soread_filtops;
1532 		sb = &so->so_rcv;
1533 		break;
1534 	case EVFILT_WRITE:
1535 		kn->kn_fop = &sowrite_filtops;
1536 		sb = &so->so_snd;
1537 		break;
1538 	default:
1539 		return (1);
1540 	}
1541 
1542 	s = splnet();
1543 	SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1544 	sb->sb_flags |= SB_KNOTE;
1545 	splx(s);
1546 	return (0);
1547 }
1548 
1549 static void
1550 filt_sordetach(struct knote *kn)
1551 {
1552 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1553 	int s = splnet();
1554 
1555 	SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1556 	if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1557 		so->so_rcv.sb_flags &= ~SB_KNOTE;
1558 	splx(s);
1559 }
1560 
1561 /*ARGSUSED*/
1562 static int
1563 filt_soread(struct knote *kn, long hint)
1564 {
1565 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1566 
1567 	kn->kn_data = so->so_rcv.sb_cc;
1568 	if (so->so_state & SS_CANTRCVMORE) {
1569 		kn->kn_flags |= EV_EOF;
1570 		kn->kn_fflags = so->so_error;
1571 		return (1);
1572 	}
1573 	if (so->so_error)	/* temporary udp error */
1574 		return (1);
1575 	if (kn->kn_sfflags & NOTE_LOWAT)
1576 		return (kn->kn_data >= kn->kn_sdata);
1577 	return (kn->kn_data >= so->so_rcv.sb_lowat);
1578 }
1579 
1580 static void
1581 filt_sowdetach(struct knote *kn)
1582 {
1583 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1584 	int s = splnet();
1585 
1586 	SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1587 	if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1588 		so->so_snd.sb_flags &= ~SB_KNOTE;
1589 	splx(s);
1590 }
1591 
1592 /*ARGSUSED*/
1593 static int
1594 filt_sowrite(struct knote *kn, long hint)
1595 {
1596 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1597 
1598 	kn->kn_data = sbspace(&so->so_snd);
1599 	if (so->so_state & SS_CANTSENDMORE) {
1600 		kn->kn_flags |= EV_EOF;
1601 		kn->kn_fflags = so->so_error;
1602 		return (1);
1603 	}
1604 	if (so->so_error)	/* temporary udp error */
1605 		return (1);
1606 	if (((so->so_state & SS_ISCONNECTED) == 0) &&
1607 	    (so->so_proto->pr_flags & PR_CONNREQUIRED))
1608 		return (0);
1609 	if (kn->kn_sfflags & NOTE_LOWAT)
1610 		return (kn->kn_data >= kn->kn_sdata);
1611 	return (kn->kn_data >= so->so_snd.sb_lowat);
1612 }
1613 
1614 /*ARGSUSED*/
1615 static int
1616 filt_solisten(struct knote *kn, long hint)
1617 {
1618 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1619 
1620 	kn->kn_data = so->so_qlen;
1621 	return (! TAILQ_EMPTY(&so->so_comp));
1622 }
1623