xref: /dflybsd-src/sys/kern/uipc_socket.c (revision bd4539cc23771f3c0b3fae4ecf80e725b613b305)
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 1.68.2.24 2003/11/11 17:18:18 silby Exp $
35  * $DragonFly: src/sys/kern/uipc_socket.c,v 1.15 2004/03/05 16:57:15 hsu 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/socketops.h>
56 #include <sys/resourcevar.h>
57 #include <sys/signalvar.h>
58 #include <sys/sysctl.h>
59 #include <sys/uio.h>
60 #include <sys/jail.h>
61 #include <vm/vm_zone.h>
62 
63 #include <machine/limits.h>
64 
65 #ifdef INET
66 static int	 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
67 #endif /* INET */
68 
69 static void 	filt_sordetach(struct knote *kn);
70 static int 	filt_soread(struct knote *kn, long hint);
71 static void 	filt_sowdetach(struct knote *kn);
72 static int	filt_sowrite(struct knote *kn, long hint);
73 static int	filt_solisten(struct knote *kn, long hint);
74 
75 static struct filterops solisten_filtops =
76 	{ 1, NULL, filt_sordetach, filt_solisten };
77 static struct filterops soread_filtops =
78 	{ 1, NULL, filt_sordetach, filt_soread };
79 static struct filterops sowrite_filtops =
80 	{ 1, NULL, filt_sowdetach, filt_sowrite };
81 
82 struct	vm_zone *socket_zone;
83 so_gen_t	so_gencnt;	/* generation count for sockets */
84 
85 MALLOC_DEFINE(M_SONAME, "soname", "socket name");
86 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");
87 
88 SYSCTL_DECL(_kern_ipc);
89 
90 static int somaxconn = SOMAXCONN;
91 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW,
92     &somaxconn, 0, "Maximum pending socket connection queue size");
93 
94 /*
95  * Socket operation routines.
96  * These routines are called by the routines in
97  * sys_socket.c or from a system process, and
98  * implement the semantics of socket operations by
99  * switching out to the protocol specific routines.
100  */
101 
102 /*
103  * Get a socket structure from our zone, and initialize it.
104  * We don't implement `waitok' yet (see comments in uipc_domain.c).
105  * Note that it would probably be better to allocate socket
106  * and PCB at the same time, but I'm not convinced that all
107  * the protocols can be easily modified to do this.
108  */
109 struct socket *
110 soalloc(waitok)
111 	int waitok;
112 {
113 	struct socket *so;
114 
115 	so = zalloc(socket_zone);
116 	if (so) {
117 		/* XXX race condition for reentrant kernel */
118 		bzero(so, sizeof *so);
119 		so->so_gencnt = ++so_gencnt;
120 		TAILQ_INIT(&so->so_aiojobq);
121 	}
122 	return so;
123 }
124 
125 int
126 socreate(int dom, struct socket **aso, int type,
127 	int proto, struct thread *td)
128 {
129 	struct proc *p = td->td_proc;
130 	struct protosw *prp;
131 	struct socket *so;
132 	struct pru_attach_info ai;
133 	int error;
134 
135 	if (proto)
136 		prp = pffindproto(dom, proto, type);
137 	else
138 		prp = pffindtype(dom, type);
139 
140 	if (prp == 0 || prp->pr_usrreqs->pru_attach == 0)
141 		return (EPROTONOSUPPORT);
142 
143 	if (p->p_ucred->cr_prison && jail_socket_unixiproute_only &&
144 	    prp->pr_domain->dom_family != PF_LOCAL &&
145 	    prp->pr_domain->dom_family != PF_INET &&
146 	    prp->pr_domain->dom_family != PF_ROUTE) {
147 		return (EPROTONOSUPPORT);
148 	}
149 
150 	if (prp->pr_type != type)
151 		return (EPROTOTYPE);
152 	so = soalloc(p != 0);
153 	if (so == 0)
154 		return (ENOBUFS);
155 
156 	TAILQ_INIT(&so->so_incomp);
157 	TAILQ_INIT(&so->so_comp);
158 	so->so_type = type;
159 	so->so_cred = crhold(p->p_ucred);
160 	so->so_proto = prp;
161 	ai.sb_rlimit = &p->p_rlimit[RLIMIT_SBSIZE];
162 	ai.p_ucred = p->p_ucred;
163 	ai.fd_rdir = p->p_fd->fd_rdir;
164 	error = so_pru_attach(so, proto, &ai);
165 	if (error) {
166 		so->so_state |= SS_NOFDREF;
167 		sofree(so);
168 		return (error);
169 	}
170 	*aso = so;
171 	return (0);
172 }
173 
174 int
175 sobind(struct socket *so, struct sockaddr *nam, struct thread *td)
176 {
177 	int s = splnet();
178 	int error;
179 
180 	error = so_pru_bind(so, nam, td);
181 	splx(s);
182 	return (error);
183 }
184 
185 void
186 sodealloc(struct socket *so)
187 {
188 
189 	so->so_gencnt = ++so_gencnt;
190 	if (so->so_rcv.sb_hiwat)
191 		(void)chgsbsize(so->so_cred->cr_uidinfo,
192 		    &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
193 	if (so->so_snd.sb_hiwat)
194 		(void)chgsbsize(so->so_cred->cr_uidinfo,
195 		    &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
196 #ifdef INET
197 	if (so->so_accf != NULL) {
198 		if (so->so_accf->so_accept_filter != NULL &&
199 			so->so_accf->so_accept_filter->accf_destroy != NULL) {
200 			so->so_accf->so_accept_filter->accf_destroy(so);
201 		}
202 		if (so->so_accf->so_accept_filter_str != NULL)
203 			FREE(so->so_accf->so_accept_filter_str, M_ACCF);
204 		FREE(so->so_accf, M_ACCF);
205 	}
206 #endif /* INET */
207 	crfree(so->so_cred);
208 	zfree(socket_zone, so);
209 }
210 
211 int
212 solisten(struct socket *so, int backlog, struct thread *td)
213 {
214 	int s, error;
215 
216 	s = splnet();
217  	if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) {
218  		splx(s);
219  		return (EINVAL);
220  	}
221 
222 	error = so_pru_listen(so, td);
223 	if (error) {
224 		splx(s);
225 		return (error);
226 	}
227 	if (TAILQ_EMPTY(&so->so_comp))
228 		so->so_options |= SO_ACCEPTCONN;
229 	if (backlog < 0 || backlog > somaxconn)
230 		backlog = somaxconn;
231 	so->so_qlimit = backlog;
232 	splx(s);
233 	return (0);
234 }
235 
236 void
237 sofree(struct socket *so)
238 {
239 	struct socket *head = so->so_head;
240 
241 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
242 		return;
243 	if (head != NULL) {
244 		if (so->so_state & SS_INCOMP) {
245 			TAILQ_REMOVE(&head->so_incomp, so, so_list);
246 			head->so_incqlen--;
247 		} else if (so->so_state & SS_COMP) {
248 			/*
249 			 * We must not decommission a socket that's
250 			 * on the accept(2) queue.  If we do, then
251 			 * accept(2) may hang after select(2) indicated
252 			 * that the listening socket was ready.
253 			 */
254 			return;
255 		} else {
256 			panic("sofree: not queued");
257 		}
258 		so->so_state &= ~SS_INCOMP;
259 		so->so_head = NULL;
260 	}
261 	sbrelease(&so->so_snd, so);
262 	sorflush(so);
263 	sodealloc(so);
264 }
265 
266 /*
267  * Close a socket on last file table reference removal.
268  * Initiate disconnect if connected.
269  * Free socket when disconnect complete.
270  */
271 int
272 soclose(struct socket *so)
273 {
274 	int s = splnet();		/* conservative */
275 	int error = 0;
276 
277 	funsetown(so->so_sigio);
278 	if (so->so_options & SO_ACCEPTCONN) {
279 		struct socket *sp, *sonext;
280 
281 		sp = TAILQ_FIRST(&so->so_incomp);
282 		for (; sp != NULL; sp = sonext) {
283 			sonext = TAILQ_NEXT(sp, so_list);
284 			(void) soabort(sp);
285 		}
286 		for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
287 			sonext = TAILQ_NEXT(sp, so_list);
288 			/* Dequeue from so_comp since sofree() won't do it */
289 			TAILQ_REMOVE(&so->so_comp, sp, so_list);
290 			so->so_qlen--;
291 			sp->so_state &= ~SS_COMP;
292 			sp->so_head = NULL;
293 			(void) soabort(sp);
294 		}
295 	}
296 	if (so->so_pcb == 0)
297 		goto discard;
298 	if (so->so_state & SS_ISCONNECTED) {
299 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
300 			error = sodisconnect(so);
301 			if (error)
302 				goto drop;
303 		}
304 		if (so->so_options & SO_LINGER) {
305 			if ((so->so_state & SS_ISDISCONNECTING) &&
306 			    (so->so_state & SS_NBIO))
307 				goto drop;
308 			while (so->so_state & SS_ISCONNECTED) {
309 				error = tsleep((caddr_t)&so->so_timeo,
310 				    PCATCH, "soclos", so->so_linger * hz);
311 				if (error)
312 					break;
313 			}
314 		}
315 	}
316 drop:
317 	if (so->so_pcb) {
318 		int error2;
319 
320 		error2 = so_pru_detach(so);
321 		if (error == 0)
322 			error = error2;
323 	}
324 discard:
325 	if (so->so_state & SS_NOFDREF)
326 		panic("soclose: NOFDREF");
327 	so->so_state |= SS_NOFDREF;
328 	sofree(so);
329 	splx(s);
330 	return (error);
331 }
332 
333 /*
334  * Must be called at splnet...
335  */
336 int
337 soabort(so)
338 	struct socket *so;
339 {
340 	int error;
341 
342 	error = so_pru_abort(so);
343 	if (error) {
344 		sofree(so);
345 		return error;
346 	}
347 	return (0);
348 }
349 
350 int
351 soaccept(struct socket *so, struct sockaddr **nam)
352 {
353 	int s = splnet();
354 	int error;
355 
356 	if ((so->so_state & SS_NOFDREF) == 0)
357 		panic("soaccept: !NOFDREF");
358 	so->so_state &= ~SS_NOFDREF;
359 	error = so_pru_accept(so, nam);
360 	splx(s);
361 	return (error);
362 }
363 
364 int
365 soconnect(struct socket *so, struct sockaddr *nam, struct thread *td)
366 {
367 	int s;
368 	int error;
369 
370 	if (so->so_options & SO_ACCEPTCONN)
371 		return (EOPNOTSUPP);
372 	s = splnet();
373 	/*
374 	 * If protocol is connection-based, can only connect once.
375 	 * Otherwise, if connected, try to disconnect first.
376 	 * This allows user to disconnect by connecting to, e.g.,
377 	 * a null address.
378 	 */
379 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
380 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
381 	    (error = sodisconnect(so))))
382 		error = EISCONN;
383 	else
384 		error = so_pru_connect(so, nam, td);
385 	splx(s);
386 	return (error);
387 }
388 
389 int
390 soconnect2(struct socket *so1, struct socket *so2)
391 {
392 	int s = splnet();
393 	int error;
394 
395 	error = so_pru_connect2(so1, so2);
396 	splx(s);
397 	return (error);
398 }
399 
400 int
401 sodisconnect(struct socket *so)
402 {
403 	int s = splnet();
404 	int error;
405 
406 	if ((so->so_state & SS_ISCONNECTED) == 0) {
407 		error = ENOTCONN;
408 		goto bad;
409 	}
410 	if (so->so_state & SS_ISDISCONNECTING) {
411 		error = EALREADY;
412 		goto bad;
413 	}
414 	error = so_pru_disconnect(so);
415 bad:
416 	splx(s);
417 	return (error);
418 }
419 
420 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
421 /*
422  * Send on a socket.
423  * If send must go all at once and message is larger than
424  * send buffering, then hard error.
425  * Lock against other senders.
426  * If must go all at once and not enough room now, then
427  * inform user that this would block and do nothing.
428  * Otherwise, if nonblocking, send as much as possible.
429  * The data to be sent is described by "uio" if nonzero,
430  * otherwise by the mbuf chain "top" (which must be null
431  * if uio is not).  Data provided in mbuf chain must be small
432  * enough to send all at once.
433  *
434  * Returns nonzero on error, timeout or signal; callers
435  * must check for short counts if EINTR/ERESTART are returned.
436  * Data and control buffers are freed on return.
437  */
438 int
439 sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
440 	struct mbuf *top, struct mbuf *control, int flags,
441 	struct thread *td)
442 {
443 	struct mbuf **mp;
444 	struct mbuf *m;
445 	long space, len, resid;
446 	int clen = 0, error, s, dontroute, mlen;
447 	int atomic = sosendallatonce(so) || top;
448 	int pru_flags;
449 
450 	if (uio)
451 		resid = uio->uio_resid;
452 	else
453 		resid = top->m_pkthdr.len;
454 	/*
455 	 * In theory resid should be unsigned.
456 	 * However, space must be signed, as it might be less than 0
457 	 * if we over-committed, and we must use a signed comparison
458 	 * of space and resid.  On the other hand, a negative resid
459 	 * causes us to loop sending 0-length segments to the protocol.
460 	 *
461 	 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
462 	 * type sockets since that's an error.
463 	 */
464 	if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
465 		error = EINVAL;
466 		goto out;
467 	}
468 
469 	dontroute =
470 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
471 	    (so->so_proto->pr_flags & PR_ATOMIC);
472 	if (td->td_proc && td->td_proc->p_stats)
473 		td->td_proc->p_stats->p_ru.ru_msgsnd++;
474 	if (control)
475 		clen = control->m_len;
476 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
477 
478 restart:
479 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
480 	if (error)
481 		goto out;
482 	do {
483 		s = splnet();
484 		if (so->so_state & SS_CANTSENDMORE)
485 			snderr(EPIPE);
486 		if (so->so_error) {
487 			error = so->so_error;
488 			so->so_error = 0;
489 			splx(s);
490 			goto release;
491 		}
492 		if ((so->so_state & SS_ISCONNECTED) == 0) {
493 			/*
494 			 * `sendto' and `sendmsg' is allowed on a connection-
495 			 * based socket if it supports implied connect.
496 			 * Return ENOTCONN if not connected and no address is
497 			 * supplied.
498 			 */
499 			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
500 			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
501 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
502 				    !(resid == 0 && clen != 0))
503 					snderr(ENOTCONN);
504 			} else if (addr == 0)
505 			    snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
506 				   ENOTCONN : EDESTADDRREQ);
507 		}
508 		space = sbspace(&so->so_snd);
509 		if (flags & MSG_OOB)
510 			space += 1024;
511 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
512 		    clen > so->so_snd.sb_hiwat)
513 			snderr(EMSGSIZE);
514 		if (space < resid + clen && uio &&
515 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
516 			if (so->so_state & SS_NBIO)
517 				snderr(EWOULDBLOCK);
518 			sbunlock(&so->so_snd);
519 			error = sbwait(&so->so_snd);
520 			splx(s);
521 			if (error)
522 				goto out;
523 			goto restart;
524 		}
525 		splx(s);
526 		mp = &top;
527 		space -= clen;
528 		do {
529 		    if (uio == NULL) {
530 			/*
531 			 * Data is prepackaged in "top".
532 			 */
533 			resid = 0;
534 			if (flags & MSG_EOR)
535 				top->m_flags |= M_EOR;
536 		    } else do {
537 			if (top == 0) {
538 				MGETHDR(m, M_WAIT, MT_DATA);
539 				if (m == NULL) {
540 					error = ENOBUFS;
541 					goto release;
542 				}
543 				mlen = MHLEN;
544 				m->m_pkthdr.len = 0;
545 				m->m_pkthdr.rcvif = (struct ifnet *)0;
546 			} else {
547 				MGET(m, M_WAIT, MT_DATA);
548 				if (m == NULL) {
549 					error = ENOBUFS;
550 					goto release;
551 				}
552 				mlen = MLEN;
553 			}
554 			if (resid >= MINCLSIZE) {
555 				MCLGET(m, M_WAIT);
556 				if ((m->m_flags & M_EXT) == 0)
557 					goto nopages;
558 				mlen = MCLBYTES;
559 				len = min(min(mlen, resid), space);
560 			} else {
561 nopages:
562 				len = min(min(mlen, resid), space);
563 				/*
564 				 * For datagram protocols, leave room
565 				 * for protocol headers in first mbuf.
566 				 */
567 				if (atomic && top == 0 && len < mlen)
568 					MH_ALIGN(m, len);
569 			}
570 			space -= len;
571 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
572 			resid = uio->uio_resid;
573 			m->m_len = len;
574 			*mp = m;
575 			top->m_pkthdr.len += len;
576 			if (error)
577 				goto release;
578 			mp = &m->m_next;
579 			if (resid <= 0) {
580 				if (flags & MSG_EOR)
581 					top->m_flags |= M_EOR;
582 				break;
583 			}
584 		    } while (space > 0 && atomic);
585 		    if (dontroute)
586 			    so->so_options |= SO_DONTROUTE;
587 		    if (flags & MSG_OOB) {
588 		    	    pru_flags = PRUS_OOB;
589 		    } else if ((flags & MSG_EOF) &&
590 		    	       (so->so_proto->pr_flags & PR_IMPLOPCL) &&
591 		    	       (resid <= 0)) {
592 			    /*
593 			     * If the user set MSG_EOF, the protocol
594 			     * understands this flag and nothing left to
595 			     * send then use PRU_SEND_EOF instead of PRU_SEND.
596 			     */
597 		    	    pru_flags = PRUS_EOF;
598 		    } else if (resid > 0 && space > 0) {
599 			    /* If there is more to send, set PRUS_MORETOCOME */
600 		    	    pru_flags = PRUS_MORETOCOME;
601 		    } else {
602 		    	    pru_flags = 0;
603 		    }
604 		    s = splnet();				/* XXX */
605 		    /*
606 		     * XXX all the SS_CANTSENDMORE checks previously
607 		     * done could be out of date.  We could have recieved
608 		     * a reset packet in an interrupt or maybe we slept
609 		     * while doing page faults in uiomove() etc. We could
610 		     * probably recheck again inside the splnet() protection
611 		     * here, but there are probably other places that this
612 		     * also happens.  We must rethink this.
613 		     */
614 		    error = so_pru_send(so, pru_flags, top, addr, control, td);
615 		    splx(s);
616 		    if (dontroute)
617 			    so->so_options &= ~SO_DONTROUTE;
618 		    clen = 0;
619 		    control = 0;
620 		    top = 0;
621 		    mp = &top;
622 		    if (error)
623 			    goto release;
624 		} while (resid && space > 0);
625 	} while (resid);
626 
627 release:
628 	sbunlock(&so->so_snd);
629 out:
630 	if (top)
631 		m_freem(top);
632 	if (control)
633 		m_freem(control);
634 	return (error);
635 }
636 
637 /*
638  * Implement receive operations on a socket.
639  * We depend on the way that records are added to the sockbuf
640  * by sbappend*.  In particular, each record (mbufs linked through m_next)
641  * must begin with an address if the protocol so specifies,
642  * followed by an optional mbuf or mbufs containing ancillary data,
643  * and then zero or more mbufs of data.
644  * In order to avoid blocking network interrupts for the entire time here,
645  * we splx() while doing the actual copy to user space.
646  * Although the sockbuf is locked, new data may still be appended,
647  * and thus we must maintain consistency of the sockbuf during that time.
648  *
649  * The caller may receive the data as a single mbuf chain by supplying
650  * an mbuf **mp0 for use in returning the chain.  The uio is then used
651  * only for the count in uio_resid.
652  */
653 int
654 soreceive(so, psa, uio, mp0, controlp, flagsp)
655 	struct socket *so;
656 	struct sockaddr **psa;
657 	struct uio *uio;
658 	struct mbuf **mp0;
659 	struct mbuf **controlp;
660 	int *flagsp;
661 {
662 	struct mbuf *m, **mp;
663 	int flags, len, error, s, offset;
664 	struct protosw *pr = so->so_proto;
665 	struct mbuf *nextrecord;
666 	int moff, type = 0;
667 	int orig_resid = uio->uio_resid;
668 
669 	mp = mp0;
670 	if (psa)
671 		*psa = 0;
672 	if (controlp)
673 		*controlp = 0;
674 	if (flagsp)
675 		flags = *flagsp &~ MSG_EOR;
676 	else
677 		flags = 0;
678 	if (flags & MSG_OOB) {
679 		m = m_get(M_WAIT, MT_DATA);
680 		if (m == NULL)
681 			return (ENOBUFS);
682 		error = so_pru_rcvoob(so, m, flags & MSG_PEEK);
683 		if (error)
684 			goto bad;
685 		do {
686 			error = uiomove(mtod(m, caddr_t),
687 			    (int) min(uio->uio_resid, m->m_len), uio);
688 			m = m_free(m);
689 		} while (uio->uio_resid && error == 0 && m);
690 bad:
691 		if (m)
692 			m_freem(m);
693 		return (error);
694 	}
695 	if (mp)
696 		*mp = (struct mbuf *)0;
697 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
698 		so_pru_rcvd(so, 0);
699 
700 restart:
701 	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
702 	if (error)
703 		return (error);
704 	s = splnet();
705 
706 	m = so->so_rcv.sb_mb;
707 	/*
708 	 * If we have less data than requested, block awaiting more
709 	 * (subject to any timeout) if:
710 	 *   1. the current count is less than the low water mark, or
711 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
712 	 *	receive operation at once if we block (resid <= hiwat).
713 	 *   3. MSG_DONTWAIT is not set
714 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
715 	 * we have to do the receive in sections, and thus risk returning
716 	 * a short count if a timeout or signal occurs after we start.
717 	 */
718 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
719 	    so->so_rcv.sb_cc < uio->uio_resid) &&
720 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
721 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
722 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
723 		KASSERT(m != 0 || !so->so_rcv.sb_cc, ("receive 1"));
724 		if (so->so_error) {
725 			if (m)
726 				goto dontblock;
727 			error = so->so_error;
728 			if ((flags & MSG_PEEK) == 0)
729 				so->so_error = 0;
730 			goto release;
731 		}
732 		if (so->so_state & SS_CANTRCVMORE) {
733 			if (m)
734 				goto dontblock;
735 			else
736 				goto release;
737 		}
738 		for (; m; m = m->m_next)
739 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
740 				m = so->so_rcv.sb_mb;
741 				goto dontblock;
742 			}
743 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
744 		    (pr->pr_flags & PR_CONNREQUIRED)) {
745 			error = ENOTCONN;
746 			goto release;
747 		}
748 		if (uio->uio_resid == 0)
749 			goto release;
750 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
751 			error = EWOULDBLOCK;
752 			goto release;
753 		}
754 		sbunlock(&so->so_rcv);
755 		error = sbwait(&so->so_rcv);
756 		splx(s);
757 		if (error)
758 			return (error);
759 		goto restart;
760 	}
761 dontblock:
762 	if (uio->uio_td && uio->uio_td->td_proc)
763 		uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++;
764 	nextrecord = m->m_nextpkt;
765 	if (pr->pr_flags & PR_ADDR) {
766 		KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
767 		orig_resid = 0;
768 		if (psa)
769 			*psa = dup_sockaddr(mtod(m, struct sockaddr *),
770 					    mp0 == 0);
771 		if (flags & MSG_PEEK) {
772 			m = m->m_next;
773 		} else {
774 			sbfree(&so->so_rcv, m);
775 			so->so_rcv.sb_mb = m_free(m);
776 			m = so->so_rcv.sb_mb;
777 		}
778 	}
779 	while (m && m->m_type == MT_CONTROL && error == 0) {
780 		if (flags & MSG_PEEK) {
781 			if (controlp)
782 				*controlp = m_copy(m, 0, m->m_len);
783 			m = m->m_next;
784 		} else {
785 			sbfree(&so->so_rcv, m);
786 			if (controlp) {
787 				if (pr->pr_domain->dom_externalize &&
788 				    mtod(m, struct cmsghdr *)->cmsg_type ==
789 				    SCM_RIGHTS)
790 				   error = (*pr->pr_domain->dom_externalize)(m);
791 				*controlp = m;
792 				so->so_rcv.sb_mb = m->m_next;
793 				m->m_next = 0;
794 				m = so->so_rcv.sb_mb;
795 			} else {
796 				so->so_rcv.sb_mb = m_free(m);
797 				m = so->so_rcv.sb_mb;
798 			}
799 		}
800 		if (controlp) {
801 			orig_resid = 0;
802 			controlp = &(*controlp)->m_next;
803 		}
804 	}
805 	if (m) {
806 		if ((flags & MSG_PEEK) == 0)
807 			m->m_nextpkt = nextrecord;
808 		type = m->m_type;
809 		if (type == MT_OOBDATA)
810 			flags |= MSG_OOB;
811 	}
812 	moff = 0;
813 	offset = 0;
814 	while (m && uio->uio_resid > 0 && error == 0) {
815 		if (m->m_type == MT_OOBDATA) {
816 			if (type != MT_OOBDATA)
817 				break;
818 		} else if (type == MT_OOBDATA)
819 			break;
820 		else
821 		    KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
822 			("receive 3"));
823 		so->so_state &= ~SS_RCVATMARK;
824 		len = uio->uio_resid;
825 		if (so->so_oobmark && len > so->so_oobmark - offset)
826 			len = so->so_oobmark - offset;
827 		if (len > m->m_len - moff)
828 			len = m->m_len - moff;
829 		/*
830 		 * If mp is set, just pass back the mbufs.
831 		 * Otherwise copy them out via the uio, then free.
832 		 * Sockbuf must be consistent here (points to current mbuf,
833 		 * it points to next record) when we drop priority;
834 		 * we must note any additions to the sockbuf when we
835 		 * block interrupts again.
836 		 */
837 		if (mp == 0) {
838 			splx(s);
839 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
840 			s = splnet();
841 			if (error)
842 				goto release;
843 		} else
844 			uio->uio_resid -= len;
845 		if (len == m->m_len - moff) {
846 			if (m->m_flags & M_EOR)
847 				flags |= MSG_EOR;
848 			if (flags & MSG_PEEK) {
849 				m = m->m_next;
850 				moff = 0;
851 			} else {
852 				nextrecord = m->m_nextpkt;
853 				sbfree(&so->so_rcv, m);
854 				if (mp) {
855 					*mp = m;
856 					mp = &m->m_next;
857 					so->so_rcv.sb_mb = m = m->m_next;
858 					*mp = (struct mbuf *)0;
859 				} else {
860 					so->so_rcv.sb_mb = m = m_free(m);
861 				}
862 				if (m)
863 					m->m_nextpkt = nextrecord;
864 			}
865 		} else {
866 			if (flags & MSG_PEEK)
867 				moff += len;
868 			else {
869 				if (mp)
870 					*mp = m_copym(m, 0, len, M_WAIT);
871 				m->m_data += len;
872 				m->m_len -= len;
873 				so->so_rcv.sb_cc -= len;
874 			}
875 		}
876 		if (so->so_oobmark) {
877 			if ((flags & MSG_PEEK) == 0) {
878 				so->so_oobmark -= len;
879 				if (so->so_oobmark == 0) {
880 					so->so_state |= SS_RCVATMARK;
881 					break;
882 				}
883 			} else {
884 				offset += len;
885 				if (offset == so->so_oobmark)
886 					break;
887 			}
888 		}
889 		if (flags & MSG_EOR)
890 			break;
891 		/*
892 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
893 		 * we must not quit until "uio->uio_resid == 0" or an error
894 		 * termination.  If a signal/timeout occurs, return
895 		 * with a short count but without error.
896 		 * Keep sockbuf locked against other readers.
897 		 */
898 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
899 		    !sosendallatonce(so) && !nextrecord) {
900 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
901 				break;
902 			/*
903 			 * The window might have closed to zero, make
904 			 * sure we send an ack now that we've drained
905 			 * the buffer or we might end up blocking until
906 			 * the idle takes over (5 seconds).
907 			 */
908 			if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
909 				so_pru_rcvd(so, flags);
910 			error = sbwait(&so->so_rcv);
911 			if (error) {
912 				sbunlock(&so->so_rcv);
913 				splx(s);
914 				return (0);
915 			}
916 			m = so->so_rcv.sb_mb;
917 			if (m)
918 				nextrecord = m->m_nextpkt;
919 		}
920 	}
921 
922 	if (m && pr->pr_flags & PR_ATOMIC) {
923 		flags |= MSG_TRUNC;
924 		if ((flags & MSG_PEEK) == 0)
925 			(void) sbdroprecord(&so->so_rcv);
926 	}
927 	if ((flags & MSG_PEEK) == 0) {
928 		if (m == 0)
929 			so->so_rcv.sb_mb = nextrecord;
930 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
931 			so_pru_rcvd(so, flags);
932 	}
933 	if (orig_resid == uio->uio_resid && orig_resid &&
934 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
935 		sbunlock(&so->so_rcv);
936 		splx(s);
937 		goto restart;
938 	}
939 
940 	if (flagsp)
941 		*flagsp |= flags;
942 release:
943 	sbunlock(&so->so_rcv);
944 	splx(s);
945 	return (error);
946 }
947 
948 int
949 soshutdown(so, how)
950 	struct socket *so;
951 	int how;
952 {
953 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
954 		return (EINVAL);
955 
956 	if (how != SHUT_WR)
957 		sorflush(so);
958 	if (how != SHUT_RD)
959 		return (so_pru_shutdown(so));
960 	return (0);
961 }
962 
963 void
964 sorflush(so)
965 	struct socket *so;
966 {
967 	struct sockbuf *sb = &so->so_rcv;
968 	struct protosw *pr = so->so_proto;
969 	int s;
970 	struct sockbuf asb;
971 
972 	sb->sb_flags |= SB_NOINTR;
973 	(void) sblock(sb, M_WAITOK);
974 	s = splimp();
975 	socantrcvmore(so);
976 	sbunlock(sb);
977 	asb = *sb;
978 	bzero((caddr_t)sb, sizeof (*sb));
979 	if (asb.sb_flags & SB_KNOTE) {
980 		sb->sb_sel.si_note = asb.sb_sel.si_note;
981 		sb->sb_flags = SB_KNOTE;
982 	}
983 	splx(s);
984 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
985 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
986 	sbrelease(&asb, so);
987 }
988 
989 #ifdef INET
990 static int
991 do_setopt_accept_filter(so, sopt)
992 	struct	socket *so;
993 	struct	sockopt *sopt;
994 {
995 	struct accept_filter_arg	*afap = NULL;
996 	struct accept_filter	*afp;
997 	struct so_accf	*af = so->so_accf;
998 	int	error = 0;
999 
1000 	/* do not set/remove accept filters on non listen sockets */
1001 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
1002 		error = EINVAL;
1003 		goto out;
1004 	}
1005 
1006 	/* removing the filter */
1007 	if (sopt == NULL) {
1008 		if (af != NULL) {
1009 			if (af->so_accept_filter != NULL &&
1010 				af->so_accept_filter->accf_destroy != NULL) {
1011 				af->so_accept_filter->accf_destroy(so);
1012 			}
1013 			if (af->so_accept_filter_str != NULL) {
1014 				FREE(af->so_accept_filter_str, M_ACCF);
1015 			}
1016 			FREE(af, M_ACCF);
1017 			so->so_accf = NULL;
1018 		}
1019 		so->so_options &= ~SO_ACCEPTFILTER;
1020 		return (0);
1021 	}
1022 	/* adding a filter */
1023 	/* must remove previous filter first */
1024 	if (af != NULL) {
1025 		error = EINVAL;
1026 		goto out;
1027 	}
1028 	/* don't put large objects on the kernel stack */
1029 	MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1030 	error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1031 	afap->af_name[sizeof(afap->af_name)-1] = '\0';
1032 	afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1033 	if (error)
1034 		goto out;
1035 	afp = accept_filt_get(afap->af_name);
1036 	if (afp == NULL) {
1037 		error = ENOENT;
1038 		goto out;
1039 	}
1040 	MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK);
1041 	bzero(af, sizeof(*af));
1042 	if (afp->accf_create != NULL) {
1043 		if (afap->af_name[0] != '\0') {
1044 			int len = strlen(afap->af_name) + 1;
1045 
1046 			MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1047 			strcpy(af->so_accept_filter_str, afap->af_name);
1048 		}
1049 		af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1050 		if (af->so_accept_filter_arg == NULL) {
1051 			FREE(af->so_accept_filter_str, M_ACCF);
1052 			FREE(af, M_ACCF);
1053 			so->so_accf = NULL;
1054 			error = EINVAL;
1055 			goto out;
1056 		}
1057 	}
1058 	af->so_accept_filter = afp;
1059 	so->so_accf = af;
1060 	so->so_options |= SO_ACCEPTFILTER;
1061 out:
1062 	if (afap != NULL)
1063 		FREE(afap, M_TEMP);
1064 	return (error);
1065 }
1066 #endif /* INET */
1067 
1068 /*
1069  * Perhaps this routine, and sooptcopyout(), below, ought to come in
1070  * an additional variant to handle the case where the option value needs
1071  * to be some kind of integer, but not a specific size.
1072  * In addition to their use here, these functions are also called by the
1073  * protocol-level pr_ctloutput() routines.
1074  */
1075 int
1076 sooptcopyin(sopt, buf, len, minlen)
1077 	struct	sockopt *sopt;
1078 	void	*buf;
1079 	size_t	len;
1080 	size_t	minlen;
1081 {
1082 	size_t	valsize;
1083 
1084 	/*
1085 	 * If the user gives us more than we wanted, we ignore it,
1086 	 * but if we don't get the minimum length the caller
1087 	 * wants, we return EINVAL.  On success, sopt->sopt_valsize
1088 	 * is set to however much we actually retrieved.
1089 	 */
1090 	if ((valsize = sopt->sopt_valsize) < minlen)
1091 		return EINVAL;
1092 	if (valsize > len)
1093 		sopt->sopt_valsize = valsize = len;
1094 
1095 	if (sopt->sopt_td != NULL)
1096 		return (copyin(sopt->sopt_val, buf, valsize));
1097 
1098 	bcopy(sopt->sopt_val, buf, valsize);
1099 	return 0;
1100 }
1101 
1102 int
1103 sosetopt(so, sopt)
1104 	struct socket *so;
1105 	struct sockopt *sopt;
1106 {
1107 	int	error, optval;
1108 	struct	linger l;
1109 	struct	timeval tv;
1110 	u_long  val;
1111 
1112 	error = 0;
1113 	if (sopt->sopt_level != SOL_SOCKET) {
1114 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1115 			return (so_pr_ctloutput(so, sopt));
1116 		}
1117 		error = ENOPROTOOPT;
1118 	} else {
1119 		switch (sopt->sopt_name) {
1120 #ifdef INET
1121 		case SO_ACCEPTFILTER:
1122 			error = do_setopt_accept_filter(so, sopt);
1123 			if (error)
1124 				goto bad;
1125 			break;
1126 #endif /* INET */
1127 		case SO_LINGER:
1128 			error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1129 			if (error)
1130 				goto bad;
1131 
1132 			so->so_linger = l.l_linger;
1133 			if (l.l_onoff)
1134 				so->so_options |= SO_LINGER;
1135 			else
1136 				so->so_options &= ~SO_LINGER;
1137 			break;
1138 
1139 		case SO_DEBUG:
1140 		case SO_KEEPALIVE:
1141 		case SO_DONTROUTE:
1142 		case SO_USELOOPBACK:
1143 		case SO_BROADCAST:
1144 		case SO_REUSEADDR:
1145 		case SO_REUSEPORT:
1146 		case SO_OOBINLINE:
1147 		case SO_TIMESTAMP:
1148 			error = sooptcopyin(sopt, &optval, sizeof optval,
1149 					    sizeof optval);
1150 			if (error)
1151 				goto bad;
1152 			if (optval)
1153 				so->so_options |= sopt->sopt_name;
1154 			else
1155 				so->so_options &= ~sopt->sopt_name;
1156 			break;
1157 
1158 		case SO_SNDBUF:
1159 		case SO_RCVBUF:
1160 		case SO_SNDLOWAT:
1161 		case SO_RCVLOWAT:
1162 			error = sooptcopyin(sopt, &optval, sizeof optval,
1163 					    sizeof optval);
1164 			if (error)
1165 				goto bad;
1166 
1167 			/*
1168 			 * Values < 1 make no sense for any of these
1169 			 * options, so disallow them.
1170 			 */
1171 			if (optval < 1) {
1172 				error = EINVAL;
1173 				goto bad;
1174 			}
1175 
1176 			switch (sopt->sopt_name) {
1177 			case SO_SNDBUF:
1178 			case SO_RCVBUF:
1179 				if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1180 				    &so->so_snd : &so->so_rcv, (u_long)optval,
1181 				    so,
1182 				    &curproc->p_rlimit[RLIMIT_SBSIZE]) == 0) {
1183 					error = ENOBUFS;
1184 					goto bad;
1185 				}
1186 				break;
1187 
1188 			/*
1189 			 * Make sure the low-water is never greater than
1190 			 * the high-water.
1191 			 */
1192 			case SO_SNDLOWAT:
1193 				so->so_snd.sb_lowat =
1194 				    (optval > so->so_snd.sb_hiwat) ?
1195 				    so->so_snd.sb_hiwat : optval;
1196 				break;
1197 			case SO_RCVLOWAT:
1198 				so->so_rcv.sb_lowat =
1199 				    (optval > so->so_rcv.sb_hiwat) ?
1200 				    so->so_rcv.sb_hiwat : optval;
1201 				break;
1202 			}
1203 			break;
1204 
1205 		case SO_SNDTIMEO:
1206 		case SO_RCVTIMEO:
1207 			error = sooptcopyin(sopt, &tv, sizeof tv,
1208 					    sizeof tv);
1209 			if (error)
1210 				goto bad;
1211 
1212 			/* assert(hz > 0); */
1213 			if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1214 			    tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1215 				error = EDOM;
1216 				goto bad;
1217 			}
1218 			/* assert(tick > 0); */
1219 			/* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1220 			val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1221 			if (val > SHRT_MAX) {
1222 				error = EDOM;
1223 				goto bad;
1224 			}
1225 			if (val == 0 && tv.tv_usec != 0)
1226 				val = 1;
1227 
1228 			switch (sopt->sopt_name) {
1229 			case SO_SNDTIMEO:
1230 				so->so_snd.sb_timeo = val;
1231 				break;
1232 			case SO_RCVTIMEO:
1233 				so->so_rcv.sb_timeo = val;
1234 				break;
1235 			}
1236 			break;
1237 		default:
1238 			error = ENOPROTOOPT;
1239 			break;
1240 		}
1241 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1242 			(void) so_pr_ctloutput(so, sopt);
1243 		}
1244 	}
1245 bad:
1246 	return (error);
1247 }
1248 
1249 /* Helper routine for getsockopt */
1250 int
1251 sooptcopyout(struct sockopt *sopt, const void *buf, size_t len)
1252 {
1253 	int	error;
1254 	size_t	valsize;
1255 
1256 	error = 0;
1257 
1258 	/*
1259 	 * Documented get behavior is that we always return a value,
1260 	 * possibly truncated to fit in the user's buffer.
1261 	 * Traditional behavior is that we always tell the user
1262 	 * precisely how much we copied, rather than something useful
1263 	 * like the total amount we had available for her.
1264 	 * Note that this interface is not idempotent; the entire answer must
1265 	 * generated ahead of time.
1266 	 */
1267 	valsize = min(len, sopt->sopt_valsize);
1268 	sopt->sopt_valsize = valsize;
1269 	if (sopt->sopt_val != 0) {
1270 		if (sopt->sopt_td != NULL)
1271 			error = copyout(buf, sopt->sopt_val, valsize);
1272 		else
1273 			bcopy(buf, sopt->sopt_val, valsize);
1274 	}
1275 	return error;
1276 }
1277 
1278 int
1279 sogetopt(so, sopt)
1280 	struct socket *so;
1281 	struct sockopt *sopt;
1282 {
1283 	int	error, optval;
1284 	struct	linger l;
1285 	struct	timeval tv;
1286 	struct accept_filter_arg *afap;
1287 
1288 	error = 0;
1289 	if (sopt->sopt_level != SOL_SOCKET) {
1290 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1291 			return (so_pr_ctloutput(so, sopt));
1292 		} else
1293 			return (ENOPROTOOPT);
1294 	} else {
1295 		switch (sopt->sopt_name) {
1296 #ifdef INET
1297 		case SO_ACCEPTFILTER:
1298 			if ((so->so_options & SO_ACCEPTCONN) == 0)
1299 				return (EINVAL);
1300 			MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1301 				M_TEMP, M_WAITOK);
1302 			bzero(afap, sizeof(*afap));
1303 			if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1304 				strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1305 				if (so->so_accf->so_accept_filter_str != NULL)
1306 					strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1307 			}
1308 			error = sooptcopyout(sopt, afap, sizeof(*afap));
1309 			FREE(afap, M_TEMP);
1310 			break;
1311 #endif /* INET */
1312 
1313 		case SO_LINGER:
1314 			l.l_onoff = so->so_options & SO_LINGER;
1315 			l.l_linger = so->so_linger;
1316 			error = sooptcopyout(sopt, &l, sizeof l);
1317 			break;
1318 
1319 		case SO_USELOOPBACK:
1320 		case SO_DONTROUTE:
1321 		case SO_DEBUG:
1322 		case SO_KEEPALIVE:
1323 		case SO_REUSEADDR:
1324 		case SO_REUSEPORT:
1325 		case SO_BROADCAST:
1326 		case SO_OOBINLINE:
1327 		case SO_TIMESTAMP:
1328 			optval = so->so_options & sopt->sopt_name;
1329 integer:
1330 			error = sooptcopyout(sopt, &optval, sizeof optval);
1331 			break;
1332 
1333 		case SO_TYPE:
1334 			optval = so->so_type;
1335 			goto integer;
1336 
1337 		case SO_ERROR:
1338 			optval = so->so_error;
1339 			so->so_error = 0;
1340 			goto integer;
1341 
1342 		case SO_SNDBUF:
1343 			optval = so->so_snd.sb_hiwat;
1344 			goto integer;
1345 
1346 		case SO_RCVBUF:
1347 			optval = so->so_rcv.sb_hiwat;
1348 			goto integer;
1349 
1350 		case SO_SNDLOWAT:
1351 			optval = so->so_snd.sb_lowat;
1352 			goto integer;
1353 
1354 		case SO_RCVLOWAT:
1355 			optval = so->so_rcv.sb_lowat;
1356 			goto integer;
1357 
1358 		case SO_SNDTIMEO:
1359 		case SO_RCVTIMEO:
1360 			optval = (sopt->sopt_name == SO_SNDTIMEO ?
1361 				  so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1362 
1363 			tv.tv_sec = optval / hz;
1364 			tv.tv_usec = (optval % hz) * tick;
1365 			error = sooptcopyout(sopt, &tv, sizeof tv);
1366 			break;
1367 
1368 		default:
1369 			error = ENOPROTOOPT;
1370 			break;
1371 		}
1372 		return (error);
1373 	}
1374 }
1375 
1376 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1377 int
1378 soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1379 {
1380 	struct mbuf *m, *m_prev;
1381 	int sopt_size = sopt->sopt_valsize;
1382 
1383 	MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1384 	if (m == 0)
1385 		return ENOBUFS;
1386 	if (sopt_size > MLEN) {
1387 		MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1388 		if ((m->m_flags & M_EXT) == 0) {
1389 			m_free(m);
1390 			return ENOBUFS;
1391 		}
1392 		m->m_len = min(MCLBYTES, sopt_size);
1393 	} else {
1394 		m->m_len = min(MLEN, sopt_size);
1395 	}
1396 	sopt_size -= m->m_len;
1397 	*mp = m;
1398 	m_prev = m;
1399 
1400 	while (sopt_size) {
1401 		MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1402 		if (m == 0) {
1403 			m_freem(*mp);
1404 			return ENOBUFS;
1405 		}
1406 		if (sopt_size > MLEN) {
1407 			MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1408 			if ((m->m_flags & M_EXT) == 0) {
1409 				m_freem(*mp);
1410 				return ENOBUFS;
1411 			}
1412 			m->m_len = min(MCLBYTES, sopt_size);
1413 		} else {
1414 			m->m_len = min(MLEN, sopt_size);
1415 		}
1416 		sopt_size -= m->m_len;
1417 		m_prev->m_next = m;
1418 		m_prev = m;
1419 	}
1420 	return 0;
1421 }
1422 
1423 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1424 int
1425 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1426 {
1427 	struct mbuf *m0 = m;
1428 
1429 	if (sopt->sopt_val == NULL)
1430 		return 0;
1431 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1432 		if (sopt->sopt_td != NULL) {
1433 			int error;
1434 
1435 			error = copyin(sopt->sopt_val, mtod(m, char *),
1436 				       m->m_len);
1437 			if (error != 0) {
1438 				m_freem(m0);
1439 				return(error);
1440 			}
1441 		} else
1442 			bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1443 		sopt->sopt_valsize -= m->m_len;
1444 		(caddr_t)sopt->sopt_val += m->m_len;
1445 		m = m->m_next;
1446 	}
1447 	if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1448 		panic("ip6_sooptmcopyin");
1449 	return 0;
1450 }
1451 
1452 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1453 int
1454 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1455 {
1456 	struct mbuf *m0 = m;
1457 	size_t valsize = 0;
1458 
1459 	if (sopt->sopt_val == NULL)
1460 		return 0;
1461 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1462 		if (sopt->sopt_td != NULL) {
1463 			int error;
1464 
1465 			error = copyout(mtod(m, char *), sopt->sopt_val,
1466 				       m->m_len);
1467 			if (error != 0) {
1468 				m_freem(m0);
1469 				return(error);
1470 			}
1471 		} else
1472 			bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1473 	       sopt->sopt_valsize -= m->m_len;
1474 	       (caddr_t)sopt->sopt_val += m->m_len;
1475 	       valsize += m->m_len;
1476 	       m = m->m_next;
1477 	}
1478 	if (m != NULL) {
1479 		/* enough soopt buffer should be given from user-land */
1480 		m_freem(m0);
1481 		return(EINVAL);
1482 	}
1483 	sopt->sopt_valsize = valsize;
1484 	return 0;
1485 }
1486 
1487 void
1488 sohasoutofband(so)
1489 	struct socket *so;
1490 {
1491 	if (so->so_sigio != NULL)
1492 		pgsigio(so->so_sigio, SIGURG, 0);
1493 	selwakeup(&so->so_rcv.sb_sel);
1494 }
1495 
1496 int
1497 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td)
1498 {
1499 	int revents = 0;
1500 	int s = splnet();
1501 
1502 	if (events & (POLLIN | POLLRDNORM))
1503 		if (soreadable(so))
1504 			revents |= events & (POLLIN | POLLRDNORM);
1505 
1506 	if (events & POLLINIGNEOF)
1507 		if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat ||
1508 			!TAILQ_EMPTY(&so->so_comp) || so->so_error)
1509 			revents |= POLLINIGNEOF;
1510 
1511 	if (events & (POLLOUT | POLLWRNORM))
1512 		if (sowriteable(so))
1513 			revents |= events & (POLLOUT | POLLWRNORM);
1514 
1515 	if (events & (POLLPRI | POLLRDBAND))
1516 		if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1517 			revents |= events & (POLLPRI | POLLRDBAND);
1518 
1519 	if (revents == 0) {
1520 		if (events &
1521 			(POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM |
1522 			 POLLRDBAND)) {
1523 			selrecord(td, &so->so_rcv.sb_sel);
1524 			so->so_rcv.sb_flags |= SB_SEL;
1525 		}
1526 
1527 		if (events & (POLLOUT | POLLWRNORM)) {
1528 			selrecord(td, &so->so_snd.sb_sel);
1529 			so->so_snd.sb_flags |= SB_SEL;
1530 		}
1531 	}
1532 
1533 	splx(s);
1534 	return (revents);
1535 }
1536 
1537 int
1538 sokqfilter(struct file *fp, struct knote *kn)
1539 {
1540 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1541 	struct sockbuf *sb;
1542 	int s;
1543 
1544 	switch (kn->kn_filter) {
1545 	case EVFILT_READ:
1546 		if (so->so_options & SO_ACCEPTCONN)
1547 			kn->kn_fop = &solisten_filtops;
1548 		else
1549 			kn->kn_fop = &soread_filtops;
1550 		sb = &so->so_rcv;
1551 		break;
1552 	case EVFILT_WRITE:
1553 		kn->kn_fop = &sowrite_filtops;
1554 		sb = &so->so_snd;
1555 		break;
1556 	default:
1557 		return (1);
1558 	}
1559 
1560 	s = splnet();
1561 	SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1562 	sb->sb_flags |= SB_KNOTE;
1563 	splx(s);
1564 	return (0);
1565 }
1566 
1567 static void
1568 filt_sordetach(struct knote *kn)
1569 {
1570 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1571 	int s = splnet();
1572 
1573 	SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1574 	if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1575 		so->so_rcv.sb_flags &= ~SB_KNOTE;
1576 	splx(s);
1577 }
1578 
1579 /*ARGSUSED*/
1580 static int
1581 filt_soread(struct knote *kn, long hint)
1582 {
1583 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1584 
1585 	kn->kn_data = so->so_rcv.sb_cc;
1586 	if (so->so_state & SS_CANTRCVMORE) {
1587 		kn->kn_flags |= EV_EOF;
1588 		kn->kn_fflags = so->so_error;
1589 		return (1);
1590 	}
1591 	if (so->so_error)	/* temporary udp error */
1592 		return (1);
1593 	if (kn->kn_sfflags & NOTE_LOWAT)
1594 		return (kn->kn_data >= kn->kn_sdata);
1595 	return (kn->kn_data >= so->so_rcv.sb_lowat);
1596 }
1597 
1598 static void
1599 filt_sowdetach(struct knote *kn)
1600 {
1601 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1602 	int s = splnet();
1603 
1604 	SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1605 	if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1606 		so->so_snd.sb_flags &= ~SB_KNOTE;
1607 	splx(s);
1608 }
1609 
1610 /*ARGSUSED*/
1611 static int
1612 filt_sowrite(struct knote *kn, long hint)
1613 {
1614 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1615 
1616 	kn->kn_data = sbspace(&so->so_snd);
1617 	if (so->so_state & SS_CANTSENDMORE) {
1618 		kn->kn_flags |= EV_EOF;
1619 		kn->kn_fflags = so->so_error;
1620 		return (1);
1621 	}
1622 	if (so->so_error)	/* temporary udp error */
1623 		return (1);
1624 	if (((so->so_state & SS_ISCONNECTED) == 0) &&
1625 	    (so->so_proto->pr_flags & PR_CONNREQUIRED))
1626 		return (0);
1627 	if (kn->kn_sfflags & NOTE_LOWAT)
1628 		return (kn->kn_data >= kn->kn_sdata);
1629 	return (kn->kn_data >= so->so_snd.sb_lowat);
1630 }
1631 
1632 /*ARGSUSED*/
1633 static int
1634 filt_solisten(struct knote *kn, long hint)
1635 {
1636 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1637 
1638 	kn->kn_data = so->so_qlen;
1639 	return (! TAILQ_EMPTY(&so->so_comp));
1640 }
1641