xref: /freebsd/sys/kern/uipc_socket.c (revision c17d43407fe04133a94055b0dbc7ea8965654a9f)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)uipc_socket.c	8.3 (Berkeley) 4/15/94
34  * $FreeBSD$
35  */
36 
37 #include "opt_inet.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/fcntl.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/mutex.h>
46 #include <sys/domain.h>
47 #include <sys/file.h>			/* for struct knote */
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
50 #include <sys/event.h>
51 #include <sys/poll.h>
52 #include <sys/proc.h>
53 #include <sys/protosw.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/resourcevar.h>
57 #include <sys/signalvar.h>
58 #include <sys/sysctl.h>
59 #include <sys/uio.h>
60 #include <sys/jail.h>
61 
62 #include <vm/uma.h>
63 
64 #include <machine/limits.h>
65 
66 #ifdef INET
67 static int	 do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
68 #endif
69 
70 static void 	filt_sordetach(struct knote *kn);
71 static int 	filt_soread(struct knote *kn, long hint);
72 static void 	filt_sowdetach(struct knote *kn);
73 static int	filt_sowrite(struct knote *kn, long hint);
74 static int	filt_solisten(struct knote *kn, long hint);
75 
76 static struct filterops solisten_filtops =
77 	{ 1, NULL, filt_sordetach, filt_solisten };
78 static struct filterops soread_filtops =
79 	{ 1, NULL, filt_sordetach, filt_soread };
80 static struct filterops sowrite_filtops =
81 	{ 1, NULL, filt_sowdetach, filt_sowrite };
82 
83 uma_zone_t socket_zone;
84 so_gen_t	so_gencnt;	/* generation count for sockets */
85 
86 MALLOC_DEFINE(M_SONAME, "soname", "socket name");
87 MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");
88 
89 SYSCTL_DECL(_kern_ipc);
90 
91 static int somaxconn = SOMAXCONN;
92 SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW,
93     &somaxconn, 0, "Maximum pending socket connection queue size");
94 static int numopensockets;
95 SYSCTL_INT(_kern_ipc, OID_AUTO, numopensockets, CTLFLAG_RD,
96     &numopensockets, 0, "Number of open sockets");
97 
98 
99 /*
100  * Socket operation routines.
101  * These routines are called by the routines in
102  * sys_socket.c or from a system process, and
103  * implement the semantics of socket operations by
104  * switching out to the protocol specific routines.
105  */
106 
107 /*
108  * Get a socket structure from our zone, and initialize it.
109  * Note that it would probably be better to allocate socket
110  * and PCB at the same time, but I'm not convinced that all
111  * the protocols can be easily modified to do this.
112  *
113  * soalloc() returns a socket with a ref count of 0.
114  */
115 struct socket *
116 soalloc(waitok)
117 	int waitok;
118 {
119 	struct socket *so;
120 	int flag;
121 
122 	if (waitok == 1)
123 		flag = M_WAITOK;
124 	else
125 		flag = M_NOWAIT;
126 
127 	so = uma_zalloc(socket_zone, flag);
128 	if (so) {
129 		/* XXX race condition for reentrant kernel */
130 		bzero(so, sizeof *so);
131 		so->so_gencnt = ++so_gencnt;
132 		so->so_zone = socket_zone;
133 		/* sx_init(&so->so_sxlock, "socket sxlock"); */
134 		TAILQ_INIT(&so->so_aiojobq);
135 		++numopensockets;
136 	}
137 	return so;
138 }
139 
140 /*
141  * socreate returns a socket with a ref count of 1.  The socket should be
142  * closed with soclose().
143  */
144 int
145 socreate(dom, aso, type, proto, cred, td)
146 	int dom;
147 	struct socket **aso;
148 	register int type;
149 	int proto;
150 	struct ucred *cred;
151 	struct thread *td;
152 {
153 	register struct protosw *prp;
154 	register struct socket *so;
155 	register int error;
156 
157 	if (proto)
158 		prp = pffindproto(dom, proto, type);
159 	else
160 		prp = pffindtype(dom, type);
161 
162 	if (prp == 0 || prp->pr_usrreqs->pru_attach == 0)
163 		return (EPROTONOSUPPORT);
164 
165 	if (jailed(td->td_ucred) && jail_socket_unixiproute_only &&
166 	    prp->pr_domain->dom_family != PF_LOCAL &&
167 	    prp->pr_domain->dom_family != PF_INET &&
168 	    prp->pr_domain->dom_family != PF_ROUTE) {
169 		return (EPROTONOSUPPORT);
170 	}
171 
172 	if (prp->pr_type != type)
173 		return (EPROTOTYPE);
174 	so = soalloc(td != 0);
175 	if (so == 0)
176 		return (ENOBUFS);
177 
178 	TAILQ_INIT(&so->so_incomp);
179 	TAILQ_INIT(&so->so_comp);
180 	so->so_type = type;
181 	so->so_cred = crhold(cred);
182 	so->so_proto = prp;
183 	soref(so);
184 	error = (*prp->pr_usrreqs->pru_attach)(so, proto, td);
185 	if (error) {
186 		so->so_state |= SS_NOFDREF;
187 		sorele(so);
188 		return (error);
189 	}
190 	*aso = so;
191 	return (0);
192 }
193 
194 int
195 sobind(so, nam, td)
196 	struct socket *so;
197 	struct sockaddr *nam;
198 	struct thread *td;
199 {
200 	int s = splnet();
201 	int error;
202 
203 	error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, td);
204 	splx(s);
205 	return (error);
206 }
207 
208 static void
209 sodealloc(struct socket *so)
210 {
211 
212 	KASSERT(so->so_count == 0, ("sodealloc(): so_count %d", so->so_count));
213 	so->so_gencnt = ++so_gencnt;
214 	if (so->so_rcv.sb_hiwat)
215 		(void)chgsbsize(so->so_cred->cr_uidinfo,
216 		    &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
217 	if (so->so_snd.sb_hiwat)
218 		(void)chgsbsize(so->so_cred->cr_uidinfo,
219 		    &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
220 #ifdef INET
221 	if (so->so_accf != NULL) {
222 		if (so->so_accf->so_accept_filter != NULL &&
223 			so->so_accf->so_accept_filter->accf_destroy != NULL) {
224 			so->so_accf->so_accept_filter->accf_destroy(so);
225 		}
226 		if (so->so_accf->so_accept_filter_str != NULL)
227 			FREE(so->so_accf->so_accept_filter_str, M_ACCF);
228 		FREE(so->so_accf, M_ACCF);
229 	}
230 #endif
231 	crfree(so->so_cred);
232 	/* sx_destroy(&so->so_sxlock); */
233 	uma_zfree(so->so_zone, so);
234 	--numopensockets;
235 }
236 
237 int
238 solisten(so, backlog, td)
239 	register struct socket *so;
240 	int backlog;
241 	struct thread *td;
242 {
243 	int s, error;
244 
245 	s = splnet();
246 	error = (*so->so_proto->pr_usrreqs->pru_listen)(so, td);
247 	if (error) {
248 		splx(s);
249 		return (error);
250 	}
251 	if (TAILQ_EMPTY(&so->so_comp))
252 		so->so_options |= SO_ACCEPTCONN;
253 	if (backlog < 0 || backlog > somaxconn)
254 		backlog = somaxconn;
255 	so->so_qlimit = backlog;
256 	splx(s);
257 	return (0);
258 }
259 
260 void
261 sofree(so)
262 	register struct socket *so;
263 {
264 	struct socket *head = so->so_head;
265 
266 	KASSERT(so->so_count == 0, ("socket %p so_count not 0", so));
267 
268 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
269 		return;
270 	if (head != NULL) {
271 		if (so->so_state & SS_INCOMP) {
272 			TAILQ_REMOVE(&head->so_incomp, so, so_list);
273 			head->so_incqlen--;
274 		} else if (so->so_state & SS_COMP) {
275 			/*
276 			 * We must not decommission a socket that's
277 			 * on the accept(2) queue.  If we do, then
278 			 * accept(2) may hang after select(2) indicated
279 			 * that the listening socket was ready.
280 			 */
281 			return;
282 		} else {
283 			panic("sofree: not queued");
284 		}
285 		head->so_qlen--;
286 		so->so_state &= ~SS_INCOMP;
287 		so->so_head = NULL;
288 	}
289 	sbrelease(&so->so_snd, so);
290 	sorflush(so);
291 	sodealloc(so);
292 }
293 
294 /*
295  * Close a socket on last file table reference removal.
296  * Initiate disconnect if connected.
297  * Free socket when disconnect complete.
298  *
299  * This function will sorele() the socket.  Note that soclose() may be
300  * called prior to the ref count reaching zero.  The actual socket
301  * structure will not be freed until the ref count reaches zero.
302  */
303 int
304 soclose(so)
305 	register struct socket *so;
306 {
307 	int s = splnet();		/* conservative */
308 	int error = 0;
309 
310 	funsetown(so->so_sigio);
311 	if (so->so_options & SO_ACCEPTCONN) {
312 		struct socket *sp, *sonext;
313 
314 		sp = TAILQ_FIRST(&so->so_incomp);
315 		for (; sp != NULL; sp = sonext) {
316 			sonext = TAILQ_NEXT(sp, so_list);
317 			(void) soabort(sp);
318 		}
319 		for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
320 			sonext = TAILQ_NEXT(sp, so_list);
321 			/* Dequeue from so_comp since sofree() won't do it */
322 			TAILQ_REMOVE(&so->so_comp, sp, so_list);
323 			so->so_qlen--;
324 			sp->so_state &= ~SS_COMP;
325 			sp->so_head = NULL;
326 			(void) soabort(sp);
327 		}
328 	}
329 	if (so->so_pcb == 0)
330 		goto discard;
331 	if (so->so_state & SS_ISCONNECTED) {
332 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
333 			error = sodisconnect(so);
334 			if (error)
335 				goto drop;
336 		}
337 		if (so->so_options & SO_LINGER) {
338 			if ((so->so_state & SS_ISDISCONNECTING) &&
339 			    (so->so_state & SS_NBIO))
340 				goto drop;
341 			while (so->so_state & SS_ISCONNECTED) {
342 				error = tsleep((caddr_t)&so->so_timeo,
343 				    PSOCK | PCATCH, "soclos", so->so_linger * hz);
344 				if (error)
345 					break;
346 			}
347 		}
348 	}
349 drop:
350 	if (so->so_pcb) {
351 		int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so);
352 		if (error == 0)
353 			error = error2;
354 	}
355 discard:
356 	if (so->so_state & SS_NOFDREF)
357 		panic("soclose: NOFDREF");
358 	so->so_state |= SS_NOFDREF;
359 	sorele(so);
360 	splx(s);
361 	return (error);
362 }
363 
364 /*
365  * Must be called at splnet...
366  */
367 int
368 soabort(so)
369 	struct socket *so;
370 {
371 	int error;
372 
373 	error = (*so->so_proto->pr_usrreqs->pru_abort)(so);
374 	if (error) {
375 		sotryfree(so);	/* note: does not decrement the ref count */
376 		return error;
377 	}
378 	return (0);
379 }
380 
381 int
382 soaccept(so, nam)
383 	register struct socket *so;
384 	struct sockaddr **nam;
385 {
386 	int s = splnet();
387 	int error;
388 
389 	if ((so->so_state & SS_NOFDREF) == 0)
390 		panic("soaccept: !NOFDREF");
391 	so->so_state &= ~SS_NOFDREF;
392 	error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam);
393 	splx(s);
394 	return (error);
395 }
396 
397 int
398 soconnect(so, nam, td)
399 	register struct socket *so;
400 	struct sockaddr *nam;
401 	struct thread *td;
402 {
403 	int s;
404 	int error;
405 
406 	if (so->so_options & SO_ACCEPTCONN)
407 		return (EOPNOTSUPP);
408 	s = splnet();
409 	/*
410 	 * If protocol is connection-based, can only connect once.
411 	 * Otherwise, if connected, try to disconnect first.
412 	 * This allows user to disconnect by connecting to, e.g.,
413 	 * a null address.
414 	 */
415 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
416 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
417 	    (error = sodisconnect(so))))
418 		error = EISCONN;
419 	else
420 		error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, td);
421 	splx(s);
422 	return (error);
423 }
424 
425 int
426 soconnect2(so1, so2)
427 	register struct socket *so1;
428 	struct socket *so2;
429 {
430 	int s = splnet();
431 	int error;
432 
433 	error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2);
434 	splx(s);
435 	return (error);
436 }
437 
438 int
439 sodisconnect(so)
440 	register struct socket *so;
441 {
442 	int s = splnet();
443 	int error;
444 
445 	if ((so->so_state & SS_ISCONNECTED) == 0) {
446 		error = ENOTCONN;
447 		goto bad;
448 	}
449 	if (so->so_state & SS_ISDISCONNECTING) {
450 		error = EALREADY;
451 		goto bad;
452 	}
453 	error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so);
454 bad:
455 	splx(s);
456 	return (error);
457 }
458 
459 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
460 /*
461  * Send on a socket.
462  * If send must go all at once and message is larger than
463  * send buffering, then hard error.
464  * Lock against other senders.
465  * If must go all at once and not enough room now, then
466  * inform user that this would block and do nothing.
467  * Otherwise, if nonblocking, send as much as possible.
468  * The data to be sent is described by "uio" if nonzero,
469  * otherwise by the mbuf chain "top" (which must be null
470  * if uio is not).  Data provided in mbuf chain must be small
471  * enough to send all at once.
472  *
473  * Returns nonzero on error, timeout or signal; callers
474  * must check for short counts if EINTR/ERESTART are returned.
475  * Data and control buffers are freed on return.
476  */
477 int
478 sosend(so, addr, uio, top, control, flags, td)
479 	register struct socket *so;
480 	struct sockaddr *addr;
481 	struct uio *uio;
482 	struct mbuf *top;
483 	struct mbuf *control;
484 	int flags;
485 	struct thread *td;
486 {
487 	struct mbuf **mp;
488 	register struct mbuf *m;
489 	register long space, len, resid;
490 	int clen = 0, error, s, dontroute, mlen;
491 	int atomic = sosendallatonce(so) || top;
492 
493 	if (uio)
494 		resid = uio->uio_resid;
495 	else
496 		resid = top->m_pkthdr.len;
497 	/*
498 	 * In theory resid should be unsigned.
499 	 * However, space must be signed, as it might be less than 0
500 	 * if we over-committed, and we must use a signed comparison
501 	 * of space and resid.  On the other hand, a negative resid
502 	 * causes us to loop sending 0-length segments to the protocol.
503 	 *
504 	 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
505 	 * type sockets since that's an error.
506 	 */
507 	if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
508 		error = EINVAL;
509 		goto out;
510 	}
511 
512 	dontroute =
513 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
514 	    (so->so_proto->pr_flags & PR_ATOMIC);
515 	if (td)
516 		td->td_proc->p_stats->p_ru.ru_msgsnd++;
517 	if (control)
518 		clen = control->m_len;
519 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
520 
521 restart:
522 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
523 	if (error)
524 		goto out;
525 	do {
526 		s = splnet();
527 		if (so->so_state & SS_CANTSENDMORE)
528 			snderr(EPIPE);
529 		if (so->so_error) {
530 			error = so->so_error;
531 			so->so_error = 0;
532 			splx(s);
533 			goto release;
534 		}
535 		if ((so->so_state & SS_ISCONNECTED) == 0) {
536 			/*
537 			 * `sendto' and `sendmsg' is allowed on a connection-
538 			 * based socket if it supports implied connect.
539 			 * Return ENOTCONN if not connected and no address is
540 			 * supplied.
541 			 */
542 			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
543 			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
544 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
545 				    !(resid == 0 && clen != 0))
546 					snderr(ENOTCONN);
547 			} else if (addr == 0)
548 			    snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
549 				   ENOTCONN : EDESTADDRREQ);
550 		}
551 		space = sbspace(&so->so_snd);
552 		if (flags & MSG_OOB)
553 			space += 1024;
554 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
555 		    clen > so->so_snd.sb_hiwat)
556 			snderr(EMSGSIZE);
557 		if (space < resid + clen &&
558 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
559 			if (so->so_state & SS_NBIO)
560 				snderr(EWOULDBLOCK);
561 			sbunlock(&so->so_snd);
562 			error = sbwait(&so->so_snd);
563 			splx(s);
564 			if (error)
565 				goto out;
566 			goto restart;
567 		}
568 		splx(s);
569 		mp = &top;
570 		space -= clen;
571 		do {
572 		    if (uio == NULL) {
573 			/*
574 			 * Data is prepackaged in "top".
575 			 */
576 			resid = 0;
577 			if (flags & MSG_EOR)
578 				top->m_flags |= M_EOR;
579 		    } else do {
580 			if (top == 0) {
581 				MGETHDR(m, M_TRYWAIT, MT_DATA);
582 				if (m == NULL) {
583 					error = ENOBUFS;
584 					goto release;
585 				}
586 				mlen = MHLEN;
587 				m->m_pkthdr.len = 0;
588 				m->m_pkthdr.rcvif = (struct ifnet *)0;
589 			} else {
590 				MGET(m, M_TRYWAIT, MT_DATA);
591 				if (m == NULL) {
592 					error = ENOBUFS;
593 					goto release;
594 				}
595 				mlen = MLEN;
596 			}
597 			if (resid >= MINCLSIZE) {
598 				MCLGET(m, M_TRYWAIT);
599 				if ((m->m_flags & M_EXT) == 0)
600 					goto nopages;
601 				mlen = MCLBYTES;
602 				len = min(min(mlen, resid), space);
603 			} else {
604 nopages:
605 				len = min(min(mlen, resid), space);
606 				/*
607 				 * For datagram protocols, leave room
608 				 * for protocol headers in first mbuf.
609 				 */
610 				if (atomic && top == 0 && len < mlen)
611 					MH_ALIGN(m, len);
612 			}
613 			space -= len;
614 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
615 			resid = uio->uio_resid;
616 			m->m_len = len;
617 			*mp = m;
618 			top->m_pkthdr.len += len;
619 			if (error)
620 				goto release;
621 			mp = &m->m_next;
622 			if (resid <= 0) {
623 				if (flags & MSG_EOR)
624 					top->m_flags |= M_EOR;
625 				break;
626 			}
627 		    } while (space > 0 && atomic);
628 		    if (dontroute)
629 			    so->so_options |= SO_DONTROUTE;
630 		    s = splnet();				/* XXX */
631 		    /*
632 		     * XXX all the SS_CANTSENDMORE checks previously
633 		     * done could be out of date.  We could have recieved
634 		     * a reset packet in an interrupt or maybe we slept
635 		     * while doing page faults in uiomove() etc. We could
636 		     * probably recheck again inside the splnet() protection
637 		     * here, but there are probably other places that this
638 		     * also happens.  We must rethink this.
639 		     */
640 		    error = (*so->so_proto->pr_usrreqs->pru_send)(so,
641 			(flags & MSG_OOB) ? PRUS_OOB :
642 			/*
643 			 * If the user set MSG_EOF, the protocol
644 			 * understands this flag and nothing left to
645 			 * send then use PRU_SEND_EOF instead of PRU_SEND.
646 			 */
647 			((flags & MSG_EOF) &&
648 			 (so->so_proto->pr_flags & PR_IMPLOPCL) &&
649 			 (resid <= 0)) ?
650 				PRUS_EOF :
651 			/* If there is more to send set PRUS_MORETOCOME */
652 			(resid > 0 && space > 0) ? PRUS_MORETOCOME : 0,
653 			top, addr, control, td);
654 		    splx(s);
655 		    if (dontroute)
656 			    so->so_options &= ~SO_DONTROUTE;
657 		    clen = 0;
658 		    control = 0;
659 		    top = 0;
660 		    mp = &top;
661 		    if (error)
662 			goto release;
663 		} while (resid && space > 0);
664 	} while (resid);
665 
666 release:
667 	sbunlock(&so->so_snd);
668 out:
669 	if (top)
670 		m_freem(top);
671 	if (control)
672 		m_freem(control);
673 	return (error);
674 }
675 
676 /*
677  * Implement receive operations on a socket.
678  * We depend on the way that records are added to the sockbuf
679  * by sbappend*.  In particular, each record (mbufs linked through m_next)
680  * must begin with an address if the protocol so specifies,
681  * followed by an optional mbuf or mbufs containing ancillary data,
682  * and then zero or more mbufs of data.
683  * In order to avoid blocking network interrupts for the entire time here,
684  * we splx() while doing the actual copy to user space.
685  * Although the sockbuf is locked, new data may still be appended,
686  * and thus we must maintain consistency of the sockbuf during that time.
687  *
688  * The caller may receive the data as a single mbuf chain by supplying
689  * an mbuf **mp0 for use in returning the chain.  The uio is then used
690  * only for the count in uio_resid.
691  */
692 int
693 soreceive(so, psa, uio, mp0, controlp, flagsp)
694 	register struct socket *so;
695 	struct sockaddr **psa;
696 	struct uio *uio;
697 	struct mbuf **mp0;
698 	struct mbuf **controlp;
699 	int *flagsp;
700 {
701 	struct mbuf *m, **mp;
702 	register int flags, len, error, s, offset;
703 	struct protosw *pr = so->so_proto;
704 	struct mbuf *nextrecord;
705 	int moff, type = 0;
706 	int orig_resid = uio->uio_resid;
707 
708 	mp = mp0;
709 	if (psa)
710 		*psa = 0;
711 	if (controlp)
712 		*controlp = 0;
713 	if (flagsp)
714 		flags = *flagsp &~ MSG_EOR;
715 	else
716 		flags = 0;
717 	if (flags & MSG_OOB) {
718 		m = m_get(M_TRYWAIT, MT_DATA);
719 		if (m == NULL)
720 			return (ENOBUFS);
721 		error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK);
722 		if (error)
723 			goto bad;
724 		do {
725 			error = uiomove(mtod(m, caddr_t),
726 			    (int) min(uio->uio_resid, m->m_len), uio);
727 			m = m_free(m);
728 		} while (uio->uio_resid && error == 0 && m);
729 bad:
730 		if (m)
731 			m_freem(m);
732 		return (error);
733 	}
734 	if (mp)
735 		*mp = (struct mbuf *)0;
736 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
737 		(*pr->pr_usrreqs->pru_rcvd)(so, 0);
738 
739 restart:
740 	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
741 	if (error)
742 		return (error);
743 	s = splnet();
744 
745 	m = so->so_rcv.sb_mb;
746 	/*
747 	 * If we have less data than requested, block awaiting more
748 	 * (subject to any timeout) if:
749 	 *   1. the current count is less than the low water mark, or
750 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
751 	 *	receive operation at once if we block (resid <= hiwat).
752 	 *   3. MSG_DONTWAIT is not set
753 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
754 	 * we have to do the receive in sections, and thus risk returning
755 	 * a short count if a timeout or signal occurs after we start.
756 	 */
757 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
758 	    so->so_rcv.sb_cc < uio->uio_resid) &&
759 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
760 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
761 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
762 		KASSERT(m != 0 || !so->so_rcv.sb_cc,
763 		    ("receive: m == %p so->so_rcv.sb_cc == %lu",
764 		    m, so->so_rcv.sb_cc));
765 		if (so->so_error) {
766 			if (m)
767 				goto dontblock;
768 			error = so->so_error;
769 			if ((flags & MSG_PEEK) == 0)
770 				so->so_error = 0;
771 			goto release;
772 		}
773 		if (so->so_state & SS_CANTRCVMORE) {
774 			if (m)
775 				goto dontblock;
776 			else
777 				goto release;
778 		}
779 		for (; m; m = m->m_next)
780 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
781 				m = so->so_rcv.sb_mb;
782 				goto dontblock;
783 			}
784 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
785 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
786 			error = ENOTCONN;
787 			goto release;
788 		}
789 		if (uio->uio_resid == 0)
790 			goto release;
791 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
792 			error = EWOULDBLOCK;
793 			goto release;
794 		}
795 		sbunlock(&so->so_rcv);
796 		error = sbwait(&so->so_rcv);
797 		splx(s);
798 		if (error)
799 			return (error);
800 		goto restart;
801 	}
802 dontblock:
803 	if (uio->uio_td)
804 		uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++;
805 	nextrecord = m->m_nextpkt;
806 	if (pr->pr_flags & PR_ADDR) {
807 		KASSERT(m->m_type == MT_SONAME,
808 		    ("m->m_type == %d", m->m_type));
809 		orig_resid = 0;
810 		if (psa)
811 			*psa = dup_sockaddr(mtod(m, struct sockaddr *),
812 					    mp0 == 0);
813 		if (flags & MSG_PEEK) {
814 			m = m->m_next;
815 		} else {
816 			sbfree(&so->so_rcv, m);
817 			so->so_rcv.sb_mb = m_free(m);
818 			m = so->so_rcv.sb_mb;
819 		}
820 	}
821 	while (m && m->m_type == MT_CONTROL && error == 0) {
822 		if (flags & MSG_PEEK) {
823 			if (controlp)
824 				*controlp = m_copy(m, 0, m->m_len);
825 			m = m->m_next;
826 		} else {
827 			sbfree(&so->so_rcv, m);
828 			so->so_rcv.sb_mb = m->m_next;
829 			m->m_next = NULL;
830 			if (pr->pr_domain->dom_externalize)
831 				error =
832 				(*pr->pr_domain->dom_externalize)(m, controlp);
833 			else if (controlp)
834 				*controlp = m;
835 			else
836 				m_freem(m);
837 			m = so->so_rcv.sb_mb;
838 		}
839 		if (controlp) {
840 			orig_resid = 0;
841 			do
842 				controlp = &(*controlp)->m_next;
843 			while (*controlp != NULL);
844 		}
845 	}
846 	if (m) {
847 		if ((flags & MSG_PEEK) == 0)
848 			m->m_nextpkt = nextrecord;
849 		type = m->m_type;
850 		if (type == MT_OOBDATA)
851 			flags |= MSG_OOB;
852 	}
853 	moff = 0;
854 	offset = 0;
855 	while (m && uio->uio_resid > 0 && error == 0) {
856 		if (m->m_type == MT_OOBDATA) {
857 			if (type != MT_OOBDATA)
858 				break;
859 		} else if (type == MT_OOBDATA)
860 			break;
861 		else
862 		    KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
863 			("m->m_type == %d", m->m_type));
864 		so->so_state &= ~SS_RCVATMARK;
865 		len = uio->uio_resid;
866 		if (so->so_oobmark && len > so->so_oobmark - offset)
867 			len = so->so_oobmark - offset;
868 		if (len > m->m_len - moff)
869 			len = m->m_len - moff;
870 		/*
871 		 * If mp is set, just pass back the mbufs.
872 		 * Otherwise copy them out via the uio, then free.
873 		 * Sockbuf must be consistent here (points to current mbuf,
874 		 * it points to next record) when we drop priority;
875 		 * we must note any additions to the sockbuf when we
876 		 * block interrupts again.
877 		 */
878 		if (mp == 0) {
879 			splx(s);
880 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
881 			s = splnet();
882 			if (error)
883 				goto release;
884 		} else
885 			uio->uio_resid -= len;
886 		if (len == m->m_len - moff) {
887 			if (m->m_flags & M_EOR)
888 				flags |= MSG_EOR;
889 			if (flags & MSG_PEEK) {
890 				m = m->m_next;
891 				moff = 0;
892 			} else {
893 				nextrecord = m->m_nextpkt;
894 				sbfree(&so->so_rcv, m);
895 				if (mp) {
896 					*mp = m;
897 					mp = &m->m_next;
898 					so->so_rcv.sb_mb = m = m->m_next;
899 					*mp = (struct mbuf *)0;
900 				} else {
901 					so->so_rcv.sb_mb = m_free(m);
902 					m = so->so_rcv.sb_mb;
903 				}
904 				if (m)
905 					m->m_nextpkt = nextrecord;
906 			}
907 		} else {
908 			if (flags & MSG_PEEK)
909 				moff += len;
910 			else {
911 				if (mp)
912 					*mp = m_copym(m, 0, len, M_TRYWAIT);
913 				m->m_data += len;
914 				m->m_len -= len;
915 				so->so_rcv.sb_cc -= len;
916 			}
917 		}
918 		if (so->so_oobmark) {
919 			if ((flags & MSG_PEEK) == 0) {
920 				so->so_oobmark -= len;
921 				if (so->so_oobmark == 0) {
922 					so->so_state |= SS_RCVATMARK;
923 					break;
924 				}
925 			} else {
926 				offset += len;
927 				if (offset == so->so_oobmark)
928 					break;
929 			}
930 		}
931 		if (flags & MSG_EOR)
932 			break;
933 		/*
934 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
935 		 * we must not quit until "uio->uio_resid == 0" or an error
936 		 * termination.  If a signal/timeout occurs, return
937 		 * with a short count but without error.
938 		 * Keep sockbuf locked against other readers.
939 		 */
940 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
941 		    !sosendallatonce(so) && !nextrecord) {
942 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
943 				break;
944 			/*
945 			 * Notify the protocol that some data has been
946 			 * drained before blocking.
947 			 */
948 			if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
949 				(*pr->pr_usrreqs->pru_rcvd)(so, flags);
950 			error = sbwait(&so->so_rcv);
951 			if (error) {
952 				sbunlock(&so->so_rcv);
953 				splx(s);
954 				return (0);
955 			}
956 			m = so->so_rcv.sb_mb;
957 			if (m)
958 				nextrecord = m->m_nextpkt;
959 		}
960 	}
961 
962 	if (m && pr->pr_flags & PR_ATOMIC) {
963 		flags |= MSG_TRUNC;
964 		if ((flags & MSG_PEEK) == 0)
965 			(void) sbdroprecord(&so->so_rcv);
966 	}
967 	if ((flags & MSG_PEEK) == 0) {
968 		if (m == 0)
969 			so->so_rcv.sb_mb = nextrecord;
970 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
971 			(*pr->pr_usrreqs->pru_rcvd)(so, flags);
972 	}
973 	if (orig_resid == uio->uio_resid && orig_resid &&
974 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
975 		sbunlock(&so->so_rcv);
976 		splx(s);
977 		goto restart;
978 	}
979 
980 	if (flagsp)
981 		*flagsp |= flags;
982 release:
983 	sbunlock(&so->so_rcv);
984 	splx(s);
985 	return (error);
986 }
987 
988 int
989 soshutdown(so, how)
990 	register struct socket *so;
991 	register int how;
992 {
993 	register struct protosw *pr = so->so_proto;
994 
995 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
996 		return (EINVAL);
997 
998 	if (how != SHUT_WR)
999 		sorflush(so);
1000 	if (how != SHUT_RD)
1001 		return ((*pr->pr_usrreqs->pru_shutdown)(so));
1002 	return (0);
1003 }
1004 
1005 void
1006 sorflush(so)
1007 	register struct socket *so;
1008 {
1009 	register struct sockbuf *sb = &so->so_rcv;
1010 	register struct protosw *pr = so->so_proto;
1011 	register int s;
1012 	struct sockbuf asb;
1013 
1014 	sb->sb_flags |= SB_NOINTR;
1015 	(void) sblock(sb, M_WAITOK);
1016 	s = splimp();
1017 	socantrcvmore(so);
1018 	sbunlock(sb);
1019 	asb = *sb;
1020 	bzero((caddr_t)sb, sizeof (*sb));
1021 	splx(s);
1022 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1023 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
1024 	sbrelease(&asb, so);
1025 }
1026 
1027 #ifdef INET
1028 static int
1029 do_setopt_accept_filter(so, sopt)
1030 	struct	socket *so;
1031 	struct	sockopt *sopt;
1032 {
1033 	struct accept_filter_arg	*afap = NULL;
1034 	struct accept_filter	*afp;
1035 	struct so_accf	*af = so->so_accf;
1036 	int	error = 0;
1037 
1038 	/* do not set/remove accept filters on non listen sockets */
1039 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
1040 		error = EINVAL;
1041 		goto out;
1042 	}
1043 
1044 	/* removing the filter */
1045 	if (sopt == NULL) {
1046 		if (af != NULL) {
1047 			if (af->so_accept_filter != NULL &&
1048 				af->so_accept_filter->accf_destroy != NULL) {
1049 				af->so_accept_filter->accf_destroy(so);
1050 			}
1051 			if (af->so_accept_filter_str != NULL) {
1052 				FREE(af->so_accept_filter_str, M_ACCF);
1053 			}
1054 			FREE(af, M_ACCF);
1055 			so->so_accf = NULL;
1056 		}
1057 		so->so_options &= ~SO_ACCEPTFILTER;
1058 		return (0);
1059 	}
1060 	/* adding a filter */
1061 	/* must remove previous filter first */
1062 	if (af != NULL) {
1063 		error = EINVAL;
1064 		goto out;
1065 	}
1066 	/* don't put large objects on the kernel stack */
1067 	MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1068 	error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1069 	afap->af_name[sizeof(afap->af_name)-1] = '\0';
1070 	afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1071 	if (error)
1072 		goto out;
1073 	afp = accept_filt_get(afap->af_name);
1074 	if (afp == NULL) {
1075 		error = ENOENT;
1076 		goto out;
1077 	}
1078 	MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO);
1079 	if (afp->accf_create != NULL) {
1080 		if (afap->af_name[0] != '\0') {
1081 			int len = strlen(afap->af_name) + 1;
1082 
1083 			MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1084 			strcpy(af->so_accept_filter_str, afap->af_name);
1085 		}
1086 		af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1087 		if (af->so_accept_filter_arg == NULL) {
1088 			FREE(af->so_accept_filter_str, M_ACCF);
1089 			FREE(af, M_ACCF);
1090 			so->so_accf = NULL;
1091 			error = EINVAL;
1092 			goto out;
1093 		}
1094 	}
1095 	af->so_accept_filter = afp;
1096 	so->so_accf = af;
1097 	so->so_options |= SO_ACCEPTFILTER;
1098 out:
1099 	if (afap != NULL)
1100 		FREE(afap, M_TEMP);
1101 	return (error);
1102 }
1103 #endif /* INET */
1104 
1105 /*
1106  * Perhaps this routine, and sooptcopyout(), below, ought to come in
1107  * an additional variant to handle the case where the option value needs
1108  * to be some kind of integer, but not a specific size.
1109  * In addition to their use here, these functions are also called by the
1110  * protocol-level pr_ctloutput() routines.
1111  */
1112 int
1113 sooptcopyin(sopt, buf, len, minlen)
1114 	struct	sockopt *sopt;
1115 	void	*buf;
1116 	size_t	len;
1117 	size_t	minlen;
1118 {
1119 	size_t	valsize;
1120 
1121 	/*
1122 	 * If the user gives us more than we wanted, we ignore it,
1123 	 * but if we don't get the minimum length the caller
1124 	 * wants, we return EINVAL.  On success, sopt->sopt_valsize
1125 	 * is set to however much we actually retrieved.
1126 	 */
1127 	if ((valsize = sopt->sopt_valsize) < minlen)
1128 		return EINVAL;
1129 	if (valsize > len)
1130 		sopt->sopt_valsize = valsize = len;
1131 
1132 	if (sopt->sopt_td != 0)
1133 		return (copyin(sopt->sopt_val, buf, valsize));
1134 
1135 	bcopy(sopt->sopt_val, buf, valsize);
1136 	return 0;
1137 }
1138 
1139 int
1140 sosetopt(so, sopt)
1141 	struct socket *so;
1142 	struct sockopt *sopt;
1143 {
1144 	int	error, optval;
1145 	struct	linger l;
1146 	struct	timeval tv;
1147 	u_long  val;
1148 
1149 	error = 0;
1150 	if (sopt->sopt_level != SOL_SOCKET) {
1151 		if (so->so_proto && so->so_proto->pr_ctloutput)
1152 			return ((*so->so_proto->pr_ctloutput)
1153 				  (so, sopt));
1154 		error = ENOPROTOOPT;
1155 	} else {
1156 		switch (sopt->sopt_name) {
1157 #ifdef INET
1158 		case SO_ACCEPTFILTER:
1159 			error = do_setopt_accept_filter(so, sopt);
1160 			if (error)
1161 				goto bad;
1162 			break;
1163 #endif
1164 		case SO_LINGER:
1165 			error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1166 			if (error)
1167 				goto bad;
1168 
1169 			so->so_linger = l.l_linger;
1170 			if (l.l_onoff)
1171 				so->so_options |= SO_LINGER;
1172 			else
1173 				so->so_options &= ~SO_LINGER;
1174 			break;
1175 
1176 		case SO_DEBUG:
1177 		case SO_KEEPALIVE:
1178 		case SO_DONTROUTE:
1179 		case SO_USELOOPBACK:
1180 		case SO_BROADCAST:
1181 		case SO_REUSEADDR:
1182 		case SO_REUSEPORT:
1183 		case SO_OOBINLINE:
1184 		case SO_TIMESTAMP:
1185 			error = sooptcopyin(sopt, &optval, sizeof optval,
1186 					    sizeof optval);
1187 			if (error)
1188 				goto bad;
1189 			if (optval)
1190 				so->so_options |= sopt->sopt_name;
1191 			else
1192 				so->so_options &= ~sopt->sopt_name;
1193 			break;
1194 
1195 		case SO_SNDBUF:
1196 		case SO_RCVBUF:
1197 		case SO_SNDLOWAT:
1198 		case SO_RCVLOWAT:
1199 			error = sooptcopyin(sopt, &optval, sizeof optval,
1200 					    sizeof optval);
1201 			if (error)
1202 				goto bad;
1203 
1204 			/*
1205 			 * Values < 1 make no sense for any of these
1206 			 * options, so disallow them.
1207 			 */
1208 			if (optval < 1) {
1209 				error = EINVAL;
1210 				goto bad;
1211 			}
1212 
1213 			switch (sopt->sopt_name) {
1214 			case SO_SNDBUF:
1215 			case SO_RCVBUF:
1216 				if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1217 				    &so->so_snd : &so->so_rcv, (u_long)optval,
1218 				    so, curthread) == 0) {
1219 					error = ENOBUFS;
1220 					goto bad;
1221 				}
1222 				break;
1223 
1224 			/*
1225 			 * Make sure the low-water is never greater than
1226 			 * the high-water.
1227 			 */
1228 			case SO_SNDLOWAT:
1229 				so->so_snd.sb_lowat =
1230 				    (optval > so->so_snd.sb_hiwat) ?
1231 				    so->so_snd.sb_hiwat : optval;
1232 				break;
1233 			case SO_RCVLOWAT:
1234 				so->so_rcv.sb_lowat =
1235 				    (optval > so->so_rcv.sb_hiwat) ?
1236 				    so->so_rcv.sb_hiwat : optval;
1237 				break;
1238 			}
1239 			break;
1240 
1241 		case SO_SNDTIMEO:
1242 		case SO_RCVTIMEO:
1243 			error = sooptcopyin(sopt, &tv, sizeof tv,
1244 					    sizeof tv);
1245 			if (error)
1246 				goto bad;
1247 
1248 			/* assert(hz > 0); */
1249 			if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1250 			    tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1251 				error = EDOM;
1252 				goto bad;
1253 			}
1254 			/* assert(tick > 0); */
1255 			/* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1256 			val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1257 			if (val > SHRT_MAX) {
1258 				error = EDOM;
1259 				goto bad;
1260 			}
1261 
1262 			switch (sopt->sopt_name) {
1263 			case SO_SNDTIMEO:
1264 				so->so_snd.sb_timeo = val;
1265 				break;
1266 			case SO_RCVTIMEO:
1267 				so->so_rcv.sb_timeo = val;
1268 				break;
1269 			}
1270 			break;
1271 		default:
1272 			error = ENOPROTOOPT;
1273 			break;
1274 		}
1275 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1276 			(void) ((*so->so_proto->pr_ctloutput)
1277 				  (so, sopt));
1278 		}
1279 	}
1280 bad:
1281 	return (error);
1282 }
1283 
1284 /* Helper routine for getsockopt */
1285 int
1286 sooptcopyout(sopt, buf, len)
1287 	struct	sockopt *sopt;
1288 	void	*buf;
1289 	size_t	len;
1290 {
1291 	int	error;
1292 	size_t	valsize;
1293 
1294 	error = 0;
1295 
1296 	/*
1297 	 * Documented get behavior is that we always return a value,
1298 	 * possibly truncated to fit in the user's buffer.
1299 	 * Traditional behavior is that we always tell the user
1300 	 * precisely how much we copied, rather than something useful
1301 	 * like the total amount we had available for her.
1302 	 * Note that this interface is not idempotent; the entire answer must
1303 	 * generated ahead of time.
1304 	 */
1305 	valsize = min(len, sopt->sopt_valsize);
1306 	sopt->sopt_valsize = valsize;
1307 	if (sopt->sopt_val != 0) {
1308 		if (sopt->sopt_td != 0)
1309 			error = copyout(buf, sopt->sopt_val, valsize);
1310 		else
1311 			bcopy(buf, sopt->sopt_val, valsize);
1312 	}
1313 	return error;
1314 }
1315 
1316 int
1317 sogetopt(so, sopt)
1318 	struct socket *so;
1319 	struct sockopt *sopt;
1320 {
1321 	int	error, optval;
1322 	struct	linger l;
1323 	struct	timeval tv;
1324 #ifdef INET
1325 	struct accept_filter_arg *afap;
1326 #endif
1327 
1328 	error = 0;
1329 	if (sopt->sopt_level != SOL_SOCKET) {
1330 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1331 			return ((*so->so_proto->pr_ctloutput)
1332 				  (so, sopt));
1333 		} else
1334 			return (ENOPROTOOPT);
1335 	} else {
1336 		switch (sopt->sopt_name) {
1337 #ifdef INET
1338 		case SO_ACCEPTFILTER:
1339 			if ((so->so_options & SO_ACCEPTCONN) == 0)
1340 				return (EINVAL);
1341 			MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1342 				M_TEMP, M_WAITOK | M_ZERO);
1343 			if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1344 				strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1345 				if (so->so_accf->so_accept_filter_str != NULL)
1346 					strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1347 			}
1348 			error = sooptcopyout(sopt, afap, sizeof(*afap));
1349 			FREE(afap, M_TEMP);
1350 			break;
1351 #endif
1352 
1353 		case SO_LINGER:
1354 			l.l_onoff = so->so_options & SO_LINGER;
1355 			l.l_linger = so->so_linger;
1356 			error = sooptcopyout(sopt, &l, sizeof l);
1357 			break;
1358 
1359 		case SO_USELOOPBACK:
1360 		case SO_DONTROUTE:
1361 		case SO_DEBUG:
1362 		case SO_KEEPALIVE:
1363 		case SO_REUSEADDR:
1364 		case SO_REUSEPORT:
1365 		case SO_BROADCAST:
1366 		case SO_OOBINLINE:
1367 		case SO_TIMESTAMP:
1368 			optval = so->so_options & sopt->sopt_name;
1369 integer:
1370 			error = sooptcopyout(sopt, &optval, sizeof optval);
1371 			break;
1372 
1373 		case SO_TYPE:
1374 			optval = so->so_type;
1375 			goto integer;
1376 
1377 		case SO_ERROR:
1378 			optval = so->so_error;
1379 			so->so_error = 0;
1380 			goto integer;
1381 
1382 		case SO_SNDBUF:
1383 			optval = so->so_snd.sb_hiwat;
1384 			goto integer;
1385 
1386 		case SO_RCVBUF:
1387 			optval = so->so_rcv.sb_hiwat;
1388 			goto integer;
1389 
1390 		case SO_SNDLOWAT:
1391 			optval = so->so_snd.sb_lowat;
1392 			goto integer;
1393 
1394 		case SO_RCVLOWAT:
1395 			optval = so->so_rcv.sb_lowat;
1396 			goto integer;
1397 
1398 		case SO_SNDTIMEO:
1399 		case SO_RCVTIMEO:
1400 			optval = (sopt->sopt_name == SO_SNDTIMEO ?
1401 				  so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1402 
1403 			tv.tv_sec = optval / hz;
1404 			tv.tv_usec = (optval % hz) * tick;
1405 			error = sooptcopyout(sopt, &tv, sizeof tv);
1406 			break;
1407 
1408 		default:
1409 			error = ENOPROTOOPT;
1410 			break;
1411 		}
1412 		return (error);
1413 	}
1414 }
1415 
1416 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1417 int
1418 soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1419 {
1420 	struct mbuf *m, *m_prev;
1421 	int sopt_size = sopt->sopt_valsize;
1422 
1423 	MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1424 	if (m == 0)
1425 		return ENOBUFS;
1426 	if (sopt_size > MLEN) {
1427 		MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT);
1428 		if ((m->m_flags & M_EXT) == 0) {
1429 			m_free(m);
1430 			return ENOBUFS;
1431 		}
1432 		m->m_len = min(MCLBYTES, sopt_size);
1433 	} else {
1434 		m->m_len = min(MLEN, sopt_size);
1435 	}
1436 	sopt_size -= m->m_len;
1437 	*mp = m;
1438 	m_prev = m;
1439 
1440 	while (sopt_size) {
1441 		MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1442 		if (m == 0) {
1443 			m_freem(*mp);
1444 			return ENOBUFS;
1445 		}
1446 		if (sopt_size > MLEN) {
1447 			MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT);
1448 			if ((m->m_flags & M_EXT) == 0) {
1449 				m_freem(*mp);
1450 				return ENOBUFS;
1451 			}
1452 			m->m_len = min(MCLBYTES, sopt_size);
1453 		} else {
1454 			m->m_len = min(MLEN, sopt_size);
1455 		}
1456 		sopt_size -= m->m_len;
1457 		m_prev->m_next = m;
1458 		m_prev = m;
1459 	}
1460 	return 0;
1461 }
1462 
1463 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1464 int
1465 soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1466 {
1467 	struct mbuf *m0 = m;
1468 
1469 	if (sopt->sopt_val == NULL)
1470 		return 0;
1471 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1472 		if (sopt->sopt_td != NULL) {
1473 			int error;
1474 
1475 			error = copyin(sopt->sopt_val, mtod(m, char *),
1476 				       m->m_len);
1477 			if (error != 0) {
1478 				m_freem(m0);
1479 				return(error);
1480 			}
1481 		} else
1482 			bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1483 		sopt->sopt_valsize -= m->m_len;
1484 		(caddr_t)sopt->sopt_val += m->m_len;
1485 		m = m->m_next;
1486 	}
1487 	if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1488 		panic("ip6_sooptmcopyin");
1489 	return 0;
1490 }
1491 
1492 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1493 int
1494 soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1495 {
1496 	struct mbuf *m0 = m;
1497 	size_t valsize = 0;
1498 
1499 	if (sopt->sopt_val == NULL)
1500 		return 0;
1501 	while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1502 		if (sopt->sopt_td != NULL) {
1503 			int error;
1504 
1505 			error = copyout(mtod(m, char *), sopt->sopt_val,
1506 				       m->m_len);
1507 			if (error != 0) {
1508 				m_freem(m0);
1509 				return(error);
1510 			}
1511 		} else
1512 			bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1513 	       sopt->sopt_valsize -= m->m_len;
1514 	       (caddr_t)sopt->sopt_val += m->m_len;
1515 	       valsize += m->m_len;
1516 	       m = m->m_next;
1517 	}
1518 	if (m != NULL) {
1519 		/* enough soopt buffer should be given from user-land */
1520 		m_freem(m0);
1521 		return(EINVAL);
1522 	}
1523 	sopt->sopt_valsize = valsize;
1524 	return 0;
1525 }
1526 
1527 void
1528 sohasoutofband(so)
1529 	register struct socket *so;
1530 {
1531 	if (so->so_sigio != NULL)
1532 		pgsigio(so->so_sigio, SIGURG, 0);
1533 	selwakeup(&so->so_rcv.sb_sel);
1534 }
1535 
1536 int
1537 sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td)
1538 {
1539 	int revents = 0;
1540 	int s = splnet();
1541 
1542 	if (events & (POLLIN | POLLRDNORM))
1543 		if (soreadable(so))
1544 			revents |= events & (POLLIN | POLLRDNORM);
1545 
1546 	if (events & POLLINIGNEOF)
1547 		if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat ||
1548 		    !TAILQ_EMPTY(&so->so_comp) || so->so_error)
1549 			revents |= POLLINIGNEOF;
1550 
1551 	if (events & (POLLOUT | POLLWRNORM))
1552 		if (sowriteable(so))
1553 			revents |= events & (POLLOUT | POLLWRNORM);
1554 
1555 	if (events & (POLLPRI | POLLRDBAND))
1556 		if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1557 			revents |= events & (POLLPRI | POLLRDBAND);
1558 
1559 	if (revents == 0) {
1560 		if (events &
1561 		    (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM |
1562 		     POLLRDBAND)) {
1563 			selrecord(td, &so->so_rcv.sb_sel);
1564 			so->so_rcv.sb_flags |= SB_SEL;
1565 		}
1566 
1567 		if (events & (POLLOUT | POLLWRNORM)) {
1568 			selrecord(td, &so->so_snd.sb_sel);
1569 			so->so_snd.sb_flags |= SB_SEL;
1570 		}
1571 	}
1572 
1573 	splx(s);
1574 	return (revents);
1575 }
1576 
1577 int
1578 sokqfilter(struct file *fp, struct knote *kn)
1579 {
1580 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1581 	struct sockbuf *sb;
1582 	int s;
1583 
1584 	switch (kn->kn_filter) {
1585 	case EVFILT_READ:
1586 		if (so->so_options & SO_ACCEPTCONN)
1587 			kn->kn_fop = &solisten_filtops;
1588 		else
1589 			kn->kn_fop = &soread_filtops;
1590 		sb = &so->so_rcv;
1591 		break;
1592 	case EVFILT_WRITE:
1593 		kn->kn_fop = &sowrite_filtops;
1594 		sb = &so->so_snd;
1595 		break;
1596 	default:
1597 		return (1);
1598 	}
1599 
1600 	s = splnet();
1601 	SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1602 	sb->sb_flags |= SB_KNOTE;
1603 	splx(s);
1604 	return (0);
1605 }
1606 
1607 static void
1608 filt_sordetach(struct knote *kn)
1609 {
1610 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1611 	int s = splnet();
1612 
1613 	SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1614 	if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1615 		so->so_rcv.sb_flags &= ~SB_KNOTE;
1616 	splx(s);
1617 }
1618 
1619 /*ARGSUSED*/
1620 static int
1621 filt_soread(struct knote *kn, long hint)
1622 {
1623 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1624 
1625 	kn->kn_data = so->so_rcv.sb_cc;
1626 	if (so->so_state & SS_CANTRCVMORE) {
1627 		kn->kn_flags |= EV_EOF;
1628 		kn->kn_fflags = so->so_error;
1629 		return (1);
1630 	}
1631 	if (so->so_error)	/* temporary udp error */
1632 		return (1);
1633 	if (kn->kn_sfflags & NOTE_LOWAT)
1634 		return (kn->kn_data >= kn->kn_sdata);
1635 	return (kn->kn_data >= so->so_rcv.sb_lowat);
1636 }
1637 
1638 static void
1639 filt_sowdetach(struct knote *kn)
1640 {
1641 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1642 	int s = splnet();
1643 
1644 	SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1645 	if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1646 		so->so_snd.sb_flags &= ~SB_KNOTE;
1647 	splx(s);
1648 }
1649 
1650 /*ARGSUSED*/
1651 static int
1652 filt_sowrite(struct knote *kn, long hint)
1653 {
1654 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1655 
1656 	kn->kn_data = sbspace(&so->so_snd);
1657 	if (so->so_state & SS_CANTSENDMORE) {
1658 		kn->kn_flags |= EV_EOF;
1659 		kn->kn_fflags = so->so_error;
1660 		return (1);
1661 	}
1662 	if (so->so_error)	/* temporary udp error */
1663 		return (1);
1664 	if (((so->so_state & SS_ISCONNECTED) == 0) &&
1665 	    (so->so_proto->pr_flags & PR_CONNREQUIRED))
1666 		return (0);
1667 	if (kn->kn_sfflags & NOTE_LOWAT)
1668 		return (kn->kn_data >= kn->kn_sdata);
1669 	return (kn->kn_data >= so->so_snd.sb_lowat);
1670 }
1671 
1672 /*ARGSUSED*/
1673 static int
1674 filt_solisten(struct knote *kn, long hint)
1675 {
1676 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1677 
1678 	kn->kn_data = so->so_qlen - so->so_incqlen;
1679 	return (! TAILQ_EMPTY(&so->so_comp));
1680 }
1681 
1682 int
1683 socheckuid(struct socket *so, uid_t uid)
1684 {
1685 
1686 	if (so == NULL)
1687 		return (EPERM);
1688 	if (so->so_cred->cr_uid == uid)
1689 		return (0);
1690 	return (EPERM);
1691 }
1692