xref: /openbsd/sys/kern/uipc_socket.c (revision 78b63d65)
1 /*	$OpenBSD: uipc_socket.c,v 1.39 2001/11/28 17:18:00 ericj Exp $	*/
2 /*	$NetBSD: uipc_socket.c,v 1.21 1996/02/04 02:17:52 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1990, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)uipc_socket.c	8.3 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/proc.h>
42 #include <sys/file.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/domain.h>
46 #include <sys/kernel.h>
47 #include <sys/event.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/signalvar.h>
52 #include <sys/resourcevar.h>
53 #include <sys/pool.h>
54 
55 void 	filt_sordetach(struct knote *kn);
56 int 	filt_soread(struct knote *kn, long hint);
57 void 	filt_sowdetach(struct knote *kn);
58 int	filt_sowrite(struct knote *kn, long hint);
59 int	filt_solisten(struct knote *kn, long hint);
60 
61 struct filterops solisten_filtops =
62 	{ 1, NULL, filt_sordetach, filt_solisten };
63 struct filterops soread_filtops =
64 	{ 1, NULL, filt_sordetach, filt_soread };
65 struct filterops sowrite_filtops =
66 	{ 1, NULL, filt_sowdetach, filt_sowrite };
67 
68 
69 #ifndef SOMINCONN
70 #define SOMINCONN 80
71 #endif /* SOMINCONN */
72 
73 int	somaxconn = SOMAXCONN;
74 int	sominconn = SOMINCONN;
75 
76 struct pool socket_pool;
77 
78 void
79 soinit(void)
80 {
81 
82 	pool_init(&socket_pool, sizeof(struct socket), 0, 0, 0,
83 	    "sockpl", 0, NULL, NULL, M_SOCKET);
84 }
85 
86 /*
87  * Socket operation routines.
88  * These routines are called by the routines in
89  * sys_socket.c or from a system process, and
90  * implement the semantics of socket operations by
91  * switching out to the protocol specific routines.
92  */
93 /*ARGSUSED*/
94 int
95 socreate(dom, aso, type, proto)
96 	int dom;
97 	struct socket **aso;
98 	register int type;
99 	int proto;
100 {
101 	struct proc *p = curproc;		/* XXX */
102 	struct protosw *prp;
103 	struct socket *so;
104 	int error, s;
105 
106 	if (proto)
107 		prp = pffindproto(dom, proto, type);
108 	else
109 		prp = pffindtype(dom, type);
110 	if (prp == 0 || prp->pr_usrreq == 0)
111 		return (EPROTONOSUPPORT);
112 	if (prp->pr_type != type)
113 		return (EPROTOTYPE);
114 	s = splsoftnet();
115 	so = pool_get(&socket_pool, PR_WAITOK);
116 	bzero((caddr_t)so, sizeof(*so));
117 	TAILQ_INIT(&so->so_q0);
118 	TAILQ_INIT(&so->so_q);
119 	so->so_type = type;
120 	if (p->p_ucred->cr_uid == 0)
121 		so->so_state = SS_PRIV;
122 	so->so_ruid = p->p_cred->p_ruid;
123 	so->so_euid = p->p_ucred->cr_uid;
124 	so->so_proto = prp;
125 	error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL,
126 	    (struct mbuf *)(long)proto, NULL);
127 	if (error) {
128 		so->so_state |= SS_NOFDREF;
129 		sofree(so);
130 		splx(s);
131 		return (error);
132 	}
133 #ifdef COMPAT_SUNOS
134 	{
135 		extern struct emul emul_sunos;
136 		if (p->p_emul == &emul_sunos && type == SOCK_DGRAM)
137 			so->so_options |= SO_BROADCAST;
138 	}
139 #endif
140 	splx(s);
141 	*aso = so;
142 	return (0);
143 }
144 
145 int
146 sobind(so, nam)
147 	struct socket *so;
148 	struct mbuf *nam;
149 {
150 	int s = splsoftnet();
151 	int error;
152 
153 	error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL);
154 	splx(s);
155 	return (error);
156 }
157 
158 int
159 solisten(so, backlog)
160 	register struct socket *so;
161 	int backlog;
162 {
163 	int s = splsoftnet(), error;
164 
165 	error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL, NULL, NULL);
166 	if (error) {
167 		splx(s);
168 		return (error);
169 	}
170 	if (TAILQ_FIRST(&so->so_q) == NULL)
171 		so->so_options |= SO_ACCEPTCONN;
172 	if (backlog < 0 || backlog > somaxconn)
173 		backlog = somaxconn;
174 	if (backlog < sominconn)
175 		backlog = sominconn;
176 	so->so_qlimit = backlog;
177 	splx(s);
178 	return (0);
179 }
180 
181 /*
182  *  Must be called at splsoftnet()
183  */
184 
185 void
186 sofree(so)
187 	register struct socket *so;
188 {
189 
190 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
191 		return;
192 	if (so->so_head) {
193 		/*
194 		 * We must not decommission a socket that's on the accept(2)
195 		 * queue.  If we do, then accept(2) may hang after select(2)
196 		 * indicated that the listening socket was ready.
197 		 */
198 		if (!soqremque(so, 0))
199 			return;
200 	}
201 	sbrelease(&so->so_snd);
202 	sorflush(so);
203 	pool_put(&socket_pool, so);
204 }
205 
206 /*
207  * Close a socket on last file table reference removal.
208  * Initiate disconnect if connected.
209  * Free socket when disconnect complete.
210  */
211 int
212 soclose(so)
213 	register struct socket *so;
214 {
215 	struct socket *so2;
216 	int s = splsoftnet();		/* conservative */
217 	int error = 0;
218 
219 	if (so->so_options & SO_ACCEPTCONN) {
220 		while ((so2 = TAILQ_FIRST(&so->so_q0)) != NULL) {
221 			(void) soqremque(so2, 0);
222 			(void) soabort(so2);
223 		}
224 		while ((so2 = TAILQ_FIRST(&so->so_q)) != NULL) {
225 			(void) soqremque(so2, 1);
226 			(void) soabort(so2);
227 		}
228 	}
229 	if (so->so_pcb == 0)
230 		goto discard;
231 	if (so->so_state & SS_ISCONNECTED) {
232 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
233 			error = sodisconnect(so);
234 			if (error)
235 				goto drop;
236 		}
237 		if (so->so_options & SO_LINGER) {
238 			if ((so->so_state & SS_ISDISCONNECTING) &&
239 			    (so->so_state & SS_NBIO))
240 				goto drop;
241 			while (so->so_state & SS_ISCONNECTED) {
242 				error = tsleep((caddr_t)&so->so_timeo,
243 				    PSOCK | PCATCH, netcls,
244 				    so->so_linger * hz);
245 				if (error)
246 					break;
247 			}
248 		}
249 	}
250 drop:
251 	if (so->so_pcb) {
252 		int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, NULL,
253 							NULL, NULL);
254 		if (error == 0)
255 			error = error2;
256 	}
257 discard:
258 	if (so->so_state & SS_NOFDREF)
259 		panic("soclose: NOFDREF");
260 	so->so_state |= SS_NOFDREF;
261 	sofree(so);
262 	splx(s);
263 	return (error);
264 }
265 
266 /*
267  * Must be called at splsoftnet...
268  */
269 int
270 soabort(so)
271 	struct socket *so;
272 {
273 
274 	return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL, NULL, NULL);
275 }
276 
277 int
278 soaccept(so, nam)
279 	register struct socket *so;
280 	struct mbuf *nam;
281 {
282 	int s = splsoftnet();
283 	int error = 0;
284 
285 	if ((so->so_state & SS_NOFDREF) == 0)
286 		panic("soaccept: !NOFDREF");
287 	so->so_state &= ~SS_NOFDREF;
288 	if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
289 	    (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
290 		error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, NULL,
291 		    nam, NULL);
292 	else
293 		error = ECONNABORTED;
294 	splx(s);
295 	return (error);
296 }
297 
298 int
299 soconnect(so, nam)
300 	register struct socket *so;
301 	struct mbuf *nam;
302 {
303 	int s;
304 	int error;
305 
306 	if (so->so_options & SO_ACCEPTCONN)
307 		return (EOPNOTSUPP);
308 	s = splsoftnet();
309 	/*
310 	 * If protocol is connection-based, can only connect once.
311 	 * Otherwise, if connected, try to disconnect first.
312 	 * This allows user to disconnect by connecting to, e.g.,
313 	 * a null address.
314 	 */
315 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
316 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
317 	    (error = sodisconnect(so))))
318 		error = EISCONN;
319 	else
320 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
321 						   NULL, nam, NULL);
322 	splx(s);
323 	return (error);
324 }
325 
326 int
327 soconnect2(so1, so2)
328 	register struct socket *so1;
329 	struct socket *so2;
330 {
331 	int s = splsoftnet();
332 	int error;
333 
334 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, NULL,
335 					    (struct mbuf *)so2, NULL);
336 	splx(s);
337 	return (error);
338 }
339 
340 int
341 sodisconnect(so)
342 	register struct socket *so;
343 {
344 	int s = splsoftnet();
345 	int error;
346 
347 	if ((so->so_state & SS_ISCONNECTED) == 0) {
348 		error = ENOTCONN;
349 		goto bad;
350 	}
351 	if (so->so_state & SS_ISDISCONNECTING) {
352 		error = EALREADY;
353 		goto bad;
354 	}
355 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, NULL, NULL,
356 					   NULL);
357 bad:
358 	splx(s);
359 	return (error);
360 }
361 
362 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
363 /*
364  * Send on a socket.
365  * If send must go all at once and message is larger than
366  * send buffering, then hard error.
367  * Lock against other senders.
368  * If must go all at once and not enough room now, then
369  * inform user that this would block and do nothing.
370  * Otherwise, if nonblocking, send as much as possible.
371  * The data to be sent is described by "uio" if nonzero,
372  * otherwise by the mbuf chain "top" (which must be null
373  * if uio is not).  Data provided in mbuf chain must be small
374  * enough to send all at once.
375  *
376  * Returns nonzero on error, timeout or signal; callers
377  * must check for short counts if EINTR/ERESTART are returned.
378  * Data and control buffers are freed on return.
379  */
380 int
381 sosend(so, addr, uio, top, control, flags)
382 	register struct socket *so;
383 	struct mbuf *addr;
384 	struct uio *uio;
385 	struct mbuf *top;
386 	struct mbuf *control;
387 	int flags;
388 {
389 	struct proc *p = curproc;		/* XXX */
390 	struct mbuf **mp;
391 	struct mbuf *m;
392 	long space, len, mlen, clen = 0;
393 	quad_t resid;
394 	int error, s, dontroute;
395 	int atomic = sosendallatonce(so) || top;
396 
397 	if (uio)
398 		resid = uio->uio_resid;
399 	else
400 		resid = top->m_pkthdr.len;
401 	/*
402 	 * In theory resid should be unsigned (since uio->uio_resid is).
403 	 * However, space must be signed, as it might be less than 0
404 	 * if we over-committed, and we must use a signed comparison
405 	 * of space and resid.  On the other hand, a negative resid
406 	 * causes us to loop sending 0-length segments to the protocol.
407 	 * MSG_EOR on a SOCK_STREAM socket is also invalid.
408 	 */
409 	if (resid < 0 ||
410 	    (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
411 		error = EINVAL;
412 		goto out;
413 	}
414 	dontroute =
415 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
416 	    (so->so_proto->pr_flags & PR_ATOMIC);
417 	p->p_stats->p_ru.ru_msgsnd++;
418 	if (control)
419 		clen = control->m_len;
420 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
421 
422 restart:
423 	if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
424 		goto out;
425 	do {
426 		s = splsoftnet();
427 		if (so->so_state & SS_CANTSENDMORE)
428 			snderr(EPIPE);
429 		if (so->so_error) {
430 			error = so->so_error;
431 			so->so_error = 0;
432 			splx(s);
433 			goto release;
434 		}
435 		if ((so->so_state & SS_ISCONNECTED) == 0) {
436 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
437 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
438 				    !(resid == 0 && clen != 0))
439 					snderr(ENOTCONN);
440 			} else if (addr == 0)
441 				snderr(EDESTADDRREQ);
442 		}
443 		space = sbspace(&so->so_snd);
444 		if (flags & MSG_OOB)
445 			space += 1024;
446 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
447 		    clen > so->so_snd.sb_hiwat)
448 			snderr(EMSGSIZE);
449 		if (space < resid + clen && uio &&
450 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
451 			if (so->so_state & SS_NBIO)
452 				snderr(EWOULDBLOCK);
453 			sbunlock(&so->so_snd);
454 			error = sbwait(&so->so_snd);
455 			splx(s);
456 			if (error)
457 				goto out;
458 			goto restart;
459 		}
460 		splx(s);
461 		mp = &top;
462 		space -= clen;
463 		do {
464 		    if (uio == NULL) {
465 			/*
466 			 * Data is prepackaged in "top".
467 			 */
468 			resid = 0;
469 			if (flags & MSG_EOR)
470 				top->m_flags |= M_EOR;
471 		    } else do {
472 				if (top == 0) {
473 					MGETHDR(m, M_WAIT, MT_DATA);
474 					mlen = MHLEN;
475 					m->m_pkthdr.len = 0;
476 					m->m_pkthdr.rcvif = (struct ifnet *)0;
477 				} else {
478 					MGET(m, M_WAIT, MT_DATA);
479 					mlen = MLEN;
480 				}
481 				if (resid >= MINCLSIZE && space >= MCLBYTES) {
482 					MCLGET(m, M_WAIT);
483 					if ((m->m_flags & M_EXT) == 0)
484 						goto nopages;
485 					mlen = MCLBYTES;
486 					if (atomic && top == 0) {
487 						len = lmin(MCLBYTES - max_hdr, resid);
488 						m->m_data += max_hdr;
489 					} else
490 						len = lmin(MCLBYTES, resid);
491 					space -= len;
492 				} else {
493 nopages:
494 					len = lmin(lmin(mlen, resid), space);
495 					space -= len;
496 					/*
497 					 * For datagram protocols, leave room
498 					 * for protocol headers in first mbuf.
499 					 */
500 					if (atomic && top == 0 && len < mlen)
501 						MH_ALIGN(m, len);
502 				}
503 				error = uiomove(mtod(m, caddr_t), (int)len,
504 				    uio);
505 				resid = uio->uio_resid;
506 				m->m_len = len;
507 				*mp = m;
508 				top->m_pkthdr.len += len;
509 				if (error)
510 					goto release;
511 				mp = &m->m_next;
512 				if (resid <= 0) {
513 					if (flags & MSG_EOR)
514 						top->m_flags |= M_EOR;
515 					break;
516 				}
517 			} while (space > 0 && atomic);
518 			if (dontroute)
519 				so->so_options |= SO_DONTROUTE;
520 			s = splsoftnet();		/* XXX */
521 			error = (*so->so_proto->pr_usrreq)(so,
522 			    (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
523 			    top, addr, control);
524 			splx(s);
525 			if (dontroute)
526 				so->so_options &= ~SO_DONTROUTE;
527 			clen = 0;
528 			control = 0;
529 			top = 0;
530 			mp = &top;
531 			if (error)
532 				goto release;
533 		} while (resid && space > 0);
534 	} while (resid);
535 
536 release:
537 	sbunlock(&so->so_snd);
538 out:
539 	if (top)
540 		m_freem(top);
541 	if (control)
542 		m_freem(control);
543 	return (error);
544 }
545 
546 /*
547  * Implement receive operations on a socket.
548  * We depend on the way that records are added to the sockbuf
549  * by sbappend*.  In particular, each record (mbufs linked through m_next)
550  * must begin with an address if the protocol so specifies,
551  * followed by an optional mbuf or mbufs containing ancillary data,
552  * and then zero or more mbufs of data.
553  * In order to avoid blocking network interrupts for the entire time here,
554  * we splx() while doing the actual copy to user space.
555  * Although the sockbuf is locked, new data may still be appended,
556  * and thus we must maintain consistency of the sockbuf during that time.
557  *
558  * The caller may receive the data as a single mbuf chain by supplying
559  * an mbuf **mp0 for use in returning the chain.  The uio is then used
560  * only for the count in uio_resid.
561  */
562 int
563 soreceive(so, paddr, uio, mp0, controlp, flagsp)
564 	register struct socket *so;
565 	struct mbuf **paddr;
566 	struct uio *uio;
567 	struct mbuf **mp0;
568 	struct mbuf **controlp;
569 	int *flagsp;
570 {
571 	register struct mbuf *m, **mp;
572 	register int flags, len, error, s, offset;
573 	struct protosw *pr = so->so_proto;
574 	struct mbuf *nextrecord;
575 	int moff, type = 0;
576 	size_t orig_resid = uio->uio_resid;
577 	int uio_error = 0;
578 	int resid;
579 
580 	mp = mp0;
581 	if (paddr)
582 		*paddr = 0;
583 	if (controlp)
584 		*controlp = 0;
585 	if (flagsp)
586 		flags = *flagsp &~ MSG_EOR;
587 	else
588 		flags = 0;
589 	if (so->so_state & SS_NBIO)
590 		flags |= MSG_DONTWAIT;
591 	if (flags & MSG_OOB) {
592 		m = m_get(M_WAIT, MT_DATA);
593 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
594 		    (struct mbuf *)(long)(flags & MSG_PEEK), NULL);
595 		if (error)
596 			goto bad;
597 		do {
598 			error = uiomove(mtod(m, caddr_t),
599 			    (int) min(uio->uio_resid, m->m_len), uio);
600 			m = m_free(m);
601 		} while (uio->uio_resid && error == 0 && m);
602 bad:
603 		if (m)
604 			m_freem(m);
605 		return (error);
606 	}
607 	if (mp)
608 		*mp = (struct mbuf *)0;
609 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
610 		(*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL);
611 
612 restart:
613 	if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
614 		return (error);
615 	s = splsoftnet();
616 
617 	m = so->so_rcv.sb_mb;
618 	/*
619 	 * If we have less data than requested, block awaiting more
620 	 * (subject to any timeout) if:
621 	 *   1. the current count is less than the low water mark,
622 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
623 	 *	receive operation at once if we block (resid <= hiwat), or
624 	 *   3. MSG_DONTWAIT is not set.
625 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
626 	 * we have to do the receive in sections, and thus risk returning
627 	 * a short count if a timeout or signal occurs after we start.
628 	 */
629 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
630 	    so->so_rcv.sb_cc < uio->uio_resid) &&
631 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
632 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
633 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
634 #ifdef DIAGNOSTIC
635 		if (m == 0 && so->so_rcv.sb_cc)
636 			panic("receive 1");
637 #endif
638 		if (so->so_error) {
639 			if (m)
640 				goto dontblock;
641 			error = so->so_error;
642 			if ((flags & MSG_PEEK) == 0)
643 				so->so_error = 0;
644 			goto release;
645 		}
646 		if (so->so_state & SS_CANTRCVMORE) {
647 			if (m)
648 				goto dontblock;
649 			else
650 				goto release;
651 		}
652 		for (; m; m = m->m_next)
653 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
654 				m = so->so_rcv.sb_mb;
655 				goto dontblock;
656 			}
657 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
658 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
659 			error = ENOTCONN;
660 			goto release;
661 		}
662 		if (uio->uio_resid == 0 && controlp == NULL)
663 			goto release;
664 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
665 			error = EWOULDBLOCK;
666 			goto release;
667 		}
668 		sbunlock(&so->so_rcv);
669 		error = sbwait(&so->so_rcv);
670 		splx(s);
671 		if (error)
672 			return (error);
673 		goto restart;
674 	}
675 dontblock:
676 #ifdef notyet /* XXXX */
677 	if (uio->uio_procp)
678 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
679 #endif
680 	nextrecord = m->m_nextpkt;
681 	if (pr->pr_flags & PR_ADDR) {
682 #ifdef DIAGNOSTIC
683 		if (m->m_type != MT_SONAME)
684 			panic("receive 1a");
685 #endif
686 		orig_resid = 0;
687 		if (flags & MSG_PEEK) {
688 			if (paddr)
689 				*paddr = m_copy(m, 0, m->m_len);
690 			m = m->m_next;
691 		} else {
692 			sbfree(&so->so_rcv, m);
693 			if (paddr) {
694 				*paddr = m;
695 				so->so_rcv.sb_mb = m->m_next;
696 				m->m_next = 0;
697 				m = so->so_rcv.sb_mb;
698 			} else {
699 				MFREE(m, so->so_rcv.sb_mb);
700 				m = so->so_rcv.sb_mb;
701 			}
702 		}
703 	}
704 	while (m && m->m_type == MT_CONTROL && error == 0) {
705 		if (flags & MSG_PEEK) {
706 			if (controlp)
707 				*controlp = m_copy(m, 0, m->m_len);
708 			m = m->m_next;
709 		} else {
710 			sbfree(&so->so_rcv, m);
711 			if (controlp) {
712 				if (pr->pr_domain->dom_externalize &&
713 				    mtod(m, struct cmsghdr *)->cmsg_type ==
714 				    SCM_RIGHTS)
715 				   error = (*pr->pr_domain->dom_externalize)(m);
716 				*controlp = m;
717 				so->so_rcv.sb_mb = m->m_next;
718 				m->m_next = 0;
719 				m = so->so_rcv.sb_mb;
720 			} else {
721 				MFREE(m, so->so_rcv.sb_mb);
722 				m = so->so_rcv.sb_mb;
723 			}
724 		}
725 		if (controlp) {
726 			orig_resid = 0;
727 			controlp = &(*controlp)->m_next;
728 		}
729 	}
730 	if (m) {
731 		if ((flags & MSG_PEEK) == 0)
732 			m->m_nextpkt = nextrecord;
733 		type = m->m_type;
734 		if (type == MT_OOBDATA)
735 			flags |= MSG_OOB;
736 		if (m->m_flags & M_BCAST)
737 			flags |= MSG_BCAST;
738 		if (m->m_flags & M_MCAST)
739 			flags |= MSG_MCAST;
740 	}
741 	moff = 0;
742 	offset = 0;
743 	while (m && uio->uio_resid > 0 && error == 0) {
744 		if (m->m_type == MT_OOBDATA) {
745 			if (type != MT_OOBDATA)
746 				break;
747 		} else if (type == MT_OOBDATA)
748 			break;
749 #ifdef DIAGNOSTIC
750 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
751 			panic("receive 3");
752 #endif
753 		so->so_state &= ~SS_RCVATMARK;
754 		len = uio->uio_resid;
755 		if (so->so_oobmark && len > so->so_oobmark - offset)
756 			len = so->so_oobmark - offset;
757 		if (len > m->m_len - moff)
758 			len = m->m_len - moff;
759 		/*
760 		 * If mp is set, just pass back the mbufs.
761 		 * Otherwise copy them out via the uio, then free.
762 		 * Sockbuf must be consistent here (points to current mbuf,
763 		 * it points to next record) when we drop priority;
764 		 * we must note any additions to the sockbuf when we
765 		 * block interrupts again.
766 		 */
767 		if (mp == 0 && uio_error == 0) {
768 			resid = uio->uio_resid;
769 			splx(s);
770 			uio_error =
771 				uiomove(mtod(m, caddr_t) + moff, (int)len,
772 					uio);
773 			s = splsoftnet();
774 			if (uio_error)
775 				uio->uio_resid = resid - len;
776 		} else
777 			uio->uio_resid -= len;
778 		if (len == m->m_len - moff) {
779 			if (m->m_flags & M_EOR)
780 				flags |= MSG_EOR;
781 			if (flags & MSG_PEEK) {
782 				m = m->m_next;
783 				moff = 0;
784 			} else {
785 				nextrecord = m->m_nextpkt;
786 				sbfree(&so->so_rcv, m);
787 				if (mp) {
788 					*mp = m;
789 					mp = &m->m_next;
790 					so->so_rcv.sb_mb = m = m->m_next;
791 					*mp = (struct mbuf *)0;
792 				} else {
793 					MFREE(m, so->so_rcv.sb_mb);
794 					m = so->so_rcv.sb_mb;
795 				}
796 				if (m)
797 					m->m_nextpkt = nextrecord;
798 			}
799 		} else {
800 			if (flags & MSG_PEEK)
801 				moff += len;
802 			else {
803 				if (mp)
804 					*mp = m_copym(m, 0, len, M_WAIT);
805 				m->m_data += len;
806 				m->m_len -= len;
807 				so->so_rcv.sb_cc -= len;
808 			}
809 		}
810 		if (so->so_oobmark) {
811 			if ((flags & MSG_PEEK) == 0) {
812 				so->so_oobmark -= len;
813 				if (so->so_oobmark == 0) {
814 					so->so_state |= SS_RCVATMARK;
815 					break;
816 				}
817 			} else {
818 				offset += len;
819 				if (offset == so->so_oobmark)
820 					break;
821 			}
822 		}
823 		if (flags & MSG_EOR)
824 			break;
825 		/*
826 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
827 		 * we must not quit until "uio->uio_resid == 0" or an error
828 		 * termination.  If a signal/timeout occurs, return
829 		 * with a short count but without error.
830 		 * Keep sockbuf locked against other readers.
831 		 */
832 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
833 		    !sosendallatonce(so) && !nextrecord) {
834 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
835 				break;
836 			error = sbwait(&so->so_rcv);
837 			if (error) {
838 				sbunlock(&so->so_rcv);
839 				splx(s);
840 				return (0);
841 			}
842 			if ((m = so->so_rcv.sb_mb) != NULL)
843 				nextrecord = m->m_nextpkt;
844 		}
845 	}
846 
847 	if (m && pr->pr_flags & PR_ATOMIC) {
848 		flags |= MSG_TRUNC;
849 		if ((flags & MSG_PEEK) == 0)
850 			(void) sbdroprecord(&so->so_rcv);
851 	}
852 	if ((flags & MSG_PEEK) == 0) {
853 		if (m == 0)
854 			so->so_rcv.sb_mb = nextrecord;
855 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
856 			(*pr->pr_usrreq)(so, PRU_RCVD, NULL,
857 					 (struct mbuf *)(long)flags, NULL);
858 	}
859 	if (orig_resid == uio->uio_resid && orig_resid &&
860 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
861 		sbunlock(&so->so_rcv);
862 		splx(s);
863 		goto restart;
864 	}
865 
866 	if (uio_error)
867 		error = uio_error;
868 
869 	if (flagsp)
870 		*flagsp |= flags;
871 release:
872 	sbunlock(&so->so_rcv);
873 	splx(s);
874 	return (error);
875 }
876 
877 int
878 soshutdown(so, how)
879 	register struct socket *so;
880 	register int how;
881 {
882 	register struct protosw *pr = so->so_proto;
883 
884 	how++;
885 	if (how & ~(FREAD|FWRITE))
886 		return (EINVAL);
887 	if (how & FREAD)
888 		sorflush(so);
889 	if (how & FWRITE)
890 		return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL, NULL, NULL);
891 	return (0);
892 }
893 
894 void
895 sorflush(so)
896 	register struct socket *so;
897 {
898 	register struct sockbuf *sb = &so->so_rcv;
899 	register struct protosw *pr = so->so_proto;
900 	register int s;
901 	struct sockbuf asb;
902 
903 	sb->sb_flags |= SB_NOINTR;
904 	(void) sblock(sb, M_WAITOK);
905 	s = splimp();
906 	socantrcvmore(so);
907 	sbunlock(sb);
908 	asb = *sb;
909 	bzero((caddr_t)sb, sizeof (*sb));
910 	/* XXX - the bzero stumps all over so_rcv */
911 	if (asb.sb_flags & SB_KNOTE) {
912 		sb->sb_sel.si_note = asb.sb_sel.si_note;
913 		sb->sb_flags = SB_KNOTE;
914 	}
915 	splx(s);
916 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
917 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
918 	sbrelease(&asb);
919 }
920 
921 int
922 sosetopt(so, level, optname, m0)
923 	register struct socket *so;
924 	int level, optname;
925 	struct mbuf *m0;
926 {
927 	int error = 0;
928 	register struct mbuf *m = m0;
929 
930 	if (level != SOL_SOCKET) {
931 		if (so->so_proto && so->so_proto->pr_ctloutput)
932 			return ((*so->so_proto->pr_ctloutput)
933 				  (PRCO_SETOPT, so, level, optname, &m0));
934 		error = ENOPROTOOPT;
935 	} else {
936 		switch (optname) {
937 
938 		case SO_LINGER:
939 			if (m == NULL || m->m_len != sizeof (struct linger)) {
940 				error = EINVAL;
941 				goto bad;
942 			}
943 			so->so_linger = mtod(m, struct linger *)->l_linger;
944 			/* fall thru... */
945 
946 		case SO_DEBUG:
947 		case SO_KEEPALIVE:
948 		case SO_DONTROUTE:
949 		case SO_USELOOPBACK:
950 		case SO_BROADCAST:
951 		case SO_REUSEADDR:
952 		case SO_REUSEPORT:
953 		case SO_OOBINLINE:
954 			if (m == NULL || m->m_len < sizeof (int)) {
955 				error = EINVAL;
956 				goto bad;
957 			}
958 			if (*mtod(m, int *))
959 				so->so_options |= optname;
960 			else
961 				so->so_options &= ~optname;
962 			break;
963 
964 		case SO_SNDBUF:
965 		case SO_RCVBUF:
966 		case SO_SNDLOWAT:
967 		case SO_RCVLOWAT:
968 		    {
969 			u_long cnt;
970 
971 			if (m == NULL || m->m_len < sizeof (int)) {
972 				error = EINVAL;
973 				goto bad;
974 			}
975 			cnt = *mtod(m, int *);
976 			if ((long)cnt <= 0)
977 				cnt = 1;
978 			switch (optname) {
979 
980 			case SO_SNDBUF:
981 			case SO_RCVBUF:
982 				if (sbreserve(optname == SO_SNDBUF ?
983 				    &so->so_snd : &so->so_rcv,
984 				    cnt) == 0) {
985 					error = ENOBUFS;
986 					goto bad;
987 				}
988 				break;
989 
990 			case SO_SNDLOWAT:
991 				so->so_snd.sb_lowat = (cnt > so->so_snd.sb_hiwat) ?
992 				    so->so_snd.sb_hiwat : cnt;
993 				break;
994 			case SO_RCVLOWAT:
995 				so->so_rcv.sb_lowat = (cnt > so->so_rcv.sb_hiwat) ?
996 				    so->so_rcv.sb_hiwat : cnt;
997 				break;
998 			}
999 			break;
1000 		    }
1001 
1002 		case SO_SNDTIMEO:
1003 		case SO_RCVTIMEO:
1004 		    {
1005 			struct timeval *tv;
1006 			short val;
1007 
1008 			if (m == NULL || m->m_len < sizeof (*tv)) {
1009 				error = EINVAL;
1010 				goto bad;
1011 			}
1012 			tv = mtod(m, struct timeval *);
1013 			if (tv->tv_sec * hz + tv->tv_usec / tick > SHRT_MAX) {
1014 				error = EDOM;
1015 				goto bad;
1016 			}
1017 			val = tv->tv_sec * hz + tv->tv_usec / tick;
1018 
1019 			switch (optname) {
1020 
1021 			case SO_SNDTIMEO:
1022 				so->so_snd.sb_timeo = val;
1023 				break;
1024 			case SO_RCVTIMEO:
1025 				so->so_rcv.sb_timeo = val;
1026 				break;
1027 			}
1028 			break;
1029 		    }
1030 
1031 		default:
1032 			error = ENOPROTOOPT;
1033 			break;
1034 		}
1035 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1036 			(void) ((*so->so_proto->pr_ctloutput)
1037 				  (PRCO_SETOPT, so, level, optname, &m0));
1038 			m = NULL;	/* freed by protocol */
1039 		}
1040 	}
1041 bad:
1042 	if (m)
1043 		(void) m_free(m);
1044 	return (error);
1045 }
1046 
1047 int
1048 sogetopt(so, level, optname, mp)
1049 	register struct socket *so;
1050 	int level, optname;
1051 	struct mbuf **mp;
1052 {
1053 	register struct mbuf *m;
1054 
1055 	if (level != SOL_SOCKET) {
1056 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1057 			return ((*so->so_proto->pr_ctloutput)
1058 				  (PRCO_GETOPT, so, level, optname, mp));
1059 		} else
1060 			return (ENOPROTOOPT);
1061 	} else {
1062 		m = m_get(M_WAIT, MT_SOOPTS);
1063 		m->m_len = sizeof (int);
1064 
1065 		switch (optname) {
1066 
1067 		case SO_LINGER:
1068 			m->m_len = sizeof (struct linger);
1069 			mtod(m, struct linger *)->l_onoff =
1070 				so->so_options & SO_LINGER;
1071 			mtod(m, struct linger *)->l_linger = so->so_linger;
1072 			break;
1073 
1074 		case SO_USELOOPBACK:
1075 		case SO_DONTROUTE:
1076 		case SO_DEBUG:
1077 		case SO_KEEPALIVE:
1078 		case SO_REUSEADDR:
1079 		case SO_REUSEPORT:
1080 		case SO_BROADCAST:
1081 		case SO_OOBINLINE:
1082 			*mtod(m, int *) = so->so_options & optname;
1083 			break;
1084 
1085 		case SO_TYPE:
1086 			*mtod(m, int *) = so->so_type;
1087 			break;
1088 
1089 		case SO_ERROR:
1090 			*mtod(m, int *) = so->so_error;
1091 			so->so_error = 0;
1092 			break;
1093 
1094 		case SO_SNDBUF:
1095 			*mtod(m, int *) = so->so_snd.sb_hiwat;
1096 			break;
1097 
1098 		case SO_RCVBUF:
1099 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
1100 			break;
1101 
1102 		case SO_SNDLOWAT:
1103 			*mtod(m, int *) = so->so_snd.sb_lowat;
1104 			break;
1105 
1106 		case SO_RCVLOWAT:
1107 			*mtod(m, int *) = so->so_rcv.sb_lowat;
1108 			break;
1109 
1110 		case SO_SNDTIMEO:
1111 		case SO_RCVTIMEO:
1112 		    {
1113 			int val = (optname == SO_SNDTIMEO ?
1114 			    so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1115 
1116 			m->m_len = sizeof(struct timeval);
1117 			mtod(m, struct timeval *)->tv_sec = val / hz;
1118 			mtod(m, struct timeval *)->tv_usec =
1119 			    (val % hz) * tick;
1120 			break;
1121 		    }
1122 
1123 		default:
1124 			(void)m_free(m);
1125 			return (ENOPROTOOPT);
1126 		}
1127 		*mp = m;
1128 		return (0);
1129 	}
1130 }
1131 
1132 void
1133 sohasoutofband(so)
1134 	register struct socket *so;
1135 {
1136 	csignal(so->so_pgid, SIGURG, so->so_siguid, so->so_sigeuid);
1137 	selwakeup(&so->so_rcv.sb_sel);
1138 }
1139 
1140 int
1141 soo_kqfilter(struct file *fp, struct knote *kn)
1142 {
1143 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1144 	struct sockbuf *sb;
1145 	int s;
1146 
1147 	switch (kn->kn_filter) {
1148 	case EVFILT_READ:
1149 		if (so->so_options & SO_ACCEPTCONN)
1150 			kn->kn_fop = &solisten_filtops;
1151 		else
1152 			kn->kn_fop = &soread_filtops;
1153 		sb = &so->so_rcv;
1154 		break;
1155 	case EVFILT_WRITE:
1156 		kn->kn_fop = &sowrite_filtops;
1157 		sb = &so->so_snd;
1158 		break;
1159 	default:
1160 		return (1);
1161 	}
1162 
1163 	s = splnet();
1164 	SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1165 	sb->sb_flags |= SB_KNOTE;
1166 	splx(s);
1167 	return (0);
1168 }
1169 
1170 void
1171 filt_sordetach(struct knote *kn)
1172 {
1173 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1174 	int s = splnet();
1175 
1176 	SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1177 	if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1178 		so->so_rcv.sb_flags &= ~SB_KNOTE;
1179 	splx(s);
1180 }
1181 
1182 /*ARGSUSED*/
1183 int
1184 filt_soread(struct knote *kn, long hint)
1185 {
1186 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1187 
1188 	kn->kn_data = so->so_rcv.sb_cc;
1189 	if (so->so_state & SS_CANTRCVMORE) {
1190 		kn->kn_flags |= EV_EOF;
1191 		kn->kn_fflags = so->so_error;
1192 		return (1);
1193 	}
1194 	if (so->so_error)	/* temporary udp error */
1195 		return (1);
1196 	if (kn->kn_sfflags & NOTE_LOWAT)
1197 		return (kn->kn_data >= kn->kn_sdata);
1198 	return (kn->kn_data >= so->so_rcv.sb_lowat);
1199 }
1200 
1201 void
1202 filt_sowdetach(struct knote *kn)
1203 {
1204 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1205 	int s = splnet();
1206 
1207 	SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1208 	if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1209 		so->so_snd.sb_flags &= ~SB_KNOTE;
1210 	splx(s);
1211 }
1212 
1213 /*ARGSUSED*/
1214 int
1215 filt_sowrite(struct knote *kn, long hint)
1216 {
1217 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1218 
1219 	kn->kn_data = sbspace(&so->so_snd);
1220 	if (so->so_state & SS_CANTSENDMORE) {
1221 		kn->kn_flags |= EV_EOF;
1222 		kn->kn_fflags = so->so_error;
1223 		return (1);
1224 	}
1225 	if (so->so_error)	/* temporary udp error */
1226 		return (1);
1227 	if (((so->so_state & SS_ISCONNECTED) == 0) &&
1228 	    (so->so_proto->pr_flags & PR_CONNREQUIRED))
1229 		return (0);
1230 	if (kn->kn_sfflags & NOTE_LOWAT)
1231 		return (kn->kn_data >= kn->kn_sdata);
1232 	return (kn->kn_data >= so->so_snd.sb_lowat);
1233 }
1234 
1235 /*ARGSUSED*/
1236 int
1237 filt_solisten(struct knote *kn, long hint)
1238 {
1239 	struct socket *so = (struct socket *)kn->kn_fp->f_data;
1240 
1241 	kn->kn_data = so->so_qlen;
1242 	return (so->so_qlen != 0);
1243 }
1244