xref: /netbsd-src/sys/kern/uipc_socket.c (revision 95d875fb90b1458e4f1de6950286ddcd6644bc61)
1 /*	$NetBSD: uipc_socket.c,v 1.48 1999/06/08 02:39:57 thorpej Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)uipc_socket.c	8.6 (Berkeley) 5/2/95
36  */
37 
38 #include "opt_compat_sunos.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/proc.h>
43 #include <sys/file.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/domain.h>
47 #include <sys/kernel.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 struct pool socket_pool;
56 
57 void
58 soinit()
59 {
60 
61 	pool_init(&socket_pool, sizeof(struct socket), 0, 0, 0,
62 	    "sockpl", 0, NULL, NULL, M_SOCKET);
63 }
64 
65 /*
66  * Socket operation routines.
67  * These routines are called by the routines in
68  * sys_socket.c or from a system process, and
69  * implement the semantics of socket operations by
70  * switching out to the protocol specific routines.
71  */
72 /*ARGSUSED*/
73 int
74 socreate(dom, aso, type, proto)
75 	int dom;
76 	struct socket **aso;
77 	register int type;
78 	int proto;
79 {
80 	struct proc *p = curproc;		/* XXX */
81 	register struct protosw *prp;
82 	register struct socket *so;
83 	register int error;
84 	int s;
85 
86 	if (proto)
87 		prp = pffindproto(dom, proto, type);
88 	else
89 		prp = pffindtype(dom, type);
90 	if (prp == 0 || prp->pr_usrreq == 0)
91 		return (EPROTONOSUPPORT);
92 	if (prp->pr_type != type)
93 		return (EPROTOTYPE);
94 	s = splsoftnet();
95 	so = pool_get(&socket_pool, PR_WAITOK);
96 	memset((caddr_t)so, 0, sizeof(*so));
97 	TAILQ_INIT(&so->so_q0);
98 	TAILQ_INIT(&so->so_q);
99 	so->so_type = type;
100 	so->so_proto = prp;
101 	so->so_send = sosend;
102 	so->so_receive = soreceive;
103 	if (p != 0)
104 		so->so_uid = p->p_ucred->cr_uid;
105 	error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
106 	    (struct mbuf *)(long)proto, (struct mbuf *)0, p);
107 	if (error) {
108 		so->so_state |= SS_NOFDREF;
109 		sofree(so);
110 		splx(s);
111 		return (error);
112 	}
113 #ifdef COMPAT_SUNOS
114 	{
115 		extern struct emul emul_sunos;
116 		if (p->p_emul == &emul_sunos && type == SOCK_DGRAM)
117 			so->so_options |= SO_BROADCAST;
118 	}
119 #endif
120 	splx(s);
121 	*aso = so;
122 	return (0);
123 }
124 
125 int
126 sobind(so, nam)
127 	struct socket *so;
128 	struct mbuf *nam;
129 {
130 	struct proc *p = curproc;		/* XXX */
131 	int s = splsoftnet();
132 	int error;
133 
134 	error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0,
135 	    nam, (struct mbuf *)0, p);
136 	splx(s);
137 	return (error);
138 }
139 
140 int
141 solisten(so, backlog)
142 	register struct socket *so;
143 	int backlog;
144 {
145 	int s = splsoftnet(), error;
146 
147 	error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0,
148 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
149 	if (error) {
150 		splx(s);
151 		return (error);
152 	}
153 	if (so->so_q.tqh_first == NULL)
154 		so->so_options |= SO_ACCEPTCONN;
155 	if (backlog < 0)
156 		backlog = 0;
157 	so->so_qlimit = min(backlog, SOMAXCONN);
158 	splx(s);
159 	return (0);
160 }
161 
162 void
163 sofree(so)
164 	register struct socket *so;
165 {
166 
167 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
168 		return;
169 	if (so->so_head) {
170 		/*
171 		 * We must not decommission a socket that's on the accept(2)
172 		 * queue.  If we do, then accept(2) may hang after select(2)
173 		 * indicated that the listening socket was ready.
174 		 */
175 		if (!soqremque(so, 0))
176 			return;
177 	}
178 	sbrelease(&so->so_snd);
179 	sorflush(so);
180 	pool_put(&socket_pool, so);
181 }
182 
183 /*
184  * Close a socket on last file table reference removal.
185  * Initiate disconnect if connected.
186  * Free socket when disconnect complete.
187  */
188 int
189 soclose(so)
190 	register struct socket *so;
191 {
192 	struct socket *so2;
193 	int s = splsoftnet();		/* conservative */
194 	int error = 0;
195 
196 	if (so->so_options & SO_ACCEPTCONN) {
197 		while ((so2 = so->so_q0.tqh_first) != 0) {
198 			(void) soqremque(so2, 0);
199 			(void) soabort(so2);
200 		}
201 		while ((so2 = so->so_q.tqh_first) != 0) {
202 			(void) soqremque(so2, 1);
203 			(void) soabort(so2);
204 		}
205 	}
206 	if (so->so_pcb == 0)
207 		goto discard;
208 	if (so->so_state & SS_ISCONNECTED) {
209 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
210 			error = sodisconnect(so);
211 			if (error)
212 				goto drop;
213 		}
214 		if (so->so_options & SO_LINGER) {
215 			if ((so->so_state & SS_ISDISCONNECTING) &&
216 			    (so->so_state & SS_NBIO))
217 				goto drop;
218 			while (so->so_state & SS_ISCONNECTED) {
219 				error = tsleep((caddr_t)&so->so_timeo,
220 					       PSOCK | PCATCH, netcls,
221 					       so->so_linger * hz);
222 				if (error)
223 					break;
224 			}
225 		}
226 	}
227 drop:
228 	if (so->so_pcb) {
229 		int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
230 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
231 		    (struct proc *)0);
232 		if (error == 0)
233 			error = error2;
234 	}
235 discard:
236 	if (so->so_state & SS_NOFDREF)
237 		panic("soclose: NOFDREF");
238 	so->so_state |= SS_NOFDREF;
239 	sofree(so);
240 	splx(s);
241 	return (error);
242 }
243 
244 /*
245  * Must be called at splsoftnet...
246  */
247 int
248 soabort(so)
249 	struct socket *so;
250 {
251 
252 	return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0,
253 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
254 }
255 
256 int
257 soaccept(so, nam)
258 	register struct socket *so;
259 	struct mbuf *nam;
260 {
261 	int s = splsoftnet();
262 	int error;
263 
264 	if ((so->so_state & SS_NOFDREF) == 0)
265 		panic("soaccept: !NOFDREF");
266 	so->so_state &= ~SS_NOFDREF;
267 	if ((so->so_state & SS_ISDISCONNECTED) == 0)
268 		error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
269 		    (struct mbuf *)0, nam, (struct mbuf *)0, (struct proc *)0);
270 	else
271 		error = 0;
272 	splx(s);
273 	return (error);
274 }
275 
276 int
277 soconnect(so, nam)
278 	register struct socket *so;
279 	struct mbuf *nam;
280 {
281 	struct proc *p = curproc;		/* XXX */
282 	int s;
283 	int error;
284 
285 	if (so->so_options & SO_ACCEPTCONN)
286 		return (EOPNOTSUPP);
287 	s = splsoftnet();
288 	/*
289 	 * If protocol is connection-based, can only connect once.
290 	 * Otherwise, if connected, try to disconnect first.
291 	 * This allows user to disconnect by connecting to, e.g.,
292 	 * a null address.
293 	 */
294 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
295 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
296 	    (error = sodisconnect(so))))
297 		error = EISCONN;
298 	else
299 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
300 		    (struct mbuf *)0, nam, (struct mbuf *)0, p);
301 	splx(s);
302 	return (error);
303 }
304 
305 int
306 soconnect2(so1, so2)
307 	register struct socket *so1;
308 	struct socket *so2;
309 {
310 	int s = splsoftnet();
311 	int error;
312 
313 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
314 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0,
315 	    (struct proc *)0);
316 	splx(s);
317 	return (error);
318 }
319 
320 int
321 sodisconnect(so)
322 	register struct socket *so;
323 {
324 	int s = splsoftnet();
325 	int error;
326 
327 	if ((so->so_state & SS_ISCONNECTED) == 0) {
328 		error = ENOTCONN;
329 		goto bad;
330 	}
331 	if (so->so_state & SS_ISDISCONNECTING) {
332 		error = EALREADY;
333 		goto bad;
334 	}
335 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
336 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
337 	    (struct proc *)0);
338 bad:
339 	splx(s);
340 	return (error);
341 }
342 
343 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
344 /*
345  * Send on a socket.
346  * If send must go all at once and message is larger than
347  * send buffering, then hard error.
348  * Lock against other senders.
349  * If must go all at once and not enough room now, then
350  * inform user that this would block and do nothing.
351  * Otherwise, if nonblocking, send as much as possible.
352  * The data to be sent is described by "uio" if nonzero,
353  * otherwise by the mbuf chain "top" (which must be null
354  * if uio is not).  Data provided in mbuf chain must be small
355  * enough to send all at once.
356  *
357  * Returns nonzero on error, timeout or signal; callers
358  * must check for short counts if EINTR/ERESTART are returned.
359  * Data and control buffers are freed on return.
360  */
361 int
362 sosend(so, addr, uio, top, control, flags)
363 	register struct socket *so;
364 	struct mbuf *addr;
365 	struct uio *uio;
366 	struct mbuf *top;
367 	struct mbuf *control;
368 	int flags;
369 {
370 	struct proc *p = curproc;		/* XXX */
371 	struct mbuf **mp;
372 	register struct mbuf *m;
373 	register long space, len, resid;
374 	int clen = 0, error, s, dontroute, mlen;
375 	int atomic = sosendallatonce(so) || top;
376 
377 	if (uio)
378 		resid = uio->uio_resid;
379 	else
380 		resid = top->m_pkthdr.len;
381 	/*
382 	 * In theory resid should be unsigned.
383 	 * However, space must be signed, as it might be less than 0
384 	 * if we over-committed, and we must use a signed comparison
385 	 * of space and resid.  On the other hand, a negative resid
386 	 * causes us to loop sending 0-length segments to the protocol.
387 	 */
388 	if (resid < 0) {
389 		error = EINVAL;
390 		goto out;
391 	}
392 	dontroute =
393 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
394 	    (so->so_proto->pr_flags & PR_ATOMIC);
395 	p->p_stats->p_ru.ru_msgsnd++;
396 	if (control)
397 		clen = control->m_len;
398 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
399 
400 restart:
401 	if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
402 		goto out;
403 	do {
404 		s = splsoftnet();
405 		if (so->so_state & SS_CANTSENDMORE)
406 			snderr(EPIPE);
407 		if (so->so_error) {
408 			error = so->so_error;
409 			so->so_error = 0;
410 			splx(s);
411 			goto release;
412 		}
413 		if ((so->so_state & SS_ISCONNECTED) == 0) {
414 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
415 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
416 				    !(resid == 0 && clen != 0))
417 					snderr(ENOTCONN);
418 			} else if (addr == 0)
419 				snderr(EDESTADDRREQ);
420 		}
421 		space = sbspace(&so->so_snd);
422 		if (flags & MSG_OOB)
423 			space += 1024;
424 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
425 		    clen > so->so_snd.sb_hiwat)
426 			snderr(EMSGSIZE);
427 		if (space < resid + clen && uio &&
428 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
429 			if (so->so_state & SS_NBIO)
430 				snderr(EWOULDBLOCK);
431 			sbunlock(&so->so_snd);
432 			error = sbwait(&so->so_snd);
433 			splx(s);
434 			if (error)
435 				goto out;
436 			goto restart;
437 		}
438 		splx(s);
439 		mp = &top;
440 		space -= clen;
441 		do {
442 			if (uio == NULL) {
443 				/*
444 				 * Data is prepackaged in "top".
445 				 */
446 				resid = 0;
447 				if (flags & MSG_EOR)
448 					top->m_flags |= M_EOR;
449 			} else do {
450 				if (top == 0) {
451 					MGETHDR(m, M_WAIT, MT_DATA);
452 					mlen = MHLEN;
453 					m->m_pkthdr.len = 0;
454 					m->m_pkthdr.rcvif = (struct ifnet *)0;
455 				} else {
456 					MGET(m, M_WAIT, MT_DATA);
457 					mlen = MLEN;
458 				}
459 				if (resid >= MINCLSIZE && space >= MCLBYTES) {
460 					MCLGET(m, M_WAIT);
461 					if ((m->m_flags & M_EXT) == 0)
462 						goto nopages;
463 					mlen = MCLBYTES;
464 #ifdef	MAPPED_MBUFS
465 					len = min(MCLBYTES, resid);
466 #else
467 					if (atomic && top == 0) {
468 						len = min(MCLBYTES - max_hdr, resid);
469 						m->m_data += max_hdr;
470 					} else
471 						len = min(MCLBYTES, resid);
472 #endif
473 					space -= len;
474 				} else {
475 nopages:
476 					len = min(min(mlen, resid), space);
477 					space -= len;
478 					/*
479 					 * For datagram protocols, leave room
480 					 * for protocol headers in first mbuf.
481 					 */
482 					if (atomic && top == 0 && len < mlen)
483 						MH_ALIGN(m, len);
484 				}
485 				error = uiomove(mtod(m, caddr_t), (int)len, uio);
486 				resid = uio->uio_resid;
487 				m->m_len = len;
488 				*mp = m;
489 				top->m_pkthdr.len += len;
490 				if (error)
491 					goto release;
492 				mp = &m->m_next;
493 				if (resid <= 0) {
494 					if (flags & MSG_EOR)
495 						top->m_flags |= M_EOR;
496 					break;
497 				}
498 			} while (space > 0 && atomic);
499 
500 			s = splsoftnet();
501 
502 			if (so->so_state & SS_CANTSENDMORE)
503 				snderr(EPIPE);
504 
505 			if (dontroute)
506 				so->so_options |= SO_DONTROUTE;
507 			if (resid > 0)
508 				so->so_state |= SS_MORETOCOME;
509 			error = (*so->so_proto->pr_usrreq)(so,
510 			    (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
511 			    top, addr, control, p);
512 			if (dontroute)
513 				so->so_options &= ~SO_DONTROUTE;
514 			if (resid > 0)
515 				so->so_state &= ~SS_MORETOCOME;
516 			splx(s);
517 
518 			clen = 0;
519 			control = 0;
520 			top = 0;
521 			mp = &top;
522 			if (error)
523 				goto release;
524 		} while (resid && space > 0);
525 	} while (resid);
526 
527 release:
528 	sbunlock(&so->so_snd);
529 out:
530 	if (top)
531 		m_freem(top);
532 	if (control)
533 		m_freem(control);
534 	return (error);
535 }
536 
537 /*
538  * Implement receive operations on a socket.
539  * We depend on the way that records are added to the sockbuf
540  * by sbappend*.  In particular, each record (mbufs linked through m_next)
541  * must begin with an address if the protocol so specifies,
542  * followed by an optional mbuf or mbufs containing ancillary data,
543  * and then zero or more mbufs of data.
544  * In order to avoid blocking network interrupts for the entire time here,
545  * we splx() while doing the actual copy to user space.
546  * Although the sockbuf is locked, new data may still be appended,
547  * and thus we must maintain consistency of the sockbuf during that time.
548  *
549  * The caller may receive the data as a single mbuf chain by supplying
550  * an mbuf **mp0 for use in returning the chain.  The uio is then used
551  * only for the count in uio_resid.
552  */
553 int
554 soreceive(so, paddr, uio, mp0, controlp, flagsp)
555 	register struct socket *so;
556 	struct mbuf **paddr;
557 	struct uio *uio;
558 	struct mbuf **mp0;
559 	struct mbuf **controlp;
560 	int *flagsp;
561 {
562 	register struct mbuf *m, **mp;
563 	register int flags, len, error, s, offset;
564 	struct protosw *pr = so->so_proto;
565 	struct mbuf *nextrecord;
566 	int moff, type = 0;
567 	int orig_resid = uio->uio_resid;
568 
569 	mp = mp0;
570 	if (paddr)
571 		*paddr = 0;
572 	if (controlp)
573 		*controlp = 0;
574 	if (flagsp)
575 		flags = *flagsp &~ MSG_EOR;
576 	else
577 		flags = 0;
578 	if (flags & MSG_OOB) {
579 		m = m_get(M_WAIT, MT_DATA);
580 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
581 		    (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0,
582 		    (struct proc *)0);
583 		if (error)
584 			goto bad;
585 		do {
586 			error = uiomove(mtod(m, caddr_t),
587 			    (int) min(uio->uio_resid, m->m_len), uio);
588 			m = m_free(m);
589 		} while (uio->uio_resid && error == 0 && m);
590 bad:
591 		if (m)
592 			m_freem(m);
593 		return (error);
594 	}
595 	if (mp)
596 		*mp = (struct mbuf *)0;
597 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
598 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
599 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
600 
601 restart:
602 	if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
603 		return (error);
604 	s = splsoftnet();
605 
606 	m = so->so_rcv.sb_mb;
607 	/*
608 	 * If we have less data than requested, block awaiting more
609 	 * (subject to any timeout) if:
610 	 *   1. the current count is less than the low water mark,
611 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
612 	 *	receive operation at once if we block (resid <= hiwat), or
613 	 *   3. MSG_DONTWAIT is not set.
614 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
615 	 * we have to do the receive in sections, and thus risk returning
616 	 * a short count if a timeout or signal occurs after we start.
617 	 */
618 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
619 	    so->so_rcv.sb_cc < uio->uio_resid) &&
620 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
621 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
622 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
623 #ifdef DIAGNOSTIC
624 		if (m == 0 && so->so_rcv.sb_cc)
625 			panic("receive 1");
626 #endif
627 		if (so->so_error) {
628 			if (m)
629 				goto dontblock;
630 			error = so->so_error;
631 			if ((flags & MSG_PEEK) == 0)
632 				so->so_error = 0;
633 			goto release;
634 		}
635 		if (so->so_state & SS_CANTRCVMORE) {
636 			if (m)
637 				goto dontblock;
638 			else
639 				goto release;
640 		}
641 		for (; m; m = m->m_next)
642 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
643 				m = so->so_rcv.sb_mb;
644 				goto dontblock;
645 			}
646 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
647 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
648 			error = ENOTCONN;
649 			goto release;
650 		}
651 		if (uio->uio_resid == 0)
652 			goto release;
653 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
654 			error = EWOULDBLOCK;
655 			goto release;
656 		}
657 		sbunlock(&so->so_rcv);
658 		error = sbwait(&so->so_rcv);
659 		splx(s);
660 		if (error)
661 			return (error);
662 		goto restart;
663 	}
664 dontblock:
665 #ifdef notyet /* XXXX */
666 	if (uio->uio_procp)
667 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
668 #endif
669 	nextrecord = m->m_nextpkt;
670 	if (pr->pr_flags & PR_ADDR) {
671 #ifdef DIAGNOSTIC
672 		if (m->m_type != MT_SONAME)
673 			panic("receive 1a");
674 #endif
675 		orig_resid = 0;
676 		if (flags & MSG_PEEK) {
677 			if (paddr)
678 				*paddr = m_copy(m, 0, m->m_len);
679 			m = m->m_next;
680 		} else {
681 			sbfree(&so->so_rcv, m);
682 			if (paddr) {
683 				*paddr = m;
684 				so->so_rcv.sb_mb = m->m_next;
685 				m->m_next = 0;
686 				m = so->so_rcv.sb_mb;
687 			} else {
688 				MFREE(m, so->so_rcv.sb_mb);
689 				m = so->so_rcv.sb_mb;
690 			}
691 		}
692 	}
693 	while (m && m->m_type == MT_CONTROL && error == 0) {
694 		if (flags & MSG_PEEK) {
695 			if (controlp)
696 				*controlp = m_copy(m, 0, m->m_len);
697 			m = m->m_next;
698 		} else {
699 			sbfree(&so->so_rcv, m);
700 			if (controlp) {
701 				if (pr->pr_domain->dom_externalize &&
702 				    mtod(m, struct cmsghdr *)->cmsg_type ==
703 				    SCM_RIGHTS)
704 					error = (*pr->pr_domain->dom_externalize)(m);
705 				*controlp = m;
706 				so->so_rcv.sb_mb = m->m_next;
707 				m->m_next = 0;
708 				m = so->so_rcv.sb_mb;
709 			} else {
710 				MFREE(m, so->so_rcv.sb_mb);
711 				m = so->so_rcv.sb_mb;
712 			}
713 		}
714 		if (controlp) {
715 			orig_resid = 0;
716 			controlp = &(*controlp)->m_next;
717 		}
718 	}
719 	if (m) {
720 		if ((flags & MSG_PEEK) == 0)
721 			m->m_nextpkt = nextrecord;
722 		type = m->m_type;
723 		if (type == MT_OOBDATA)
724 			flags |= MSG_OOB;
725 	}
726 	moff = 0;
727 	offset = 0;
728 	while (m && uio->uio_resid > 0 && error == 0) {
729 		if (m->m_type == MT_OOBDATA) {
730 			if (type != MT_OOBDATA)
731 				break;
732 		} else if (type == MT_OOBDATA)
733 			break;
734 #ifdef DIAGNOSTIC
735 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
736 			panic("receive 3");
737 #endif
738 		so->so_state &= ~SS_RCVATMARK;
739 		len = uio->uio_resid;
740 		if (so->so_oobmark && len > so->so_oobmark - offset)
741 			len = so->so_oobmark - offset;
742 		if (len > m->m_len - moff)
743 			len = m->m_len - moff;
744 		/*
745 		 * If mp is set, just pass back the mbufs.
746 		 * Otherwise copy them out via the uio, then free.
747 		 * Sockbuf must be consistent here (points to current mbuf,
748 		 * it points to next record) when we drop priority;
749 		 * we must note any additions to the sockbuf when we
750 		 * block interrupts again.
751 		 */
752 		if (mp == 0) {
753 			splx(s);
754 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
755 			s = splsoftnet();
756 		} else
757 			uio->uio_resid -= len;
758 		if (len == m->m_len - moff) {
759 			if (m->m_flags & M_EOR)
760 				flags |= MSG_EOR;
761 			if (flags & MSG_PEEK) {
762 				m = m->m_next;
763 				moff = 0;
764 			} else {
765 				nextrecord = m->m_nextpkt;
766 				sbfree(&so->so_rcv, m);
767 				if (mp) {
768 					*mp = m;
769 					mp = &m->m_next;
770 					so->so_rcv.sb_mb = m = m->m_next;
771 					*mp = (struct mbuf *)0;
772 				} else {
773 					MFREE(m, so->so_rcv.sb_mb);
774 					m = so->so_rcv.sb_mb;
775 				}
776 				if (m)
777 					m->m_nextpkt = nextrecord;
778 			}
779 		} else {
780 			if (flags & MSG_PEEK)
781 				moff += len;
782 			else {
783 				if (mp)
784 					*mp = m_copym(m, 0, len, M_WAIT);
785 				m->m_data += len;
786 				m->m_len -= len;
787 				so->so_rcv.sb_cc -= len;
788 			}
789 		}
790 		if (so->so_oobmark) {
791 			if ((flags & MSG_PEEK) == 0) {
792 				so->so_oobmark -= len;
793 				if (so->so_oobmark == 0) {
794 					so->so_state |= SS_RCVATMARK;
795 					break;
796 				}
797 			} else {
798 				offset += len;
799 				if (offset == so->so_oobmark)
800 					break;
801 			}
802 		}
803 		if (flags & MSG_EOR)
804 			break;
805 		/*
806 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
807 		 * we must not quit until "uio->uio_resid == 0" or an error
808 		 * termination.  If a signal/timeout occurs, return
809 		 * with a short count but without error.
810 		 * Keep sockbuf locked against other readers.
811 		 */
812 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
813 		    !sosendallatonce(so) && !nextrecord) {
814 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
815 				break;
816 			error = sbwait(&so->so_rcv);
817 			if (error) {
818 				sbunlock(&so->so_rcv);
819 				splx(s);
820 				return (0);
821 			}
822 			if ((m = so->so_rcv.sb_mb) != NULL)
823 				nextrecord = m->m_nextpkt;
824 		}
825 	}
826 
827 	if (m && pr->pr_flags & PR_ATOMIC) {
828 		flags |= MSG_TRUNC;
829 		if ((flags & MSG_PEEK) == 0)
830 			(void) sbdroprecord(&so->so_rcv);
831 	}
832 	if ((flags & MSG_PEEK) == 0) {
833 		if (m == 0)
834 			so->so_rcv.sb_mb = nextrecord;
835 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
836 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
837 			    (struct mbuf *)(long)flags, (struct mbuf *)0,
838 			    (struct proc *)0);
839 	}
840 	if (orig_resid == uio->uio_resid && orig_resid &&
841 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
842 		sbunlock(&so->so_rcv);
843 		splx(s);
844 		goto restart;
845 	}
846 
847 	if (flagsp)
848 		*flagsp |= flags;
849 release:
850 	sbunlock(&so->so_rcv);
851 	splx(s);
852 	return (error);
853 }
854 
855 int
856 soshutdown(so, how)
857 	struct socket *so;
858 	int how;
859 {
860 	struct protosw *pr = so->so_proto;
861 
862 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
863 		return (EINVAL);
864 
865 	if (how == SHUT_RD || how == SHUT_RDWR)
866 		sorflush(so);
867 	if (how == SHUT_WR || how == SHUT_RDWR)
868 		return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0,
869 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
870 	return (0);
871 }
872 
873 void
874 sorflush(so)
875 	register struct socket *so;
876 {
877 	register struct sockbuf *sb = &so->so_rcv;
878 	register struct protosw *pr = so->so_proto;
879 	register int s;
880 	struct sockbuf asb;
881 
882 	sb->sb_flags |= SB_NOINTR;
883 	(void) sblock(sb, M_WAITOK);
884 	s = splimp();
885 	socantrcvmore(so);
886 	sbunlock(sb);
887 	asb = *sb;
888 	memset((caddr_t)sb, 0, sizeof(*sb));
889 	splx(s);
890 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
891 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
892 	sbrelease(&asb);
893 }
894 
895 int
896 sosetopt(so, level, optname, m0)
897 	register struct socket *so;
898 	int level, optname;
899 	struct mbuf *m0;
900 {
901 	int error = 0;
902 	register struct mbuf *m = m0;
903 
904 	if (level != SOL_SOCKET) {
905 		if (so->so_proto && so->so_proto->pr_ctloutput)
906 			return ((*so->so_proto->pr_ctloutput)
907 				  (PRCO_SETOPT, so, level, optname, &m0));
908 		error = ENOPROTOOPT;
909 	} else {
910 		switch (optname) {
911 
912 		case SO_LINGER:
913 			if (m == NULL || m->m_len != sizeof(struct linger)) {
914 				error = EINVAL;
915 				goto bad;
916 			}
917 			so->so_linger = mtod(m, struct linger *)->l_linger;
918 			/* fall thru... */
919 
920 		case SO_DEBUG:
921 		case SO_KEEPALIVE:
922 		case SO_DONTROUTE:
923 		case SO_USELOOPBACK:
924 		case SO_BROADCAST:
925 		case SO_REUSEADDR:
926 		case SO_REUSEPORT:
927 		case SO_OOBINLINE:
928 		case SO_TIMESTAMP:
929 			if (m == NULL || m->m_len < sizeof(int)) {
930 				error = EINVAL;
931 				goto bad;
932 			}
933 			if (*mtod(m, int *))
934 				so->so_options |= optname;
935 			else
936 				so->so_options &= ~optname;
937 			break;
938 
939 		case SO_SNDBUF:
940 		case SO_RCVBUF:
941 		case SO_SNDLOWAT:
942 		case SO_RCVLOWAT:
943 		    {
944 			int optval;
945 
946 			if (m == NULL || m->m_len < sizeof(int)) {
947 				error = EINVAL;
948 				goto bad;
949 			}
950 
951 			/*
952 			 * Values < 1 make no sense for any of these
953 			 * options, so disallow them.
954 			 */
955 			optval = *mtod(m, int *);
956 			if (optval < 1) {
957 				error = EINVAL;
958 				goto bad;
959 			}
960 
961 			switch (optname) {
962 
963 			case SO_SNDBUF:
964 			case SO_RCVBUF:
965 				if (sbreserve(optname == SO_SNDBUF ?
966 				    &so->so_snd : &so->so_rcv,
967 				    (u_long) optval) == 0) {
968 					error = ENOBUFS;
969 					goto bad;
970 				}
971 				break;
972 
973 			/*
974 			 * Make sure the low-water is never greater than
975 			 * the high-water.
976 			 */
977 			case SO_SNDLOWAT:
978 				so->so_snd.sb_lowat =
979 				    (optval > so->so_snd.sb_hiwat) ?
980 				    so->so_snd.sb_hiwat : optval;
981 				break;
982 			case SO_RCVLOWAT:
983 				so->so_rcv.sb_lowat =
984 				    (optval > so->so_rcv.sb_hiwat) ?
985 				    so->so_rcv.sb_hiwat : optval;
986 				break;
987 			}
988 			break;
989 		    }
990 
991 		case SO_SNDTIMEO:
992 		case SO_RCVTIMEO:
993 		    {
994 			struct timeval *tv;
995 			short val;
996 
997 			if (m == NULL || m->m_len < sizeof(*tv)) {
998 				error = EINVAL;
999 				goto bad;
1000 			}
1001 			tv = mtod(m, struct timeval *);
1002 			if (tv->tv_sec * hz + tv->tv_usec / tick > SHRT_MAX) {
1003 				error = EDOM;
1004 				goto bad;
1005 			}
1006 			val = tv->tv_sec * hz + tv->tv_usec / tick;
1007 
1008 			switch (optname) {
1009 
1010 			case SO_SNDTIMEO:
1011 				so->so_snd.sb_timeo = val;
1012 				break;
1013 			case SO_RCVTIMEO:
1014 				so->so_rcv.sb_timeo = val;
1015 				break;
1016 			}
1017 			break;
1018 		    }
1019 
1020 		default:
1021 			error = ENOPROTOOPT;
1022 			break;
1023 		}
1024 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1025 			(void) ((*so->so_proto->pr_ctloutput)
1026 				  (PRCO_SETOPT, so, level, optname, &m0));
1027 			m = NULL;	/* freed by protocol */
1028 		}
1029 	}
1030 bad:
1031 	if (m)
1032 		(void) m_free(m);
1033 	return (error);
1034 }
1035 
1036 int
1037 sogetopt(so, level, optname, mp)
1038 	register struct socket *so;
1039 	int level, optname;
1040 	struct mbuf **mp;
1041 {
1042 	register struct mbuf *m;
1043 
1044 	if (level != SOL_SOCKET) {
1045 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1046 			return ((*so->so_proto->pr_ctloutput)
1047 				  (PRCO_GETOPT, so, level, optname, mp));
1048 		} else
1049 			return (ENOPROTOOPT);
1050 	} else {
1051 		m = m_get(M_WAIT, MT_SOOPTS);
1052 		m->m_len = sizeof(int);
1053 
1054 		switch (optname) {
1055 
1056 		case SO_LINGER:
1057 			m->m_len = sizeof(struct linger);
1058 			mtod(m, struct linger *)->l_onoff =
1059 				so->so_options & SO_LINGER;
1060 			mtod(m, struct linger *)->l_linger = so->so_linger;
1061 			break;
1062 
1063 		case SO_USELOOPBACK:
1064 		case SO_DONTROUTE:
1065 		case SO_DEBUG:
1066 		case SO_KEEPALIVE:
1067 		case SO_REUSEADDR:
1068 		case SO_REUSEPORT:
1069 		case SO_BROADCAST:
1070 		case SO_OOBINLINE:
1071 		case SO_TIMESTAMP:
1072 			*mtod(m, int *) = so->so_options & optname;
1073 			break;
1074 
1075 		case SO_TYPE:
1076 			*mtod(m, int *) = so->so_type;
1077 			break;
1078 
1079 		case SO_ERROR:
1080 			*mtod(m, int *) = so->so_error;
1081 			so->so_error = 0;
1082 			break;
1083 
1084 		case SO_SNDBUF:
1085 			*mtod(m, int *) = so->so_snd.sb_hiwat;
1086 			break;
1087 
1088 		case SO_RCVBUF:
1089 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
1090 			break;
1091 
1092 		case SO_SNDLOWAT:
1093 			*mtod(m, int *) = so->so_snd.sb_lowat;
1094 			break;
1095 
1096 		case SO_RCVLOWAT:
1097 			*mtod(m, int *) = so->so_rcv.sb_lowat;
1098 			break;
1099 
1100 		case SO_SNDTIMEO:
1101 		case SO_RCVTIMEO:
1102 		    {
1103 			int val = (optname == SO_SNDTIMEO ?
1104 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1105 
1106 			m->m_len = sizeof(struct timeval);
1107 			mtod(m, struct timeval *)->tv_sec = val / hz;
1108 			mtod(m, struct timeval *)->tv_usec =
1109 			    (val % hz) * tick;
1110 			break;
1111 		    }
1112 
1113 		default:
1114 			(void)m_free(m);
1115 			return (ENOPROTOOPT);
1116 		}
1117 		*mp = m;
1118 		return (0);
1119 	}
1120 }
1121 
1122 void
1123 sohasoutofband(so)
1124 	register struct socket *so;
1125 {
1126 	struct proc *p;
1127 
1128 	if (so->so_pgid < 0)
1129 		gsignal(-so->so_pgid, SIGURG);
1130 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1131 		psignal(p, SIGURG);
1132 	selwakeup(&so->so_rcv.sb_sel);
1133 }
1134