xref: /netbsd-src/sys/kern/uipc_socket.c (revision bada23909e740596d0a3785a73bd3583a9807fb8)
1 /*	$NetBSD: uipc_socket.c,v 1.43 1999/01/21 22:09:10 mycroft 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 	error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
104 	    (struct mbuf *)(long)proto, (struct mbuf *)0, p);
105 	if (error) {
106 		so->so_state |= SS_NOFDREF;
107 		sofree(so);
108 		splx(s);
109 		return (error);
110 	}
111 #ifdef COMPAT_SUNOS
112 	{
113 		extern struct emul emul_sunos;
114 		if (p->p_emul == &emul_sunos && type == SOCK_DGRAM)
115 			so->so_options |= SO_BROADCAST;
116 	}
117 #endif
118 	splx(s);
119 	*aso = so;
120 	return (0);
121 }
122 
123 int
124 sobind(so, nam)
125 	struct socket *so;
126 	struct mbuf *nam;
127 {
128 	struct proc *p = curproc;		/* XXX */
129 	int s = splsoftnet();
130 	int error;
131 
132 	error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0,
133 	    nam, (struct mbuf *)0, p);
134 	splx(s);
135 	return (error);
136 }
137 
138 int
139 solisten(so, backlog)
140 	register struct socket *so;
141 	int backlog;
142 {
143 	int s = splsoftnet(), error;
144 
145 	error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0,
146 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
147 	if (error) {
148 		splx(s);
149 		return (error);
150 	}
151 	if (so->so_q.tqh_first == NULL)
152 		so->so_options |= SO_ACCEPTCONN;
153 	if (backlog < 0)
154 		backlog = 0;
155 	so->so_qlimit = min(backlog, SOMAXCONN);
156 	splx(s);
157 	return (0);
158 }
159 
160 void
161 sofree(so)
162 	register struct socket *so;
163 {
164 
165 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
166 		return;
167 	if (so->so_head) {
168 		/*
169 		 * We must not decommission a socket that's on the accept(2)
170 		 * queue.  If we do, then accept(2) may hang after select(2)
171 		 * indicated that the listening socket was ready.
172 		 */
173 		if (!soqremque(so, 0))
174 			return;
175 	}
176 	sbrelease(&so->so_snd);
177 	sorflush(so);
178 	pool_put(&socket_pool, so);
179 }
180 
181 /*
182  * Close a socket on last file table reference removal.
183  * Initiate disconnect if connected.
184  * Free socket when disconnect complete.
185  */
186 int
187 soclose(so)
188 	register struct socket *so;
189 {
190 	struct socket *so2;
191 	int s = splsoftnet();		/* conservative */
192 	int error = 0;
193 
194 	if (so->so_options & SO_ACCEPTCONN) {
195 		while ((so2 = so->so_q0.tqh_first) != 0) {
196 			(void) soqremque(so2, 0);
197 			(void) soabort(so2);
198 		}
199 		while ((so2 = so->so_q.tqh_first) != 0) {
200 			(void) soqremque(so2, 1);
201 			(void) soabort(so2);
202 		}
203 	}
204 	if (so->so_pcb == 0)
205 		goto discard;
206 	if (so->so_state & SS_ISCONNECTED) {
207 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
208 			error = sodisconnect(so);
209 			if (error)
210 				goto drop;
211 		}
212 		if (so->so_options & SO_LINGER) {
213 			if ((so->so_state & SS_ISDISCONNECTING) &&
214 			    (so->so_state & SS_NBIO))
215 				goto drop;
216 			while (so->so_state & SS_ISCONNECTED) {
217 				error = tsleep((caddr_t)&so->so_timeo,
218 					       PSOCK | PCATCH, netcls,
219 					       so->so_linger * hz);
220 				if (error)
221 					break;
222 			}
223 		}
224 	}
225 drop:
226 	if (so->so_pcb) {
227 		int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
228 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
229 		    (struct proc *)0);
230 		if (error == 0)
231 			error = error2;
232 	}
233 discard:
234 	if (so->so_state & SS_NOFDREF)
235 		panic("soclose: NOFDREF");
236 	so->so_state |= SS_NOFDREF;
237 	sofree(so);
238 	splx(s);
239 	return (error);
240 }
241 
242 /*
243  * Must be called at splsoftnet...
244  */
245 int
246 soabort(so)
247 	struct socket *so;
248 {
249 
250 	return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0,
251 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
252 }
253 
254 int
255 soaccept(so, nam)
256 	register struct socket *so;
257 	struct mbuf *nam;
258 {
259 	int s = splsoftnet();
260 	int error;
261 
262 	if ((so->so_state & SS_NOFDREF) == 0)
263 		panic("soaccept: !NOFDREF");
264 	so->so_state &= ~SS_NOFDREF;
265 	if ((so->so_state & SS_ISDISCONNECTED) == 0)
266 		error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
267 		    (struct mbuf *)0, nam, (struct mbuf *)0, (struct proc *)0);
268 	else
269 		error = 0;
270 	splx(s);
271 	return (error);
272 }
273 
274 int
275 soconnect(so, nam)
276 	register struct socket *so;
277 	struct mbuf *nam;
278 {
279 	struct proc *p = curproc;		/* XXX */
280 	int s;
281 	int error;
282 
283 	if (so->so_options & SO_ACCEPTCONN)
284 		return (EOPNOTSUPP);
285 	s = splsoftnet();
286 	/*
287 	 * If protocol is connection-based, can only connect once.
288 	 * Otherwise, if connected, try to disconnect first.
289 	 * This allows user to disconnect by connecting to, e.g.,
290 	 * a null address.
291 	 */
292 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
293 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
294 	    (error = sodisconnect(so))))
295 		error = EISCONN;
296 	else
297 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
298 		    (struct mbuf *)0, nam, (struct mbuf *)0, p);
299 	splx(s);
300 	return (error);
301 }
302 
303 int
304 soconnect2(so1, so2)
305 	register struct socket *so1;
306 	struct socket *so2;
307 {
308 	int s = splsoftnet();
309 	int error;
310 
311 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
312 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0,
313 	    (struct proc *)0);
314 	splx(s);
315 	return (error);
316 }
317 
318 int
319 sodisconnect(so)
320 	register struct socket *so;
321 {
322 	int s = splsoftnet();
323 	int error;
324 
325 	if ((so->so_state & SS_ISCONNECTED) == 0) {
326 		error = ENOTCONN;
327 		goto bad;
328 	}
329 	if (so->so_state & SS_ISDISCONNECTING) {
330 		error = EALREADY;
331 		goto bad;
332 	}
333 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
334 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
335 	    (struct proc *)0);
336 bad:
337 	splx(s);
338 	return (error);
339 }
340 
341 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
342 /*
343  * Send on a socket.
344  * If send must go all at once and message is larger than
345  * send buffering, then hard error.
346  * Lock against other senders.
347  * If must go all at once and not enough room now, then
348  * inform user that this would block and do nothing.
349  * Otherwise, if nonblocking, send as much as possible.
350  * The data to be sent is described by "uio" if nonzero,
351  * otherwise by the mbuf chain "top" (which must be null
352  * if uio is not).  Data provided in mbuf chain must be small
353  * enough to send all at once.
354  *
355  * Returns nonzero on error, timeout or signal; callers
356  * must check for short counts if EINTR/ERESTART are returned.
357  * Data and control buffers are freed on return.
358  */
359 int
360 sosend(so, addr, uio, top, control, flags)
361 	register struct socket *so;
362 	struct mbuf *addr;
363 	struct uio *uio;
364 	struct mbuf *top;
365 	struct mbuf *control;
366 	int flags;
367 {
368 	struct proc *p = curproc;		/* XXX */
369 	struct mbuf **mp;
370 	register struct mbuf *m;
371 	register long space, len, resid;
372 	int clen = 0, error, s, dontroute, mlen;
373 	int atomic = sosendallatonce(so) || top;
374 
375 	if (uio)
376 		resid = uio->uio_resid;
377 	else
378 		resid = top->m_pkthdr.len;
379 	/*
380 	 * In theory resid should be unsigned.
381 	 * However, space must be signed, as it might be less than 0
382 	 * if we over-committed, and we must use a signed comparison
383 	 * of space and resid.  On the other hand, a negative resid
384 	 * causes us to loop sending 0-length segments to the protocol.
385 	 */
386 	if (resid < 0) {
387 		error = EINVAL;
388 		goto out;
389 	}
390 	dontroute =
391 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
392 	    (so->so_proto->pr_flags & PR_ATOMIC);
393 	p->p_stats->p_ru.ru_msgsnd++;
394 	if (control)
395 		clen = control->m_len;
396 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
397 
398 restart:
399 	if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
400 		goto out;
401 	do {
402 		s = splsoftnet();
403 		if (so->so_state & SS_CANTSENDMORE)
404 			snderr(EPIPE);
405 		if (so->so_error)
406 			snderr(so->so_error);
407 		if ((so->so_state & SS_ISCONNECTED) == 0) {
408 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
409 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
410 				    !(resid == 0 && clen != 0))
411 					snderr(ENOTCONN);
412 			} else if (addr == 0)
413 				snderr(EDESTADDRREQ);
414 		}
415 		space = sbspace(&so->so_snd);
416 		if (flags & MSG_OOB)
417 			space += 1024;
418 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
419 		    clen > so->so_snd.sb_hiwat)
420 			snderr(EMSGSIZE);
421 		if (space < resid + clen && uio &&
422 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
423 			if (so->so_state & SS_NBIO)
424 				snderr(EWOULDBLOCK);
425 			sbunlock(&so->so_snd);
426 			error = sbwait(&so->so_snd);
427 			splx(s);
428 			if (error)
429 				goto out;
430 			goto restart;
431 		}
432 		splx(s);
433 		mp = &top;
434 		space -= clen;
435 		do {
436 		    if (uio == NULL) {
437 			/*
438 			 * Data is prepackaged in "top".
439 			 */
440 			resid = 0;
441 			if (flags & MSG_EOR)
442 				top->m_flags |= M_EOR;
443 		    } else do {
444 			if (top == 0) {
445 				MGETHDR(m, M_WAIT, MT_DATA);
446 				mlen = MHLEN;
447 				m->m_pkthdr.len = 0;
448 				m->m_pkthdr.rcvif = (struct ifnet *)0;
449 			} else {
450 				MGET(m, M_WAIT, MT_DATA);
451 				mlen = MLEN;
452 			}
453 			if (resid >= MINCLSIZE && space >= MCLBYTES) {
454 				MCLGET(m, M_WAIT);
455 				if ((m->m_flags & M_EXT) == 0)
456 					goto nopages;
457 				mlen = MCLBYTES;
458 #ifdef	MAPPED_MBUFS
459 				len = min(MCLBYTES, resid);
460 #else
461 				if (atomic && top == 0) {
462 					len = min(MCLBYTES - max_hdr, resid);
463 					m->m_data += max_hdr;
464 				} else
465 					len = min(MCLBYTES, resid);
466 #endif
467 				space -= len;
468 			} else {
469 nopages:
470 				len = min(min(mlen, resid), space);
471 				space -= len;
472 				/*
473 				 * For datagram protocols, leave room
474 				 * for protocol headers in first mbuf.
475 				 */
476 				if (atomic && top == 0 && len < mlen)
477 					MH_ALIGN(m, len);
478 			}
479 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
480 			resid = uio->uio_resid;
481 			m->m_len = len;
482 			*mp = m;
483 			top->m_pkthdr.len += len;
484 			if (error)
485 				goto release;
486 			mp = &m->m_next;
487 			if (resid <= 0) {
488 				if (flags & MSG_EOR)
489 					top->m_flags |= M_EOR;
490 				break;
491 			}
492 		    } while (space > 0 && atomic);
493 		    if (dontroute)
494 			    so->so_options |= SO_DONTROUTE;
495 		    if (resid > 0)
496 			    so->so_state |= SS_MORETOCOME;
497 		    s = splsoftnet();				/* XXX */
498 		    error = (*so->so_proto->pr_usrreq)(so,
499 			(flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
500 			top, addr, control, p);
501 		    splx(s);
502 		    if (dontroute)
503 			    so->so_options &= ~SO_DONTROUTE;
504 		    if (resid > 0)
505 			    so->so_state &= ~SS_MORETOCOME;
506 		    clen = 0;
507 		    control = 0;
508 		    top = 0;
509 		    mp = &top;
510 		    if (error)
511 			goto release;
512 		} while (resid && space > 0);
513 	} while (resid);
514 
515 release:
516 	sbunlock(&so->so_snd);
517 out:
518 	if (top)
519 		m_freem(top);
520 	if (control)
521 		m_freem(control);
522 	return (error);
523 }
524 
525 /*
526  * Implement receive operations on a socket.
527  * We depend on the way that records are added to the sockbuf
528  * by sbappend*.  In particular, each record (mbufs linked through m_next)
529  * must begin with an address if the protocol so specifies,
530  * followed by an optional mbuf or mbufs containing ancillary data,
531  * and then zero or more mbufs of data.
532  * In order to avoid blocking network interrupts for the entire time here,
533  * we splx() while doing the actual copy to user space.
534  * Although the sockbuf is locked, new data may still be appended,
535  * and thus we must maintain consistency of the sockbuf during that time.
536  *
537  * The caller may receive the data as a single mbuf chain by supplying
538  * an mbuf **mp0 for use in returning the chain.  The uio is then used
539  * only for the count in uio_resid.
540  */
541 int
542 soreceive(so, paddr, uio, mp0, controlp, flagsp)
543 	register struct socket *so;
544 	struct mbuf **paddr;
545 	struct uio *uio;
546 	struct mbuf **mp0;
547 	struct mbuf **controlp;
548 	int *flagsp;
549 {
550 	register struct mbuf *m, **mp;
551 	register int flags, len, error, s, offset;
552 	struct protosw *pr = so->so_proto;
553 	struct mbuf *nextrecord;
554 	int moff, type = 0;
555 	int orig_resid = uio->uio_resid;
556 
557 	mp = mp0;
558 	if (paddr)
559 		*paddr = 0;
560 	if (controlp)
561 		*controlp = 0;
562 	if (flagsp)
563 		flags = *flagsp &~ MSG_EOR;
564 	else
565 		flags = 0;
566 	if (flags & MSG_OOB) {
567 		m = m_get(M_WAIT, MT_DATA);
568 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
569 		    (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0,
570 		    (struct proc *)0);
571 		if (error)
572 			goto bad;
573 		do {
574 			error = uiomove(mtod(m, caddr_t),
575 			    (int) min(uio->uio_resid, m->m_len), uio);
576 			m = m_free(m);
577 		} while (uio->uio_resid && error == 0 && m);
578 bad:
579 		if (m)
580 			m_freem(m);
581 		return (error);
582 	}
583 	if (mp)
584 		*mp = (struct mbuf *)0;
585 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
586 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
587 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
588 
589 restart:
590 	if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
591 		return (error);
592 	s = splsoftnet();
593 
594 	m = so->so_rcv.sb_mb;
595 	/*
596 	 * If we have less data than requested, block awaiting more
597 	 * (subject to any timeout) if:
598 	 *   1. the current count is less than the low water mark,
599 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
600 	 *	receive operation at once if we block (resid <= hiwat), or
601 	 *   3. MSG_DONTWAIT is not set.
602 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
603 	 * we have to do the receive in sections, and thus risk returning
604 	 * a short count if a timeout or signal occurs after we start.
605 	 */
606 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
607 	    so->so_rcv.sb_cc < uio->uio_resid) &&
608 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
609 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
610 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
611 #ifdef DIAGNOSTIC
612 		if (m == 0 && so->so_rcv.sb_cc)
613 			panic("receive 1");
614 #endif
615 		if (so->so_error) {
616 			if (m)
617 				goto dontblock;
618 			error = so->so_error;
619 			if ((flags & MSG_PEEK) == 0)
620 				so->so_error = 0;
621 			goto release;
622 		}
623 		if (so->so_state & SS_CANTRCVMORE) {
624 			if (m)
625 				goto dontblock;
626 			else
627 				goto release;
628 		}
629 		for (; m; m = m->m_next)
630 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
631 				m = so->so_rcv.sb_mb;
632 				goto dontblock;
633 			}
634 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
635 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
636 			error = ENOTCONN;
637 			goto release;
638 		}
639 		if (uio->uio_resid == 0)
640 			goto release;
641 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
642 			error = EWOULDBLOCK;
643 			goto release;
644 		}
645 		sbunlock(&so->so_rcv);
646 		error = sbwait(&so->so_rcv);
647 		splx(s);
648 		if (error)
649 			return (error);
650 		goto restart;
651 	}
652 dontblock:
653 #ifdef notyet /* XXXX */
654 	if (uio->uio_procp)
655 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
656 #endif
657 	nextrecord = m->m_nextpkt;
658 	if (pr->pr_flags & PR_ADDR) {
659 #ifdef DIAGNOSTIC
660 		if (m->m_type != MT_SONAME)
661 			panic("receive 1a");
662 #endif
663 		orig_resid = 0;
664 		if (flags & MSG_PEEK) {
665 			if (paddr)
666 				*paddr = m_copy(m, 0, m->m_len);
667 			m = m->m_next;
668 		} else {
669 			sbfree(&so->so_rcv, m);
670 			if (paddr) {
671 				*paddr = m;
672 				so->so_rcv.sb_mb = m->m_next;
673 				m->m_next = 0;
674 				m = so->so_rcv.sb_mb;
675 			} else {
676 				MFREE(m, so->so_rcv.sb_mb);
677 				m = so->so_rcv.sb_mb;
678 			}
679 		}
680 	}
681 	while (m && m->m_type == MT_CONTROL && error == 0) {
682 		if (flags & MSG_PEEK) {
683 			if (controlp)
684 				*controlp = m_copy(m, 0, m->m_len);
685 			m = m->m_next;
686 		} else {
687 			sbfree(&so->so_rcv, m);
688 			if (controlp) {
689 				if (pr->pr_domain->dom_externalize &&
690 				    mtod(m, struct cmsghdr *)->cmsg_type ==
691 				    SCM_RIGHTS)
692 				   error = (*pr->pr_domain->dom_externalize)(m);
693 				*controlp = m;
694 				so->so_rcv.sb_mb = m->m_next;
695 				m->m_next = 0;
696 				m = so->so_rcv.sb_mb;
697 			} else {
698 				MFREE(m, so->so_rcv.sb_mb);
699 				m = so->so_rcv.sb_mb;
700 			}
701 		}
702 		if (controlp) {
703 			orig_resid = 0;
704 			controlp = &(*controlp)->m_next;
705 		}
706 	}
707 	if (m) {
708 		if ((flags & MSG_PEEK) == 0)
709 			m->m_nextpkt = nextrecord;
710 		type = m->m_type;
711 		if (type == MT_OOBDATA)
712 			flags |= MSG_OOB;
713 	}
714 	moff = 0;
715 	offset = 0;
716 	while (m && uio->uio_resid > 0 && error == 0) {
717 		if (m->m_type == MT_OOBDATA) {
718 			if (type != MT_OOBDATA)
719 				break;
720 		} else if (type == MT_OOBDATA)
721 			break;
722 #ifdef DIAGNOSTIC
723 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
724 			panic("receive 3");
725 #endif
726 		so->so_state &= ~SS_RCVATMARK;
727 		len = uio->uio_resid;
728 		if (so->so_oobmark && len > so->so_oobmark - offset)
729 			len = so->so_oobmark - offset;
730 		if (len > m->m_len - moff)
731 			len = m->m_len - moff;
732 		/*
733 		 * If mp is set, just pass back the mbufs.
734 		 * Otherwise copy them out via the uio, then free.
735 		 * Sockbuf must be consistent here (points to current mbuf,
736 		 * it points to next record) when we drop priority;
737 		 * we must note any additions to the sockbuf when we
738 		 * block interrupts again.
739 		 */
740 		if (mp == 0) {
741 			splx(s);
742 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
743 			s = splsoftnet();
744 		} else
745 			uio->uio_resid -= len;
746 		if (len == m->m_len - moff) {
747 			if (m->m_flags & M_EOR)
748 				flags |= MSG_EOR;
749 			if (flags & MSG_PEEK) {
750 				m = m->m_next;
751 				moff = 0;
752 			} else {
753 				nextrecord = m->m_nextpkt;
754 				sbfree(&so->so_rcv, m);
755 				if (mp) {
756 					*mp = m;
757 					mp = &m->m_next;
758 					so->so_rcv.sb_mb = m = m->m_next;
759 					*mp = (struct mbuf *)0;
760 				} else {
761 					MFREE(m, so->so_rcv.sb_mb);
762 					m = so->so_rcv.sb_mb;
763 				}
764 				if (m)
765 					m->m_nextpkt = nextrecord;
766 			}
767 		} else {
768 			if (flags & MSG_PEEK)
769 				moff += len;
770 			else {
771 				if (mp)
772 					*mp = m_copym(m, 0, len, M_WAIT);
773 				m->m_data += len;
774 				m->m_len -= len;
775 				so->so_rcv.sb_cc -= len;
776 			}
777 		}
778 		if (so->so_oobmark) {
779 			if ((flags & MSG_PEEK) == 0) {
780 				so->so_oobmark -= len;
781 				if (so->so_oobmark == 0) {
782 					so->so_state |= SS_RCVATMARK;
783 					break;
784 				}
785 			} else {
786 				offset += len;
787 				if (offset == so->so_oobmark)
788 					break;
789 			}
790 		}
791 		if (flags & MSG_EOR)
792 			break;
793 		/*
794 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
795 		 * we must not quit until "uio->uio_resid == 0" or an error
796 		 * termination.  If a signal/timeout occurs, return
797 		 * with a short count but without error.
798 		 * Keep sockbuf locked against other readers.
799 		 */
800 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
801 		    !sosendallatonce(so) && !nextrecord) {
802 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
803 				break;
804 			error = sbwait(&so->so_rcv);
805 			if (error) {
806 				sbunlock(&so->so_rcv);
807 				splx(s);
808 				return (0);
809 			}
810 			if ((m = so->so_rcv.sb_mb) != NULL)
811 				nextrecord = m->m_nextpkt;
812 		}
813 	}
814 
815 	if (m && pr->pr_flags & PR_ATOMIC) {
816 		flags |= MSG_TRUNC;
817 		if ((flags & MSG_PEEK) == 0)
818 			(void) sbdroprecord(&so->so_rcv);
819 	}
820 	if ((flags & MSG_PEEK) == 0) {
821 		if (m == 0)
822 			so->so_rcv.sb_mb = nextrecord;
823 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
824 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
825 			    (struct mbuf *)(long)flags, (struct mbuf *)0,
826 			    (struct proc *)0);
827 	}
828 	if (orig_resid == uio->uio_resid && orig_resid &&
829 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
830 		sbunlock(&so->so_rcv);
831 		splx(s);
832 		goto restart;
833 	}
834 
835 	if (flagsp)
836 		*flagsp |= flags;
837 release:
838 	sbunlock(&so->so_rcv);
839 	splx(s);
840 	return (error);
841 }
842 
843 int
844 soshutdown(so, how)
845 	struct socket *so;
846 	int how;
847 {
848 	struct protosw *pr = so->so_proto;
849 
850 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
851 		return (EINVAL);
852 
853 	if (how == SHUT_RD || how == SHUT_RDWR)
854 		sorflush(so);
855 	if (how == SHUT_WR || how == SHUT_RDWR)
856 		return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0,
857 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
858 	return (0);
859 }
860 
861 void
862 sorflush(so)
863 	register struct socket *so;
864 {
865 	register struct sockbuf *sb = &so->so_rcv;
866 	register struct protosw *pr = so->so_proto;
867 	register int s;
868 	struct sockbuf asb;
869 
870 	sb->sb_flags |= SB_NOINTR;
871 	(void) sblock(sb, M_WAITOK);
872 	s = splimp();
873 	socantrcvmore(so);
874 	sbunlock(sb);
875 	asb = *sb;
876 	memset((caddr_t)sb, 0, sizeof(*sb));
877 	splx(s);
878 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
879 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
880 	sbrelease(&asb);
881 }
882 
883 int
884 sosetopt(so, level, optname, m0)
885 	register struct socket *so;
886 	int level, optname;
887 	struct mbuf *m0;
888 {
889 	int error = 0;
890 	register struct mbuf *m = m0;
891 
892 	if (level != SOL_SOCKET) {
893 		if (so->so_proto && so->so_proto->pr_ctloutput)
894 			return ((*so->so_proto->pr_ctloutput)
895 				  (PRCO_SETOPT, so, level, optname, &m0));
896 		error = ENOPROTOOPT;
897 	} else {
898 		switch (optname) {
899 
900 		case SO_LINGER:
901 			if (m == NULL || m->m_len != sizeof(struct linger)) {
902 				error = EINVAL;
903 				goto bad;
904 			}
905 			so->so_linger = mtod(m, struct linger *)->l_linger;
906 			/* fall thru... */
907 
908 		case SO_DEBUG:
909 		case SO_KEEPALIVE:
910 		case SO_DONTROUTE:
911 		case SO_USELOOPBACK:
912 		case SO_BROADCAST:
913 		case SO_REUSEADDR:
914 		case SO_REUSEPORT:
915 		case SO_OOBINLINE:
916 		case SO_TIMESTAMP:
917 			if (m == NULL || m->m_len < sizeof(int)) {
918 				error = EINVAL;
919 				goto bad;
920 			}
921 			if (*mtod(m, int *))
922 				so->so_options |= optname;
923 			else
924 				so->so_options &= ~optname;
925 			break;
926 
927 		case SO_SNDBUF:
928 		case SO_RCVBUF:
929 		case SO_SNDLOWAT:
930 		case SO_RCVLOWAT:
931 		    {
932 			int optval;
933 
934 			if (m == NULL || m->m_len < sizeof(int)) {
935 				error = EINVAL;
936 				goto bad;
937 			}
938 
939 			/*
940 			 * Values < 1 make no sense for any of these
941 			 * options, so disallow them.
942 			 */
943 			optval = *mtod(m, int *);
944 			if (optval < 1) {
945 				error = EINVAL;
946 				goto bad;
947 			}
948 
949 			switch (optname) {
950 
951 			case SO_SNDBUF:
952 			case SO_RCVBUF:
953 				if (sbreserve(optname == SO_SNDBUF ?
954 				    &so->so_snd : &so->so_rcv,
955 				    (u_long) optval) == 0) {
956 					error = ENOBUFS;
957 					goto bad;
958 				}
959 				break;
960 
961 			/*
962 			 * Make sure the low-water is never greater than
963 			 * the high-water.
964 			 */
965 			case SO_SNDLOWAT:
966 				so->so_snd.sb_lowat =
967 				    (optval > so->so_snd.sb_hiwat) ?
968 				    so->so_snd.sb_hiwat : optval;
969 				break;
970 			case SO_RCVLOWAT:
971 				so->so_rcv.sb_lowat =
972 				    (optval > so->so_rcv.sb_hiwat) ?
973 				    so->so_rcv.sb_hiwat : optval;
974 				break;
975 			}
976 			break;
977 		    }
978 
979 		case SO_SNDTIMEO:
980 		case SO_RCVTIMEO:
981 		    {
982 			struct timeval *tv;
983 			short val;
984 
985 			if (m == NULL || m->m_len < sizeof(*tv)) {
986 				error = EINVAL;
987 				goto bad;
988 			}
989 			tv = mtod(m, struct timeval *);
990 			if (tv->tv_sec * hz + tv->tv_usec / tick > SHRT_MAX) {
991 				error = EDOM;
992 				goto bad;
993 			}
994 			val = tv->tv_sec * hz + tv->tv_usec / tick;
995 
996 			switch (optname) {
997 
998 			case SO_SNDTIMEO:
999 				so->so_snd.sb_timeo = val;
1000 				break;
1001 			case SO_RCVTIMEO:
1002 				so->so_rcv.sb_timeo = val;
1003 				break;
1004 			}
1005 			break;
1006 		    }
1007 
1008 		default:
1009 			error = ENOPROTOOPT;
1010 			break;
1011 		}
1012 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1013 			(void) ((*so->so_proto->pr_ctloutput)
1014 				  (PRCO_SETOPT, so, level, optname, &m0));
1015 			m = NULL;	/* freed by protocol */
1016 		}
1017 	}
1018 bad:
1019 	if (m)
1020 		(void) m_free(m);
1021 	return (error);
1022 }
1023 
1024 int
1025 sogetopt(so, level, optname, mp)
1026 	register struct socket *so;
1027 	int level, optname;
1028 	struct mbuf **mp;
1029 {
1030 	register struct mbuf *m;
1031 
1032 	if (level != SOL_SOCKET) {
1033 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1034 			return ((*so->so_proto->pr_ctloutput)
1035 				  (PRCO_GETOPT, so, level, optname, mp));
1036 		} else
1037 			return (ENOPROTOOPT);
1038 	} else {
1039 		m = m_get(M_WAIT, MT_SOOPTS);
1040 		m->m_len = sizeof(int);
1041 
1042 		switch (optname) {
1043 
1044 		case SO_LINGER:
1045 			m->m_len = sizeof(struct linger);
1046 			mtod(m, struct linger *)->l_onoff =
1047 				so->so_options & SO_LINGER;
1048 			mtod(m, struct linger *)->l_linger = so->so_linger;
1049 			break;
1050 
1051 		case SO_USELOOPBACK:
1052 		case SO_DONTROUTE:
1053 		case SO_DEBUG:
1054 		case SO_KEEPALIVE:
1055 		case SO_REUSEADDR:
1056 		case SO_REUSEPORT:
1057 		case SO_BROADCAST:
1058 		case SO_OOBINLINE:
1059 		case SO_TIMESTAMP:
1060 			*mtod(m, int *) = so->so_options & optname;
1061 			break;
1062 
1063 		case SO_TYPE:
1064 			*mtod(m, int *) = so->so_type;
1065 			break;
1066 
1067 		case SO_ERROR:
1068 			*mtod(m, int *) = so->so_error;
1069 			so->so_error = 0;
1070 			break;
1071 
1072 		case SO_SNDBUF:
1073 			*mtod(m, int *) = so->so_snd.sb_hiwat;
1074 			break;
1075 
1076 		case SO_RCVBUF:
1077 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
1078 			break;
1079 
1080 		case SO_SNDLOWAT:
1081 			*mtod(m, int *) = so->so_snd.sb_lowat;
1082 			break;
1083 
1084 		case SO_RCVLOWAT:
1085 			*mtod(m, int *) = so->so_rcv.sb_lowat;
1086 			break;
1087 
1088 		case SO_SNDTIMEO:
1089 		case SO_RCVTIMEO:
1090 		    {
1091 			int val = (optname == SO_SNDTIMEO ?
1092 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1093 
1094 			m->m_len = sizeof(struct timeval);
1095 			mtod(m, struct timeval *)->tv_sec = val / hz;
1096 			mtod(m, struct timeval *)->tv_usec =
1097 			    (val % hz) * tick;
1098 			break;
1099 		    }
1100 
1101 		default:
1102 			(void)m_free(m);
1103 			return (ENOPROTOOPT);
1104 		}
1105 		*mp = m;
1106 		return (0);
1107 	}
1108 }
1109 
1110 void
1111 sohasoutofband(so)
1112 	register struct socket *so;
1113 {
1114 	struct proc *p;
1115 
1116 	if (so->so_pgid < 0)
1117 		gsignal(-so->so_pgid, SIGURG);
1118 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1119 		psignal(p, SIGURG);
1120 	selwakeup(&so->so_rcv.sb_sel);
1121 }
1122