xref: /netbsd-src/sys/kern/uipc_socket.c (revision 4472dbe5e3bd91ef2540bada7a7ca7384627ff9b)
1 /*	$NetBSD: uipc_socket.c,v 1.50 2000/03/30 09:27:14 augustss 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 extern int somaxconn;			/* patchable (XXX sysctl) */
58 int somaxconn = SOMAXCONN;
59 
60 void
61 soinit()
62 {
63 
64 	pool_init(&socket_pool, sizeof(struct socket), 0, 0, 0,
65 	    "sockpl", 0, NULL, NULL, M_SOCKET);
66 }
67 
68 /*
69  * Socket operation routines.
70  * These routines are called by the routines in
71  * sys_socket.c or from a system process, and
72  * implement the semantics of socket operations by
73  * switching out to the protocol specific routines.
74  */
75 /*ARGSUSED*/
76 int
77 socreate(dom, aso, type, proto)
78 	int dom;
79 	struct socket **aso;
80 	int type;
81 	int proto;
82 {
83 	struct proc *p = curproc;		/* XXX */
84 	struct protosw *prp;
85 	struct socket *so;
86 	int error;
87 	int s;
88 
89 	if (proto)
90 		prp = pffindproto(dom, proto, type);
91 	else
92 		prp = pffindtype(dom, type);
93 	if (prp == 0 || prp->pr_usrreq == 0)
94 		return (EPROTONOSUPPORT);
95 	if (prp->pr_type != type)
96 		return (EPROTOTYPE);
97 	s = splsoftnet();
98 	so = pool_get(&socket_pool, PR_WAITOK);
99 	memset((caddr_t)so, 0, sizeof(*so));
100 	TAILQ_INIT(&so->so_q0);
101 	TAILQ_INIT(&so->so_q);
102 	so->so_type = type;
103 	so->so_proto = prp;
104 	so->so_send = sosend;
105 	so->so_receive = soreceive;
106 	if (p != 0)
107 		so->so_uid = p->p_ucred->cr_uid;
108 	error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
109 	    (struct mbuf *)(long)proto, (struct mbuf *)0, p);
110 	if (error) {
111 		so->so_state |= SS_NOFDREF;
112 		sofree(so);
113 		splx(s);
114 		return (error);
115 	}
116 #ifdef COMPAT_SUNOS
117 	{
118 		extern struct emul emul_sunos;
119 		if (p->p_emul == &emul_sunos && type == SOCK_DGRAM)
120 			so->so_options |= SO_BROADCAST;
121 	}
122 #endif
123 	splx(s);
124 	*aso = so;
125 	return (0);
126 }
127 
128 int
129 sobind(so, nam)
130 	struct socket *so;
131 	struct mbuf *nam;
132 {
133 	struct proc *p = curproc;		/* XXX */
134 	int s = splsoftnet();
135 	int error;
136 
137 	error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0,
138 	    nam, (struct mbuf *)0, p);
139 	splx(s);
140 	return (error);
141 }
142 
143 int
144 solisten(so, backlog)
145 	struct socket *so;
146 	int backlog;
147 {
148 	int s = splsoftnet(), error;
149 
150 	error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0,
151 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
152 	if (error) {
153 		splx(s);
154 		return (error);
155 	}
156 	if (so->so_q.tqh_first == NULL)
157 		so->so_options |= SO_ACCEPTCONN;
158 	if (backlog < 0)
159 		backlog = 0;
160 	so->so_qlimit = min(backlog, somaxconn);
161 	splx(s);
162 	return (0);
163 }
164 
165 void
166 sofree(so)
167 	struct socket *so;
168 {
169 
170 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
171 		return;
172 	if (so->so_head) {
173 		/*
174 		 * We must not decommission a socket that's on the accept(2)
175 		 * queue.  If we do, then accept(2) may hang after select(2)
176 		 * indicated that the listening socket was ready.
177 		 */
178 		if (!soqremque(so, 0))
179 			return;
180 	}
181 	sbrelease(&so->so_snd);
182 	sorflush(so);
183 	pool_put(&socket_pool, so);
184 }
185 
186 /*
187  * Close a socket on last file table reference removal.
188  * Initiate disconnect if connected.
189  * Free socket when disconnect complete.
190  */
191 int
192 soclose(so)
193 	struct socket *so;
194 {
195 	struct socket *so2;
196 	int s = splsoftnet();		/* conservative */
197 	int error = 0;
198 
199 	if (so->so_options & SO_ACCEPTCONN) {
200 		while ((so2 = so->so_q0.tqh_first) != 0) {
201 			(void) soqremque(so2, 0);
202 			(void) soabort(so2);
203 		}
204 		while ((so2 = so->so_q.tqh_first) != 0) {
205 			(void) soqremque(so2, 1);
206 			(void) soabort(so2);
207 		}
208 	}
209 	if (so->so_pcb == 0)
210 		goto discard;
211 	if (so->so_state & SS_ISCONNECTED) {
212 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
213 			error = sodisconnect(so);
214 			if (error)
215 				goto drop;
216 		}
217 		if (so->so_options & SO_LINGER) {
218 			if ((so->so_state & SS_ISDISCONNECTING) &&
219 			    (so->so_state & SS_NBIO))
220 				goto drop;
221 			while (so->so_state & SS_ISCONNECTED) {
222 				error = tsleep((caddr_t)&so->so_timeo,
223 					       PSOCK | PCATCH, netcls,
224 					       so->so_linger * hz);
225 				if (error)
226 					break;
227 			}
228 		}
229 	}
230 drop:
231 	if (so->so_pcb) {
232 		int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
233 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
234 		    (struct proc *)0);
235 		if (error == 0)
236 			error = error2;
237 	}
238 discard:
239 	if (so->so_state & SS_NOFDREF)
240 		panic("soclose: NOFDREF");
241 	so->so_state |= SS_NOFDREF;
242 	sofree(so);
243 	splx(s);
244 	return (error);
245 }
246 
247 /*
248  * Must be called at splsoftnet...
249  */
250 int
251 soabort(so)
252 	struct socket *so;
253 {
254 
255 	return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0,
256 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
257 }
258 
259 int
260 soaccept(so, nam)
261 	struct socket *so;
262 	struct mbuf *nam;
263 {
264 	int s = splsoftnet();
265 	int error;
266 
267 	if ((so->so_state & SS_NOFDREF) == 0)
268 		panic("soaccept: !NOFDREF");
269 	so->so_state &= ~SS_NOFDREF;
270 	if ((so->so_state & SS_ISDISCONNECTED) == 0)
271 		error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
272 		    (struct mbuf *)0, nam, (struct mbuf *)0, (struct proc *)0);
273 	else
274 		error = 0;
275 	splx(s);
276 	return (error);
277 }
278 
279 int
280 soconnect(so, nam)
281 	struct socket *so;
282 	struct mbuf *nam;
283 {
284 	struct proc *p = curproc;		/* XXX */
285 	int s;
286 	int error;
287 
288 	if (so->so_options & SO_ACCEPTCONN)
289 		return (EOPNOTSUPP);
290 	s = splsoftnet();
291 	/*
292 	 * If protocol is connection-based, can only connect once.
293 	 * Otherwise, if connected, try to disconnect first.
294 	 * This allows user to disconnect by connecting to, e.g.,
295 	 * a null address.
296 	 */
297 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
298 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
299 	    (error = sodisconnect(so))))
300 		error = EISCONN;
301 	else
302 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
303 		    (struct mbuf *)0, nam, (struct mbuf *)0, p);
304 	splx(s);
305 	return (error);
306 }
307 
308 int
309 soconnect2(so1, so2)
310 	struct socket *so1;
311 	struct socket *so2;
312 {
313 	int s = splsoftnet();
314 	int error;
315 
316 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
317 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0,
318 	    (struct proc *)0);
319 	splx(s);
320 	return (error);
321 }
322 
323 int
324 sodisconnect(so)
325 	struct socket *so;
326 {
327 	int s = splsoftnet();
328 	int error;
329 
330 	if ((so->so_state & SS_ISCONNECTED) == 0) {
331 		error = ENOTCONN;
332 		goto bad;
333 	}
334 	if (so->so_state & SS_ISDISCONNECTING) {
335 		error = EALREADY;
336 		goto bad;
337 	}
338 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
339 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
340 	    (struct proc *)0);
341 bad:
342 	splx(s);
343 	return (error);
344 }
345 
346 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
347 /*
348  * Send on a socket.
349  * If send must go all at once and message is larger than
350  * send buffering, then hard error.
351  * Lock against other senders.
352  * If must go all at once and not enough room now, then
353  * inform user that this would block and do nothing.
354  * Otherwise, if nonblocking, send as much as possible.
355  * The data to be sent is described by "uio" if nonzero,
356  * otherwise by the mbuf chain "top" (which must be null
357  * if uio is not).  Data provided in mbuf chain must be small
358  * enough to send all at once.
359  *
360  * Returns nonzero on error, timeout or signal; callers
361  * must check for short counts if EINTR/ERESTART are returned.
362  * Data and control buffers are freed on return.
363  */
364 int
365 sosend(so, addr, uio, top, control, flags)
366 	struct socket *so;
367 	struct mbuf *addr;
368 	struct uio *uio;
369 	struct mbuf *top;
370 	struct mbuf *control;
371 	int flags;
372 {
373 	struct proc *p = curproc;		/* XXX */
374 	struct mbuf **mp;
375 	struct mbuf *m;
376 	long space, len, resid;
377 	int clen = 0, error, s, dontroute, mlen;
378 	int atomic = sosendallatonce(so) || top;
379 
380 	if (uio)
381 		resid = uio->uio_resid;
382 	else
383 		resid = top->m_pkthdr.len;
384 	/*
385 	 * In theory resid should be unsigned.
386 	 * However, space must be signed, as it might be less than 0
387 	 * if we over-committed, and we must use a signed comparison
388 	 * of space and resid.  On the other hand, a negative resid
389 	 * causes us to loop sending 0-length segments to the protocol.
390 	 */
391 	if (resid < 0) {
392 		error = EINVAL;
393 		goto out;
394 	}
395 	dontroute =
396 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
397 	    (so->so_proto->pr_flags & PR_ATOMIC);
398 	p->p_stats->p_ru.ru_msgsnd++;
399 	if (control)
400 		clen = control->m_len;
401 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
402 
403 restart:
404 	if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
405 		goto out;
406 	do {
407 		s = splsoftnet();
408 		if (so->so_state & SS_CANTSENDMORE)
409 			snderr(EPIPE);
410 		if (so->so_error) {
411 			error = so->so_error;
412 			so->so_error = 0;
413 			splx(s);
414 			goto release;
415 		}
416 		if ((so->so_state & SS_ISCONNECTED) == 0) {
417 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
418 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
419 				    !(resid == 0 && clen != 0))
420 					snderr(ENOTCONN);
421 			} else if (addr == 0)
422 				snderr(EDESTADDRREQ);
423 		}
424 		space = sbspace(&so->so_snd);
425 		if (flags & MSG_OOB)
426 			space += 1024;
427 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
428 		    clen > so->so_snd.sb_hiwat)
429 			snderr(EMSGSIZE);
430 		if (space < resid + clen && uio &&
431 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
432 			if (so->so_state & SS_NBIO)
433 				snderr(EWOULDBLOCK);
434 			sbunlock(&so->so_snd);
435 			error = sbwait(&so->so_snd);
436 			splx(s);
437 			if (error)
438 				goto out;
439 			goto restart;
440 		}
441 		splx(s);
442 		mp = &top;
443 		space -= clen;
444 		do {
445 			if (uio == NULL) {
446 				/*
447 				 * Data is prepackaged in "top".
448 				 */
449 				resid = 0;
450 				if (flags & MSG_EOR)
451 					top->m_flags |= M_EOR;
452 			} else do {
453 				if (top == 0) {
454 					MGETHDR(m, M_WAIT, MT_DATA);
455 					mlen = MHLEN;
456 					m->m_pkthdr.len = 0;
457 					m->m_pkthdr.rcvif = (struct ifnet *)0;
458 				} else {
459 					MGET(m, M_WAIT, MT_DATA);
460 					mlen = MLEN;
461 				}
462 				if (resid >= MINCLSIZE && space >= MCLBYTES) {
463 					MCLGET(m, M_WAIT);
464 					if ((m->m_flags & M_EXT) == 0)
465 						goto nopages;
466 					mlen = MCLBYTES;
467 #ifdef	MAPPED_MBUFS
468 					len = min(MCLBYTES, resid);
469 #else
470 					if (atomic && top == 0) {
471 						len = min(MCLBYTES - max_hdr, resid);
472 						m->m_data += max_hdr;
473 					} else
474 						len = min(MCLBYTES, resid);
475 #endif
476 					space -= len;
477 				} else {
478 nopages:
479 					len = min(min(mlen, resid), space);
480 					space -= len;
481 					/*
482 					 * For datagram protocols, leave room
483 					 * for protocol headers in first mbuf.
484 					 */
485 					if (atomic && top == 0 && len < mlen)
486 						MH_ALIGN(m, len);
487 				}
488 				error = uiomove(mtod(m, caddr_t), (int)len, uio);
489 				resid = uio->uio_resid;
490 				m->m_len = len;
491 				*mp = m;
492 				top->m_pkthdr.len += len;
493 				if (error)
494 					goto release;
495 				mp = &m->m_next;
496 				if (resid <= 0) {
497 					if (flags & MSG_EOR)
498 						top->m_flags |= M_EOR;
499 					break;
500 				}
501 			} while (space > 0 && atomic);
502 
503 			s = splsoftnet();
504 
505 			if (so->so_state & SS_CANTSENDMORE)
506 				snderr(EPIPE);
507 
508 			if (dontroute)
509 				so->so_options |= SO_DONTROUTE;
510 			if (resid > 0)
511 				so->so_state |= SS_MORETOCOME;
512 			error = (*so->so_proto->pr_usrreq)(so,
513 			    (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
514 			    top, addr, control, p);
515 			if (dontroute)
516 				so->so_options &= ~SO_DONTROUTE;
517 			if (resid > 0)
518 				so->so_state &= ~SS_MORETOCOME;
519 			splx(s);
520 
521 			clen = 0;
522 			control = 0;
523 			top = 0;
524 			mp = &top;
525 			if (error)
526 				goto release;
527 		} while (resid && space > 0);
528 	} while (resid);
529 
530 release:
531 	sbunlock(&so->so_snd);
532 out:
533 	if (top)
534 		m_freem(top);
535 	if (control)
536 		m_freem(control);
537 	return (error);
538 }
539 
540 /*
541  * Implement receive operations on a socket.
542  * We depend on the way that records are added to the sockbuf
543  * by sbappend*.  In particular, each record (mbufs linked through m_next)
544  * must begin with an address if the protocol so specifies,
545  * followed by an optional mbuf or mbufs containing ancillary data,
546  * and then zero or more mbufs of data.
547  * In order to avoid blocking network interrupts for the entire time here,
548  * we splx() while doing the actual copy to user space.
549  * Although the sockbuf is locked, new data may still be appended,
550  * and thus we must maintain consistency of the sockbuf during that time.
551  *
552  * The caller may receive the data as a single mbuf chain by supplying
553  * an mbuf **mp0 for use in returning the chain.  The uio is then used
554  * only for the count in uio_resid.
555  */
556 int
557 soreceive(so, paddr, uio, mp0, controlp, flagsp)
558 	struct socket *so;
559 	struct mbuf **paddr;
560 	struct uio *uio;
561 	struct mbuf **mp0;
562 	struct mbuf **controlp;
563 	int *flagsp;
564 {
565 	struct mbuf *m, **mp;
566 	int flags, len, error, s, offset;
567 	struct protosw *pr = so->so_proto;
568 	struct mbuf *nextrecord;
569 	int moff, type = 0;
570 	int orig_resid = uio->uio_resid;
571 
572 	mp = mp0;
573 	if (paddr)
574 		*paddr = 0;
575 	if (controlp)
576 		*controlp = 0;
577 	if (flagsp)
578 		flags = *flagsp &~ MSG_EOR;
579 	else
580 		flags = 0;
581 	if (flags & MSG_OOB) {
582 		m = m_get(M_WAIT, MT_DATA);
583 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
584 		    (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0,
585 		    (struct proc *)0);
586 		if (error)
587 			goto bad;
588 		do {
589 			error = uiomove(mtod(m, caddr_t),
590 			    (int) min(uio->uio_resid, m->m_len), uio);
591 			m = m_free(m);
592 		} while (uio->uio_resid && error == 0 && m);
593 bad:
594 		if (m)
595 			m_freem(m);
596 		return (error);
597 	}
598 	if (mp)
599 		*mp = (struct mbuf *)0;
600 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
601 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
602 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
603 
604 restart:
605 	if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
606 		return (error);
607 	s = splsoftnet();
608 
609 	m = so->so_rcv.sb_mb;
610 	/*
611 	 * If we have less data than requested, block awaiting more
612 	 * (subject to any timeout) if:
613 	 *   1. the current count is less than the low water mark,
614 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
615 	 *	receive operation at once if we block (resid <= hiwat), or
616 	 *   3. MSG_DONTWAIT is not set.
617 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
618 	 * we have to do the receive in sections, and thus risk returning
619 	 * a short count if a timeout or signal occurs after we start.
620 	 */
621 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
622 	    so->so_rcv.sb_cc < uio->uio_resid) &&
623 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
624 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
625 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
626 #ifdef DIAGNOSTIC
627 		if (m == 0 && so->so_rcv.sb_cc)
628 			panic("receive 1");
629 #endif
630 		if (so->so_error) {
631 			if (m)
632 				goto dontblock;
633 			error = so->so_error;
634 			if ((flags & MSG_PEEK) == 0)
635 				so->so_error = 0;
636 			goto release;
637 		}
638 		if (so->so_state & SS_CANTRCVMORE) {
639 			if (m)
640 				goto dontblock;
641 			else
642 				goto release;
643 		}
644 		for (; m; m = m->m_next)
645 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
646 				m = so->so_rcv.sb_mb;
647 				goto dontblock;
648 			}
649 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
650 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
651 			error = ENOTCONN;
652 			goto release;
653 		}
654 		if (uio->uio_resid == 0)
655 			goto release;
656 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
657 			error = EWOULDBLOCK;
658 			goto release;
659 		}
660 		sbunlock(&so->so_rcv);
661 		error = sbwait(&so->so_rcv);
662 		splx(s);
663 		if (error)
664 			return (error);
665 		goto restart;
666 	}
667 dontblock:
668 #ifdef notyet /* XXXX */
669 	if (uio->uio_procp)
670 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
671 #endif
672 	nextrecord = m->m_nextpkt;
673 	if (pr->pr_flags & PR_ADDR) {
674 #ifdef DIAGNOSTIC
675 		if (m->m_type != MT_SONAME)
676 			panic("receive 1a");
677 #endif
678 		orig_resid = 0;
679 		if (flags & MSG_PEEK) {
680 			if (paddr)
681 				*paddr = m_copy(m, 0, m->m_len);
682 			m = m->m_next;
683 		} else {
684 			sbfree(&so->so_rcv, m);
685 			if (paddr) {
686 				*paddr = m;
687 				so->so_rcv.sb_mb = m->m_next;
688 				m->m_next = 0;
689 				m = so->so_rcv.sb_mb;
690 			} else {
691 				MFREE(m, so->so_rcv.sb_mb);
692 				m = so->so_rcv.sb_mb;
693 			}
694 		}
695 	}
696 	while (m && m->m_type == MT_CONTROL && error == 0) {
697 		if (flags & MSG_PEEK) {
698 			if (controlp)
699 				*controlp = m_copy(m, 0, m->m_len);
700 			m = m->m_next;
701 		} else {
702 			sbfree(&so->so_rcv, m);
703 			if (controlp) {
704 				if (pr->pr_domain->dom_externalize &&
705 				    mtod(m, struct cmsghdr *)->cmsg_type ==
706 				    SCM_RIGHTS)
707 					error = (*pr->pr_domain->dom_externalize)(m);
708 				*controlp = m;
709 				so->so_rcv.sb_mb = m->m_next;
710 				m->m_next = 0;
711 				m = so->so_rcv.sb_mb;
712 			} else {
713 				MFREE(m, so->so_rcv.sb_mb);
714 				m = so->so_rcv.sb_mb;
715 			}
716 		}
717 		if (controlp) {
718 			orig_resid = 0;
719 			controlp = &(*controlp)->m_next;
720 		}
721 	}
722 	if (m) {
723 		if ((flags & MSG_PEEK) == 0)
724 			m->m_nextpkt = nextrecord;
725 		type = m->m_type;
726 		if (type == MT_OOBDATA)
727 			flags |= MSG_OOB;
728 	}
729 	moff = 0;
730 	offset = 0;
731 	while (m && uio->uio_resid > 0 && error == 0) {
732 		if (m->m_type == MT_OOBDATA) {
733 			if (type != MT_OOBDATA)
734 				break;
735 		} else if (type == MT_OOBDATA)
736 			break;
737 #ifdef DIAGNOSTIC
738 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
739 			panic("receive 3");
740 #endif
741 		so->so_state &= ~SS_RCVATMARK;
742 		len = uio->uio_resid;
743 		if (so->so_oobmark && len > so->so_oobmark - offset)
744 			len = so->so_oobmark - offset;
745 		if (len > m->m_len - moff)
746 			len = m->m_len - moff;
747 		/*
748 		 * If mp is set, just pass back the mbufs.
749 		 * Otherwise copy them out via the uio, then free.
750 		 * Sockbuf must be consistent here (points to current mbuf,
751 		 * it points to next record) when we drop priority;
752 		 * we must note any additions to the sockbuf when we
753 		 * block interrupts again.
754 		 */
755 		if (mp == 0) {
756 			splx(s);
757 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
758 			s = splsoftnet();
759 		} else
760 			uio->uio_resid -= len;
761 		if (len == m->m_len - moff) {
762 			if (m->m_flags & M_EOR)
763 				flags |= MSG_EOR;
764 			if (flags & MSG_PEEK) {
765 				m = m->m_next;
766 				moff = 0;
767 			} else {
768 				nextrecord = m->m_nextpkt;
769 				sbfree(&so->so_rcv, m);
770 				if (mp) {
771 					*mp = m;
772 					mp = &m->m_next;
773 					so->so_rcv.sb_mb = m = m->m_next;
774 					*mp = (struct mbuf *)0;
775 				} else {
776 					MFREE(m, so->so_rcv.sb_mb);
777 					m = so->so_rcv.sb_mb;
778 				}
779 				if (m)
780 					m->m_nextpkt = nextrecord;
781 			}
782 		} else {
783 			if (flags & MSG_PEEK)
784 				moff += len;
785 			else {
786 				if (mp)
787 					*mp = m_copym(m, 0, len, M_WAIT);
788 				m->m_data += len;
789 				m->m_len -= len;
790 				so->so_rcv.sb_cc -= len;
791 			}
792 		}
793 		if (so->so_oobmark) {
794 			if ((flags & MSG_PEEK) == 0) {
795 				so->so_oobmark -= len;
796 				if (so->so_oobmark == 0) {
797 					so->so_state |= SS_RCVATMARK;
798 					break;
799 				}
800 			} else {
801 				offset += len;
802 				if (offset == so->so_oobmark)
803 					break;
804 			}
805 		}
806 		if (flags & MSG_EOR)
807 			break;
808 		/*
809 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
810 		 * we must not quit until "uio->uio_resid == 0" or an error
811 		 * termination.  If a signal/timeout occurs, return
812 		 * with a short count but without error.
813 		 * Keep sockbuf locked against other readers.
814 		 */
815 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
816 		    !sosendallatonce(so) && !nextrecord) {
817 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
818 				break;
819 			error = sbwait(&so->so_rcv);
820 			if (error) {
821 				sbunlock(&so->so_rcv);
822 				splx(s);
823 				return (0);
824 			}
825 			if ((m = so->so_rcv.sb_mb) != NULL)
826 				nextrecord = m->m_nextpkt;
827 		}
828 	}
829 
830 	if (m && pr->pr_flags & PR_ATOMIC) {
831 		flags |= MSG_TRUNC;
832 		if ((flags & MSG_PEEK) == 0)
833 			(void) sbdroprecord(&so->so_rcv);
834 	}
835 	if ((flags & MSG_PEEK) == 0) {
836 		if (m == 0)
837 			so->so_rcv.sb_mb = nextrecord;
838 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
839 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
840 			    (struct mbuf *)(long)flags, (struct mbuf *)0,
841 			    (struct proc *)0);
842 	}
843 	if (orig_resid == uio->uio_resid && orig_resid &&
844 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
845 		sbunlock(&so->so_rcv);
846 		splx(s);
847 		goto restart;
848 	}
849 
850 	if (flagsp)
851 		*flagsp |= flags;
852 release:
853 	sbunlock(&so->so_rcv);
854 	splx(s);
855 	return (error);
856 }
857 
858 int
859 soshutdown(so, how)
860 	struct socket *so;
861 	int how;
862 {
863 	struct protosw *pr = so->so_proto;
864 
865 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
866 		return (EINVAL);
867 
868 	if (how == SHUT_RD || how == SHUT_RDWR)
869 		sorflush(so);
870 	if (how == SHUT_WR || how == SHUT_RDWR)
871 		return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0,
872 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
873 	return (0);
874 }
875 
876 void
877 sorflush(so)
878 	struct socket *so;
879 {
880 	struct sockbuf *sb = &so->so_rcv;
881 	struct protosw *pr = so->so_proto;
882 	int s;
883 	struct sockbuf asb;
884 
885 	sb->sb_flags |= SB_NOINTR;
886 	(void) sblock(sb, M_WAITOK);
887 	s = splimp();
888 	socantrcvmore(so);
889 	sbunlock(sb);
890 	asb = *sb;
891 	memset((caddr_t)sb, 0, sizeof(*sb));
892 	splx(s);
893 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
894 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
895 	sbrelease(&asb);
896 }
897 
898 int
899 sosetopt(so, level, optname, m0)
900 	struct socket *so;
901 	int level, optname;
902 	struct mbuf *m0;
903 {
904 	int error = 0;
905 	struct mbuf *m = m0;
906 
907 	if (level != SOL_SOCKET) {
908 		if (so->so_proto && so->so_proto->pr_ctloutput)
909 			return ((*so->so_proto->pr_ctloutput)
910 				  (PRCO_SETOPT, so, level, optname, &m0));
911 		error = ENOPROTOOPT;
912 	} else {
913 		switch (optname) {
914 
915 		case SO_LINGER:
916 			if (m == NULL || m->m_len != sizeof(struct linger)) {
917 				error = EINVAL;
918 				goto bad;
919 			}
920 			so->so_linger = mtod(m, struct linger *)->l_linger;
921 			/* fall thru... */
922 
923 		case SO_DEBUG:
924 		case SO_KEEPALIVE:
925 		case SO_DONTROUTE:
926 		case SO_USELOOPBACK:
927 		case SO_BROADCAST:
928 		case SO_REUSEADDR:
929 		case SO_REUSEPORT:
930 		case SO_OOBINLINE:
931 		case SO_TIMESTAMP:
932 			if (m == NULL || m->m_len < sizeof(int)) {
933 				error = EINVAL;
934 				goto bad;
935 			}
936 			if (*mtod(m, int *))
937 				so->so_options |= optname;
938 			else
939 				so->so_options &= ~optname;
940 			break;
941 
942 		case SO_SNDBUF:
943 		case SO_RCVBUF:
944 		case SO_SNDLOWAT:
945 		case SO_RCVLOWAT:
946 		    {
947 			int optval;
948 
949 			if (m == NULL || m->m_len < sizeof(int)) {
950 				error = EINVAL;
951 				goto bad;
952 			}
953 
954 			/*
955 			 * Values < 1 make no sense for any of these
956 			 * options, so disallow them.
957 			 */
958 			optval = *mtod(m, int *);
959 			if (optval < 1) {
960 				error = EINVAL;
961 				goto bad;
962 			}
963 
964 			switch (optname) {
965 
966 			case SO_SNDBUF:
967 			case SO_RCVBUF:
968 				if (sbreserve(optname == SO_SNDBUF ?
969 				    &so->so_snd : &so->so_rcv,
970 				    (u_long) optval) == 0) {
971 					error = ENOBUFS;
972 					goto bad;
973 				}
974 				break;
975 
976 			/*
977 			 * Make sure the low-water is never greater than
978 			 * the high-water.
979 			 */
980 			case SO_SNDLOWAT:
981 				so->so_snd.sb_lowat =
982 				    (optval > so->so_snd.sb_hiwat) ?
983 				    so->so_snd.sb_hiwat : optval;
984 				break;
985 			case SO_RCVLOWAT:
986 				so->so_rcv.sb_lowat =
987 				    (optval > so->so_rcv.sb_hiwat) ?
988 				    so->so_rcv.sb_hiwat : optval;
989 				break;
990 			}
991 			break;
992 		    }
993 
994 		case SO_SNDTIMEO:
995 		case SO_RCVTIMEO:
996 		    {
997 			struct timeval *tv;
998 			short val;
999 
1000 			if (m == NULL || m->m_len < sizeof(*tv)) {
1001 				error = EINVAL;
1002 				goto bad;
1003 			}
1004 			tv = mtod(m, struct timeval *);
1005 			if (tv->tv_sec * hz + tv->tv_usec / tick > SHRT_MAX) {
1006 				error = EDOM;
1007 				goto bad;
1008 			}
1009 			val = tv->tv_sec * hz + tv->tv_usec / tick;
1010 
1011 			switch (optname) {
1012 
1013 			case SO_SNDTIMEO:
1014 				so->so_snd.sb_timeo = val;
1015 				break;
1016 			case SO_RCVTIMEO:
1017 				so->so_rcv.sb_timeo = val;
1018 				break;
1019 			}
1020 			break;
1021 		    }
1022 
1023 		default:
1024 			error = ENOPROTOOPT;
1025 			break;
1026 		}
1027 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1028 			(void) ((*so->so_proto->pr_ctloutput)
1029 				  (PRCO_SETOPT, so, level, optname, &m0));
1030 			m = NULL;	/* freed by protocol */
1031 		}
1032 	}
1033 bad:
1034 	if (m)
1035 		(void) m_free(m);
1036 	return (error);
1037 }
1038 
1039 int
1040 sogetopt(so, level, optname, mp)
1041 	struct socket *so;
1042 	int level, optname;
1043 	struct mbuf **mp;
1044 {
1045 	struct mbuf *m;
1046 
1047 	if (level != SOL_SOCKET) {
1048 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1049 			return ((*so->so_proto->pr_ctloutput)
1050 				  (PRCO_GETOPT, so, level, optname, mp));
1051 		} else
1052 			return (ENOPROTOOPT);
1053 	} else {
1054 		m = m_get(M_WAIT, MT_SOOPTS);
1055 		m->m_len = sizeof(int);
1056 
1057 		switch (optname) {
1058 
1059 		case SO_LINGER:
1060 			m->m_len = sizeof(struct linger);
1061 			mtod(m, struct linger *)->l_onoff =
1062 				so->so_options & SO_LINGER;
1063 			mtod(m, struct linger *)->l_linger = so->so_linger;
1064 			break;
1065 
1066 		case SO_USELOOPBACK:
1067 		case SO_DONTROUTE:
1068 		case SO_DEBUG:
1069 		case SO_KEEPALIVE:
1070 		case SO_REUSEADDR:
1071 		case SO_REUSEPORT:
1072 		case SO_BROADCAST:
1073 		case SO_OOBINLINE:
1074 		case SO_TIMESTAMP:
1075 			*mtod(m, int *) = so->so_options & optname;
1076 			break;
1077 
1078 		case SO_TYPE:
1079 			*mtod(m, int *) = so->so_type;
1080 			break;
1081 
1082 		case SO_ERROR:
1083 			*mtod(m, int *) = so->so_error;
1084 			so->so_error = 0;
1085 			break;
1086 
1087 		case SO_SNDBUF:
1088 			*mtod(m, int *) = so->so_snd.sb_hiwat;
1089 			break;
1090 
1091 		case SO_RCVBUF:
1092 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
1093 			break;
1094 
1095 		case SO_SNDLOWAT:
1096 			*mtod(m, int *) = so->so_snd.sb_lowat;
1097 			break;
1098 
1099 		case SO_RCVLOWAT:
1100 			*mtod(m, int *) = so->so_rcv.sb_lowat;
1101 			break;
1102 
1103 		case SO_SNDTIMEO:
1104 		case SO_RCVTIMEO:
1105 		    {
1106 			int val = (optname == SO_SNDTIMEO ?
1107 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1108 
1109 			m->m_len = sizeof(struct timeval);
1110 			mtod(m, struct timeval *)->tv_sec = val / hz;
1111 			mtod(m, struct timeval *)->tv_usec =
1112 			    (val % hz) * tick;
1113 			break;
1114 		    }
1115 
1116 		default:
1117 			(void)m_free(m);
1118 			return (ENOPROTOOPT);
1119 		}
1120 		*mp = m;
1121 		return (0);
1122 	}
1123 }
1124 
1125 void
1126 sohasoutofband(so)
1127 	struct socket *so;
1128 {
1129 	struct proc *p;
1130 
1131 	if (so->so_pgid < 0)
1132 		gsignal(-so->so_pgid, SIGURG);
1133 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1134 		psignal(p, SIGURG);
1135 	selwakeup(&so->so_rcv.sb_sel);
1136 }
1137