xref: /netbsd-src/sys/kern/uipc_socket.c (revision d9158b13b5dfe46201430699a3f7a235ecf28df3)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	from: @(#)uipc_socket.c	8.3 (Berkeley) 4/15/94
34  *	$Id: uipc_socket.c,v 1.15 1994/05/13 06:01:37 mycroft Exp $
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/proc.h>
40 #include <sys/file.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/domain.h>
44 #include <sys/kernel.h>
45 #include <sys/protosw.h>
46 #include <sys/socket.h>
47 #include <sys/socketvar.h>
48 #include <sys/resourcevar.h>
49 
50 /*
51  * Socket operation routines.
52  * These routines are called by the routines in
53  * sys_socket.c or from a system process, and
54  * implement the semantics of socket operations by
55  * switching out to the protocol specific routines.
56  */
57 /*ARGSUSED*/
58 int
59 socreate(dom, aso, type, proto)
60 	int dom;
61 	struct socket **aso;
62 	register int type;
63 	int proto;
64 {
65 	struct proc *p = curproc;		/* XXX */
66 	register struct protosw *prp;
67 	register struct socket *so;
68 	register int error;
69 
70 	if (proto)
71 		prp = pffindproto(dom, proto, type);
72 	else
73 		prp = pffindtype(dom, type);
74 	if (prp == 0 || prp->pr_usrreq == 0)
75 		return (EPROTONOSUPPORT);
76 	if (prp->pr_type != type)
77 		return (EPROTOTYPE);
78 	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_WAIT);
79 	bzero((caddr_t)so, sizeof(*so));
80 	so->so_type = type;
81 	if (p->p_ucred->cr_uid == 0)
82 		so->so_state = SS_PRIV;
83 	so->so_proto = prp;
84 	error =
85 	    (*prp->pr_usrreq)(so, PRU_ATTACH,
86 		(struct mbuf *)0, (struct mbuf *)proto, (struct mbuf *)0);
87 	if (error) {
88 		so->so_state |= SS_NOFDREF;
89 		sofree(so);
90 		return (error);
91 	}
92 #ifdef COMPAT_SUNOS
93 	if (p->p_emul == EMUL_SUNOS && type == SOCK_DGRAM)
94 		so->so_options |= SO_BROADCAST;
95 #endif
96 	*aso = so;
97 	return (0);
98 }
99 
100 int
101 sobind(so, nam)
102 	struct socket *so;
103 	struct mbuf *nam;
104 {
105 	int s = splnet();
106 	int error;
107 
108 	error =
109 	    (*so->so_proto->pr_usrreq)(so, PRU_BIND,
110 		(struct mbuf *)0, nam, (struct mbuf *)0);
111 	splx(s);
112 	return (error);
113 }
114 
115 int
116 solisten(so, backlog)
117 	register struct socket *so;
118 	int backlog;
119 {
120 	int s = splnet(), error;
121 
122 	error =
123 	    (*so->so_proto->pr_usrreq)(so, PRU_LISTEN,
124 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
125 	if (error) {
126 		splx(s);
127 		return (error);
128 	}
129 	if (so->so_q == 0)
130 		so->so_options |= SO_ACCEPTCONN;
131 	if (backlog < 0)
132 		backlog = 0;
133 	so->so_qlimit = min(backlog, SOMAXCONN);
134 	splx(s);
135 	return (0);
136 }
137 
138 int
139 sofree(so)
140 	register struct socket *so;
141 {
142 
143 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
144 		return;
145 	if (so->so_head) {
146 		if (!soqremque(so, 0) && !soqremque(so, 1))
147 			panic("sofree dq");
148 		so->so_head = 0;
149 	}
150 	sbrelease(&so->so_snd);
151 	sorflush(so);
152 	FREE(so, M_SOCKET);
153 }
154 
155 /*
156  * Close a socket on last file table reference removal.
157  * Initiate disconnect if connected.
158  * Free socket when disconnect complete.
159  */
160 int
161 soclose(so)
162 	register struct socket *so;
163 {
164 	int s = splnet();		/* conservative */
165 	int error = 0;
166 
167 	if (so->so_options & SO_ACCEPTCONN) {
168 		while (so->so_q0)
169 			(void) soabort(so->so_q0);
170 		while (so->so_q)
171 			(void) soabort(so->so_q);
172 	}
173 	if (so->so_pcb == 0)
174 		goto discard;
175 	if (so->so_state & SS_ISCONNECTED) {
176 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
177 			error = sodisconnect(so);
178 			if (error)
179 				goto drop;
180 		}
181 		if (so->so_options & SO_LINGER) {
182 			if ((so->so_state & SS_ISDISCONNECTING) &&
183 			    (so->so_state & SS_NBIO))
184 				goto drop;
185 			while (so->so_state & SS_ISCONNECTED)
186 				if (error = tsleep((caddr_t)&so->so_timeo,
187 				    PSOCK | PCATCH, netcls, so->so_linger))
188 					break;
189 		}
190 	}
191 drop:
192 	if (so->so_pcb) {
193 		int error2 =
194 		    (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
195 			(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
196 		if (error == 0)
197 			error = error2;
198 	}
199 discard:
200 	if (so->so_state & SS_NOFDREF)
201 		panic("soclose: NOFDREF");
202 	so->so_state |= SS_NOFDREF;
203 	sofree(so);
204 	splx(s);
205 	return (error);
206 }
207 
208 /*
209  * Must be called at splnet...
210  */
211 int
212 soabort(so)
213 	struct socket *so;
214 {
215 
216 	return (
217 	    (*so->so_proto->pr_usrreq)(so, PRU_ABORT,
218 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
219 }
220 
221 int
222 soaccept(so, nam)
223 	register struct socket *so;
224 	struct mbuf *nam;
225 {
226 	int s = splnet();
227 	int error;
228 
229 	if ((so->so_state & SS_NOFDREF) == 0)
230 		panic("soaccept: !NOFDREF");
231 	so->so_state &= ~SS_NOFDREF;
232 	error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
233 	    (struct mbuf *)0, nam, (struct mbuf *)0);
234 	splx(s);
235 	return (error);
236 }
237 
238 int
239 soconnect(so, nam)
240 	register struct socket *so;
241 	struct mbuf *nam;
242 {
243 	int s;
244 	int error;
245 
246 	if (so->so_options & SO_ACCEPTCONN)
247 		return (EOPNOTSUPP);
248 	s = splnet();
249 	/*
250 	 * If protocol is connection-based, can only connect once.
251 	 * Otherwise, if connected, try to disconnect first.
252 	 * This allows user to disconnect by connecting to, e.g.,
253 	 * a null address.
254 	 */
255 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
256 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
257 	    (error = sodisconnect(so))))
258 		error = EISCONN;
259 	else
260 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
261 		    (struct mbuf *)0, nam, (struct mbuf *)0);
262 	splx(s);
263 	return (error);
264 }
265 
266 int
267 soconnect2(so1, so2)
268 	register struct socket *so1;
269 	struct socket *so2;
270 {
271 	int s = splnet();
272 	int error;
273 
274 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
275 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0);
276 	splx(s);
277 	return (error);
278 }
279 
280 int
281 sodisconnect(so)
282 	register struct socket *so;
283 {
284 	int s = splnet();
285 	int error;
286 
287 	if ((so->so_state & SS_ISCONNECTED) == 0) {
288 		error = ENOTCONN;
289 		goto bad;
290 	}
291 	if (so->so_state & SS_ISDISCONNECTING) {
292 		error = EALREADY;
293 		goto bad;
294 	}
295 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
296 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
297 bad:
298 	splx(s);
299 	return (error);
300 }
301 
302 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
303 /*
304  * Send on a socket.
305  * If send must go all at once and message is larger than
306  * send buffering, then hard error.
307  * Lock against other senders.
308  * If must go all at once and not enough room now, then
309  * inform user that this would block and do nothing.
310  * Otherwise, if nonblocking, send as much as possible.
311  * The data to be sent is described by "uio" if nonzero,
312  * otherwise by the mbuf chain "top" (which must be null
313  * if uio is not).  Data provided in mbuf chain must be small
314  * enough to send all at once.
315  *
316  * Returns nonzero on error, timeout or signal; callers
317  * must check for short counts if EINTR/ERESTART are returned.
318  * Data and control buffers are freed on return.
319  */
320 int
321 sosend(so, addr, uio, top, control, flags)
322 	register struct socket *so;
323 	struct mbuf *addr;
324 	struct uio *uio;
325 	struct mbuf *top;
326 	struct mbuf *control;
327 	int flags;
328 {
329 	struct proc *p = curproc;		/* XXX */
330 	struct mbuf **mp;
331 	register struct mbuf *m;
332 	register long space, len, resid;
333 	int clen = 0, error, s, dontroute, mlen;
334 	int atomic = sosendallatonce(so) || top;
335 
336 	if (uio)
337 		resid = uio->uio_resid;
338 	else
339 		resid = top->m_pkthdr.len;
340 	/*
341 	 * In theory resid should be unsigned.
342 	 * However, space must be signed, as it might be less than 0
343 	 * if we over-committed, and we must use a signed comparison
344 	 * of space and resid.  On the other hand, a negative resid
345 	 * causes us to loop sending 0-length segments to the protocol.
346 	 */
347 	if (resid < 0)
348 		return (EINVAL);
349 	dontroute =
350 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
351 	    (so->so_proto->pr_flags & PR_ATOMIC);
352 	p->p_stats->p_ru.ru_msgsnd++;
353 	if (control)
354 		clen = control->m_len;
355 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
356 
357 restart:
358 	if (error = sblock(&so->so_snd, SBLOCKWAIT(flags)))
359 		goto out;
360 	do {
361 		s = splnet();
362 		if (so->so_state & SS_CANTSENDMORE)
363 			snderr(EPIPE);
364 		if (so->so_error)
365 			snderr(so->so_error);
366 		if ((so->so_state & SS_ISCONNECTED) == 0) {
367 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
368 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
369 				    !(resid == 0 && clen != 0))
370 					snderr(ENOTCONN);
371 			} else if (addr == 0)
372 				snderr(EDESTADDRREQ);
373 		}
374 		space = sbspace(&so->so_snd);
375 		if (flags & MSG_OOB)
376 			space += 1024;
377 		if (atomic && resid > so->so_snd.sb_hiwat ||
378 		    clen > so->so_snd.sb_hiwat)
379 			snderr(EMSGSIZE);
380 		if (space < resid + clen && uio &&
381 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
382 			if (so->so_state & SS_NBIO)
383 				snderr(EWOULDBLOCK);
384 			sbunlock(&so->so_snd);
385 			error = sbwait(&so->so_snd);
386 			splx(s);
387 			if (error)
388 				goto out;
389 			goto restart;
390 		}
391 		splx(s);
392 		mp = &top;
393 		space -= clen;
394 		do {
395 		    if (uio == NULL) {
396 			/*
397 			 * Data is prepackaged in "top".
398 			 */
399 			resid = 0;
400 			if (flags & MSG_EOR)
401 				top->m_flags |= M_EOR;
402 		    } else do {
403 			if (top == 0) {
404 				MGETHDR(m, M_WAIT, MT_DATA);
405 				mlen = MHLEN;
406 				m->m_pkthdr.len = 0;
407 				m->m_pkthdr.rcvif = (struct ifnet *)0;
408 			} else {
409 				MGET(m, M_WAIT, MT_DATA);
410 				mlen = MLEN;
411 			}
412 			if (resid >= MINCLSIZE && space >= MCLBYTES) {
413 				MCLGET(m, M_WAIT);
414 				if ((m->m_flags & M_EXT) == 0)
415 					goto nopages;
416 				mlen = MCLBYTES;
417 #ifdef	MAPPED_MBUFS
418 				len = min(MCLBYTES, resid);
419 #else
420 				if (atomic && top == 0) {
421 					len = min(MCLBYTES - max_hdr, resid);
422 					m->m_data += max_hdr;
423 				} else
424 					len = min(MCLBYTES, resid);
425 #endif
426 				space -= MCLBYTES;
427 			} else {
428 nopages:
429 				len = min(min(mlen, resid), space);
430 				space -= len;
431 				/*
432 				 * For datagram protocols, leave room
433 				 * for protocol headers in first mbuf.
434 				 */
435 				if (atomic && top == 0 && len < mlen)
436 					MH_ALIGN(m, len);
437 			}
438 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
439 			resid = uio->uio_resid;
440 			m->m_len = len;
441 			*mp = m;
442 			top->m_pkthdr.len += len;
443 			if (error)
444 				goto release;
445 			mp = &m->m_next;
446 			if (resid <= 0) {
447 				if (flags & MSG_EOR)
448 					top->m_flags |= M_EOR;
449 				break;
450 			}
451 		    } while (space > 0 && atomic);
452 		    if (dontroute)
453 			    so->so_options |= SO_DONTROUTE;
454 		    s = splnet();				/* XXX */
455 		    error = (*so->so_proto->pr_usrreq)(so,
456 			(flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
457 			top, addr, control);
458 		    splx(s);
459 		    if (dontroute)
460 			    so->so_options &= ~SO_DONTROUTE;
461 		    clen = 0;
462 		    control = 0;
463 		    top = 0;
464 		    mp = &top;
465 		    if (error)
466 			goto release;
467 		} while (resid && space > 0);
468 	} while (resid);
469 
470 release:
471 	sbunlock(&so->so_snd);
472 out:
473 	if (top)
474 		m_freem(top);
475 	if (control)
476 		m_freem(control);
477 	return (error);
478 }
479 
480 /*
481  * Implement receive operations on a socket.
482  * We depend on the way that records are added to the sockbuf
483  * by sbappend*.  In particular, each record (mbufs linked through m_next)
484  * must begin with an address if the protocol so specifies,
485  * followed by an optional mbuf or mbufs containing ancillary data,
486  * and then zero or more mbufs of data.
487  * In order to avoid blocking network interrupts for the entire time here,
488  * we splx() while doing the actual copy to user space.
489  * Although the sockbuf is locked, new data may still be appended,
490  * and thus we must maintain consistency of the sockbuf during that time.
491  *
492  * The caller may receive the data as a single mbuf chain by supplying
493  * an mbuf **mp0 for use in returning the chain.  The uio is then used
494  * only for the count in uio_resid.
495  */
496 int
497 soreceive(so, paddr, uio, mp0, controlp, flagsp)
498 	register struct socket *so;
499 	struct mbuf **paddr;
500 	struct uio *uio;
501 	struct mbuf **mp0;
502 	struct mbuf **controlp;
503 	int *flagsp;
504 {
505 	register struct mbuf *m, **mp;
506 	register int flags, len, error, s, offset;
507 	struct protosw *pr = so->so_proto;
508 	struct mbuf *nextrecord;
509 	int moff, type;
510 	int orig_resid = uio->uio_resid;
511 
512 	mp = mp0;
513 	if (paddr)
514 		*paddr = 0;
515 	if (controlp)
516 		*controlp = 0;
517 	if (flagsp)
518 		flags = *flagsp &~ MSG_EOR;
519 	else
520 		flags = 0;
521 	if (flags & MSG_OOB) {
522 		m = m_get(M_WAIT, MT_DATA);
523 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB,
524 		    m, (struct mbuf *)(flags & MSG_PEEK), (struct mbuf *)0);
525 		if (error)
526 			goto bad;
527 		do {
528 			error = uiomove(mtod(m, caddr_t),
529 			    (int) min(uio->uio_resid, m->m_len), uio);
530 			m = m_free(m);
531 		} while (uio->uio_resid && error == 0 && m);
532 bad:
533 		if (m)
534 			m_freem(m);
535 		return (error);
536 	}
537 	if (mp)
538 		*mp = (struct mbuf *)0;
539 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
540 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
541 		    (struct mbuf *)0, (struct mbuf *)0);
542 
543 restart:
544 	if (error = sblock(&so->so_rcv, SBLOCKWAIT(flags)))
545 		return (error);
546 	s = splnet();
547 
548 	m = so->so_rcv.sb_mb;
549 	/*
550 	 * If we have less data than requested, block awaiting more
551 	 * (subject to any timeout) if:
552 	 *   1. the current count is less than the low water mark,
553 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
554 	 *	receive operation at once if we block (resid <= hiwat), or
555 	 *   3. MSG_DONTWAIT is not set.
556 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
557 	 * we have to do the receive in sections, and thus risk returning
558 	 * a short count if a timeout or signal occurs after we start.
559 	 */
560 	if (m == 0 || ((flags & MSG_DONTWAIT) == 0 &&
561 	    so->so_rcv.sb_cc < uio->uio_resid) &&
562 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
563 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
564 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0) {
565 #ifdef DIAGNOSTIC
566 		if (m == 0 && so->so_rcv.sb_cc)
567 			panic("receive 1");
568 #endif
569 		if (so->so_error) {
570 			if (m)
571 				goto dontblock;
572 			error = so->so_error;
573 			if ((flags & MSG_PEEK) == 0)
574 				so->so_error = 0;
575 			goto release;
576 		}
577 		if (so->so_state & SS_CANTRCVMORE) {
578 			if (m)
579 				goto dontblock;
580 			else
581 				goto release;
582 		}
583 		for (; m; m = m->m_next)
584 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
585 				m = so->so_rcv.sb_mb;
586 				goto dontblock;
587 			}
588 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
589 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
590 			error = ENOTCONN;
591 			goto release;
592 		}
593 		if (uio->uio_resid == 0)
594 			goto release;
595 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
596 			error = EWOULDBLOCK;
597 			goto release;
598 		}
599 		sbunlock(&so->so_rcv);
600 		error = sbwait(&so->so_rcv);
601 		splx(s);
602 		if (error)
603 			return (error);
604 		goto restart;
605 	}
606 dontblock:
607 #ifdef notyet /* XXXX */
608 	if (uio->uio_procp)
609 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
610 #endif
611 	nextrecord = m->m_nextpkt;
612 	if (pr->pr_flags & PR_ADDR) {
613 #ifdef DIAGNOSTIC
614 		if (m->m_type != MT_SONAME)
615 			panic("receive 1a");
616 #endif
617 		orig_resid = 0;
618 		if (flags & MSG_PEEK) {
619 			if (paddr)
620 				*paddr = m_copy(m, 0, m->m_len);
621 			m = m->m_next;
622 		} else {
623 			sbfree(&so->so_rcv, m);
624 			if (paddr) {
625 				*paddr = m;
626 				so->so_rcv.sb_mb = m->m_next;
627 				m->m_next = 0;
628 				m = so->so_rcv.sb_mb;
629 			} else {
630 				MFREE(m, so->so_rcv.sb_mb);
631 				m = so->so_rcv.sb_mb;
632 			}
633 		}
634 	}
635 	while (m && m->m_type == MT_CONTROL && error == 0) {
636 		if (flags & MSG_PEEK) {
637 			if (controlp)
638 				*controlp = m_copy(m, 0, m->m_len);
639 			m = m->m_next;
640 		} else {
641 			sbfree(&so->so_rcv, m);
642 			if (controlp) {
643 				if (pr->pr_domain->dom_externalize &&
644 				    mtod(m, struct cmsghdr *)->cmsg_type ==
645 				    SCM_RIGHTS)
646 				   error = (*pr->pr_domain->dom_externalize)(m);
647 				*controlp = m;
648 				so->so_rcv.sb_mb = m->m_next;
649 				m->m_next = 0;
650 				m = so->so_rcv.sb_mb;
651 			} else {
652 				MFREE(m, so->so_rcv.sb_mb);
653 				m = so->so_rcv.sb_mb;
654 			}
655 		}
656 		if (controlp) {
657 			orig_resid = 0;
658 			controlp = &(*controlp)->m_next;
659 		}
660 	}
661 	if (m) {
662 		if ((flags & MSG_PEEK) == 0)
663 			m->m_nextpkt = nextrecord;
664 		type = m->m_type;
665 		if (type == MT_OOBDATA)
666 			flags |= MSG_OOB;
667 	}
668 	moff = 0;
669 	offset = 0;
670 	while (m && uio->uio_resid > 0 && error == 0) {
671 		if (m->m_type == MT_OOBDATA) {
672 			if (type != MT_OOBDATA)
673 				break;
674 		} else if (type == MT_OOBDATA)
675 			break;
676 #ifdef DIAGNOSTIC
677 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
678 			panic("receive 3");
679 #endif
680 		so->so_state &= ~SS_RCVATMARK;
681 		len = uio->uio_resid;
682 		if (so->so_oobmark && len > so->so_oobmark - offset)
683 			len = so->so_oobmark - offset;
684 		if (len > m->m_len - moff)
685 			len = m->m_len - moff;
686 		/*
687 		 * If mp is set, just pass back the mbufs.
688 		 * Otherwise copy them out via the uio, then free.
689 		 * Sockbuf must be consistent here (points to current mbuf,
690 		 * it points to next record) when we drop priority;
691 		 * we must note any additions to the sockbuf when we
692 		 * block interrupts again.
693 		 */
694 		if (mp == 0) {
695 			splx(s);
696 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
697 			s = splnet();
698 		} else
699 			uio->uio_resid -= len;
700 		if (len == m->m_len - moff) {
701 			if (m->m_flags & M_EOR)
702 				flags |= MSG_EOR;
703 			if (flags & MSG_PEEK) {
704 				m = m->m_next;
705 				moff = 0;
706 			} else {
707 				nextrecord = m->m_nextpkt;
708 				sbfree(&so->so_rcv, m);
709 				if (mp) {
710 					*mp = m;
711 					mp = &m->m_next;
712 					so->so_rcv.sb_mb = m = m->m_next;
713 					*mp = (struct mbuf *)0;
714 				} else {
715 					MFREE(m, so->so_rcv.sb_mb);
716 					m = so->so_rcv.sb_mb;
717 				}
718 				if (m)
719 					m->m_nextpkt = nextrecord;
720 			}
721 		} else {
722 			if (flags & MSG_PEEK)
723 				moff += len;
724 			else {
725 				if (mp)
726 					*mp = m_copym(m, 0, len, M_WAIT);
727 				m->m_data += len;
728 				m->m_len -= len;
729 				so->so_rcv.sb_cc -= len;
730 			}
731 		}
732 		if (so->so_oobmark) {
733 			if ((flags & MSG_PEEK) == 0) {
734 				so->so_oobmark -= len;
735 				if (so->so_oobmark == 0) {
736 					so->so_state |= SS_RCVATMARK;
737 					break;
738 				}
739 			} else {
740 				offset += len;
741 				if (offset == so->so_oobmark)
742 					break;
743 			}
744 		}
745 		if (flags & MSG_EOR)
746 			break;
747 		/*
748 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
749 		 * we must not quit until "uio->uio_resid == 0" or an error
750 		 * termination.  If a signal/timeout occurs, return
751 		 * with a short count but without error.
752 		 * Keep sockbuf locked against other readers.
753 		 */
754 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
755 		    !sosendallatonce(so) && !nextrecord) {
756 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
757 				break;
758 			error = sbwait(&so->so_rcv);
759 			if (error) {
760 				sbunlock(&so->so_rcv);
761 				splx(s);
762 				return (0);
763 			}
764 			if (m = so->so_rcv.sb_mb)
765 				nextrecord = m->m_nextpkt;
766 		}
767 	}
768 
769 	if (m && pr->pr_flags & PR_ATOMIC) {
770 		flags |= MSG_TRUNC;
771 		if ((flags & MSG_PEEK) == 0)
772 			(void) sbdroprecord(&so->so_rcv);
773 	}
774 	if ((flags & MSG_PEEK) == 0) {
775 		if (m == 0)
776 			so->so_rcv.sb_mb = nextrecord;
777 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
778 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
779 			    (struct mbuf *)flags, (struct mbuf *)0,
780 			    (struct mbuf *)0);
781 	}
782 	if (orig_resid == uio->uio_resid && orig_resid &&
783 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
784 		sbunlock(&so->so_rcv);
785 		splx(s);
786 		goto restart;
787 	}
788 
789 	if (flagsp)
790 		*flagsp |= flags;
791 release:
792 	sbunlock(&so->so_rcv);
793 	splx(s);
794 	return (error);
795 }
796 
797 int
798 soshutdown(so, how)
799 	register struct socket *so;
800 	register int how;
801 {
802 	register struct protosw *pr = so->so_proto;
803 
804 	how++;
805 	if (how & FREAD)
806 		sorflush(so);
807 	if (how & FWRITE)
808 		return ((*pr->pr_usrreq)(so, PRU_SHUTDOWN,
809 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
810 	return (0);
811 }
812 
813 void
814 sorflush(so)
815 	register struct socket *so;
816 {
817 	register struct sockbuf *sb = &so->so_rcv;
818 	register struct protosw *pr = so->so_proto;
819 	register int s;
820 	struct sockbuf asb;
821 
822 	sb->sb_flags |= SB_NOINTR;
823 	(void) sblock(sb, M_WAITOK);
824 	s = splimp();
825 	socantrcvmore(so);
826 	sbunlock(sb);
827 	asb = *sb;
828 	bzero((caddr_t)sb, sizeof (*sb));
829 	splx(s);
830 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
831 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
832 	sbrelease(&asb);
833 }
834 
835 int
836 sosetopt(so, level, optname, m0)
837 	register struct socket *so;
838 	int level, optname;
839 	struct mbuf *m0;
840 {
841 	int error = 0;
842 	register struct mbuf *m = m0;
843 
844 	if (level != SOL_SOCKET) {
845 		if (so->so_proto && so->so_proto->pr_ctloutput)
846 			return ((*so->so_proto->pr_ctloutput)
847 				  (PRCO_SETOPT, so, level, optname, &m0));
848 		error = ENOPROTOOPT;
849 	} else {
850 		switch (optname) {
851 
852 		case SO_LINGER:
853 			if (m == NULL || m->m_len != sizeof (struct linger)) {
854 				error = EINVAL;
855 				goto bad;
856 			}
857 			so->so_linger = mtod(m, struct linger *)->l_linger;
858 			/* fall thru... */
859 
860 		case SO_DEBUG:
861 		case SO_KEEPALIVE:
862 		case SO_DONTROUTE:
863 		case SO_USELOOPBACK:
864 		case SO_BROADCAST:
865 		case SO_REUSEADDR:
866 		case SO_REUSEPORT:
867 		case SO_OOBINLINE:
868 			if (m == NULL || m->m_len < sizeof (int)) {
869 				error = EINVAL;
870 				goto bad;
871 			}
872 			if (*mtod(m, int *))
873 				so->so_options |= optname;
874 			else
875 				so->so_options &= ~optname;
876 			break;
877 
878 		case SO_SNDBUF:
879 		case SO_RCVBUF:
880 		case SO_SNDLOWAT:
881 		case SO_RCVLOWAT:
882 			if (m == NULL || m->m_len < sizeof (int)) {
883 				error = EINVAL;
884 				goto bad;
885 			}
886 			switch (optname) {
887 
888 			case SO_SNDBUF:
889 			case SO_RCVBUF:
890 				if (sbreserve(optname == SO_SNDBUF ?
891 				    &so->so_snd : &so->so_rcv,
892 				    (u_long) *mtod(m, int *)) == 0) {
893 					error = ENOBUFS;
894 					goto bad;
895 				}
896 				break;
897 
898 			case SO_SNDLOWAT:
899 				so->so_snd.sb_lowat = *mtod(m, int *);
900 				break;
901 			case SO_RCVLOWAT:
902 				so->so_rcv.sb_lowat = *mtod(m, int *);
903 				break;
904 			}
905 			break;
906 
907 		case SO_SNDTIMEO:
908 		case SO_RCVTIMEO:
909 		    {
910 			struct timeval *tv;
911 			short val;
912 
913 			if (m == NULL || m->m_len < sizeof (*tv)) {
914 				error = EINVAL;
915 				goto bad;
916 			}
917 			tv = mtod(m, struct timeval *);
918 			if (tv->tv_sec > SHRT_MAX / hz - hz) {
919 				error = EDOM;
920 				goto bad;
921 			}
922 			val = tv->tv_sec * hz + tv->tv_usec / tick;
923 
924 			switch (optname) {
925 
926 			case SO_SNDTIMEO:
927 				so->so_snd.sb_timeo = val;
928 				break;
929 			case SO_RCVTIMEO:
930 				so->so_rcv.sb_timeo = val;
931 				break;
932 			}
933 			break;
934 		    }
935 
936 		default:
937 			error = ENOPROTOOPT;
938 			break;
939 		}
940 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
941 			(void) ((*so->so_proto->pr_ctloutput)
942 				  (PRCO_SETOPT, so, level, optname, &m0));
943 			m = NULL;	/* freed by protocol */
944 		}
945 	}
946 bad:
947 	if (m)
948 		(void) m_free(m);
949 	return (error);
950 }
951 
952 int
953 sogetopt(so, level, optname, mp)
954 	register struct socket *so;
955 	int level, optname;
956 	struct mbuf **mp;
957 {
958 	register struct mbuf *m;
959 
960 	if (level != SOL_SOCKET) {
961 		if (so->so_proto && so->so_proto->pr_ctloutput) {
962 			return ((*so->so_proto->pr_ctloutput)
963 				  (PRCO_GETOPT, so, level, optname, mp));
964 		} else
965 			return (ENOPROTOOPT);
966 	} else {
967 		m = m_get(M_WAIT, MT_SOOPTS);
968 		m->m_len = sizeof (int);
969 
970 		switch (optname) {
971 
972 		case SO_LINGER:
973 			m->m_len = sizeof (struct linger);
974 			mtod(m, struct linger *)->l_onoff =
975 				so->so_options & SO_LINGER;
976 			mtod(m, struct linger *)->l_linger = so->so_linger;
977 			break;
978 
979 		case SO_USELOOPBACK:
980 		case SO_DONTROUTE:
981 		case SO_DEBUG:
982 		case SO_KEEPALIVE:
983 		case SO_REUSEADDR:
984 		case SO_REUSEPORT:
985 		case SO_BROADCAST:
986 		case SO_OOBINLINE:
987 			*mtod(m, int *) = so->so_options & optname;
988 			break;
989 
990 		case SO_TYPE:
991 			*mtod(m, int *) = so->so_type;
992 			break;
993 
994 		case SO_ERROR:
995 			*mtod(m, int *) = so->so_error;
996 			so->so_error = 0;
997 			break;
998 
999 		case SO_SNDBUF:
1000 			*mtod(m, int *) = so->so_snd.sb_hiwat;
1001 			break;
1002 
1003 		case SO_RCVBUF:
1004 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
1005 			break;
1006 
1007 		case SO_SNDLOWAT:
1008 			*mtod(m, int *) = so->so_snd.sb_lowat;
1009 			break;
1010 
1011 		case SO_RCVLOWAT:
1012 			*mtod(m, int *) = so->so_rcv.sb_lowat;
1013 			break;
1014 
1015 		case SO_SNDTIMEO:
1016 		case SO_RCVTIMEO:
1017 		    {
1018 			int val = (optname == SO_SNDTIMEO ?
1019 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1020 
1021 			m->m_len = sizeof(struct timeval);
1022 			mtod(m, struct timeval *)->tv_sec = val / hz;
1023 			mtod(m, struct timeval *)->tv_usec =
1024 			    (val % hz) / tick;
1025 			break;
1026 		    }
1027 
1028 		default:
1029 			(void)m_free(m);
1030 			return (ENOPROTOOPT);
1031 		}
1032 		*mp = m;
1033 		return (0);
1034 	}
1035 }
1036 
1037 void
1038 sohasoutofband(so)
1039 	register struct socket *so;
1040 {
1041 	struct proc *p;
1042 
1043 	if (so->so_pgid < 0)
1044 		gsignal(-so->so_pgid, SIGURG);
1045 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1046 		psignal(p, SIGURG);
1047 	selwakeup(&so->so_rcv.sb_sel);
1048 }
1049