xref: /original-bsd/sys/kern/uipc_socket.c (revision e718337e)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990 Regents of the University of California.
3  * All rights reserved.
4  *
5  * %sccs.include.redist.c%
6  *
7  *	@(#)uipc_socket.c	7.24 (Berkeley) 11/05/90
8  */
9 
10 #include "param.h"
11 #include "user.h"
12 #include "proc.h"
13 #include "file.h"
14 #include "malloc.h"
15 #include "mbuf.h"
16 #include "domain.h"
17 #include "kernel.h"
18 #include "protosw.h"
19 #include "socket.h"
20 #include "socketvar.h"
21 #include "time.h"
22 
23 /*
24  * Socket operation routines.
25  * These routines are called by the routines in
26  * sys_socket.c or from a system process, and
27  * implement the semantics of socket operations by
28  * switching out to the protocol specific routines.
29  *
30  * TODO:
31  *	test socketpair
32  *	clean up async
33  *	out-of-band is a kludge
34  */
35 /*ARGSUSED*/
36 socreate(dom, aso, type, proto)
37 	struct socket **aso;
38 	register int type;
39 	int proto;
40 {
41 	register struct protosw *prp;
42 	register struct socket *so;
43 	register int error;
44 
45 	if (proto)
46 		prp = pffindproto(dom, proto, type);
47 	else
48 		prp = pffindtype(dom, type);
49 	if (prp == 0)
50 		return (EPROTONOSUPPORT);
51 	if (prp->pr_type != type)
52 		return (EPROTOTYPE);
53 	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_WAIT);
54 	bzero((caddr_t)so, sizeof(*so));
55 	so->so_type = type;
56 	if (u.u_uid == 0)
57 		so->so_state = SS_PRIV;
58 	so->so_proto = prp;
59 	error =
60 	    (*prp->pr_usrreq)(so, PRU_ATTACH,
61 		(struct mbuf *)0, (struct mbuf *)proto, (struct mbuf *)0);
62 	if (error) {
63 		so->so_state |= SS_NOFDREF;
64 		sofree(so);
65 		return (error);
66 	}
67 	*aso = so;
68 	return (0);
69 }
70 
71 sobind(so, nam)
72 	struct socket *so;
73 	struct mbuf *nam;
74 {
75 	int s = splnet();
76 	int error;
77 
78 	error =
79 	    (*so->so_proto->pr_usrreq)(so, PRU_BIND,
80 		(struct mbuf *)0, nam, (struct mbuf *)0);
81 	splx(s);
82 	return (error);
83 }
84 
85 solisten(so, backlog)
86 	register struct socket *so;
87 	int backlog;
88 {
89 	int s = splnet(), error;
90 
91 	error =
92 	    (*so->so_proto->pr_usrreq)(so, PRU_LISTEN,
93 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
94 	if (error) {
95 		splx(s);
96 		return (error);
97 	}
98 	if (so->so_q == 0)
99 		so->so_options |= SO_ACCEPTCONN;
100 	if (backlog < 0)
101 		backlog = 0;
102 	so->so_qlimit = min(backlog, SOMAXCONN);
103 	splx(s);
104 	return (0);
105 }
106 
107 sofree(so)
108 	register struct socket *so;
109 {
110 
111 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
112 		return;
113 	if (so->so_head) {
114 		if (!soqremque(so, 0) && !soqremque(so, 1))
115 			panic("sofree dq");
116 		so->so_head = 0;
117 	}
118 	sbrelease(&so->so_snd);
119 	sorflush(so);
120 	FREE(so, M_SOCKET);
121 }
122 
123 /*
124  * Close a socket on last file table reference removal.
125  * Initiate disconnect if connected.
126  * Free socket when disconnect complete.
127  */
128 soclose(so)
129 	register struct socket *so;
130 {
131 	int s = splnet();		/* conservative */
132 	int error = 0;
133 
134 	if (so->so_options & SO_ACCEPTCONN) {
135 		while (so->so_q0)
136 			(void) soabort(so->so_q0);
137 		while (so->so_q)
138 			(void) soabort(so->so_q);
139 	}
140 	if (so->so_pcb == 0)
141 		goto discard;
142 	if (so->so_state & SS_ISCONNECTED) {
143 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
144 			error = sodisconnect(so);
145 			if (error)
146 				goto drop;
147 		}
148 		if (so->so_options & SO_LINGER) {
149 			if ((so->so_state & SS_ISDISCONNECTING) &&
150 			    (so->so_state & SS_NBIO))
151 				goto drop;
152 			while (so->so_state & SS_ISCONNECTED)
153 				if (error = tsleep((caddr_t)&so->so_timeo,
154 				    PSOCK | PCATCH, netcls, so->so_linger))
155 					break;
156 		}
157 	}
158 drop:
159 	if (so->so_pcb) {
160 		int error2 =
161 		    (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
162 			(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
163 		if (error == 0)
164 			error = error2;
165 	}
166 discard:
167 	if (so->so_state & SS_NOFDREF)
168 		panic("soclose: NOFDREF");
169 	so->so_state |= SS_NOFDREF;
170 	sofree(so);
171 	splx(s);
172 	return (error);
173 }
174 
175 /*
176  * Must be called at splnet...
177  */
178 soabort(so)
179 	struct socket *so;
180 {
181 
182 	return (
183 	    (*so->so_proto->pr_usrreq)(so, PRU_ABORT,
184 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
185 }
186 
187 soaccept(so, nam)
188 	register struct socket *so;
189 	struct mbuf *nam;
190 {
191 	int s = splnet();
192 	int error;
193 
194 	if ((so->so_state & SS_NOFDREF) == 0)
195 		panic("soaccept: !NOFDREF");
196 	so->so_state &= ~SS_NOFDREF;
197 	error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
198 	    (struct mbuf *)0, nam, (struct mbuf *)0);
199 	splx(s);
200 	return (error);
201 }
202 
203 soconnect(so, nam)
204 	register struct socket *so;
205 	struct mbuf *nam;
206 {
207 	int s;
208 	int error;
209 
210 	if (so->so_options & SO_ACCEPTCONN)
211 		return (EOPNOTSUPP);
212 	s = splnet();
213 	/*
214 	 * If protocol is connection-based, can only connect once.
215 	 * Otherwise, if connected, try to disconnect first.
216 	 * This allows user to disconnect by connecting to, e.g.,
217 	 * a null address.
218 	 */
219 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
220 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
221 	    (error = sodisconnect(so))))
222 		error = EISCONN;
223 	else
224 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
225 		    (struct mbuf *)0, nam, (struct mbuf *)0);
226 	splx(s);
227 	return (error);
228 }
229 
230 soconnect2(so1, so2)
231 	register struct socket *so1;
232 	struct socket *so2;
233 {
234 	int s = splnet();
235 	int error;
236 
237 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
238 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0);
239 	splx(s);
240 	return (error);
241 }
242 
243 sodisconnect(so)
244 	register struct socket *so;
245 {
246 	int s = splnet();
247 	int error;
248 
249 	if ((so->so_state & SS_ISCONNECTED) == 0) {
250 		error = ENOTCONN;
251 		goto bad;
252 	}
253 	if (so->so_state & SS_ISDISCONNECTING) {
254 		error = EALREADY;
255 		goto bad;
256 	}
257 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
258 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
259 bad:
260 	splx(s);
261 	return (error);
262 }
263 
264 /*
265  * Send on a socket.
266  * If send must go all at once and message is larger than
267  * send buffering, then hard error.
268  * Lock against other senders.
269  * If must go all at once and not enough room now, then
270  * inform user that this would block and do nothing.
271  * Otherwise, if nonblocking, send as much as possible.
272  * The data to be sent is described by "uio" if nonzero,
273  * otherwise by the mbuf chain "top" (which must be null
274  * if uio is not).  Data provided in mbuf chain must be small
275  * enough to send all at once.
276  *
277  * Returns nonzero on error, timeout or signal; callers
278  * must check for short counts if EINTR/ERESTART are returned.
279  * Data and control buffers are freed on return.
280  */
281 sosend(so, addr, uio, top, control, flags)
282 	register struct socket *so;
283 	struct mbuf *addr;
284 	struct uio *uio;
285 	struct mbuf *top;
286 	struct mbuf *control;
287 	int flags;
288 {
289 	struct mbuf **mp;
290 	register struct mbuf *m;
291 	register long space, len, resid;
292 	int clen = 0, error, s, dontroute, mlen;
293 	int atomic = sosendallatonce(so) || top;
294 
295 	if (uio)
296 		resid = uio->uio_resid;
297 	else
298 		resid = top->m_pkthdr.len;
299 	dontroute =
300 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
301 	    (so->so_proto->pr_flags & PR_ATOMIC);
302 	u.u_ru.ru_msgsnd++;
303 	if (control)
304 		clen = control->m_len;
305 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
306 
307 restart:
308 	if (error = sblock(&so->so_snd))
309 		goto out;
310 	do {
311 		s = splnet();
312 		if (so->so_state & SS_CANTSENDMORE)
313 			snderr(EPIPE);
314 		if (so->so_error)
315 			snderr(so->so_error);
316 		if ((so->so_state & SS_ISCONNECTED) == 0) {
317 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
318 				if ((so->so_state & SS_ISCONFIRMING) == 0)
319 					snderr(ENOTCONN);
320 			} else if (addr == 0)
321 				snderr(EDESTADDRREQ);
322 		}
323 		space = sbspace(&so->so_snd);
324 		if (flags & MSG_OOB)
325 			space += 1024;
326 		if (space < resid + clen &&
327 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
328 			if (atomic && resid > so->so_snd.sb_hiwat ||
329 			    clen > so->so_snd.sb_hiwat)
330 				snderr(EMSGSIZE);
331 			if (so->so_state & SS_NBIO)
332 				snderr(EWOULDBLOCK);
333 			sbunlock(&so->so_snd);
334 			error = sbwait(&so->so_snd);
335 			splx(s);
336 			if (error)
337 				goto out;
338 			goto restart;
339 		}
340 		splx(s);
341 		mp = &top;
342 		space -= clen;
343 		do {
344 		    if (uio == NULL) {
345 			/*
346 			 * Data is prepackaged in "top".
347 			 */
348 			resid = 0;
349 			if (flags & MSG_EOR)
350 				top->m_flags |= M_EOR;
351 		    } else do {
352 			if (top == 0) {
353 				MGETHDR(m, M_WAIT, MT_DATA);
354 				mlen = MHLEN;
355 				m->m_pkthdr.len = 0;
356 				m->m_pkthdr.rcvif = (struct ifnet *)0;
357 			} else {
358 				MGET(m, M_WAIT, MT_DATA);
359 				mlen = MLEN;
360 			}
361 			if (resid >= MINCLSIZE && space >= MCLBYTES) {
362 				MCLGET(m, M_WAIT);
363 				if ((m->m_flags & M_EXT) == 0)
364 					goto nopages;
365 				mlen = MCLBYTES;
366 #ifdef	MAPPED_MBUFS
367 				len = min(MCLBYTES, resid);
368 #else
369 				if (top == 0) {
370 					len = min(MCLBYTES - max_hdr, resid);
371 					m->m_data += max_hdr;
372 				}
373 #endif
374 				space -= MCLBYTES;
375 			} else {
376 nopages:
377 				len = min(min(mlen, resid), space);
378 				space -= len;
379 				/*
380 				 * For datagram protocols, leave room
381 				 * for protocol headers in first mbuf.
382 				 */
383 				if (atomic && top == 0 && len < mlen)
384 					MH_ALIGN(m, len);
385 			}
386 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
387 			resid = uio->uio_resid;
388 			m->m_len = len;
389 			*mp = m;
390 			top->m_pkthdr.len += len;
391 			if (error)
392 				goto release;
393 			mp = &m->m_next;
394 			if (resid <= 0) {
395 				if (flags & MSG_EOR)
396 					top->m_flags |= M_EOR;
397 				break;
398 			}
399 		    } while (space > 0 && atomic);
400 		    if (dontroute)
401 			    so->so_options |= SO_DONTROUTE;
402 		    s = splnet();				/* XXX */
403 		    error = (*so->so_proto->pr_usrreq)(so,
404 			(flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
405 			top, addr, control);
406 		    splx(s);
407 		    if (dontroute)
408 			    so->so_options &= ~SO_DONTROUTE;
409 		    clen = 0;
410 		    control = 0;
411 		    top = 0;
412 		    mp = &top;
413 		    if (error)
414 			goto release;
415 		} while (resid && space > 0);
416 	} while (resid);
417 
418 release:
419 	sbunlock(&so->so_snd);
420 out:
421 	if (top)
422 		m_freem(top);
423 	if (control)
424 		m_freem(control);
425 	return (error);
426 }
427 
428 /*
429  * Implement receive operations on a socket.
430  * We depend on the way that records are added to the sockbuf
431  * by sbappend*.  In particular, each record (mbufs linked through m_next)
432  * must begin with an address if the protocol so specifies,
433  * followed by an optional mbuf or mbufs containing ancillary data,
434  * and then zero or more mbufs of data.
435  * In order to avoid blocking network interrupts for the entire time here,
436  * we splx() while doing the actual copy to user space.
437  * Although the sockbuf is locked, new data may still be appended,
438  * and thus we must maintain consistency of the sockbuf during that time.
439  *
440  * The caller may receive the data as a single mbuf chain by supplying
441  * an mbuf **mp0 for use in returning the chain.  The uio is then used
442  * only for the count in uio_resid.
443  */
444 soreceive(so, paddr, uio, mp0, controlp, flagsp)
445 	register struct socket *so;
446 	struct mbuf **paddr;
447 	struct uio *uio;
448 	struct mbuf **mp0;
449 	struct mbuf **controlp;
450 	int *flagsp;
451 {
452 	register struct mbuf *m, **mp;
453 	register int flags, len, error, s, offset;
454 	struct protosw *pr = so->so_proto;
455 	struct mbuf *nextrecord;
456 	int moff, type;
457 
458 	mp = mp0;
459 	if (paddr)
460 		*paddr = 0;
461 	if (controlp)
462 		*controlp = 0;
463 	if (flagsp)
464 		flags = *flagsp &~ MSG_EOR;
465 	else
466 		flags = 0;
467 	if (flags & MSG_OOB) {
468 		m = m_get(M_WAIT, MT_DATA);
469 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB,
470 		    m, (struct mbuf *)(flags & MSG_PEEK), (struct mbuf *)0);
471 		if (error)
472 			goto bad;
473 		do {
474 			error = uiomove(mtod(m, caddr_t),
475 			    (int) min(uio->uio_resid, m->m_len), uio);
476 			m = m_free(m);
477 		} while (uio->uio_resid && error == 0 && m);
478 bad:
479 		if (m)
480 			m_freem(m);
481 		return (error);
482 	}
483 	if (mp)
484 		*mp = (struct mbuf *)0;
485 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
486 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
487 		    (struct mbuf *)0, (struct mbuf *)0);
488 
489 restart:
490 	if (error = sblock(&so->so_rcv))
491 		return (error);
492 	s = splnet();
493 
494 	m = so->so_rcv.sb_mb;
495 	/*
496 	 * If we have less data than requested, block awaiting more
497 	 * (subject to any timeout) if:
498 	 *   1. the current count is less than the low water mark, or
499 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
500 	 *	receive operation at once if we block (resid <= hiwat).
501 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
502 	 * we have to do the receive in sections, and thus risk returning
503 	 * a short count if a timeout or signal occurs after we start.
504 	 */
505 	if (m == 0 || so->so_rcv.sb_cc < uio->uio_resid &&
506 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
507 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat))) {
508 #ifdef DIAGNOSTIC
509 		if (m == 0 && so->so_rcv.sb_cc)
510 			panic("receive 1");
511 #endif
512 		if (so->so_error) {
513 			error = so->so_error;
514 			so->so_error = 0;
515 			goto release;
516 		}
517 		if (so->so_state & SS_CANTRCVMORE)
518 			goto release;
519 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
520 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
521 			error = ENOTCONN;
522 			goto release;
523 		}
524 		if (uio->uio_resid == 0)
525 			goto release;
526 		if (so->so_state & SS_NBIO) {
527 			error = EWOULDBLOCK;
528 			goto release;
529 		}
530 		sbunlock(&so->so_rcv);
531 		error = sbwait(&so->so_rcv);
532 		splx(s);
533 		if (error)
534 			return (error);
535 		goto restart;
536 	}
537 	u.u_ru.ru_msgrcv++;
538 	nextrecord = m->m_nextpkt;
539 	if (pr->pr_flags & PR_ADDR) {
540 #ifdef DIAGNOSTIC
541 		if (m->m_type != MT_SONAME)
542 			panic("receive 1a");
543 #endif
544 		if (flags & MSG_PEEK) {
545 			if (paddr)
546 				*paddr = m_copy(m, 0, m->m_len);
547 			m = m->m_next;
548 		} else {
549 			sbfree(&so->so_rcv, m);
550 			if (paddr) {
551 				*paddr = m;
552 				so->so_rcv.sb_mb = m->m_next;
553 				m->m_next = 0;
554 				m = so->so_rcv.sb_mb;
555 			} else {
556 				MFREE(m, so->so_rcv.sb_mb);
557 				m = so->so_rcv.sb_mb;
558 			}
559 		}
560 	}
561 	while (m && m->m_type == MT_CONTROL && error == 0) {
562 		if (flags & MSG_PEEK) {
563 			if (controlp)
564 				*controlp = m_copy(m, 0, m->m_len);
565 			m = m->m_next;
566 		} else {
567 			sbfree(&so->so_rcv, m);
568 			if (controlp) {
569 				if (pr->pr_domain->dom_externalize &&
570 				    mtod(m, struct cmsghdr *)->cmsg_type ==
571 				    SCM_RIGHTS)
572 				   error = (*pr->pr_domain->dom_externalize)(m);
573 				*controlp = m;
574 				so->so_rcv.sb_mb = m->m_next;
575 				m->m_next = 0;
576 				m = so->so_rcv.sb_mb;
577 			} else {
578 				MFREE(m, so->so_rcv.sb_mb);
579 				m = so->so_rcv.sb_mb;
580 			}
581 		}
582 		if (controlp)
583 			controlp = &(*controlp)->m_next;
584 	}
585 	if (m) {
586 		if ((flags & MSG_PEEK) == 0)
587 			m->m_nextpkt = nextrecord;
588 		type = m->m_type;
589 	}
590 	moff = 0;
591 	offset = 0;
592 	while (m && m->m_type == type && uio->uio_resid > 0 && error == 0) {
593 		if (m->m_type == MT_OOBDATA)
594 			flags |= MSG_OOB;
595 #ifdef DIAGNOSTIC
596 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
597 			panic("receive 3");
598 #endif
599 		so->so_state &= ~SS_RCVATMARK;
600 		len = uio->uio_resid;
601 		if (so->so_oobmark && len > so->so_oobmark - offset)
602 			len = so->so_oobmark - offset;
603 		if (len > m->m_len - moff)
604 			len = m->m_len - moff;
605 		/*
606 		 * If mp is set, just pass back the mbufs.
607 		 * Otherwise copy them out via the uio, then free.
608 		 * Sockbuf must be consistent here (points to current mbuf,
609 		 * it points to next record) when we drop priority;
610 		 * we must note any additions to the sockbuf when we
611 		 * block interrupts again.
612 		 */
613 		if (mp == 0) {
614 			splx(s);
615 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
616 			s = splnet();
617 		} else
618 			uio->uio_resid -= len;
619 		if (len == m->m_len - moff) {
620 			if (m->m_flags & M_EOR)
621 				flags |= MSG_EOR;
622 			if (flags & MSG_PEEK) {
623 				m = m->m_next;
624 				moff = 0;
625 			} else {
626 				nextrecord = m->m_nextpkt;
627 				sbfree(&so->so_rcv, m);
628 				if (mp) {
629 					*mp = m;
630 					mp = &m->m_next;
631 					so->so_rcv.sb_mb = m = m->m_next;
632 					*mp = (struct mbuf *)0;
633 				} else {
634 					MFREE(m, so->so_rcv.sb_mb);
635 					m = so->so_rcv.sb_mb;
636 				}
637 				if (m)
638 					m->m_nextpkt = nextrecord;
639 			}
640 		} else {
641 			if (flags & MSG_PEEK)
642 				moff += len;
643 			else {
644 				if (mp)
645 					*mp = m_copym(m, 0, len, M_WAIT);
646 				m->m_data += len;
647 				m->m_len -= len;
648 				so->so_rcv.sb_cc -= len;
649 			}
650 		}
651 		if (so->so_oobmark) {
652 			if ((flags & MSG_PEEK) == 0) {
653 				so->so_oobmark -= len;
654 				if (so->so_oobmark == 0) {
655 					so->so_state |= SS_RCVATMARK;
656 					break;
657 				}
658 			} else
659 				offset += len;
660 		}
661 		if (flags & MSG_EOR)
662 			break;
663 		/*
664 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
665 		 * we must not quit until "uio->uio_resid == 0" or an error
666 		 * termination.  If a signal/timeout occurs, return
667 		 * with a short count but without error.
668 		 * Keep sockbuf locked against other readers.
669 		 */
670 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
671 		    !sosendallatonce(so)) {
672 			error = sbwait(&so->so_rcv);
673 			if (error) {
674 				sbunlock(&so->so_rcv);
675 				splx(s);
676 				return (0);
677 			}
678 			if (m = so->so_rcv.sb_mb)
679 				nextrecord = m->m_nextpkt;
680 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
681 				break;
682 			continue;
683 		}
684 	}
685 	if ((flags & MSG_PEEK) == 0) {
686 		if (m == 0)
687 			so->so_rcv.sb_mb = nextrecord;
688 		else if (pr->pr_flags & PR_ATOMIC) {
689 			flags |= MSG_TRUNC;
690 			(void) sbdroprecord(&so->so_rcv);
691 		}
692 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
693 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
694 			    (struct mbuf *)flags, (struct mbuf *)0,
695 			    (struct mbuf *)0);
696 	}
697 	if (flagsp)
698 		*flagsp |= flags;
699 release:
700 	sbunlock(&so->so_rcv);
701 	splx(s);
702 	return (error);
703 }
704 
705 soshutdown(so, how)
706 	register struct socket *so;
707 	register int how;
708 {
709 	register struct protosw *pr = so->so_proto;
710 
711 	how++;
712 	if (how & FREAD)
713 		sorflush(so);
714 	if (how & FWRITE)
715 		return ((*pr->pr_usrreq)(so, PRU_SHUTDOWN,
716 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
717 	return (0);
718 }
719 
720 sorflush(so)
721 	register struct socket *so;
722 {
723 	register struct sockbuf *sb = &so->so_rcv;
724 	register struct protosw *pr = so->so_proto;
725 	register int s;
726 	struct sockbuf asb;
727 
728 	sb->sb_flags |= SB_NOINTR;
729 	(void) sblock(sb);
730 	s = splimp();
731 	socantrcvmore(so);
732 	sbunlock(sb);
733 	asb = *sb;
734 	bzero((caddr_t)sb, sizeof (*sb));
735 	splx(s);
736 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
737 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
738 	sbrelease(&asb);
739 }
740 
741 sosetopt(so, level, optname, m0)
742 	register struct socket *so;
743 	int level, optname;
744 	struct mbuf *m0;
745 {
746 	int error = 0;
747 	register struct mbuf *m = m0;
748 
749 	if (level != SOL_SOCKET) {
750 		if (so->so_proto && so->so_proto->pr_ctloutput)
751 			return ((*so->so_proto->pr_ctloutput)
752 				  (PRCO_SETOPT, so, level, optname, &m0));
753 		error = ENOPROTOOPT;
754 	} else {
755 		switch (optname) {
756 
757 		case SO_LINGER:
758 			if (m == NULL || m->m_len != sizeof (struct linger)) {
759 				error = EINVAL;
760 				goto bad;
761 			}
762 			so->so_linger = mtod(m, struct linger *)->l_linger;
763 			/* fall thru... */
764 
765 		case SO_DEBUG:
766 		case SO_KEEPALIVE:
767 		case SO_DONTROUTE:
768 		case SO_USELOOPBACK:
769 		case SO_BROADCAST:
770 		case SO_REUSEADDR:
771 		case SO_OOBINLINE:
772 			if (m == NULL || m->m_len < sizeof (int)) {
773 				error = EINVAL;
774 				goto bad;
775 			}
776 			if (*mtod(m, int *))
777 				so->so_options |= optname;
778 			else
779 				so->so_options &= ~optname;
780 			break;
781 
782 		case SO_SNDBUF:
783 		case SO_RCVBUF:
784 		case SO_SNDLOWAT:
785 		case SO_RCVLOWAT:
786 			if (m == NULL || m->m_len < sizeof (int)) {
787 				error = EINVAL;
788 				goto bad;
789 			}
790 			switch (optname) {
791 
792 			case SO_SNDBUF:
793 			case SO_RCVBUF:
794 				if (sbreserve(optname == SO_SNDBUF ?
795 				    &so->so_snd : &so->so_rcv,
796 				    (u_long) *mtod(m, int *)) == 0) {
797 					error = ENOBUFS;
798 					goto bad;
799 				}
800 				break;
801 
802 			case SO_SNDLOWAT:
803 				so->so_snd.sb_lowat = *mtod(m, int *);
804 				break;
805 			case SO_RCVLOWAT:
806 				so->so_rcv.sb_lowat = *mtod(m, int *);
807 				break;
808 			}
809 			break;
810 
811 		case SO_SNDTIMEO:
812 		case SO_RCVTIMEO:
813 		    {
814 			struct timeval *tv;
815 			short val;
816 
817 			if (m == NULL || m->m_len < sizeof (*tv)) {
818 				error = EINVAL;
819 				goto bad;
820 			}
821 			tv = mtod(m, struct timeval *);
822 			if (tv->tv_sec > SHRT_MAX / hz - hz) {
823 				error = EDOM;
824 				goto bad;
825 			}
826 			val = tv->tv_sec * hz + tv->tv_usec / tick;
827 
828 			switch (optname) {
829 
830 			case SO_SNDTIMEO:
831 				so->so_snd.sb_timeo = val;
832 				break;
833 			case SO_RCVTIMEO:
834 				so->so_rcv.sb_timeo = val;
835 				break;
836 			}
837 			break;
838 		    }
839 
840 		default:
841 			error = ENOPROTOOPT;
842 			break;
843 		}
844 	}
845 bad:
846 	if (m)
847 		(void) m_free(m);
848 	return (error);
849 }
850 
851 sogetopt(so, level, optname, mp)
852 	register struct socket *so;
853 	int level, optname;
854 	struct mbuf **mp;
855 {
856 	register struct mbuf *m;
857 
858 	if (level != SOL_SOCKET) {
859 		if (so->so_proto && so->so_proto->pr_ctloutput) {
860 			return ((*so->so_proto->pr_ctloutput)
861 				  (PRCO_GETOPT, so, level, optname, mp));
862 		} else
863 			return (ENOPROTOOPT);
864 	} else {
865 		m = m_get(M_WAIT, MT_SOOPTS);
866 		m->m_len = sizeof (int);
867 
868 		switch (optname) {
869 
870 		case SO_LINGER:
871 			m->m_len = sizeof (struct linger);
872 			mtod(m, struct linger *)->l_onoff =
873 				so->so_options & SO_LINGER;
874 			mtod(m, struct linger *)->l_linger = so->so_linger;
875 			break;
876 
877 		case SO_USELOOPBACK:
878 		case SO_DONTROUTE:
879 		case SO_DEBUG:
880 		case SO_KEEPALIVE:
881 		case SO_REUSEADDR:
882 		case SO_BROADCAST:
883 		case SO_OOBINLINE:
884 			*mtod(m, int *) = so->so_options & optname;
885 			break;
886 
887 		case SO_TYPE:
888 			*mtod(m, int *) = so->so_type;
889 			break;
890 
891 		case SO_ERROR:
892 			*mtod(m, int *) = so->so_error;
893 			so->so_error = 0;
894 			break;
895 
896 		case SO_SNDBUF:
897 			*mtod(m, int *) = so->so_snd.sb_hiwat;
898 			break;
899 
900 		case SO_RCVBUF:
901 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
902 			break;
903 
904 		case SO_SNDLOWAT:
905 			*mtod(m, int *) = so->so_snd.sb_lowat;
906 			break;
907 
908 		case SO_RCVLOWAT:
909 			*mtod(m, int *) = so->so_rcv.sb_lowat;
910 			break;
911 
912 		case SO_SNDTIMEO:
913 		case SO_RCVTIMEO:
914 		    {
915 			int val = (optname == SO_SNDTIMEO ?
916 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
917 
918 			m->m_len = sizeof(struct timeval);
919 			mtod(m, struct timeval *)->tv_sec = val / hz;
920 			mtod(m, struct timeval *)->tv_usec =
921 			    (val % hz) / tick;
922 			break;
923 		    }
924 
925 		default:
926 			(void)m_free(m);
927 			return (ENOPROTOOPT);
928 		}
929 		*mp = m;
930 		return (0);
931 	}
932 }
933 
934 sohasoutofband(so)
935 	register struct socket *so;
936 {
937 	struct proc *p;
938 
939 	if (so->so_pgid < 0)
940 		gsignal(-so->so_pgid, SIGURG);
941 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
942 		psignal(p, SIGURG);
943 	if (so->so_rcv.sb_sel) {
944 		selwakeup(so->so_rcv.sb_sel, so->so_rcv.sb_flags & SB_COLL);
945 		so->so_rcv.sb_sel = 0;
946 		so->so_rcv.sb_flags &= ~SB_COLL;
947 	}
948 }
949