xref: /csrg-svn/sys/kern/uipc_socket.c (revision 44450)
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	7.22 (Berkeley) 06/28/90
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 "protosw.h"
18 #include "socket.h"
19 #include "socketvar.h"
20 
21 /*
22  * Socket operation routines.
23  * These routines are called by the routines in
24  * sys_socket.c or from a system process, and
25  * implement the semantics of socket operations by
26  * switching out to the protocol specific routines.
27  *
28  * TODO:
29  *	test socketpair
30  *	clean up async
31  *	out-of-band is a kludge
32  */
33 /*ARGSUSED*/
34 socreate(dom, aso, type, proto)
35 	struct socket **aso;
36 	register int type;
37 	int proto;
38 {
39 	register struct protosw *prp;
40 	register struct socket *so;
41 	register int error;
42 
43 	if (proto)
44 		prp = pffindproto(dom, proto, type);
45 	else
46 		prp = pffindtype(dom, type);
47 	if (prp == 0)
48 		return (EPROTONOSUPPORT);
49 	if (prp->pr_type != type)
50 		return (EPROTOTYPE);
51 	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_WAIT);
52 	bzero((caddr_t)so, sizeof(*so));
53 	so->so_type = type;
54 	if (u.u_uid == 0)
55 		so->so_state = SS_PRIV;
56 	so->so_proto = prp;
57 	error =
58 	    (*prp->pr_usrreq)(so, PRU_ATTACH,
59 		(struct mbuf *)0, (struct mbuf *)proto, (struct mbuf *)0);
60 	if (error) {
61 		so->so_state |= SS_NOFDREF;
62 		sofree(so);
63 		return (error);
64 	}
65 	*aso = so;
66 	return (0);
67 }
68 
69 sobind(so, nam)
70 	struct socket *so;
71 	struct mbuf *nam;
72 {
73 	int s = splnet();
74 	int error;
75 
76 	error =
77 	    (*so->so_proto->pr_usrreq)(so, PRU_BIND,
78 		(struct mbuf *)0, nam, (struct mbuf *)0);
79 	splx(s);
80 	return (error);
81 }
82 
83 solisten(so, backlog)
84 	register struct socket *so;
85 	int backlog;
86 {
87 	int s = splnet(), error;
88 
89 	error =
90 	    (*so->so_proto->pr_usrreq)(so, PRU_LISTEN,
91 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
92 	if (error) {
93 		splx(s);
94 		return (error);
95 	}
96 	if (so->so_q == 0)
97 		so->so_options |= SO_ACCEPTCONN;
98 	if (backlog < 0)
99 		backlog = 0;
100 	so->so_qlimit = min(backlog, SOMAXCONN);
101 	splx(s);
102 	return (0);
103 }
104 
105 sofree(so)
106 	register struct socket *so;
107 {
108 
109 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
110 		return;
111 	if (so->so_head) {
112 		if (!soqremque(so, 0) && !soqremque(so, 1))
113 			panic("sofree dq");
114 		so->so_head = 0;
115 	}
116 	sbrelease(&so->so_snd);
117 	sorflush(so);
118 	FREE(so, M_SOCKET);
119 }
120 
121 /*
122  * Close a socket on last file table reference removal.
123  * Initiate disconnect if connected.
124  * Free socket when disconnect complete.
125  */
126 soclose(so)
127 	register struct socket *so;
128 {
129 	int s = splnet();		/* conservative */
130 	int error = 0;
131 
132 	if (so->so_options & SO_ACCEPTCONN) {
133 		while (so->so_q0)
134 			(void) soabort(so->so_q0);
135 		while (so->so_q)
136 			(void) soabort(so->so_q);
137 	}
138 	if (so->so_pcb == 0)
139 		goto discard;
140 	if (so->so_state & SS_ISCONNECTED) {
141 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
142 			error = sodisconnect(so);
143 			if (error)
144 				goto drop;
145 		}
146 		if (so->so_options & SO_LINGER) {
147 			if ((so->so_state & SS_ISDISCONNECTING) &&
148 			    (so->so_state & SS_NBIO))
149 				goto drop;
150 			while (so->so_state & SS_ISCONNECTED)
151 				if (error = tsleep((caddr_t)&so->so_timeo,
152 				    PSOCK | PCATCH, netcls, so->so_linger))
153 					break;
154 		}
155 	}
156 drop:
157 	if (so->so_pcb) {
158 		int error2 =
159 		    (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
160 			(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
161 		if (error == 0)
162 			error = error2;
163 	}
164 discard:
165 	if (so->so_state & SS_NOFDREF)
166 		panic("soclose: NOFDREF");
167 	so->so_state |= SS_NOFDREF;
168 	sofree(so);
169 	splx(s);
170 	return (error);
171 }
172 
173 /*
174  * Must be called at splnet...
175  */
176 soabort(so)
177 	struct socket *so;
178 {
179 
180 	return (
181 	    (*so->so_proto->pr_usrreq)(so, PRU_ABORT,
182 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
183 }
184 
185 soaccept(so, nam)
186 	register struct socket *so;
187 	struct mbuf *nam;
188 {
189 	int s = splnet();
190 	int error;
191 
192 	if ((so->so_state & SS_NOFDREF) == 0)
193 		panic("soaccept: !NOFDREF");
194 	so->so_state &= ~SS_NOFDREF;
195 	error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
196 	    (struct mbuf *)0, nam, (struct mbuf *)0);
197 	splx(s);
198 	return (error);
199 }
200 
201 soconnect(so, nam)
202 	register struct socket *so;
203 	struct mbuf *nam;
204 {
205 	int s;
206 	int error;
207 
208 	if (so->so_options & SO_ACCEPTCONN)
209 		return (EOPNOTSUPP);
210 	s = splnet();
211 	/*
212 	 * If protocol is connection-based, can only connect once.
213 	 * Otherwise, if connected, try to disconnect first.
214 	 * This allows user to disconnect by connecting to, e.g.,
215 	 * a null address.
216 	 */
217 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
218 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
219 	    (error = sodisconnect(so))))
220 		error = EISCONN;
221 	else
222 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
223 		    (struct mbuf *)0, nam, (struct mbuf *)0);
224 	splx(s);
225 	return (error);
226 }
227 
228 soconnect2(so1, so2)
229 	register struct socket *so1;
230 	struct socket *so2;
231 {
232 	int s = splnet();
233 	int error;
234 
235 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
236 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0);
237 	splx(s);
238 	return (error);
239 }
240 
241 sodisconnect(so)
242 	register struct socket *so;
243 {
244 	int s = splnet();
245 	int error;
246 
247 	if ((so->so_state & SS_ISCONNECTED) == 0) {
248 		error = ENOTCONN;
249 		goto bad;
250 	}
251 	if (so->so_state & SS_ISDISCONNECTING) {
252 		error = EALREADY;
253 		goto bad;
254 	}
255 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
256 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
257 bad:
258 	splx(s);
259 	return (error);
260 }
261 
262 /*
263  * Send on a socket.
264  * If send must go all at once and message is larger than
265  * send buffering, then hard error.
266  * Lock against other senders.
267  * If must go all at once and not enough room now, then
268  * inform user that this would block and do nothing.
269  * Otherwise, if nonblocking, send as much as possible.
270  * The data to be sent is described by "uio" if nonzero,
271  * otherwise by the mbuf chain "top" (which must be null
272  * if uio is not).  Data provided in mbuf chain must be small
273  * enough to send all at once.
274  *
275  * Returns nonzero on error, timeout or signal; callers
276  * must check for short counts if EINTR/ERESTART are returned.
277  * Data and control buffers are freed on return.
278  */
279 sosend(so, addr, uio, top, control, flags)
280 	register struct socket *so;
281 	struct mbuf *addr;
282 	struct uio *uio;
283 	struct mbuf *top;
284 	struct mbuf *control;
285 	int flags;
286 {
287 	struct mbuf **mp;
288 	register struct mbuf *m;
289 	register long space, len, resid;
290 	int clen = 0, error, s, dontroute, mlen;
291 	int atomic = sosendallatonce(so) || top;
292 
293 	if (uio)
294 		resid = uio->uio_resid;
295 	else
296 		resid = top->m_pkthdr.len;
297 	dontroute =
298 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
299 	    (so->so_proto->pr_flags & PR_ATOMIC);
300 	u.u_ru.ru_msgsnd++;
301 	if (control)
302 		clen = control->m_len;
303 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
304 
305 restart:
306 	if (error = sblock(&so->so_snd))
307 		goto out;
308 	do {
309 		s = splnet();
310 		if (so->so_state & SS_CANTSENDMORE)
311 			snderr(EPIPE);
312 		if (so->so_error)
313 			snderr(so->so_error);
314 		if ((so->so_state & SS_ISCONNECTED) == 0) {
315 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
316 				if ((so->so_state & SS_ISCONFIRMING) == 0)
317 					snderr(ENOTCONN);
318 			} else if (addr == 0)
319 				snderr(EDESTADDRREQ);
320 		}
321 		space = sbspace(&so->so_snd);
322 		if (flags & MSG_OOB)
323 			space += 1024;
324 		if (space < resid + clen &&
325 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
326 			if (atomic && resid > so->so_snd.sb_hiwat ||
327 			    clen > so->so_snd.sb_hiwat)
328 				snderr(EMSGSIZE);
329 			if (so->so_state & SS_NBIO)
330 				snderr(EWOULDBLOCK);
331 			sbunlock(&so->so_snd);
332 			error = sbwait(&so->so_snd);
333 			splx(s);
334 			if (error)
335 				goto out;
336 			goto restart;
337 		}
338 		splx(s);
339 		mp = &top;
340 		space -= clen;
341 		do {
342 		    if (uio == NULL) {
343 			/*
344 			 * Data is prepackaged in "top".
345 			 */
346 			resid = 0;
347 			if (flags & MSG_EOR)
348 				top->m_flags |= M_EOR;
349 		    } else do {
350 			if (top == 0) {
351 				MGETHDR(m, M_WAIT, MT_DATA);
352 				mlen = MHLEN;
353 				m->m_pkthdr.len = 0;
354 				m->m_pkthdr.rcvif = (struct ifnet *)0;
355 			} else {
356 				MGET(m, M_WAIT, MT_DATA);
357 				mlen = MLEN;
358 			}
359 			if (resid >= MINCLSIZE && space >= MCLBYTES) {
360 				MCLGET(m, M_WAIT);
361 				if ((m->m_flags & M_EXT) == 0)
362 					goto nopages;
363 				mlen = MCLBYTES;
364 #ifdef	MAPPED_MBUFS
365 				len = min(MCLBYTES, resid);
366 #else
367 				if (top == 0) {
368 					len = min(MCLBYTES - max_hdr, resid);
369 					m->m_data += max_hdr;
370 				}
371 #endif
372 				space -= MCLBYTES;
373 			} else {
374 nopages:
375 				len = min(min(mlen, resid), space);
376 				space -= len;
377 				/*
378 				 * For datagram protocols, leave room
379 				 * for protocol headers in first mbuf.
380 				 */
381 				if (atomic && top == 0 && len < mlen)
382 					MH_ALIGN(m, len);
383 			}
384 			error = uiomove(mtod(m, caddr_t), len, uio);
385 			resid = uio->uio_resid;
386 			m->m_len = len;
387 			*mp = m;
388 			top->m_pkthdr.len += len;
389 			if (error)
390 				goto release;
391 			mp = &m->m_next;
392 			if (resid <= 0) {
393 				if (flags & MSG_EOR)
394 					top->m_flags |= M_EOR;
395 				break;
396 			}
397 		    } while (space > 0 && atomic);
398 		    if (dontroute)
399 			    so->so_options |= SO_DONTROUTE;
400 		    s = splnet();				/* XXX */
401 		    error = (*so->so_proto->pr_usrreq)(so,
402 			(flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
403 			top, addr, control);
404 		    splx(s);
405 		    if (dontroute)
406 			    so->so_options &= ~SO_DONTROUTE;
407 		    clen = 0;
408 		    control = 0;
409 		    top = 0;
410 		    mp = &top;
411 		    if (error)
412 			goto release;
413 		} while (resid && space > 0);
414 	} while (resid);
415 
416 release:
417 	sbunlock(&so->so_snd);
418 out:
419 	if (top)
420 		m_freem(top);
421 	if (control)
422 		m_freem(control);
423 	return (error);
424 }
425 
426 /*
427  * Implement receive operations on a socket.
428  * We depend on the way that records are added to the sockbuf
429  * by sbappend*.  In particular, each record (mbufs linked through m_next)
430  * must begin with an address if the protocol so specifies,
431  * followed by an optional mbuf or mbufs containing ancillary data,
432  * and then zero or more mbufs of data.
433  * In order to avoid blocking network interrupts for the entire time here,
434  * we splx() while doing the actual copy to user space.
435  * Although the sockbuf is locked, new data may still be appended,
436  * and thus we must maintain consistency of the sockbuf during that time.
437  *
438  * The caller may receive the data as a single mbuf chain by supplying
439  * an mbuf **mp0 for use in returning the chain.  The uio is then used
440  * only for the count in uio_resid.
441  */
442 soreceive(so, paddr, uio, mp0, controlp, flagsp)
443 	register struct socket *so;
444 	struct mbuf **paddr;
445 	struct uio *uio;
446 	struct mbuf **mp0;
447 	struct mbuf **controlp;
448 	int *flagsp;
449 {
450 	register struct mbuf *m, **mp;
451 	register int flags, len, error, s, offset;
452 	struct protosw *pr = so->so_proto;
453 	struct mbuf *nextrecord;
454 	int moff, type;
455 
456 	mp = mp0;
457 	if (paddr)
458 		*paddr = 0;
459 	if (controlp)
460 		*controlp = 0;
461 	if (flagsp)
462 		flags = *flagsp &~ MSG_EOR;
463 	else
464 		flags = 0;
465 	if (flags & MSG_OOB) {
466 		m = m_get(M_WAIT, MT_DATA);
467 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB,
468 		    m, (struct mbuf *)(flags & MSG_PEEK), (struct mbuf *)0);
469 		if (error)
470 			goto bad;
471 		do {
472 			error = uiomove(mtod(m, caddr_t),
473 			    (int) min(uio->uio_resid, m->m_len), uio);
474 			m = m_free(m);
475 		} while (uio->uio_resid && error == 0 && m);
476 bad:
477 		if (m)
478 			m_freem(m);
479 		return (error);
480 	}
481 	if (mp)
482 		*mp = (struct mbuf *)0;
483 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
484 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
485 		    (struct mbuf *)0, (struct mbuf *)0);
486 
487 restart:
488 	if (error = sblock(&so->so_rcv))
489 		return (error);
490 	s = splnet();
491 
492 	m = so->so_rcv.sb_mb;
493 	/*
494 	 * If we have less data than requested, block awaiting more
495 	 * (subject to any timeout) if:
496 	 *   1. the current count is less than the low water mark, or
497 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
498 	 *	receive operation at once if we block (resid <= hiwat).
499 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
500 	 * we have to do the receive in sections, and thus risk returning
501 	 * a short count if a timeout or signal occurs after we start.
502 	 */
503 	if (m == 0 || so->so_rcv.sb_cc < uio->uio_resid &&
504 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
505 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat))) {
506 #ifdef DIAGNOSTIC
507 		if (m == 0 && so->so_rcv.sb_cc)
508 			panic("receive 1");
509 #endif
510 		if (so->so_error) {
511 			error = so->so_error;
512 			so->so_error = 0;
513 			goto release;
514 		}
515 		if (so->so_state & SS_CANTRCVMORE)
516 			goto release;
517 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
518 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
519 			error = ENOTCONN;
520 			goto release;
521 		}
522 		if (uio->uio_resid == 0)
523 			goto release;
524 		if (so->so_state & SS_NBIO) {
525 			error = EWOULDBLOCK;
526 			goto release;
527 		}
528 		sbunlock(&so->so_rcv);
529 		error = sbwait(&so->so_rcv);
530 		splx(s);
531 		if (error)
532 			return (error);
533 		goto restart;
534 	}
535 	u.u_ru.ru_msgrcv++;
536 	nextrecord = m->m_nextpkt;
537 	if (pr->pr_flags & PR_ADDR) {
538 #ifdef DIAGNOSTIC
539 		if (m->m_type != MT_SONAME)
540 			panic("receive 1a");
541 #endif
542 		if (flags & MSG_PEEK) {
543 			if (paddr)
544 				*paddr = m_copy(m, 0, m->m_len);
545 			m = m->m_next;
546 		} else {
547 			sbfree(&so->so_rcv, m);
548 			if (paddr) {
549 				*paddr = m;
550 				so->so_rcv.sb_mb = m->m_next;
551 				m->m_next = 0;
552 				m = so->so_rcv.sb_mb;
553 			} else {
554 				MFREE(m, so->so_rcv.sb_mb);
555 				m = so->so_rcv.sb_mb;
556 			}
557 		}
558 	}
559 	while (m && m->m_type == MT_CONTROL && error == 0) {
560 		if (flags & MSG_PEEK) {
561 			if (controlp)
562 				*controlp = m_copy(m, 0, m->m_len);
563 			m = m->m_next;
564 		} else {
565 			sbfree(&so->so_rcv, m);
566 			if (controlp) {
567 				if (pr->pr_domain->dom_externalize &&
568 				    mtod(m, struct cmsghdr *)->cmsg_type ==
569 				    SCM_RIGHTS)
570 				   error = (*pr->pr_domain->dom_externalize)(m);
571 				*controlp = m;
572 				so->so_rcv.sb_mb = m->m_next;
573 				m->m_next = 0;
574 				m = so->so_rcv.sb_mb;
575 			} else {
576 				MFREE(m, so->so_rcv.sb_mb);
577 				m = so->so_rcv.sb_mb;
578 			}
579 		}
580 		if (controlp)
581 			controlp = &(*controlp)->m_next;
582 	}
583 	if (m) {
584 		if ((flags & MSG_PEEK) == 0)
585 			m->m_nextpkt = nextrecord;
586 		type = m->m_type;
587 	}
588 	moff = 0;
589 	offset = 0;
590 	while (m && m->m_type == type && uio->uio_resid > 0 && error == 0) {
591 		if (m->m_type == MT_OOBDATA)
592 			flags |= MSG_OOB;
593 #ifdef DIAGNOSTIC
594 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
595 			panic("receive 3");
596 #endif
597 		type = m->m_type;
598 		so->so_state &= ~SS_RCVATMARK;
599 		len = uio->uio_resid;
600 		if (so->so_oobmark && len > so->so_oobmark - offset)
601 			len = so->so_oobmark - offset;
602 		if (len > m->m_len - moff)
603 			len = m->m_len - moff;
604 		/*
605 		 * If mp is set, just pass back the mbufs.
606 		 * Otherwise copy them out via the uio, then free.
607 		 * Sockbuf must be consistent here (points to current mbuf,
608 		 * it points to next record) when we drop priority;
609 		 * we must note any additions to the sockbuf when we
610 		 * block interrupts again.
611 		 */
612 		if (mp == 0) {
613 			splx(s);
614 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
615 			s = splnet();
616 		} else
617 			uio->uio_resid -= len;
618 		if (len == m->m_len - moff) {
619 			if (m->m_flags & M_EOR)
620 				flags |= MSG_EOR;
621 			if (flags & MSG_PEEK) {
622 				m = m->m_next;
623 				moff = 0;
624 			} else {
625 				nextrecord = m->m_nextpkt;
626 				sbfree(&so->so_rcv, m);
627 				if (mp) {
628 					*mp = m;
629 					mp = &m->m_next;
630 					so->so_rcv.sb_mb = m = m->m_next;
631 					*mp = (struct mbuf *)0;
632 				} else {
633 					MFREE(m, so->so_rcv.sb_mb);
634 					m = so->so_rcv.sb_mb;
635 				}
636 				if (m)
637 					m->m_nextpkt = nextrecord;
638 			}
639 		} else {
640 			if (flags & MSG_PEEK)
641 				moff += len;
642 			else {
643 				if (mp)
644 					*mp = m_copym(m, 0, len, M_WAIT);
645 				m->m_data += len;
646 				m->m_len -= len;
647 				so->so_rcv.sb_cc -= len;
648 			}
649 		}
650 		if (so->so_oobmark) {
651 			if ((flags & MSG_PEEK) == 0) {
652 				so->so_oobmark -= len;
653 				if (so->so_oobmark == 0) {
654 					so->so_state |= SS_RCVATMARK;
655 					break;
656 				}
657 			} else
658 				offset += len;
659 		}
660 		if (flags & MSG_EOR)
661 			break;
662 		/*
663 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
664 		 * we must not quit until "uio->uio_resid == 0" or an error
665 		 * termination.  If a signal/timeout occurs, return
666 		 * with a short count but without error.
667 		 * Keep sockbuf locked against other readers.
668 		 */
669 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
670 		    !sosendallatonce(so)) {
671 			error = sbwait(&so->so_rcv);
672 			if (error) {
673 				sbunlock(&so->so_rcv);
674 				splx(s);
675 				return (0);
676 			}
677 			if (m = so->so_rcv.sb_mb)
678 				nextrecord = m->m_nextpkt;
679 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
680 				break;
681 			continue;
682 		}
683 	}
684 	if ((flags & MSG_PEEK) == 0) {
685 		if (m == 0)
686 			so->so_rcv.sb_mb = nextrecord;
687 		else if (pr->pr_flags & PR_ATOMIC) {
688 			flags |= MSG_TRUNC;
689 			(void) sbdroprecord(&so->so_rcv);
690 		}
691 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
692 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
693 			    (struct mbuf *)flags, (struct mbuf *)0,
694 			    (struct mbuf *)0);
695 	}
696 	if (flagsp)
697 		*flagsp |= flags;
698 release:
699 	sbunlock(&so->so_rcv);
700 	splx(s);
701 	return (error);
702 }
703 
704 soshutdown(so, how)
705 	register struct socket *so;
706 	register int how;
707 {
708 	register struct protosw *pr = so->so_proto;
709 
710 	how++;
711 	if (how & FREAD)
712 		sorflush(so);
713 	if (how & FWRITE)
714 		return ((*pr->pr_usrreq)(so, PRU_SHUTDOWN,
715 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
716 	return (0);
717 }
718 
719 sorflush(so)
720 	register struct socket *so;
721 {
722 	register struct sockbuf *sb = &so->so_rcv;
723 	register struct protosw *pr = so->so_proto;
724 	register int s;
725 	struct sockbuf asb;
726 
727 	sb->sb_flags |= SB_NOINTR;
728 	(void) sblock(sb);
729 	s = splimp();
730 	socantrcvmore(so);
731 	sbunlock(sb);
732 	asb = *sb;
733 	bzero((caddr_t)sb, sizeof (*sb));
734 	splx(s);
735 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
736 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
737 	sbrelease(&asb);
738 }
739 
740 sosetopt(so, level, optname, m0)
741 	register struct socket *so;
742 	int level, optname;
743 	struct mbuf *m0;
744 {
745 	int error = 0;
746 	register struct mbuf *m = m0;
747 
748 	if (level != SOL_SOCKET) {
749 		if (so->so_proto && so->so_proto->pr_ctloutput)
750 			return ((*so->so_proto->pr_ctloutput)
751 				  (PRCO_SETOPT, so, level, optname, &m0));
752 		error = ENOPROTOOPT;
753 	} else {
754 		switch (optname) {
755 
756 		case SO_LINGER:
757 			if (m == NULL || m->m_len != sizeof (struct linger)) {
758 				error = EINVAL;
759 				goto bad;
760 			}
761 			so->so_linger = mtod(m, struct linger *)->l_linger;
762 			/* fall thru... */
763 
764 		case SO_DEBUG:
765 		case SO_KEEPALIVE:
766 		case SO_DONTROUTE:
767 		case SO_USELOOPBACK:
768 		case SO_BROADCAST:
769 		case SO_REUSEADDR:
770 		case SO_OOBINLINE:
771 			if (m == NULL || m->m_len < sizeof (int)) {
772 				error = EINVAL;
773 				goto bad;
774 			}
775 			if (*mtod(m, int *))
776 				so->so_options |= optname;
777 			else
778 				so->so_options &= ~optname;
779 			break;
780 
781 		case SO_SNDBUF:
782 		case SO_RCVBUF:
783 		case SO_SNDLOWAT:
784 		case SO_RCVLOWAT:
785 		case SO_SNDTIMEO:
786 		case SO_RCVTIMEO:
787 			if (m == NULL || m->m_len < sizeof (int)) {
788 				error = EINVAL;
789 				goto bad;
790 			}
791 			switch (optname) {
792 
793 			case SO_SNDBUF:
794 			case SO_RCVBUF:
795 				if (sbreserve(optname == SO_SNDBUF ?
796 				    &so->so_snd : &so->so_rcv,
797 				    (u_long) *mtod(m, int *)) == 0) {
798 					error = ENOBUFS;
799 					goto bad;
800 				}
801 				break;
802 
803 			case SO_SNDLOWAT:
804 				so->so_snd.sb_lowat = *mtod(m, int *);
805 				break;
806 			case SO_RCVLOWAT:
807 				so->so_rcv.sb_lowat = *mtod(m, int *);
808 				break;
809 			case SO_SNDTIMEO:
810 				so->so_snd.sb_timeo = *mtod(m, int *);
811 				break;
812 			case SO_RCVTIMEO:
813 				so->so_rcv.sb_timeo = *mtod(m, int *);
814 				break;
815 			}
816 			break;
817 
818 		default:
819 			error = ENOPROTOOPT;
820 			break;
821 		}
822 	}
823 bad:
824 	if (m)
825 		(void) m_free(m);
826 	return (error);
827 }
828 
829 sogetopt(so, level, optname, mp)
830 	register struct socket *so;
831 	int level, optname;
832 	struct mbuf **mp;
833 {
834 	register struct mbuf *m;
835 
836 	if (level != SOL_SOCKET) {
837 		if (so->so_proto && so->so_proto->pr_ctloutput) {
838 			return ((*so->so_proto->pr_ctloutput)
839 				  (PRCO_GETOPT, so, level, optname, mp));
840 		} else
841 			return (ENOPROTOOPT);
842 	} else {
843 		m = m_get(M_WAIT, MT_SOOPTS);
844 		m->m_len = sizeof (int);
845 
846 		switch (optname) {
847 
848 		case SO_LINGER:
849 			m->m_len = sizeof (struct linger);
850 			mtod(m, struct linger *)->l_onoff =
851 				so->so_options & SO_LINGER;
852 			mtod(m, struct linger *)->l_linger = so->so_linger;
853 			break;
854 
855 		case SO_USELOOPBACK:
856 		case SO_DONTROUTE:
857 		case SO_DEBUG:
858 		case SO_KEEPALIVE:
859 		case SO_REUSEADDR:
860 		case SO_BROADCAST:
861 		case SO_OOBINLINE:
862 			*mtod(m, int *) = so->so_options & optname;
863 			break;
864 
865 		case SO_TYPE:
866 			*mtod(m, int *) = so->so_type;
867 			break;
868 
869 		case SO_ERROR:
870 			*mtod(m, int *) = so->so_error;
871 			so->so_error = 0;
872 			break;
873 
874 		case SO_SNDBUF:
875 			*mtod(m, int *) = so->so_snd.sb_hiwat;
876 			break;
877 
878 		case SO_RCVBUF:
879 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
880 			break;
881 
882 		case SO_SNDLOWAT:
883 			*mtod(m, int *) = so->so_snd.sb_lowat;
884 			break;
885 
886 		case SO_RCVLOWAT:
887 			*mtod(m, int *) = so->so_rcv.sb_lowat;
888 			break;
889 
890 		case SO_SNDTIMEO:
891 			*mtod(m, int *) = so->so_snd.sb_timeo;
892 			break;
893 
894 		case SO_RCVTIMEO:
895 			*mtod(m, int *) = so->so_rcv.sb_timeo;
896 			break;
897 
898 		default:
899 			(void)m_free(m);
900 			return (ENOPROTOOPT);
901 		}
902 		*mp = m;
903 		return (0);
904 	}
905 }
906 
907 sohasoutofband(so)
908 	register struct socket *so;
909 {
910 	struct proc *p;
911 
912 	if (so->so_pgid < 0)
913 		gsignal(-so->so_pgid, SIGURG);
914 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
915 		psignal(p, SIGURG);
916 	if (so->so_rcv.sb_sel) {
917 		selwakeup(so->so_rcv.sb_sel, so->so_rcv.sb_flags & SB_COLL);
918 		so->so_rcv.sb_sel = 0;
919 		so->so_rcv.sb_flags &= ~SB_COLL;
920 	}
921 }
922