xref: /csrg-svn/sys/kern/uipc_socket.c (revision 50942)
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	7.29 (Berkeley) 08/30/91
8  */
9 
10 #include "param.h"
11 #include "proc.h"
12 #include "file.h"
13 #include "malloc.h"
14 #include "mbuf.h"
15 #include "domain.h"
16 #include "kernel.h"
17 #include "protosw.h"
18 #include "socket.h"
19 #include "socketvar.h"
20 #include "resourcevar.h"
21 
22 /*
23  * Socket operation routines.
24  * These routines are called by the routines in
25  * sys_socket.c or from a system process, and
26  * implement the semantics of socket operations by
27  * switching out to the protocol specific routines.
28  */
29 /*ARGSUSED*/
30 socreate(dom, aso, type, proto)
31 	struct socket **aso;
32 	register int type;
33 	int proto;
34 {
35 	struct proc *p = curproc;		/* XXX */
36 	register struct protosw *prp;
37 	register struct socket *so;
38 	register int error;
39 
40 	if (proto)
41 		prp = pffindproto(dom, proto, type);
42 	else
43 		prp = pffindtype(dom, type);
44 	if (prp == 0)
45 		return (EPROTONOSUPPORT);
46 	if (prp->pr_type != type)
47 		return (EPROTOTYPE);
48 	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_WAIT);
49 	bzero((caddr_t)so, sizeof(*so));
50 	so->so_type = type;
51 	if (p->p_ucred->cr_uid == 0)
52 		so->so_state = SS_PRIV;
53 	so->so_proto = prp;
54 	error =
55 	    (*prp->pr_usrreq)(so, PRU_ATTACH,
56 		(struct mbuf *)0, (struct mbuf *)proto, (struct mbuf *)0);
57 	if (error) {
58 		so->so_state |= SS_NOFDREF;
59 		sofree(so);
60 		return (error);
61 	}
62 	*aso = so;
63 	return (0);
64 }
65 
66 sobind(so, nam)
67 	struct socket *so;
68 	struct mbuf *nam;
69 {
70 	int s = splnet();
71 	int error;
72 
73 	error =
74 	    (*so->so_proto->pr_usrreq)(so, PRU_BIND,
75 		(struct mbuf *)0, nam, (struct mbuf *)0);
76 	splx(s);
77 	return (error);
78 }
79 
80 solisten(so, backlog)
81 	register struct socket *so;
82 	int backlog;
83 {
84 	int s = splnet(), error;
85 
86 	error =
87 	    (*so->so_proto->pr_usrreq)(so, PRU_LISTEN,
88 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
89 	if (error) {
90 		splx(s);
91 		return (error);
92 	}
93 	if (so->so_q == 0)
94 		so->so_options |= SO_ACCEPTCONN;
95 	if (backlog < 0)
96 		backlog = 0;
97 	so->so_qlimit = min(backlog, SOMAXCONN);
98 	splx(s);
99 	return (0);
100 }
101 
102 sofree(so)
103 	register struct socket *so;
104 {
105 
106 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
107 		return;
108 	if (so->so_head) {
109 		if (!soqremque(so, 0) && !soqremque(so, 1))
110 			panic("sofree dq");
111 		so->so_head = 0;
112 	}
113 	sbrelease(&so->so_snd);
114 	sorflush(so);
115 	FREE(so, M_SOCKET);
116 }
117 
118 /*
119  * Close a socket on last file table reference removal.
120  * Initiate disconnect if connected.
121  * Free socket when disconnect complete.
122  */
123 soclose(so)
124 	register struct socket *so;
125 {
126 	int s = splnet();		/* conservative */
127 	int error = 0;
128 
129 	if (so->so_options & SO_ACCEPTCONN) {
130 		while (so->so_q0)
131 			(void) soabort(so->so_q0);
132 		while (so->so_q)
133 			(void) soabort(so->so_q);
134 	}
135 	if (so->so_pcb == 0)
136 		goto discard;
137 	if (so->so_state & SS_ISCONNECTED) {
138 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
139 			error = sodisconnect(so);
140 			if (error)
141 				goto drop;
142 		}
143 		if (so->so_options & SO_LINGER) {
144 			if ((so->so_state & SS_ISDISCONNECTING) &&
145 			    (so->so_state & SS_NBIO))
146 				goto drop;
147 			while (so->so_state & SS_ISCONNECTED)
148 				if (error = tsleep((caddr_t)&so->so_timeo,
149 				    PSOCK | PCATCH, netcls, so->so_linger))
150 					break;
151 		}
152 	}
153 drop:
154 	if (so->so_pcb) {
155 		int error2 =
156 		    (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
157 			(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
158 		if (error == 0)
159 			error = error2;
160 	}
161 discard:
162 	if (so->so_state & SS_NOFDREF)
163 		panic("soclose: NOFDREF");
164 	so->so_state |= SS_NOFDREF;
165 	sofree(so);
166 	splx(s);
167 	return (error);
168 }
169 
170 /*
171  * Must be called at splnet...
172  */
173 soabort(so)
174 	struct socket *so;
175 {
176 
177 	return (
178 	    (*so->so_proto->pr_usrreq)(so, PRU_ABORT,
179 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
180 }
181 
182 soaccept(so, nam)
183 	register struct socket *so;
184 	struct mbuf *nam;
185 {
186 	int s = splnet();
187 	int error;
188 
189 	if ((so->so_state & SS_NOFDREF) == 0)
190 		panic("soaccept: !NOFDREF");
191 	so->so_state &= ~SS_NOFDREF;
192 	error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
193 	    (struct mbuf *)0, nam, (struct mbuf *)0);
194 	splx(s);
195 	return (error);
196 }
197 
198 soconnect(so, nam)
199 	register struct socket *so;
200 	struct mbuf *nam;
201 {
202 	int s;
203 	int error;
204 
205 	if (so->so_options & SO_ACCEPTCONN)
206 		return (EOPNOTSUPP);
207 	s = splnet();
208 	/*
209 	 * If protocol is connection-based, can only connect once.
210 	 * Otherwise, if connected, try to disconnect first.
211 	 * This allows user to disconnect by connecting to, e.g.,
212 	 * a null address.
213 	 */
214 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
215 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
216 	    (error = sodisconnect(so))))
217 		error = EISCONN;
218 	else
219 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
220 		    (struct mbuf *)0, nam, (struct mbuf *)0);
221 	splx(s);
222 	return (error);
223 }
224 
225 soconnect2(so1, so2)
226 	register struct socket *so1;
227 	struct socket *so2;
228 {
229 	int s = splnet();
230 	int error;
231 
232 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
233 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0);
234 	splx(s);
235 	return (error);
236 }
237 
238 sodisconnect(so)
239 	register struct socket *so;
240 {
241 	int s = splnet();
242 	int error;
243 
244 	if ((so->so_state & SS_ISCONNECTED) == 0) {
245 		error = ENOTCONN;
246 		goto bad;
247 	}
248 	if (so->so_state & SS_ISDISCONNECTING) {
249 		error = EALREADY;
250 		goto bad;
251 	}
252 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
253 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
254 bad:
255 	splx(s);
256 	return (error);
257 }
258 
259 /*
260  * Send on a socket.
261  * If send must go all at once and message is larger than
262  * send buffering, then hard error.
263  * Lock against other senders.
264  * If must go all at once and not enough room now, then
265  * inform user that this would block and do nothing.
266  * Otherwise, if nonblocking, send as much as possible.
267  * The data to be sent is described by "uio" if nonzero,
268  * otherwise by the mbuf chain "top" (which must be null
269  * if uio is not).  Data provided in mbuf chain must be small
270  * enough to send all at once.
271  *
272  * Returns nonzero on error, timeout or signal; callers
273  * must check for short counts if EINTR/ERESTART are returned.
274  * Data and control buffers are freed on return.
275  */
276 sosend(so, addr, uio, top, control, flags)
277 	register struct socket *so;
278 	struct mbuf *addr;
279 	struct uio *uio;
280 	struct mbuf *top;
281 	struct mbuf *control;
282 	int flags;
283 {
284 	struct proc *p = curproc;		/* XXX */
285 	struct mbuf **mp;
286 	register struct mbuf *m;
287 	register long space, len, resid;
288 	int clen = 0, error, s, dontroute, mlen;
289 	int atomic = sosendallatonce(so) || top;
290 
291 	if (uio)
292 		resid = uio->uio_resid;
293 	else
294 		resid = top->m_pkthdr.len;
295 	dontroute =
296 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
297 	    (so->so_proto->pr_flags & PR_ATOMIC);
298 	p->p_stats->p_ru.ru_msgsnd++;
299 	if (control)
300 		clen = control->m_len;
301 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
302 
303 restart:
304 	if (error = sblock(&so->so_snd))
305 		goto out;
306 	do {
307 		s = splnet();
308 		if (so->so_state & SS_CANTSENDMORE)
309 			snderr(EPIPE);
310 		if (so->so_error)
311 			snderr(so->so_error);
312 		if ((so->so_state & SS_ISCONNECTED) == 0) {
313 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
314 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
315 				    !(resid == 0 && clen != 0))
316 					snderr(ENOTCONN);
317 			} else if (addr == 0)
318 				snderr(EDESTADDRREQ);
319 		}
320 		space = sbspace(&so->so_snd);
321 		if (flags & MSG_OOB)
322 			space += 1024;
323 		if (space < resid + clen &&
324 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
325 			if (atomic && resid > so->so_snd.sb_hiwat ||
326 			    clen > so->so_snd.sb_hiwat)
327 				snderr(EMSGSIZE);
328 			if (so->so_state & SS_NBIO)
329 				snderr(EWOULDBLOCK);
330 			sbunlock(&so->so_snd);
331 			error = sbwait(&so->so_snd);
332 			splx(s);
333 			if (error)
334 				goto out;
335 			goto restart;
336 		}
337 		splx(s);
338 		mp = &top;
339 		space -= clen;
340 		do {
341 		    if (uio == NULL) {
342 			/*
343 			 * Data is prepackaged in "top".
344 			 */
345 			resid = 0;
346 			if (flags & MSG_EOR)
347 				top->m_flags |= M_EOR;
348 		    } else do {
349 			if (top == 0) {
350 				MGETHDR(m, M_WAIT, MT_DATA);
351 				mlen = MHLEN;
352 				m->m_pkthdr.len = 0;
353 				m->m_pkthdr.rcvif = (struct ifnet *)0;
354 			} else {
355 				MGET(m, M_WAIT, MT_DATA);
356 				mlen = MLEN;
357 			}
358 			if (resid >= MINCLSIZE && space >= MCLBYTES) {
359 				MCLGET(m, M_WAIT);
360 				if ((m->m_flags & M_EXT) == 0)
361 					goto nopages;
362 				mlen = MCLBYTES;
363 #ifdef	MAPPED_MBUFS
364 				len = min(MCLBYTES, resid);
365 #else
366 				if (top == 0) {
367 					len = min(MCLBYTES - max_hdr, resid);
368 					m->m_data += max_hdr;
369 				} else
370 					len = min(MCLBYTES, resid);
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), (int)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 	struct proc *p = curproc;		/* XXX */
451 	register struct mbuf *m, **mp;
452 	register int flags, len, error, s, offset;
453 	struct protosw *pr = so->so_proto;
454 	struct mbuf *nextrecord;
455 	int moff, type;
456 
457 	mp = mp0;
458 	if (paddr)
459 		*paddr = 0;
460 	if (controlp)
461 		*controlp = 0;
462 	if (flagsp)
463 		flags = *flagsp &~ MSG_EOR;
464 	else
465 		flags = 0;
466 	if (flags & MSG_OOB) {
467 		m = m_get(M_WAIT, MT_DATA);
468 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB,
469 		    m, (struct mbuf *)(flags & MSG_PEEK), (struct mbuf *)0);
470 		if (error)
471 			goto bad;
472 		do {
473 			error = uiomove(mtod(m, caddr_t),
474 			    (int) min(uio->uio_resid, m->m_len), uio);
475 			m = m_free(m);
476 		} while (uio->uio_resid && error == 0 && m);
477 bad:
478 		if (m)
479 			m_freem(m);
480 		return (error);
481 	}
482 	if (mp)
483 		*mp = (struct mbuf *)0;
484 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
485 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
486 		    (struct mbuf *)0, (struct mbuf *)0);
487 
488 restart:
489 	if (error = sblock(&so->so_rcv))
490 		return (error);
491 	s = splnet();
492 
493 	m = so->so_rcv.sb_mb;
494 	/*
495 	 * If we have less data than requested, block awaiting more
496 	 * (subject to any timeout) if:
497 	 *   1. the current count is less than the low water mark, or
498 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
499 	 *	receive operation at once if we block (resid <= hiwat).
500 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
501 	 * we have to do the receive in sections, and thus risk returning
502 	 * a short count if a timeout or signal occurs after we start.
503 	 */
504 	while (m == 0 || so->so_rcv.sb_cc < uio->uio_resid &&
505 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
506 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
507 	    m->m_nextpkt == 0) {
508 #ifdef DIAGNOSTIC
509 		if (m == 0 && so->so_rcv.sb_cc)
510 			panic("receive 1");
511 #endif
512 		if (so->so_error) {
513 			if (m)
514 				break;
515 			error = so->so_error;
516 			if ((flags & MSG_PEEK) == 0)
517 				so->so_error = 0;
518 			goto release;
519 		}
520 		if (so->so_state & SS_CANTRCVMORE) {
521 			if (m)
522 				break;
523 			else
524 				goto release;
525 		}
526 		for (; m; m = m->m_next)
527 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
528 				m = so->so_rcv.sb_mb;
529 				goto dontblock;
530 			}
531 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
532 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
533 			error = ENOTCONN;
534 			goto release;
535 		}
536 		if (uio->uio_resid == 0)
537 			goto release;
538 		if (so->so_state & SS_NBIO) {
539 			error = EWOULDBLOCK;
540 			goto release;
541 		}
542 		sbunlock(&so->so_rcv);
543 		error = sbwait(&so->so_rcv);
544 		splx(s);
545 		if (error)
546 			return (error);
547 		goto restart;
548 	}
549 dontblock:
550 	p->p_stats->p_ru.ru_msgrcv++;
551 	nextrecord = m->m_nextpkt;
552 	if (pr->pr_flags & PR_ADDR) {
553 #ifdef DIAGNOSTIC
554 		if (m->m_type != MT_SONAME)
555 			panic("receive 1a");
556 #endif
557 		if (flags & MSG_PEEK) {
558 			if (paddr)
559 				*paddr = m_copy(m, 0, m->m_len);
560 			m = m->m_next;
561 		} else {
562 			sbfree(&so->so_rcv, m);
563 			if (paddr) {
564 				*paddr = m;
565 				so->so_rcv.sb_mb = m->m_next;
566 				m->m_next = 0;
567 				m = so->so_rcv.sb_mb;
568 			} else {
569 				MFREE(m, so->so_rcv.sb_mb);
570 				m = so->so_rcv.sb_mb;
571 			}
572 		}
573 	}
574 	while (m && m->m_type == MT_CONTROL && error == 0) {
575 		if (flags & MSG_PEEK) {
576 			if (controlp)
577 				*controlp = m_copy(m, 0, m->m_len);
578 			m = m->m_next;
579 		} else {
580 			sbfree(&so->so_rcv, m);
581 			if (controlp) {
582 				if (pr->pr_domain->dom_externalize &&
583 				    mtod(m, struct cmsghdr *)->cmsg_type ==
584 				    SCM_RIGHTS)
585 				   error = (*pr->pr_domain->dom_externalize)(m);
586 				*controlp = m;
587 				so->so_rcv.sb_mb = m->m_next;
588 				m->m_next = 0;
589 				m = so->so_rcv.sb_mb;
590 			} else {
591 				MFREE(m, so->so_rcv.sb_mb);
592 				m = so->so_rcv.sb_mb;
593 			}
594 		}
595 		if (controlp)
596 			controlp = &(*controlp)->m_next;
597 	}
598 	if (m) {
599 		if ((flags & MSG_PEEK) == 0)
600 			m->m_nextpkt = nextrecord;
601 		type = m->m_type;
602 		if (type == MT_OOBDATA)
603 			flags |= MSG_OOB;
604 	}
605 	moff = 0;
606 	offset = 0;
607 	while (m && uio->uio_resid > 0 && error == 0) {
608 		if (m->m_type == MT_OOBDATA) {
609 			if (type != MT_OOBDATA)
610 				break;
611 		} else if (type == MT_OOBDATA)
612 			break;
613 #ifdef DIAGNOSTIC
614 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
615 			panic("receive 3");
616 #endif
617 		so->so_state &= ~SS_RCVATMARK;
618 		len = uio->uio_resid;
619 		if (so->so_oobmark && len > so->so_oobmark - offset)
620 			len = so->so_oobmark - offset;
621 		if (len > m->m_len - moff)
622 			len = m->m_len - moff;
623 		/*
624 		 * If mp is set, just pass back the mbufs.
625 		 * Otherwise copy them out via the uio, then free.
626 		 * Sockbuf must be consistent here (points to current mbuf,
627 		 * it points to next record) when we drop priority;
628 		 * we must note any additions to the sockbuf when we
629 		 * block interrupts again.
630 		 */
631 		if (mp == 0) {
632 			splx(s);
633 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
634 			s = splnet();
635 		} else
636 			uio->uio_resid -= len;
637 		if (len == m->m_len - moff) {
638 			if (m->m_flags & M_EOR)
639 				flags |= MSG_EOR;
640 			if (flags & MSG_PEEK) {
641 				m = m->m_next;
642 				moff = 0;
643 			} else {
644 				nextrecord = m->m_nextpkt;
645 				sbfree(&so->so_rcv, m);
646 				if (mp) {
647 					*mp = m;
648 					mp = &m->m_next;
649 					so->so_rcv.sb_mb = m = m->m_next;
650 					*mp = (struct mbuf *)0;
651 				} else {
652 					MFREE(m, so->so_rcv.sb_mb);
653 					m = so->so_rcv.sb_mb;
654 				}
655 				if (m)
656 					m->m_nextpkt = nextrecord;
657 			}
658 		} else {
659 			if (flags & MSG_PEEK)
660 				moff += len;
661 			else {
662 				if (mp)
663 					*mp = m_copym(m, 0, len, M_WAIT);
664 				m->m_data += len;
665 				m->m_len -= len;
666 				so->so_rcv.sb_cc -= len;
667 			}
668 		}
669 		if (so->so_oobmark) {
670 			if ((flags & MSG_PEEK) == 0) {
671 				so->so_oobmark -= len;
672 				if (so->so_oobmark == 0) {
673 					so->so_state |= SS_RCVATMARK;
674 					break;
675 				}
676 			} else
677 				offset += len;
678 		}
679 		if (flags & MSG_EOR)
680 			break;
681 		/*
682 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
683 		 * we must not quit until "uio->uio_resid == 0" or an error
684 		 * termination.  If a signal/timeout occurs, return
685 		 * with a short count but without error.
686 		 * Keep sockbuf locked against other readers.
687 		 */
688 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
689 		    !sosendallatonce(so)) {
690 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
691 				break;
692 			error = sbwait(&so->so_rcv);
693 			if (error) {
694 				sbunlock(&so->so_rcv);
695 				splx(s);
696 				return (0);
697 			}
698 			if (m = so->so_rcv.sb_mb)
699 				nextrecord = m->m_nextpkt;
700 		}
701 	}
702 	if ((flags & MSG_PEEK) == 0) {
703 		if (m == 0)
704 			so->so_rcv.sb_mb = nextrecord;
705 		else if (pr->pr_flags & PR_ATOMIC) {
706 			flags |= MSG_TRUNC;
707 			(void) sbdroprecord(&so->so_rcv);
708 		}
709 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
710 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
711 			    (struct mbuf *)flags, (struct mbuf *)0,
712 			    (struct mbuf *)0);
713 	}
714 	if (flagsp)
715 		*flagsp |= flags;
716 release:
717 	sbunlock(&so->so_rcv);
718 	splx(s);
719 	return (error);
720 }
721 
722 soshutdown(so, how)
723 	register struct socket *so;
724 	register int how;
725 {
726 	register struct protosw *pr = so->so_proto;
727 
728 	how++;
729 	if (how & FREAD)
730 		sorflush(so);
731 	if (how & FWRITE)
732 		return ((*pr->pr_usrreq)(so, PRU_SHUTDOWN,
733 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
734 	return (0);
735 }
736 
737 sorflush(so)
738 	register struct socket *so;
739 {
740 	register struct sockbuf *sb = &so->so_rcv;
741 	register struct protosw *pr = so->so_proto;
742 	register int s;
743 	struct sockbuf asb;
744 
745 	sb->sb_flags |= SB_NOINTR;
746 	(void) sblock(sb);
747 	s = splimp();
748 	socantrcvmore(so);
749 	sbunlock(sb);
750 	asb = *sb;
751 	bzero((caddr_t)sb, sizeof (*sb));
752 	splx(s);
753 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
754 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
755 	sbrelease(&asb);
756 }
757 
758 sosetopt(so, level, optname, m0)
759 	register struct socket *so;
760 	int level, optname;
761 	struct mbuf *m0;
762 {
763 	int error = 0;
764 	register struct mbuf *m = m0;
765 
766 	if (level != SOL_SOCKET) {
767 		if (so->so_proto && so->so_proto->pr_ctloutput)
768 			return ((*so->so_proto->pr_ctloutput)
769 				  (PRCO_SETOPT, so, level, optname, &m0));
770 		error = ENOPROTOOPT;
771 	} else {
772 		switch (optname) {
773 
774 		case SO_LINGER:
775 			if (m == NULL || m->m_len != sizeof (struct linger)) {
776 				error = EINVAL;
777 				goto bad;
778 			}
779 			so->so_linger = mtod(m, struct linger *)->l_linger;
780 			/* fall thru... */
781 
782 		case SO_DEBUG:
783 		case SO_KEEPALIVE:
784 		case SO_DONTROUTE:
785 		case SO_USELOOPBACK:
786 		case SO_BROADCAST:
787 		case SO_REUSEADDR:
788 		case SO_OOBINLINE:
789 			if (m == NULL || m->m_len < sizeof (int)) {
790 				error = EINVAL;
791 				goto bad;
792 			}
793 			if (*mtod(m, int *))
794 				so->so_options |= optname;
795 			else
796 				so->so_options &= ~optname;
797 			break;
798 
799 		case SO_SNDBUF:
800 		case SO_RCVBUF:
801 		case SO_SNDLOWAT:
802 		case SO_RCVLOWAT:
803 			if (m == NULL || m->m_len < sizeof (int)) {
804 				error = EINVAL;
805 				goto bad;
806 			}
807 			switch (optname) {
808 
809 			case SO_SNDBUF:
810 			case SO_RCVBUF:
811 				if (sbreserve(optname == SO_SNDBUF ?
812 				    &so->so_snd : &so->so_rcv,
813 				    (u_long) *mtod(m, int *)) == 0) {
814 					error = ENOBUFS;
815 					goto bad;
816 				}
817 				break;
818 
819 			case SO_SNDLOWAT:
820 				so->so_snd.sb_lowat = *mtod(m, int *);
821 				break;
822 			case SO_RCVLOWAT:
823 				so->so_rcv.sb_lowat = *mtod(m, int *);
824 				break;
825 			}
826 			break;
827 
828 		case SO_SNDTIMEO:
829 		case SO_RCVTIMEO:
830 		    {
831 			struct timeval *tv;
832 			short val;
833 
834 			if (m == NULL || m->m_len < sizeof (*tv)) {
835 				error = EINVAL;
836 				goto bad;
837 			}
838 			tv = mtod(m, struct timeval *);
839 			if (tv->tv_sec > SHRT_MAX / hz - hz) {
840 				error = EDOM;
841 				goto bad;
842 			}
843 			val = tv->tv_sec * hz + tv->tv_usec / tick;
844 
845 			switch (optname) {
846 
847 			case SO_SNDTIMEO:
848 				so->so_snd.sb_timeo = val;
849 				break;
850 			case SO_RCVTIMEO:
851 				so->so_rcv.sb_timeo = val;
852 				break;
853 			}
854 			break;
855 		    }
856 
857 		default:
858 			error = ENOPROTOOPT;
859 			break;
860 		}
861 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput)
862 			(void) ((*so->so_proto->pr_ctloutput)
863 				  (PRCO_SETOPT, so, level, optname, &m0));
864 	}
865 bad:
866 	if (m)
867 		(void) m_free(m);
868 	return (error);
869 }
870 
871 sogetopt(so, level, optname, mp)
872 	register struct socket *so;
873 	int level, optname;
874 	struct mbuf **mp;
875 {
876 	register struct mbuf *m;
877 
878 	if (level != SOL_SOCKET) {
879 		if (so->so_proto && so->so_proto->pr_ctloutput) {
880 			return ((*so->so_proto->pr_ctloutput)
881 				  (PRCO_GETOPT, so, level, optname, mp));
882 		} else
883 			return (ENOPROTOOPT);
884 	} else {
885 		m = m_get(M_WAIT, MT_SOOPTS);
886 		m->m_len = sizeof (int);
887 
888 		switch (optname) {
889 
890 		case SO_LINGER:
891 			m->m_len = sizeof (struct linger);
892 			mtod(m, struct linger *)->l_onoff =
893 				so->so_options & SO_LINGER;
894 			mtod(m, struct linger *)->l_linger = so->so_linger;
895 			break;
896 
897 		case SO_USELOOPBACK:
898 		case SO_DONTROUTE:
899 		case SO_DEBUG:
900 		case SO_KEEPALIVE:
901 		case SO_REUSEADDR:
902 		case SO_BROADCAST:
903 		case SO_OOBINLINE:
904 			*mtod(m, int *) = so->so_options & optname;
905 			break;
906 
907 		case SO_TYPE:
908 			*mtod(m, int *) = so->so_type;
909 			break;
910 
911 		case SO_ERROR:
912 			*mtod(m, int *) = so->so_error;
913 			so->so_error = 0;
914 			break;
915 
916 		case SO_SNDBUF:
917 			*mtod(m, int *) = so->so_snd.sb_hiwat;
918 			break;
919 
920 		case SO_RCVBUF:
921 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
922 			break;
923 
924 		case SO_SNDLOWAT:
925 			*mtod(m, int *) = so->so_snd.sb_lowat;
926 			break;
927 
928 		case SO_RCVLOWAT:
929 			*mtod(m, int *) = so->so_rcv.sb_lowat;
930 			break;
931 
932 		case SO_SNDTIMEO:
933 		case SO_RCVTIMEO:
934 		    {
935 			int val = (optname == SO_SNDTIMEO ?
936 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
937 
938 			m->m_len = sizeof(struct timeval);
939 			mtod(m, struct timeval *)->tv_sec = val / hz;
940 			mtod(m, struct timeval *)->tv_usec =
941 			    (val % hz) / tick;
942 			break;
943 		    }
944 
945 		default:
946 			(void)m_free(m);
947 			return (ENOPROTOOPT);
948 		}
949 		*mp = m;
950 		return (0);
951 	}
952 }
953 
954 sohasoutofband(so)
955 	register struct socket *so;
956 {
957 	struct proc *p;
958 
959 	if (so->so_pgid < 0)
960 		gsignal(-so->so_pgid, SIGURG);
961 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
962 		psignal(p, SIGURG);
963 	if (so->so_rcv.sb_sel) {
964 		selwakeup(so->so_rcv.sb_sel, so->so_rcv.sb_flags & SB_COLL);
965 		so->so_rcv.sb_sel = 0;
966 		so->so_rcv.sb_flags &= ~SB_COLL;
967 	}
968 }
969