xref: /freebsd/sys/kern/uipc_socket.c (revision e627b39baccd1ec9129690167cf5e6d860509655)
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  *	@(#)uipc_socket.c	8.3 (Berkeley) 4/15/94
34  * $Id: uipc_socket.c,v 1.18 1996/05/09 20:14:57 wollman Exp $
35  */
36 
37 #include <sys/param.h>
38 #include <sys/queue.h>
39 #include <sys/systm.h>
40 #include <sys/proc.h>
41 #include <sys/file.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/domain.h>
45 #include <sys/kernel.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/resourcevar.h>
50 #include <sys/signalvar.h>
51 #include <sys/sysctl.h>
52 
53 static int somaxconn = SOMAXCONN;
54 SYSCTL_INT(_kern, KERN_SOMAXCONN, somaxconn, CTLFLAG_RW, &somaxconn, 0, "");
55 
56 /*
57  * Socket operation routines.
58  * These routines are called by the routines in
59  * sys_socket.c or from a system process, and
60  * implement the semantics of socket operations by
61  * switching out to the protocol specific routines.
62  */
63 /*ARGSUSED*/
64 int
65 socreate(dom, aso, type, proto, p)
66 	int dom;
67 	struct socket **aso;
68 	register int type;
69 	int proto;
70 	struct proc *p;
71 {
72 	register struct protosw *prp;
73 	register struct socket *so;
74 	register int error;
75 
76 	if (proto)
77 		prp = pffindproto(dom, proto, type);
78 	else
79 		prp = pffindtype(dom, type);
80 	if (prp == 0 || prp->pr_usrreqs == 0)
81 		return (EPROTONOSUPPORT);
82 	if (prp->pr_type != type)
83 		return (EPROTOTYPE);
84 	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_WAIT);
85 	bzero((caddr_t)so, sizeof(*so));
86 	TAILQ_INIT(&so->so_incomp);
87 	TAILQ_INIT(&so->so_comp);
88 	so->so_type = type;
89 	if (p->p_ucred->cr_uid == 0)
90 		so->so_state = SS_PRIV;
91 	so->so_proto = prp;
92 	error = (*prp->pr_usrreqs->pru_attach)(so, proto);
93 	if (error) {
94 		so->so_state |= SS_NOFDREF;
95 		sofree(so);
96 		return (error);
97 	}
98 	*aso = so;
99 	return (0);
100 }
101 
102 int
103 sobind(so, nam)
104 	struct socket *so;
105 	struct mbuf *nam;
106 {
107 	int s = splnet();
108 	int error;
109 
110 	error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam);
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 = (*so->so_proto->pr_usrreqs->pru_listen)(so);
123 	if (error) {
124 		splx(s);
125 		return (error);
126 	}
127 	if (so->so_comp.tqh_first == NULL)
128 		so->so_options |= SO_ACCEPTCONN;
129 	if (backlog < 0 || backlog > somaxconn)
130 		backlog = somaxconn;
131 	so->so_qlimit = backlog;
132 	splx(s);
133 	return (0);
134 }
135 
136 void
137 sofree(so)
138 	register struct socket *so;
139 {
140 	struct socket *head = so->so_head;
141 
142 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
143 		return;
144 	if (head != NULL) {
145 		if (so->so_state & SS_INCOMP) {
146 			TAILQ_REMOVE(&head->so_incomp, so, so_list);
147 		} else if (so->so_state & SS_COMP) {
148 			TAILQ_REMOVE(&head->so_comp, so, so_list);
149 		} else {
150 			panic("sofree: not queued");
151 		}
152 		head->so_qlen--;
153 		so->so_state &= ~(SS_INCOMP|SS_COMP);
154 		so->so_head = NULL;
155 	}
156 	sbrelease(&so->so_snd);
157 	sorflush(so);
158 	FREE(so, M_SOCKET);
159 }
160 
161 /*
162  * Close a socket on last file table reference removal.
163  * Initiate disconnect if connected.
164  * Free socket when disconnect complete.
165  */
166 int
167 soclose(so)
168 	register struct socket *so;
169 {
170 	int s = splnet();		/* conservative */
171 	int error = 0;
172 
173 	if (so->so_options & SO_ACCEPTCONN) {
174 		struct socket *sp, *sonext;
175 
176 		for (sp = so->so_incomp.tqh_first; sp != NULL; sp = sonext) {
177 			sonext = sp->so_list.tqe_next;
178 			(void) soabort(sp);
179 		}
180 		for (sp = so->so_comp.tqh_first; sp != NULL; sp = sonext) {
181 			sonext = sp->so_list.tqe_next;
182 			(void) soabort(sp);
183 		}
184 	}
185 	if (so->so_pcb == 0)
186 		goto discard;
187 	if (so->so_state & SS_ISCONNECTED) {
188 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
189 			error = sodisconnect(so);
190 			if (error)
191 				goto drop;
192 		}
193 		if (so->so_options & SO_LINGER) {
194 			if ((so->so_state & SS_ISDISCONNECTING) &&
195 			    (so->so_state & SS_NBIO))
196 				goto drop;
197 			while (so->so_state & SS_ISCONNECTED) {
198 				error = tsleep((caddr_t)&so->so_timeo,
199 				    PSOCK | PCATCH, "soclos", so->so_linger);
200 				if (error)
201 					break;
202 			}
203 		}
204 	}
205 drop:
206 	if (so->so_pcb) {
207 		int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so);
208 		if (error == 0)
209 			error = error2;
210 	}
211 discard:
212 	if (so->so_state & SS_NOFDREF)
213 		panic("soclose: NOFDREF");
214 	so->so_state |= SS_NOFDREF;
215 	sofree(so);
216 	splx(s);
217 	return (error);
218 }
219 
220 /*
221  * Must be called at splnet...
222  */
223 int
224 soabort(so)
225 	struct socket *so;
226 {
227 
228 	return (*so->so_proto->pr_usrreqs->pru_abort)(so);
229 }
230 
231 int
232 soaccept(so, nam)
233 	register struct socket *so;
234 	struct mbuf *nam;
235 {
236 	int s = splnet();
237 	int error;
238 
239 	if ((so->so_state & SS_NOFDREF) == 0)
240 		panic("soaccept: !NOFDREF");
241 	so->so_state &= ~SS_NOFDREF;
242 	error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam);
243 	splx(s);
244 	return (error);
245 }
246 
247 int
248 soconnect(so, nam)
249 	register struct socket *so;
250 	struct mbuf *nam;
251 {
252 	int s;
253 	int error;
254 
255 	if (so->so_options & SO_ACCEPTCONN)
256 		return (EOPNOTSUPP);
257 	s = splnet();
258 	/*
259 	 * If protocol is connection-based, can only connect once.
260 	 * Otherwise, if connected, try to disconnect first.
261 	 * This allows user to disconnect by connecting to, e.g.,
262 	 * a null address.
263 	 */
264 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
265 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
266 	    (error = sodisconnect(so))))
267 		error = EISCONN;
268 	else
269 		error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam);
270 	splx(s);
271 	return (error);
272 }
273 
274 int
275 soconnect2(so1, so2)
276 	register struct socket *so1;
277 	struct socket *so2;
278 {
279 	int s = splnet();
280 	int error;
281 
282 	error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2);
283 	splx(s);
284 	return (error);
285 }
286 
287 int
288 sodisconnect(so)
289 	register struct socket *so;
290 {
291 	int s = splnet();
292 	int error;
293 
294 	if ((so->so_state & SS_ISCONNECTED) == 0) {
295 		error = ENOTCONN;
296 		goto bad;
297 	}
298 	if (so->so_state & SS_ISDISCONNECTING) {
299 		error = EALREADY;
300 		goto bad;
301 	}
302 	error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so);
303 bad:
304 	splx(s);
305 	return (error);
306 }
307 
308 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
309 /*
310  * Send on a socket.
311  * If send must go all at once and message is larger than
312  * send buffering, then hard error.
313  * Lock against other senders.
314  * If must go all at once and not enough room now, then
315  * inform user that this would block and do nothing.
316  * Otherwise, if nonblocking, send as much as possible.
317  * The data to be sent is described by "uio" if nonzero,
318  * otherwise by the mbuf chain "top" (which must be null
319  * if uio is not).  Data provided in mbuf chain must be small
320  * enough to send all at once.
321  *
322  * Returns nonzero on error, timeout or signal; callers
323  * must check for short counts if EINTR/ERESTART are returned.
324  * Data and control buffers are freed on return.
325  */
326 int
327 sosend(so, addr, uio, top, control, flags)
328 	register struct socket *so;
329 	struct mbuf *addr;
330 	struct uio *uio;
331 	struct mbuf *top;
332 	struct mbuf *control;
333 	int flags;
334 {
335 	struct proc *p = curproc;		/* XXX */
336 	struct mbuf **mp;
337 	register struct mbuf *m;
338 	register long space, len, resid;
339 	int clen = 0, error, s, dontroute, mlen;
340 	int atomic = sosendallatonce(so) || top;
341 
342 	if (uio)
343 		resid = uio->uio_resid;
344 	else
345 		resid = top->m_pkthdr.len;
346 	/*
347 	 * In theory resid should be unsigned.
348 	 * However, space must be signed, as it might be less than 0
349 	 * if we over-committed, and we must use a signed comparison
350 	 * of space and resid.  On the other hand, a negative resid
351 	 * causes us to loop sending 0-length segments to the protocol.
352 	 */
353 	if (resid < 0)
354 		return (EINVAL);
355 	dontroute =
356 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
357 	    (so->so_proto->pr_flags & PR_ATOMIC);
358 	p->p_stats->p_ru.ru_msgsnd++;
359 	if (control)
360 		clen = control->m_len;
361 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
362 
363 restart:
364 	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
365 	if (error)
366 		goto out;
367 	do {
368 		s = splnet();
369 		if (so->so_state & SS_CANTSENDMORE)
370 			snderr(EPIPE);
371 		if (so->so_error)
372 			snderr(so->so_error);
373 		if ((so->so_state & SS_ISCONNECTED) == 0) {
374 			/*
375 			 * `sendto' and `sendmsg' is allowed on a connection-
376 			 * based socket if it supports implied connect.
377 			 * Return ENOTCONN if not connected and no address is
378 			 * supplied.
379 			 */
380 			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
381 			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
382 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
383 				    !(resid == 0 && clen != 0))
384 					snderr(ENOTCONN);
385 			} else if (addr == 0)
386 			    snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
387 				   ENOTCONN : EDESTADDRREQ);
388 		}
389 		space = sbspace(&so->so_snd);
390 		if (flags & MSG_OOB)
391 			space += 1024;
392 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
393 		    clen > so->so_snd.sb_hiwat)
394 			snderr(EMSGSIZE);
395 		if (space < resid + clen && uio &&
396 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
397 			if (so->so_state & SS_NBIO)
398 				snderr(EWOULDBLOCK);
399 			sbunlock(&so->so_snd);
400 			error = sbwait(&so->so_snd);
401 			splx(s);
402 			if (error)
403 				goto out;
404 			goto restart;
405 		}
406 		splx(s);
407 		mp = &top;
408 		space -= clen;
409 		do {
410 		    if (uio == NULL) {
411 			/*
412 			 * Data is prepackaged in "top".
413 			 */
414 			resid = 0;
415 			if (flags & MSG_EOR)
416 				top->m_flags |= M_EOR;
417 		    } else do {
418 			if (top == 0) {
419 				MGETHDR(m, M_WAIT, MT_DATA);
420 				mlen = MHLEN;
421 				m->m_pkthdr.len = 0;
422 				m->m_pkthdr.rcvif = (struct ifnet *)0;
423 			} else {
424 				MGET(m, M_WAIT, MT_DATA);
425 				mlen = MLEN;
426 			}
427 			if (resid >= MINCLSIZE) {
428 				MCLGET(m, M_WAIT);
429 				if ((m->m_flags & M_EXT) == 0)
430 					goto nopages;
431 				mlen = MCLBYTES;
432 				len = min(min(mlen, resid), space);
433 			} else {
434 nopages:
435 				len = min(min(mlen, resid), space);
436 				/*
437 				 * For datagram protocols, leave room
438 				 * for protocol headers in first mbuf.
439 				 */
440 				if (atomic && top == 0 && len < mlen)
441 					MH_ALIGN(m, len);
442 			}
443 			space -= len;
444 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
445 			resid = uio->uio_resid;
446 			m->m_len = len;
447 			*mp = m;
448 			top->m_pkthdr.len += len;
449 			if (error)
450 				goto release;
451 			mp = &m->m_next;
452 			if (resid <= 0) {
453 				if (flags & MSG_EOR)
454 					top->m_flags |= M_EOR;
455 				break;
456 			}
457 		    } while (space > 0 && atomic);
458 		    if (dontroute)
459 			    so->so_options |= SO_DONTROUTE;
460 		    s = splnet();				/* XXX */
461 		    error = (*so->so_proto->pr_usrreqs->pru_send)(so,
462 			(flags & MSG_OOB) ? PRUS_OOB :
463 			/*
464 			 * If the user set MSG_EOF, the protocol
465 			 * understands this flag and nothing left to
466 			 * send then use PRU_SEND_EOF instead of PRU_SEND.
467 			 */
468 			((flags & MSG_EOF) &&
469 			 (so->so_proto->pr_flags & PR_IMPLOPCL) &&
470 			 (resid <= 0)) ?
471 				PRUS_EOF : 0,
472 			top, addr, control);
473 		    splx(s);
474 		    if (dontroute)
475 			    so->so_options &= ~SO_DONTROUTE;
476 		    clen = 0;
477 		    control = 0;
478 		    top = 0;
479 		    mp = &top;
480 		    if (error)
481 			goto release;
482 		} while (resid && space > 0);
483 	} while (resid);
484 
485 release:
486 	sbunlock(&so->so_snd);
487 out:
488 	if (top)
489 		m_freem(top);
490 	if (control)
491 		m_freem(control);
492 	return (error);
493 }
494 
495 /*
496  * Implement receive operations on a socket.
497  * We depend on the way that records are added to the sockbuf
498  * by sbappend*.  In particular, each record (mbufs linked through m_next)
499  * must begin with an address if the protocol so specifies,
500  * followed by an optional mbuf or mbufs containing ancillary data,
501  * and then zero or more mbufs of data.
502  * In order to avoid blocking network interrupts for the entire time here,
503  * we splx() while doing the actual copy to user space.
504  * Although the sockbuf is locked, new data may still be appended,
505  * and thus we must maintain consistency of the sockbuf during that time.
506  *
507  * The caller may receive the data as a single mbuf chain by supplying
508  * an mbuf **mp0 for use in returning the chain.  The uio is then used
509  * only for the count in uio_resid.
510  */
511 int
512 soreceive(so, paddr, uio, mp0, controlp, flagsp)
513 	register struct socket *so;
514 	struct mbuf **paddr;
515 	struct uio *uio;
516 	struct mbuf **mp0;
517 	struct mbuf **controlp;
518 	int *flagsp;
519 {
520 	register struct mbuf *m, **mp;
521 	register int flags, len, error, s, offset;
522 	struct protosw *pr = so->so_proto;
523 	struct mbuf *nextrecord;
524 	int moff, type = 0;
525 	int orig_resid = uio->uio_resid;
526 
527 	mp = mp0;
528 	if (paddr)
529 		*paddr = 0;
530 	if (controlp)
531 		*controlp = 0;
532 	if (flagsp)
533 		flags = *flagsp &~ MSG_EOR;
534 	else
535 		flags = 0;
536 	if (flags & MSG_OOB) {
537 		m = m_get(M_WAIT, MT_DATA);
538 		error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK);
539 		if (error)
540 			goto bad;
541 		do {
542 			error = uiomove(mtod(m, caddr_t),
543 			    (int) min(uio->uio_resid, m->m_len), uio);
544 			m = m_free(m);
545 		} while (uio->uio_resid && error == 0 && m);
546 bad:
547 		if (m)
548 			m_freem(m);
549 		return (error);
550 	}
551 	if (mp)
552 		*mp = (struct mbuf *)0;
553 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
554 		(*pr->pr_usrreqs->pru_rcvd)(so, 0);
555 
556 restart:
557 	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
558 	if (error)
559 		return (error);
560 	s = splnet();
561 
562 	m = so->so_rcv.sb_mb;
563 	/*
564 	 * If we have less data than requested, block awaiting more
565 	 * (subject to any timeout) if:
566 	 *   1. the current count is less than the low water mark, or
567 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
568 	 *	receive operation at once if we block (resid <= hiwat).
569 	 *   3. MSG_DONTWAIT is not set
570 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
571 	 * we have to do the receive in sections, and thus risk returning
572 	 * a short count if a timeout or signal occurs after we start.
573 	 */
574 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
575 	    so->so_rcv.sb_cc < uio->uio_resid) &&
576 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
577 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
578 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
579 #ifdef DIAGNOSTIC
580 		if (m == 0 && so->so_rcv.sb_cc)
581 			panic("receive 1");
582 #endif
583 		if (so->so_error) {
584 			if (m)
585 				goto dontblock;
586 			error = so->so_error;
587 			if ((flags & MSG_PEEK) == 0)
588 				so->so_error = 0;
589 			goto release;
590 		}
591 		if (so->so_state & SS_CANTRCVMORE) {
592 			if (m)
593 				goto dontblock;
594 			else
595 				goto release;
596 		}
597 		for (; m; m = m->m_next)
598 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
599 				m = so->so_rcv.sb_mb;
600 				goto dontblock;
601 			}
602 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
603 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
604 			error = ENOTCONN;
605 			goto release;
606 		}
607 		if (uio->uio_resid == 0)
608 			goto release;
609 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
610 			error = EWOULDBLOCK;
611 			goto release;
612 		}
613 		sbunlock(&so->so_rcv);
614 		error = sbwait(&so->so_rcv);
615 		splx(s);
616 		if (error)
617 			return (error);
618 		goto restart;
619 	}
620 dontblock:
621 	if (uio->uio_procp)
622 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
623 	nextrecord = m->m_nextpkt;
624 	if (pr->pr_flags & PR_ADDR) {
625 #ifdef DIAGNOSTIC
626 		if (m->m_type != MT_SONAME)
627 			panic("receive 1a");
628 #endif
629 		orig_resid = 0;
630 		if (flags & MSG_PEEK) {
631 			if (paddr)
632 				*paddr = m_copy(m, 0, m->m_len);
633 			m = m->m_next;
634 		} else {
635 			sbfree(&so->so_rcv, m);
636 			if (paddr) {
637 				*paddr = m;
638 				so->so_rcv.sb_mb = m->m_next;
639 				m->m_next = 0;
640 				m = so->so_rcv.sb_mb;
641 			} else {
642 				MFREE(m, so->so_rcv.sb_mb);
643 				m = so->so_rcv.sb_mb;
644 			}
645 		}
646 	}
647 	while (m && m->m_type == MT_CONTROL && error == 0) {
648 		if (flags & MSG_PEEK) {
649 			if (controlp)
650 				*controlp = m_copy(m, 0, m->m_len);
651 			m = m->m_next;
652 		} else {
653 			sbfree(&so->so_rcv, m);
654 			if (controlp) {
655 				if (pr->pr_domain->dom_externalize &&
656 				    mtod(m, struct cmsghdr *)->cmsg_type ==
657 				    SCM_RIGHTS)
658 				   error = (*pr->pr_domain->dom_externalize)(m);
659 				*controlp = m;
660 				so->so_rcv.sb_mb = m->m_next;
661 				m->m_next = 0;
662 				m = so->so_rcv.sb_mb;
663 			} else {
664 				MFREE(m, so->so_rcv.sb_mb);
665 				m = so->so_rcv.sb_mb;
666 			}
667 		}
668 		if (controlp) {
669 			orig_resid = 0;
670 			controlp = &(*controlp)->m_next;
671 		}
672 	}
673 	if (m) {
674 		if ((flags & MSG_PEEK) == 0)
675 			m->m_nextpkt = nextrecord;
676 		type = m->m_type;
677 		if (type == MT_OOBDATA)
678 			flags |= MSG_OOB;
679 	}
680 	moff = 0;
681 	offset = 0;
682 	while (m && uio->uio_resid > 0 && error == 0) {
683 		if (m->m_type == MT_OOBDATA) {
684 			if (type != MT_OOBDATA)
685 				break;
686 		} else if (type == MT_OOBDATA)
687 			break;
688 #ifdef DIAGNOSTIC
689 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
690 			panic("receive 3");
691 #endif
692 		so->so_state &= ~SS_RCVATMARK;
693 		len = uio->uio_resid;
694 		if (so->so_oobmark && len > so->so_oobmark - offset)
695 			len = so->so_oobmark - offset;
696 		if (len > m->m_len - moff)
697 			len = m->m_len - moff;
698 		/*
699 		 * If mp is set, just pass back the mbufs.
700 		 * Otherwise copy them out via the uio, then free.
701 		 * Sockbuf must be consistent here (points to current mbuf,
702 		 * it points to next record) when we drop priority;
703 		 * we must note any additions to the sockbuf when we
704 		 * block interrupts again.
705 		 */
706 		if (mp == 0) {
707 			splx(s);
708 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
709 			s = splnet();
710 		} else
711 			uio->uio_resid -= len;
712 		if (len == m->m_len - moff) {
713 			if (m->m_flags & M_EOR)
714 				flags |= MSG_EOR;
715 			if (flags & MSG_PEEK) {
716 				m = m->m_next;
717 				moff = 0;
718 			} else {
719 				nextrecord = m->m_nextpkt;
720 				sbfree(&so->so_rcv, m);
721 				if (mp) {
722 					*mp = m;
723 					mp = &m->m_next;
724 					so->so_rcv.sb_mb = m = m->m_next;
725 					*mp = (struct mbuf *)0;
726 				} else {
727 					MFREE(m, so->so_rcv.sb_mb);
728 					m = so->so_rcv.sb_mb;
729 				}
730 				if (m)
731 					m->m_nextpkt = nextrecord;
732 			}
733 		} else {
734 			if (flags & MSG_PEEK)
735 				moff += len;
736 			else {
737 				if (mp)
738 					*mp = m_copym(m, 0, len, M_WAIT);
739 				m->m_data += len;
740 				m->m_len -= len;
741 				so->so_rcv.sb_cc -= len;
742 			}
743 		}
744 		if (so->so_oobmark) {
745 			if ((flags & MSG_PEEK) == 0) {
746 				so->so_oobmark -= len;
747 				if (so->so_oobmark == 0) {
748 					so->so_state |= SS_RCVATMARK;
749 					break;
750 				}
751 			} else {
752 				offset += len;
753 				if (offset == so->so_oobmark)
754 					break;
755 			}
756 		}
757 		if (flags & MSG_EOR)
758 			break;
759 		/*
760 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
761 		 * we must not quit until "uio->uio_resid == 0" or an error
762 		 * termination.  If a signal/timeout occurs, return
763 		 * with a short count but without error.
764 		 * Keep sockbuf locked against other readers.
765 		 */
766 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
767 		    !sosendallatonce(so) && !nextrecord) {
768 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
769 				break;
770 			error = sbwait(&so->so_rcv);
771 			if (error) {
772 				sbunlock(&so->so_rcv);
773 				splx(s);
774 				return (0);
775 			}
776 			m = so->so_rcv.sb_mb;
777 			if (m)
778 				nextrecord = m->m_nextpkt;
779 		}
780 	}
781 
782 	if (m && pr->pr_flags & PR_ATOMIC) {
783 		flags |= MSG_TRUNC;
784 		if ((flags & MSG_PEEK) == 0)
785 			(void) sbdroprecord(&so->so_rcv);
786 	}
787 	if ((flags & MSG_PEEK) == 0) {
788 		if (m == 0)
789 			so->so_rcv.sb_mb = nextrecord;
790 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
791 			(*pr->pr_usrreqs->pru_rcvd)(so, flags);
792 	}
793 	if (orig_resid == uio->uio_resid && orig_resid &&
794 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
795 		sbunlock(&so->so_rcv);
796 		splx(s);
797 		goto restart;
798 	}
799 
800 	if (flagsp)
801 		*flagsp |= flags;
802 release:
803 	sbunlock(&so->so_rcv);
804 	splx(s);
805 	return (error);
806 }
807 
808 int
809 soshutdown(so, how)
810 	register struct socket *so;
811 	register int how;
812 {
813 	register struct protosw *pr = so->so_proto;
814 
815 	how++;
816 	if (how & FREAD)
817 		sorflush(so);
818 	if (how & FWRITE)
819 		return ((*pr->pr_usrreqs->pru_shutdown)(so));
820 	return (0);
821 }
822 
823 void
824 sorflush(so)
825 	register struct socket *so;
826 {
827 	register struct sockbuf *sb = &so->so_rcv;
828 	register struct protosw *pr = so->so_proto;
829 	register int s;
830 	struct sockbuf asb;
831 
832 	sb->sb_flags |= SB_NOINTR;
833 	(void) sblock(sb, M_WAITOK);
834 	s = splimp();
835 	socantrcvmore(so);
836 	sbunlock(sb);
837 	asb = *sb;
838 	bzero((caddr_t)sb, sizeof (*sb));
839 	splx(s);
840 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
841 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
842 	sbrelease(&asb);
843 }
844 
845 int
846 sosetopt(so, level, optname, m0)
847 	register struct socket *so;
848 	int level, optname;
849 	struct mbuf *m0;
850 {
851 	int error = 0;
852 	register struct mbuf *m = m0;
853 
854 	if (level != SOL_SOCKET) {
855 		if (so->so_proto && so->so_proto->pr_ctloutput)
856 			return ((*so->so_proto->pr_ctloutput)
857 				  (PRCO_SETOPT, so, level, optname, &m0));
858 		error = ENOPROTOOPT;
859 	} else {
860 		switch (optname) {
861 
862 		case SO_LINGER:
863 			if (m == NULL || m->m_len != sizeof (struct linger)) {
864 				error = EINVAL;
865 				goto bad;
866 			}
867 			so->so_linger = mtod(m, struct linger *)->l_linger;
868 			/* fall thru... */
869 
870 		case SO_DEBUG:
871 		case SO_KEEPALIVE:
872 		case SO_DONTROUTE:
873 		case SO_USELOOPBACK:
874 		case SO_BROADCAST:
875 		case SO_REUSEADDR:
876 		case SO_REUSEPORT:
877 		case SO_OOBINLINE:
878 		case SO_TIMESTAMP:
879 			if (m == NULL || m->m_len < sizeof (int)) {
880 				error = EINVAL;
881 				goto bad;
882 			}
883 			if (*mtod(m, int *))
884 				so->so_options |= optname;
885 			else
886 				so->so_options &= ~optname;
887 			break;
888 
889 		case SO_SNDBUF:
890 		case SO_RCVBUF:
891 		case SO_SNDLOWAT:
892 		case SO_RCVLOWAT:
893 			if (m == NULL || m->m_len < sizeof (int)) {
894 				error = EINVAL;
895 				goto bad;
896 			}
897 			switch (optname) {
898 
899 			case SO_SNDBUF:
900 			case SO_RCVBUF:
901 				if (sbreserve(optname == SO_SNDBUF ?
902 				    &so->so_snd : &so->so_rcv,
903 				    (u_long) *mtod(m, int *)) == 0) {
904 					error = ENOBUFS;
905 					goto bad;
906 				}
907 				break;
908 
909 			case SO_SNDLOWAT:
910 				so->so_snd.sb_lowat = *mtod(m, int *);
911 				break;
912 			case SO_RCVLOWAT:
913 				so->so_rcv.sb_lowat = *mtod(m, int *);
914 				break;
915 			}
916 			break;
917 
918 		case SO_SNDTIMEO:
919 		case SO_RCVTIMEO:
920 		    {
921 			struct timeval *tv;
922 			short val;
923 
924 			if (m == NULL || m->m_len < sizeof (*tv)) {
925 				error = EINVAL;
926 				goto bad;
927 			}
928 			tv = mtod(m, struct timeval *);
929 			if (tv->tv_sec > SHRT_MAX / hz - hz) {
930 				error = EDOM;
931 				goto bad;
932 			}
933 			val = tv->tv_sec * hz + tv->tv_usec / tick;
934 
935 			switch (optname) {
936 
937 			case SO_SNDTIMEO:
938 				so->so_snd.sb_timeo = val;
939 				break;
940 			case SO_RCVTIMEO:
941 				so->so_rcv.sb_timeo = val;
942 				break;
943 			}
944 			break;
945 		    }
946 
947 		case SO_PRIVSTATE:
948 			/* we don't care what the parameter is... */
949 			so->so_state &= ~SS_PRIV;
950 			break;
951 
952 		default:
953 			error = ENOPROTOOPT;
954 			break;
955 		}
956 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
957 			(void) ((*so->so_proto->pr_ctloutput)
958 				  (PRCO_SETOPT, so, level, optname, &m0));
959 			m = NULL;	/* freed by protocol */
960 		}
961 	}
962 bad:
963 	if (m)
964 		(void) m_free(m);
965 	return (error);
966 }
967 
968 int
969 sogetopt(so, level, optname, mp)
970 	register struct socket *so;
971 	int level, optname;
972 	struct mbuf **mp;
973 {
974 	register struct mbuf *m;
975 
976 	if (level != SOL_SOCKET) {
977 		if (so->so_proto && so->so_proto->pr_ctloutput) {
978 			return ((*so->so_proto->pr_ctloutput)
979 				  (PRCO_GETOPT, so, level, optname, mp));
980 		} else
981 			return (ENOPROTOOPT);
982 	} else {
983 		m = m_get(M_WAIT, MT_SOOPTS);
984 		m->m_len = sizeof (int);
985 
986 		switch (optname) {
987 
988 		case SO_LINGER:
989 			m->m_len = sizeof (struct linger);
990 			mtod(m, struct linger *)->l_onoff =
991 				so->so_options & SO_LINGER;
992 			mtod(m, struct linger *)->l_linger = so->so_linger;
993 			break;
994 
995 		case SO_USELOOPBACK:
996 		case SO_DONTROUTE:
997 		case SO_DEBUG:
998 		case SO_KEEPALIVE:
999 		case SO_REUSEADDR:
1000 		case SO_REUSEPORT:
1001 		case SO_BROADCAST:
1002 		case SO_OOBINLINE:
1003 		case SO_TIMESTAMP:
1004 			*mtod(m, int *) = so->so_options & optname;
1005 			break;
1006 
1007 		case SO_PRIVSTATE:
1008 			*mtod(m, int *) = so->so_state & SS_PRIV;
1009 			break;
1010 
1011 		case SO_TYPE:
1012 			*mtod(m, int *) = so->so_type;
1013 			break;
1014 
1015 		case SO_ERROR:
1016 			*mtod(m, int *) = so->so_error;
1017 			so->so_error = 0;
1018 			break;
1019 
1020 		case SO_SNDBUF:
1021 			*mtod(m, int *) = so->so_snd.sb_hiwat;
1022 			break;
1023 
1024 		case SO_RCVBUF:
1025 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
1026 			break;
1027 
1028 		case SO_SNDLOWAT:
1029 			*mtod(m, int *) = so->so_snd.sb_lowat;
1030 			break;
1031 
1032 		case SO_RCVLOWAT:
1033 			*mtod(m, int *) = so->so_rcv.sb_lowat;
1034 			break;
1035 
1036 		case SO_SNDTIMEO:
1037 		case SO_RCVTIMEO:
1038 		    {
1039 			int val = (optname == SO_SNDTIMEO ?
1040 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1041 
1042 			m->m_len = sizeof(struct timeval);
1043 			mtod(m, struct timeval *)->tv_sec = val / hz;
1044 			mtod(m, struct timeval *)->tv_usec =
1045 			    (val % hz) * tick;
1046 			break;
1047 		    }
1048 
1049 		default:
1050 			(void)m_free(m);
1051 			return (ENOPROTOOPT);
1052 		}
1053 		*mp = m;
1054 		return (0);
1055 	}
1056 }
1057 
1058 void
1059 sohasoutofband(so)
1060 	register struct socket *so;
1061 {
1062 	struct proc *p;
1063 
1064 	if (so->so_pgid < 0)
1065 		gsignal(-so->so_pgid, SIGURG);
1066 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1067 		psignal(p, SIGURG);
1068 	selwakeup(&so->so_rcv.sb_sel);
1069 }
1070