xref: /netbsd-src/sys/kern/uipc_socket.c (revision dc306354b0b29af51801a7632f1e95265a68cd81)
1 /*	$NetBSD: uipc_socket.c,v 1.40 1998/12/16 00:26:10 thorpej Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)uipc_socket.c	8.6 (Berkeley) 5/2/95
36  */
37 
38 #include "opt_compat_sunos.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/proc.h>
43 #include <sys/file.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/domain.h>
47 #include <sys/kernel.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/signalvar.h>
52 #include <sys/resourcevar.h>
53 #include <sys/pool.h>
54 
55 struct pool socket_pool;
56 
57 void
58 soinit()
59 {
60 
61 	pool_init(&socket_pool, sizeof(struct socket), 0, 0, 0,
62 	    "sockpl", 0, NULL, NULL, M_SOCKET);
63 }
64 
65 /*
66  * Socket operation routines.
67  * These routines are called by the routines in
68  * sys_socket.c or from a system process, and
69  * implement the semantics of socket operations by
70  * switching out to the protocol specific routines.
71  */
72 /*ARGSUSED*/
73 int
74 socreate(dom, aso, type, proto)
75 	int dom;
76 	struct socket **aso;
77 	register int type;
78 	int proto;
79 {
80 	struct proc *p = curproc;		/* XXX */
81 	register struct protosw *prp;
82 	register struct socket *so;
83 	register int error;
84 	int s;
85 
86 	if (proto)
87 		prp = pffindproto(dom, proto, type);
88 	else
89 		prp = pffindtype(dom, type);
90 	if (prp == 0 || prp->pr_usrreq == 0)
91 		return (EPROTONOSUPPORT);
92 	if (prp->pr_type != type)
93 		return (EPROTOTYPE);
94 	s = splsoftnet();
95 	so = pool_get(&socket_pool, PR_WAITOK);
96 	memset((caddr_t)so, 0, sizeof(*so));
97 	TAILQ_INIT(&so->so_q0);
98 	TAILQ_INIT(&so->so_q);
99 	so->so_type = type;
100 	so->so_proto = prp;
101 	so->so_send = sosend;
102 	so->so_receive = soreceive;
103 	error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
104 	    (struct mbuf *)(long)proto, (struct mbuf *)0, p);
105 	if (error) {
106 		so->so_state |= SS_NOFDREF;
107 		sofree(so);
108 		splx(s);
109 		return (error);
110 	}
111 #ifdef COMPAT_SUNOS
112 	{
113 		extern struct emul emul_sunos;
114 		if (p->p_emul == &emul_sunos && type == SOCK_DGRAM)
115 			so->so_options |= SO_BROADCAST;
116 	}
117 #endif
118 	splx(s);
119 	*aso = so;
120 	return (0);
121 }
122 
123 int
124 sobind(so, nam)
125 	struct socket *so;
126 	struct mbuf *nam;
127 {
128 	struct proc *p = curproc;		/* XXX */
129 	int s = splsoftnet();
130 	int error;
131 
132 	error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0,
133 	    nam, (struct mbuf *)0, p);
134 	splx(s);
135 	return (error);
136 }
137 
138 int
139 solisten(so, backlog)
140 	register struct socket *so;
141 	int backlog;
142 {
143 	int s = splsoftnet(), error;
144 
145 	error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0,
146 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
147 	if (error) {
148 		splx(s);
149 		return (error);
150 	}
151 	if (so->so_q.tqh_first == NULL)
152 		so->so_options |= SO_ACCEPTCONN;
153 	if (backlog < 0)
154 		backlog = 0;
155 	so->so_qlimit = min(backlog, SOMAXCONN);
156 	splx(s);
157 	return (0);
158 }
159 
160 void
161 sofree(so)
162 	register struct socket *so;
163 {
164 
165 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
166 		return;
167 	if (so->so_head) {
168 		if (!soqremque(so, 0) && !soqremque(so, 1))
169 			panic("sofree dq");
170 		so->so_head = 0;
171 	}
172 	sbrelease(&so->so_snd);
173 	sorflush(so);
174 	pool_put(&socket_pool, so);
175 }
176 
177 /*
178  * Close a socket on last file table reference removal.
179  * Initiate disconnect if connected.
180  * Free socket when disconnect complete.
181  */
182 int
183 soclose(so)
184 	register struct socket *so;
185 {
186 	int s = splsoftnet();		/* conservative */
187 	int error = 0;
188 
189 	if (so->so_options & SO_ACCEPTCONN) {
190 		while (so->so_q0.tqh_first)
191 			(void) soabort(so->so_q0.tqh_first);
192 		while (so->so_q.tqh_first)
193 			(void) soabort(so->so_q.tqh_first);
194 	}
195 	if (so->so_pcb == 0)
196 		goto discard;
197 	if (so->so_state & SS_ISCONNECTED) {
198 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
199 			error = sodisconnect(so);
200 			if (error)
201 				goto drop;
202 		}
203 		if (so->so_options & SO_LINGER) {
204 			if ((so->so_state & SS_ISDISCONNECTING) &&
205 			    (so->so_state & SS_NBIO))
206 				goto drop;
207 			while (so->so_state & SS_ISCONNECTED) {
208 				error = tsleep((caddr_t)&so->so_timeo,
209 					       PSOCK | PCATCH, netcls,
210 					       so->so_linger * hz);
211 				if (error)
212 					break;
213 			}
214 		}
215 	}
216 drop:
217 	if (so->so_pcb) {
218 		int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
219 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
220 		    (struct proc *)0);
221 		if (error == 0)
222 			error = error2;
223 	}
224 discard:
225 	if (so->so_state & SS_NOFDREF)
226 		panic("soclose: NOFDREF");
227 	so->so_state |= SS_NOFDREF;
228 	sofree(so);
229 	splx(s);
230 	return (error);
231 }
232 
233 /*
234  * Must be called at splsoftnet...
235  */
236 int
237 soabort(so)
238 	struct socket *so;
239 {
240 
241 	return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0,
242 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
243 }
244 
245 int
246 soaccept(so, nam)
247 	register struct socket *so;
248 	struct mbuf *nam;
249 {
250 	int s = splsoftnet();
251 	int error;
252 
253 	if ((so->so_state & SS_NOFDREF) == 0)
254 		panic("soaccept: !NOFDREF");
255 	so->so_state &= ~SS_NOFDREF;
256 	error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, (struct mbuf *)0,
257 	    nam, (struct mbuf *)0, (struct proc *)0);
258 	splx(s);
259 	return (error);
260 }
261 
262 int
263 soconnect(so, nam)
264 	register struct socket *so;
265 	struct mbuf *nam;
266 {
267 	struct proc *p = curproc;		/* XXX */
268 	int s;
269 	int error;
270 
271 	if (so->so_options & SO_ACCEPTCONN)
272 		return (EOPNOTSUPP);
273 	s = splsoftnet();
274 	/*
275 	 * If protocol is connection-based, can only connect once.
276 	 * Otherwise, if connected, try to disconnect first.
277 	 * This allows user to disconnect by connecting to, e.g.,
278 	 * a null address.
279 	 */
280 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
281 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
282 	    (error = sodisconnect(so))))
283 		error = EISCONN;
284 	else
285 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
286 		    (struct mbuf *)0, nam, (struct mbuf *)0, p);
287 	splx(s);
288 	return (error);
289 }
290 
291 int
292 soconnect2(so1, so2)
293 	register struct socket *so1;
294 	struct socket *so2;
295 {
296 	int s = splsoftnet();
297 	int error;
298 
299 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
300 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0,
301 	    (struct proc *)0);
302 	splx(s);
303 	return (error);
304 }
305 
306 int
307 sodisconnect(so)
308 	register struct socket *so;
309 {
310 	int s = splsoftnet();
311 	int error;
312 
313 	if ((so->so_state & SS_ISCONNECTED) == 0) {
314 		error = ENOTCONN;
315 		goto bad;
316 	}
317 	if (so->so_state & SS_ISDISCONNECTING) {
318 		error = EALREADY;
319 		goto bad;
320 	}
321 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
322 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
323 	    (struct proc *)0);
324 bad:
325 	splx(s);
326 	return (error);
327 }
328 
329 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
330 /*
331  * Send on a socket.
332  * If send must go all at once and message is larger than
333  * send buffering, then hard error.
334  * Lock against other senders.
335  * If must go all at once and not enough room now, then
336  * inform user that this would block and do nothing.
337  * Otherwise, if nonblocking, send as much as possible.
338  * The data to be sent is described by "uio" if nonzero,
339  * otherwise by the mbuf chain "top" (which must be null
340  * if uio is not).  Data provided in mbuf chain must be small
341  * enough to send all at once.
342  *
343  * Returns nonzero on error, timeout or signal; callers
344  * must check for short counts if EINTR/ERESTART are returned.
345  * Data and control buffers are freed on return.
346  */
347 int
348 sosend(so, addr, uio, top, control, flags)
349 	register struct socket *so;
350 	struct mbuf *addr;
351 	struct uio *uio;
352 	struct mbuf *top;
353 	struct mbuf *control;
354 	int flags;
355 {
356 	struct proc *p = curproc;		/* XXX */
357 	struct mbuf **mp;
358 	register struct mbuf *m;
359 	register long space, len, resid;
360 	int clen = 0, error, s, dontroute, mlen;
361 	int atomic = sosendallatonce(so) || top;
362 
363 	if (uio)
364 		resid = uio->uio_resid;
365 	else
366 		resid = top->m_pkthdr.len;
367 	/*
368 	 * In theory resid should be unsigned.
369 	 * However, space must be signed, as it might be less than 0
370 	 * if we over-committed, and we must use a signed comparison
371 	 * of space and resid.  On the other hand, a negative resid
372 	 * causes us to loop sending 0-length segments to the protocol.
373 	 */
374 	if (resid < 0) {
375 		error = EINVAL;
376 		goto out;
377 	}
378 	dontroute =
379 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
380 	    (so->so_proto->pr_flags & PR_ATOMIC);
381 	p->p_stats->p_ru.ru_msgsnd++;
382 	if (control)
383 		clen = control->m_len;
384 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
385 
386 restart:
387 	if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
388 		goto out;
389 	do {
390 		s = splsoftnet();
391 		if (so->so_state & SS_CANTSENDMORE)
392 			snderr(EPIPE);
393 		if (so->so_error)
394 			snderr(so->so_error);
395 		if ((so->so_state & SS_ISCONNECTED) == 0) {
396 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
397 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
398 				    !(resid == 0 && clen != 0))
399 					snderr(ENOTCONN);
400 			} else if (addr == 0)
401 				snderr(EDESTADDRREQ);
402 		}
403 		space = sbspace(&so->so_snd);
404 		if (flags & MSG_OOB)
405 			space += 1024;
406 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
407 		    clen > so->so_snd.sb_hiwat)
408 			snderr(EMSGSIZE);
409 		if (space < resid + clen && uio &&
410 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
411 			if (so->so_state & SS_NBIO)
412 				snderr(EWOULDBLOCK);
413 			sbunlock(&so->so_snd);
414 			error = sbwait(&so->so_snd);
415 			splx(s);
416 			if (error)
417 				goto out;
418 			goto restart;
419 		}
420 		splx(s);
421 		mp = &top;
422 		space -= clen;
423 		do {
424 		    if (uio == NULL) {
425 			/*
426 			 * Data is prepackaged in "top".
427 			 */
428 			resid = 0;
429 			if (flags & MSG_EOR)
430 				top->m_flags |= M_EOR;
431 		    } else do {
432 			if (top == 0) {
433 				MGETHDR(m, M_WAIT, MT_DATA);
434 				mlen = MHLEN;
435 				m->m_pkthdr.len = 0;
436 				m->m_pkthdr.rcvif = (struct ifnet *)0;
437 			} else {
438 				MGET(m, M_WAIT, MT_DATA);
439 				mlen = MLEN;
440 			}
441 			if (resid >= MINCLSIZE && space >= MCLBYTES) {
442 				MCLGET(m, M_WAIT);
443 				if ((m->m_flags & M_EXT) == 0)
444 					goto nopages;
445 				mlen = MCLBYTES;
446 #ifdef	MAPPED_MBUFS
447 				len = min(MCLBYTES, resid);
448 #else
449 				if (atomic && top == 0) {
450 					len = min(MCLBYTES - max_hdr, resid);
451 					m->m_data += max_hdr;
452 				} else
453 					len = min(MCLBYTES, resid);
454 #endif
455 				space -= len;
456 			} else {
457 nopages:
458 				len = min(min(mlen, resid), space);
459 				space -= len;
460 				/*
461 				 * For datagram protocols, leave room
462 				 * for protocol headers in first mbuf.
463 				 */
464 				if (atomic && top == 0 && len < mlen)
465 					MH_ALIGN(m, len);
466 			}
467 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
468 			resid = uio->uio_resid;
469 			m->m_len = len;
470 			*mp = m;
471 			top->m_pkthdr.len += len;
472 			if (error)
473 				goto release;
474 			mp = &m->m_next;
475 			if (resid <= 0) {
476 				if (flags & MSG_EOR)
477 					top->m_flags |= M_EOR;
478 				break;
479 			}
480 		    } while (space > 0 && atomic);
481 		    if (dontroute)
482 			    so->so_options |= SO_DONTROUTE;
483 		    if (resid > 0)
484 			    so->so_state |= SS_MORETOCOME;
485 		    s = splsoftnet();				/* XXX */
486 		    error = (*so->so_proto->pr_usrreq)(so,
487 			(flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
488 			top, addr, control, p);
489 		    splx(s);
490 		    if (dontroute)
491 			    so->so_options &= ~SO_DONTROUTE;
492 		    if (resid > 0)
493 			    so->so_state &= ~SS_MORETOCOME;
494 		    clen = 0;
495 		    control = 0;
496 		    top = 0;
497 		    mp = &top;
498 		    if (error)
499 			goto release;
500 		} while (resid && space > 0);
501 	} while (resid);
502 
503 release:
504 	sbunlock(&so->so_snd);
505 out:
506 	if (top)
507 		m_freem(top);
508 	if (control)
509 		m_freem(control);
510 	return (error);
511 }
512 
513 /*
514  * Implement receive operations on a socket.
515  * We depend on the way that records are added to the sockbuf
516  * by sbappend*.  In particular, each record (mbufs linked through m_next)
517  * must begin with an address if the protocol so specifies,
518  * followed by an optional mbuf or mbufs containing ancillary data,
519  * and then zero or more mbufs of data.
520  * In order to avoid blocking network interrupts for the entire time here,
521  * we splx() while doing the actual copy to user space.
522  * Although the sockbuf is locked, new data may still be appended,
523  * and thus we must maintain consistency of the sockbuf during that time.
524  *
525  * The caller may receive the data as a single mbuf chain by supplying
526  * an mbuf **mp0 for use in returning the chain.  The uio is then used
527  * only for the count in uio_resid.
528  */
529 int
530 soreceive(so, paddr, uio, mp0, controlp, flagsp)
531 	register struct socket *so;
532 	struct mbuf **paddr;
533 	struct uio *uio;
534 	struct mbuf **mp0;
535 	struct mbuf **controlp;
536 	int *flagsp;
537 {
538 	register struct mbuf *m, **mp;
539 	register int flags, len, error, s, offset;
540 	struct protosw *pr = so->so_proto;
541 	struct mbuf *nextrecord;
542 	int moff, type = 0;
543 	int orig_resid = uio->uio_resid;
544 
545 	mp = mp0;
546 	if (paddr)
547 		*paddr = 0;
548 	if (controlp)
549 		*controlp = 0;
550 	if (flagsp)
551 		flags = *flagsp &~ MSG_EOR;
552 	else
553 		flags = 0;
554 	if (flags & MSG_OOB) {
555 		m = m_get(M_WAIT, MT_DATA);
556 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
557 		    (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0,
558 		    (struct proc *)0);
559 		if (error)
560 			goto bad;
561 		do {
562 			error = uiomove(mtod(m, caddr_t),
563 			    (int) min(uio->uio_resid, m->m_len), uio);
564 			m = m_free(m);
565 		} while (uio->uio_resid && error == 0 && m);
566 bad:
567 		if (m)
568 			m_freem(m);
569 		return (error);
570 	}
571 	if (mp)
572 		*mp = (struct mbuf *)0;
573 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
574 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
575 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
576 
577 restart:
578 	if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
579 		return (error);
580 	s = splsoftnet();
581 
582 	m = so->so_rcv.sb_mb;
583 	/*
584 	 * If we have less data than requested, block awaiting more
585 	 * (subject to any timeout) if:
586 	 *   1. the current count is less than the low water mark,
587 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
588 	 *	receive operation at once if we block (resid <= hiwat), or
589 	 *   3. MSG_DONTWAIT is not set.
590 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
591 	 * we have to do the receive in sections, and thus risk returning
592 	 * a short count if a timeout or signal occurs after we start.
593 	 */
594 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
595 	    so->so_rcv.sb_cc < uio->uio_resid) &&
596 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
597 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
598 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
599 #ifdef DIAGNOSTIC
600 		if (m == 0 && so->so_rcv.sb_cc)
601 			panic("receive 1");
602 #endif
603 		if (so->so_error) {
604 			if (m)
605 				goto dontblock;
606 			error = so->so_error;
607 			if ((flags & MSG_PEEK) == 0)
608 				so->so_error = 0;
609 			goto release;
610 		}
611 		if (so->so_state & SS_CANTRCVMORE) {
612 			if (m)
613 				goto dontblock;
614 			else
615 				goto release;
616 		}
617 		for (; m; m = m->m_next)
618 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
619 				m = so->so_rcv.sb_mb;
620 				goto dontblock;
621 			}
622 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
623 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
624 			error = ENOTCONN;
625 			goto release;
626 		}
627 		if (uio->uio_resid == 0)
628 			goto release;
629 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
630 			error = EWOULDBLOCK;
631 			goto release;
632 		}
633 		sbunlock(&so->so_rcv);
634 		error = sbwait(&so->so_rcv);
635 		splx(s);
636 		if (error)
637 			return (error);
638 		goto restart;
639 	}
640 dontblock:
641 #ifdef notyet /* XXXX */
642 	if (uio->uio_procp)
643 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
644 #endif
645 	nextrecord = m->m_nextpkt;
646 	if (pr->pr_flags & PR_ADDR) {
647 #ifdef DIAGNOSTIC
648 		if (m->m_type != MT_SONAME)
649 			panic("receive 1a");
650 #endif
651 		orig_resid = 0;
652 		if (flags & MSG_PEEK) {
653 			if (paddr)
654 				*paddr = m_copy(m, 0, m->m_len);
655 			m = m->m_next;
656 		} else {
657 			sbfree(&so->so_rcv, m);
658 			if (paddr) {
659 				*paddr = 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 	}
669 	while (m && m->m_type == MT_CONTROL && error == 0) {
670 		if (flags & MSG_PEEK) {
671 			if (controlp)
672 				*controlp = m_copy(m, 0, m->m_len);
673 			m = m->m_next;
674 		} else {
675 			sbfree(&so->so_rcv, m);
676 			if (controlp) {
677 				if (pr->pr_domain->dom_externalize &&
678 				    mtod(m, struct cmsghdr *)->cmsg_type ==
679 				    SCM_RIGHTS)
680 				   error = (*pr->pr_domain->dom_externalize)(m);
681 				*controlp = m;
682 				so->so_rcv.sb_mb = m->m_next;
683 				m->m_next = 0;
684 				m = so->so_rcv.sb_mb;
685 			} else {
686 				MFREE(m, so->so_rcv.sb_mb);
687 				m = so->so_rcv.sb_mb;
688 			}
689 		}
690 		if (controlp) {
691 			orig_resid = 0;
692 			controlp = &(*controlp)->m_next;
693 		}
694 	}
695 	if (m) {
696 		if ((flags & MSG_PEEK) == 0)
697 			m->m_nextpkt = nextrecord;
698 		type = m->m_type;
699 		if (type == MT_OOBDATA)
700 			flags |= MSG_OOB;
701 	}
702 	moff = 0;
703 	offset = 0;
704 	while (m && uio->uio_resid > 0 && error == 0) {
705 		if (m->m_type == MT_OOBDATA) {
706 			if (type != MT_OOBDATA)
707 				break;
708 		} else if (type == MT_OOBDATA)
709 			break;
710 #ifdef DIAGNOSTIC
711 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
712 			panic("receive 3");
713 #endif
714 		so->so_state &= ~SS_RCVATMARK;
715 		len = uio->uio_resid;
716 		if (so->so_oobmark && len > so->so_oobmark - offset)
717 			len = so->so_oobmark - offset;
718 		if (len > m->m_len - moff)
719 			len = m->m_len - moff;
720 		/*
721 		 * If mp is set, just pass back the mbufs.
722 		 * Otherwise copy them out via the uio, then free.
723 		 * Sockbuf must be consistent here (points to current mbuf,
724 		 * it points to next record) when we drop priority;
725 		 * we must note any additions to the sockbuf when we
726 		 * block interrupts again.
727 		 */
728 		if (mp == 0) {
729 			splx(s);
730 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
731 			s = splsoftnet();
732 		} else
733 			uio->uio_resid -= len;
734 		if (len == m->m_len - moff) {
735 			if (m->m_flags & M_EOR)
736 				flags |= MSG_EOR;
737 			if (flags & MSG_PEEK) {
738 				m = m->m_next;
739 				moff = 0;
740 			} else {
741 				nextrecord = m->m_nextpkt;
742 				sbfree(&so->so_rcv, m);
743 				if (mp) {
744 					*mp = m;
745 					mp = &m->m_next;
746 					so->so_rcv.sb_mb = m = m->m_next;
747 					*mp = (struct mbuf *)0;
748 				} else {
749 					MFREE(m, so->so_rcv.sb_mb);
750 					m = so->so_rcv.sb_mb;
751 				}
752 				if (m)
753 					m->m_nextpkt = nextrecord;
754 			}
755 		} else {
756 			if (flags & MSG_PEEK)
757 				moff += len;
758 			else {
759 				if (mp)
760 					*mp = m_copym(m, 0, len, M_WAIT);
761 				m->m_data += len;
762 				m->m_len -= len;
763 				so->so_rcv.sb_cc -= len;
764 			}
765 		}
766 		if (so->so_oobmark) {
767 			if ((flags & MSG_PEEK) == 0) {
768 				so->so_oobmark -= len;
769 				if (so->so_oobmark == 0) {
770 					so->so_state |= SS_RCVATMARK;
771 					break;
772 				}
773 			} else {
774 				offset += len;
775 				if (offset == so->so_oobmark)
776 					break;
777 			}
778 		}
779 		if (flags & MSG_EOR)
780 			break;
781 		/*
782 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
783 		 * we must not quit until "uio->uio_resid == 0" or an error
784 		 * termination.  If a signal/timeout occurs, return
785 		 * with a short count but without error.
786 		 * Keep sockbuf locked against other readers.
787 		 */
788 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
789 		    !sosendallatonce(so) && !nextrecord) {
790 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
791 				break;
792 			error = sbwait(&so->so_rcv);
793 			if (error) {
794 				sbunlock(&so->so_rcv);
795 				splx(s);
796 				return (0);
797 			}
798 			if ((m = so->so_rcv.sb_mb) != NULL)
799 				nextrecord = m->m_nextpkt;
800 		}
801 	}
802 
803 	if (m && pr->pr_flags & PR_ATOMIC) {
804 		flags |= MSG_TRUNC;
805 		if ((flags & MSG_PEEK) == 0)
806 			(void) sbdroprecord(&so->so_rcv);
807 	}
808 	if ((flags & MSG_PEEK) == 0) {
809 		if (m == 0)
810 			so->so_rcv.sb_mb = nextrecord;
811 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
812 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
813 			    (struct mbuf *)(long)flags, (struct mbuf *)0,
814 			    (struct proc *)0);
815 	}
816 	if (orig_resid == uio->uio_resid && orig_resid &&
817 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
818 		sbunlock(&so->so_rcv);
819 		splx(s);
820 		goto restart;
821 	}
822 
823 	if (flagsp)
824 		*flagsp |= flags;
825 release:
826 	sbunlock(&so->so_rcv);
827 	splx(s);
828 	return (error);
829 }
830 
831 int
832 soshutdown(so, how)
833 	struct socket *so;
834 	int how;
835 {
836 	struct protosw *pr = so->so_proto;
837 
838 	if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
839 		return (EINVAL);
840 
841 	if (how == SHUT_RD || how == SHUT_RDWR)
842 		sorflush(so);
843 	if (how == SHUT_WR || how == SHUT_RDWR)
844 		return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0,
845 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
846 	return (0);
847 }
848 
849 void
850 sorflush(so)
851 	register struct socket *so;
852 {
853 	register struct sockbuf *sb = &so->so_rcv;
854 	register struct protosw *pr = so->so_proto;
855 	register int s;
856 	struct sockbuf asb;
857 
858 	sb->sb_flags |= SB_NOINTR;
859 	(void) sblock(sb, M_WAITOK);
860 	s = splimp();
861 	socantrcvmore(so);
862 	sbunlock(sb);
863 	asb = *sb;
864 	memset((caddr_t)sb, 0, sizeof(*sb));
865 	splx(s);
866 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
867 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
868 	sbrelease(&asb);
869 }
870 
871 int
872 sosetopt(so, level, optname, m0)
873 	register struct socket *so;
874 	int level, optname;
875 	struct mbuf *m0;
876 {
877 	int error = 0;
878 	register struct mbuf *m = m0;
879 
880 	if (level != SOL_SOCKET) {
881 		if (so->so_proto && so->so_proto->pr_ctloutput)
882 			return ((*so->so_proto->pr_ctloutput)
883 				  (PRCO_SETOPT, so, level, optname, &m0));
884 		error = ENOPROTOOPT;
885 	} else {
886 		switch (optname) {
887 
888 		case SO_LINGER:
889 			if (m == NULL || m->m_len != sizeof(struct linger)) {
890 				error = EINVAL;
891 				goto bad;
892 			}
893 			so->so_linger = mtod(m, struct linger *)->l_linger;
894 			/* fall thru... */
895 
896 		case SO_DEBUG:
897 		case SO_KEEPALIVE:
898 		case SO_DONTROUTE:
899 		case SO_USELOOPBACK:
900 		case SO_BROADCAST:
901 		case SO_REUSEADDR:
902 		case SO_REUSEPORT:
903 		case SO_OOBINLINE:
904 		case SO_TIMESTAMP:
905 			if (m == NULL || m->m_len < sizeof(int)) {
906 				error = EINVAL;
907 				goto bad;
908 			}
909 			if (*mtod(m, int *))
910 				so->so_options |= optname;
911 			else
912 				so->so_options &= ~optname;
913 			break;
914 
915 		case SO_SNDBUF:
916 		case SO_RCVBUF:
917 		case SO_SNDLOWAT:
918 		case SO_RCVLOWAT:
919 		    {
920 			int optval;
921 
922 			if (m == NULL || m->m_len < sizeof(int)) {
923 				error = EINVAL;
924 				goto bad;
925 			}
926 
927 			/*
928 			 * Values < 1 make no sense for any of these
929 			 * options, so disallow them.
930 			 */
931 			optval = *mtod(m, int *);
932 			if (optval < 1) {
933 				error = EINVAL;
934 				goto bad;
935 			}
936 
937 			switch (optname) {
938 
939 			case SO_SNDBUF:
940 			case SO_RCVBUF:
941 				if (sbreserve(optname == SO_SNDBUF ?
942 				    &so->so_snd : &so->so_rcv,
943 				    (u_long) optval) == 0) {
944 					error = ENOBUFS;
945 					goto bad;
946 				}
947 				break;
948 
949 			/*
950 			 * Make sure the low-water is never greater than
951 			 * the high-water.
952 			 */
953 			case SO_SNDLOWAT:
954 				so->so_snd.sb_lowat =
955 				    (optval > so->so_snd.sb_hiwat) ?
956 				    so->so_snd.sb_hiwat : optval;
957 				break;
958 			case SO_RCVLOWAT:
959 				so->so_rcv.sb_lowat =
960 				    (optval > so->so_rcv.sb_hiwat) ?
961 				    so->so_rcv.sb_hiwat : optval;
962 				break;
963 			}
964 			break;
965 		    }
966 
967 		case SO_SNDTIMEO:
968 		case SO_RCVTIMEO:
969 		    {
970 			struct timeval *tv;
971 			short val;
972 
973 			if (m == NULL || m->m_len < sizeof(*tv)) {
974 				error = EINVAL;
975 				goto bad;
976 			}
977 			tv = mtod(m, struct timeval *);
978 			if (tv->tv_sec * hz + tv->tv_usec / tick > SHRT_MAX) {
979 				error = EDOM;
980 				goto bad;
981 			}
982 			val = tv->tv_sec * hz + tv->tv_usec / tick;
983 
984 			switch (optname) {
985 
986 			case SO_SNDTIMEO:
987 				so->so_snd.sb_timeo = val;
988 				break;
989 			case SO_RCVTIMEO:
990 				so->so_rcv.sb_timeo = val;
991 				break;
992 			}
993 			break;
994 		    }
995 
996 		default:
997 			error = ENOPROTOOPT;
998 			break;
999 		}
1000 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1001 			(void) ((*so->so_proto->pr_ctloutput)
1002 				  (PRCO_SETOPT, so, level, optname, &m0));
1003 			m = NULL;	/* freed by protocol */
1004 		}
1005 	}
1006 bad:
1007 	if (m)
1008 		(void) m_free(m);
1009 	return (error);
1010 }
1011 
1012 int
1013 sogetopt(so, level, optname, mp)
1014 	register struct socket *so;
1015 	int level, optname;
1016 	struct mbuf **mp;
1017 {
1018 	register struct mbuf *m;
1019 
1020 	if (level != SOL_SOCKET) {
1021 		if (so->so_proto && so->so_proto->pr_ctloutput) {
1022 			return ((*so->so_proto->pr_ctloutput)
1023 				  (PRCO_GETOPT, so, level, optname, mp));
1024 		} else
1025 			return (ENOPROTOOPT);
1026 	} else {
1027 		m = m_get(M_WAIT, MT_SOOPTS);
1028 		m->m_len = sizeof(int);
1029 
1030 		switch (optname) {
1031 
1032 		case SO_LINGER:
1033 			m->m_len = sizeof(struct linger);
1034 			mtod(m, struct linger *)->l_onoff =
1035 				so->so_options & SO_LINGER;
1036 			mtod(m, struct linger *)->l_linger = so->so_linger;
1037 			break;
1038 
1039 		case SO_USELOOPBACK:
1040 		case SO_DONTROUTE:
1041 		case SO_DEBUG:
1042 		case SO_KEEPALIVE:
1043 		case SO_REUSEADDR:
1044 		case SO_REUSEPORT:
1045 		case SO_BROADCAST:
1046 		case SO_OOBINLINE:
1047 		case SO_TIMESTAMP:
1048 			*mtod(m, int *) = so->so_options & optname;
1049 			break;
1050 
1051 		case SO_TYPE:
1052 			*mtod(m, int *) = so->so_type;
1053 			break;
1054 
1055 		case SO_ERROR:
1056 			*mtod(m, int *) = so->so_error;
1057 			so->so_error = 0;
1058 			break;
1059 
1060 		case SO_SNDBUF:
1061 			*mtod(m, int *) = so->so_snd.sb_hiwat;
1062 			break;
1063 
1064 		case SO_RCVBUF:
1065 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
1066 			break;
1067 
1068 		case SO_SNDLOWAT:
1069 			*mtod(m, int *) = so->so_snd.sb_lowat;
1070 			break;
1071 
1072 		case SO_RCVLOWAT:
1073 			*mtod(m, int *) = so->so_rcv.sb_lowat;
1074 			break;
1075 
1076 		case SO_SNDTIMEO:
1077 		case SO_RCVTIMEO:
1078 		    {
1079 			int val = (optname == SO_SNDTIMEO ?
1080 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1081 
1082 			m->m_len = sizeof(struct timeval);
1083 			mtod(m, struct timeval *)->tv_sec = val / hz;
1084 			mtod(m, struct timeval *)->tv_usec =
1085 			    (val % hz) * tick;
1086 			break;
1087 		    }
1088 
1089 		default:
1090 			(void)m_free(m);
1091 			return (ENOPROTOOPT);
1092 		}
1093 		*mp = m;
1094 		return (0);
1095 	}
1096 }
1097 
1098 void
1099 sohasoutofband(so)
1100 	register struct socket *so;
1101 {
1102 	struct proc *p;
1103 
1104 	if (so->so_pgid < 0)
1105 		gsignal(-so->so_pgid, SIGURG);
1106 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1107 		psignal(p, SIGURG);
1108 	selwakeup(&so->so_rcv.sb_sel);
1109 }
1110