xref: /netbsd-src/sys/kern/uipc_socket.c (revision 76dfffe33547c37f8bdd446e3e4ab0f3c16cea4b)
1 /*	$NetBSD: uipc_socket.c,v 1.25 1996/08/14 05:53:18 explorer 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	8.3 (Berkeley) 4/15/94
36  */
37 
38 #include <sys/param.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/signalvar.h>
50 #include <sys/resourcevar.h>
51 
52 /*
53  * Socket operation routines.
54  * These routines are called by the routines in
55  * sys_socket.c or from a system process, and
56  * implement the semantics of socket operations by
57  * switching out to the protocol specific routines.
58  */
59 /*ARGSUSED*/
60 int
61 socreate(dom, aso, type, proto)
62 	int dom;
63 	struct socket **aso;
64 	register int type;
65 	int proto;
66 {
67 	struct proc *p = curproc;		/* XXX */
68 	register struct protosw *prp;
69 	register struct socket *so;
70 	register int error;
71 
72 	if (proto)
73 		prp = pffindproto(dom, proto, type);
74 	else
75 		prp = pffindtype(dom, type);
76 	if (prp == 0 || prp->pr_usrreq == 0)
77 		return (EPROTONOSUPPORT);
78 	if (prp->pr_type != type)
79 		return (EPROTOTYPE);
80 	MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_WAIT);
81 	bzero((caddr_t)so, sizeof(*so));
82 	so->so_type = type;
83 	so->so_proto = prp;
84 	error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
85 	    (struct mbuf *)(long)proto, (struct mbuf *)0, p);
86 	if (error) {
87 		so->so_state |= SS_NOFDREF;
88 		sofree(so);
89 		return (error);
90 	}
91 #ifdef COMPAT_SUNOS
92 	{
93 		extern struct emul emul_sunos;
94 		if (p->p_emul == &emul_sunos && type == SOCK_DGRAM)
95 			so->so_options |= SO_BROADCAST;
96 	}
97 #endif
98 	*aso = so;
99 	return (0);
100 }
101 
102 int
103 sobind(so, nam)
104 	struct socket *so;
105 	struct mbuf *nam;
106 {
107 	struct proc *p = curproc;		/* XXX */
108 	int s = splsoftnet();
109 	int error;
110 
111 	error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0,
112 	    nam, (struct mbuf *)0, p);
113 	splx(s);
114 	return (error);
115 }
116 
117 int
118 solisten(so, backlog)
119 	register struct socket *so;
120 	int backlog;
121 {
122 	int s = splsoftnet(), error;
123 
124 	error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0,
125 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
126 	if (error) {
127 		splx(s);
128 		return (error);
129 	}
130 	if (so->so_q == 0)
131 		so->so_options |= SO_ACCEPTCONN;
132 	if (backlog < 0)
133 		backlog = 0;
134 	so->so_qlimit = min(backlog, SOMAXCONN);
135 	splx(s);
136 	return (0);
137 }
138 
139 void
140 sofree(so)
141 	register struct socket *so;
142 {
143 
144 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
145 		return;
146 	if (so->so_head) {
147 		if (!soqremque(so, 0) && !soqremque(so, 1))
148 			panic("sofree dq");
149 		so->so_head = 0;
150 	}
151 	sbrelease(&so->so_snd);
152 	sorflush(so);
153 	FREE(so, M_SOCKET);
154 }
155 
156 /*
157  * Close a socket on last file table reference removal.
158  * Initiate disconnect if connected.
159  * Free socket when disconnect complete.
160  */
161 int
162 soclose(so)
163 	register struct socket *so;
164 {
165 	int s = splsoftnet();		/* conservative */
166 	int error = 0;
167 
168 	if (so->so_options & SO_ACCEPTCONN) {
169 		while (so->so_q0)
170 			(void) soabort(so->so_q0);
171 		while (so->so_q)
172 			(void) soabort(so->so_q);
173 	}
174 	if (so->so_pcb == 0)
175 		goto discard;
176 	if (so->so_state & SS_ISCONNECTED) {
177 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
178 			error = sodisconnect(so);
179 			if (error)
180 				goto drop;
181 		}
182 		if (so->so_options & SO_LINGER) {
183 			if ((so->so_state & SS_ISDISCONNECTING) &&
184 			    (so->so_state & SS_NBIO))
185 				goto drop;
186 			while (so->so_state & SS_ISCONNECTED) {
187 				error = tsleep((caddr_t)&so->so_timeo,
188 					       PSOCK | PCATCH, netcls,
189 					       so->so_linger);
190 				if (error)
191 					break;
192 			}
193 		}
194 	}
195 drop:
196 	if (so->so_pcb) {
197 		int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
198 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
199 		    (struct proc *)0);
200 		if (error == 0)
201 			error = error2;
202 	}
203 discard:
204 	if (so->so_state & SS_NOFDREF)
205 		panic("soclose: NOFDREF");
206 	so->so_state |= SS_NOFDREF;
207 	sofree(so);
208 	splx(s);
209 	return (error);
210 }
211 
212 /*
213  * Must be called at splsoftnet...
214  */
215 int
216 soabort(so)
217 	struct socket *so;
218 {
219 
220 	return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0,
221 	    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
222 }
223 
224 int
225 soaccept(so, nam)
226 	register struct socket *so;
227 	struct mbuf *nam;
228 {
229 	int s = splsoftnet();
230 	int error;
231 
232 	if ((so->so_state & SS_NOFDREF) == 0)
233 		panic("soaccept: !NOFDREF");
234 	so->so_state &= ~SS_NOFDREF;
235 	error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, (struct mbuf *)0,
236 	    nam, (struct mbuf *)0, (struct proc *)0);
237 	splx(s);
238 	return (error);
239 }
240 
241 int
242 soconnect(so, nam)
243 	register struct socket *so;
244 	struct mbuf *nam;
245 {
246 	struct proc *p = curproc;		/* XXX */
247 	int s;
248 	int error;
249 
250 	if (so->so_options & SO_ACCEPTCONN)
251 		return (EOPNOTSUPP);
252 	s = splsoftnet();
253 	/*
254 	 * If protocol is connection-based, can only connect once.
255 	 * Otherwise, if connected, try to disconnect first.
256 	 * This allows user to disconnect by connecting to, e.g.,
257 	 * a null address.
258 	 */
259 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
260 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
261 	    (error = sodisconnect(so))))
262 		error = EISCONN;
263 	else
264 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
265 		    (struct mbuf *)0, nam, (struct mbuf *)0, p);
266 	splx(s);
267 	return (error);
268 }
269 
270 int
271 soconnect2(so1, so2)
272 	register struct socket *so1;
273 	struct socket *so2;
274 {
275 	int s = splsoftnet();
276 	int error;
277 
278 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
279 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0,
280 	    (struct proc *)0);
281 	splx(s);
282 	return (error);
283 }
284 
285 int
286 sodisconnect(so)
287 	register struct socket *so;
288 {
289 	int s = splsoftnet();
290 	int error;
291 
292 	if ((so->so_state & SS_ISCONNECTED) == 0) {
293 		error = ENOTCONN;
294 		goto bad;
295 	}
296 	if (so->so_state & SS_ISDISCONNECTING) {
297 		error = EALREADY;
298 		goto bad;
299 	}
300 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
301 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
302 	    (struct proc *)0);
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 	if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
365 		goto out;
366 	do {
367 		s = splsoftnet();
368 		if (so->so_state & SS_CANTSENDMORE)
369 			snderr(EPIPE);
370 		if (so->so_error)
371 			snderr(so->so_error);
372 		if ((so->so_state & SS_ISCONNECTED) == 0) {
373 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
374 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
375 				    !(resid == 0 && clen != 0))
376 					snderr(ENOTCONN);
377 			} else if (addr == 0)
378 				snderr(EDESTADDRREQ);
379 		}
380 		space = sbspace(&so->so_snd);
381 		if (flags & MSG_OOB)
382 			space += 1024;
383 		if ((atomic && resid > so->so_snd.sb_hiwat) ||
384 		    clen > so->so_snd.sb_hiwat)
385 			snderr(EMSGSIZE);
386 		if (space < resid + clen && uio &&
387 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
388 			if (so->so_state & SS_NBIO)
389 				snderr(EWOULDBLOCK);
390 			sbunlock(&so->so_snd);
391 			error = sbwait(&so->so_snd);
392 			splx(s);
393 			if (error)
394 				goto out;
395 			goto restart;
396 		}
397 		splx(s);
398 		mp = &top;
399 		space -= clen;
400 		do {
401 		    if (uio == NULL) {
402 			/*
403 			 * Data is prepackaged in "top".
404 			 */
405 			resid = 0;
406 			if (flags & MSG_EOR)
407 				top->m_flags |= M_EOR;
408 		    } else do {
409 			if (top == 0) {
410 				MGETHDR(m, M_WAIT, MT_DATA);
411 				mlen = MHLEN;
412 				m->m_pkthdr.len = 0;
413 				m->m_pkthdr.rcvif = (struct ifnet *)0;
414 			} else {
415 				MGET(m, M_WAIT, MT_DATA);
416 				mlen = MLEN;
417 			}
418 			if (resid >= MINCLSIZE && space >= MCLBYTES) {
419 				MCLGET(m, M_WAIT);
420 				if ((m->m_flags & M_EXT) == 0)
421 					goto nopages;
422 				mlen = MCLBYTES;
423 #ifdef	MAPPED_MBUFS
424 				len = min(MCLBYTES, resid);
425 #else
426 				if (atomic && top == 0) {
427 					len = min(MCLBYTES - max_hdr, resid);
428 					m->m_data += max_hdr;
429 				} else
430 					len = min(MCLBYTES, resid);
431 #endif
432 				space -= len;
433 			} else {
434 nopages:
435 				len = min(min(mlen, resid), space);
436 				space -= len;
437 				/*
438 				 * For datagram protocols, leave room
439 				 * for protocol headers in first mbuf.
440 				 */
441 				if (atomic && top == 0 && len < mlen)
442 					MH_ALIGN(m, len);
443 			}
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 = splsoftnet();				/* XXX */
461 		    error = (*so->so_proto->pr_usrreq)(so,
462 			(flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
463 			top, addr, control, p);
464 		    splx(s);
465 		    if (dontroute)
466 			    so->so_options &= ~SO_DONTROUTE;
467 		    clen = 0;
468 		    control = 0;
469 		    top = 0;
470 		    mp = &top;
471 		    if (error)
472 			goto release;
473 		} while (resid && space > 0);
474 	} while (resid);
475 
476 release:
477 	sbunlock(&so->so_snd);
478 out:
479 	if (top)
480 		m_freem(top);
481 	if (control)
482 		m_freem(control);
483 	return (error);
484 }
485 
486 /*
487  * Implement receive operations on a socket.
488  * We depend on the way that records are added to the sockbuf
489  * by sbappend*.  In particular, each record (mbufs linked through m_next)
490  * must begin with an address if the protocol so specifies,
491  * followed by an optional mbuf or mbufs containing ancillary data,
492  * and then zero or more mbufs of data.
493  * In order to avoid blocking network interrupts for the entire time here,
494  * we splx() while doing the actual copy to user space.
495  * Although the sockbuf is locked, new data may still be appended,
496  * and thus we must maintain consistency of the sockbuf during that time.
497  *
498  * The caller may receive the data as a single mbuf chain by supplying
499  * an mbuf **mp0 for use in returning the chain.  The uio is then used
500  * only for the count in uio_resid.
501  */
502 int
503 soreceive(so, paddr, uio, mp0, controlp, flagsp)
504 	register struct socket *so;
505 	struct mbuf **paddr;
506 	struct uio *uio;
507 	struct mbuf **mp0;
508 	struct mbuf **controlp;
509 	int *flagsp;
510 {
511 	register struct mbuf *m, **mp;
512 	register int flags, len, error, s, offset;
513 	struct protosw *pr = so->so_proto;
514 	struct mbuf *nextrecord;
515 	int moff, type = 0;
516 	int orig_resid = uio->uio_resid;
517 
518 	mp = mp0;
519 	if (paddr)
520 		*paddr = 0;
521 	if (controlp)
522 		*controlp = 0;
523 	if (flagsp)
524 		flags = *flagsp &~ MSG_EOR;
525 	else
526 		flags = 0;
527 	if (flags & MSG_OOB) {
528 		m = m_get(M_WAIT, MT_DATA);
529 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
530 		    (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0,
531 		    (struct proc *)0);
532 		if (error)
533 			goto bad;
534 		do {
535 			error = uiomove(mtod(m, caddr_t),
536 			    (int) min(uio->uio_resid, m->m_len), uio);
537 			m = m_free(m);
538 		} while (uio->uio_resid && error == 0 && m);
539 bad:
540 		if (m)
541 			m_freem(m);
542 		return (error);
543 	}
544 	if (mp)
545 		*mp = (struct mbuf *)0;
546 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
547 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
548 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
549 
550 restart:
551 	if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
552 		return (error);
553 	s = splsoftnet();
554 
555 	m = so->so_rcv.sb_mb;
556 	/*
557 	 * If we have less data than requested, block awaiting more
558 	 * (subject to any timeout) if:
559 	 *   1. the current count is less than the low water mark,
560 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
561 	 *	receive operation at once if we block (resid <= hiwat), or
562 	 *   3. MSG_DONTWAIT is not set.
563 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
564 	 * we have to do the receive in sections, and thus risk returning
565 	 * a short count if a timeout or signal occurs after we start.
566 	 */
567 	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
568 	    so->so_rcv.sb_cc < uio->uio_resid) &&
569 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
570 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
571 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
572 #ifdef DIAGNOSTIC
573 		if (m == 0 && so->so_rcv.sb_cc)
574 			panic("receive 1");
575 #endif
576 		if (so->so_error) {
577 			if (m)
578 				goto dontblock;
579 			error = so->so_error;
580 			if ((flags & MSG_PEEK) == 0)
581 				so->so_error = 0;
582 			goto release;
583 		}
584 		if (so->so_state & SS_CANTRCVMORE) {
585 			if (m)
586 				goto dontblock;
587 			else
588 				goto release;
589 		}
590 		for (; m; m = m->m_next)
591 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
592 				m = so->so_rcv.sb_mb;
593 				goto dontblock;
594 			}
595 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
596 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
597 			error = ENOTCONN;
598 			goto release;
599 		}
600 		if (uio->uio_resid == 0)
601 			goto release;
602 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
603 			error = EWOULDBLOCK;
604 			goto release;
605 		}
606 		sbunlock(&so->so_rcv);
607 		error = sbwait(&so->so_rcv);
608 		splx(s);
609 		if (error)
610 			return (error);
611 		goto restart;
612 	}
613 dontblock:
614 #ifdef notyet /* XXXX */
615 	if (uio->uio_procp)
616 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
617 #endif
618 	nextrecord = m->m_nextpkt;
619 	if (pr->pr_flags & PR_ADDR) {
620 #ifdef DIAGNOSTIC
621 		if (m->m_type != MT_SONAME)
622 			panic("receive 1a");
623 #endif
624 		orig_resid = 0;
625 		if (flags & MSG_PEEK) {
626 			if (paddr)
627 				*paddr = m_copy(m, 0, m->m_len);
628 			m = m->m_next;
629 		} else {
630 			sbfree(&so->so_rcv, m);
631 			if (paddr) {
632 				*paddr = m;
633 				so->so_rcv.sb_mb = m->m_next;
634 				m->m_next = 0;
635 				m = so->so_rcv.sb_mb;
636 			} else {
637 				MFREE(m, so->so_rcv.sb_mb);
638 				m = so->so_rcv.sb_mb;
639 			}
640 		}
641 	}
642 	while (m && m->m_type == MT_CONTROL && error == 0) {
643 		if (flags & MSG_PEEK) {
644 			if (controlp)
645 				*controlp = m_copy(m, 0, m->m_len);
646 			m = m->m_next;
647 		} else {
648 			sbfree(&so->so_rcv, m);
649 			if (controlp) {
650 				if (pr->pr_domain->dom_externalize &&
651 				    mtod(m, struct cmsghdr *)->cmsg_type ==
652 				    SCM_RIGHTS)
653 				   error = (*pr->pr_domain->dom_externalize)(m);
654 				*controlp = m;
655 				so->so_rcv.sb_mb = m->m_next;
656 				m->m_next = 0;
657 				m = so->so_rcv.sb_mb;
658 			} else {
659 				MFREE(m, so->so_rcv.sb_mb);
660 				m = so->so_rcv.sb_mb;
661 			}
662 		}
663 		if (controlp) {
664 			orig_resid = 0;
665 			controlp = &(*controlp)->m_next;
666 		}
667 	}
668 	if (m) {
669 		if ((flags & MSG_PEEK) == 0)
670 			m->m_nextpkt = nextrecord;
671 		type = m->m_type;
672 		if (type == MT_OOBDATA)
673 			flags |= MSG_OOB;
674 	}
675 	moff = 0;
676 	offset = 0;
677 	while (m && uio->uio_resid > 0 && error == 0) {
678 		if (m->m_type == MT_OOBDATA) {
679 			if (type != MT_OOBDATA)
680 				break;
681 		} else if (type == MT_OOBDATA)
682 			break;
683 #ifdef DIAGNOSTIC
684 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
685 			panic("receive 3");
686 #endif
687 		so->so_state &= ~SS_RCVATMARK;
688 		len = uio->uio_resid;
689 		if (so->so_oobmark && len > so->so_oobmark - offset)
690 			len = so->so_oobmark - offset;
691 		if (len > m->m_len - moff)
692 			len = m->m_len - moff;
693 		/*
694 		 * If mp is set, just pass back the mbufs.
695 		 * Otherwise copy them out via the uio, then free.
696 		 * Sockbuf must be consistent here (points to current mbuf,
697 		 * it points to next record) when we drop priority;
698 		 * we must note any additions to the sockbuf when we
699 		 * block interrupts again.
700 		 */
701 		if (mp == 0) {
702 			splx(s);
703 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
704 			s = splsoftnet();
705 		} else
706 			uio->uio_resid -= len;
707 		if (len == m->m_len - moff) {
708 			if (m->m_flags & M_EOR)
709 				flags |= MSG_EOR;
710 			if (flags & MSG_PEEK) {
711 				m = m->m_next;
712 				moff = 0;
713 			} else {
714 				nextrecord = m->m_nextpkt;
715 				sbfree(&so->so_rcv, m);
716 				if (mp) {
717 					*mp = m;
718 					mp = &m->m_next;
719 					so->so_rcv.sb_mb = m = m->m_next;
720 					*mp = (struct mbuf *)0;
721 				} else {
722 					MFREE(m, so->so_rcv.sb_mb);
723 					m = so->so_rcv.sb_mb;
724 				}
725 				if (m)
726 					m->m_nextpkt = nextrecord;
727 			}
728 		} else {
729 			if (flags & MSG_PEEK)
730 				moff += len;
731 			else {
732 				if (mp)
733 					*mp = m_copym(m, 0, len, M_WAIT);
734 				m->m_data += len;
735 				m->m_len -= len;
736 				so->so_rcv.sb_cc -= len;
737 			}
738 		}
739 		if (so->so_oobmark) {
740 			if ((flags & MSG_PEEK) == 0) {
741 				so->so_oobmark -= len;
742 				if (so->so_oobmark == 0) {
743 					so->so_state |= SS_RCVATMARK;
744 					break;
745 				}
746 			} else {
747 				offset += len;
748 				if (offset == so->so_oobmark)
749 					break;
750 			}
751 		}
752 		if (flags & MSG_EOR)
753 			break;
754 		/*
755 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
756 		 * we must not quit until "uio->uio_resid == 0" or an error
757 		 * termination.  If a signal/timeout occurs, return
758 		 * with a short count but without error.
759 		 * Keep sockbuf locked against other readers.
760 		 */
761 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
762 		    !sosendallatonce(so) && !nextrecord) {
763 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
764 				break;
765 			error = sbwait(&so->so_rcv);
766 			if (error) {
767 				sbunlock(&so->so_rcv);
768 				splx(s);
769 				return (0);
770 			}
771 			if ((m = so->so_rcv.sb_mb) != NULL)
772 				nextrecord = m->m_nextpkt;
773 		}
774 	}
775 
776 	if (m && pr->pr_flags & PR_ATOMIC) {
777 		flags |= MSG_TRUNC;
778 		if ((flags & MSG_PEEK) == 0)
779 			(void) sbdroprecord(&so->so_rcv);
780 	}
781 	if ((flags & MSG_PEEK) == 0) {
782 		if (m == 0)
783 			so->so_rcv.sb_mb = nextrecord;
784 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
785 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
786 			    (struct mbuf *)(long)flags, (struct mbuf *)0,
787 			    (struct proc *)0);
788 	}
789 	if (orig_resid == uio->uio_resid && orig_resid &&
790 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
791 		sbunlock(&so->so_rcv);
792 		splx(s);
793 		goto restart;
794 	}
795 
796 	if (flagsp)
797 		*flagsp |= flags;
798 release:
799 	sbunlock(&so->so_rcv);
800 	splx(s);
801 	return (error);
802 }
803 
804 int
805 soshutdown(so, how)
806 	register struct socket *so;
807 	register int how;
808 {
809 	register struct protosw *pr = so->so_proto;
810 
811 	how++;
812 	if (how & FREAD)
813 		sorflush(so);
814 	if (how & FWRITE)
815 		return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0,
816 		    (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
817 	return (0);
818 }
819 
820 void
821 sorflush(so)
822 	register struct socket *so;
823 {
824 	register struct sockbuf *sb = &so->so_rcv;
825 	register struct protosw *pr = so->so_proto;
826 	register int s;
827 	struct sockbuf asb;
828 
829 	sb->sb_flags |= SB_NOINTR;
830 	(void) sblock(sb, M_WAITOK);
831 	s = splimp();
832 	socantrcvmore(so);
833 	sbunlock(sb);
834 	asb = *sb;
835 	bzero((caddr_t)sb, sizeof (*sb));
836 	splx(s);
837 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
838 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
839 	sbrelease(&asb);
840 }
841 
842 int
843 sosetopt(so, level, optname, m0)
844 	register struct socket *so;
845 	int level, optname;
846 	struct mbuf *m0;
847 {
848 	int error = 0;
849 	register struct mbuf *m = m0;
850 
851 	if (level != SOL_SOCKET) {
852 		if (so->so_proto && so->so_proto->pr_ctloutput)
853 			return ((*so->so_proto->pr_ctloutput)
854 				  (PRCO_SETOPT, so, level, optname, &m0));
855 		error = ENOPROTOOPT;
856 	} else {
857 		switch (optname) {
858 
859 		case SO_LINGER:
860 			if (m == NULL || m->m_len != sizeof (struct linger)) {
861 				error = EINVAL;
862 				goto bad;
863 			}
864 			so->so_linger = mtod(m, struct linger *)->l_linger;
865 			/* fall thru... */
866 
867 		case SO_DEBUG:
868 		case SO_KEEPALIVE:
869 		case SO_DONTROUTE:
870 		case SO_USELOOPBACK:
871 		case SO_BROADCAST:
872 		case SO_REUSEADDR:
873 		case SO_REUSEPORT:
874 		case SO_OOBINLINE:
875 			if (m == NULL || m->m_len < sizeof (int)) {
876 				error = EINVAL;
877 				goto bad;
878 			}
879 			if (*mtod(m, int *))
880 				so->so_options |= optname;
881 			else
882 				so->so_options &= ~optname;
883 			break;
884 
885 		case SO_SNDBUF:
886 		case SO_RCVBUF:
887 		case SO_SNDLOWAT:
888 		case SO_RCVLOWAT:
889 			if (m == NULL || m->m_len < sizeof (int)) {
890 				error = EINVAL;
891 				goto bad;
892 			}
893 			switch (optname) {
894 
895 			case SO_SNDBUF:
896 			case SO_RCVBUF:
897 				if (sbreserve(optname == SO_SNDBUF ?
898 				    &so->so_snd : &so->so_rcv,
899 				    (u_long) *mtod(m, int *)) == 0) {
900 					error = ENOBUFS;
901 					goto bad;
902 				}
903 				break;
904 
905 			case SO_SNDLOWAT:
906 				so->so_snd.sb_lowat = *mtod(m, int *);
907 				break;
908 			case SO_RCVLOWAT:
909 				so->so_rcv.sb_lowat = *mtod(m, int *);
910 				break;
911 			}
912 			break;
913 
914 		case SO_SNDTIMEO:
915 		case SO_RCVTIMEO:
916 		    {
917 			struct timeval *tv;
918 			short val;
919 
920 			if (m == NULL || m->m_len < sizeof (*tv)) {
921 				error = EINVAL;
922 				goto bad;
923 			}
924 			tv = mtod(m, struct timeval *);
925 			if (tv->tv_sec * hz + tv->tv_usec / tick > SHRT_MAX) {
926 				error = EDOM;
927 				goto bad;
928 			}
929 			val = tv->tv_sec * hz + tv->tv_usec / tick;
930 
931 			switch (optname) {
932 
933 			case SO_SNDTIMEO:
934 				so->so_snd.sb_timeo = val;
935 				break;
936 			case SO_RCVTIMEO:
937 				so->so_rcv.sb_timeo = val;
938 				break;
939 			}
940 			break;
941 		    }
942 
943 		default:
944 			error = ENOPROTOOPT;
945 			break;
946 		}
947 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
948 			(void) ((*so->so_proto->pr_ctloutput)
949 				  (PRCO_SETOPT, so, level, optname, &m0));
950 			m = NULL;	/* freed by protocol */
951 		}
952 	}
953 bad:
954 	if (m)
955 		(void) m_free(m);
956 	return (error);
957 }
958 
959 int
960 sogetopt(so, level, optname, mp)
961 	register struct socket *so;
962 	int level, optname;
963 	struct mbuf **mp;
964 {
965 	register struct mbuf *m;
966 
967 	if (level != SOL_SOCKET) {
968 		if (so->so_proto && so->so_proto->pr_ctloutput) {
969 			return ((*so->so_proto->pr_ctloutput)
970 				  (PRCO_GETOPT, so, level, optname, mp));
971 		} else
972 			return (ENOPROTOOPT);
973 	} else {
974 		m = m_get(M_WAIT, MT_SOOPTS);
975 		m->m_len = sizeof (int);
976 
977 		switch (optname) {
978 
979 		case SO_LINGER:
980 			m->m_len = sizeof (struct linger);
981 			mtod(m, struct linger *)->l_onoff =
982 				so->so_options & SO_LINGER;
983 			mtod(m, struct linger *)->l_linger = so->so_linger;
984 			break;
985 
986 		case SO_USELOOPBACK:
987 		case SO_DONTROUTE:
988 		case SO_DEBUG:
989 		case SO_KEEPALIVE:
990 		case SO_REUSEADDR:
991 		case SO_REUSEPORT:
992 		case SO_BROADCAST:
993 		case SO_OOBINLINE:
994 			*mtod(m, int *) = so->so_options & optname;
995 			break;
996 
997 		case SO_TYPE:
998 			*mtod(m, int *) = so->so_type;
999 			break;
1000 
1001 		case SO_ERROR:
1002 			*mtod(m, int *) = so->so_error;
1003 			so->so_error = 0;
1004 			break;
1005 
1006 		case SO_SNDBUF:
1007 			*mtod(m, int *) = so->so_snd.sb_hiwat;
1008 			break;
1009 
1010 		case SO_RCVBUF:
1011 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
1012 			break;
1013 
1014 		case SO_SNDLOWAT:
1015 			*mtod(m, int *) = so->so_snd.sb_lowat;
1016 			break;
1017 
1018 		case SO_RCVLOWAT:
1019 			*mtod(m, int *) = so->so_rcv.sb_lowat;
1020 			break;
1021 
1022 		case SO_SNDTIMEO:
1023 		case SO_RCVTIMEO:
1024 		    {
1025 			int val = (optname == SO_SNDTIMEO ?
1026 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1027 
1028 			m->m_len = sizeof(struct timeval);
1029 			mtod(m, struct timeval *)->tv_sec = val / hz;
1030 			mtod(m, struct timeval *)->tv_usec =
1031 			    (val % hz) / tick;
1032 			break;
1033 		    }
1034 
1035 		default:
1036 			(void)m_free(m);
1037 			return (ENOPROTOOPT);
1038 		}
1039 		*mp = m;
1040 		return (0);
1041 	}
1042 }
1043 
1044 void
1045 sohasoutofband(so)
1046 	register struct socket *so;
1047 {
1048 	struct proc *p;
1049 
1050 	if (so->so_pgid < 0)
1051 		gsignal(-so->so_pgid, SIGURG);
1052 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1053 		psignal(p, SIGURG);
1054 	selwakeup(&so->so_rcv.sb_sel);
1055 }
1056