xref: /freebsd/sys/kern/uipc_socket.c (revision afe61c15161c324a7af299a9b8457aba5afc92db)
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  */
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/proc.h>
39 #include <sys/file.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/domain.h>
43 #include <sys/kernel.h>
44 #include <sys/protosw.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/resourcevar.h>
48 
49 void	sofree		__P((struct socket *));
50 void	sorflush	__P((struct socket *));
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 	if (p->p_ucred->cr_uid == 0)
84 		so->so_state = SS_PRIV;
85 	so->so_proto = prp;
86 	error =
87 	    (*prp->pr_usrreq)(so, PRU_ATTACH,
88 		(struct mbuf *)0, (struct mbuf *)proto, (struct mbuf *)0);
89 	if (error) {
90 		so->so_state |= SS_NOFDREF;
91 		sofree(so);
92 		return (error);
93 	}
94 	*aso = so;
95 	return (0);
96 }
97 
98 int
99 sobind(so, nam)
100 	struct socket *so;
101 	struct mbuf *nam;
102 {
103 	int s = splnet();
104 	int error;
105 
106 	error =
107 	    (*so->so_proto->pr_usrreq)(so, PRU_BIND,
108 		(struct mbuf *)0, nam, (struct mbuf *)0);
109 	splx(s);
110 	return (error);
111 }
112 
113 int
114 solisten(so, backlog)
115 	register struct socket *so;
116 	int backlog;
117 {
118 	int s = splnet(), error;
119 
120 	error =
121 	    (*so->so_proto->pr_usrreq)(so, PRU_LISTEN,
122 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
123 	if (error) {
124 		splx(s);
125 		return (error);
126 	}
127 	if (so->so_q == 0)
128 		so->so_options |= SO_ACCEPTCONN;
129 	if (backlog < 0)
130 		backlog = 0;
131 	so->so_qlimit = min(backlog, SOMAXCONN);
132 	splx(s);
133 	return (0);
134 }
135 
136 void
137 sofree(so)
138 	register struct socket *so;
139 {
140 
141 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
142 		return;
143 	if (so->so_head) {
144 		if (!soqremque(so, 0) && !soqremque(so, 1))
145 			panic("sofree dq");
146 		so->so_head = 0;
147 	}
148 	sbrelease(&so->so_snd);
149 	sorflush(so);
150 	FREE(so, M_SOCKET);
151 }
152 
153 /*
154  * Close a socket on last file table reference removal.
155  * Initiate disconnect if connected.
156  * Free socket when disconnect complete.
157  */
158 int
159 soclose(so)
160 	register struct socket *so;
161 {
162 	int s = splnet();		/* conservative */
163 	int error = 0;
164 
165 	if (so->so_options & SO_ACCEPTCONN) {
166 		while (so->so_q0)
167 			(void) soabort(so->so_q0);
168 		while (so->so_q)
169 			(void) soabort(so->so_q);
170 	}
171 	if (so->so_pcb == 0)
172 		goto discard;
173 	if (so->so_state & SS_ISCONNECTED) {
174 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
175 			error = sodisconnect(so);
176 			if (error)
177 				goto drop;
178 		}
179 		if (so->so_options & SO_LINGER) {
180 			if ((so->so_state & SS_ISDISCONNECTING) &&
181 			    (so->so_state & SS_NBIO))
182 				goto drop;
183 			while (so->so_state & SS_ISCONNECTED)
184 				if (error = tsleep((caddr_t)&so->so_timeo,
185 				    PSOCK | PCATCH, netcls, so->so_linger))
186 					break;
187 		}
188 	}
189 drop:
190 	if (so->so_pcb) {
191 		int error2 =
192 		    (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
193 			(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
194 		if (error == 0)
195 			error = error2;
196 	}
197 discard:
198 	if (so->so_state & SS_NOFDREF)
199 		panic("soclose: NOFDREF");
200 	so->so_state |= SS_NOFDREF;
201 	sofree(so);
202 	splx(s);
203 	return (error);
204 }
205 
206 /*
207  * Must be called at splnet...
208  */
209 int
210 soabort(so)
211 	struct socket *so;
212 {
213 
214 	return (
215 	    (*so->so_proto->pr_usrreq)(so, PRU_ABORT,
216 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
217 }
218 
219 int
220 soaccept(so, nam)
221 	register struct socket *so;
222 	struct mbuf *nam;
223 {
224 	int s = splnet();
225 	int error;
226 
227 	if ((so->so_state & SS_NOFDREF) == 0)
228 		panic("soaccept: !NOFDREF");
229 	so->so_state &= ~SS_NOFDREF;
230 	error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
231 	    (struct mbuf *)0, nam, (struct mbuf *)0);
232 	splx(s);
233 	return (error);
234 }
235 
236 int
237 soconnect(so, nam)
238 	register struct socket *so;
239 	struct mbuf *nam;
240 {
241 	int s;
242 	int error;
243 
244 	if (so->so_options & SO_ACCEPTCONN)
245 		return (EOPNOTSUPP);
246 	s = splnet();
247 	/*
248 	 * If protocol is connection-based, can only connect once.
249 	 * Otherwise, if connected, try to disconnect first.
250 	 * This allows user to disconnect by connecting to, e.g.,
251 	 * a null address.
252 	 */
253 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
254 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
255 	    (error = sodisconnect(so))))
256 		error = EISCONN;
257 	else
258 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
259 		    (struct mbuf *)0, nam, (struct mbuf *)0);
260 	splx(s);
261 	return (error);
262 }
263 
264 int
265 soconnect2(so1, so2)
266 	register struct socket *so1;
267 	struct socket *so2;
268 {
269 	int s = splnet();
270 	int error;
271 
272 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
273 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0);
274 	splx(s);
275 	return (error);
276 }
277 
278 int
279 sodisconnect(so)
280 	register struct socket *so;
281 {
282 	int s = splnet();
283 	int error;
284 
285 	if ((so->so_state & SS_ISCONNECTED) == 0) {
286 		error = ENOTCONN;
287 		goto bad;
288 	}
289 	if (so->so_state & SS_ISDISCONNECTING) {
290 		error = EALREADY;
291 		goto bad;
292 	}
293 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
294 	    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
295 bad:
296 	splx(s);
297 	return (error);
298 }
299 
300 #define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
301 /*
302  * Send on a socket.
303  * If send must go all at once and message is larger than
304  * send buffering, then hard error.
305  * Lock against other senders.
306  * If must go all at once and not enough room now, then
307  * inform user that this would block and do nothing.
308  * Otherwise, if nonblocking, send as much as possible.
309  * The data to be sent is described by "uio" if nonzero,
310  * otherwise by the mbuf chain "top" (which must be null
311  * if uio is not).  Data provided in mbuf chain must be small
312  * enough to send all at once.
313  *
314  * Returns nonzero on error, timeout or signal; callers
315  * must check for short counts if EINTR/ERESTART are returned.
316  * Data and control buffers are freed on return.
317  */
318 int
319 sosend(so, addr, uio, top, control, flags)
320 	register struct socket *so;
321 	struct mbuf *addr;
322 	struct uio *uio;
323 	struct mbuf *top;
324 	struct mbuf *control;
325 	int flags;
326 {
327 	struct proc *p = curproc;		/* XXX */
328 	struct mbuf **mp;
329 	register struct mbuf *m;
330 	register long space, len, resid;
331 	int clen = 0, error, s, dontroute, mlen;
332 	int atomic = sosendallatonce(so) || top;
333 
334 	if (uio)
335 		resid = uio->uio_resid;
336 	else
337 		resid = top->m_pkthdr.len;
338 	/*
339 	 * In theory resid should be unsigned.
340 	 * However, space must be signed, as it might be less than 0
341 	 * if we over-committed, and we must use a signed comparison
342 	 * of space and resid.  On the other hand, a negative resid
343 	 * causes us to loop sending 0-length segments to the protocol.
344 	 */
345 	if (resid < 0)
346 		return (EINVAL);
347 	dontroute =
348 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
349 	    (so->so_proto->pr_flags & PR_ATOMIC);
350 	p->p_stats->p_ru.ru_msgsnd++;
351 	if (control)
352 		clen = control->m_len;
353 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
354 
355 restart:
356 	if (error = sblock(&so->so_snd, SBLOCKWAIT(flags)))
357 		goto out;
358 	do {
359 		s = splnet();
360 		if (so->so_state & SS_CANTSENDMORE)
361 			snderr(EPIPE);
362 		if (so->so_error)
363 			snderr(so->so_error);
364 		if ((so->so_state & SS_ISCONNECTED) == 0) {
365 			if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
366 				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
367 				    !(resid == 0 && clen != 0))
368 					snderr(ENOTCONN);
369 			} else if (addr == 0)
370 				snderr(EDESTADDRREQ);
371 		}
372 		space = sbspace(&so->so_snd);
373 		if (flags & MSG_OOB)
374 			space += 1024;
375 		if (atomic && resid > so->so_snd.sb_hiwat ||
376 		    clen > so->so_snd.sb_hiwat)
377 			snderr(EMSGSIZE);
378 		if (space < resid + clen && uio &&
379 		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
380 			if (so->so_state & SS_NBIO)
381 				snderr(EWOULDBLOCK);
382 			sbunlock(&so->so_snd);
383 			error = sbwait(&so->so_snd);
384 			splx(s);
385 			if (error)
386 				goto out;
387 			goto restart;
388 		}
389 		splx(s);
390 		mp = &top;
391 		space -= clen;
392 		do {
393 		    if (uio == NULL) {
394 			/*
395 			 * Data is prepackaged in "top".
396 			 */
397 			resid = 0;
398 			if (flags & MSG_EOR)
399 				top->m_flags |= M_EOR;
400 		    } else do {
401 			if (top == 0) {
402 				MGETHDR(m, M_WAIT, MT_DATA);
403 				mlen = MHLEN;
404 				m->m_pkthdr.len = 0;
405 				m->m_pkthdr.rcvif = (struct ifnet *)0;
406 			} else {
407 				MGET(m, M_WAIT, MT_DATA);
408 				mlen = MLEN;
409 			}
410 			if (resid >= MINCLSIZE) {
411 				MCLGET(m, M_WAIT);
412 				if ((m->m_flags & M_EXT) == 0)
413 					goto nopages;
414 				mlen = MCLBYTES;
415 				len = min(min(mlen, resid), space);
416 			} else {
417 nopages:
418 				len = min(min(mlen, resid), space);
419 				/*
420 				 * For datagram protocols, leave room
421 				 * for protocol headers in first mbuf.
422 				 */
423 				if (atomic && top == 0 && len < mlen)
424 					MH_ALIGN(m, len);
425 			}
426 			space -= len;
427 			error = uiomove(mtod(m, caddr_t), (int)len, uio);
428 			resid = uio->uio_resid;
429 			m->m_len = len;
430 			*mp = m;
431 			top->m_pkthdr.len += len;
432 			if (error)
433 				goto release;
434 			mp = &m->m_next;
435 			if (resid <= 0) {
436 				if (flags & MSG_EOR)
437 					top->m_flags |= M_EOR;
438 				break;
439 			}
440 		    } while (space > 0 && atomic);
441 		    if (dontroute)
442 			    so->so_options |= SO_DONTROUTE;
443 		    s = splnet();				/* XXX */
444 		    error = (*so->so_proto->pr_usrreq)(so,
445 			(flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
446 			top, addr, control);
447 		    splx(s);
448 		    if (dontroute)
449 			    so->so_options &= ~SO_DONTROUTE;
450 		    clen = 0;
451 		    control = 0;
452 		    top = 0;
453 		    mp = &top;
454 		    if (error)
455 			goto release;
456 		} while (resid && space > 0);
457 	} while (resid);
458 
459 release:
460 	sbunlock(&so->so_snd);
461 out:
462 	if (top)
463 		m_freem(top);
464 	if (control)
465 		m_freem(control);
466 	return (error);
467 }
468 
469 /*
470  * Implement receive operations on a socket.
471  * We depend on the way that records are added to the sockbuf
472  * by sbappend*.  In particular, each record (mbufs linked through m_next)
473  * must begin with an address if the protocol so specifies,
474  * followed by an optional mbuf or mbufs containing ancillary data,
475  * and then zero or more mbufs of data.
476  * In order to avoid blocking network interrupts for the entire time here,
477  * we splx() while doing the actual copy to user space.
478  * Although the sockbuf is locked, new data may still be appended,
479  * and thus we must maintain consistency of the sockbuf during that time.
480  *
481  * The caller may receive the data as a single mbuf chain by supplying
482  * an mbuf **mp0 for use in returning the chain.  The uio is then used
483  * only for the count in uio_resid.
484  */
485 int
486 soreceive(so, paddr, uio, mp0, controlp, flagsp)
487 	register struct socket *so;
488 	struct mbuf **paddr;
489 	struct uio *uio;
490 	struct mbuf **mp0;
491 	struct mbuf **controlp;
492 	int *flagsp;
493 {
494 	register struct mbuf *m, **mp;
495 	register int flags, len, error, s, offset;
496 	struct protosw *pr = so->so_proto;
497 	struct mbuf *nextrecord;
498 	int moff, type = 0;
499 	int orig_resid = uio->uio_resid;
500 
501 	mp = mp0;
502 	if (paddr)
503 		*paddr = 0;
504 	if (controlp)
505 		*controlp = 0;
506 	if (flagsp)
507 		flags = *flagsp &~ MSG_EOR;
508 	else
509 		flags = 0;
510 	if (flags & MSG_OOB) {
511 		m = m_get(M_WAIT, MT_DATA);
512 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB,
513 		    m, (struct mbuf *)(flags & MSG_PEEK), (struct mbuf *)0);
514 		if (error)
515 			goto bad;
516 		do {
517 			error = uiomove(mtod(m, caddr_t),
518 			    (int) min(uio->uio_resid, m->m_len), uio);
519 			m = m_free(m);
520 		} while (uio->uio_resid && error == 0 && m);
521 bad:
522 		if (m)
523 			m_freem(m);
524 		return (error);
525 	}
526 	if (mp)
527 		*mp = (struct mbuf *)0;
528 	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
529 		(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
530 		    (struct mbuf *)0, (struct mbuf *)0);
531 
532 restart:
533 	if (error = sblock(&so->so_rcv, SBLOCKWAIT(flags)))
534 		return (error);
535 	s = splnet();
536 
537 	m = so->so_rcv.sb_mb;
538 	/*
539 	 * If we have less data than requested, block awaiting more
540 	 * (subject to any timeout) if:
541 	 *   1. the current count is less than the low water mark, or
542 	 *   2. MSG_WAITALL is set, and it is possible to do the entire
543 	 *	receive operation at once if we block (resid <= hiwat).
544 	 *   3. MSG_DONTWAIT is not set
545 	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
546 	 * we have to do the receive in sections, and thus risk returning
547 	 * a short count if a timeout or signal occurs after we start.
548 	 */
549 	if (m == 0 || ((flags & MSG_DONTWAIT) == 0 &&
550 	    so->so_rcv.sb_cc < uio->uio_resid) &&
551 	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
552 	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
553 	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0) {
554 #ifdef DIAGNOSTIC
555 		if (m == 0 && so->so_rcv.sb_cc)
556 			panic("receive 1");
557 #endif
558 		if (so->so_error) {
559 			if (m)
560 				goto dontblock;
561 			error = so->so_error;
562 			if ((flags & MSG_PEEK) == 0)
563 				so->so_error = 0;
564 			goto release;
565 		}
566 		if (so->so_state & SS_CANTRCVMORE) {
567 			if (m)
568 				goto dontblock;
569 			else
570 				goto release;
571 		}
572 		for (; m; m = m->m_next)
573 			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
574 				m = so->so_rcv.sb_mb;
575 				goto dontblock;
576 			}
577 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
578 		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
579 			error = ENOTCONN;
580 			goto release;
581 		}
582 		if (uio->uio_resid == 0)
583 			goto release;
584 		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
585 			error = EWOULDBLOCK;
586 			goto release;
587 		}
588 		sbunlock(&so->so_rcv);
589 		error = sbwait(&so->so_rcv);
590 		splx(s);
591 		if (error)
592 			return (error);
593 		goto restart;
594 	}
595 dontblock:
596 	if (uio->uio_procp)
597 		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
598 	nextrecord = m->m_nextpkt;
599 	if (pr->pr_flags & PR_ADDR) {
600 #ifdef DIAGNOSTIC
601 		if (m->m_type != MT_SONAME)
602 			panic("receive 1a");
603 #endif
604 		orig_resid = 0;
605 		if (flags & MSG_PEEK) {
606 			if (paddr)
607 				*paddr = m_copy(m, 0, m->m_len);
608 			m = m->m_next;
609 		} else {
610 			sbfree(&so->so_rcv, m);
611 			if (paddr) {
612 				*paddr = m;
613 				so->so_rcv.sb_mb = m->m_next;
614 				m->m_next = 0;
615 				m = so->so_rcv.sb_mb;
616 			} else {
617 				MFREE(m, so->so_rcv.sb_mb);
618 				m = so->so_rcv.sb_mb;
619 			}
620 		}
621 	}
622 	while (m && m->m_type == MT_CONTROL && error == 0) {
623 		if (flags & MSG_PEEK) {
624 			if (controlp)
625 				*controlp = m_copy(m, 0, m->m_len);
626 			m = m->m_next;
627 		} else {
628 			sbfree(&so->so_rcv, m);
629 			if (controlp) {
630 				if (pr->pr_domain->dom_externalize &&
631 				    mtod(m, struct cmsghdr *)->cmsg_type ==
632 				    SCM_RIGHTS)
633 				   error = (*pr->pr_domain->dom_externalize)(m);
634 				*controlp = m;
635 				so->so_rcv.sb_mb = m->m_next;
636 				m->m_next = 0;
637 				m = so->so_rcv.sb_mb;
638 			} else {
639 				MFREE(m, so->so_rcv.sb_mb);
640 				m = so->so_rcv.sb_mb;
641 			}
642 		}
643 		if (controlp) {
644 			orig_resid = 0;
645 			controlp = &(*controlp)->m_next;
646 		}
647 	}
648 	if (m) {
649 		if ((flags & MSG_PEEK) == 0)
650 			m->m_nextpkt = nextrecord;
651 		type = m->m_type;
652 		if (type == MT_OOBDATA)
653 			flags |= MSG_OOB;
654 	}
655 	moff = 0;
656 	offset = 0;
657 	while (m && uio->uio_resid > 0 && error == 0) {
658 		if (m->m_type == MT_OOBDATA) {
659 			if (type != MT_OOBDATA)
660 				break;
661 		} else if (type == MT_OOBDATA)
662 			break;
663 #ifdef DIAGNOSTIC
664 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
665 			panic("receive 3");
666 #endif
667 		so->so_state &= ~SS_RCVATMARK;
668 		len = uio->uio_resid;
669 		if (so->so_oobmark && len > so->so_oobmark - offset)
670 			len = so->so_oobmark - offset;
671 		if (len > m->m_len - moff)
672 			len = m->m_len - moff;
673 		/*
674 		 * If mp is set, just pass back the mbufs.
675 		 * Otherwise copy them out via the uio, then free.
676 		 * Sockbuf must be consistent here (points to current mbuf,
677 		 * it points to next record) when we drop priority;
678 		 * we must note any additions to the sockbuf when we
679 		 * block interrupts again.
680 		 */
681 		if (mp == 0) {
682 			splx(s);
683 			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
684 			s = splnet();
685 		} else
686 			uio->uio_resid -= len;
687 		if (len == m->m_len - moff) {
688 			if (m->m_flags & M_EOR)
689 				flags |= MSG_EOR;
690 			if (flags & MSG_PEEK) {
691 				m = m->m_next;
692 				moff = 0;
693 			} else {
694 				nextrecord = m->m_nextpkt;
695 				sbfree(&so->so_rcv, m);
696 				if (mp) {
697 					*mp = m;
698 					mp = &m->m_next;
699 					so->so_rcv.sb_mb = m = m->m_next;
700 					*mp = (struct mbuf *)0;
701 				} else {
702 					MFREE(m, so->so_rcv.sb_mb);
703 					m = so->so_rcv.sb_mb;
704 				}
705 				if (m)
706 					m->m_nextpkt = nextrecord;
707 			}
708 		} else {
709 			if (flags & MSG_PEEK)
710 				moff += len;
711 			else {
712 				if (mp)
713 					*mp = m_copym(m, 0, len, M_WAIT);
714 				m->m_data += len;
715 				m->m_len -= len;
716 				so->so_rcv.sb_cc -= len;
717 			}
718 		}
719 		if (so->so_oobmark) {
720 			if ((flags & MSG_PEEK) == 0) {
721 				so->so_oobmark -= len;
722 				if (so->so_oobmark == 0) {
723 					so->so_state |= SS_RCVATMARK;
724 					break;
725 				}
726 			} else {
727 				offset += len;
728 				if (offset == so->so_oobmark)
729 					break;
730 			}
731 		}
732 		if (flags & MSG_EOR)
733 			break;
734 		/*
735 		 * If the MSG_WAITALL flag is set (for non-atomic socket),
736 		 * we must not quit until "uio->uio_resid == 0" or an error
737 		 * termination.  If a signal/timeout occurs, return
738 		 * with a short count but without error.
739 		 * Keep sockbuf locked against other readers.
740 		 */
741 		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
742 		    !sosendallatonce(so) && !nextrecord) {
743 			if (so->so_error || so->so_state & SS_CANTRCVMORE)
744 				break;
745 			error = sbwait(&so->so_rcv);
746 			if (error) {
747 				sbunlock(&so->so_rcv);
748 				splx(s);
749 				return (0);
750 			}
751 			if (m = so->so_rcv.sb_mb)
752 				nextrecord = m->m_nextpkt;
753 		}
754 	}
755 
756 	if (m && pr->pr_flags & PR_ATOMIC) {
757 		flags |= MSG_TRUNC;
758 		if ((flags & MSG_PEEK) == 0)
759 			(void) sbdroprecord(&so->so_rcv);
760 	}
761 	if ((flags & MSG_PEEK) == 0) {
762 		if (m == 0)
763 			so->so_rcv.sb_mb = nextrecord;
764 		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
765 			(*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
766 			    (struct mbuf *)flags, (struct mbuf *)0,
767 			    (struct mbuf *)0);
768 	}
769 	if (orig_resid == uio->uio_resid && orig_resid &&
770 	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
771 		sbunlock(&so->so_rcv);
772 		splx(s);
773 		goto restart;
774 	}
775 
776 	if (flagsp)
777 		*flagsp |= flags;
778 release:
779 	sbunlock(&so->so_rcv);
780 	splx(s);
781 	return (error);
782 }
783 
784 int
785 soshutdown(so, how)
786 	register struct socket *so;
787 	register int how;
788 {
789 	register struct protosw *pr = so->so_proto;
790 
791 	how++;
792 	if (how & FREAD)
793 		sorflush(so);
794 	if (how & FWRITE)
795 		return ((*pr->pr_usrreq)(so, PRU_SHUTDOWN,
796 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
797 	return (0);
798 }
799 
800 void
801 sorflush(so)
802 	register struct socket *so;
803 {
804 	register struct sockbuf *sb = &so->so_rcv;
805 	register struct protosw *pr = so->so_proto;
806 	register int s;
807 	struct sockbuf asb;
808 
809 	sb->sb_flags |= SB_NOINTR;
810 	(void) sblock(sb, M_WAITOK);
811 	s = splimp();
812 	socantrcvmore(so);
813 	sbunlock(sb);
814 	asb = *sb;
815 	bzero((caddr_t)sb, sizeof (*sb));
816 	splx(s);
817 	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
818 		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
819 	sbrelease(&asb);
820 }
821 
822 int
823 sosetopt(so, level, optname, m0)
824 	register struct socket *so;
825 	int level, optname;
826 	struct mbuf *m0;
827 {
828 	int error = 0;
829 	register struct mbuf *m = m0;
830 
831 	if (level != SOL_SOCKET) {
832 		if (so->so_proto && so->so_proto->pr_ctloutput)
833 			return ((*so->so_proto->pr_ctloutput)
834 				  (PRCO_SETOPT, so, level, optname, &m0));
835 		error = ENOPROTOOPT;
836 	} else {
837 		switch (optname) {
838 
839 		case SO_LINGER:
840 			if (m == NULL || m->m_len != sizeof (struct linger)) {
841 				error = EINVAL;
842 				goto bad;
843 			}
844 			so->so_linger = mtod(m, struct linger *)->l_linger;
845 			/* fall thru... */
846 
847 		case SO_DEBUG:
848 		case SO_KEEPALIVE:
849 		case SO_DONTROUTE:
850 		case SO_USELOOPBACK:
851 		case SO_BROADCAST:
852 		case SO_REUSEADDR:
853 		case SO_REUSEPORT:
854 		case SO_OOBINLINE:
855 			if (m == NULL || m->m_len < sizeof (int)) {
856 				error = EINVAL;
857 				goto bad;
858 			}
859 			if (*mtod(m, int *))
860 				so->so_options |= optname;
861 			else
862 				so->so_options &= ~optname;
863 			break;
864 
865 		case SO_SNDBUF:
866 		case SO_RCVBUF:
867 		case SO_SNDLOWAT:
868 		case SO_RCVLOWAT:
869 			if (m == NULL || m->m_len < sizeof (int)) {
870 				error = EINVAL;
871 				goto bad;
872 			}
873 			switch (optname) {
874 
875 			case SO_SNDBUF:
876 			case SO_RCVBUF:
877 				if (sbreserve(optname == SO_SNDBUF ?
878 				    &so->so_snd : &so->so_rcv,
879 				    (u_long) *mtod(m, int *)) == 0) {
880 					error = ENOBUFS;
881 					goto bad;
882 				}
883 				break;
884 
885 			case SO_SNDLOWAT:
886 				so->so_snd.sb_lowat = *mtod(m, int *);
887 				break;
888 			case SO_RCVLOWAT:
889 				so->so_rcv.sb_lowat = *mtod(m, int *);
890 				break;
891 			}
892 			break;
893 
894 		case SO_SNDTIMEO:
895 		case SO_RCVTIMEO:
896 		    {
897 			struct timeval *tv;
898 			short val;
899 
900 			if (m == NULL || m->m_len < sizeof (*tv)) {
901 				error = EINVAL;
902 				goto bad;
903 			}
904 			tv = mtod(m, struct timeval *);
905 			if (tv->tv_sec > SHRT_MAX / hz - hz) {
906 				error = EDOM;
907 				goto bad;
908 			}
909 			val = tv->tv_sec * hz + tv->tv_usec / tick;
910 
911 			switch (optname) {
912 
913 			case SO_SNDTIMEO:
914 				so->so_snd.sb_timeo = val;
915 				break;
916 			case SO_RCVTIMEO:
917 				so->so_rcv.sb_timeo = val;
918 				break;
919 			}
920 			break;
921 		    }
922 
923 		default:
924 			error = ENOPROTOOPT;
925 			break;
926 		}
927 		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
928 			(void) ((*so->so_proto->pr_ctloutput)
929 				  (PRCO_SETOPT, so, level, optname, &m0));
930 			m = NULL;	/* freed by protocol */
931 		}
932 	}
933 bad:
934 	if (m)
935 		(void) m_free(m);
936 	return (error);
937 }
938 
939 int
940 sogetopt(so, level, optname, mp)
941 	register struct socket *so;
942 	int level, optname;
943 	struct mbuf **mp;
944 {
945 	register struct mbuf *m;
946 
947 	if (level != SOL_SOCKET) {
948 		if (so->so_proto && so->so_proto->pr_ctloutput) {
949 			return ((*so->so_proto->pr_ctloutput)
950 				  (PRCO_GETOPT, so, level, optname, mp));
951 		} else
952 			return (ENOPROTOOPT);
953 	} else {
954 		m = m_get(M_WAIT, MT_SOOPTS);
955 		m->m_len = sizeof (int);
956 
957 		switch (optname) {
958 
959 		case SO_LINGER:
960 			m->m_len = sizeof (struct linger);
961 			mtod(m, struct linger *)->l_onoff =
962 				so->so_options & SO_LINGER;
963 			mtod(m, struct linger *)->l_linger = so->so_linger;
964 			break;
965 
966 		case SO_USELOOPBACK:
967 		case SO_DONTROUTE:
968 		case SO_DEBUG:
969 		case SO_KEEPALIVE:
970 		case SO_REUSEADDR:
971 		case SO_REUSEPORT:
972 		case SO_BROADCAST:
973 		case SO_OOBINLINE:
974 			*mtod(m, int *) = so->so_options & optname;
975 			break;
976 
977 		case SO_TYPE:
978 			*mtod(m, int *) = so->so_type;
979 			break;
980 
981 		case SO_ERROR:
982 			*mtod(m, int *) = so->so_error;
983 			so->so_error = 0;
984 			break;
985 
986 		case SO_SNDBUF:
987 			*mtod(m, int *) = so->so_snd.sb_hiwat;
988 			break;
989 
990 		case SO_RCVBUF:
991 			*mtod(m, int *) = so->so_rcv.sb_hiwat;
992 			break;
993 
994 		case SO_SNDLOWAT:
995 			*mtod(m, int *) = so->so_snd.sb_lowat;
996 			break;
997 
998 		case SO_RCVLOWAT:
999 			*mtod(m, int *) = so->so_rcv.sb_lowat;
1000 			break;
1001 
1002 		case SO_SNDTIMEO:
1003 		case SO_RCVTIMEO:
1004 		    {
1005 			int val = (optname == SO_SNDTIMEO ?
1006 			     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1007 
1008 			m->m_len = sizeof(struct timeval);
1009 			mtod(m, struct timeval *)->tv_sec = val / hz;
1010 			mtod(m, struct timeval *)->tv_usec =
1011 			    (val % hz) / tick;
1012 			break;
1013 		    }
1014 
1015 		default:
1016 			(void)m_free(m);
1017 			return (ENOPROTOOPT);
1018 		}
1019 		*mp = m;
1020 		return (0);
1021 	}
1022 }
1023 
1024 void
1025 sohasoutofband(so)
1026 	register struct socket *so;
1027 {
1028 	struct proc *p;
1029 
1030 	if (so->so_pgid < 0)
1031 		gsignal(-so->so_pgid, SIGURG);
1032 	else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1033 		psignal(p, SIGURG);
1034 	selwakeup(&so->so_rcv.sb_sel);
1035 }
1036