xref: /csrg-svn/sys/kern/uipc_socket.c (revision 16427)
1 /*	uipc_socket.c	6.4	84/04/30	*/
2 
3 #include "../h/param.h"
4 #include "../h/systm.h"
5 #include "../h/dir.h"
6 #include "../h/user.h"
7 #include "../h/proc.h"
8 #include "../h/file.h"
9 #include "../h/inode.h"
10 #include "../h/buf.h"
11 #include "../h/mbuf.h"
12 #include "../h/un.h"
13 #include "../h/protosw.h"
14 #include "../h/socket.h"
15 #include "../h/socketvar.h"
16 #include "../h/stat.h"
17 #include "../h/ioctl.h"
18 #include "../h/uio.h"
19 #include "../net/route.h"
20 #include "../netinet/in.h"
21 #include "../net/if.h"
22 
23 /*
24  * Socket operation routines.
25  * These routines are called by the routines in
26  * sys_socket.c or from a system process, and
27  * implement the semantics of socket operations by
28  * switching out to the protocol specific routines.
29  *
30  * TODO:
31  *	sostat
32  *	test socketpair
33  *	PR_RIGHTS
34  *	clean up select, async
35  *	out-of-band is a kludge
36  */
37 /*ARGSUSED*/
38 socreate(dom, aso, type, proto)
39 	struct socket **aso;
40 	register int type;
41 	int proto;
42 {
43 	register struct protosw *prp;
44 	register struct socket *so;
45 	register struct mbuf *m;
46 	register int error;
47 
48 	if (proto)
49 		prp = pffindproto(dom, proto);
50 	else
51 		prp = pffindtype(dom, type);
52 	if (prp == 0)
53 		return (EPROTONOSUPPORT);
54 	if (prp->pr_type != type)
55 		return (EPROTOTYPE);
56 	m = m_getclr(M_WAIT, MT_SOCKET);
57 	if (m == 0)
58 		return (ENOBUFS);
59 	so = mtod(m, struct socket *);
60 	so->so_options = 0;
61 	so->so_state = 0;
62 	so->so_type = type;
63 	if (u.u_uid == 0)
64 		so->so_state = SS_PRIV;
65 	so->so_proto = prp;
66 	error =
67 	    (*prp->pr_usrreq)(so, PRU_ATTACH,
68 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
69 	if (error) {
70 		so->so_state |= SS_NOFDREF;
71 		sofree(so);
72 		return (error);
73 	}
74 	*aso = so;
75 	return (0);
76 }
77 
78 sobind(so, nam)
79 	struct socket *so;
80 	struct mbuf *nam;
81 {
82 	int s = splnet();
83 	int error;
84 
85 	error =
86 	    (*so->so_proto->pr_usrreq)(so, PRU_BIND,
87 		(struct mbuf *)0, nam, (struct mbuf *)0);
88 	splx(s);
89 	return (error);
90 }
91 
92 solisten(so, backlog)
93 	register struct socket *so;
94 	int backlog;
95 {
96 	int s = splnet(), error;
97 
98 	error =
99 	    (*so->so_proto->pr_usrreq)(so, PRU_LISTEN,
100 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
101 	if (error) {
102 		splx(s);
103 		return (error);
104 	}
105 	if (so->so_q == 0) {
106 		so->so_q = so;
107 		so->so_q0 = so;
108 		so->so_options |= SO_ACCEPTCONN;
109 	}
110 	if (backlog < 0)
111 		backlog = 0;
112 	so->so_qlimit = MIN(backlog, SOMAXCONN);
113 	splx(s);
114 	return (0);
115 }
116 
117 sofree(so)
118 	register struct socket *so;
119 {
120 
121 	if (so->so_head) {
122 		if (!soqremque(so, 0) && !soqremque(so, 1))
123 			panic("sofree dq");
124 		so->so_head = 0;
125 	}
126 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
127 		return;
128 	sbrelease(&so->so_snd);
129 	sorflush(so);
130 	(void) m_free(dtom(so));
131 }
132 
133 /*
134  * Close a socket on last file table reference removal.
135  * Initiate disconnect if connected.
136  * Free socket when disconnect complete.
137  */
138 soclose(so)
139 	register struct socket *so;
140 {
141 	int s = splnet();		/* conservative */
142 	int error;
143 
144 	if (so->so_options & SO_ACCEPTCONN) {
145 		while (so->so_q0 != so)
146 			(void) soabort(so->so_q0);
147 		while (so->so_q != so)
148 			(void) soabort(so->so_q);
149 	}
150 	if (so->so_pcb == 0)
151 		goto discard;
152 	if (so->so_state & SS_ISCONNECTED) {
153 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
154 			error = sodisconnect(so, (struct mbuf *)0);
155 			if (error)
156 				goto drop;
157 		}
158 		if (so->so_options & SO_LINGER) {
159 			if ((so->so_state & SS_ISDISCONNECTING) &&
160 			    (so->so_state & SS_NBIO))
161 				goto drop;
162 			while (so->so_state & SS_ISCONNECTED)
163 				sleep((caddr_t)&so->so_timeo, PZERO+1);
164 		}
165 	}
166 drop:
167 	if (so->so_pcb) {
168 		int error2 =
169 		    (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
170 			(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
171 		if (error == 0)
172 			error = error2;
173 	}
174 discard:
175 	if (so->so_state & SS_NOFDREF)
176 		panic("soclose: NOFDREF");
177 	so->so_state |= SS_NOFDREF;
178 	sofree(so);
179 	splx(s);
180 	return (error);
181 }
182 
183 /*
184  * Must be called at splnet...
185  */
186 soabort(so)
187 	struct socket *so;
188 {
189 
190 	return (
191 	    (*so->so_proto->pr_usrreq)(so, PRU_ABORT,
192 		(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
193 }
194 
195 soaccept(so, nam)
196 	register struct socket *so;
197 	struct mbuf *nam;
198 {
199 	int s = splnet();
200 	int error;
201 
202 	if ((so->so_state & SS_NOFDREF) == 0)
203 		panic("soaccept: !NOFDREF");
204 	so->so_state &= ~SS_NOFDREF;
205 	error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
206 	    (struct mbuf *)0, nam, (struct mbuf *)0);
207 	splx(s);
208 	return (error);
209 }
210 
211 soconnect(so, nam)
212 	register struct socket *so;
213 	struct mbuf *nam;
214 {
215 	int s = splnet();
216 	int error;
217 
218 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) {
219 		error = EISCONN;
220 		goto bad;
221 	}
222 	error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
223 	    (struct mbuf *)0, nam, (struct mbuf *)0);
224 bad:
225 	splx(s);
226 	return (error);
227 }
228 
229 soconnect2(so1, so2)
230 	register struct socket *so1;
231 	struct socket *so2;
232 {
233 	int s = splnet();
234 	int error;
235 
236 	error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
237 	    (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0);
238 	splx(s);
239 	return (error);
240 }
241 
242 sodisconnect(so, nam)
243 	register struct socket *so;
244 	struct mbuf *nam;
245 {
246 	int s = splnet();
247 	int error;
248 
249 	if ((so->so_state & SS_ISCONNECTED) == 0) {
250 		error = ENOTCONN;
251 		goto bad;
252 	}
253 	if (so->so_state & SS_ISDISCONNECTING) {
254 		error = EALREADY;
255 		goto bad;
256 	}
257 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
258 	    (struct mbuf *)0, nam, (struct mbuf *)0);
259 bad:
260 	splx(s);
261 	return (error);
262 }
263 
264 /*
265  * Send on a socket.
266  * If send must go all at once and message is larger than
267  * send buffering, then hard error.
268  * Lock against other senders.
269  * If must go all at once and not enough room now, then
270  * inform user that this would block and do nothing.
271  * Otherwise, if nonblocking, send as much as possible.
272  */
273 sosend(so, nam, uio, flags, rights)
274 	register struct socket *so;
275 	struct mbuf *nam;
276 	register struct uio *uio;
277 	int flags;
278 	struct mbuf *rights;
279 {
280 	struct mbuf *top = 0;
281 	register struct mbuf *m, **mp;
282 	register int space;
283 	int len, error = 0, s, dontroute, first = 1;
284 
285 	if (sosendallatonce(so) && uio->uio_resid > so->so_snd.sb_hiwat)
286 		return (EMSGSIZE);
287 	dontroute =
288 	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
289 	    (so->so_proto->pr_flags & PR_ATOMIC);
290 	u.u_ru.ru_msgsnd++;
291 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
292 
293 restart:
294 	sblock(&so->so_snd);
295 	do {
296 		s = splnet();
297 		if (so->so_state & SS_CANTSENDMORE) {
298 			psignal(u.u_procp, SIGPIPE);
299 			snderr(EPIPE);
300 		}
301 		if (so->so_error) {
302 			error = so->so_error;
303 			so->so_error = 0;			/* ??? */
304 			splx(s);
305 			goto release;
306 		}
307 		if ((so->so_state & SS_ISCONNECTED) == 0) {
308 			if (so->so_proto->pr_flags & PR_CONNREQUIRED)
309 				snderr(ENOTCONN);
310 			if (nam == 0)
311 				snderr(EDESTADDRREQ);
312 		}
313 		if (flags & MSG_OOB)
314 			space = 1024;
315 		else {
316 			space = sbspace(&so->so_snd);
317 			if (space <= 0 ||
318 			    sosendallatonce(so) && space < uio->uio_resid) {
319 				if (so->so_state & SS_NBIO) {
320 					if (first)
321 						error = EWOULDBLOCK;
322 					splx(s);
323 					goto release;
324 				}
325 				sbunlock(&so->so_snd);
326 				sbwait(&so->so_snd);
327 				splx(s);
328 				goto restart;
329 			}
330 		}
331 		splx(s);
332 		mp = &top;
333 		while (uio->uio_resid > 0 && space > 0) {
334 			register struct iovec *iov = uio->uio_iov;
335 
336 			if (iov->iov_len == 0) {
337 				uio->uio_iov++;
338 				uio->uio_iovcnt--;
339 				if (uio->uio_iovcnt < 0)
340 					panic("sosend");
341 				continue;
342 			}
343 			MGET(m, M_WAIT, MT_DATA);
344 			if (m == NULL) {
345 				error = ENOBUFS;		/* SIGPIPE? */
346 				goto release;
347 			}
348 			if (iov->iov_len >= CLBYTES && space >= CLBYTES) {
349 				register struct mbuf *p;
350 				MCLGET(p, 1);
351 				if (p == 0)
352 					goto nopages;
353 				m->m_off = (int)p - (int)m;
354 				len = CLBYTES;
355 			} else {
356 nopages:
357 				len = MIN(MLEN, iov->iov_len);
358 			}
359 			error = uiomove(mtod(m, caddr_t), len, UIO_WRITE, uio);
360 			m->m_len = len;
361 			*mp = m;
362 			if (error)
363 				goto release;
364 			mp = &m->m_next;
365 			space -= len;
366 		}
367 		if (dontroute)
368 			so->so_options |= SO_DONTROUTE;
369 		s = splnet();
370 		error = (*so->so_proto->pr_usrreq)(so,
371 		    (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
372 		    top, (caddr_t)nam, rights);
373 		splx(s);
374 		if (dontroute)
375 			so->so_options &= ~SO_DONTROUTE;
376 		top = 0;
377 		first = 0;
378 		if (error)
379 			break;
380 	} while (uio->uio_resid);
381 
382 release:
383 	sbunlock(&so->so_snd);
384 	if (top)
385 		m_freem(top);
386 	return (error);
387 }
388 
389 soreceive(so, aname, uio, flags, rightsp)
390 	register struct socket *so;
391 	struct mbuf **aname;
392 	register struct uio *uio;
393 	int flags;
394 	struct mbuf **rightsp;
395 {
396 	register struct mbuf *m, *n;
397 	register int len, error = 0, s, eor, tomark;
398 	struct protosw *pr = so->so_proto;
399 	int moff;
400 
401 	if (rightsp)
402 		*rightsp = 0;
403 	if (aname)
404 		*aname = 0;
405 	if (flags & MSG_OOB) {
406 		m = m_get(M_WAIT, MT_DATA);
407 		if (m == 0)
408 			return (ENOBUFS);
409 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB,
410 		    m, (struct mbuf *)0, (struct mbuf *)0);
411 		if (error)
412 			goto bad;
413 		do {
414 			len = uio->uio_resid;
415 			if (len > m->m_len)
416 				len = m->m_len;
417 			error =
418 			    uiomove(mtod(m, caddr_t), (int)len, UIO_READ, uio);
419 			m = m_free(m);
420 		} while (uio->uio_resid && error == 0 && m);
421 bad:
422 		if (m)
423 			m_freem(m);
424 		return (error);
425 	}
426 
427 restart:
428 	sblock(&so->so_rcv);
429 	s = splnet();
430 
431 #define	rcverr(errno)	{ error = errno; splx(s); goto release; }
432 	if (so->so_rcv.sb_cc == 0) {
433 		if (so->so_error) {
434 			error = so->so_error;
435 			so->so_error = 0;
436 			splx(s);
437 			goto release;
438 		}
439 		if (so->so_state & SS_CANTRCVMORE) {
440 			splx(s);
441 			goto release;
442 		}
443 		if ((so->so_state & SS_ISCONNECTED) == 0 &&
444 		    (so->so_proto->pr_flags & PR_CONNREQUIRED))
445 			rcverr(ENOTCONN);
446 		if (so->so_state & SS_NBIO)
447 			rcverr(EWOULDBLOCK);
448 		sbunlock(&so->so_rcv);
449 		sbwait(&so->so_rcv);
450 		splx(s);
451 		goto restart;
452 	}
453 	u.u_ru.ru_msgrcv++;
454 	m = so->so_rcv.sb_mb;
455 	if (m == 0)
456 		panic("receive");
457 	if (pr->pr_flags & PR_ADDR) {
458 		if ((flags & MSG_PEEK) == 0) {
459 			so->so_rcv.sb_cc -= m->m_len;
460 			so->so_rcv.sb_mbcnt -= MSIZE;
461 		}
462 		if (aname) {
463 			if (flags & MSG_PEEK) {
464 				*aname = m_copy(m, 0, m->m_len);
465 				if (*aname == NULL)
466 					panic("receive 2");
467 			} else
468 				*aname = m;
469 			m = m->m_next;
470 			(*aname)->m_next = 0;
471 		} else
472 			if (flags & MSG_PEEK)
473 				m = m->m_next;
474 			else
475 				m = m_free(m);
476 		if (m == 0)
477 			panic("receive 2a");
478 		if (rightsp) {
479 			if (m->m_len)
480 				*rightsp = m_copy(m, 0, m->m_len);
481 			else {
482 				*rightsp = m_get(M_DONTWAIT, MT_SONAME);
483 				if (*rightsp)
484 					(*rightsp)->m_len = 0;
485 			}
486 #ifdef notdef
487 			if (*rightsp == NULL)
488 				panic("receive 2b");
489 #endif
490 		}
491 		if (flags & MSG_PEEK)
492 			m = m->m_next;
493 		else {
494 			so->so_rcv.sb_cc -= m->m_len;
495 			so->so_rcv.sb_mbcnt -= MSIZE;
496 			m = m_free(m);
497 		}
498 		if (m == 0)
499 			panic("receive 3");
500 		if ((flags & MSG_PEEK) == 0)
501 			so->so_rcv.sb_mb = m;
502 	}
503 	eor = 0;
504 	moff = 0;
505 	tomark = so->so_oobmark;
506 	do {
507 		if (uio->uio_resid <= 0)
508 			break;
509 		len = uio->uio_resid;
510 		so->so_state &= ~SS_RCVATMARK;
511 		if (tomark && len > tomark)
512 			len = tomark;
513 		if (moff+len > m->m_len - moff)
514 			len = m->m_len - moff;
515 		splx(s);
516 		error =
517 		    uiomove(mtod(m, caddr_t) + moff, (int)len, UIO_READ, uio);
518 		s = splnet();
519 		if (len == m->m_len) {
520 			eor = (int)m->m_act;
521 			if (flags & MSG_PEEK)
522 				m = m->m_next;
523 			else {
524 				sbfree(&so->so_rcv, m);
525 				MFREE(m, n);
526 				m = n;
527 				so->so_rcv.sb_mb = m;
528 			}
529 			moff = 0;
530 		} else {
531 			if (flags & MSG_PEEK)
532 				moff += len;
533 			else {
534 				m->m_off += len;
535 				m->m_len -= len;
536 				so->so_rcv.sb_cc -= len;
537 			}
538 		}
539 		if ((flags & MSG_PEEK) == 0 && so->so_oobmark) {
540 			so->so_oobmark -= len;
541 			if (so->so_oobmark == 0) {
542 				so->so_state |= SS_RCVATMARK;
543 				break;
544 			}
545 		}
546 		if (tomark) {
547 			tomark -= len;
548 			if (tomark == 0)
549 				break;
550 		}
551 	} while (m && error == 0 && !eor);
552 	if (flags & MSG_PEEK)
553 		goto release;
554 	if ((so->so_proto->pr_flags & PR_ATOMIC) && eor == 0)
555 		do {
556 			if (m == 0)
557 				panic("receive 4");
558 			sbfree(&so->so_rcv, m);
559 			eor = (int)m->m_act;
560 			so->so_rcv.sb_mb = m->m_next;
561 			MFREE(m, n);
562 			m = n;
563 		} while (eor == 0);
564 	if ((so->so_proto->pr_flags & PR_WANTRCVD) && so->so_pcb)
565 		(*so->so_proto->pr_usrreq)(so, PRU_RCVD,
566 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
567 release:
568 	sbunlock(&so->so_rcv);
569 	if (error == 0 && rightsp &&
570 	    *rightsp && so->so_proto->pr_family == AF_UNIX)
571 		error = unp_externalize(*rightsp);
572 	splx(s);
573 	return (error);
574 }
575 
576 soshutdown(so, how)
577 	register struct socket *so;
578 	register int how;
579 {
580 	register struct protosw *pr = so->so_proto;
581 
582 	how++;
583 	if (how & FREAD)
584 		sorflush(so);
585 	if (how & FWRITE)
586 		return ((*pr->pr_usrreq)(so, PRU_SHUTDOWN,
587 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
588 	return (0);
589 }
590 
591 sorflush(so)
592 	register struct socket *so;
593 {
594 	register struct sockbuf *sb = &so->so_rcv;
595 	register struct protosw *pr = so->so_proto;
596 	register int s;
597 	struct sockbuf asb;
598 
599 	sblock(sb);
600 	s = splimp();
601 	socantrcvmore(so);
602 	sbunlock(sb);
603 	asb = *sb;
604 	bzero((caddr_t)sb, sizeof (*sb));
605 	splx(s);
606 	if (pr->pr_family == AF_UNIX && (pr->pr_flags & PR_RIGHTS))
607 		unp_scan(asb.sb_mb, unp_discard);
608 	sbrelease(&asb);
609 }
610 
611 sosetopt(so, level, optname, m)
612 	register struct socket *so;
613 	int level, optname;
614 	register struct mbuf *m;
615 {
616 
617 	if (level != SOL_SOCKET)
618 		return (EINVAL);	/* XXX */
619 	switch (optname) {
620 
621 	case SO_DEBUG:
622 	case SO_KEEPALIVE:
623 	case SO_DONTROUTE:
624 	case SO_USELOOPBACK:
625 	case SO_REUSEADDR:
626 		so->so_options |= optname;
627 		break;
628 
629 	case SO_LINGER:
630 		if (m == NULL || m->m_len != sizeof (int))
631 			return (EINVAL);
632 		so->so_options |= SO_LINGER;
633 		so->so_linger = *mtod(m, int *);
634 		break;
635 
636 	case SO_DONTLINGER:
637 		so->so_options &= ~SO_LINGER;
638 		so->so_linger = 0;
639 		break;
640 
641 	default:
642 		return (EINVAL);
643 	}
644 	return (0);
645 }
646 
647 sogetopt(so, level, optname, m)
648 	register struct socket *so;
649 	int level, optname;
650 	register struct mbuf *m;
651 {
652 
653 	if (level != SOL_SOCKET)
654 		return (EINVAL);	/* XXX */
655 	switch (optname) {
656 
657 	case SO_USELOOPBACK:
658 	case SO_DONTROUTE:
659 	case SO_DEBUG:
660 	case SO_KEEPALIVE:
661 	case SO_LINGER:
662 	case SO_REUSEADDR:
663 		if ((so->so_options & optname) == 0)
664 			return (ENOPROTOOPT);
665 		if (optname == SO_LINGER && m != NULL) {
666 			*mtod(m, int *) = so->so_linger;
667 			m->m_len = sizeof (so->so_linger);
668 		}
669 		break;
670 
671 	default:
672 		return (EINVAL);
673 	}
674 	return (0);
675 }
676 
677 sohasoutofband(so)
678 	register struct socket *so;
679 {
680 
681 	if (so->so_pgrp == 0)
682 		return;
683 	if (so->so_pgrp > 0)
684 		gsignal(so->so_pgrp, SIGURG);
685 	else {
686 		struct proc *p = pfind(-so->so_pgrp);
687 
688 		if (p)
689 			psignal(p, SIGURG);
690 	}
691 }
692