xref: /csrg-svn/sys/kern/uipc_socket.c (revision 15123)
1 /*	uipc_socket.c	6.2	83/09/29	*/
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  */
272 sosend(so, nam, uio, flags, rights)
273 	register struct socket *so;
274 	struct mbuf *nam;
275 	register struct uio *uio;
276 	int flags;
277 	struct mbuf *rights;
278 {
279 	struct mbuf *top = 0;
280 	register struct mbuf *m, **mp = &top;
281 	register int space;
282 	int len, error = 0, s, dontroute;
283 	struct sockbuf sendtempbuf;
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 restart:
291 	sblock(&so->so_snd);
292 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
293 
294 	u.u_ru.ru_msgsnd++;
295 again:
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 (top) {
314 		if (dontroute)
315 			so->so_options |= SO_DONTROUTE;
316 		error = (*so->so_proto->pr_usrreq)(so,
317 		    (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
318 		    top, (caddr_t)nam, rights);
319 		if (dontroute)
320 			so->so_options &= ~SO_DONTROUTE;
321 		top = 0;
322 		if (error) {
323 			splx(s);
324 			goto release;
325 		}
326 		mp = &top;
327 	}
328 	if (uio->uio_resid == 0) {
329 		splx(s);
330 		goto release;
331 	}
332 	if (flags & MSG_OOB)
333 		space = 1024;
334 	else {
335 		space = sbspace(&so->so_snd);
336 		if (space <= 0 ||
337 		    sosendallatonce(so) && space < uio->uio_resid) {
338 			if (so->so_state & SS_NBIO)
339 				snderr(EWOULDBLOCK);
340 			sbunlock(&so->so_snd);
341 			sbwait(&so->so_snd);
342 			splx(s);
343 			goto restart;
344 		}
345 	}
346 	splx(s);
347 	/*
348 	 * Temporary kludge-- don't want to update so_snd in this loop
349 	 * (will be done when sent), but need to recalculate
350 	 * space on each iteration.  For now, copy so_snd into a tmp.
351 	 */
352 	sendtempbuf = so->so_snd;
353 	while (uio->uio_resid > 0 && space > 0) {
354 		register struct iovec *iov = uio->uio_iov;
355 
356 		if (iov->iov_len == 0) {
357 			uio->uio_iov++;
358 			uio->uio_iovcnt--;
359 			if (uio->uio_iovcnt < 0)
360 				panic("sosend");
361 			continue;
362 		}
363 		MGET(m, M_WAIT, MT_DATA);
364 		if (m == NULL) {
365 			error = ENOBUFS;			/* SIGPIPE? */
366 			goto release;
367 		}
368 		if (iov->iov_len >= CLBYTES && space >= CLBYTES) {
369 			register struct mbuf *p;
370 			MCLGET(p, 1);
371 			if (p == 0)
372 				goto nopages;
373 			m->m_off = (int)p - (int)m;
374 			len = CLBYTES;
375 		} else {
376 nopages:
377 			len = MIN(MLEN, iov->iov_len);
378 		}
379 		error = uiomove(mtod(m, caddr_t), len, UIO_WRITE, uio);
380 		m->m_len = len;
381 		*mp = m;
382 		if (error)
383 			goto release;
384 		mp = &m->m_next;
385 		if (flags & MSG_OOB)
386 			space -= len;
387 		else {
388 			sballoc(&sendtempbuf, m);
389 			space = sbspace(&sendtempbuf);
390 		}
391 	}
392 	goto again;
393 
394 release:
395 	sbunlock(&so->so_snd);
396 	if (top)
397 		m_freem(top);
398 	return (error);
399 }
400 
401 soreceive(so, aname, uio, flags, rightsp)
402 	register struct socket *so;
403 	struct mbuf **aname;
404 	register struct uio *uio;
405 	int flags;
406 	struct mbuf **rightsp;
407 {
408 	register struct mbuf *m, *n;
409 	register int len, error = 0, s, eor, tomark;
410 	struct protosw *pr = so->so_proto;
411 	int moff;
412 
413 	if (rightsp)
414 		*rightsp = 0;
415 	if (aname)
416 		*aname = 0;
417 	if (flags & MSG_OOB) {
418 		m = m_get(M_WAIT, MT_DATA);
419 		if (m == 0)
420 			return (ENOBUFS);
421 		error = (*pr->pr_usrreq)(so, PRU_RCVOOB,
422 		    m, (struct mbuf *)0, (struct mbuf *)0);
423 		if (error)
424 			goto bad;
425 		do {
426 			len = uio->uio_resid;
427 			if (len > m->m_len)
428 				len = m->m_len;
429 			error =
430 			    uiomove(mtod(m, caddr_t), (int)len, UIO_READ, uio);
431 			m = m_free(m);
432 		} while (uio->uio_resid && error == 0 && m);
433 bad:
434 		if (m)
435 			m_freem(m);
436 		return (error);
437 	}
438 
439 restart:
440 	sblock(&so->so_rcv);
441 	s = splnet();
442 
443 #define	rcverr(errno)	{ error = errno; splx(s); goto release; }
444 	if (so->so_rcv.sb_cc == 0) {
445 		if (so->so_error) {
446 			error = so->so_error;
447 			so->so_error = 0;
448 			splx(s);
449 			goto release;
450 		}
451 		if (so->so_state & SS_CANTRCVMORE) {
452 			splx(s);
453 			goto release;
454 		}
455 		if ((so->so_state & SS_ISCONNECTED) == 0 &&
456 		    (so->so_proto->pr_flags & PR_CONNREQUIRED))
457 			rcverr(ENOTCONN);
458 		if (so->so_state & SS_NBIO)
459 			rcverr(EWOULDBLOCK);
460 		sbunlock(&so->so_rcv);
461 		sbwait(&so->so_rcv);
462 		splx(s);
463 		goto restart;
464 	}
465 	u.u_ru.ru_msgrcv++;
466 	m = so->so_rcv.sb_mb;
467 	if (m == 0)
468 		panic("receive");
469 	if (pr->pr_flags & PR_ADDR) {
470 		if ((flags & MSG_PEEK) == 0) {
471 			so->so_rcv.sb_cc -= m->m_len;
472 			so->so_rcv.sb_mbcnt -= MSIZE;
473 		}
474 		if (aname) {
475 			if (flags & MSG_PEEK) {
476 				*aname = m_copy(m, 0, m->m_len);
477 				if (*aname == NULL)
478 					panic("receive 2");
479 			} else
480 				*aname = m;
481 			m = m->m_next;
482 			(*aname)->m_next = 0;
483 		} else
484 			if (flags & MSG_PEEK)
485 				m = m->m_next;
486 			else
487 				m = m_free(m);
488 		if (m == 0)
489 			panic("receive 2a");
490 		if (rightsp) {
491 			if (m->m_len)
492 				*rightsp = m_copy(m, 0, m->m_len);
493 			else {
494 				*rightsp = m_get(M_DONTWAIT, MT_SONAME);
495 				if (*rightsp)
496 					(*rightsp)->m_len = 0;
497 			}
498 #ifdef notdef
499 			if (*rightsp == NULL)
500 				panic("receive 2b");
501 #endif
502 		}
503 		if (flags & MSG_PEEK)
504 			m = m->m_next;
505 		else {
506 			so->so_rcv.sb_cc -= m->m_len;
507 			so->so_rcv.sb_mbcnt -= MSIZE;
508 			m = m_free(m);
509 		}
510 		if (m == 0)
511 			panic("receive 3");
512 		if ((flags & MSG_PEEK) == 0)
513 			so->so_rcv.sb_mb = m;
514 	}
515 	eor = 0;
516 	moff = 0;
517 	tomark = so->so_oobmark;
518 	do {
519 		if (uio->uio_resid <= 0)
520 			break;
521 		len = uio->uio_resid;
522 		so->so_state &= ~SS_RCVATMARK;
523 		if (tomark && len > tomark)
524 			len = tomark;
525 		if (moff+len > m->m_len - moff)
526 			len = m->m_len - moff;
527 		splx(s);
528 		error =
529 		    uiomove(mtod(m, caddr_t) + moff, (int)len, UIO_READ, uio);
530 		s = splnet();
531 		if (len == m->m_len) {
532 			eor = (int)m->m_act;
533 			if (flags & MSG_PEEK)
534 				m = m->m_next;
535 			else {
536 				sbfree(&so->so_rcv, m);
537 				MFREE(m, n);
538 				m = n;
539 				so->so_rcv.sb_mb = m;
540 			}
541 			moff = 0;
542 		} else {
543 			if (flags & MSG_PEEK)
544 				moff += len;
545 			else {
546 				m->m_off += len;
547 				m->m_len -= len;
548 				so->so_rcv.sb_cc -= len;
549 			}
550 		}
551 		if ((flags & MSG_PEEK) == 0 && so->so_oobmark) {
552 			so->so_oobmark -= len;
553 			if (so->so_oobmark == 0) {
554 				so->so_state |= SS_RCVATMARK;
555 				break;
556 			}
557 		}
558 		if (tomark) {
559 			tomark -= len;
560 			if (tomark == 0)
561 				break;
562 		}
563 	} while (m && error == 0 && !eor);
564 	if (flags & MSG_PEEK)
565 		goto release;
566 	if ((so->so_proto->pr_flags & PR_ATOMIC) && eor == 0)
567 		do {
568 			if (m == 0)
569 				panic("receive 4");
570 			sbfree(&so->so_rcv, m);
571 			eor = (int)m->m_act;
572 			so->so_rcv.sb_mb = m->m_next;
573 			MFREE(m, n);
574 			m = n;
575 		} while (eor == 0);
576 	if ((so->so_proto->pr_flags & PR_WANTRCVD) && so->so_pcb)
577 		(*so->so_proto->pr_usrreq)(so, PRU_RCVD,
578 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0);
579 release:
580 	sbunlock(&so->so_rcv);
581 	if (error == 0 && rightsp &&
582 	    *rightsp && so->so_proto->pr_family == AF_UNIX)
583 		error = unp_externalize(*rightsp);
584 	splx(s);
585 	return (error);
586 }
587 
588 soshutdown(so, how)
589 	register struct socket *so;
590 	register int how;
591 {
592 	register struct protosw *pr = so->so_proto;
593 
594 	how++;
595 	if (how & FREAD)
596 		sorflush(so);
597 	if (how & FWRITE)
598 		return ((*pr->pr_usrreq)(so, PRU_SHUTDOWN,
599 		    (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0));
600 	return (0);
601 }
602 
603 sorflush(so)
604 	register struct socket *so;
605 {
606 	register struct sockbuf *sb = &so->so_rcv;
607 	register struct protosw *pr = so->so_proto;
608 	register int s;
609 	struct sockbuf asb;
610 
611 	sblock(sb);
612 	s = splimp();
613 	socantrcvmore(so);
614 	sbunlock(sb);
615 	asb = *sb;
616 	bzero((caddr_t)sb, sizeof (*sb));
617 	splx(s);
618 	if (pr->pr_family == AF_UNIX && (pr->pr_flags & PR_RIGHTS))
619 		unp_scan(asb.sb_mb, unp_discard);
620 	sbrelease(&asb);
621 }
622 
623 sosetopt(so, level, optname, m)
624 	register struct socket *so;
625 	int level, optname;
626 	register struct mbuf *m;
627 {
628 
629 	if (level != SOL_SOCKET)
630 		return (EINVAL);	/* XXX */
631 	switch (optname) {
632 
633 	case SO_DEBUG:
634 	case SO_KEEPALIVE:
635 	case SO_DONTROUTE:
636 	case SO_USELOOPBACK:
637 	case SO_REUSEADDR:
638 		so->so_options |= optname;
639 		break;
640 
641 	case SO_LINGER:
642 		if (m == NULL || m->m_len != sizeof (int))
643 			return (EINVAL);
644 		so->so_options |= SO_LINGER;
645 		so->so_linger = *mtod(m, int *);
646 		break;
647 
648 	case SO_DONTLINGER:
649 		so->so_options &= ~SO_LINGER;
650 		so->so_linger = 0;
651 		break;
652 
653 	default:
654 		return (EINVAL);
655 	}
656 	return (0);
657 }
658 
659 sogetopt(so, level, optname, m)
660 	register struct socket *so;
661 	int level, optname;
662 	register struct mbuf *m;
663 {
664 
665 	if (level != SOL_SOCKET)
666 		return (EINVAL);	/* XXX */
667 	switch (optname) {
668 
669 	case SO_USELOOPBACK:
670 	case SO_DONTROUTE:
671 	case SO_DEBUG:
672 	case SO_KEEPALIVE:
673 	case SO_LINGER:
674 	case SO_REUSEADDR:
675 		if ((so->so_options & optname) == 0)
676 			return (ENOPROTOOPT);
677 		if (optname == SO_LINGER && m != NULL) {
678 			*mtod(m, int *) = so->so_linger;
679 			m->m_len = sizeof (so->so_linger);
680 		}
681 		break;
682 
683 	default:
684 		return (EINVAL);
685 	}
686 	return (0);
687 }
688 
689 sohasoutofband(so)
690 	register struct socket *so;
691 {
692 
693 	if (so->so_pgrp == 0)
694 		return;
695 	if (so->so_pgrp > 0)
696 		gsignal(so->so_pgrp, SIGURG);
697 	else {
698 		struct proc *p = pfind(-so->so_pgrp);
699 
700 		if (p)
701 			psignal(p, SIGURG);
702 	}
703 }
704