xref: /csrg-svn/sys/kern/uipc_socket.c (revision 8391)
1 /*	uipc_socket.c	4.52	82/10/09	*/
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/protosw.h"
13 #include "../h/socket.h"
14 #include "../h/socketvar.h"
15 #include "../h/stat.h"
16 #include "../h/ioctl.h"
17 #include "../h/uio.h"
18 #include "../net/route.h"
19 
20 /*
21  * Socket operation routines.
22  * These routines are called by the routines in
23  * sys_socket.c or from a system process, and
24  * implement the semantics of socket operations by
25  * switching out to the protocol specific routines.
26  */
27 
28 socreate(dom, aso, type, proto, opt)
29 	struct socket **aso;
30 	int type, proto;
31 	struct socketopt *opt;
32 {
33 	register struct protosw *prp;
34 	register struct socket *so;
35 	struct mbuf *m;
36 	int pf, error;
37 
38 	pf = dom ? PF_UNIX : PF_INET;		/* should be u.u_protof */
39 	if (proto)
40 		prp = pffindproto(pf, proto);
41 	else
42 		prp = pffindtype(pf, type);
43 	if (prp == 0)
44 		return (EPROTONOSUPPORT);
45 	if (prp->pr_type != type)
46 		return (EPROTOTYPE);
47 	m = m_getclr(M_WAIT);
48 	if (m == 0)
49 		return (ENOBUFS);
50 	so = mtod(m, struct socket *);
51 	so->so_options = 0;
52 	so->so_state = 0;
53 	if (u.u_uid == 0)
54 		so->so_state = SS_PRIV;
55 	so->so_proto = prp;
56 	error = (*prp->pr_usrreq)(so, PRU_ATTACH,
57 	    (struct mbuf *)0, (struct mbuf *)0, (struct socketopt *)0);
58 	if (error) {
59 		so->so_state |= SS_NOFDREF;
60 		sofree(so);
61 		return (error);
62 	}
63 	*aso = so;
64 	return (0);
65 }
66 
67 sobind(so, nam, opt)
68 	struct socket *so;
69 	struct mbuf *nam;
70 	struct socketopt *opt;
71 {
72 	int s = splnet();
73 	int error;
74 
75 	error =
76 	    (*so->so_proto->pr_usrreq)(so, PRU_BIND,
77 		(struct mbuf *)0, nam, opt);
78 	splx(s);
79 	return (error);
80 }
81 
82 solisten(so, backlog)
83 	struct socket *so;
84 	int backlog;
85 {
86 	int s = splnet();
87 	int error;
88 
89 	error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN,
90 	    (struct mbuf *)0, (struct mbuf *)0, (struct socketopt *)0);
91 	if (error) {
92 		splx(s);
93 		return (error);
94 	}
95 	if (so->so_q == 0) {
96 		so->so_q = so;
97 		so->so_q0 = so;
98 		so->so_options |= SO_ACCEPTCONN;
99 	}
100 	if (backlog < 0)
101 		backlog = 0;
102 	so->so_qlimit = backlog < 5 ? backlog : 5;
103 	so->so_options |= SO_NEWFDONCONN;
104 	return (0);
105 }
106 
107 sofree(so)
108 	struct socket *so;
109 {
110 
111 	if (so->so_head) {
112 		if (!soqremque(so, 0) && !soqremque(so, 1))
113 			panic("sofree dq");
114 		so->so_head = 0;
115 	}
116 	if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
117 		return;
118 	sbrelease(&so->so_snd);
119 	sbrelease(&so->so_rcv);
120 	(void) m_free(dtom(so));
121 }
122 
123 /*
124  * Close a socket on last file table reference removal.
125  * Initiate disconnect if connected.
126  * Free socket when disconnect complete.
127  */
128 soclose(so, exiting)
129 	register struct socket *so;
130 	int exiting;
131 {
132 	int s = splnet();		/* conservative */
133 
134 	if (so->so_options & SO_ACCEPTCONN) {
135 		while (so->so_q0 != so)
136 			soclose(so->so_q0, 1);
137 		while (so->so_q != so)
138 			soclose(so->so_q, 1);
139 	}
140 	if (so->so_pcb == 0)
141 		goto discard;
142 	if (exiting)
143 		so->so_options |= SO_KEEPALIVE;
144 	if (so->so_state & SS_ISCONNECTED) {
145 		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
146 			u.u_error = sodisconnect(so, (struct sockaddr *)0);
147 			if (u.u_error) {
148 				if (exiting)
149 					goto drop;
150 				splx(s);
151 				return;
152 			}
153 		}
154 		if ((so->so_options & SO_DONTLINGER) == 0) {
155 			if ((so->so_state & SS_ISDISCONNECTING) &&
156 			    (so->so_state & SS_NBIO) &&
157 			    exiting == 0) {
158 				u.u_error = EINPROGRESS;
159 				splx(s);
160 				return;
161 			}
162 			/* should use tsleep here, for at most linger */
163 			while (so->so_state & SS_ISCONNECTED)
164 				sleep((caddr_t)&so->so_timeo, PZERO+1);
165 		}
166 	}
167 drop:
168 	if (so->so_pcb) {
169 		u.u_error = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
170 		    (struct mbuf *)0, (struct mbuf *)0, (struct socketopt *)0);
171 		if (exiting == 0 && u.u_error) {
172 			splx(s);
173 			return;
174 		}
175 	}
176 discard:
177 	so->so_state |= SS_NOFDREF;
178 	sofree(so);
179 	splx(s);
180 }
181 
182 /*ARGSUSED*/
183 sostat(so, sb)
184 	struct socket *so;
185 	struct stat *sb;
186 {
187 
188 	bzero((caddr_t)sb, sizeof (*sb));		/* XXX */
189 	return (0);					/* XXX */
190 }
191 
192 soaccept(so, nam, opt)
193 	struct socket *so;
194 	struct mbuf *nam;
195 	struct socketopt *opt;
196 {
197 	int s = splnet();
198 	int error;
199 
200 	error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
201 	    (struct mbuf *)0, nam, opt);
202 	splx(s);
203 	return (error);
204 }
205 
206 soconnect(so, nam, opt)
207 	struct socket *so;
208 	struct mbuf *nam;
209 	struct socketopt *opt;
210 {
211 	int s = splnet();
212 	int error;
213 
214 	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) {
215 		error = EISCONN;
216 		goto bad;
217 	}
218 	error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
219 	    (struct mbuf *)0, nam, opt);
220 bad:
221 	splx(s);
222 	return (error);
223 }
224 
225 sodisconnect(so, nam)
226 	struct socket *so;
227 	struct mbuf *nam;
228 {
229 	int s = splnet();
230 	int error;
231 
232 	if ((so->so_state & SS_ISCONNECTED) == 0) {
233 		error = ENOTCONN;
234 		goto bad;
235 	}
236 	if (so->so_state & SS_ISDISCONNECTING) {
237 		error = EALREADY;
238 		goto bad;
239 	}
240 	error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
241 	    (struct mbuf *)0, nam, (struct socketopt *)0);
242 bad:
243 	splx(s);
244 	return (error);
245 }
246 
247 /*
248  * Send on a socket.
249  * If send must go all at once and message is larger than
250  * send buffering, then hard error.
251  * Lock against other senders.
252  * If must go all at once and not enough room now, then
253  * inform user that this would block and do nothing.
254  */
255 sosend(so, nam, uio, flags)
256 	register struct socket *so;
257 	struct mbuf *nam;
258 	struct uio *uio;
259 	int flags;
260 {
261 	struct mbuf *top = 0;
262 	register struct mbuf *m, **mp = &top;
263 	register u_int len;
264 	int error = 0, space, s;
265 
266 	if (sosendallatonce(so) && uio->uio_resid > so->so_snd.sb_hiwat)
267 		return (EMSGSIZE);
268 restart:
269 	sblock(&so->so_snd);
270 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
271 
272 	u.u_ru.ru_msgsnd++;
273 again:
274 	s = splnet();
275 	if (so->so_state & SS_CANTSENDMORE) {
276 		psignal(u.u_procp, SIGPIPE);
277 		snderr(EPIPE);
278 	}
279 	if (so->so_error) {
280 		error = so->so_error;
281 		so->so_error = 0;				/* ??? */
282 		splx(s);
283 		goto release;
284 	}
285 	if ((so->so_state & SS_ISCONNECTED) == 0) {
286 		if (so->so_proto->pr_flags & PR_CONNREQUIRED)
287 			snderr(ENOTCONN);
288 		if (nam == 0)
289 			snderr(EDESTADDRREQ);
290 	}
291 	if (top) {
292 		error = (*so->so_proto->pr_usrreq)(so,
293 		    (flags & SOF_OOB) ? PRU_SENDOOB : PRU_SEND,
294 		    top, (caddr_t)nam, (struct socketopt *)0);
295 		top = 0;
296 		if (error) {
297 			splx(s);
298 			goto release;
299 		}
300 		mp = &top;
301 	}
302 	if (uio->uio_resid == 0) {
303 		splx(s);
304 		goto release;
305 	}
306 	if (flags & SOF_OOB)
307 		space = 1024;
308 	else {
309 		space = sbspace(&so->so_snd);
310 		if (space <= 0 ||
311 		    sosendallatonce(so) && space < uio->uio_resid) {
312 			if (so->so_state & SS_NBIO)
313 				snderr(EWOULDBLOCK);
314 			sbunlock(&so->so_snd);
315 			sbwait(&so->so_snd);
316 			splx(s);
317 			goto restart;
318 		}
319 	}
320 	splx(s);
321 	while (uio->uio_resid > 0 && space > 0) {
322 		register struct iovec *iov = uio->uio_iov;
323 
324 		if (iov->iov_len == 0) {
325 			uio->uio_iov++;
326 			uio->uio_iovcnt--;
327 			if (uio->uio_iovcnt < 0)
328 				panic("sosend");
329 			continue;
330 		}
331 		MGET(m, 1);
332 		if (m == NULL) {
333 			error = ENOBUFS;			/* SIGPIPE? */
334 			goto release;
335 		}
336 		if (iov->iov_len >= CLBYTES && space >= CLBYTES) {
337 			register struct mbuf *p;
338 			MCLGET(p, 1);
339 			if (p == 0)
340 				goto nopages;
341 			m->m_off = (int)p - (int)m;
342 			len = CLBYTES;
343 		} else {
344 nopages:
345 			len = MIN(MLEN, iov->iov_len);
346 		}
347 		uiomove(mtod(m, caddr_t), len, UIO_WRITE, uio);
348 		m->m_len = len;
349 		*mp = m;
350 		mp = &m->m_next;
351 		if (flags & SOF_OOB)
352 			space -= len;
353 		else
354 			space = sbspace(&so->so_snd);
355 	}
356 	goto again;
357 
358 release:
359 	sbunlock(&so->so_snd);
360 	if (top)
361 		m_freem(top);
362 	return (error);
363 }
364 
365 soreceive(so, aname, uio, flags)
366 	register struct socket *so;
367 	struct mbuf **aname;
368 	struct uio *uio;
369 	int flags;
370 {
371 	register struct iovec *iov;
372 	register struct mbuf *m, *n;
373 	u_int len;
374 	int eor, s, error = 0, moff, tomark;
375 
376 	if (flags & SOF_OOB) {
377 		struct mbuf *m = m_get(M_WAIT);
378 
379 		(*so->so_proto->pr_usrreq)(so, PRU_RCVOOB,
380 		    m, (struct mbuf *)0, (struct socketopt *)0);
381 		len = uio->uio_resid;
382 		do {
383 			if (len > m->m_len)
384 				len = m->m_len;
385 			uiomove(mtod(m, caddr_t), (int)len, UIO_READ, uio);
386 			m = m_free(m);
387 		} while (uio->uio_resid && u.u_error == 0 && m);
388 		if (m)
389 			(void) m_freem(m);
390 		return;
391 	}
392 
393 restart:
394 	sblock(&so->so_rcv);
395 	s = splnet();
396 
397 #define	rcverr(errno)	{ error = errno; splx(s); goto release; }
398 	if (so->so_rcv.sb_cc == 0) {
399 		if (so->so_error) {
400 			error = so->so_error;
401 			so->so_error = 0;
402 			splx(s);
403 			goto release;
404 		}
405 		if (so->so_state & SS_CANTRCVMORE) {
406 			splx(s);
407 			goto release;
408 		}
409 		if ((so->so_state & SS_ISCONNECTED) == 0 &&
410 		    (so->so_proto->pr_flags & PR_CONNREQUIRED))
411 			rcverr(ENOTCONN);
412 		if (so->so_state & SS_NBIO)
413 			rcverr(EWOULDBLOCK);
414 		sbunlock(&so->so_rcv);
415 		sbwait(&so->so_rcv);
416 		splx(s);
417 		goto restart;
418 	}
419 	u.u_ru.ru_msgrcv++;
420 	m = so->so_rcv.sb_mb;
421 	if (m == 0)
422 		panic("receive");
423 	if (so->so_proto->pr_flags & PR_ADDR) {
424 		if ((flags & SOF_PREVIEW) == 0) {
425 			so->so_rcv.sb_cc -= m->m_len;
426 			so->so_rcv.sb_mbcnt -= MSIZE;
427 		}
428 		if (aname) {
429 			if (flags & SOF_PREVIEW)
430 				*aname = m_copy(m, 0, m->m_len);
431 			else
432 				*aname = m;
433 			m = m->m_next;
434 			(*aname)->m_next = 0;
435 		} else
436 			if (flags & SOF_PREVIEW)
437 				m = m->m_next;
438 			else
439 				m = m_free(m);
440 		if (m == 0)
441 			panic("receive 2");
442 	}
443 	eor = 0;
444 	moff = 0;
445 	tomark = so->so_oobmark;
446 	do {
447 		if (uio->uio_resid <= 0)
448 			break;
449 		len = uio->uio_resid;
450 		so->so_state &= ~SS_RCVATMARK;
451 		if (tomark && len > tomark)
452 			len = tomark;
453 		if (len > m->m_len - moff)
454 			len = m->m_len - moff;
455 		splx(s);
456 		uiomove(mtod(m, caddr_t) + moff, (int)len, UIO_READ, uio);
457 		s = splnet();
458 		if (len == m->m_len) {
459 			eor = (int)m->m_act;
460 			if (flags & SOF_PREVIEW)
461 				m = m->m_next;
462 			else {
463 				sbfree(&so->so_rcv, m);
464 				MFREE(m, n);
465 				m = n;
466 			}
467 			moff = 0;
468 		} else {
469 			if (flags & SOF_PREVIEW)
470 				moff += len;
471 			else {
472 				m->m_off += len;
473 				m->m_len -= len;
474 				so->so_rcv.sb_cc -= len;
475 			}
476 		}
477 		if ((flags & SOF_PREVIEW) == 0 && so->so_oobmark) {
478 			so->so_oobmark -= len;
479 			if (so->so_oobmark == 0) {
480 				so->so_state |= SS_RCVATMARK;
481 				break;
482 			}
483 		}
484 		if (tomark) {
485 			tomark -= len;
486 			if (tomark == 0)
487 				break;
488 		}
489 	} while (m && !eor);
490 	if (flags & SOF_PREVIEW)
491 		goto release;
492 	so->so_rcv.sb_mb = m;
493 	if ((so->so_proto->pr_flags & PR_ATOMIC) && eor == 0)
494 		do {
495 			if (m == 0)
496 				panic("receive 3");
497 			sbfree(&so->so_rcv, m);
498 			eor = (int)m->m_act;
499 			so->so_rcv.sb_mb = m->m_next;
500 			MFREE(m, n);
501 			m = n;
502 		} while (eor == 0);
503 	if ((so->so_proto->pr_flags & PR_WANTRCVD) && so->so_pcb)
504 		(*so->so_proto->pr_usrreq)(so, PRU_RCVD,
505 		    (struct mbuf *)0, (struct mbuf *)0, (struct socketopt *)0);
506 release:
507 	sbunlock(&so->so_rcv);
508 	splx(s);
509 	return (error);
510 }
511 
512 sohasoutofband(so)
513 	struct socket *so;
514 {
515 
516 	if (so->so_pgrp == 0)
517 		return;
518 	if (so->so_pgrp > 0)
519 		gsignal(so->so_pgrp, SIGURG);
520 	else {
521 		struct proc *p = pfind(-so->so_pgrp);
522 
523 		if (p)
524 			psignal(p, SIGURG);
525 	}
526 }
527 
528 /*ARGSUSED*/
529 soioctl(so, cmd, data)
530 	register struct socket *so;
531 	int cmd;
532 	register char *data;
533 {
534 
535 	switch (cmd) {
536 
537 	case FIONBIO:
538 		if (*(int *)data)
539 			so->so_state |= SS_NBIO;
540 		else
541 			so->so_state &= ~SS_NBIO;
542 		return;
543 
544 	case FIOASYNC:
545 		if (*(int *)data)
546 			so->so_state |= SS_ASYNC;
547 		else
548 			so->so_state &= ~SS_ASYNC;
549 		return;
550 
551 	case SIOCSKEEP:
552 		if (*(int *)data)
553 			so->so_options &= ~SO_KEEPALIVE;
554 		else
555 			so->so_options |= SO_KEEPALIVE;
556 		return;
557 
558 	case SIOCGKEEP:
559 		*(int *)data = (so->so_options & SO_KEEPALIVE) != 0;
560 		return;
561 
562 	case SIOCSLINGER:
563 		so->so_linger = *(int *)data;
564 		if (so->so_linger)
565 			so->so_options &= ~SO_DONTLINGER;
566 		else
567 			so->so_options |= SO_DONTLINGER;
568 		return;
569 
570 	case SIOCGLINGER:
571 		*(int *)data = so->so_linger;
572 		return;
573 
574 	case SIOCSPGRP:
575 		so->so_pgrp = *(int *)data;
576 		return;
577 
578 	case SIOCGPGRP:
579 		*(int *)data = so->so_pgrp;
580 		return;
581 
582 	case SIOCDONE: {
583 		int flags = *(int *)data;
584 
585 		flags++;
586 		if (flags & FREAD) {
587 			int s = splimp();
588 			socantrcvmore(so);
589 			sbflush(&so->so_rcv);
590 			splx(s);
591 		}
592 		if (flags & FWRITE)
593 			u.u_error = (*so->so_proto->pr_usrreq)(so, PRU_SHUTDOWN,
594 			    (struct mbuf *)0, (struct mbuf *)0,
595 			    (struct socketopt *)0);
596 		return;
597 	}
598 
599 	case SIOCSENDOOB: {
600 		char oob = *(char *)data;
601 		struct mbuf *m = m_get(M_DONTWAIT);
602 
603 		if (m == 0) {
604 			u.u_error = ENOBUFS;
605 			return;
606 		}
607 		m->m_len = 1;
608 		*mtod(m, char *) = oob;
609 		(*so->so_proto->pr_usrreq)(so, PRU_SENDOOB,
610 		    m, (struct mbuf *)0, (struct socketopt *)0);
611 		return;
612 	}
613 
614 	case SIOCRCVOOB: {
615 		struct mbuf *m = m_get(M_WAIT);
616 
617 		if (m == 0) {
618 			u.u_error = ENOBUFS;
619 			return;
620 		}
621 		*mtod(m, caddr_t) = 0;
622 		(*so->so_proto->pr_usrreq)(so, PRU_RCVOOB,
623 		    m, (struct mbuf *)0, (struct socketopt *)0);
624 		*(char *)data = *mtod(m, char *);
625 		(void) m_free(m);
626 		return;
627 	}
628 
629 	case SIOCATMARK:
630 		*(int *)data = (so->so_state&SS_RCVATMARK) != 0;
631 		return;
632 
633 	/* routing table update calls */
634 	case SIOCADDRT:
635 	case SIOCDELRT:
636 		if (!suser())
637 			return;
638 		u.u_error = rtrequest(cmd, (struct rtentry *)data);
639 		return;
640 
641 	/* type/protocol specific ioctls */
642 	}
643 	u.u_error = EOPNOTSUPP;
644 }
645