xref: /csrg-svn/sys/kern/uipc_usrreq.c (revision 37616)
1 /*
2  * Copyright (c) 1982, 1986, 1989 Regents of the University of California.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms are permitted
6  * provided that the above copyright notice and this paragraph are
7  * duplicated in all such forms and that any documentation,
8  * advertising materials, and other materials related to such
9  * distribution and use acknowledge that the software was developed
10  * by the University of California, Berkeley.  The name of the
11  * University may not be used to endorse or promote products derived
12  * from this software without specific prior written permission.
13  * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
14  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
15  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
16  *
17  *	@(#)uipc_usrreq.c	7.2.1.1 (Berkeley) 05/02/89
18  */
19 
20 #include "param.h"
21 #include "dir.h"
22 #include "user.h"
23 #include "mbuf.h"
24 #include "domain.h"
25 #include "protosw.h"
26 #include "socket.h"
27 #include "socketvar.h"
28 #include "unpcb.h"
29 #include "un.h"
30 #include "vnode.h"
31 #include "mount.h"
32 #include "file.h"
33 #include "stat.h"
34 
35 /*
36  * Unix communications domain.
37  *
38  * TODO:
39  *	SEQPACKET, RDM
40  *	rethink name space problems
41  *	need a proper out-of-band
42  */
43 struct	sockaddr sun_noname = { AF_UNIX };
44 ino_t	unp_vno;			/* prototype for fake vnode numbers */
45 
46 /*ARGSUSED*/
47 uipc_usrreq(so, req, m, nam, rights)
48 	struct socket *so;
49 	int req;
50 	struct mbuf *m, *nam, *rights;
51 {
52 	struct unpcb *unp = sotounpcb(so);
53 	register struct socket *so2;
54 	int error = 0;
55 
56 	if (req == PRU_CONTROL)
57 		return (EOPNOTSUPP);
58 	if (req != PRU_SEND && rights && rights->m_len) {
59 		error = EOPNOTSUPP;
60 		goto release;
61 	}
62 	if (unp == 0 && req != PRU_ATTACH) {
63 		error = EINVAL;
64 		goto release;
65 	}
66 	switch (req) {
67 
68 	case PRU_ATTACH:
69 		if (unp) {
70 			error = EISCONN;
71 			break;
72 		}
73 		error = unp_attach(so);
74 		break;
75 
76 	case PRU_DETACH:
77 		unp_detach(unp);
78 		break;
79 
80 	case PRU_BIND:
81 		error = unp_bind(unp, nam);
82 		break;
83 
84 	case PRU_LISTEN:
85 		if (unp->unp_vnode == 0)
86 			error = EINVAL;
87 		break;
88 
89 	case PRU_CONNECT:
90 		error = unp_connect(so, nam);
91 		break;
92 
93 	case PRU_CONNECT2:
94 		error = unp_connect2(so, (struct socket *)nam);
95 		break;
96 
97 	case PRU_DISCONNECT:
98 		unp_disconnect(unp);
99 		break;
100 
101 	case PRU_ACCEPT:
102 		/*
103 		 * Pass back name of connected socket,
104 		 * if it was bound and we are still connected
105 		 * (our peer may have closed already!).
106 		 */
107 		if (unp->unp_conn && unp->unp_conn->unp_addr) {
108 			nam->m_len = unp->unp_conn->unp_addr->m_len;
109 			bcopy(mtod(unp->unp_conn->unp_addr, caddr_t),
110 			    mtod(nam, caddr_t), (unsigned)nam->m_len);
111 		} else {
112 			nam->m_len = sizeof(sun_noname);
113 			*(mtod(nam, struct sockaddr *)) = sun_noname;
114 		}
115 		break;
116 
117 	case PRU_SHUTDOWN:
118 		socantsendmore(so);
119 		unp_usrclosed(unp);
120 		break;
121 
122 	case PRU_RCVD:
123 		switch (so->so_type) {
124 
125 		case SOCK_DGRAM:
126 			panic("uipc 1");
127 			/*NOTREACHED*/
128 
129 		case SOCK_STREAM:
130 #define	rcv (&so->so_rcv)
131 #define snd (&so2->so_snd)
132 			if (unp->unp_conn == 0)
133 				break;
134 			so2 = unp->unp_conn->unp_socket;
135 			/*
136 			 * Adjust backpressure on sender
137 			 * and wakeup any waiting to write.
138 			 */
139 			snd->sb_mbmax += unp->unp_mbcnt - rcv->sb_mbcnt;
140 			unp->unp_mbcnt = rcv->sb_mbcnt;
141 			snd->sb_hiwat += unp->unp_cc - rcv->sb_cc;
142 			unp->unp_cc = rcv->sb_cc;
143 			sowwakeup(so2);
144 #undef snd
145 #undef rcv
146 			break;
147 
148 		default:
149 			panic("uipc 2");
150 		}
151 		break;
152 
153 	case PRU_SEND:
154 		if (rights) {
155 			error = unp_internalize(rights);
156 			if (error)
157 				break;
158 		}
159 		switch (so->so_type) {
160 
161 		case SOCK_DGRAM: {
162 			struct sockaddr *from;
163 
164 			if (nam) {
165 				if (unp->unp_conn) {
166 					error = EISCONN;
167 					break;
168 				}
169 				error = unp_connect(so, nam);
170 				if (error)
171 					break;
172 			} else {
173 				if (unp->unp_conn == 0) {
174 					error = ENOTCONN;
175 					break;
176 				}
177 			}
178 			so2 = unp->unp_conn->unp_socket;
179 			if (unp->unp_addr)
180 				from = mtod(unp->unp_addr, struct sockaddr *);
181 			else
182 				from = &sun_noname;
183 			if (sbspace(&so2->so_rcv) > 0 &&
184 			    sbappendaddr(&so2->so_rcv, from, m, rights)) {
185 				sorwakeup(so2);
186 				m = 0;
187 			} else
188 				error = ENOBUFS;
189 			if (nam)
190 				unp_disconnect(unp);
191 			break;
192 		}
193 
194 		case SOCK_STREAM:
195 #define	rcv (&so2->so_rcv)
196 #define	snd (&so->so_snd)
197 			if (so->so_state & SS_CANTSENDMORE) {
198 				error = EPIPE;
199 				break;
200 			}
201 			if (unp->unp_conn == 0)
202 				panic("uipc 3");
203 			so2 = unp->unp_conn->unp_socket;
204 			/*
205 			 * Send to paired receive port, and then reduce
206 			 * send buffer hiwater marks to maintain backpressure.
207 			 * Wake up readers.
208 			 */
209 			if (rights)
210 				(void)sbappendrights(rcv, m, rights);
211 			else
212 				sbappend(rcv, m);
213 			snd->sb_mbmax -=
214 			    rcv->sb_mbcnt - unp->unp_conn->unp_mbcnt;
215 			unp->unp_conn->unp_mbcnt = rcv->sb_mbcnt;
216 			snd->sb_hiwat -= rcv->sb_cc - unp->unp_conn->unp_cc;
217 			unp->unp_conn->unp_cc = rcv->sb_cc;
218 			sorwakeup(so2);
219 			m = 0;
220 #undef snd
221 #undef rcv
222 			break;
223 
224 		default:
225 			panic("uipc 4");
226 		}
227 		break;
228 
229 	case PRU_ABORT:
230 		unp_drop(unp, ECONNABORTED);
231 		break;
232 
233 	case PRU_SENSE:
234 		((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat;
235 		if (so->so_type == SOCK_STREAM && unp->unp_conn != 0) {
236 			so2 = unp->unp_conn->unp_socket;
237 			((struct stat *) m)->st_blksize += so2->so_rcv.sb_cc;
238 		}
239 		((struct stat *) m)->st_dev = NODEV;
240 		if (unp->unp_vno == 0)
241 			unp->unp_vno = unp_vno++;
242 		((struct stat *) m)->st_ino = unp->unp_vno;
243 		return (0);
244 
245 	case PRU_RCVOOB:
246 		return (EOPNOTSUPP);
247 
248 	case PRU_SENDOOB:
249 		error = EOPNOTSUPP;
250 		break;
251 
252 	case PRU_SOCKADDR:
253 		break;
254 
255 	case PRU_PEERADDR:
256 		if (unp->unp_conn && unp->unp_conn->unp_addr) {
257 			nam->m_len = unp->unp_conn->unp_addr->m_len;
258 			bcopy(mtod(unp->unp_conn->unp_addr, caddr_t),
259 			    mtod(nam, caddr_t), (unsigned)nam->m_len);
260 		}
261 		break;
262 
263 	case PRU_SLOWTIMO:
264 		break;
265 
266 	default:
267 		panic("piusrreq");
268 	}
269 release:
270 	if (m)
271 		m_freem(m);
272 	return (error);
273 }
274 
275 /*
276  * Both send and receive buffers are allocated PIPSIZ bytes of buffering
277  * for stream sockets, although the total for sender and receiver is
278  * actually only PIPSIZ.
279  * Datagram sockets really use the sendspace as the maximum datagram size,
280  * and don't really want to reserve the sendspace.  Their recvspace should
281  * be large enough for at least one max-size datagram plus address.
282  */
283 #define	PIPSIZ	4096
284 int	unpst_sendspace = PIPSIZ;
285 int	unpst_recvspace = PIPSIZ;
286 int	unpdg_sendspace = 2*1024;	/* really max datagram size */
287 int	unpdg_recvspace = 4*1024;
288 
289 int	unp_rights;			/* file descriptors in flight */
290 
291 unp_attach(so)
292 	struct socket *so;
293 {
294 	register struct mbuf *m;
295 	register struct unpcb *unp;
296 	int error;
297 
298 	switch (so->so_type) {
299 
300 	case SOCK_STREAM:
301 		error = soreserve(so, unpst_sendspace, unpst_recvspace);
302 		break;
303 
304 	case SOCK_DGRAM:
305 		error = soreserve(so, unpdg_sendspace, unpdg_recvspace);
306 		break;
307 	}
308 	if (error)
309 		return (error);
310 	m = m_getclr(M_DONTWAIT, MT_PCB);
311 	if (m == NULL)
312 		return (ENOBUFS);
313 	unp = mtod(m, struct unpcb *);
314 	so->so_pcb = (caddr_t)unp;
315 	unp->unp_socket = so;
316 	return (0);
317 }
318 
319 unp_detach(unp)
320 	register struct unpcb *unp;
321 {
322 
323 	if (unp->unp_vnode) {
324 		unp->unp_vnode->v_socket = 0;
325 		vrele(unp->unp_vnode);
326 		unp->unp_vnode = 0;
327 	}
328 	if (unp->unp_conn)
329 		unp_disconnect(unp);
330 	while (unp->unp_refs)
331 		unp_drop(unp->unp_refs, ECONNRESET);
332 	soisdisconnected(unp->unp_socket);
333 	unp->unp_socket->so_pcb = 0;
334 	m_freem(unp->unp_addr);
335 	(void) m_free(dtom(unp));
336 	if (unp_rights)
337 		unp_gc();
338 }
339 
340 unp_bind(unp, nam)
341 	struct unpcb *unp;
342 	struct mbuf *nam;
343 {
344 	struct sockaddr_un *soun = mtod(nam, struct sockaddr_un *);
345 	register struct vnode *vp;
346 	register struct nameidata *ndp = &u.u_nd;
347 	struct vattr vattr;
348 	int error;
349 
350 	ndp->ni_dirp = soun->sun_path;
351 	if (unp->unp_vnode != NULL || nam->m_len == MLEN)
352 		return (EINVAL);
353 	*(mtod(nam, caddr_t) + nam->m_len) = 0;
354 /* SHOULD BE ABLE TO ADOPT EXISTING AND wakeup() ALA FIFO's */
355 	ndp->ni_nameiop = CREATE | FOLLOW | LOCKPARENT;
356 	ndp->ni_segflg = UIO_SYSSPACE;
357 	if (error = namei(ndp))
358 		return (error);
359 	vp = ndp->ni_vp;
360 	if (vp != NULL) {
361 		vop_abortop(ndp);
362 		return (EADDRINUSE);
363 	}
364 	vattr_null(&vattr);
365 	vattr.va_type = VSOCK;
366 	vattr.va_mode = 0777;
367 	if (error = vop_create(ndp, &vattr))
368 		return (error);
369 	vp = ndp->ni_vp;
370 	vp->v_socket = unp->unp_socket;
371 	unp->unp_vnode = vp;
372 	unp->unp_addr = m_copy(nam, 0, (int)M_COPYALL);
373 	vop_unlock(vp);
374 	return (0);
375 }
376 
377 unp_connect(so, nam)
378 	struct socket *so;
379 	struct mbuf *nam;
380 {
381 	register struct sockaddr_un *soun = mtod(nam, struct sockaddr_un *);
382 	register struct vnode *vp;
383 	int error;
384 	register struct socket *so2;
385 	register struct nameidata *ndp = &u.u_nd;
386 
387 	ndp->ni_dirp = soun->sun_path;
388 	if (nam->m_len + (nam->m_off - MMINOFF) == MLEN)
389 		return (EMSGSIZE);
390 	*(mtod(nam, caddr_t) + nam->m_len) = 0;
391 	ndp->ni_nameiop = LOOKUP | FOLLOW;
392 	ndp->ni_segflg = UIO_SYSSPACE;
393 	if (error = namei(ndp))
394 		return (error);
395 	vp = ndp->ni_vp;
396 	if (error = vn_access(vp, VWRITE, u.u_cred))
397 		goto bad;
398 	if (vp->v_type != VSOCK) {
399 		error = ENOTSOCK;
400 		goto bad;
401 	}
402 	so2 = vp->v_socket;
403 	if (so2 == 0) {
404 		error = ECONNREFUSED;
405 		goto bad;
406 	}
407 	if (so->so_type != so2->so_type) {
408 		error = EPROTOTYPE;
409 		goto bad;
410 	}
411 	if (so->so_proto->pr_flags & PR_CONNREQUIRED &&
412 	    ((so2->so_options&SO_ACCEPTCONN) == 0 ||
413 	     (so2 = sonewconn(so2)) == 0)) {
414 		error = ECONNREFUSED;
415 		goto bad;
416 	}
417 	error = unp_connect2(so, so2);
418 bad:
419 	vrele(vp);
420 	return (error);
421 }
422 
423 unp_connect2(so, so2)
424 	register struct socket *so;
425 	register struct socket *so2;
426 {
427 	register struct unpcb *unp = sotounpcb(so);
428 	register struct unpcb *unp2;
429 
430 	if (so2->so_type != so->so_type)
431 		return (EPROTOTYPE);
432 	unp2 = sotounpcb(so2);
433 	unp->unp_conn = unp2;
434 	switch (so->so_type) {
435 
436 	case SOCK_DGRAM:
437 		unp->unp_nextref = unp2->unp_refs;
438 		unp2->unp_refs = unp;
439 		soisconnected(so);
440 		break;
441 
442 	case SOCK_STREAM:
443 		unp2->unp_conn = unp;
444 		soisconnected(so2);
445 		soisconnected(so);
446 		break;
447 
448 	default:
449 		panic("unp_connect2");
450 	}
451 	return (0);
452 }
453 
454 unp_disconnect(unp)
455 	struct unpcb *unp;
456 {
457 	register struct unpcb *unp2 = unp->unp_conn;
458 
459 	if (unp2 == 0)
460 		return;
461 	unp->unp_conn = 0;
462 	switch (unp->unp_socket->so_type) {
463 
464 	case SOCK_DGRAM:
465 		if (unp2->unp_refs == unp)
466 			unp2->unp_refs = unp->unp_nextref;
467 		else {
468 			unp2 = unp2->unp_refs;
469 			for (;;) {
470 				if (unp2 == 0)
471 					panic("unp_disconnect");
472 				if (unp2->unp_nextref == unp)
473 					break;
474 				unp2 = unp2->unp_nextref;
475 			}
476 			unp2->unp_nextref = unp->unp_nextref;
477 		}
478 		unp->unp_nextref = 0;
479 		unp->unp_socket->so_state &= ~SS_ISCONNECTED;
480 		break;
481 
482 	case SOCK_STREAM:
483 		soisdisconnected(unp->unp_socket);
484 		unp2->unp_conn = 0;
485 		soisdisconnected(unp2->unp_socket);
486 		break;
487 	}
488 }
489 
490 #ifdef notdef
491 unp_abort(unp)
492 	struct unpcb *unp;
493 {
494 
495 	unp_detach(unp);
496 }
497 #endif
498 
499 /*ARGSUSED*/
500 unp_usrclosed(unp)
501 	struct unpcb *unp;
502 {
503 
504 }
505 
506 unp_drop(unp, errno)
507 	struct unpcb *unp;
508 	int errno;
509 {
510 	struct socket *so = unp->unp_socket;
511 
512 	so->so_error = errno;
513 	unp_disconnect(unp);
514 	if (so->so_head) {
515 		so->so_pcb = (caddr_t) 0;
516 		m_freem(unp->unp_addr);
517 		(void) m_free(dtom(unp));
518 		sofree(so);
519 	}
520 }
521 
522 #ifdef notdef
523 unp_drain()
524 {
525 
526 }
527 #endif
528 
529 unp_externalize(rights)
530 	struct mbuf *rights;
531 {
532 	int newfds = rights->m_len / sizeof (int);
533 	register int i;
534 	register struct file **rp = mtod(rights, struct file **);
535 	register struct file *fp;
536 	int f;
537 
538 	if (newfds > ufavail()) {
539 		for (i = 0; i < newfds; i++) {
540 			fp = *rp;
541 			unp_discard(fp);
542 			*rp++ = 0;
543 		}
544 		return (EMSGSIZE);
545 	}
546 	for (i = 0; i < newfds; i++) {
547 		if (ufalloc(0, &f))
548 			panic("unp_externalize");
549 		fp = *rp;
550 		u.u_ofile[f] = fp;
551 		fp->f_msgcount--;
552 		unp_rights--;
553 		*(int *)rp++ = f;
554 	}
555 	return (0);
556 }
557 
558 unp_internalize(rights)
559 	struct mbuf *rights;
560 {
561 	register struct file **rp;
562 	int oldfds = rights->m_len / sizeof (int);
563 	register int i;
564 	register struct file *fp;
565 
566 	rp = mtod(rights, struct file **);
567 	for (i = 0; i < oldfds; i++)
568 		if (getf(*(int *)rp++) == 0)
569 			return (EBADF);
570 	rp = mtod(rights, struct file **);
571 	for (i = 0; i < oldfds; i++) {
572 		fp = getf(*(int *)rp);
573 		*rp++ = fp;
574 		fp->f_count++;
575 		fp->f_msgcount++;
576 		unp_rights++;
577 	}
578 	return (0);
579 }
580 
581 int	unp_defer, unp_gcing;
582 int	unp_mark();
583 extern	struct domain unixdomain;
584 
585 unp_gc()
586 {
587 	register struct file *fp;
588 	register struct socket *so;
589 
590 	if (unp_gcing)
591 		return;
592 	unp_gcing = 1;
593 restart:
594 	unp_defer = 0;
595 	for (fp = file; fp < fileNFILE; fp++)
596 		fp->f_flag &= ~(FMARK|FDEFER);
597 	do {
598 		for (fp = file; fp < fileNFILE; fp++) {
599 			if (fp->f_count == 0)
600 				continue;
601 			if (fp->f_flag & FDEFER) {
602 				fp->f_flag &= ~FDEFER;
603 				unp_defer--;
604 			} else {
605 				if (fp->f_flag & FMARK)
606 					continue;
607 				if (fp->f_count == fp->f_msgcount)
608 					continue;
609 				fp->f_flag |= FMARK;
610 			}
611 			if (fp->f_type != DTYPE_SOCKET)
612 				continue;
613 			so = (struct socket *)fp->f_data;
614 			if (so->so_proto->pr_domain != &unixdomain ||
615 			    (so->so_proto->pr_flags&PR_RIGHTS) == 0)
616 				continue;
617 			if (so->so_rcv.sb_flags & SB_LOCK) {
618 				sbwait(&so->so_rcv);
619 				goto restart;
620 			}
621 			unp_scan(so->so_rcv.sb_mb, unp_mark);
622 		}
623 	} while (unp_defer);
624 	for (fp = file; fp < fileNFILE; fp++) {
625 		if (fp->f_count == 0)
626 			continue;
627 		if (fp->f_count == fp->f_msgcount && (fp->f_flag & FMARK) == 0)
628 			while (fp->f_msgcount)
629 				unp_discard(fp);
630 	}
631 	unp_gcing = 0;
632 }
633 
634 unp_dispose(m)
635 	struct mbuf *m;
636 {
637 	int unp_discard();
638 
639 	if (m)
640 		unp_scan(m, unp_discard);
641 }
642 
643 unp_scan(m0, op)
644 	register struct mbuf *m0;
645 	int (*op)();
646 {
647 	register struct mbuf *m;
648 	register struct file **rp;
649 	register int i;
650 	int qfds;
651 
652 	while (m0) {
653 		for (m = m0; m; m = m->m_next)
654 			if (m->m_type == MT_RIGHTS && m->m_len) {
655 				qfds = m->m_len / sizeof (struct file *);
656 				rp = mtod(m, struct file **);
657 				for (i = 0; i < qfds; i++)
658 					(*op)(*rp++);
659 				break;		/* XXX, but saves time */
660 			}
661 		m0 = m0->m_act;
662 	}
663 }
664 
665 unp_mark(fp)
666 	struct file *fp;
667 {
668 
669 	if (fp->f_flag & FMARK)
670 		return;
671 	unp_defer++;
672 	fp->f_flag |= (FMARK|FDEFER);
673 }
674 
675 unp_discard(fp)
676 	struct file *fp;
677 {
678 
679 	fp->f_msgcount--;
680 	unp_rights--;
681 	closef(fp);
682 }
683