xref: /freebsd-11-stable/sys/kern/uipc_syscalls.c (revision e44b2b6db6e78f348c1062529eafe2e5f8f61c1e)
1 /*-
2  * Copyright (c) 1982, 1986, 1989, 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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)uipc_syscalls.c	8.4 (Berkeley) 2/21/94
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_capsicum.h"
36 #include "opt_inet.h"
37 #include "opt_inet6.h"
38 #include "opt_compat.h"
39 #include "opt_ktrace.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/capsicum.h>
44 #include <sys/kernel.h>
45 #include <sys/lock.h>
46 #include <sys/mutex.h>
47 #include <sys/sysproto.h>
48 #include <sys/malloc.h>
49 #include <sys/filedesc.h>
50 #include <sys/proc.h>
51 #include <sys/filio.h>
52 #include <sys/jail.h>
53 #include <sys/mbuf.h>
54 #include <sys/protosw.h>
55 #include <sys/rwlock.h>
56 #include <sys/socket.h>
57 #include <sys/socketvar.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/un.h>
60 #include <sys/unpcb.h>
61 #ifdef KTRACE
62 #include <sys/ktrace.h>
63 #endif
64 #ifdef COMPAT_FREEBSD32
65 #include <compat/freebsd32/freebsd32_util.h>
66 #endif
67 
68 #include <net/vnet.h>
69 
70 #include <security/audit/audit.h>
71 #include <security/mac/mac_framework.h>
72 
73 /*
74  * Flags for accept1() and kern_accept4(), in addition to SOCK_CLOEXEC
75  * and SOCK_NONBLOCK.
76  */
77 #define	ACCEPT4_INHERIT	0x1
78 #define	ACCEPT4_COMPAT	0x2
79 
80 static int sendit(struct thread *td, int s, struct msghdr *mp, int flags);
81 static int recvit(struct thread *td, int s, struct msghdr *mp, void *namelenp);
82 
83 static int accept1(struct thread *td, int s, struct sockaddr *uname,
84 		   socklen_t *anamelen, int flags);
85 static int getsockname1(struct thread *td, struct getsockname_args *uap,
86 			int compat);
87 static int getpeername1(struct thread *td, struct getpeername_args *uap,
88 			int compat);
89 static int sockargs(struct mbuf **, char *, socklen_t, int);
90 
91 /*
92  * Convert a user file descriptor to a kernel file entry and check if required
93  * capability rights are present.
94  * If required copy of current set of capability rights is returned.
95  * A reference on the file entry is held upon returning.
96  */
97 int
getsock_cap(struct thread * td,int fd,cap_rights_t * rightsp,struct file ** fpp,u_int * fflagp,struct filecaps * havecapsp)98 getsock_cap(struct thread *td, int fd, cap_rights_t *rightsp,
99     struct file **fpp, u_int *fflagp, struct filecaps *havecapsp)
100 {
101 	struct file *fp;
102 	int error;
103 
104 	error = fget_cap(td, fd, rightsp, &fp, havecapsp);
105 	if (error != 0)
106 		return (error);
107 	if (fp->f_type != DTYPE_SOCKET) {
108 		fdrop(fp, td);
109 		if (havecapsp != NULL)
110 			filecaps_free(havecapsp);
111 		return (ENOTSOCK);
112 	}
113 	if (fflagp != NULL)
114 		*fflagp = fp->f_flag;
115 	*fpp = fp;
116 	return (0);
117 }
118 
119 /*
120  * System call interface to the socket abstraction.
121  */
122 #if defined(COMPAT_43)
123 #define COMPAT_OLDSOCK
124 #endif
125 
126 int
sys_socket(td,uap)127 sys_socket(td, uap)
128 	struct thread *td;
129 	struct socket_args /* {
130 		int	domain;
131 		int	type;
132 		int	protocol;
133 	} */ *uap;
134 {
135 
136 	return (kern_socket(td, uap->domain, uap->type, uap->protocol));
137 }
138 
139 int
kern_socket(struct thread * td,int domain,int type,int protocol)140 kern_socket(struct thread *td, int domain, int type, int protocol)
141 {
142 	struct socket *so;
143 	struct file *fp;
144 	int fd, error, oflag, fflag;
145 
146 	AUDIT_ARG_SOCKET(domain, type, protocol);
147 
148 	oflag = 0;
149 	fflag = 0;
150 	if ((type & SOCK_CLOEXEC) != 0) {
151 		type &= ~SOCK_CLOEXEC;
152 		oflag |= O_CLOEXEC;
153 	}
154 	if ((type & SOCK_NONBLOCK) != 0) {
155 		type &= ~SOCK_NONBLOCK;
156 		fflag |= FNONBLOCK;
157 	}
158 
159 #ifdef MAC
160 	error = mac_socket_check_create(td->td_ucred, domain, type, protocol);
161 	if (error != 0)
162 		return (error);
163 #endif
164 	error = falloc(td, &fp, &fd, oflag);
165 	if (error != 0)
166 		return (error);
167 	/* An extra reference on `fp' has been held for us by falloc(). */
168 	error = socreate(domain, &so, type, protocol, td->td_ucred, td);
169 	if (error != 0) {
170 		fdclose(td, fp, fd);
171 	} else {
172 		finit(fp, FREAD | FWRITE | fflag, DTYPE_SOCKET, so, &socketops);
173 		if ((fflag & FNONBLOCK) != 0)
174 			(void) fo_ioctl(fp, FIONBIO, &fflag, td->td_ucred, td);
175 		td->td_retval[0] = fd;
176 	}
177 	fdrop(fp, td);
178 	return (error);
179 }
180 
181 /* ARGSUSED */
182 int
sys_bind(td,uap)183 sys_bind(td, uap)
184 	struct thread *td;
185 	struct bind_args /* {
186 		int	s;
187 		caddr_t	name;
188 		int	namelen;
189 	} */ *uap;
190 {
191 	struct sockaddr *sa;
192 	int error;
193 
194 	error = getsockaddr(&sa, uap->name, uap->namelen);
195 	if (error == 0) {
196 		error = kern_bindat(td, AT_FDCWD, uap->s, sa);
197 		free(sa, M_SONAME);
198 	}
199 	return (error);
200 }
201 
202 int
kern_bindat(struct thread * td,int dirfd,int fd,struct sockaddr * sa)203 kern_bindat(struct thread *td, int dirfd, int fd, struct sockaddr *sa)
204 {
205 	struct socket *so;
206 	struct file *fp;
207 	cap_rights_t rights;
208 	int error;
209 
210 	AUDIT_ARG_FD(fd);
211 	AUDIT_ARG_SOCKADDR(td, dirfd, sa);
212 	error = getsock_cap(td, fd, cap_rights_init(&rights, CAP_BIND),
213 	    &fp, NULL, NULL);
214 	if (error != 0)
215 		return (error);
216 	so = fp->f_data;
217 #ifdef KTRACE
218 	if (KTRPOINT(td, KTR_STRUCT))
219 		ktrsockaddr(sa);
220 #endif
221 #ifdef MAC
222 	error = mac_socket_check_bind(td->td_ucred, so, sa);
223 	if (error == 0) {
224 #endif
225 		if (dirfd == AT_FDCWD)
226 			error = sobind(so, sa, td);
227 		else
228 			error = sobindat(dirfd, so, sa, td);
229 #ifdef MAC
230 	}
231 #endif
232 	fdrop(fp, td);
233 	return (error);
234 }
235 
236 /* ARGSUSED */
237 int
sys_bindat(td,uap)238 sys_bindat(td, uap)
239 	struct thread *td;
240 	struct bindat_args /* {
241 		int	fd;
242 		int	s;
243 		caddr_t	name;
244 		int	namelen;
245 	} */ *uap;
246 {
247 	struct sockaddr *sa;
248 	int error;
249 
250 	error = getsockaddr(&sa, uap->name, uap->namelen);
251 	if (error == 0) {
252 		error = kern_bindat(td, uap->fd, uap->s, sa);
253 		free(sa, M_SONAME);
254 	}
255 	return (error);
256 }
257 
258 /* ARGSUSED */
259 int
sys_listen(td,uap)260 sys_listen(td, uap)
261 	struct thread *td;
262 	struct listen_args /* {
263 		int	s;
264 		int	backlog;
265 	} */ *uap;
266 {
267 
268 	return (kern_listen(td, uap->s, uap->backlog));
269 }
270 
271 int
kern_listen(struct thread * td,int s,int backlog)272 kern_listen(struct thread *td, int s, int backlog)
273 {
274 	struct socket *so;
275 	struct file *fp;
276 	cap_rights_t rights;
277 	int error;
278 
279 	AUDIT_ARG_FD(s);
280 	error = getsock_cap(td, s, cap_rights_init(&rights, CAP_LISTEN),
281 	    &fp, NULL, NULL);
282 	if (error == 0) {
283 		so = fp->f_data;
284 #ifdef MAC
285 		error = mac_socket_check_listen(td->td_ucred, so);
286 		if (error == 0)
287 #endif
288 			error = solisten(so, backlog, td);
289 		fdrop(fp, td);
290 	}
291 	return (error);
292 }
293 
294 /*
295  * accept1()
296  */
297 static int
accept1(td,s,uname,anamelen,flags)298 accept1(td, s, uname, anamelen, flags)
299 	struct thread *td;
300 	int s;
301 	struct sockaddr *uname;
302 	socklen_t *anamelen;
303 	int flags;
304 {
305 	struct sockaddr *name;
306 	socklen_t namelen;
307 	struct file *fp;
308 	int error;
309 
310 	if (uname == NULL)
311 		return (kern_accept4(td, s, NULL, NULL, flags, NULL));
312 
313 	error = copyin(anamelen, &namelen, sizeof (namelen));
314 	if (error != 0)
315 		return (error);
316 
317 	error = kern_accept4(td, s, &name, &namelen, flags, &fp);
318 
319 	if (error != 0)
320 		return (error);
321 
322 	if (error == 0 && uname != NULL) {
323 #ifdef COMPAT_OLDSOCK
324 		if (flags & ACCEPT4_COMPAT)
325 			((struct osockaddr *)name)->sa_family =
326 			    name->sa_family;
327 #endif
328 		error = copyout(name, uname, namelen);
329 	}
330 	if (error == 0)
331 		error = copyout(&namelen, anamelen,
332 		    sizeof(namelen));
333 	if (error != 0)
334 		fdclose(td, fp, td->td_retval[0]);
335 	fdrop(fp, td);
336 	free(name, M_SONAME);
337 	return (error);
338 }
339 
340 int
kern_accept(struct thread * td,int s,struct sockaddr ** name,socklen_t * namelen,struct file ** fp)341 kern_accept(struct thread *td, int s, struct sockaddr **name,
342     socklen_t *namelen, struct file **fp)
343 {
344 	return (kern_accept4(td, s, name, namelen, ACCEPT4_INHERIT, fp));
345 }
346 
347 int
kern_accept4(struct thread * td,int s,struct sockaddr ** name,socklen_t * namelen,int flags,struct file ** fp)348 kern_accept4(struct thread *td, int s, struct sockaddr **name,
349     socklen_t *namelen, int flags, struct file **fp)
350 {
351 	struct file *headfp, *nfp = NULL;
352 	struct sockaddr *sa = NULL;
353 	struct socket *head, *so;
354 	struct filecaps fcaps;
355 	cap_rights_t rights;
356 	u_int fflag;
357 	pid_t pgid;
358 	int error, fd, tmp;
359 
360 	if (name != NULL)
361 		*name = NULL;
362 
363 	AUDIT_ARG_FD(s);
364 	error = getsock_cap(td, s, cap_rights_init(&rights, CAP_ACCEPT),
365 	    &headfp, &fflag, &fcaps);
366 	if (error != 0)
367 		return (error);
368 	head = headfp->f_data;
369 	if ((head->so_options & SO_ACCEPTCONN) == 0) {
370 		error = EINVAL;
371 		goto done;
372 	}
373 #ifdef MAC
374 	error = mac_socket_check_accept(td->td_ucred, head);
375 	if (error != 0)
376 		goto done;
377 #endif
378 	error = falloc_caps(td, &nfp, &fd,
379 	    (flags & SOCK_CLOEXEC) ? O_CLOEXEC : 0, &fcaps);
380 	if (error != 0)
381 		goto done;
382 	ACCEPT_LOCK();
383 	if ((head->so_state & SS_NBIO) && TAILQ_EMPTY(&head->so_comp)) {
384 		ACCEPT_UNLOCK();
385 		error = EWOULDBLOCK;
386 		goto noconnection;
387 	}
388 	while (TAILQ_EMPTY(&head->so_comp) && head->so_error == 0) {
389 		if (head->so_rcv.sb_state & SBS_CANTRCVMORE) {
390 			head->so_error = ECONNABORTED;
391 			break;
392 		}
393 		error = msleep(&head->so_timeo, &accept_mtx, PSOCK | PCATCH,
394 		    "accept", 0);
395 		if (error != 0) {
396 			ACCEPT_UNLOCK();
397 			goto noconnection;
398 		}
399 	}
400 	if (head->so_error) {
401 		error = head->so_error;
402 		head->so_error = 0;
403 		ACCEPT_UNLOCK();
404 		goto noconnection;
405 	}
406 	so = TAILQ_FIRST(&head->so_comp);
407 	KASSERT(!(so->so_qstate & SQ_INCOMP), ("accept1: so SQ_INCOMP"));
408 	KASSERT(so->so_qstate & SQ_COMP, ("accept1: so not SQ_COMP"));
409 
410 	/*
411 	 * Before changing the flags on the socket, we have to bump the
412 	 * reference count.  Otherwise, if the protocol calls sofree(),
413 	 * the socket will be released due to a zero refcount.
414 	 */
415 	SOCK_LOCK(so);			/* soref() and so_state update */
416 	soref(so);			/* file descriptor reference */
417 
418 	TAILQ_REMOVE(&head->so_comp, so, so_list);
419 	head->so_qlen--;
420 	if (flags & ACCEPT4_INHERIT)
421 		so->so_state |= (head->so_state & SS_NBIO);
422 	else
423 		so->so_state |= (flags & SOCK_NONBLOCK) ? SS_NBIO : 0;
424 	so->so_qstate &= ~SQ_COMP;
425 	so->so_head = NULL;
426 
427 	SOCK_UNLOCK(so);
428 	ACCEPT_UNLOCK();
429 
430 	/* An extra reference on `nfp' has been held for us by falloc(). */
431 	td->td_retval[0] = fd;
432 
433 	/* connection has been removed from the listen queue */
434 	KNOTE_UNLOCKED(&head->so_rcv.sb_sel.si_note, 0);
435 
436 	if (flags & ACCEPT4_INHERIT) {
437 		pgid = fgetown(&head->so_sigio);
438 		if (pgid != 0)
439 			fsetown(pgid, &so->so_sigio);
440 	} else {
441 		fflag &= ~(FNONBLOCK | FASYNC);
442 		if (flags & SOCK_NONBLOCK)
443 			fflag |= FNONBLOCK;
444 	}
445 
446 	finit(nfp, fflag, DTYPE_SOCKET, so, &socketops);
447 	/* Sync socket nonblocking/async state with file flags */
448 	tmp = fflag & FNONBLOCK;
449 	(void) fo_ioctl(nfp, FIONBIO, &tmp, td->td_ucred, td);
450 	tmp = fflag & FASYNC;
451 	(void) fo_ioctl(nfp, FIOASYNC, &tmp, td->td_ucred, td);
452 	sa = NULL;
453 	error = soaccept(so, &sa);
454 	if (error != 0)
455 		goto noconnection;
456 	if (sa == NULL) {
457 		if (name)
458 			*namelen = 0;
459 		goto done;
460 	}
461 	AUDIT_ARG_SOCKADDR(td, AT_FDCWD, sa);
462 	if (name) {
463 		/* check sa_len before it is destroyed */
464 		if (*namelen > sa->sa_len)
465 			*namelen = sa->sa_len;
466 #ifdef KTRACE
467 		if (KTRPOINT(td, KTR_STRUCT))
468 			ktrsockaddr(sa);
469 #endif
470 		*name = sa;
471 		sa = NULL;
472 	}
473 noconnection:
474 	free(sa, M_SONAME);
475 
476 	/*
477 	 * close the new descriptor, assuming someone hasn't ripped it
478 	 * out from under us.
479 	 */
480 	if (error != 0)
481 		fdclose(td, nfp, fd);
482 
483 	/*
484 	 * Release explicitly held references before returning.  We return
485 	 * a reference on nfp to the caller on success if they request it.
486 	 */
487 done:
488 	if (nfp == NULL)
489 		filecaps_free(&fcaps);
490 	if (fp != NULL) {
491 		if (error == 0) {
492 			*fp = nfp;
493 			nfp = NULL;
494 		} else
495 			*fp = NULL;
496 	}
497 	if (nfp != NULL)
498 		fdrop(nfp, td);
499 	fdrop(headfp, td);
500 	return (error);
501 }
502 
503 int
sys_accept(td,uap)504 sys_accept(td, uap)
505 	struct thread *td;
506 	struct accept_args *uap;
507 {
508 
509 	return (accept1(td, uap->s, uap->name, uap->anamelen, ACCEPT4_INHERIT));
510 }
511 
512 int
sys_accept4(td,uap)513 sys_accept4(td, uap)
514 	struct thread *td;
515 	struct accept4_args *uap;
516 {
517 
518 	if (uap->flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
519 		return (EINVAL);
520 
521 	return (accept1(td, uap->s, uap->name, uap->anamelen, uap->flags));
522 }
523 
524 #ifdef COMPAT_OLDSOCK
525 int
oaccept(td,uap)526 oaccept(td, uap)
527 	struct thread *td;
528 	struct accept_args *uap;
529 {
530 
531 	return (accept1(td, uap->s, uap->name, uap->anamelen,
532 	    ACCEPT4_INHERIT | ACCEPT4_COMPAT));
533 }
534 #endif /* COMPAT_OLDSOCK */
535 
536 /* ARGSUSED */
537 int
sys_connect(td,uap)538 sys_connect(td, uap)
539 	struct thread *td;
540 	struct connect_args /* {
541 		int	s;
542 		caddr_t	name;
543 		int	namelen;
544 	} */ *uap;
545 {
546 	struct sockaddr *sa;
547 	int error;
548 
549 	error = getsockaddr(&sa, uap->name, uap->namelen);
550 	if (error == 0) {
551 		error = kern_connectat(td, AT_FDCWD, uap->s, sa);
552 		free(sa, M_SONAME);
553 	}
554 	return (error);
555 }
556 
557 int
kern_connectat(struct thread * td,int dirfd,int fd,struct sockaddr * sa)558 kern_connectat(struct thread *td, int dirfd, int fd, struct sockaddr *sa)
559 {
560 	struct socket *so;
561 	struct file *fp;
562 	cap_rights_t rights;
563 	int error, interrupted = 0;
564 
565 	AUDIT_ARG_FD(fd);
566 	AUDIT_ARG_SOCKADDR(td, dirfd, sa);
567 	error = getsock_cap(td, fd, cap_rights_init(&rights, CAP_CONNECT),
568 	    &fp, NULL, NULL);
569 	if (error != 0)
570 		return (error);
571 	so = fp->f_data;
572 	if (so->so_state & SS_ISCONNECTING) {
573 		error = EALREADY;
574 		goto done1;
575 	}
576 #ifdef KTRACE
577 	if (KTRPOINT(td, KTR_STRUCT))
578 		ktrsockaddr(sa);
579 #endif
580 #ifdef MAC
581 	error = mac_socket_check_connect(td->td_ucred, so, sa);
582 	if (error != 0)
583 		goto bad;
584 #endif
585 	if (dirfd == AT_FDCWD)
586 		error = soconnect(so, sa, td);
587 	else
588 		error = soconnectat(dirfd, so, sa, td);
589 	if (error != 0)
590 		goto bad;
591 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
592 		error = EINPROGRESS;
593 		goto done1;
594 	}
595 	SOCK_LOCK(so);
596 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
597 		error = msleep(&so->so_timeo, SOCK_MTX(so), PSOCK | PCATCH,
598 		    "connec", 0);
599 		if (error != 0) {
600 			if (error == EINTR || error == ERESTART)
601 				interrupted = 1;
602 			break;
603 		}
604 	}
605 	if (error == 0) {
606 		error = so->so_error;
607 		so->so_error = 0;
608 	}
609 	SOCK_UNLOCK(so);
610 bad:
611 	if (!interrupted)
612 		so->so_state &= ~SS_ISCONNECTING;
613 	if (error == ERESTART)
614 		error = EINTR;
615 done1:
616 	fdrop(fp, td);
617 	return (error);
618 }
619 
620 /* ARGSUSED */
621 int
sys_connectat(td,uap)622 sys_connectat(td, uap)
623 	struct thread *td;
624 	struct connectat_args /* {
625 		int	fd;
626 		int	s;
627 		caddr_t	name;
628 		int	namelen;
629 	} */ *uap;
630 {
631 	struct sockaddr *sa;
632 	int error;
633 
634 	error = getsockaddr(&sa, uap->name, uap->namelen);
635 	if (error == 0) {
636 		error = kern_connectat(td, uap->fd, uap->s, sa);
637 		free(sa, M_SONAME);
638 	}
639 	return (error);
640 }
641 
642 int
kern_socketpair(struct thread * td,int domain,int type,int protocol,int * rsv)643 kern_socketpair(struct thread *td, int domain, int type, int protocol,
644     int *rsv)
645 {
646 	struct file *fp1, *fp2;
647 	struct socket *so1, *so2;
648 	int fd, error, oflag, fflag;
649 
650 	AUDIT_ARG_SOCKET(domain, type, protocol);
651 
652 	oflag = 0;
653 	fflag = 0;
654 	if ((type & SOCK_CLOEXEC) != 0) {
655 		type &= ~SOCK_CLOEXEC;
656 		oflag |= O_CLOEXEC;
657 	}
658 	if ((type & SOCK_NONBLOCK) != 0) {
659 		type &= ~SOCK_NONBLOCK;
660 		fflag |= FNONBLOCK;
661 	}
662 #ifdef MAC
663 	/* We might want to have a separate check for socket pairs. */
664 	error = mac_socket_check_create(td->td_ucred, domain, type,
665 	    protocol);
666 	if (error != 0)
667 		return (error);
668 #endif
669 	error = socreate(domain, &so1, type, protocol, td->td_ucred, td);
670 	if (error != 0)
671 		return (error);
672 	error = socreate(domain, &so2, type, protocol, td->td_ucred, td);
673 	if (error != 0)
674 		goto free1;
675 	/* On success extra reference to `fp1' and 'fp2' is set by falloc. */
676 	error = falloc(td, &fp1, &fd, oflag);
677 	if (error != 0)
678 		goto free2;
679 	rsv[0] = fd;
680 	fp1->f_data = so1;	/* so1 already has ref count */
681 	error = falloc(td, &fp2, &fd, oflag);
682 	if (error != 0)
683 		goto free3;
684 	fp2->f_data = so2;	/* so2 already has ref count */
685 	rsv[1] = fd;
686 	error = soconnect2(so1, so2);
687 	if (error != 0)
688 		goto free4;
689 	if (type == SOCK_DGRAM) {
690 		/*
691 		 * Datagram socket connection is asymmetric.
692 		 */
693 		 error = soconnect2(so2, so1);
694 		 if (error != 0)
695 			goto free4;
696 	} else if (so1->so_proto->pr_flags & PR_CONNREQUIRED) {
697 		struct unpcb *unp, *unp2;
698 		unp = sotounpcb(so1);
699 		unp2 = sotounpcb(so2);
700 		/*
701 		 * No need to lock the unps, because the sockets are brand-new.
702 		 * No other threads can be using them yet
703 		 */
704 		unp_copy_peercred(td, unp, unp2, unp);
705 	}
706 	finit(fp1, FREAD | FWRITE | fflag, DTYPE_SOCKET, fp1->f_data,
707 	    &socketops);
708 	finit(fp2, FREAD | FWRITE | fflag, DTYPE_SOCKET, fp2->f_data,
709 	    &socketops);
710 	if ((fflag & FNONBLOCK) != 0) {
711 		(void) fo_ioctl(fp1, FIONBIO, &fflag, td->td_ucred, td);
712 		(void) fo_ioctl(fp2, FIONBIO, &fflag, td->td_ucred, td);
713 	}
714 	fdrop(fp1, td);
715 	fdrop(fp2, td);
716 	return (0);
717 free4:
718 	fdclose(td, fp2, rsv[1]);
719 	fdrop(fp2, td);
720 free3:
721 	fdclose(td, fp1, rsv[0]);
722 	fdrop(fp1, td);
723 free2:
724 	if (so2 != NULL)
725 		(void)soclose(so2);
726 free1:
727 	if (so1 != NULL)
728 		(void)soclose(so1);
729 	return (error);
730 }
731 
732 int
sys_socketpair(struct thread * td,struct socketpair_args * uap)733 sys_socketpair(struct thread *td, struct socketpair_args *uap)
734 {
735 	int error, sv[2];
736 
737 	error = kern_socketpair(td, uap->domain, uap->type,
738 	    uap->protocol, sv);
739 	if (error != 0)
740 		return (error);
741 	error = copyout(sv, uap->rsv, 2 * sizeof(int));
742 	if (error != 0) {
743 		(void)kern_close(td, sv[0]);
744 		(void)kern_close(td, sv[1]);
745 	}
746 	return (error);
747 }
748 
749 static int
sendit(td,s,mp,flags)750 sendit(td, s, mp, flags)
751 	struct thread *td;
752 	int s;
753 	struct msghdr *mp;
754 	int flags;
755 {
756 	struct mbuf *control;
757 	struct sockaddr *to;
758 	int error;
759 
760 #ifdef CAPABILITY_MODE
761 	if (IN_CAPABILITY_MODE(td) && (mp->msg_name != NULL))
762 		return (ECAPMODE);
763 #endif
764 
765 	if (mp->msg_name != NULL) {
766 		error = getsockaddr(&to, mp->msg_name, mp->msg_namelen);
767 		if (error != 0) {
768 			to = NULL;
769 			goto bad;
770 		}
771 		mp->msg_name = to;
772 	} else {
773 		to = NULL;
774 	}
775 
776 	if (mp->msg_control) {
777 		if (mp->msg_controllen < sizeof(struct cmsghdr)
778 #ifdef COMPAT_OLDSOCK
779 		    && mp->msg_flags != MSG_COMPAT
780 #endif
781 		) {
782 			error = EINVAL;
783 			goto bad;
784 		}
785 		error = sockargs(&control, mp->msg_control,
786 		    mp->msg_controllen, MT_CONTROL);
787 		if (error != 0)
788 			goto bad;
789 #ifdef COMPAT_OLDSOCK
790 		if (mp->msg_flags == MSG_COMPAT) {
791 			struct cmsghdr *cm;
792 
793 			M_PREPEND(control, sizeof(*cm), M_WAITOK);
794 			cm = mtod(control, struct cmsghdr *);
795 			cm->cmsg_len = control->m_len;
796 			cm->cmsg_level = SOL_SOCKET;
797 			cm->cmsg_type = SCM_RIGHTS;
798 		}
799 #endif
800 	} else {
801 		control = NULL;
802 	}
803 
804 	error = kern_sendit(td, s, mp, flags, control, UIO_USERSPACE);
805 
806 bad:
807 	free(to, M_SONAME);
808 	return (error);
809 }
810 
811 int
kern_sendit(td,s,mp,flags,control,segflg)812 kern_sendit(td, s, mp, flags, control, segflg)
813 	struct thread *td;
814 	int s;
815 	struct msghdr *mp;
816 	int flags;
817 	struct mbuf *control;
818 	enum uio_seg segflg;
819 {
820 	struct file *fp;
821 	struct uio auio;
822 	struct iovec *iov;
823 	struct socket *so;
824 	cap_rights_t rights;
825 #ifdef KTRACE
826 	struct uio *ktruio = NULL;
827 #endif
828 	ssize_t len;
829 	int i, error;
830 
831 	AUDIT_ARG_FD(s);
832 	cap_rights_init(&rights, CAP_SEND);
833 	if (mp->msg_name != NULL) {
834 		AUDIT_ARG_SOCKADDR(td, AT_FDCWD, mp->msg_name);
835 		cap_rights_set(&rights, CAP_CONNECT);
836 	}
837 	error = getsock_cap(td, s, &rights, &fp, NULL, NULL);
838 	if (error != 0) {
839 		m_freem(control);
840 		return (error);
841 	}
842 	so = (struct socket *)fp->f_data;
843 
844 #ifdef KTRACE
845 	if (mp->msg_name != NULL && KTRPOINT(td, KTR_STRUCT))
846 		ktrsockaddr(mp->msg_name);
847 #endif
848 #ifdef MAC
849 	if (mp->msg_name != NULL) {
850 		error = mac_socket_check_connect(td->td_ucred, so,
851 		    mp->msg_name);
852 		if (error != 0) {
853 			m_freem(control);
854 			goto bad;
855 		}
856 	}
857 	error = mac_socket_check_send(td->td_ucred, so);
858 	if (error != 0) {
859 		m_freem(control);
860 		goto bad;
861 	}
862 #endif
863 
864 	auio.uio_iov = mp->msg_iov;
865 	auio.uio_iovcnt = mp->msg_iovlen;
866 	auio.uio_segflg = segflg;
867 	auio.uio_rw = UIO_WRITE;
868 	auio.uio_td = td;
869 	auio.uio_offset = 0;			/* XXX */
870 	auio.uio_resid = 0;
871 	iov = mp->msg_iov;
872 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
873 		if ((auio.uio_resid += iov->iov_len) < 0) {
874 			error = EINVAL;
875 			m_freem(control);
876 			goto bad;
877 		}
878 	}
879 #ifdef KTRACE
880 	if (KTRPOINT(td, KTR_GENIO))
881 		ktruio = cloneuio(&auio);
882 #endif
883 	len = auio.uio_resid;
884 	error = sosend(so, mp->msg_name, &auio, 0, control, flags, td);
885 	if (error != 0) {
886 		if (auio.uio_resid != len && (error == ERESTART ||
887 		    error == EINTR || error == EWOULDBLOCK))
888 			error = 0;
889 		/* Generation of SIGPIPE can be controlled per socket */
890 		if (error == EPIPE && !(so->so_options & SO_NOSIGPIPE) &&
891 		    !(flags & MSG_NOSIGNAL)) {
892 			PROC_LOCK(td->td_proc);
893 			tdsignal(td, SIGPIPE);
894 			PROC_UNLOCK(td->td_proc);
895 		}
896 	}
897 	if (error == 0)
898 		td->td_retval[0] = len - auio.uio_resid;
899 #ifdef KTRACE
900 	if (ktruio != NULL) {
901 		ktruio->uio_resid = td->td_retval[0];
902 		ktrgenio(s, UIO_WRITE, ktruio, error);
903 	}
904 #endif
905 bad:
906 	fdrop(fp, td);
907 	return (error);
908 }
909 
910 int
sys_sendto(td,uap)911 sys_sendto(td, uap)
912 	struct thread *td;
913 	struct sendto_args /* {
914 		int	s;
915 		caddr_t	buf;
916 		size_t	len;
917 		int	flags;
918 		caddr_t	to;
919 		int	tolen;
920 	} */ *uap;
921 {
922 	struct msghdr msg;
923 	struct iovec aiov;
924 
925 	msg.msg_name = uap->to;
926 	msg.msg_namelen = uap->tolen;
927 	msg.msg_iov = &aiov;
928 	msg.msg_iovlen = 1;
929 	msg.msg_control = 0;
930 #ifdef COMPAT_OLDSOCK
931 	msg.msg_flags = 0;
932 #endif
933 	aiov.iov_base = uap->buf;
934 	aiov.iov_len = uap->len;
935 	return (sendit(td, uap->s, &msg, uap->flags));
936 }
937 
938 #ifdef COMPAT_OLDSOCK
939 int
osend(td,uap)940 osend(td, uap)
941 	struct thread *td;
942 	struct osend_args /* {
943 		int	s;
944 		caddr_t	buf;
945 		int	len;
946 		int	flags;
947 	} */ *uap;
948 {
949 	struct msghdr msg;
950 	struct iovec aiov;
951 
952 	msg.msg_name = 0;
953 	msg.msg_namelen = 0;
954 	msg.msg_iov = &aiov;
955 	msg.msg_iovlen = 1;
956 	aiov.iov_base = uap->buf;
957 	aiov.iov_len = uap->len;
958 	msg.msg_control = 0;
959 	msg.msg_flags = 0;
960 	return (sendit(td, uap->s, &msg, uap->flags));
961 }
962 
963 int
osendmsg(td,uap)964 osendmsg(td, uap)
965 	struct thread *td;
966 	struct osendmsg_args /* {
967 		int	s;
968 		caddr_t	msg;
969 		int	flags;
970 	} */ *uap;
971 {
972 	struct msghdr msg;
973 	struct iovec *iov;
974 	int error;
975 
976 	error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
977 	if (error != 0)
978 		return (error);
979 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
980 	if (error != 0)
981 		return (error);
982 	msg.msg_iov = iov;
983 	msg.msg_flags = MSG_COMPAT;
984 	error = sendit(td, uap->s, &msg, uap->flags);
985 	free(iov, M_IOV);
986 	return (error);
987 }
988 #endif
989 
990 int
sys_sendmsg(td,uap)991 sys_sendmsg(td, uap)
992 	struct thread *td;
993 	struct sendmsg_args /* {
994 		int	s;
995 		caddr_t	msg;
996 		int	flags;
997 	} */ *uap;
998 {
999 	struct msghdr msg;
1000 	struct iovec *iov;
1001 	int error;
1002 
1003 	error = copyin(uap->msg, &msg, sizeof (msg));
1004 	if (error != 0)
1005 		return (error);
1006 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1007 	if (error != 0)
1008 		return (error);
1009 	msg.msg_iov = iov;
1010 #ifdef COMPAT_OLDSOCK
1011 	msg.msg_flags = 0;
1012 #endif
1013 	error = sendit(td, uap->s, &msg, uap->flags);
1014 	free(iov, M_IOV);
1015 	return (error);
1016 }
1017 
1018 int
kern_recvit(td,s,mp,fromseg,controlp)1019 kern_recvit(td, s, mp, fromseg, controlp)
1020 	struct thread *td;
1021 	int s;
1022 	struct msghdr *mp;
1023 	enum uio_seg fromseg;
1024 	struct mbuf **controlp;
1025 {
1026 	struct uio auio;
1027 	struct iovec *iov;
1028 	struct mbuf *control, *m;
1029 	caddr_t ctlbuf;
1030 	struct file *fp;
1031 	struct socket *so;
1032 	struct sockaddr *fromsa = NULL;
1033 	cap_rights_t rights;
1034 #ifdef KTRACE
1035 	struct uio *ktruio = NULL;
1036 #endif
1037 	ssize_t len;
1038 	int error, i;
1039 
1040 	if (controlp != NULL)
1041 		*controlp = NULL;
1042 
1043 	AUDIT_ARG_FD(s);
1044 	error = getsock_cap(td, s, cap_rights_init(&rights, CAP_RECV),
1045 	    &fp, NULL, NULL);
1046 	if (error != 0)
1047 		return (error);
1048 	so = fp->f_data;
1049 
1050 #ifdef MAC
1051 	error = mac_socket_check_receive(td->td_ucred, so);
1052 	if (error != 0) {
1053 		fdrop(fp, td);
1054 		return (error);
1055 	}
1056 #endif
1057 
1058 	auio.uio_iov = mp->msg_iov;
1059 	auio.uio_iovcnt = mp->msg_iovlen;
1060 	auio.uio_segflg = UIO_USERSPACE;
1061 	auio.uio_rw = UIO_READ;
1062 	auio.uio_td = td;
1063 	auio.uio_offset = 0;			/* XXX */
1064 	auio.uio_resid = 0;
1065 	iov = mp->msg_iov;
1066 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
1067 		if ((auio.uio_resid += iov->iov_len) < 0) {
1068 			fdrop(fp, td);
1069 			return (EINVAL);
1070 		}
1071 	}
1072 #ifdef KTRACE
1073 	if (KTRPOINT(td, KTR_GENIO))
1074 		ktruio = cloneuio(&auio);
1075 #endif
1076 	control = NULL;
1077 	len = auio.uio_resid;
1078 	error = soreceive(so, &fromsa, &auio, NULL,
1079 	    (mp->msg_control || controlp) ? &control : NULL,
1080 	    &mp->msg_flags);
1081 	if (error != 0) {
1082 		if (auio.uio_resid != len && (error == ERESTART ||
1083 		    error == EINTR || error == EWOULDBLOCK))
1084 			error = 0;
1085 	}
1086 	if (fromsa != NULL)
1087 		AUDIT_ARG_SOCKADDR(td, AT_FDCWD, fromsa);
1088 #ifdef KTRACE
1089 	if (ktruio != NULL) {
1090 		ktruio->uio_resid = len - auio.uio_resid;
1091 		ktrgenio(s, UIO_READ, ktruio, error);
1092 	}
1093 #endif
1094 	if (error != 0)
1095 		goto out;
1096 	td->td_retval[0] = len - auio.uio_resid;
1097 	if (mp->msg_name) {
1098 		len = mp->msg_namelen;
1099 		if (len <= 0 || fromsa == NULL)
1100 			len = 0;
1101 		else {
1102 			/* save sa_len before it is destroyed by MSG_COMPAT */
1103 			len = MIN(len, fromsa->sa_len);
1104 #ifdef COMPAT_OLDSOCK
1105 			if (mp->msg_flags & MSG_COMPAT)
1106 				((struct osockaddr *)fromsa)->sa_family =
1107 				    fromsa->sa_family;
1108 #endif
1109 			if (fromseg == UIO_USERSPACE) {
1110 				error = copyout(fromsa, mp->msg_name,
1111 				    (unsigned)len);
1112 				if (error != 0)
1113 					goto out;
1114 			} else
1115 				bcopy(fromsa, mp->msg_name, len);
1116 		}
1117 		mp->msg_namelen = len;
1118 	}
1119 	if (mp->msg_control && controlp == NULL) {
1120 #ifdef COMPAT_OLDSOCK
1121 		/*
1122 		 * We assume that old recvmsg calls won't receive access
1123 		 * rights and other control info, esp. as control info
1124 		 * is always optional and those options didn't exist in 4.3.
1125 		 * If we receive rights, trim the cmsghdr; anything else
1126 		 * is tossed.
1127 		 */
1128 		if (control && mp->msg_flags & MSG_COMPAT) {
1129 			if (mtod(control, struct cmsghdr *)->cmsg_level !=
1130 			    SOL_SOCKET ||
1131 			    mtod(control, struct cmsghdr *)->cmsg_type !=
1132 			    SCM_RIGHTS) {
1133 				mp->msg_controllen = 0;
1134 				goto out;
1135 			}
1136 			control->m_len -= sizeof (struct cmsghdr);
1137 			control->m_data += sizeof (struct cmsghdr);
1138 		}
1139 #endif
1140 		ctlbuf = mp->msg_control;
1141 		len = mp->msg_controllen;
1142 		mp->msg_controllen = 0;
1143 		for (m = control; m != NULL && len >= m->m_len; m = m->m_next) {
1144 			if ((error = copyout(mtod(m, caddr_t), ctlbuf,
1145 			    m->m_len)) != 0)
1146 				goto out;
1147 
1148 			ctlbuf += m->m_len;
1149 			len -= m->m_len;
1150 			mp->msg_controllen += m->m_len;
1151 		}
1152 		if (m != NULL) {
1153 			mp->msg_flags |= MSG_CTRUNC;
1154 			m_dispose_extcontrolm(m);
1155 		}
1156 	}
1157 out:
1158 	fdrop(fp, td);
1159 #ifdef KTRACE
1160 	if (fromsa && KTRPOINT(td, KTR_STRUCT))
1161 		ktrsockaddr(fromsa);
1162 #endif
1163 	free(fromsa, M_SONAME);
1164 
1165 	if (error == 0 && controlp != NULL)
1166 		*controlp = control;
1167 	else if (control != NULL) {
1168 		if (error != 0)
1169 			m_dispose_extcontrolm(control);
1170 		m_freem(control);
1171 	}
1172 
1173 	return (error);
1174 }
1175 
1176 static int
recvit(td,s,mp,namelenp)1177 recvit(td, s, mp, namelenp)
1178 	struct thread *td;
1179 	int s;
1180 	struct msghdr *mp;
1181 	void *namelenp;
1182 {
1183 	int error;
1184 
1185 	error = kern_recvit(td, s, mp, UIO_USERSPACE, NULL);
1186 	if (error != 0)
1187 		return (error);
1188 	if (namelenp != NULL) {
1189 		error = copyout(&mp->msg_namelen, namelenp, sizeof (socklen_t));
1190 #ifdef COMPAT_OLDSOCK
1191 		if (mp->msg_flags & MSG_COMPAT)
1192 			error = 0;	/* old recvfrom didn't check */
1193 #endif
1194 	}
1195 	return (error);
1196 }
1197 
1198 int
sys_recvfrom(td,uap)1199 sys_recvfrom(td, uap)
1200 	struct thread *td;
1201 	struct recvfrom_args /* {
1202 		int	s;
1203 		caddr_t	buf;
1204 		size_t	len;
1205 		int	flags;
1206 		struct sockaddr * __restrict	from;
1207 		socklen_t * __restrict fromlenaddr;
1208 	} */ *uap;
1209 {
1210 	struct msghdr msg;
1211 	struct iovec aiov;
1212 	int error;
1213 
1214 	if (uap->fromlenaddr) {
1215 		error = copyin(uap->fromlenaddr,
1216 		    &msg.msg_namelen, sizeof (msg.msg_namelen));
1217 		if (error != 0)
1218 			goto done2;
1219 	} else {
1220 		msg.msg_namelen = 0;
1221 	}
1222 	msg.msg_name = uap->from;
1223 	msg.msg_iov = &aiov;
1224 	msg.msg_iovlen = 1;
1225 	aiov.iov_base = uap->buf;
1226 	aiov.iov_len = uap->len;
1227 	msg.msg_control = 0;
1228 	msg.msg_flags = uap->flags;
1229 	error = recvit(td, uap->s, &msg, uap->fromlenaddr);
1230 done2:
1231 	return (error);
1232 }
1233 
1234 #ifdef COMPAT_OLDSOCK
1235 int
orecvfrom(td,uap)1236 orecvfrom(td, uap)
1237 	struct thread *td;
1238 	struct recvfrom_args *uap;
1239 {
1240 
1241 	uap->flags |= MSG_COMPAT;
1242 	return (sys_recvfrom(td, uap));
1243 }
1244 #endif
1245 
1246 #ifdef COMPAT_OLDSOCK
1247 int
orecv(td,uap)1248 orecv(td, uap)
1249 	struct thread *td;
1250 	struct orecv_args /* {
1251 		int	s;
1252 		caddr_t	buf;
1253 		int	len;
1254 		int	flags;
1255 	} */ *uap;
1256 {
1257 	struct msghdr msg;
1258 	struct iovec aiov;
1259 
1260 	msg.msg_name = 0;
1261 	msg.msg_namelen = 0;
1262 	msg.msg_iov = &aiov;
1263 	msg.msg_iovlen = 1;
1264 	aiov.iov_base = uap->buf;
1265 	aiov.iov_len = uap->len;
1266 	msg.msg_control = 0;
1267 	msg.msg_flags = uap->flags;
1268 	return (recvit(td, uap->s, &msg, NULL));
1269 }
1270 
1271 /*
1272  * Old recvmsg.  This code takes advantage of the fact that the old msghdr
1273  * overlays the new one, missing only the flags, and with the (old) access
1274  * rights where the control fields are now.
1275  */
1276 int
orecvmsg(td,uap)1277 orecvmsg(td, uap)
1278 	struct thread *td;
1279 	struct orecvmsg_args /* {
1280 		int	s;
1281 		struct	omsghdr *msg;
1282 		int	flags;
1283 	} */ *uap;
1284 {
1285 	struct msghdr msg;
1286 	struct iovec *iov;
1287 	int error;
1288 
1289 	error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
1290 	if (error != 0)
1291 		return (error);
1292 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1293 	if (error != 0)
1294 		return (error);
1295 	msg.msg_flags = uap->flags | MSG_COMPAT;
1296 	msg.msg_iov = iov;
1297 	error = recvit(td, uap->s, &msg, &uap->msg->msg_namelen);
1298 	if (msg.msg_controllen && error == 0)
1299 		error = copyout(&msg.msg_controllen,
1300 		    &uap->msg->msg_accrightslen, sizeof (int));
1301 	free(iov, M_IOV);
1302 	return (error);
1303 }
1304 #endif
1305 
1306 int
sys_recvmsg(td,uap)1307 sys_recvmsg(td, uap)
1308 	struct thread *td;
1309 	struct recvmsg_args /* {
1310 		int	s;
1311 		struct	msghdr *msg;
1312 		int	flags;
1313 	} */ *uap;
1314 {
1315 	struct msghdr msg;
1316 	struct iovec *uiov, *iov;
1317 	int error;
1318 
1319 	error = copyin(uap->msg, &msg, sizeof (msg));
1320 	if (error != 0)
1321 		return (error);
1322 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1323 	if (error != 0)
1324 		return (error);
1325 	msg.msg_flags = uap->flags;
1326 #ifdef COMPAT_OLDSOCK
1327 	msg.msg_flags &= ~MSG_COMPAT;
1328 #endif
1329 	uiov = msg.msg_iov;
1330 	msg.msg_iov = iov;
1331 	error = recvit(td, uap->s, &msg, NULL);
1332 	if (error == 0) {
1333 		msg.msg_iov = uiov;
1334 		error = copyout(&msg, uap->msg, sizeof(msg));
1335 	}
1336 	free(iov, M_IOV);
1337 	return (error);
1338 }
1339 
1340 /* ARGSUSED */
1341 int
sys_shutdown(td,uap)1342 sys_shutdown(td, uap)
1343 	struct thread *td;
1344 	struct shutdown_args /* {
1345 		int	s;
1346 		int	how;
1347 	} */ *uap;
1348 {
1349 
1350 	return (kern_shutdown(td, uap->s, uap->how));
1351 }
1352 
1353 int
kern_shutdown(struct thread * td,int s,int how)1354 kern_shutdown(struct thread *td, int s, int how)
1355 {
1356 	struct socket *so;
1357 	struct file *fp;
1358 	cap_rights_t rights;
1359 	int error;
1360 
1361 	AUDIT_ARG_FD(s);
1362 	error = getsock_cap(td, s, cap_rights_init(&rights, CAP_SHUTDOWN),
1363 	    &fp, NULL, NULL);
1364 	if (error == 0) {
1365 		so = fp->f_data;
1366 		error = soshutdown(so, how);
1367 		/*
1368 		 * Previous versions did not return ENOTCONN, but 0 in
1369 		 * case the socket was not connected. Some important
1370 		 * programs like syslogd up to r279016, 2015-02-19,
1371 		 * still depend on this behavior.
1372 		 */
1373 		if (error == ENOTCONN &&
1374 		    td->td_proc->p_osrel < P_OSREL_SHUTDOWN_ENOTCONN)
1375 			error = 0;
1376 		fdrop(fp, td);
1377 	}
1378 	return (error);
1379 }
1380 
1381 /* ARGSUSED */
1382 int
sys_setsockopt(td,uap)1383 sys_setsockopt(td, uap)
1384 	struct thread *td;
1385 	struct setsockopt_args /* {
1386 		int	s;
1387 		int	level;
1388 		int	name;
1389 		caddr_t	val;
1390 		int	valsize;
1391 	} */ *uap;
1392 {
1393 
1394 	return (kern_setsockopt(td, uap->s, uap->level, uap->name,
1395 	    uap->val, UIO_USERSPACE, uap->valsize));
1396 }
1397 
1398 int
kern_setsockopt(td,s,level,name,val,valseg,valsize)1399 kern_setsockopt(td, s, level, name, val, valseg, valsize)
1400 	struct thread *td;
1401 	int s;
1402 	int level;
1403 	int name;
1404 	void *val;
1405 	enum uio_seg valseg;
1406 	socklen_t valsize;
1407 {
1408 	struct socket *so;
1409 	struct file *fp;
1410 	struct sockopt sopt;
1411 	cap_rights_t rights;
1412 	int error;
1413 
1414 	if (val == NULL && valsize != 0)
1415 		return (EFAULT);
1416 	if ((int)valsize < 0)
1417 		return (EINVAL);
1418 
1419 	sopt.sopt_dir = SOPT_SET;
1420 	sopt.sopt_level = level;
1421 	sopt.sopt_name = name;
1422 	sopt.sopt_val = val;
1423 	sopt.sopt_valsize = valsize;
1424 	switch (valseg) {
1425 	case UIO_USERSPACE:
1426 		sopt.sopt_td = td;
1427 		break;
1428 	case UIO_SYSSPACE:
1429 		sopt.sopt_td = NULL;
1430 		break;
1431 	default:
1432 		panic("kern_setsockopt called with bad valseg");
1433 	}
1434 
1435 	AUDIT_ARG_FD(s);
1436 	error = getsock_cap(td, s, cap_rights_init(&rights, CAP_SETSOCKOPT),
1437 	    &fp, NULL, NULL);
1438 	if (error == 0) {
1439 		so = fp->f_data;
1440 		error = sosetopt(so, &sopt);
1441 		fdrop(fp, td);
1442 	}
1443 	return(error);
1444 }
1445 
1446 /* ARGSUSED */
1447 int
sys_getsockopt(td,uap)1448 sys_getsockopt(td, uap)
1449 	struct thread *td;
1450 	struct getsockopt_args /* {
1451 		int	s;
1452 		int	level;
1453 		int	name;
1454 		void * __restrict	val;
1455 		socklen_t * __restrict avalsize;
1456 	} */ *uap;
1457 {
1458 	socklen_t valsize;
1459 	int error;
1460 
1461 	if (uap->val) {
1462 		error = copyin(uap->avalsize, &valsize, sizeof (valsize));
1463 		if (error != 0)
1464 			return (error);
1465 	}
1466 
1467 	error = kern_getsockopt(td, uap->s, uap->level, uap->name,
1468 	    uap->val, UIO_USERSPACE, &valsize);
1469 
1470 	if (error == 0)
1471 		error = copyout(&valsize, uap->avalsize, sizeof (valsize));
1472 	return (error);
1473 }
1474 
1475 /*
1476  * Kernel version of getsockopt.
1477  * optval can be a userland or userspace. optlen is always a kernel pointer.
1478  */
1479 int
kern_getsockopt(td,s,level,name,val,valseg,valsize)1480 kern_getsockopt(td, s, level, name, val, valseg, valsize)
1481 	struct thread *td;
1482 	int s;
1483 	int level;
1484 	int name;
1485 	void *val;
1486 	enum uio_seg valseg;
1487 	socklen_t *valsize;
1488 {
1489 	struct socket *so;
1490 	struct file *fp;
1491 	struct sockopt sopt;
1492 	cap_rights_t rights;
1493 	int error;
1494 
1495 	if (val == NULL)
1496 		*valsize = 0;
1497 	if ((int)*valsize < 0)
1498 		return (EINVAL);
1499 
1500 	sopt.sopt_dir = SOPT_GET;
1501 	sopt.sopt_level = level;
1502 	sopt.sopt_name = name;
1503 	sopt.sopt_val = val;
1504 	sopt.sopt_valsize = (size_t)*valsize; /* checked non-negative above */
1505 	switch (valseg) {
1506 	case UIO_USERSPACE:
1507 		sopt.sopt_td = td;
1508 		break;
1509 	case UIO_SYSSPACE:
1510 		sopt.sopt_td = NULL;
1511 		break;
1512 	default:
1513 		panic("kern_getsockopt called with bad valseg");
1514 	}
1515 
1516 	AUDIT_ARG_FD(s);
1517 	error = getsock_cap(td, s, cap_rights_init(&rights, CAP_GETSOCKOPT),
1518 	    &fp, NULL, NULL);
1519 	if (error == 0) {
1520 		so = fp->f_data;
1521 		error = sogetopt(so, &sopt);
1522 		*valsize = sopt.sopt_valsize;
1523 		fdrop(fp, td);
1524 	}
1525 	return (error);
1526 }
1527 
1528 /*
1529  * getsockname1() - Get socket name.
1530  */
1531 /* ARGSUSED */
1532 static int
getsockname1(td,uap,compat)1533 getsockname1(td, uap, compat)
1534 	struct thread *td;
1535 	struct getsockname_args /* {
1536 		int	fdes;
1537 		struct sockaddr * __restrict asa;
1538 		socklen_t * __restrict alen;
1539 	} */ *uap;
1540 	int compat;
1541 {
1542 	struct sockaddr *sa;
1543 	socklen_t len;
1544 	int error;
1545 
1546 	error = copyin(uap->alen, &len, sizeof(len));
1547 	if (error != 0)
1548 		return (error);
1549 
1550 	error = kern_getsockname(td, uap->fdes, &sa, &len);
1551 	if (error != 0)
1552 		return (error);
1553 
1554 	if (len != 0) {
1555 #ifdef COMPAT_OLDSOCK
1556 		if (compat)
1557 			((struct osockaddr *)sa)->sa_family = sa->sa_family;
1558 #endif
1559 		error = copyout(sa, uap->asa, (u_int)len);
1560 	}
1561 	free(sa, M_SONAME);
1562 	if (error == 0)
1563 		error = copyout(&len, uap->alen, sizeof(len));
1564 	return (error);
1565 }
1566 
1567 int
kern_getsockname(struct thread * td,int fd,struct sockaddr ** sa,socklen_t * alen)1568 kern_getsockname(struct thread *td, int fd, struct sockaddr **sa,
1569     socklen_t *alen)
1570 {
1571 	struct socket *so;
1572 	struct file *fp;
1573 	cap_rights_t rights;
1574 	socklen_t len;
1575 	int error;
1576 
1577 	AUDIT_ARG_FD(fd);
1578 	error = getsock_cap(td, fd, cap_rights_init(&rights, CAP_GETSOCKNAME),
1579 	    &fp, NULL, NULL);
1580 	if (error != 0)
1581 		return (error);
1582 	so = fp->f_data;
1583 	*sa = NULL;
1584 	CURVNET_SET(so->so_vnet);
1585 	error = (*so->so_proto->pr_usrreqs->pru_sockaddr)(so, sa);
1586 	CURVNET_RESTORE();
1587 	if (error != 0)
1588 		goto bad;
1589 	if (*sa == NULL)
1590 		len = 0;
1591 	else
1592 		len = MIN(*alen, (*sa)->sa_len);
1593 	*alen = len;
1594 #ifdef KTRACE
1595 	if (KTRPOINT(td, KTR_STRUCT))
1596 		ktrsockaddr(*sa);
1597 #endif
1598 bad:
1599 	fdrop(fp, td);
1600 	if (error != 0 && *sa != NULL) {
1601 		free(*sa, M_SONAME);
1602 		*sa = NULL;
1603 	}
1604 	return (error);
1605 }
1606 
1607 int
sys_getsockname(td,uap)1608 sys_getsockname(td, uap)
1609 	struct thread *td;
1610 	struct getsockname_args *uap;
1611 {
1612 
1613 	return (getsockname1(td, uap, 0));
1614 }
1615 
1616 #ifdef COMPAT_OLDSOCK
1617 int
ogetsockname(td,uap)1618 ogetsockname(td, uap)
1619 	struct thread *td;
1620 	struct getsockname_args *uap;
1621 {
1622 
1623 	return (getsockname1(td, uap, 1));
1624 }
1625 #endif /* COMPAT_OLDSOCK */
1626 
1627 /*
1628  * getpeername1() - Get name of peer for connected socket.
1629  */
1630 /* ARGSUSED */
1631 static int
getpeername1(td,uap,compat)1632 getpeername1(td, uap, compat)
1633 	struct thread *td;
1634 	struct getpeername_args /* {
1635 		int	fdes;
1636 		struct sockaddr * __restrict	asa;
1637 		socklen_t * __restrict	alen;
1638 	} */ *uap;
1639 	int compat;
1640 {
1641 	struct sockaddr *sa;
1642 	socklen_t len;
1643 	int error;
1644 
1645 	error = copyin(uap->alen, &len, sizeof (len));
1646 	if (error != 0)
1647 		return (error);
1648 
1649 	error = kern_getpeername(td, uap->fdes, &sa, &len);
1650 	if (error != 0)
1651 		return (error);
1652 
1653 	if (len != 0) {
1654 #ifdef COMPAT_OLDSOCK
1655 		if (compat)
1656 			((struct osockaddr *)sa)->sa_family = sa->sa_family;
1657 #endif
1658 		error = copyout(sa, uap->asa, (u_int)len);
1659 	}
1660 	free(sa, M_SONAME);
1661 	if (error == 0)
1662 		error = copyout(&len, uap->alen, sizeof(len));
1663 	return (error);
1664 }
1665 
1666 int
kern_getpeername(struct thread * td,int fd,struct sockaddr ** sa,socklen_t * alen)1667 kern_getpeername(struct thread *td, int fd, struct sockaddr **sa,
1668     socklen_t *alen)
1669 {
1670 	struct socket *so;
1671 	struct file *fp;
1672 	cap_rights_t rights;
1673 	socklen_t len;
1674 	int error;
1675 
1676 	AUDIT_ARG_FD(fd);
1677 	error = getsock_cap(td, fd, cap_rights_init(&rights, CAP_GETPEERNAME),
1678 	    &fp, NULL, NULL);
1679 	if (error != 0)
1680 		return (error);
1681 	so = fp->f_data;
1682 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1683 		error = ENOTCONN;
1684 		goto done;
1685 	}
1686 	*sa = NULL;
1687 	CURVNET_SET(so->so_vnet);
1688 	error = (*so->so_proto->pr_usrreqs->pru_peeraddr)(so, sa);
1689 	CURVNET_RESTORE();
1690 	if (error != 0)
1691 		goto bad;
1692 	if (*sa == NULL)
1693 		len = 0;
1694 	else
1695 		len = MIN(*alen, (*sa)->sa_len);
1696 	*alen = len;
1697 #ifdef KTRACE
1698 	if (KTRPOINT(td, KTR_STRUCT))
1699 		ktrsockaddr(*sa);
1700 #endif
1701 bad:
1702 	if (error != 0 && *sa != NULL) {
1703 		free(*sa, M_SONAME);
1704 		*sa = NULL;
1705 	}
1706 done:
1707 	fdrop(fp, td);
1708 	return (error);
1709 }
1710 
1711 int
sys_getpeername(td,uap)1712 sys_getpeername(td, uap)
1713 	struct thread *td;
1714 	struct getpeername_args *uap;
1715 {
1716 
1717 	return (getpeername1(td, uap, 0));
1718 }
1719 
1720 #ifdef COMPAT_OLDSOCK
1721 int
ogetpeername(td,uap)1722 ogetpeername(td, uap)
1723 	struct thread *td;
1724 	struct ogetpeername_args *uap;
1725 {
1726 
1727 	/* XXX uap should have type `getpeername_args *' to begin with. */
1728 	return (getpeername1(td, (struct getpeername_args *)uap, 1));
1729 }
1730 #endif /* COMPAT_OLDSOCK */
1731 
1732 static int
sockargs(struct mbuf ** mp,char * buf,socklen_t buflen,int type)1733 sockargs(struct mbuf **mp, char *buf, socklen_t buflen, int type)
1734 {
1735 	struct sockaddr *sa;
1736 	struct mbuf *m;
1737 	int error;
1738 
1739 	if (buflen > MLEN) {
1740 #ifdef COMPAT_OLDSOCK
1741 		if (type == MT_SONAME && buflen <= 112)
1742 			buflen = MLEN;		/* unix domain compat. hack */
1743 		else
1744 #endif
1745 			if (buflen > MCLBYTES)
1746 				return (EINVAL);
1747 	}
1748 	m = m_get2(buflen, M_WAITOK, type, 0);
1749 	m->m_len = buflen;
1750 	error = copyin(buf, mtod(m, void *), buflen);
1751 	if (error != 0)
1752 		(void) m_free(m);
1753 	else {
1754 		*mp = m;
1755 		if (type == MT_SONAME) {
1756 			sa = mtod(m, struct sockaddr *);
1757 
1758 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
1759 			if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1760 				sa->sa_family = sa->sa_len;
1761 #endif
1762 			sa->sa_len = buflen;
1763 		}
1764 	}
1765 	return (error);
1766 }
1767 
1768 int
getsockaddr(namp,uaddr,len)1769 getsockaddr(namp, uaddr, len)
1770 	struct sockaddr **namp;
1771 	caddr_t uaddr;
1772 	size_t len;
1773 {
1774 	struct sockaddr *sa;
1775 	int error;
1776 
1777 	if (len > SOCK_MAXADDRLEN)
1778 		return (ENAMETOOLONG);
1779 	if (len < offsetof(struct sockaddr, sa_data[0]))
1780 		return (EINVAL);
1781 	sa = malloc(len, M_SONAME, M_WAITOK);
1782 	error = copyin(uaddr, sa, len);
1783 	if (error != 0) {
1784 		free(sa, M_SONAME);
1785 	} else {
1786 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
1787 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1788 			sa->sa_family = sa->sa_len;
1789 #endif
1790 		sa->sa_len = len;
1791 		*namp = sa;
1792 	}
1793 	return (error);
1794 }
1795 
1796 /*
1797  * Dispose of externalized rights from an SCM_RIGHTS message.  This function
1798  * should be used in error or truncation cases to avoid leaking file descriptors
1799  * into the recipient's (the current thread's) table.
1800  */
1801 void
m_dispose_extcontrolm(struct mbuf * m)1802 m_dispose_extcontrolm(struct mbuf *m)
1803 {
1804 	struct cmsghdr *cm;
1805 	struct file *fp;
1806 	struct thread *td;
1807 	cap_rights_t rights;
1808 	socklen_t clen, datalen;
1809 	int error, fd, *fds, nfd;
1810 
1811 	td = curthread;
1812 	for (; m != NULL; m = m->m_next) {
1813 		if (m->m_type != MT_EXTCONTROL)
1814 			continue;
1815 		cm = mtod(m, struct cmsghdr *);
1816 		clen = m->m_len;
1817 		while (clen > 0) {
1818 			if (clen < sizeof(*cm))
1819 				panic("%s: truncated mbuf %p", __func__, m);
1820 			datalen = CMSG_SPACE(cm->cmsg_len - CMSG_SPACE(0));
1821 			if (clen < datalen)
1822 				panic("%s: truncated mbuf %p", __func__, m);
1823 
1824 			if (cm->cmsg_level == SOL_SOCKET &&
1825 			    cm->cmsg_type == SCM_RIGHTS) {
1826 				fds = (int *)CMSG_DATA(cm);
1827 				nfd = (cm->cmsg_len - CMSG_SPACE(0)) /
1828 				    sizeof(int);
1829 
1830 				while (nfd-- > 0) {
1831 					fd = *fds++;
1832 					error = fget(td, fd,
1833 					    cap_rights_init(&rights), &fp);
1834 					if (error == 0) {
1835 						fdclose(td, fp, fd);
1836 						fdrop(fp, td);
1837 					}
1838 				}
1839 			}
1840 			clen -= datalen;
1841 			cm = (struct cmsghdr *)((uint8_t *)cm + datalen);
1842 		}
1843 		m_chtype(m, MT_CONTROL);
1844 	}
1845 }
1846