xref: /netbsd/sys/kern/uipc_syscalls.c (revision 3b433a78)
1 /*	$NetBSD: uipc_syscalls.c,v 1.206 2022/07/01 22:30:51 riastradh Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Andrew Doran.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1982, 1986, 1989, 1990, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)uipc_syscalls.c	8.6 (Berkeley) 2/14/95
61  */
62 
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: uipc_syscalls.c,v 1.206 2022/07/01 22:30:51 riastradh Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_pipe.h"
68 #include "opt_sctp.h"
69 #endif
70 
71 #define MBUFTYPES
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/filedesc.h>
75 #include <sys/proc.h>
76 #include <sys/file.h>
77 #include <sys/buf.h>
78 #include <sys/mbuf.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/signalvar.h>
83 #include <sys/un.h>
84 #include <sys/ktrace.h>
85 #include <sys/event.h>
86 #include <sys/atomic.h>
87 #include <sys/kauth.h>
88 
89 #ifdef SCTP
90 #include <netinet/sctp_uio.h>
91 #include <netinet/sctp_peeloff.h>
92 #endif
93 
94 #include <sys/mount.h>
95 #include <sys/syscallargs.h>
96 
97 /*
98  * System call interface to the socket abstraction.
99  */
100 extern const struct fileops socketops;
101 
102 static int	sockargs_sb(struct sockaddr_big *, const void *, socklen_t);
103 
104 int
sys___socket30(struct lwp * l,const struct sys___socket30_args * uap,register_t * retval)105 sys___socket30(struct lwp *l, const struct sys___socket30_args *uap,
106     register_t *retval)
107 {
108 	/* {
109 		syscallarg(int)	domain;
110 		syscallarg(int)	type;
111 		syscallarg(int)	protocol;
112 	} */
113 	int fd, error;
114 
115 	error = fsocreate(SCARG(uap, domain), NULL, SCARG(uap, type),
116 	    SCARG(uap, protocol), &fd);
117 	if (error == 0) {
118 		*retval = fd;
119 	}
120 	return error;
121 }
122 
123 int
sys_bind(struct lwp * l,const struct sys_bind_args * uap,register_t * retval)124 sys_bind(struct lwp *l, const struct sys_bind_args *uap, register_t *retval)
125 {
126 	/* {
127 		syscallarg(int)				s;
128 		syscallarg(const struct sockaddr *)	name;
129 		syscallarg(unsigned int)		namelen;
130 	} */
131 	int		error;
132 	struct sockaddr_big sb;
133 
134 	error = sockargs_sb(&sb, SCARG(uap, name), SCARG(uap, namelen));
135 	if (error)
136 		return error;
137 
138 	return do_sys_bind(l, SCARG(uap, s), (struct sockaddr *)&sb);
139 }
140 
141 int
do_sys_bind(struct lwp * l,int fd,struct sockaddr * nam)142 do_sys_bind(struct lwp *l, int fd, struct sockaddr *nam)
143 {
144 	struct socket	*so;
145 	int		error;
146 
147 	if ((error = fd_getsock(fd, &so)) != 0)
148 		return error;
149 	error = sobind(so, nam, l);
150 	fd_putfile(fd);
151 	return error;
152 }
153 
154 int
sys_listen(struct lwp * l,const struct sys_listen_args * uap,register_t * retval)155 sys_listen(struct lwp *l, const struct sys_listen_args *uap, register_t *retval)
156 {
157 	/* {
158 		syscallarg(int)	s;
159 		syscallarg(int)	backlog;
160 	} */
161 	struct socket	*so;
162 	int		error;
163 
164 	if ((error = fd_getsock(SCARG(uap, s), &so)) != 0)
165 		return (error);
166 	error = solisten(so, SCARG(uap, backlog), l);
167 	fd_putfile(SCARG(uap, s));
168 	return error;
169 }
170 
171 int
do_sys_accept(struct lwp * l,int sock,struct sockaddr * name,register_t * new_sock,const sigset_t * mask,int flags,int clrflags)172 do_sys_accept(struct lwp *l, int sock, struct sockaddr *name,
173     register_t *new_sock, const sigset_t *mask, int flags, int clrflags)
174 {
175 	file_t		*fp, *fp2;
176 	int		error, fd;
177 	struct socket	*so, *so2;
178 	short		wakeup_state = 0;
179 
180 	if ((fp = fd_getfile(sock)) == NULL)
181 		return EBADF;
182 	if (fp->f_type != DTYPE_SOCKET) {
183 		fd_putfile(sock);
184 		return ENOTSOCK;
185 	}
186 	if ((error = fd_allocfile(&fp2, &fd)) != 0) {
187 		fd_putfile(sock);
188 		return error;
189 	}
190 	*new_sock = fd;
191 	so = fp->f_socket;
192 	solock(so);
193 
194 	if (__predict_false(mask))
195 		sigsuspendsetup(l, mask);
196 
197 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
198 		error = EOPNOTSUPP;
199 		goto bad;
200 	}
201 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
202 		error = EINVAL;
203 		goto bad;
204 	}
205 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
206 		error = EWOULDBLOCK;
207 		goto bad;
208 	}
209 	while (so->so_qlen == 0 && so->so_error == 0) {
210 		if (so->so_state & SS_CANTRCVMORE) {
211 			so->so_error = ECONNABORTED;
212 			break;
213 		}
214 		if (wakeup_state & SS_RESTARTSYS) {
215 			error = ERESTART;
216 			goto bad;
217 		}
218 		error = sowait(so, true, 0);
219 		if (error) {
220 			goto bad;
221 		}
222 		wakeup_state = so->so_state;
223 	}
224 	if (so->so_error) {
225 		error = so->so_error;
226 		so->so_error = 0;
227 		goto bad;
228 	}
229 	/* connection has been removed from the listen queue */
230 	KNOTE(&so->so_rcv.sb_sel.sel_klist, NOTE_SUBMIT);
231 	so2 = TAILQ_FIRST(&so->so_q);
232 	if (soqremque(so2, 1) == 0)
233 		panic("accept");
234 	fp2->f_type = DTYPE_SOCKET;
235 	fp2->f_flag = (fp->f_flag & ~clrflags) |
236 	    ((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)|
237 	    ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0);
238 	fp2->f_ops = &socketops;
239 	fp2->f_socket = so2;
240 	if (fp2->f_flag & FNONBLOCK)
241 		so2->so_state |= SS_NBIO;
242 	else
243 		so2->so_state &= ~SS_NBIO;
244 	error = soaccept(so2, name);
245 	so2->so_cred = kauth_cred_dup(so->so_cred);
246 	sounlock(so);
247 	if (error) {
248 		/* an error occurred, free the file descriptor and mbuf */
249 		mutex_enter(&fp2->f_lock);
250 		fp2->f_count++;
251 		mutex_exit(&fp2->f_lock);
252 		closef(fp2);
253 		fd_abort(curproc, NULL, fd);
254 	} else {
255 		fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0);
256 		fd_affix(curproc, fp2, fd);
257 	}
258 	fd_putfile(sock);
259 	if (__predict_false(mask))
260 		sigsuspendteardown(l);
261 	return error;
262  bad:
263 	sounlock(so);
264 	fd_putfile(sock);
265 	fd_abort(curproc, fp2, fd);
266 	if (__predict_false(mask))
267 		sigsuspendteardown(l);
268 	return error;
269 }
270 
271 int
sys_accept(struct lwp * l,const struct sys_accept_args * uap,register_t * retval)272 sys_accept(struct lwp *l, const struct sys_accept_args *uap, register_t *retval)
273 {
274 	/* {
275 		syscallarg(int)			s;
276 		syscallarg(struct sockaddr *)	name;
277 		syscallarg(unsigned int *)	anamelen;
278 	} */
279 	int error, fd;
280 	struct sockaddr_big name;
281 
282 	name.sb_len = UCHAR_MAX;
283 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
284 	    retval, NULL, 0, 0);
285 	if (error != 0)
286 		return error;
287 	error = copyout_sockname_sb(SCARG(uap, name), SCARG(uap, anamelen),
288 	    MSG_LENUSRSPACE, &name);
289 	if (error != 0) {
290 		fd = (int)*retval;
291 		if (fd_getfile(fd) != NULL)
292 			(void)fd_close(fd);
293 	}
294 	return error;
295 }
296 
297 int
sys_paccept(struct lwp * l,const struct sys_paccept_args * uap,register_t * retval)298 sys_paccept(struct lwp *l, const struct sys_paccept_args *uap,
299     register_t *retval)
300 {
301 	/* {
302 		syscallarg(int)			s;
303 		syscallarg(struct sockaddr *)	name;
304 		syscallarg(unsigned int *)	anamelen;
305 		syscallarg(const sigset_t *)	mask;
306 		syscallarg(int)			flags;
307 	} */
308 	int error, fd;
309 	struct sockaddr_big name;
310 	sigset_t *mask, amask;
311 
312 	if (SCARG(uap, mask) != NULL) {
313 		error = copyin(SCARG(uap, mask), &amask, sizeof(amask));
314 		if (error)
315 			return error;
316 		mask = &amask;
317 	} else
318 		mask = NULL;
319 
320 	name.sb_len = UCHAR_MAX;
321 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
322 	    retval, mask, SCARG(uap, flags), FNONBLOCK);
323 	if (error != 0)
324 		return error;
325 	error = copyout_sockname_sb(SCARG(uap, name), SCARG(uap, anamelen),
326 	    MSG_LENUSRSPACE, &name);
327 	if (error != 0) {
328 		fd = (int)*retval;
329 		if (fd_getfile(fd) != NULL)
330 			(void)fd_close(fd);
331 	}
332 	return error;
333 }
334 
335 int
sys_connect(struct lwp * l,const struct sys_connect_args * uap,register_t * retval)336 sys_connect(struct lwp *l, const struct sys_connect_args *uap,
337     register_t *retval)
338 {
339 	/* {
340 		syscallarg(int)				s;
341 		syscallarg(const struct sockaddr *)	name;
342 		syscallarg(unsigned int)		namelen;
343 	} */
344 	int		error;
345 	struct sockaddr_big sbig;
346 
347 	error = sockargs_sb(&sbig, SCARG(uap, name), SCARG(uap, namelen));
348 	if (error)
349 		return error;
350 	return do_sys_connect(l, SCARG(uap, s), (struct sockaddr *)&sbig);
351 }
352 
353 int
do_sys_connect(struct lwp * l,int fd,struct sockaddr * nam)354 do_sys_connect(struct lwp *l, int fd, struct sockaddr *nam)
355 {
356 	struct socket	*so;
357 	int		error;
358 	int		interrupted = 0;
359 
360 	if ((error = fd_getsock(fd, &so)) != 0) {
361 		return (error);
362 	}
363 	solock(so);
364 	if ((so->so_state & SS_ISCONNECTING) != 0) {
365 		error = EALREADY;
366 		goto out;
367 	}
368 
369 	error = soconnect(so, nam, l);
370 	if (error)
371 		goto bad;
372 	if ((so->so_state & (SS_NBIO|SS_ISCONNECTING)) ==
373 	    (SS_NBIO|SS_ISCONNECTING)) {
374 		error = EINPROGRESS;
375 		goto out;
376 	}
377 	while ((so->so_state & SS_ISCONNECTING) != 0 && so->so_error == 0) {
378 		error = sowait(so, true, 0);
379 		if (__predict_false((so->so_state & SS_ISABORTING) != 0)) {
380 			error = EPIPE;
381 			interrupted = 1;
382 			break;
383 		}
384 		if (error) {
385 			if (error == EINTR || error == ERESTART)
386 				interrupted = 1;
387 			break;
388 		}
389 	}
390 	if (error == 0) {
391 		error = so->so_error;
392 		so->so_error = 0;
393 	}
394  bad:
395 	if (!interrupted)
396 		so->so_state &= ~SS_ISCONNECTING;
397 	if (error == ERESTART)
398 		error = EINTR;
399  out:
400 	sounlock(so);
401 	fd_putfile(fd);
402 	return error;
403 }
404 
405 static int
makesocket(struct lwp * l,file_t ** fp,int * fd,int flags,int type,int domain,int proto,struct socket * soo)406 makesocket(struct lwp *l, file_t **fp, int *fd, int flags, int type,
407     int domain, int proto, struct socket *soo)
408 {
409 	struct socket *so;
410 	int error;
411 
412 	if ((error = socreate(domain, &so, type, proto, l, soo)) != 0) {
413 		return error;
414 	}
415 	if (flags & SOCK_NONBLOCK) {
416 		so->so_state |= SS_NBIO;
417 	}
418 
419 	if ((error = fd_allocfile(fp, fd)) != 0) {
420 		soclose(so);
421 		return error;
422 	}
423 	fd_set_exclose(l, *fd, (flags & SOCK_CLOEXEC) != 0);
424 	(*fp)->f_flag = FREAD|FWRITE|
425 	    ((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)|
426 	    ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0);
427 	(*fp)->f_type = DTYPE_SOCKET;
428 	(*fp)->f_ops = &socketops;
429 	(*fp)->f_socket = so;
430 	return 0;
431 }
432 
433 int
sys_socketpair(struct lwp * l,const struct sys_socketpair_args * uap,register_t * retval)434 sys_socketpair(struct lwp *l, const struct sys_socketpair_args *uap,
435     register_t *retval)
436 {
437 	/* {
438 		syscallarg(int)		domain;
439 		syscallarg(int)		type;
440 		syscallarg(int)		protocol;
441 		syscallarg(int *)	rsv;
442 	} */
443 	file_t		*fp1, *fp2;
444 	struct socket	*so1, *so2;
445 	int		fd, error, sv[2];
446 	proc_t		*p = curproc;
447 	int		flags = SCARG(uap, type) & SOCK_FLAGS_MASK;
448 	int		type = SCARG(uap, type) & ~SOCK_FLAGS_MASK;
449 	int		domain = SCARG(uap, domain);
450 	int		proto = SCARG(uap, protocol);
451 
452 	error = makesocket(l, &fp1, &fd, flags, type, domain, proto, NULL);
453 	if (error)
454 		return error;
455 	so1 = fp1->f_socket;
456 	sv[0] = fd;
457 
458 	error = makesocket(l, &fp2, &fd, flags, type, domain, proto, so1);
459 	if (error)
460 		goto out;
461 	so2 = fp2->f_socket;
462 	sv[1] = fd;
463 
464 	solock(so1);
465 	error = soconnect2(so1, so2);
466 	if (error == 0 && type == SOCK_DGRAM) {
467 		/*
468 		 * Datagram socket connection is asymmetric.
469 		 */
470 		error = soconnect2(so2, so1);
471 	}
472 	sounlock(so1);
473 
474 	if (error == 0)
475 		error = copyout(sv, SCARG(uap, rsv), sizeof(sv));
476 	if (error == 0) {
477 		fd_affix(p, fp2, sv[1]);
478 		fd_affix(p, fp1, sv[0]);
479 		return 0;
480 	}
481 	fd_abort(p, fp2, sv[1]);
482 	(void)soclose(so2);
483 out:
484 	fd_abort(p, fp1, sv[0]);
485 	(void)soclose(so1);
486 	return error;
487 }
488 
489 int
sys_sendto(struct lwp * l,const struct sys_sendto_args * uap,register_t * retval)490 sys_sendto(struct lwp *l, const struct sys_sendto_args *uap,
491     register_t *retval)
492 {
493 	/* {
494 		syscallarg(int)				s;
495 		syscallarg(const void *)		buf;
496 		syscallarg(size_t)			len;
497 		syscallarg(int)				flags;
498 		syscallarg(const struct sockaddr *)	to;
499 		syscallarg(unsigned int)		tolen;
500 	} */
501 	struct msghdr	msg = {0};
502 	struct iovec	aiov;
503 
504 	msg.msg_name = __UNCONST(SCARG(uap, to)); /* XXXUNCONST kills const */
505 	msg.msg_namelen = SCARG(uap, tolen);
506 	msg.msg_iov = &aiov;
507 	msg.msg_iovlen = 1;
508 	msg.msg_control = NULL;
509 	msg.msg_flags = 0;
510 	aiov.iov_base = __UNCONST(SCARG(uap, buf)); /* XXXUNCONST kills const */
511 	aiov.iov_len = SCARG(uap, len);
512 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, SCARG(uap, flags),
513 	    retval);
514 }
515 
516 int
sys_sendmsg(struct lwp * l,const struct sys_sendmsg_args * uap,register_t * retval)517 sys_sendmsg(struct lwp *l, const struct sys_sendmsg_args *uap,
518     register_t *retval)
519 {
520 	/* {
521 		syscallarg(int)				s;
522 		syscallarg(const struct msghdr *)	msg;
523 		syscallarg(int)				flags;
524 	} */
525 	struct msghdr	msg;
526 	int		error;
527 
528 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
529 	if (error)
530 		return (error);
531 
532 	msg.msg_flags = MSG_IOVUSRSPACE;
533 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, SCARG(uap, flags),
534 	    retval);
535 }
536 
537 int
do_sys_sendmsg_so(struct lwp * l,int s,struct socket * so,file_t * fp,struct msghdr * mp,int flags,register_t * retsize)538 do_sys_sendmsg_so(struct lwp *l, int s, struct socket *so, file_t *fp,
539     struct msghdr *mp, int flags, register_t *retsize)
540 {
541 
542 	struct iovec	aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
543 	struct sockaddr *sa = NULL;
544 	struct mbuf	*to, *control;
545 	struct uio	auio;
546 	size_t		len, iovsz;
547 	int		i, error;
548 
549 	ktrkuser("msghdr", mp, sizeof(*mp));
550 
551 	/* If the caller passed us stuff in mbufs, we must free them. */
552 	to = (mp->msg_flags & MSG_NAMEMBUF) ? mp->msg_name : NULL;
553 	control = (mp->msg_flags & MSG_CONTROLMBUF) ? mp->msg_control : NULL;
554 	iovsz = mp->msg_iovlen * sizeof(struct iovec);
555 
556 	if (mp->msg_flags & MSG_IOVUSRSPACE) {
557 		if ((unsigned int)mp->msg_iovlen > UIO_SMALLIOV) {
558 			if ((unsigned int)mp->msg_iovlen > IOV_MAX) {
559 				error = EMSGSIZE;
560 				goto bad;
561 			}
562 			iov = kmem_alloc(iovsz, KM_SLEEP);
563 		}
564 		if (mp->msg_iovlen != 0) {
565 			error = copyin(mp->msg_iov, iov, iovsz);
566 			if (error)
567 				goto bad;
568 		}
569 		auio.uio_iov = iov;
570 	} else
571 		auio.uio_iov = mp->msg_iov;
572 
573 	auio.uio_iovcnt = mp->msg_iovlen;
574 	auio.uio_rw = UIO_WRITE;
575 	auio.uio_offset = 0;			/* XXX */
576 	auio.uio_resid = 0;
577 	KASSERT(l == curlwp);
578 	auio.uio_vmspace = l->l_proc->p_vmspace;
579 
580 	tiov = auio.uio_iov;
581 	for (i = 0; i < auio.uio_iovcnt; i++, tiov++) {
582 		/*
583 		 * Writes return ssize_t because -1 is returned on error.
584 		 * Therefore, we must restrict the length to SSIZE_MAX to
585 		 * avoid garbage return values.
586 		 */
587 		auio.uio_resid += tiov->iov_len;
588 		if (tiov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
589 			error = EINVAL;
590 			goto bad;
591 		}
592 	}
593 
594 	if (mp->msg_name && to == NULL) {
595 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
596 		    UIO_USERSPACE, MT_SONAME);
597 		if (error)
598 			goto bad;
599 	}
600 
601 	if (mp->msg_control) {
602 		if (mp->msg_controllen < CMSG_ALIGN(sizeof(struct cmsghdr))) {
603 			error = EINVAL;
604 			goto bad;
605 		}
606 		if (control == NULL) {
607 			error = sockargs(&control, mp->msg_control,
608 			    mp->msg_controllen, UIO_USERSPACE, MT_CONTROL);
609 			if (error)
610 				goto bad;
611 		}
612 	}
613 
614 	if (ktrpoint(KTR_GENIO) && iovsz > 0) {
615 		ktriov = kmem_alloc(iovsz, KM_SLEEP);
616 		memcpy(ktriov, auio.uio_iov, iovsz);
617 	}
618 
619 	if (mp->msg_name)
620 		MCLAIM(to, so->so_mowner);
621 	if (mp->msg_control)
622 		MCLAIM(control, so->so_mowner);
623 
624 	if (to) {
625 		sa = mtod(to, struct sockaddr *);
626 	}
627 
628 	len = auio.uio_resid;
629 	error = (*so->so_send)(so, sa, &auio, NULL, control, flags, l);
630 	/* Protocol is responsible for freeing 'control' */
631 	control = NULL;
632 
633 	if (error) {
634 		if (auio.uio_resid != len && (error == ERESTART ||
635 		    error == EINTR || error == EWOULDBLOCK))
636 			error = 0;
637 		if (error == EPIPE && (fp->f_flag & FNOSIGPIPE) == 0 &&
638 		    (flags & MSG_NOSIGNAL) == 0) {
639 			mutex_enter(&proc_lock);
640 			psignal(l->l_proc, SIGPIPE);
641 			mutex_exit(&proc_lock);
642 		}
643 	}
644 	if (error == 0)
645 		*retsize = len - auio.uio_resid;
646 
647 bad:
648 	if (ktriov != NULL) {
649 		ktrgeniov(s, UIO_WRITE, ktriov, *retsize, error);
650 		kmem_free(ktriov, iovsz);
651 	}
652 
653 	if (iov != aiov)
654 		kmem_free(iov, iovsz);
655 	if (to)
656 		m_freem(to);
657 	if (control)
658 		m_freem(control);
659 
660 	return error;
661 }
662 
663 int
do_sys_sendmsg(struct lwp * l,int s,struct msghdr * mp,int flags,register_t * retsize)664 do_sys_sendmsg(struct lwp *l, int s, struct msghdr *mp, int flags,
665     register_t *retsize)
666 {
667 	int		error;
668 	struct socket	*so;
669 	file_t		*fp;
670 
671 	if ((error = fd_getsock1(s, &so, &fp)) != 0) {
672 		/* We have to free msg_name and msg_control ourselves */
673 		if (mp->msg_flags & MSG_NAMEMBUF)
674 			m_freem(mp->msg_name);
675 		if (mp->msg_flags & MSG_CONTROLMBUF)
676 			m_freem(mp->msg_control);
677 		return error;
678 	}
679 	error = do_sys_sendmsg_so(l, s, so, fp, mp, flags, retsize);
680 	/* msg_name and msg_control freed */
681 	fd_putfile(s);
682 	return error;
683 }
684 
685 int
sys_recvfrom(struct lwp * l,const struct sys_recvfrom_args * uap,register_t * retval)686 sys_recvfrom(struct lwp *l, const struct sys_recvfrom_args *uap,
687     register_t *retval)
688 {
689 	/* {
690 		syscallarg(int)			s;
691 		syscallarg(void *)		buf;
692 		syscallarg(size_t)		len;
693 		syscallarg(int)			flags;
694 		syscallarg(struct sockaddr *)	from;
695 		syscallarg(unsigned int *)	fromlenaddr;
696 	} */
697 	struct msghdr	msg = {0};
698 	struct iovec	aiov;
699 	int		error;
700 	struct mbuf	*from;
701 
702 	msg.msg_name = NULL;
703 	msg.msg_iov = &aiov;
704 	msg.msg_iovlen = 1;
705 	aiov.iov_base = SCARG(uap, buf);
706 	aiov.iov_len = SCARG(uap, len);
707 	msg.msg_control = NULL;
708 	msg.msg_flags = SCARG(uap, flags) & MSG_USERFLAGS;
709 
710 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from, NULL, retval);
711 	if (error != 0)
712 		return error;
713 
714 	error = copyout_sockname(SCARG(uap, from), SCARG(uap, fromlenaddr),
715 	    MSG_LENUSRSPACE, from);
716 	if (from != NULL)
717 		m_free(from);
718 	return error;
719 }
720 
721 int
sys_recvmsg(struct lwp * l,const struct sys_recvmsg_args * uap,register_t * retval)722 sys_recvmsg(struct lwp *l, const struct sys_recvmsg_args *uap,
723     register_t *retval)
724 {
725 	/* {
726 		syscallarg(int)			s;
727 		syscallarg(struct msghdr *)	msg;
728 		syscallarg(int)			flags;
729 	} */
730 	struct msghdr	msg;
731 	int		error;
732 	struct mbuf	*from, *control;
733 
734 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
735 	if (error)
736 		return error;
737 
738 	msg.msg_flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
739 
740 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
741 	    msg.msg_control != NULL ? &control : NULL, retval);
742 	if (error != 0)
743 		return error;
744 
745 	if (msg.msg_control != NULL)
746 		error = copyout_msg_control(l, &msg, control);
747 
748 	if (error == 0)
749 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
750 			from);
751 	if (from != NULL)
752 		m_free(from);
753 	if (error == 0) {
754 		ktrkuser("msghdr", &msg, sizeof(msg));
755 		error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
756 	}
757 
758 	return error;
759 }
760 
761 int
sys_sendmmsg(struct lwp * l,const struct sys_sendmmsg_args * uap,register_t * retval)762 sys_sendmmsg(struct lwp *l, const struct sys_sendmmsg_args *uap,
763     register_t *retval)
764 {
765 	/* {
766 		syscallarg(int)			s;
767 		syscallarg(struct mmsghdr *)	mmsg;
768 		syscallarg(unsigned int)	vlen;
769 		syscallarg(unsigned int)	flags;
770 	} */
771 	struct mmsghdr mmsg;
772 	struct socket *so;
773 	file_t *fp;
774 	struct msghdr *msg = &mmsg.msg_hdr;
775 	int error, s;
776 	unsigned int vlen, flags, dg;
777 
778 	s = SCARG(uap, s);
779 	if ((error = fd_getsock1(s, &so, &fp)) != 0)
780 		return error;
781 
782 	vlen = SCARG(uap, vlen);
783 	if (vlen > 1024)
784 		vlen = 1024;
785 
786 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
787 
788 	for (dg = 0; dg < vlen;) {
789 		error = copyin(SCARG(uap, mmsg) + dg, &mmsg, sizeof(mmsg));
790 		if (error)
791 			break;
792 
793 		msg->msg_flags = flags;
794 
795 		error = do_sys_sendmsg_so(l, s, so, fp, msg, flags, retval);
796 		if (error)
797 			break;
798 
799 		ktrkuser("msghdr", msg, sizeof(*msg));
800 		mmsg.msg_len = *retval;
801 		error = copyout(&mmsg, SCARG(uap, mmsg) + dg, sizeof(mmsg));
802 		if (error)
803 			break;
804 		dg++;
805 
806 	}
807 
808 	*retval = dg;
809 
810 	fd_putfile(s);
811 
812 	/*
813 	 * If we succeeded at least once, return 0.
814 	 */
815 	if (dg)
816 		return 0;
817 	return error;
818 }
819 
820 /*
821  * Adjust for a truncated SCM_RIGHTS control message.
822  *  This means closing any file descriptors that aren't present
823  *  in the returned buffer.
824  *  m is the mbuf holding the (already externalized) SCM_RIGHTS message.
825  */
826 static void
free_rights(struct mbuf * m)827 free_rights(struct mbuf *m)
828 {
829 	struct cmsghdr *cm;
830 	int *fdv;
831 	unsigned int nfds, i;
832 
833 	KASSERT(sizeof(*cm) <= m->m_len);
834 	cm = mtod(m, struct cmsghdr *);
835 
836 	KASSERT(CMSG_ALIGN(sizeof(*cm)) <= cm->cmsg_len);
837 	KASSERT(cm->cmsg_len <= m->m_len);
838 	nfds = (cm->cmsg_len - CMSG_ALIGN(sizeof(*cm))) / sizeof(int);
839 	fdv = (int *)CMSG_DATA(cm);
840 
841 	for (i = 0; i < nfds; i++)
842 		if (fd_getfile(fdv[i]) != NULL)
843 			(void)fd_close(fdv[i]);
844 }
845 
846 void
free_control_mbuf(struct lwp * l,struct mbuf * control,struct mbuf * uncopied)847 free_control_mbuf(struct lwp *l, struct mbuf *control, struct mbuf *uncopied)
848 {
849 	struct mbuf *next;
850 	struct cmsghdr *cmsg;
851 	bool do_free_rights = false;
852 
853 	while (control != NULL) {
854 		cmsg = mtod(control, struct cmsghdr *);
855 		if (control == uncopied)
856 			do_free_rights = true;
857 		if (do_free_rights && cmsg->cmsg_level == SOL_SOCKET
858 		    && cmsg->cmsg_type == SCM_RIGHTS)
859 			free_rights(control);
860 		next = control->m_next;
861 		m_free(control);
862 		control = next;
863 	}
864 }
865 
866 /* Copy socket control/CMSG data to user buffer, frees the mbuf */
867 int
copyout_msg_control(struct lwp * l,struct msghdr * mp,struct mbuf * control)868 copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
869 {
870 	int i, len, error = 0;
871 	struct cmsghdr *cmsg;
872 	struct mbuf *m;
873 	char *q;
874 
875 	len = mp->msg_controllen;
876 	if (len <= 0 || control == 0) {
877 		mp->msg_controllen = 0;
878 		free_control_mbuf(l, control, control);
879 		return 0;
880 	}
881 
882 	q = (char *)mp->msg_control;
883 
884 	for (m = control; m != NULL; ) {
885 		cmsg = mtod(m, struct cmsghdr *);
886 		i = m->m_len;
887 		if (len < i) {
888 			mp->msg_flags |= MSG_CTRUNC;
889 			if (cmsg->cmsg_level == SOL_SOCKET
890 			    && cmsg->cmsg_type == SCM_RIGHTS)
891 				/* Do not truncate me ... */
892 				break;
893 			i = len;
894 		}
895 		error = copyout(mtod(m, void *), q, i);
896 		ktrkuser(mbuftypes[MT_CONTROL], cmsg, cmsg->cmsg_len);
897 		if (error != 0) {
898 			/* We must free all the SCM_RIGHTS */
899 			m = control;
900 			break;
901 		}
902 		m = m->m_next;
903 		if (m)
904 			i = ALIGN(i);
905 		q += i;
906 		len -= i;
907 		if (len <= 0)
908 			break;
909 	}
910 
911 	free_control_mbuf(l, control, m);
912 
913 	mp->msg_controllen = q - (char *)mp->msg_control;
914 	return error;
915 }
916 
917 int
do_sys_recvmsg_so(struct lwp * l,int s,struct socket * so,struct msghdr * mp,struct mbuf ** from,struct mbuf ** control,register_t * retsize)918 do_sys_recvmsg_so(struct lwp *l, int s, struct socket *so, struct msghdr *mp,
919     struct mbuf **from, struct mbuf **control, register_t *retsize)
920 {
921 	struct iovec	aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
922 	struct uio	auio;
923 	size_t		len, iovsz;
924 	int		i, error;
925 
926 	ktrkuser("msghdr", mp, sizeof(*mp));
927 
928 	*from = NULL;
929 	if (control != NULL)
930 		*control = NULL;
931 
932 	iovsz = mp->msg_iovlen * sizeof(struct iovec);
933 
934 	if (mp->msg_flags & MSG_IOVUSRSPACE) {
935 		if ((unsigned int)mp->msg_iovlen > UIO_SMALLIOV) {
936 			if ((unsigned int)mp->msg_iovlen > IOV_MAX) {
937 				error = EMSGSIZE;
938 				goto out;
939 			}
940 			iov = kmem_alloc(iovsz, KM_SLEEP);
941 		}
942 		if (mp->msg_iovlen != 0) {
943 			error = copyin(mp->msg_iov, iov, iovsz);
944 			if (error)
945 				goto out;
946 		}
947 		auio.uio_iov = iov;
948 	} else
949 		auio.uio_iov = mp->msg_iov;
950 	auio.uio_iovcnt = mp->msg_iovlen;
951 	auio.uio_rw = UIO_READ;
952 	auio.uio_offset = 0;			/* XXX */
953 	auio.uio_resid = 0;
954 	KASSERT(l == curlwp);
955 	auio.uio_vmspace = l->l_proc->p_vmspace;
956 
957 	tiov = auio.uio_iov;
958 	for (i = 0; i < auio.uio_iovcnt; i++, tiov++) {
959 		/*
960 		 * Reads return ssize_t because -1 is returned on error.
961 		 * Therefore we must restrict the length to SSIZE_MAX to
962 		 * avoid garbage return values.
963 		 */
964 		auio.uio_resid += tiov->iov_len;
965 		if (tiov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
966 			error = EINVAL;
967 			goto out;
968 		}
969 	}
970 
971 	if (ktrpoint(KTR_GENIO) && iovsz > 0) {
972 		ktriov = kmem_alloc(iovsz, KM_SLEEP);
973 		memcpy(ktriov, auio.uio_iov, iovsz);
974 	}
975 
976 	len = auio.uio_resid;
977 	mp->msg_flags &= MSG_USERFLAGS;
978 	error = (*so->so_receive)(so, from, &auio, NULL, control,
979 	    &mp->msg_flags);
980 	KASSERT(*from == NULL || (*from)->m_next == NULL);
981 	len -= auio.uio_resid;
982 	*retsize = len;
983 	if (error != 0 && len != 0
984 	    && (error == ERESTART || error == EINTR || error == EWOULDBLOCK))
985 		/* Some data transferred */
986 		error = 0;
987 
988 	if (ktriov != NULL) {
989 		ktrgeniov(s, UIO_READ, ktriov, len, error);
990 		kmem_free(ktriov, iovsz);
991 	}
992 
993 	if (error != 0) {
994 		m_freem(*from);
995 		*from = NULL;
996 		if (control != NULL) {
997 			free_control_mbuf(l, *control, *control);
998 			*control = NULL;
999 		}
1000 	}
1001  out:
1002 	if (iov != aiov)
1003 		kmem_free(iov, iovsz);
1004 	return error;
1005 }
1006 
1007 
1008 int
do_sys_recvmsg(struct lwp * l,int s,struct msghdr * mp,struct mbuf ** from,struct mbuf ** control,register_t * retsize)1009 do_sys_recvmsg(struct lwp *l, int s, struct msghdr *mp,
1010     struct mbuf **from, struct mbuf **control, register_t *retsize)
1011 {
1012 	int error;
1013 	struct socket *so;
1014 
1015 	if ((error = fd_getsock(s, &so)) != 0)
1016 		return error;
1017 	error = do_sys_recvmsg_so(l, s, so, mp, from, control, retsize);
1018 	fd_putfile(s);
1019 	return error;
1020 }
1021 
1022 int
sys_recvmmsg(struct lwp * l,const struct sys_recvmmsg_args * uap,register_t * retval)1023 sys_recvmmsg(struct lwp *l, const struct sys_recvmmsg_args *uap,
1024     register_t *retval)
1025 {
1026 	/* {
1027 		syscallarg(int)			s;
1028 		syscallarg(struct mmsghdr *)	mmsg;
1029 		syscallarg(unsigned int)	vlen;
1030 		syscallarg(unsigned int)	flags;
1031 		syscallarg(struct timespec *)	timeout;
1032 	} */
1033 	struct mmsghdr mmsg;
1034 	struct socket *so;
1035 	struct msghdr *msg = &mmsg.msg_hdr;
1036 	int error, s;
1037 	struct mbuf *from, *control;
1038 	struct timespec ts, now;
1039 	unsigned int vlen, flags, dg;
1040 
1041 	if (SCARG(uap, timeout)) {
1042 		if ((error = copyin(SCARG(uap, timeout), &ts, sizeof(ts))) != 0)
1043 			return error;
1044 		if (ts.tv_nsec < 0 || ts.tv_nsec >= 1000000000L)
1045 			return EINVAL;
1046 		getnanotime(&now);
1047 		if (timespecaddok(&now, &ts)) {
1048 			timespecadd(&now, &ts, &ts);
1049 		} else {
1050 			ts.tv_sec = __type_max(time_t);
1051 			ts.tv_nsec = 999999999L;
1052 		}
1053 	}
1054 
1055 	s = SCARG(uap, s);
1056 	if ((error = fd_getsock(s, &so)) != 0)
1057 		return error;
1058 
1059 	/*
1060 	 * If so->so_rerror holds a deferred error return it now.
1061 	 */
1062 	if (so->so_rerror) {
1063 		error = so->so_rerror;
1064 		so->so_rerror = 0;
1065 		fd_putfile(s);
1066 		return error;
1067 	}
1068 
1069 	vlen = SCARG(uap, vlen);
1070 	if (vlen > 1024)
1071 		vlen = 1024;
1072 
1073 	from = NULL;
1074 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1075 
1076 	for (dg = 0; dg < vlen;) {
1077 		error = copyin(SCARG(uap, mmsg) + dg, &mmsg, sizeof(mmsg));
1078 		if (error)
1079 			break;
1080 
1081 		msg->msg_flags = flags & ~MSG_WAITFORONE;
1082 
1083 		if (from != NULL) {
1084 			m_free(from);
1085 			from = NULL;
1086 		}
1087 
1088 		error = do_sys_recvmsg_so(l, s, so, msg, &from,
1089 		    msg->msg_control != NULL ? &control : NULL, retval);
1090 		if (error) {
1091 			if (error == EAGAIN && dg > 0)
1092 				error = 0;
1093 			break;
1094 		}
1095 
1096 		if (msg->msg_control != NULL)
1097 			error = copyout_msg_control(l, msg, control);
1098 		if (error)
1099 			break;
1100 
1101 		error = copyout_sockname(msg->msg_name, &msg->msg_namelen, 0,
1102 		    from);
1103 		if (error)
1104 			break;
1105 
1106 		ktrkuser("msghdr", msg, sizeof *msg);
1107 		mmsg.msg_len = *retval;
1108 
1109 		error = copyout(&mmsg, SCARG(uap, mmsg) + dg, sizeof(mmsg));
1110 		if (error)
1111 			break;
1112 
1113 		dg++;
1114 		if (msg->msg_flags & MSG_OOB)
1115 			break;
1116 
1117 		if (SCARG(uap, timeout)) {
1118 			getnanotime(&now);
1119 			if (timespeccmp(&ts, &now, <))
1120 				break;
1121 		}
1122 
1123 		if (flags & MSG_WAITFORONE)
1124 			flags |= MSG_DONTWAIT;
1125 
1126 	}
1127 
1128 	if (from != NULL)
1129 		m_free(from);
1130 
1131 	*retval = dg;
1132 
1133 	/*
1134 	 * If we succeeded at least once, return 0, hopefully so->so_rerror
1135 	 * will catch it next time.
1136 	 */
1137 	if (error && dg > 0) {
1138 		so->so_rerror = error;
1139 		error = 0;
1140 	}
1141 
1142 	fd_putfile(s);
1143 
1144 	return error;
1145 }
1146 
1147 int
sys_shutdown(struct lwp * l,const struct sys_shutdown_args * uap,register_t * retval)1148 sys_shutdown(struct lwp *l, const struct sys_shutdown_args *uap,
1149     register_t *retval)
1150 {
1151 	/* {
1152 		syscallarg(int)	s;
1153 		syscallarg(int)	how;
1154 	} */
1155 	struct socket	*so;
1156 	int		error;
1157 
1158 	if ((error = fd_getsock(SCARG(uap, s), &so)) != 0)
1159 		return error;
1160 	solock(so);
1161 	error = soshutdown(so, SCARG(uap, how));
1162 	sounlock(so);
1163 	fd_putfile(SCARG(uap, s));
1164 	return error;
1165 }
1166 
1167 int
sys_setsockopt(struct lwp * l,const struct sys_setsockopt_args * uap,register_t * retval)1168 sys_setsockopt(struct lwp *l, const struct sys_setsockopt_args *uap,
1169     register_t *retval)
1170 {
1171 	/* {
1172 		syscallarg(int)			s;
1173 		syscallarg(int)			level;
1174 		syscallarg(int)			name;
1175 		syscallarg(const void *)	val;
1176 		syscallarg(unsigned int)	valsize;
1177 	} */
1178 	struct sockopt	sopt;
1179 	struct socket	*so;
1180 	file_t		*fp;
1181 	int		error;
1182 	unsigned int	len;
1183 
1184 	len = SCARG(uap, valsize);
1185 	if (len > 0 && SCARG(uap, val) == NULL)
1186 		return EINVAL;
1187 
1188 	if (len > MCLBYTES)
1189 		return EINVAL;
1190 
1191 	if ((error = fd_getsock1(SCARG(uap, s), &so, &fp)) != 0)
1192 		return (error);
1193 
1194 	sockopt_init(&sopt, SCARG(uap, level), SCARG(uap, name), len);
1195 
1196 	if (len > 0) {
1197 		error = copyin(SCARG(uap, val), sopt.sopt_data, len);
1198 		if (error)
1199 			goto out;
1200 	}
1201 
1202 	error = sosetopt(so, &sopt);
1203 	if (so->so_options & SO_NOSIGPIPE)
1204 		atomic_or_uint(&fp->f_flag, FNOSIGPIPE);
1205 	else
1206 		atomic_and_uint(&fp->f_flag, ~FNOSIGPIPE);
1207 
1208  out:
1209 	sockopt_destroy(&sopt);
1210 	fd_putfile(SCARG(uap, s));
1211 	return error;
1212 }
1213 
1214 static int
getsockopt(struct lwp * l,const struct sys_getsockopt_args * uap,register_t * retval,bool copyarg)1215 getsockopt(struct lwp *l, const struct sys_getsockopt_args *uap,
1216     register_t *retval, bool copyarg)
1217 {
1218 	struct sockopt	sopt;
1219 	struct socket	*so;
1220 	file_t		*fp;
1221 	unsigned int	valsize, len;
1222 	int		error;
1223 
1224 	if (SCARG(uap, val) != NULL) {
1225 		error = copyin(SCARG(uap, avalsize), &valsize, sizeof(valsize));
1226 		if (error)
1227 			return error;
1228 	} else
1229 		valsize = 0;
1230 
1231 	if (valsize > MCLBYTES)
1232 		return EINVAL;
1233 
1234 	if ((error = fd_getsock1(SCARG(uap, s), &so, &fp)) != 0)
1235 		return error;
1236 
1237 	sockopt_init(&sopt, SCARG(uap, level), SCARG(uap, name), valsize);
1238 	if (copyarg && valsize > 0) {
1239 		error = copyin(SCARG(uap, val), sopt.sopt_data, valsize);
1240 		if (error)
1241 			goto out;
1242 	}
1243 
1244 	if (fp->f_flag & FNOSIGPIPE)
1245 		so->so_options |= SO_NOSIGPIPE;
1246 	else
1247 		so->so_options &= ~SO_NOSIGPIPE;
1248 
1249 	error = sogetopt(so, &sopt);
1250 	if (error || valsize == 0)
1251 		goto out;
1252 
1253 	len = uimin(valsize, sopt.sopt_retsize);
1254 	error = copyout(sopt.sopt_data, SCARG(uap, val), len);
1255 	if (error)
1256 		goto out;
1257 
1258 	error = copyout(&len, SCARG(uap, avalsize), sizeof(len));
1259  out:
1260 	sockopt_destroy(&sopt);
1261 	fd_putfile(SCARG(uap, s));
1262 	return error;
1263 }
1264 
1265 int
sys_getsockopt(struct lwp * l,const struct sys_getsockopt_args * uap,register_t * retval)1266 sys_getsockopt(struct lwp *l, const struct sys_getsockopt_args *uap,
1267     register_t *retval)
1268 {
1269 	/* {
1270 		syscallarg(int)			s;
1271 		syscallarg(int)			level;
1272 		syscallarg(int)			name;
1273 		syscallarg(void *)		val;
1274 		syscallarg(unsigned int *)	avalsize;
1275 	} */
1276 	return getsockopt(l, uap, retval, false);
1277 }
1278 
1279 int
sys_getsockopt2(struct lwp * l,const struct sys_getsockopt2_args * uap,register_t * retval)1280 sys_getsockopt2(struct lwp *l, const struct sys_getsockopt2_args *uap,
1281     register_t *retval)
1282 {
1283 	/* {
1284 		syscallarg(int)			s;
1285 		syscallarg(int)			level;
1286 		syscallarg(int)			name;
1287 		syscallarg(void *)		val;
1288 		syscallarg(unsigned int *)	avalsize;
1289 	} */
1290 	return getsockopt(l, (const struct sys_getsockopt_args *) uap, retval, true);
1291 }
1292 
1293 #ifdef PIPE_SOCKETPAIR
1294 
1295 int
pipe1(struct lwp * l,int * fildes,int flags)1296 pipe1(struct lwp *l, int *fildes, int flags)
1297 {
1298 	file_t		*rf, *wf;
1299 	struct socket	*rso, *wso;
1300 	int		fd, error;
1301 	proc_t		*p;
1302 
1303 	if (flags & ~(O_CLOEXEC|O_NONBLOCK|O_NOSIGPIPE))
1304 		return EINVAL;
1305 	p = curproc;
1306 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, l, NULL)) != 0)
1307 		return error;
1308 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, l, rso)) != 0)
1309 		goto free1;
1310 	/* remember this socket pair implements a pipe */
1311 	wso->so_state |= SS_ISAPIPE;
1312 	rso->so_state |= SS_ISAPIPE;
1313 	if ((error = fd_allocfile(&rf, &fd)) != 0)
1314 		goto free2;
1315 	fildes[0] = fd;
1316 	rf->f_flag = FREAD | flags;
1317 	rf->f_type = DTYPE_SOCKET;
1318 	rf->f_ops = &socketops;
1319 	rf->f_socket = rso;
1320 	if ((error = fd_allocfile(&wf, &fd)) != 0)
1321 		goto free3;
1322 	wf->f_flag = FWRITE | flags;
1323 	wf->f_type = DTYPE_SOCKET;
1324 	wf->f_ops = &socketops;
1325 	wf->f_socket = wso;
1326 	fildes[1] = fd;
1327 	solock(wso);
1328 	/*
1329 	 * Pipes must be readable when there is at least 1
1330 	 * byte of data available in the receive buffer.
1331 	 *
1332 	 * Pipes must be writable when there is space for
1333 	 * at least PIPE_BUF bytes in the send buffer.
1334 	 * If we're increasing the low water mark for the
1335 	 * send buffer, then mimic how soreserve() would
1336 	 * have set the high water mark.
1337 	 */
1338 	rso->so_rcv.sb_lowat = 1;
1339 	if (wso->so_snd.sb_lowat < PIPE_BUF) {
1340 		wso->so_snd.sb_hiwat = PIPE_BUF * 2;
1341 	}
1342 	wso->so_snd.sb_lowat = PIPE_BUF;
1343 	error = unp_connect2(wso, rso);
1344 	sounlock(wso);
1345 	if (error != 0)
1346 		goto free4;
1347 	fd_affix(p, wf, fildes[1]);
1348 	fd_affix(p, rf, fildes[0]);
1349 	return (0);
1350  free4:
1351 	fd_abort(p, wf, fildes[1]);
1352  free3:
1353 	fd_abort(p, rf, fildes[0]);
1354  free2:
1355 	(void)soclose(wso);
1356  free1:
1357 	(void)soclose(rso);
1358 	return error;
1359 }
1360 #endif /* PIPE_SOCKETPAIR */
1361 
1362 /*
1363  * Get peer socket name.
1364  */
1365 int
do_sys_getpeername(int fd,struct sockaddr * nam)1366 do_sys_getpeername(int fd, struct sockaddr *nam)
1367 {
1368 	struct socket	*so;
1369 	int		error;
1370 
1371 	if ((error = fd_getsock(fd, &so)) != 0)
1372 		return error;
1373 
1374 	solock(so);
1375 	if ((so->so_state & SS_ISCONNECTED) == 0)
1376 		error = ENOTCONN;
1377 	else {
1378 		error = (*so->so_proto->pr_usrreqs->pr_peeraddr)(so, nam);
1379 	}
1380 	sounlock(so);
1381 	fd_putfile(fd);
1382 	return error;
1383 }
1384 
1385 /*
1386  * Get local socket name.
1387  */
1388 int
do_sys_getsockname(int fd,struct sockaddr * nam)1389 do_sys_getsockname(int fd, struct sockaddr *nam)
1390 {
1391 	struct socket	*so;
1392 	int		error;
1393 
1394 	if ((error = fd_getsock(fd, &so)) != 0)
1395 		return error;
1396 
1397 	solock(so);
1398 	error = (*so->so_proto->pr_usrreqs->pr_sockaddr)(so, nam);
1399 	sounlock(so);
1400 	fd_putfile(fd);
1401 	return error;
1402 }
1403 
1404 int
copyout_sockname_sb(struct sockaddr * asa,unsigned int * alen,int flags,struct sockaddr_big * addr)1405 copyout_sockname_sb(struct sockaddr *asa, unsigned int *alen, int flags,
1406     struct sockaddr_big *addr)
1407 {
1408 	unsigned int len;
1409 	int error;
1410 
1411 	if (asa == NULL)
1412 		/* Assume application not interested */
1413 		return 0;
1414 
1415 	if (flags & MSG_LENUSRSPACE) {
1416 		error = copyin(alen, &len, sizeof(len));
1417 		if (error)
1418 			return error;
1419 	} else
1420 		len = *alen;
1421 
1422 	if (addr == NULL) {
1423 		len = 0;
1424 		error = 0;
1425 	} else {
1426 		if (len > addr->sb_len)
1427 			len = addr->sb_len;
1428 		/* XXX addr isn't an mbuf... */
1429 		ktrkuser(mbuftypes[MT_SONAME], addr, len);
1430 		error = copyout(addr, asa, len);
1431 	}
1432 
1433 	if (error == 0) {
1434 		if (flags & MSG_LENUSRSPACE)
1435 			error = copyout(&len, alen, sizeof(len));
1436 		else
1437 			*alen = len;
1438 	}
1439 
1440 	return error;
1441 }
1442 
1443 int
copyout_sockname(struct sockaddr * asa,unsigned int * alen,int flags,struct mbuf * addr)1444 copyout_sockname(struct sockaddr *asa, unsigned int *alen, int flags,
1445     struct mbuf *addr)
1446 {
1447 	int len;
1448 	int error;
1449 
1450 	if (asa == NULL)
1451 		/* Assume application not interested */
1452 		return 0;
1453 
1454 	if (flags & MSG_LENUSRSPACE) {
1455 		error = copyin(alen, &len, sizeof(len));
1456 		if (error)
1457 			return error;
1458 	} else
1459 		len = *alen;
1460 	if (len < 0)
1461 		return EINVAL;
1462 
1463 	if (addr == NULL) {
1464 		len = 0;
1465 		error = 0;
1466 	} else {
1467 		if (len > addr->m_len)
1468 			len = addr->m_len;
1469 		/* Maybe this ought to copy a chain ? */
1470 		ktrkuser(mbuftypes[MT_SONAME], mtod(addr, void *), len);
1471 		error = copyout(mtod(addr, void *), asa, len);
1472 	}
1473 
1474 	if (error == 0) {
1475 		if (flags & MSG_LENUSRSPACE)
1476 			error = copyout(&len, alen, sizeof(len));
1477 		else
1478 			*alen = len;
1479 	}
1480 
1481 	return error;
1482 }
1483 
1484 /*
1485  * Get socket name.
1486  */
1487 int
sys_getsockname(struct lwp * l,const struct sys_getsockname_args * uap,register_t * retval)1488 sys_getsockname(struct lwp *l, const struct sys_getsockname_args *uap,
1489     register_t *retval)
1490 {
1491 	/* {
1492 		syscallarg(int)			fdes;
1493 		syscallarg(struct sockaddr *)	asa;
1494 		syscallarg(unsigned int *)	alen;
1495 	} */
1496 	struct sockaddr_big sbig;
1497 	int		    error;
1498 
1499 	sbig.sb_len = UCHAR_MAX;
1500 	error = do_sys_getsockname(SCARG(uap, fdes), (struct sockaddr *)&sbig);
1501 	if (error != 0)
1502 		return error;
1503 
1504 	error = copyout_sockname_sb(SCARG(uap, asa), SCARG(uap, alen),
1505 	    MSG_LENUSRSPACE, &sbig);
1506 	return error;
1507 }
1508 
1509 /*
1510  * Get name of peer for connected socket.
1511  */
1512 int
sys_getpeername(struct lwp * l,const struct sys_getpeername_args * uap,register_t * retval)1513 sys_getpeername(struct lwp *l, const struct sys_getpeername_args *uap,
1514     register_t *retval)
1515 {
1516 	/* {
1517 		syscallarg(int)			fdes;
1518 		syscallarg(struct sockaddr *)	asa;
1519 		syscallarg(unsigned int *)	alen;
1520 	} */
1521 	struct sockaddr_big sbig;
1522 	int		    error;
1523 
1524 	sbig.sb_len = UCHAR_MAX;
1525 	error = do_sys_getpeername(SCARG(uap, fdes), (struct sockaddr *)&sbig);
1526 	if (error != 0)
1527 		return error;
1528 
1529 	error = copyout_sockname_sb(SCARG(uap, asa), SCARG(uap, alen),
1530 	    MSG_LENUSRSPACE, &sbig);
1531 	return error;
1532 }
1533 
1534 static int
sockargs_sb(struct sockaddr_big * sb,const void * name,socklen_t buflen)1535 sockargs_sb(struct sockaddr_big *sb, const void *name, socklen_t buflen)
1536 {
1537 	int error;
1538 
1539 	/*
1540 	 * We can't allow socket names > UCHAR_MAX in length, since that
1541 	 * will overflow sb_len. Further no reasonable buflen is <=
1542 	 * offsetof(sockaddr_big, sb_data) since it shall be at least
1543 	 * the size of the preamble sb_len and sb_family members.
1544 	 */
1545 	if (buflen > UCHAR_MAX ||
1546 	    buflen <= offsetof(struct sockaddr_big, sb_data))
1547 		return EINVAL;
1548 
1549 	error = copyin(name, (void *)sb, buflen);
1550 	if (error)
1551 		return error;
1552 
1553 	ktrkuser(mbuftypes[MT_SONAME], sb, buflen);
1554 #if BYTE_ORDER != BIG_ENDIAN
1555 	/*
1556 	 * 4.3BSD compat thing - need to stay, since bind(2),
1557 	 * connect(2), sendto(2) were not versioned for COMPAT_43.
1558 	 */
1559 	if (sb->sb_family == 0 && sb->sb_len < AF_MAX)
1560 		sb->sb_family = sb->sb_len;
1561 #endif
1562 	sb->sb_len = buflen;
1563 	return 0;
1564 }
1565 
1566 /*
1567  * XXX In a perfect world, we wouldn't pass around socket control
1568  * XXX arguments in mbufs, and this could go away.
1569  */
1570 int
sockargs(struct mbuf ** mp,const void * bf,size_t buflen,enum uio_seg seg,int type)1571 sockargs(struct mbuf **mp, const void *bf, size_t buflen, enum uio_seg seg,
1572     int type)
1573 {
1574 	struct mbuf	*m;
1575 	int		error;
1576 
1577 	/*
1578 	 * We can't allow socket names > UCHAR_MAX in length, since that
1579 	 * will overflow sa_len.  Control data more than a page size in
1580 	 * length is just too much.
1581 	 */
1582 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
1583 		return EINVAL;
1584 
1585 	/*
1586 	 * length must greater than sizeof(sa_family) + sizeof(sa_len)
1587 	 */
1588 	if (type == MT_SONAME && buflen <= 2)
1589 		return EINVAL;
1590 
1591 	/* Allocate an mbuf to hold the arguments. */
1592 	m = m_get(M_WAIT, type);
1593 	/* can't claim.  don't who to assign it to. */
1594 	if (buflen > MLEN) {
1595 		/*
1596 		 * Won't fit into a regular mbuf, so we allocate just
1597 		 * enough external storage to hold the argument.
1598 		 */
1599 		MEXTMALLOC(m, buflen, M_WAITOK);
1600 	}
1601 	m->m_len = buflen;
1602 	if (seg == UIO_USERSPACE) {
1603 		error = copyin(bf, mtod(m, void *), buflen);
1604 		if (error) {
1605 			(void)m_free(m);
1606 			return error;
1607 		}
1608 	} else {
1609 		memcpy(mtod(m, void *), bf, buflen);
1610 	}
1611 	*mp = m;
1612 	switch (type) {
1613 	case MT_SONAME:
1614 		ktrkuser(mbuftypes[type], mtod(m, void *), buflen);
1615 
1616 		struct sockaddr *sa = mtod(m, struct sockaddr *);
1617 #if BYTE_ORDER != BIG_ENDIAN
1618 		/*
1619 		 * 4.3BSD compat thing - need to stay, since bind(2),
1620 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
1621 		 */
1622 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1623 			sa->sa_family = sa->sa_len;
1624 #endif
1625 		sa->sa_len = buflen;
1626 		return 0;
1627 	case MT_CONTROL:
1628 		if (!KTRPOINT(curproc, KTR_USER))
1629 			return 0;
1630 
1631 		struct msghdr mhdr;
1632 		mhdr.msg_control = mtod(m, void *);
1633 		mhdr.msg_controllen = buflen;
1634 		for (struct cmsghdr *cmsg = CMSG_FIRSTHDR(&mhdr); cmsg;
1635 		    cmsg = CMSG_NXTHDR(&mhdr, cmsg)) {
1636 			KASSERT(((char *)cmsg - mtod(m, char *)) <= buflen);
1637 			if (cmsg->cmsg_len >
1638 			    buflen - ((char *)cmsg - mtod(m, char *)))
1639 				break;
1640 			ktrkuser(mbuftypes[type], cmsg, cmsg->cmsg_len);
1641 		}
1642 		return 0;
1643 	default:
1644 		return EINVAL;
1645 	}
1646 }
1647 
1648 int
do_sys_peeloff(struct socket * head,void * data)1649 do_sys_peeloff(struct socket *head, void *data)
1650 {
1651 #ifdef SCTP
1652 	/*file_t *lfp = NULL;*/
1653 	file_t *nfp = NULL;
1654 	int error;
1655 	struct socket *so;
1656 	int fd;
1657 	uint32_t name;
1658 	/*short fflag;*/		/* type must match fp->f_flag */
1659 
1660 	name = *(uint32_t *) data;
1661 	error = sctp_can_peel_off(head, name);
1662 	if (error) {
1663 		printf("peeloff failed\n");
1664 		return error;
1665 	}
1666 	/*
1667 	 * At this point we know we do have a assoc to pull
1668 	 * we proceed to get the fd setup. This may block
1669 	 * but that is ok.
1670 	 */
1671 	error = fd_allocfile(&nfp, &fd);
1672 	if (error) {
1673 		/*
1674 		 * Probably ran out of file descriptors. Put the
1675 		 * unaccepted connection back onto the queue and
1676 		 * do another wakeup so some other process might
1677 		 * have a chance at it.
1678 		 */
1679 		return error;
1680 	}
1681 	*(int *) data = fd;
1682 
1683 	so = sctp_get_peeloff(head, name, &error);
1684 	if (so == NULL) {
1685 		/*
1686 		 * Either someone else peeled it off OR
1687 		 * we can't get a socket.
1688 		 * close the new descriptor, assuming someone hasn't ripped it
1689 		 * out from under us.
1690 		 */
1691 		mutex_enter(&nfp->f_lock);
1692 		nfp->f_count++;
1693 		mutex_exit(&nfp->f_lock);
1694 		fd_abort(curlwp->l_proc, nfp, fd);
1695 		return error;
1696 	}
1697 	so->so_state &= ~SS_NOFDREF;
1698 	so->so_state &= ~SS_ISCONNECTING;
1699 	so->so_head = NULL;
1700 	so->so_cred = kauth_cred_dup(head->so_cred);
1701 	nfp->f_socket = so;
1702 	nfp->f_flag = FREAD|FWRITE;
1703 	nfp->f_ops = &socketops;
1704 	nfp->f_type = DTYPE_SOCKET;
1705 
1706 	fd_affix(curlwp->l_proc, nfp, fd);
1707 
1708 	return error;
1709 #else
1710 	return EOPNOTSUPP;
1711 #endif
1712 }
1713