xref: /netbsd-src/sys/kern/uipc_syscalls.c (revision 2e2322c9c07009df921d11b1268f8506affbb8ba)
1 /*	$NetBSD: uipc_syscalls.c,v 1.184 2016/12/03 22:28:16 christos 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 1.184 2016/12/03 22:28:16 christos Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_pipe.h"
68 #endif
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/filedesc.h>
73 #include <sys/proc.h>
74 #include <sys/file.h>
75 #include <sys/buf.h>
76 #define MBUFTYPES
77 #include <sys/mbuf.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/signalvar.h>
82 #include <sys/un.h>
83 #include <sys/ktrace.h>
84 #include <sys/event.h>
85 #include <sys/atomic.h>
86 #include <sys/kauth.h>
87 
88 #include <sys/mount.h>
89 #include <sys/syscallargs.h>
90 
91 /*
92  * System call interface to the socket abstraction.
93  */
94 extern const struct fileops socketops;
95 
96 static int	sockargs_sb(struct sockaddr_big *, const void *, socklen_t);
97 static int	copyout_sockname_sb(struct sockaddr *, unsigned int *,
98 		    int , struct sockaddr_big *);
99 
100 int
101 sys___socket30(struct lwp *l, const struct sys___socket30_args *uap,
102     register_t *retval)
103 {
104 	/* {
105 		syscallarg(int)	domain;
106 		syscallarg(int)	type;
107 		syscallarg(int)	protocol;
108 	} */
109 	int fd, error;
110 
111 	error = fsocreate(SCARG(uap, domain), NULL, SCARG(uap, type),
112 	    SCARG(uap, protocol), &fd);
113 	if (error == 0) {
114 		*retval = fd;
115 	}
116 	return error;
117 }
118 
119 int
120 sys_bind(struct lwp *l, const struct sys_bind_args *uap, register_t *retval)
121 {
122 	/* {
123 		syscallarg(int)				s;
124 		syscallarg(const struct sockaddr *)	name;
125 		syscallarg(unsigned int)		namelen;
126 	} */
127 	int		error;
128 	struct sockaddr_big sb;
129 
130 	error = sockargs_sb(&sb, SCARG(uap, name), SCARG(uap, namelen));
131 	if (error)
132 		return error;
133 
134 	return do_sys_bind(l, SCARG(uap, s), (struct sockaddr *)&sb);
135 }
136 
137 int
138 do_sys_bind(struct lwp *l, int fd, struct sockaddr *nam)
139 {
140 	struct socket	*so;
141 	int		error;
142 
143 	if ((error = fd_getsock(fd, &so)) != 0)
144 		return error;
145 	error = sobind(so, nam, l);
146 	fd_putfile(fd);
147 	return error;
148 }
149 
150 int
151 sys_listen(struct lwp *l, const struct sys_listen_args *uap, register_t *retval)
152 {
153 	/* {
154 		syscallarg(int)	s;
155 		syscallarg(int)	backlog;
156 	} */
157 	struct socket	*so;
158 	int		error;
159 
160 	if ((error = fd_getsock(SCARG(uap, s), &so)) != 0)
161 		return (error);
162 	error = solisten(so, SCARG(uap, backlog), l);
163 	fd_putfile(SCARG(uap, s));
164 	return error;
165 }
166 
167 int
168 do_sys_accept(struct lwp *l, int sock, struct sockaddr *name,
169     register_t *new_sock, const sigset_t *mask, int flags, int clrflags)
170 {
171 	file_t		*fp, *fp2;
172 	int		error, fd;
173 	struct socket	*so, *so2;
174 	short		wakeup_state = 0;
175 
176 	if ((fp = fd_getfile(sock)) == NULL)
177 		return EBADF;
178 	if (fp->f_type != DTYPE_SOCKET) {
179 		fd_putfile(sock);
180 		return ENOTSOCK;
181 	}
182 	if ((error = fd_allocfile(&fp2, &fd)) != 0) {
183 		fd_putfile(sock);
184 		return error;
185 	}
186 	*new_sock = fd;
187 	so = fp->f_socket;
188 	solock(so);
189 
190 	if (__predict_false(mask))
191 		sigsuspendsetup(l, mask);
192 
193 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
194 		error = EOPNOTSUPP;
195 		goto bad;
196 	}
197 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
198 		error = EINVAL;
199 		goto bad;
200 	}
201 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
202 		error = EWOULDBLOCK;
203 		goto bad;
204 	}
205 	while (so->so_qlen == 0 && so->so_error == 0) {
206 		if (so->so_state & SS_CANTRCVMORE) {
207 			so->so_error = ECONNABORTED;
208 			break;
209 		}
210 		if (wakeup_state & SS_RESTARTSYS) {
211 			error = ERESTART;
212 			goto bad;
213 		}
214 		error = sowait(so, true, 0);
215 		if (error) {
216 			goto bad;
217 		}
218 		wakeup_state = so->so_state;
219 	}
220 	if (so->so_error) {
221 		error = so->so_error;
222 		so->so_error = 0;
223 		goto bad;
224 	}
225 	/* connection has been removed from the listen queue */
226 	KNOTE(&so->so_rcv.sb_sel.sel_klist, NOTE_SUBMIT);
227 	so2 = TAILQ_FIRST(&so->so_q);
228 	if (soqremque(so2, 1) == 0)
229 		panic("accept");
230 	fp2->f_type = DTYPE_SOCKET;
231 	fp2->f_flag = (fp->f_flag & ~clrflags) |
232 	    ((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)|
233 	    ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0);
234 	fp2->f_ops = &socketops;
235 	fp2->f_socket = so2;
236 	if (fp2->f_flag & FNONBLOCK)
237 		so2->so_state |= SS_NBIO;
238 	else
239 		so2->so_state &= ~SS_NBIO;
240 	error = soaccept(so2, name);
241 	so2->so_cred = kauth_cred_dup(so->so_cred);
242 	sounlock(so);
243 	if (error) {
244 		/* an error occurred, free the file descriptor and mbuf */
245 		mutex_enter(&fp2->f_lock);
246 		fp2->f_count++;
247 		mutex_exit(&fp2->f_lock);
248 		closef(fp2);
249 		fd_abort(curproc, NULL, fd);
250 	} else {
251 		fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0);
252 		fd_affix(curproc, fp2, fd);
253 	}
254 	fd_putfile(sock);
255 	if (__predict_false(mask))
256 		sigsuspendteardown(l);
257 	return error;
258  bad:
259 	sounlock(so);
260 	fd_putfile(sock);
261 	fd_abort(curproc, fp2, fd);
262 	if (__predict_false(mask))
263 		sigsuspendteardown(l);
264 	return error;
265 }
266 
267 int
268 sys_accept(struct lwp *l, const struct sys_accept_args *uap, register_t *retval)
269 {
270 	/* {
271 		syscallarg(int)			s;
272 		syscallarg(struct sockaddr *)	name;
273 		syscallarg(unsigned int *)	anamelen;
274 	} */
275 	int error, fd;
276 	struct sockaddr_big name;
277 
278 	name.sb_len = UCHAR_MAX;
279 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
280 	    retval, NULL, 0, 0);
281 	if (error != 0)
282 		return error;
283 	error = copyout_sockname_sb(SCARG(uap, name), SCARG(uap, anamelen),
284 	    MSG_LENUSRSPACE, &name);
285 	if (error != 0) {
286 		fd = (int)*retval;
287 		if (fd_getfile(fd) != NULL)
288 			(void)fd_close(fd);
289 	}
290 	return error;
291 }
292 
293 int
294 sys_paccept(struct lwp *l, const struct sys_paccept_args *uap,
295     register_t *retval)
296 {
297 	/* {
298 		syscallarg(int)			s;
299 		syscallarg(struct sockaddr *)	name;
300 		syscallarg(unsigned int *)	anamelen;
301 		syscallarg(const sigset_t *)	mask;
302 		syscallarg(int)			flags;
303 	} */
304 	int error, fd;
305 	struct sockaddr_big name;
306 	sigset_t *mask, amask;
307 
308 	if (SCARG(uap, mask) != NULL) {
309 		error = copyin(SCARG(uap, mask), &amask, sizeof(amask));
310 		if (error)
311 			return error;
312 		mask = &amask;
313 	} else
314 		mask = NULL;
315 
316 	name.sb_len = UCHAR_MAX;
317 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
318 	    retval, mask, SCARG(uap, flags), FNONBLOCK);
319 	if (error != 0)
320 		return error;
321 	error = copyout_sockname_sb(SCARG(uap, name), SCARG(uap, anamelen),
322 	    MSG_LENUSRSPACE, &name);
323 	if (error != 0) {
324 		fd = (int)*retval;
325 		if (fd_getfile(fd) != NULL)
326 			(void)fd_close(fd);
327 	}
328 	return error;
329 }
330 
331 int
332 sys_connect(struct lwp *l, const struct sys_connect_args *uap,
333     register_t *retval)
334 {
335 	/* {
336 		syscallarg(int)				s;
337 		syscallarg(const struct sockaddr *)	name;
338 		syscallarg(unsigned int)		namelen;
339 	} */
340 	int		error;
341 	struct sockaddr_big sbig;
342 
343 	error = sockargs_sb(&sbig, SCARG(uap, name), SCARG(uap, namelen));
344 	if (error)
345 		return error;
346 	return do_sys_connect(l, SCARG(uap, s), (struct sockaddr *)&sbig);
347 }
348 
349 int
350 do_sys_connect(struct lwp *l, int fd, struct sockaddr *nam)
351 {
352 	struct socket	*so;
353 	int		error;
354 	int		interrupted = 0;
355 
356 	if ((error = fd_getsock(fd, &so)) != 0) {
357 		return (error);
358 	}
359 	solock(so);
360 	if ((so->so_state & SS_ISCONNECTING) != 0) {
361 		error = EALREADY;
362 		goto out;
363 	}
364 
365 	error = soconnect(so, nam, l);
366 	if (error)
367 		goto bad;
368 	if ((so->so_state & (SS_NBIO|SS_ISCONNECTING)) ==
369 	    (SS_NBIO|SS_ISCONNECTING)) {
370 		error = EINPROGRESS;
371 		goto out;
372 	}
373 	while ((so->so_state & SS_ISCONNECTING) != 0 && so->so_error == 0) {
374 		error = sowait(so, true, 0);
375 		if (__predict_false((so->so_state & SS_ISABORTING) != 0)) {
376 			error = EPIPE;
377 			interrupted = 1;
378 			break;
379 		}
380 		if (error) {
381 			if (error == EINTR || error == ERESTART)
382 				interrupted = 1;
383 			break;
384 		}
385 	}
386 	if (error == 0) {
387 		error = so->so_error;
388 		so->so_error = 0;
389 	}
390  bad:
391 	if (!interrupted)
392 		so->so_state &= ~SS_ISCONNECTING;
393 	if (error == ERESTART)
394 		error = EINTR;
395  out:
396 	sounlock(so);
397 	fd_putfile(fd);
398 	return error;
399 }
400 
401 static int
402 makesocket(struct lwp *l, file_t **fp, int *fd, int flags, int type,
403     int domain, int proto, struct socket *soo)
404 {
405 	struct socket *so;
406 	int error;
407 
408 	if ((error = socreate(domain, &so, type, proto, l, soo)) != 0) {
409 		return error;
410 	}
411 	if (flags & SOCK_NONBLOCK) {
412 		so->so_state |= SS_NBIO;
413 	}
414 
415 	if ((error = fd_allocfile(fp, fd)) != 0) {
416 		soclose(so);
417 		return error;
418 	}
419 	fd_set_exclose(l, *fd, (flags & SOCK_CLOEXEC) != 0);
420 	(*fp)->f_flag = FREAD|FWRITE|
421 	    ((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)|
422 	    ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0);
423 	(*fp)->f_type = DTYPE_SOCKET;
424 	(*fp)->f_ops = &socketops;
425 	(*fp)->f_socket = so;
426 	return 0;
427 }
428 
429 int
430 sys_socketpair(struct lwp *l, const struct sys_socketpair_args *uap,
431     register_t *retval)
432 {
433 	/* {
434 		syscallarg(int)		domain;
435 		syscallarg(int)		type;
436 		syscallarg(int)		protocol;
437 		syscallarg(int *)	rsv;
438 	} */
439 	file_t		*fp1, *fp2;
440 	struct socket	*so1, *so2;
441 	int		fd, error, sv[2];
442 	proc_t		*p = curproc;
443 	int		flags = SCARG(uap, type) & SOCK_FLAGS_MASK;
444 	int		type = SCARG(uap, type) & ~SOCK_FLAGS_MASK;
445 	int		domain = SCARG(uap, domain);
446 	int		proto = SCARG(uap, protocol);
447 
448 	error = makesocket(l, &fp1, &fd, flags, type, domain, proto, NULL);
449 	if (error)
450 		return error;
451 	so1 = fp1->f_socket;
452 	sv[0] = fd;
453 
454 	error = makesocket(l, &fp2, &fd, flags, type, domain, proto, so1);
455 	if (error)
456 		goto out;
457 	so2 = fp2->f_socket;
458 	sv[1] = fd;
459 
460 	solock(so1);
461 	error = soconnect2(so1, so2);
462 	if (error == 0 && type == SOCK_DGRAM) {
463 		/*
464 		 * Datagram socket connection is asymmetric.
465 		 */
466 		error = soconnect2(so2, so1);
467 	}
468 	sounlock(so1);
469 
470 	if (error == 0)
471 		error = copyout(sv, SCARG(uap, rsv), sizeof(sv));
472 	if (error == 0) {
473 		fd_affix(p, fp2, sv[1]);
474 		fd_affix(p, fp1, sv[0]);
475 		return 0;
476 	}
477 	fd_abort(p, fp2, sv[1]);
478 	(void)soclose(so2);
479 out:
480 	fd_abort(p, fp1, sv[0]);
481 	(void)soclose(so1);
482 	return error;
483 }
484 
485 int
486 sys_sendto(struct lwp *l, const struct sys_sendto_args *uap,
487     register_t *retval)
488 {
489 	/* {
490 		syscallarg(int)				s;
491 		syscallarg(const void *)		buf;
492 		syscallarg(size_t)			len;
493 		syscallarg(int)				flags;
494 		syscallarg(const struct sockaddr *)	to;
495 		syscallarg(unsigned int)		tolen;
496 	} */
497 	struct msghdr	msg;
498 	struct iovec	aiov;
499 
500 	msg.msg_name = __UNCONST(SCARG(uap, to)); /* XXXUNCONST kills const */
501 	msg.msg_namelen = SCARG(uap, tolen);
502 	msg.msg_iov = &aiov;
503 	msg.msg_iovlen = 1;
504 	msg.msg_control = NULL;
505 	msg.msg_flags = 0;
506 	aiov.iov_base = __UNCONST(SCARG(uap, buf)); /* XXXUNCONST kills const */
507 	aiov.iov_len = SCARG(uap, len);
508 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, SCARG(uap, flags),
509 	    NULL, 0, retval);
510 }
511 
512 int
513 sys_sendmsg(struct lwp *l, const struct sys_sendmsg_args *uap,
514     register_t *retval)
515 {
516 	/* {
517 		syscallarg(int)				s;
518 		syscallarg(const struct msghdr *)	msg;
519 		syscallarg(int)				flags;
520 	} */
521 	struct msghdr	msg;
522 	int		error;
523 
524 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
525 	if (error)
526 		return (error);
527 
528 	msg.msg_flags = MSG_IOVUSRSPACE;
529 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, SCARG(uap, flags),
530 	    NULL, 0, retval);
531 }
532 
533 static int
534 do_sys_sendmsg_so(struct lwp *l, int s, struct socket *so, file_t *fp,
535     struct msghdr *mp, int flags, const void *kthdr, size_t ktsize,
536     register_t *retsize)
537 {
538 
539 	struct iovec	aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
540 	struct sockaddr *sa = NULL;
541 	struct mbuf	*to, *control;
542 	struct uio	auio;
543 	size_t		len, iovsz;
544 	int		i, error;
545 
546 	if (__predict_false(kthdr == NULL && ktsize == 0)) {
547 		kthdr = mp;
548 		ktsize = sizeof(*mp);
549 	}
550 	if (__predict_true(kthdr != NULL))
551 		ktrkuser("msghdr", kthdr, ktsize);
552 
553 	/* If the caller passed us stuff in mbufs, we must free them. */
554 	to = (mp->msg_flags & MSG_NAMEMBUF) ? mp->msg_name : NULL;
555 	control = (mp->msg_flags & MSG_CONTROLMBUF) ? mp->msg_control : NULL;
556 	iovsz = mp->msg_iovlen * sizeof(struct iovec);
557 
558 	if (mp->msg_flags & MSG_IOVUSRSPACE) {
559 		if ((unsigned int)mp->msg_iovlen > UIO_SMALLIOV) {
560 			if ((unsigned int)mp->msg_iovlen > IOV_MAX) {
561 				error = EMSGSIZE;
562 				goto bad;
563 			}
564 			iov = kmem_alloc(iovsz, KM_SLEEP);
565 		}
566 		if (mp->msg_iovlen != 0) {
567 			error = copyin(mp->msg_iov, iov, iovsz);
568 			if (error)
569 				goto bad;
570 		}
571 		auio.uio_iov = iov;
572 	} else
573 		auio.uio_iov = mp->msg_iov;
574 
575 	auio.uio_iovcnt = mp->msg_iovlen;
576 	auio.uio_rw = UIO_WRITE;
577 	auio.uio_offset = 0;			/* XXX */
578 	auio.uio_resid = 0;
579 	KASSERT(l == curlwp);
580 	auio.uio_vmspace = l->l_proc->p_vmspace;
581 
582 	tiov = auio.uio_iov;
583 	for (i = 0; i < auio.uio_iovcnt; i++, tiov++) {
584 		/*
585 		 * Writes return ssize_t because -1 is returned on error.
586 		 * Therefore, we must restrict the length to SSIZE_MAX to
587 		 * avoid garbage return values.
588 		 */
589 		auio.uio_resid += tiov->iov_len;
590 		if (tiov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
591 			error = EINVAL;
592 			goto bad;
593 		}
594 	}
595 
596 	if (mp->msg_name && to == NULL) {
597 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
598 		    MT_SONAME);
599 		if (error)
600 			goto bad;
601 	}
602 
603 	if (mp->msg_control) {
604 		if (mp->msg_controllen < CMSG_ALIGN(sizeof(struct cmsghdr))) {
605 			error = EINVAL;
606 			goto bad;
607 		}
608 		if (control == NULL) {
609 			error = sockargs(&control, mp->msg_control,
610 			    mp->msg_controllen, MT_CONTROL);
611 			if (error)
612 				goto bad;
613 		}
614 	}
615 
616 	if (ktrpoint(KTR_GENIO) && iovsz > 0) {
617 		ktriov = kmem_alloc(iovsz, KM_SLEEP);
618 		memcpy(ktriov, auio.uio_iov, iovsz);
619 	}
620 
621 	if (mp->msg_name)
622 		MCLAIM(to, so->so_mowner);
623 	if (mp->msg_control)
624 		MCLAIM(control, so->so_mowner);
625 
626 	if (to) {
627 		sa = mtod(to, struct sockaddr *);
628 	}
629 
630 	len = auio.uio_resid;
631 	error = (*so->so_send)(so, sa, &auio, NULL, control, flags, l);
632 	/* Protocol is responsible for freeing 'control' */
633 	control = NULL;
634 
635 	if (error) {
636 		if (auio.uio_resid != len && (error == ERESTART ||
637 		    error == EINTR || error == EWOULDBLOCK))
638 			error = 0;
639 		if (error == EPIPE && (fp->f_flag & FNOSIGPIPE) == 0 &&
640 		    (flags & MSG_NOSIGNAL) == 0) {
641 			mutex_enter(proc_lock);
642 			psignal(l->l_proc, SIGPIPE);
643 			mutex_exit(proc_lock);
644 		}
645 	}
646 	if (error == 0)
647 		*retsize = len - auio.uio_resid;
648 
649 bad:
650 	if (ktriov != NULL) {
651 		ktrgeniov(s, UIO_WRITE, ktriov, *retsize, error);
652 		kmem_free(ktriov, iovsz);
653 	}
654 
655 	if (iov != aiov)
656 		kmem_free(iov, iovsz);
657 	if (to)
658 		m_freem(to);
659 	if (control)
660 		m_freem(control);
661 
662 	return error;
663 }
664 
665 int
666 do_sys_sendmsg(struct lwp *l, int s, struct msghdr *mp, int flags,
667     const void *kthdr, size_t ktsize, register_t *retsize)
668 {
669 	int		error;
670 	struct socket	*so;
671 	file_t		*fp;
672 
673 	if ((error = fd_getsock1(s, &so, &fp)) != 0) {
674 		/* We have to free msg_name and msg_control ourselves */
675 		if (mp->msg_flags & MSG_NAMEMBUF)
676 			m_freem(mp->msg_name);
677 		if (mp->msg_flags & MSG_CONTROLMBUF)
678 			m_freem(mp->msg_control);
679 		return error;
680 	}
681 	error = do_sys_sendmsg_so(l, s, so, fp, mp, flags, kthdr, ktsize,
682 	    retsize);
683 	/* msg_name and msg_control freed */
684 	fd_putfile(s);
685 	return error;
686 }
687 
688 int
689 sys_recvfrom(struct lwp *l, const struct sys_recvfrom_args *uap,
690     register_t *retval)
691 {
692 	/* {
693 		syscallarg(int)			s;
694 		syscallarg(void *)		buf;
695 		syscallarg(size_t)		len;
696 		syscallarg(int)			flags;
697 		syscallarg(struct sockaddr *)	from;
698 		syscallarg(unsigned int *)	fromlenaddr;
699 	} */
700 	struct msghdr	msg;
701 	struct iovec	aiov;
702 	int		error;
703 	struct mbuf	*from;
704 
705 	msg.msg_name = NULL;
706 	msg.msg_iov = &aiov;
707 	msg.msg_iovlen = 1;
708 	aiov.iov_base = SCARG(uap, buf);
709 	aiov.iov_len = SCARG(uap, len);
710 	msg.msg_control = NULL;
711 	msg.msg_flags = SCARG(uap, flags) & MSG_USERFLAGS;
712 
713 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, NULL, 0, &from,
714 	    NULL, retval);
715 	if (error != 0)
716 		return error;
717 
718 	error = copyout_sockname(SCARG(uap, from), SCARG(uap, fromlenaddr),
719 	    MSG_LENUSRSPACE, from);
720 	if (from != NULL)
721 		m_free(from);
722 	return error;
723 }
724 
725 int
726 sys_recvmsg(struct lwp *l, const struct sys_recvmsg_args *uap,
727     register_t *retval)
728 {
729 	/* {
730 		syscallarg(int)			s;
731 		syscallarg(struct msghdr *)	msg;
732 		syscallarg(int)			flags;
733 	} */
734 	struct msghdr	msg;
735 	int		error;
736 	struct mbuf	*from, *control;
737 
738 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
739 	if (error)
740 		return error;
741 
742 	msg.msg_flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
743 
744 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, NULL, 0, &from,
745 	    msg.msg_control != NULL ? &control : NULL, retval);
746 	if (error != 0)
747 		return error;
748 
749 	if (msg.msg_control != NULL)
750 		error = copyout_msg_control(l, &msg, control);
751 
752 	if (error == 0)
753 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
754 			from);
755 	if (from != NULL)
756 		m_free(from);
757 	if (error == 0) {
758 		ktrkuser("msghdr", &msg, sizeof msg);
759 		error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
760 	}
761 
762 	return error;
763 }
764 
765 int
766 sys_sendmmsg(struct lwp *l, const struct sys_sendmmsg_args *uap,
767     register_t *retval)
768 {
769 	/* {
770 		syscallarg(int)			s;
771 		syscallarg(struct mmsghdr *)	mmsg;
772 		syscallarg(unsigned int)	vlen;
773 		syscallarg(unsigned int)	flags;
774 	} */
775 	struct mmsghdr mmsg;
776 	struct socket *so;
777 	file_t *fp;
778 	struct msghdr *msg = &mmsg.msg_hdr;
779 	int error, s;
780 	unsigned int vlen, flags, dg;
781 
782 	s = SCARG(uap, s);
783 	if ((error = fd_getsock1(s, &so, &fp)) != 0)
784 		return error;
785 
786 	vlen = SCARG(uap, vlen);
787 	if (vlen > 1024)
788 		vlen = 1024;
789 
790 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
791 
792 	for (dg = 0; dg < vlen;) {
793 		error = copyin(SCARG(uap, mmsg) + dg, &mmsg, sizeof(mmsg));
794 		if (error)
795 			break;
796 
797 		msg->msg_flags = flags;
798 
799 		error = do_sys_sendmsg_so(l, s, so, fp, msg, flags,
800 		    &msg, sizeof(msg), retval);
801 		if (error)
802 			break;
803 
804 		ktrkuser("msghdr", msg, sizeof *msg);
805 		mmsg.msg_len = *retval;
806 		error = copyout(&mmsg, SCARG(uap, mmsg) + dg, sizeof(mmsg));
807 		if (error)
808 			break;
809 		dg++;
810 
811 	}
812 
813 	*retval = dg;
814 	if (error)
815 		so->so_error = error;
816 
817 	fd_putfile(s);
818 
819 	/*
820 	 * If we succeeded at least once, return 0, hopefully so->so_error
821 	 * will catch it next time.
822 	 */
823 	if (dg)
824 		return 0;
825 	return error;
826 }
827 
828 /*
829  * Adjust for a truncated SCM_RIGHTS control message.
830  *  This means closing any file descriptors that aren't present
831  *  in the returned buffer.
832  *  m is the mbuf holding the (already externalized) SCM_RIGHTS message.
833  */
834 static void
835 free_rights(struct mbuf *m)
836 {
837 	struct cmsghdr *cm;
838 	int *fdv;
839 	unsigned int nfds, i;
840 
841 	KASSERT(sizeof(*cm) <= m->m_len);
842 	cm = mtod(m, struct cmsghdr *);
843 
844 	KASSERT(CMSG_ALIGN(sizeof(*cm)) <= cm->cmsg_len);
845 	KASSERT(cm->cmsg_len <= m->m_len);
846 	nfds = (cm->cmsg_len - CMSG_ALIGN(sizeof(*cm))) / sizeof(int);
847 	fdv = (int *)CMSG_DATA(cm);
848 
849 	for (i = 0; i < nfds; i++)
850 		if (fd_getfile(fdv[i]) != NULL)
851 			(void)fd_close(fdv[i]);
852 }
853 
854 void
855 free_control_mbuf(struct lwp *l, struct mbuf *control, struct mbuf *uncopied)
856 {
857 	struct mbuf *next;
858 	struct cmsghdr *cmsg;
859 	bool do_free_rights = false;
860 
861 	while (control != NULL) {
862 		cmsg = mtod(control, struct cmsghdr *);
863 		if (control == uncopied)
864 			do_free_rights = true;
865 		if (do_free_rights && cmsg->cmsg_level == SOL_SOCKET
866 		    && cmsg->cmsg_type == SCM_RIGHTS)
867 			free_rights(control);
868 		next = control->m_next;
869 		m_free(control);
870 		control = next;
871 	}
872 }
873 
874 /* Copy socket control/CMSG data to user buffer, frees the mbuf */
875 int
876 copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
877 {
878 	int i, len, error = 0;
879 	struct cmsghdr *cmsg;
880 	struct mbuf *m;
881 	char *q;
882 
883 	len = mp->msg_controllen;
884 	if (len <= 0 || control == 0) {
885 		mp->msg_controllen = 0;
886 		free_control_mbuf(l, control, control);
887 		return 0;
888 	}
889 
890 	q = (char *)mp->msg_control;
891 
892 	for (m = control; m != NULL; ) {
893 		cmsg = mtod(m, struct cmsghdr *);
894 		i = m->m_len;
895 		if (len < i) {
896 			mp->msg_flags |= MSG_CTRUNC;
897 			if (cmsg->cmsg_level == SOL_SOCKET
898 			    && cmsg->cmsg_type == SCM_RIGHTS)
899 				/* Do not truncate me ... */
900 				break;
901 			i = len;
902 		}
903 		error = copyout(mtod(m, void *), q, i);
904 		ktrkuser("msgcontrol", mtod(m, void *), i);
905 		if (error != 0) {
906 			/* We must free all the SCM_RIGHTS */
907 			m = control;
908 			break;
909 		}
910 		m = m->m_next;
911 		if (m)
912 			i = ALIGN(i);
913 		q += i;
914 		len -= i;
915 		if (len <= 0)
916 			break;
917 	}
918 
919 	free_control_mbuf(l, control, m);
920 
921 	mp->msg_controllen = q - (char *)mp->msg_control;
922 	return error;
923 }
924 
925 static int
926 do_sys_recvmsg_so(struct lwp *l, int s, struct socket *so, struct msghdr *mp,
927     const void *ktrhdr, size_t ktsize, struct mbuf **from,
928     struct mbuf **control, register_t *retsize)
929 {
930 	struct iovec	aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
931 	struct uio	auio;
932 	size_t		len, iovsz;
933 	int		i, error;
934 
935 	if (__predict_false(ktrhdr == NULL && ktsize == 0)) {
936 		ktrhdr = mp;
937 		ktsize = sizeof *mp;
938 	}
939 	if (__predict_true(ktrhdr != NULL))
940 		ktrkuser("msghdr", ktrhdr, ktsize);
941 
942 	*from = NULL;
943 	if (control != NULL)
944 		*control = NULL;
945 
946 	iovsz = mp->msg_iovlen * sizeof(struct iovec);
947 
948 	if (mp->msg_flags & MSG_IOVUSRSPACE) {
949 		if ((unsigned int)mp->msg_iovlen > UIO_SMALLIOV) {
950 			if ((unsigned int)mp->msg_iovlen > IOV_MAX) {
951 				error = EMSGSIZE;
952 				goto out;
953 			}
954 			iov = kmem_alloc(iovsz, KM_SLEEP);
955 		}
956 		if (mp->msg_iovlen != 0) {
957 			error = copyin(mp->msg_iov, iov, iovsz);
958 			if (error)
959 				goto out;
960 		}
961 		auio.uio_iov = iov;
962 	} else
963 		auio.uio_iov = mp->msg_iov;
964 	auio.uio_iovcnt = mp->msg_iovlen;
965 	auio.uio_rw = UIO_READ;
966 	auio.uio_offset = 0;			/* XXX */
967 	auio.uio_resid = 0;
968 	KASSERT(l == curlwp);
969 	auio.uio_vmspace = l->l_proc->p_vmspace;
970 
971 	tiov = auio.uio_iov;
972 	for (i = 0; i < auio.uio_iovcnt; i++, tiov++) {
973 		/*
974 		 * Reads return ssize_t because -1 is returned on error.
975 		 * Therefore we must restrict the length to SSIZE_MAX to
976 		 * avoid garbage return values.
977 		 */
978 		auio.uio_resid += tiov->iov_len;
979 		if (tiov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
980 			error = EINVAL;
981 			goto out;
982 		}
983 	}
984 
985 	if (ktrpoint(KTR_GENIO) && iovsz > 0) {
986 		ktriov = kmem_alloc(iovsz, KM_SLEEP);
987 		memcpy(ktriov, auio.uio_iov, iovsz);
988 	}
989 
990 	len = auio.uio_resid;
991 	mp->msg_flags &= MSG_USERFLAGS;
992 	error = (*so->so_receive)(so, from, &auio, NULL, control,
993 	    &mp->msg_flags);
994 	len -= auio.uio_resid;
995 	*retsize = len;
996 	if (error != 0 && len != 0
997 	    && (error == ERESTART || error == EINTR || error == EWOULDBLOCK))
998 		/* Some data transferred */
999 		error = 0;
1000 
1001 	if (ktriov != NULL) {
1002 		ktrgeniov(s, UIO_READ, ktriov, len, error);
1003 		kmem_free(ktriov, iovsz);
1004 	}
1005 
1006 	if (error != 0) {
1007 		m_freem(*from);
1008 		*from = NULL;
1009 		if (control != NULL) {
1010 			free_control_mbuf(l, *control, *control);
1011 			*control = NULL;
1012 		}
1013 	}
1014  out:
1015 	if (iov != aiov)
1016 		kmem_free(iov, iovsz);
1017 	return error;
1018 }
1019 
1020 
1021 int
1022 do_sys_recvmsg(struct lwp *l, int s, struct msghdr *mp,
1023     const void *ktrhdr, size_t ktrsize,
1024     struct mbuf **from, struct mbuf **control, register_t *retsize)
1025 {
1026 	int error;
1027 	struct socket *so;
1028 
1029 	if ((error = fd_getsock(s, &so)) != 0)
1030 		return error;
1031 	error = do_sys_recvmsg_so(l, s, so, mp, ktrhdr, ktrsize, from,
1032 	    control, retsize);
1033 	fd_putfile(s);
1034 	return error;
1035 }
1036 
1037 int
1038 sys_recvmmsg(struct lwp *l, const struct sys_recvmmsg_args *uap,
1039     register_t *retval)
1040 {
1041 	/* {
1042 		syscallarg(int)			s;
1043 		syscallarg(struct mmsghdr *)	mmsg;
1044 		syscallarg(unsigned int)	vlen;
1045 		syscallarg(unsigned int)	flags;
1046 		syscallarg(struct timespec *)	timeout;
1047 	} */
1048 	struct mmsghdr mmsg;
1049 	struct socket *so;
1050 	struct msghdr *msg = &mmsg.msg_hdr;
1051 	int error, s;
1052 	struct mbuf *from, *control;
1053 	struct timespec ts, now;
1054 	unsigned int vlen, flags, dg;
1055 
1056 	if (SCARG(uap, timeout)) {
1057 		if ((error = copyin(SCARG(uap, timeout), &ts, sizeof(ts))) != 0)
1058 			return error;
1059 		getnanotime(&now);
1060 		timespecadd(&now, &ts, &ts);
1061 	}
1062 
1063 	s = SCARG(uap, s);
1064 	if ((error = fd_getsock(s, &so)) != 0)
1065 		return error;
1066 
1067 	vlen = SCARG(uap, vlen);
1068 	if (vlen > 1024)
1069 		vlen = 1024;
1070 
1071 	from = NULL;
1072 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1073 
1074 	for (dg = 0; dg < vlen;) {
1075 		error = copyin(SCARG(uap, mmsg) + dg, &mmsg, sizeof(mmsg));
1076 		if (error)
1077 			break;
1078 
1079 		msg->msg_flags = flags & ~MSG_WAITFORONE;
1080 
1081 		if (from != NULL) {
1082 			m_free(from);
1083 			from = NULL;
1084 		}
1085 
1086 		error = do_sys_recvmsg_so(l, s, so, msg, NULL, 0, &from,
1087 		    msg->msg_control != NULL ? &control : NULL, retval);
1088 		if (error) {
1089 			if (error == EAGAIN && dg > 0)
1090 				error = 0;
1091 			break;
1092 		}
1093 
1094 		if (msg->msg_control != NULL)
1095 			error = copyout_msg_control(l, msg, control);
1096 		if (error)
1097 			break;
1098 
1099 		error = copyout_sockname(msg->msg_name, &msg->msg_namelen, 0,
1100 		    from);
1101 		if (error)
1102 			break;
1103 
1104 		ktrkuser("msghdr", msg, sizeof *msg);
1105 		mmsg.msg_len = *retval;
1106 
1107 		error = copyout(&mmsg, SCARG(uap, mmsg) + dg, sizeof(mmsg));
1108 		if (error)
1109 			break;
1110 
1111 		dg++;
1112 		if (msg->msg_flags & MSG_OOB)
1113 			break;
1114 
1115 		if (SCARG(uap, timeout)) {
1116 			getnanotime(&now);
1117 			timespecsub(&now, &ts, &now);
1118 			if (now.tv_sec > 0)
1119 				break;
1120 		}
1121 
1122 		if (flags & MSG_WAITFORONE)
1123 			flags |= MSG_DONTWAIT;
1124 
1125 	}
1126 
1127 	if (from != NULL)
1128 		m_free(from);
1129 
1130 	*retval = dg;
1131 	if (error)
1132 		so->so_error = error;
1133 
1134 	fd_putfile(s);
1135 
1136 	/*
1137 	 * If we succeeded at least once, return 0, hopefully so->so_error
1138 	 * will catch it next time.
1139 	 */
1140 	if (dg)
1141 		return 0;
1142 
1143 	return error;
1144 }
1145 
1146 int
1147 sys_shutdown(struct lwp *l, const struct sys_shutdown_args *uap,
1148     register_t *retval)
1149 {
1150 	/* {
1151 		syscallarg(int)	s;
1152 		syscallarg(int)	how;
1153 	} */
1154 	struct socket	*so;
1155 	int		error;
1156 
1157 	if ((error = fd_getsock(SCARG(uap, s), &so)) != 0)
1158 		return error;
1159 	solock(so);
1160 	error = soshutdown(so, SCARG(uap, how));
1161 	sounlock(so);
1162 	fd_putfile(SCARG(uap, s));
1163 	return error;
1164 }
1165 
1166 int
1167 sys_setsockopt(struct lwp *l, const struct sys_setsockopt_args *uap,
1168     register_t *retval)
1169 {
1170 	/* {
1171 		syscallarg(int)			s;
1172 		syscallarg(int)			level;
1173 		syscallarg(int)			name;
1174 		syscallarg(const void *)	val;
1175 		syscallarg(unsigned int)	valsize;
1176 	} */
1177 	struct sockopt	sopt;
1178 	struct socket	*so;
1179 	file_t		*fp;
1180 	int		error;
1181 	unsigned int	len;
1182 
1183 	len = SCARG(uap, valsize);
1184 	if (len > 0 && SCARG(uap, val) == NULL)
1185 		return EINVAL;
1186 
1187 	if (len > MCLBYTES)
1188 		return EINVAL;
1189 
1190 	if ((error = fd_getsock1(SCARG(uap, s), &so, &fp)) != 0)
1191 		return (error);
1192 
1193 	sockopt_init(&sopt, SCARG(uap, level), SCARG(uap, name), len);
1194 
1195 	if (len > 0) {
1196 		error = copyin(SCARG(uap, val), sopt.sopt_data, len);
1197 		if (error)
1198 			goto out;
1199 	}
1200 
1201 	error = sosetopt(so, &sopt);
1202 	if (so->so_options & SO_NOSIGPIPE)
1203 		atomic_or_uint(&fp->f_flag, FNOSIGPIPE);
1204 	else
1205 		atomic_and_uint(&fp->f_flag, ~FNOSIGPIPE);
1206 
1207  out:
1208 	sockopt_destroy(&sopt);
1209 	fd_putfile(SCARG(uap, s));
1210 	return error;
1211 }
1212 
1213 int
1214 sys_getsockopt(struct lwp *l, const struct sys_getsockopt_args *uap,
1215     register_t *retval)
1216 {
1217 	/* {
1218 		syscallarg(int)			s;
1219 		syscallarg(int)			level;
1220 		syscallarg(int)			name;
1221 		syscallarg(void *)		val;
1222 		syscallarg(unsigned int *)	avalsize;
1223 	} */
1224 	struct sockopt	sopt;
1225 	struct socket	*so;
1226 	file_t		*fp;
1227 	unsigned int	valsize, len;
1228 	int		error;
1229 
1230 	if (SCARG(uap, val) != NULL) {
1231 		error = copyin(SCARG(uap, avalsize), &valsize, sizeof(valsize));
1232 		if (error)
1233 			return error;
1234 	} else
1235 		valsize = 0;
1236 
1237 	if ((error = fd_getsock1(SCARG(uap, s), &so, &fp)) != 0)
1238 		return (error);
1239 
1240 	sockopt_init(&sopt, SCARG(uap, level), SCARG(uap, name), 0);
1241 
1242 	if (fp->f_flag & FNOSIGPIPE)
1243 		so->so_options |= SO_NOSIGPIPE;
1244 	else
1245 		so->so_options &= ~SO_NOSIGPIPE;
1246 	error = sogetopt(so, &sopt);
1247 	if (error)
1248 		goto out;
1249 
1250 	if (valsize > 0) {
1251 		len = min(valsize, sopt.sopt_size);
1252 		error = copyout(sopt.sopt_data, SCARG(uap, val), len);
1253 		if (error)
1254 			goto out;
1255 
1256 		error = copyout(&len, SCARG(uap, avalsize), sizeof(len));
1257 		if (error)
1258 			goto out;
1259 	}
1260 
1261  out:
1262 	sockopt_destroy(&sopt);
1263 	fd_putfile(SCARG(uap, s));
1264 	return error;
1265 }
1266 
1267 #ifdef PIPE_SOCKETPAIR
1268 
1269 int
1270 pipe1(struct lwp *l, register_t *retval, int flags)
1271 {
1272 	file_t		*rf, *wf;
1273 	struct socket	*rso, *wso;
1274 	int		fd, error;
1275 	proc_t		*p;
1276 
1277 	if (flags & ~(O_CLOEXEC|O_NONBLOCK|O_NOSIGPIPE))
1278 		return EINVAL;
1279 	p = curproc;
1280 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, l, NULL)) != 0)
1281 		return error;
1282 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, l, rso)) != 0)
1283 		goto free1;
1284 	/* remember this socket pair implements a pipe */
1285 	wso->so_state |= SS_ISAPIPE;
1286 	rso->so_state |= SS_ISAPIPE;
1287 	if ((error = fd_allocfile(&rf, &fd)) != 0)
1288 		goto free2;
1289 	retval[0] = fd;
1290 	rf->f_flag = FREAD | flags;
1291 	rf->f_type = DTYPE_SOCKET;
1292 	rf->f_ops = &socketops;
1293 	rf->f_socket = rso;
1294 	if ((error = fd_allocfile(&wf, &fd)) != 0)
1295 		goto free3;
1296 	wf->f_flag = FWRITE | flags;
1297 	wf->f_type = DTYPE_SOCKET;
1298 	wf->f_ops = &socketops;
1299 	wf->f_socket = wso;
1300 	retval[1] = fd;
1301 	solock(wso);
1302 	error = unp_connect2(wso, rso);
1303 	sounlock(wso);
1304 	if (error != 0)
1305 		goto free4;
1306 	fd_affix(p, wf, (int)retval[1]);
1307 	fd_affix(p, rf, (int)retval[0]);
1308 	return (0);
1309  free4:
1310 	fd_abort(p, wf, (int)retval[1]);
1311  free3:
1312 	fd_abort(p, rf, (int)retval[0]);
1313  free2:
1314 	(void)soclose(wso);
1315  free1:
1316 	(void)soclose(rso);
1317 	return error;
1318 }
1319 #endif /* PIPE_SOCKETPAIR */
1320 
1321 /*
1322  * Get peer socket name.
1323  */
1324 int
1325 do_sys_getpeername(int fd, struct sockaddr *nam)
1326 {
1327 	struct socket	*so;
1328 	int		error;
1329 
1330 	if ((error = fd_getsock(fd, &so)) != 0)
1331 		return error;
1332 
1333 	solock(so);
1334 	if ((so->so_state & SS_ISCONNECTED) == 0)
1335 		error = ENOTCONN;
1336 	else {
1337 		error = (*so->so_proto->pr_usrreqs->pr_peeraddr)(so, nam);
1338 	}
1339 	sounlock(so);
1340 	fd_putfile(fd);
1341 	return error;
1342 }
1343 
1344 /*
1345  * Get local socket name.
1346  */
1347 int
1348 do_sys_getsockname(int fd, struct sockaddr *nam)
1349 {
1350 	struct socket	*so;
1351 	int		error;
1352 
1353 	if ((error = fd_getsock(fd, &so)) != 0)
1354 		return error;
1355 
1356 	solock(so);
1357 	error = (*so->so_proto->pr_usrreqs->pr_sockaddr)(so, nam);
1358 	sounlock(so);
1359 	fd_putfile(fd);
1360 	return error;
1361 }
1362 
1363 int
1364 copyout_sockname_sb(struct sockaddr *asa, unsigned int *alen, int flags,
1365     struct sockaddr_big *addr)
1366 {
1367 	int len;
1368 	int error;
1369 
1370 	if (asa == NULL)
1371 		/* Assume application not interested */
1372 		return 0;
1373 
1374 	if (flags & MSG_LENUSRSPACE) {
1375 		error = copyin(alen, &len, sizeof(len));
1376 		if (error)
1377 			return error;
1378 	} else
1379 		len = *alen;
1380 	if (len < 0)
1381 		return EINVAL;
1382 
1383 	if (addr == NULL) {
1384 		len = 0;
1385 		error = 0;
1386 	} else {
1387 		if (len > addr->sb_len)
1388 			len = addr->sb_len;
1389 		/* XXX addr isn't an mbuf... */
1390 		ktrkuser(mbuftypes[MT_SONAME], addr, len);
1391 		error = copyout(addr, asa, len);
1392 	}
1393 
1394 	if (error == 0) {
1395 		if (flags & MSG_LENUSRSPACE)
1396 			error = copyout(&len, alen, sizeof(len));
1397 		else
1398 			*alen = len;
1399 	}
1400 
1401 	return error;
1402 }
1403 
1404 int
1405 copyout_sockname(struct sockaddr *asa, unsigned int *alen, int flags,
1406     struct mbuf *addr)
1407 {
1408 	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 	if (len < 0)
1422 		return EINVAL;
1423 
1424 	if (addr == NULL) {
1425 		len = 0;
1426 		error = 0;
1427 	} else {
1428 		if (len > addr->m_len)
1429 			len = addr->m_len;
1430 		/* Maybe this ought to copy a chain ? */
1431 		ktrkuser(mbuftypes[MT_SONAME], mtod(addr, void *), len);
1432 		error = copyout(mtod(addr, void *), asa, len);
1433 	}
1434 
1435 	if (error == 0) {
1436 		if (flags & MSG_LENUSRSPACE)
1437 			error = copyout(&len, alen, sizeof(len));
1438 		else
1439 			*alen = len;
1440 	}
1441 
1442 	return error;
1443 }
1444 
1445 /*
1446  * Get socket name.
1447  */
1448 int
1449 sys_getsockname(struct lwp *l, const struct sys_getsockname_args *uap,
1450     register_t *retval)
1451 {
1452 	/* {
1453 		syscallarg(int)			fdes;
1454 		syscallarg(struct sockaddr *)	asa;
1455 		syscallarg(unsigned int *)	alen;
1456 	} */
1457 	struct sockaddr_big sbig;
1458 	int		    error;
1459 
1460 	sbig.sb_len = UCHAR_MAX;
1461 	error = do_sys_getsockname(SCARG(uap, fdes), (struct sockaddr *)&sbig);
1462 	if (error != 0)
1463 		return error;
1464 
1465 	error = copyout_sockname_sb(SCARG(uap, asa), SCARG(uap, alen),
1466 	    MSG_LENUSRSPACE, &sbig);
1467 	return error;
1468 }
1469 
1470 /*
1471  * Get name of peer for connected socket.
1472  */
1473 int
1474 sys_getpeername(struct lwp *l, const struct sys_getpeername_args *uap,
1475     register_t *retval)
1476 {
1477 	/* {
1478 		syscallarg(int)			fdes;
1479 		syscallarg(struct sockaddr *)	asa;
1480 		syscallarg(unsigned int *)	alen;
1481 	} */
1482 	struct sockaddr_big sbig;
1483 	int		    error;
1484 
1485 	sbig.sb_len = UCHAR_MAX;
1486 	error = do_sys_getpeername(SCARG(uap, fdes), (struct sockaddr *)&sbig);
1487 	if (error != 0)
1488 		return error;
1489 
1490 	error = copyout_sockname_sb(SCARG(uap, asa), SCARG(uap, alen),
1491 	    MSG_LENUSRSPACE, &sbig);
1492 	return error;
1493 }
1494 
1495 static int
1496 sockargs_sb(struct sockaddr_big *sb, const void *name, socklen_t buflen)
1497 {
1498 	int error;
1499 
1500 	/*
1501 	 * We can't allow socket names > UCHAR_MAX in length, since that
1502 	 * will overflow sb_len. Further no reasonable buflen is <=
1503 	 * offsetof(sockaddr_big, sb_data) since it shall be at least
1504 	 * the size of the preamble sb_len and sb_family members.
1505 	 */
1506 	if (buflen > UCHAR_MAX ||
1507 	    buflen <= offsetof(struct sockaddr_big, sb_data))
1508 		return EINVAL;
1509 
1510 	error = copyin(name, (void *)sb, buflen);
1511 	if (error)
1512 		return error;
1513 
1514 	ktrkuser(mbuftypes[MT_SONAME], sb, buflen);
1515 #if BYTE_ORDER != BIG_ENDIAN
1516 	/*
1517 	 * 4.3BSD compat thing - need to stay, since bind(2),
1518 	 * connect(2), sendto(2) were not versioned for COMPAT_43.
1519 	 */
1520 	if (sb->sb_family == 0 && sb->sb_len < AF_MAX)
1521 		sb->sb_family = sb->sb_len;
1522 #endif
1523 	sb->sb_len = buflen;
1524 	return 0;
1525 }
1526 
1527 /*
1528  * XXX In a perfect world, we wouldn't pass around socket control
1529  * XXX arguments in mbufs, and this could go away.
1530  */
1531 int
1532 sockargs(struct mbuf **mp, const void *bf, size_t buflen, int type)
1533 {
1534 	struct sockaddr	*sa;
1535 	struct mbuf	*m;
1536 	int		error;
1537 
1538 	/*
1539 	 * We can't allow socket names > UCHAR_MAX in length, since that
1540 	 * will overflow sa_len.  Control data more than a page size in
1541 	 * length is just too much.
1542 	 */
1543 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
1544 		return EINVAL;
1545 
1546 	/*
1547 	 * length must greater than sizeof(sa_family) + sizeof(sa_len)
1548 	 */
1549 	if (type == MT_SONAME && buflen <= 2)
1550 		return EINVAL;
1551 
1552 	/* Allocate an mbuf to hold the arguments. */
1553 	m = m_get(M_WAIT, type);
1554 	/* can't claim.  don't who to assign it to. */
1555 	if (buflen > MLEN) {
1556 		/*
1557 		 * Won't fit into a regular mbuf, so we allocate just
1558 		 * enough external storage to hold the argument.
1559 		 */
1560 		MEXTMALLOC(m, buflen, M_WAITOK);
1561 	}
1562 	m->m_len = buflen;
1563 	error = copyin(bf, mtod(m, void *), buflen);
1564 	if (error) {
1565 		(void)m_free(m);
1566 		return error;
1567 	}
1568 	ktrkuser(mbuftypes[type], mtod(m, void *), buflen);
1569 	*mp = m;
1570 	if (type == MT_SONAME) {
1571 		sa = mtod(m, struct sockaddr *);
1572 #if BYTE_ORDER != BIG_ENDIAN
1573 		/*
1574 		 * 4.3BSD compat thing - need to stay, since bind(2),
1575 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
1576 		 */
1577 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1578 			sa->sa_family = sa->sa_len;
1579 #endif
1580 		sa->sa_len = buflen;
1581 	}
1582 	return 0;
1583 }
1584