xref: /netbsd-src/sys/kern/uipc_syscalls.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: uipc_syscalls.c,v 1.194 2018/05/04 08:47:55 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 1.194 2018/05/04 08:47:55 christos Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_pipe.h"
68 #endif
69 
70 #define MBUFTYPES
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/filedesc.h>
74 #include <sys/proc.h>
75 #include <sys/file.h>
76 #include <sys/buf.h>
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 
98 int
99 sys___socket30(struct lwp *l, const struct sys___socket30_args *uap,
100     register_t *retval)
101 {
102 	/* {
103 		syscallarg(int)	domain;
104 		syscallarg(int)	type;
105 		syscallarg(int)	protocol;
106 	} */
107 	int fd, error;
108 
109 	error = fsocreate(SCARG(uap, domain), NULL, SCARG(uap, type),
110 	    SCARG(uap, protocol), &fd);
111 	if (error == 0) {
112 		*retval = fd;
113 	}
114 	return error;
115 }
116 
117 int
118 sys_bind(struct lwp *l, const struct sys_bind_args *uap, register_t *retval)
119 {
120 	/* {
121 		syscallarg(int)				s;
122 		syscallarg(const struct sockaddr *)	name;
123 		syscallarg(unsigned int)		namelen;
124 	} */
125 	int		error;
126 	struct sockaddr_big sb;
127 
128 	error = sockargs_sb(&sb, SCARG(uap, name), SCARG(uap, namelen));
129 	if (error)
130 		return error;
131 
132 	return do_sys_bind(l, SCARG(uap, s), (struct sockaddr *)&sb);
133 }
134 
135 int
136 do_sys_bind(struct lwp *l, int fd, struct sockaddr *nam)
137 {
138 	struct socket	*so;
139 	int		error;
140 
141 	if ((error = fd_getsock(fd, &so)) != 0)
142 		return error;
143 	error = sobind(so, nam, l);
144 	fd_putfile(fd);
145 	return error;
146 }
147 
148 int
149 sys_listen(struct lwp *l, const struct sys_listen_args *uap, register_t *retval)
150 {
151 	/* {
152 		syscallarg(int)	s;
153 		syscallarg(int)	backlog;
154 	} */
155 	struct socket	*so;
156 	int		error;
157 
158 	if ((error = fd_getsock(SCARG(uap, s), &so)) != 0)
159 		return (error);
160 	error = solisten(so, SCARG(uap, backlog), l);
161 	fd_putfile(SCARG(uap, s));
162 	return error;
163 }
164 
165 int
166 do_sys_accept(struct lwp *l, int sock, struct sockaddr *name,
167     register_t *new_sock, const sigset_t *mask, int flags, int clrflags)
168 {
169 	file_t		*fp, *fp2;
170 	int		error, fd;
171 	struct socket	*so, *so2;
172 	short		wakeup_state = 0;
173 
174 	if ((fp = fd_getfile(sock)) == NULL)
175 		return EBADF;
176 	if (fp->f_type != DTYPE_SOCKET) {
177 		fd_putfile(sock);
178 		return ENOTSOCK;
179 	}
180 	if ((error = fd_allocfile(&fp2, &fd)) != 0) {
181 		fd_putfile(sock);
182 		return error;
183 	}
184 	*new_sock = fd;
185 	so = fp->f_socket;
186 	solock(so);
187 
188 	if (__predict_false(mask))
189 		sigsuspendsetup(l, mask);
190 
191 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
192 		error = EOPNOTSUPP;
193 		goto bad;
194 	}
195 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
196 		error = EINVAL;
197 		goto bad;
198 	}
199 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
200 		error = EWOULDBLOCK;
201 		goto bad;
202 	}
203 	while (so->so_qlen == 0 && so->so_error == 0) {
204 		if (so->so_state & SS_CANTRCVMORE) {
205 			so->so_error = ECONNABORTED;
206 			break;
207 		}
208 		if (wakeup_state & SS_RESTARTSYS) {
209 			error = ERESTART;
210 			goto bad;
211 		}
212 		error = sowait(so, true, 0);
213 		if (error) {
214 			goto bad;
215 		}
216 		wakeup_state = so->so_state;
217 	}
218 	if (so->so_error) {
219 		error = so->so_error;
220 		so->so_error = 0;
221 		goto bad;
222 	}
223 	/* connection has been removed from the listen queue */
224 	KNOTE(&so->so_rcv.sb_sel.sel_klist, NOTE_SUBMIT);
225 	so2 = TAILQ_FIRST(&so->so_q);
226 	if (soqremque(so2, 1) == 0)
227 		panic("accept");
228 	fp2->f_type = DTYPE_SOCKET;
229 	fp2->f_flag = (fp->f_flag & ~clrflags) |
230 	    ((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)|
231 	    ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0);
232 	fp2->f_ops = &socketops;
233 	fp2->f_socket = so2;
234 	if (fp2->f_flag & FNONBLOCK)
235 		so2->so_state |= SS_NBIO;
236 	else
237 		so2->so_state &= ~SS_NBIO;
238 	error = soaccept(so2, name);
239 	so2->so_cred = kauth_cred_dup(so->so_cred);
240 	sounlock(so);
241 	if (error) {
242 		/* an error occurred, free the file descriptor and mbuf */
243 		mutex_enter(&fp2->f_lock);
244 		fp2->f_count++;
245 		mutex_exit(&fp2->f_lock);
246 		closef(fp2);
247 		fd_abort(curproc, NULL, fd);
248 	} else {
249 		fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0);
250 		fd_affix(curproc, fp2, fd);
251 	}
252 	fd_putfile(sock);
253 	if (__predict_false(mask))
254 		sigsuspendteardown(l);
255 	return error;
256  bad:
257 	sounlock(so);
258 	fd_putfile(sock);
259 	fd_abort(curproc, fp2, fd);
260 	if (__predict_false(mask))
261 		sigsuspendteardown(l);
262 	return error;
263 }
264 
265 int
266 sys_accept(struct lwp *l, const struct sys_accept_args *uap, register_t *retval)
267 {
268 	/* {
269 		syscallarg(int)			s;
270 		syscallarg(struct sockaddr *)	name;
271 		syscallarg(unsigned int *)	anamelen;
272 	} */
273 	int error, fd;
274 	struct sockaddr_big name;
275 
276 	name.sb_len = UCHAR_MAX;
277 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
278 	    retval, NULL, 0, 0);
279 	if (error != 0)
280 		return error;
281 	error = copyout_sockname_sb(SCARG(uap, name), SCARG(uap, anamelen),
282 	    MSG_LENUSRSPACE, &name);
283 	if (error != 0) {
284 		fd = (int)*retval;
285 		if (fd_getfile(fd) != NULL)
286 			(void)fd_close(fd);
287 	}
288 	return error;
289 }
290 
291 int
292 sys_paccept(struct lwp *l, const struct sys_paccept_args *uap,
293     register_t *retval)
294 {
295 	/* {
296 		syscallarg(int)			s;
297 		syscallarg(struct sockaddr *)	name;
298 		syscallarg(unsigned int *)	anamelen;
299 		syscallarg(const sigset_t *)	mask;
300 		syscallarg(int)			flags;
301 	} */
302 	int error, fd;
303 	struct sockaddr_big name;
304 	sigset_t *mask, amask;
305 
306 	if (SCARG(uap, mask) != NULL) {
307 		error = copyin(SCARG(uap, mask), &amask, sizeof(amask));
308 		if (error)
309 			return error;
310 		mask = &amask;
311 	} else
312 		mask = NULL;
313 
314 	name.sb_len = UCHAR_MAX;
315 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
316 	    retval, mask, SCARG(uap, flags), FNONBLOCK);
317 	if (error != 0)
318 		return error;
319 	error = copyout_sockname_sb(SCARG(uap, name), SCARG(uap, anamelen),
320 	    MSG_LENUSRSPACE, &name);
321 	if (error != 0) {
322 		fd = (int)*retval;
323 		if (fd_getfile(fd) != NULL)
324 			(void)fd_close(fd);
325 	}
326 	return error;
327 }
328 
329 int
330 sys_connect(struct lwp *l, const struct sys_connect_args *uap,
331     register_t *retval)
332 {
333 	/* {
334 		syscallarg(int)				s;
335 		syscallarg(const struct sockaddr *)	name;
336 		syscallarg(unsigned int)		namelen;
337 	} */
338 	int		error;
339 	struct sockaddr_big sbig;
340 
341 	error = sockargs_sb(&sbig, SCARG(uap, name), SCARG(uap, namelen));
342 	if (error)
343 		return error;
344 	return do_sys_connect(l, SCARG(uap, s), (struct sockaddr *)&sbig);
345 }
346 
347 int
348 do_sys_connect(struct lwp *l, int fd, struct sockaddr *nam)
349 {
350 	struct socket	*so;
351 	int		error;
352 	int		interrupted = 0;
353 
354 	if ((error = fd_getsock(fd, &so)) != 0) {
355 		return (error);
356 	}
357 	solock(so);
358 	if ((so->so_state & SS_ISCONNECTING) != 0) {
359 		error = EALREADY;
360 		goto out;
361 	}
362 
363 	error = soconnect(so, nam, l);
364 	if (error)
365 		goto bad;
366 	if ((so->so_state & (SS_NBIO|SS_ISCONNECTING)) ==
367 	    (SS_NBIO|SS_ISCONNECTING)) {
368 		error = EINPROGRESS;
369 		goto out;
370 	}
371 	while ((so->so_state & SS_ISCONNECTING) != 0 && so->so_error == 0) {
372 		error = sowait(so, true, 0);
373 		if (__predict_false((so->so_state & SS_ISABORTING) != 0)) {
374 			error = EPIPE;
375 			interrupted = 1;
376 			break;
377 		}
378 		if (error) {
379 			if (error == EINTR || error == ERESTART)
380 				interrupted = 1;
381 			break;
382 		}
383 	}
384 	if (error == 0) {
385 		error = so->so_error;
386 		so->so_error = 0;
387 	}
388  bad:
389 	if (!interrupted)
390 		so->so_state &= ~SS_ISCONNECTING;
391 	if (error == ERESTART)
392 		error = EINTR;
393  out:
394 	sounlock(so);
395 	fd_putfile(fd);
396 	return error;
397 }
398 
399 static int
400 makesocket(struct lwp *l, file_t **fp, int *fd, int flags, int type,
401     int domain, int proto, struct socket *soo)
402 {
403 	struct socket *so;
404 	int error;
405 
406 	if ((error = socreate(domain, &so, type, proto, l, soo)) != 0) {
407 		return error;
408 	}
409 	if (flags & SOCK_NONBLOCK) {
410 		so->so_state |= SS_NBIO;
411 	}
412 
413 	if ((error = fd_allocfile(fp, fd)) != 0) {
414 		soclose(so);
415 		return error;
416 	}
417 	fd_set_exclose(l, *fd, (flags & SOCK_CLOEXEC) != 0);
418 	(*fp)->f_flag = FREAD|FWRITE|
419 	    ((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)|
420 	    ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0);
421 	(*fp)->f_type = DTYPE_SOCKET;
422 	(*fp)->f_ops = &socketops;
423 	(*fp)->f_socket = so;
424 	return 0;
425 }
426 
427 int
428 sys_socketpair(struct lwp *l, const struct sys_socketpair_args *uap,
429     register_t *retval)
430 {
431 	/* {
432 		syscallarg(int)		domain;
433 		syscallarg(int)		type;
434 		syscallarg(int)		protocol;
435 		syscallarg(int *)	rsv;
436 	} */
437 	file_t		*fp1, *fp2;
438 	struct socket	*so1, *so2;
439 	int		fd, error, sv[2];
440 	proc_t		*p = curproc;
441 	int		flags = SCARG(uap, type) & SOCK_FLAGS_MASK;
442 	int		type = SCARG(uap, type) & ~SOCK_FLAGS_MASK;
443 	int		domain = SCARG(uap, domain);
444 	int		proto = SCARG(uap, protocol);
445 
446 	error = makesocket(l, &fp1, &fd, flags, type, domain, proto, NULL);
447 	if (error)
448 		return error;
449 	so1 = fp1->f_socket;
450 	sv[0] = fd;
451 
452 	error = makesocket(l, &fp2, &fd, flags, type, domain, proto, so1);
453 	if (error)
454 		goto out;
455 	so2 = fp2->f_socket;
456 	sv[1] = fd;
457 
458 	solock(so1);
459 	error = soconnect2(so1, so2);
460 	if (error == 0 && type == SOCK_DGRAM) {
461 		/*
462 		 * Datagram socket connection is asymmetric.
463 		 */
464 		error = soconnect2(so2, so1);
465 	}
466 	sounlock(so1);
467 
468 	if (error == 0)
469 		error = copyout(sv, SCARG(uap, rsv), sizeof(sv));
470 	if (error == 0) {
471 		fd_affix(p, fp2, sv[1]);
472 		fd_affix(p, fp1, sv[0]);
473 		return 0;
474 	}
475 	fd_abort(p, fp2, sv[1]);
476 	(void)soclose(so2);
477 out:
478 	fd_abort(p, fp1, sv[0]);
479 	(void)soclose(so1);
480 	return error;
481 }
482 
483 int
484 sys_sendto(struct lwp *l, const struct sys_sendto_args *uap,
485     register_t *retval)
486 {
487 	/* {
488 		syscallarg(int)				s;
489 		syscallarg(const void *)		buf;
490 		syscallarg(size_t)			len;
491 		syscallarg(int)				flags;
492 		syscallarg(const struct sockaddr *)	to;
493 		syscallarg(unsigned int)		tolen;
494 	} */
495 	struct msghdr	msg;
496 	struct iovec	aiov;
497 
498 	msg.msg_name = __UNCONST(SCARG(uap, to)); /* XXXUNCONST kills const */
499 	msg.msg_namelen = SCARG(uap, tolen);
500 	msg.msg_iov = &aiov;
501 	msg.msg_iovlen = 1;
502 	msg.msg_control = NULL;
503 	msg.msg_flags = 0;
504 	aiov.iov_base = __UNCONST(SCARG(uap, buf)); /* XXXUNCONST kills const */
505 	aiov.iov_len = SCARG(uap, len);
506 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, SCARG(uap, flags),
507 	    retval);
508 }
509 
510 int
511 sys_sendmsg(struct lwp *l, const struct sys_sendmsg_args *uap,
512     register_t *retval)
513 {
514 	/* {
515 		syscallarg(int)				s;
516 		syscallarg(const struct msghdr *)	msg;
517 		syscallarg(int)				flags;
518 	} */
519 	struct msghdr	msg;
520 	int		error;
521 
522 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
523 	if (error)
524 		return (error);
525 
526 	msg.msg_flags = MSG_IOVUSRSPACE;
527 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, SCARG(uap, flags),
528 	    retval);
529 }
530 
531 int
532 do_sys_sendmsg_so(struct lwp *l, int s, struct socket *so, file_t *fp,
533     struct msghdr *mp, int flags, register_t *retsize)
534 {
535 
536 	struct iovec	aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
537 	struct sockaddr *sa = NULL;
538 	struct mbuf	*to, *control;
539 	struct uio	auio;
540 	size_t		len, iovsz;
541 	int		i, error;
542 
543 	ktrkuser("msghdr", mp, sizeof(*mp));
544 
545 	/* If the caller passed us stuff in mbufs, we must free them. */
546 	to = (mp->msg_flags & MSG_NAMEMBUF) ? mp->msg_name : NULL;
547 	control = (mp->msg_flags & MSG_CONTROLMBUF) ? mp->msg_control : NULL;
548 	iovsz = mp->msg_iovlen * sizeof(struct iovec);
549 
550 	if (mp->msg_flags & MSG_IOVUSRSPACE) {
551 		if ((unsigned int)mp->msg_iovlen > UIO_SMALLIOV) {
552 			if ((unsigned int)mp->msg_iovlen > IOV_MAX) {
553 				error = EMSGSIZE;
554 				goto bad;
555 			}
556 			iov = kmem_alloc(iovsz, KM_SLEEP);
557 		}
558 		if (mp->msg_iovlen != 0) {
559 			error = copyin(mp->msg_iov, iov, iovsz);
560 			if (error)
561 				goto bad;
562 		}
563 		auio.uio_iov = iov;
564 	} else
565 		auio.uio_iov = mp->msg_iov;
566 
567 	auio.uio_iovcnt = mp->msg_iovlen;
568 	auio.uio_rw = UIO_WRITE;
569 	auio.uio_offset = 0;			/* XXX */
570 	auio.uio_resid = 0;
571 	KASSERT(l == curlwp);
572 	auio.uio_vmspace = l->l_proc->p_vmspace;
573 
574 	tiov = auio.uio_iov;
575 	for (i = 0; i < auio.uio_iovcnt; i++, tiov++) {
576 		/*
577 		 * Writes return ssize_t because -1 is returned on error.
578 		 * Therefore, we must restrict the length to SSIZE_MAX to
579 		 * avoid garbage return values.
580 		 */
581 		auio.uio_resid += tiov->iov_len;
582 		if (tiov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
583 			error = EINVAL;
584 			goto bad;
585 		}
586 	}
587 
588 	if (mp->msg_name && to == NULL) {
589 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
590 		    UIO_USERSPACE, MT_SONAME);
591 		if (error)
592 			goto bad;
593 	}
594 
595 	if (mp->msg_control) {
596 		if (mp->msg_controllen < CMSG_ALIGN(sizeof(struct cmsghdr))) {
597 			error = EINVAL;
598 			goto bad;
599 		}
600 		if (control == NULL) {
601 			error = sockargs(&control, mp->msg_control,
602 			    mp->msg_controllen, UIO_USERSPACE, MT_CONTROL);
603 			if (error)
604 				goto bad;
605 		}
606 	}
607 
608 	if (ktrpoint(KTR_GENIO) && iovsz > 0) {
609 		ktriov = kmem_alloc(iovsz, KM_SLEEP);
610 		memcpy(ktriov, auio.uio_iov, iovsz);
611 	}
612 
613 	if (mp->msg_name)
614 		MCLAIM(to, so->so_mowner);
615 	if (mp->msg_control)
616 		MCLAIM(control, so->so_mowner);
617 
618 	if (to) {
619 		sa = mtod(to, struct sockaddr *);
620 	}
621 
622 	len = auio.uio_resid;
623 	error = (*so->so_send)(so, sa, &auio, NULL, control, flags, l);
624 	/* Protocol is responsible for freeing 'control' */
625 	control = NULL;
626 
627 	if (error) {
628 		if (auio.uio_resid != len && (error == ERESTART ||
629 		    error == EINTR || error == EWOULDBLOCK))
630 			error = 0;
631 		if (error == EPIPE && (fp->f_flag & FNOSIGPIPE) == 0 &&
632 		    (flags & MSG_NOSIGNAL) == 0) {
633 			mutex_enter(proc_lock);
634 			psignal(l->l_proc, SIGPIPE);
635 			mutex_exit(proc_lock);
636 		}
637 	}
638 	if (error == 0)
639 		*retsize = len - auio.uio_resid;
640 
641 bad:
642 	if (ktriov != NULL) {
643 		ktrgeniov(s, UIO_WRITE, ktriov, *retsize, error);
644 		kmem_free(ktriov, iovsz);
645 	}
646 
647 	if (iov != aiov)
648 		kmem_free(iov, iovsz);
649 	if (to)
650 		m_freem(to);
651 	if (control)
652 		m_freem(control);
653 
654 	return error;
655 }
656 
657 int
658 do_sys_sendmsg(struct lwp *l, int s, struct msghdr *mp, int flags,
659     register_t *retsize)
660 {
661 	int		error;
662 	struct socket	*so;
663 	file_t		*fp;
664 
665 	if ((error = fd_getsock1(s, &so, &fp)) != 0) {
666 		/* We have to free msg_name and msg_control ourselves */
667 		if (mp->msg_flags & MSG_NAMEMBUF)
668 			m_freem(mp->msg_name);
669 		if (mp->msg_flags & MSG_CONTROLMBUF)
670 			m_freem(mp->msg_control);
671 		return error;
672 	}
673 	error = do_sys_sendmsg_so(l, s, so, fp, mp, flags, retsize);
674 	/* msg_name and msg_control freed */
675 	fd_putfile(s);
676 	return error;
677 }
678 
679 int
680 sys_recvfrom(struct lwp *l, const struct sys_recvfrom_args *uap,
681     register_t *retval)
682 {
683 	/* {
684 		syscallarg(int)			s;
685 		syscallarg(void *)		buf;
686 		syscallarg(size_t)		len;
687 		syscallarg(int)			flags;
688 		syscallarg(struct sockaddr *)	from;
689 		syscallarg(unsigned int *)	fromlenaddr;
690 	} */
691 	struct msghdr	msg;
692 	struct iovec	aiov;
693 	int		error;
694 	struct mbuf	*from;
695 
696 	msg.msg_name = NULL;
697 	msg.msg_iov = &aiov;
698 	msg.msg_iovlen = 1;
699 	aiov.iov_base = SCARG(uap, buf);
700 	aiov.iov_len = SCARG(uap, len);
701 	msg.msg_control = NULL;
702 	msg.msg_flags = SCARG(uap, flags) & MSG_USERFLAGS;
703 
704 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from, NULL, retval);
705 	if (error != 0)
706 		return error;
707 
708 	error = copyout_sockname(SCARG(uap, from), SCARG(uap, fromlenaddr),
709 	    MSG_LENUSRSPACE, from);
710 	if (from != NULL)
711 		m_free(from);
712 	return error;
713 }
714 
715 int
716 sys_recvmsg(struct lwp *l, const struct sys_recvmsg_args *uap,
717     register_t *retval)
718 {
719 	/* {
720 		syscallarg(int)			s;
721 		syscallarg(struct msghdr *)	msg;
722 		syscallarg(int)			flags;
723 	} */
724 	struct msghdr	msg;
725 	int		error;
726 	struct mbuf	*from, *control;
727 
728 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
729 	if (error)
730 		return error;
731 
732 	msg.msg_flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
733 
734 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
735 	    msg.msg_control != NULL ? &control : NULL, retval);
736 	if (error != 0)
737 		return error;
738 
739 	if (msg.msg_control != NULL)
740 		error = copyout_msg_control(l, &msg, control);
741 
742 	if (error == 0)
743 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
744 			from);
745 	if (from != NULL)
746 		m_free(from);
747 	if (error == 0) {
748 		ktrkuser("msghdr", &msg, sizeof(msg));
749 		error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
750 	}
751 
752 	return error;
753 }
754 
755 int
756 sys_sendmmsg(struct lwp *l, const struct sys_sendmmsg_args *uap,
757     register_t *retval)
758 {
759 	/* {
760 		syscallarg(int)			s;
761 		syscallarg(struct mmsghdr *)	mmsg;
762 		syscallarg(unsigned int)	vlen;
763 		syscallarg(unsigned int)	flags;
764 	} */
765 	struct mmsghdr mmsg;
766 	struct socket *so;
767 	file_t *fp;
768 	struct msghdr *msg = &mmsg.msg_hdr;
769 	int error, s;
770 	unsigned int vlen, flags, dg;
771 
772 	s = SCARG(uap, s);
773 	if ((error = fd_getsock1(s, &so, &fp)) != 0)
774 		return error;
775 
776 	vlen = SCARG(uap, vlen);
777 	if (vlen > 1024)
778 		vlen = 1024;
779 
780 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
781 
782 	for (dg = 0; dg < vlen;) {
783 		error = copyin(SCARG(uap, mmsg) + dg, &mmsg, sizeof(mmsg));
784 		if (error)
785 			break;
786 
787 		msg->msg_flags = flags;
788 
789 		error = do_sys_sendmsg_so(l, s, so, fp, msg, flags, retval);
790 		if (error)
791 			break;
792 
793 		ktrkuser("msghdr", msg, sizeof(*msg));
794 		mmsg.msg_len = *retval;
795 		error = copyout(&mmsg, SCARG(uap, mmsg) + dg, sizeof(mmsg));
796 		if (error)
797 			break;
798 		dg++;
799 
800 	}
801 
802 	*retval = dg;
803 	if (error)
804 		so->so_error = error;
805 
806 	fd_putfile(s);
807 
808 	/*
809 	 * If we succeeded at least once, return 0, hopefully so->so_error
810 	 * will catch it next time.
811 	 */
812 	if (dg)
813 		return 0;
814 	return error;
815 }
816 
817 /*
818  * Adjust for a truncated SCM_RIGHTS control message.
819  *  This means closing any file descriptors that aren't present
820  *  in the returned buffer.
821  *  m is the mbuf holding the (already externalized) SCM_RIGHTS message.
822  */
823 static void
824 free_rights(struct mbuf *m)
825 {
826 	struct cmsghdr *cm;
827 	int *fdv;
828 	unsigned int nfds, i;
829 
830 	KASSERT(sizeof(*cm) <= m->m_len);
831 	cm = mtod(m, struct cmsghdr *);
832 
833 	KASSERT(CMSG_ALIGN(sizeof(*cm)) <= cm->cmsg_len);
834 	KASSERT(cm->cmsg_len <= m->m_len);
835 	nfds = (cm->cmsg_len - CMSG_ALIGN(sizeof(*cm))) / sizeof(int);
836 	fdv = (int *)CMSG_DATA(cm);
837 
838 	for (i = 0; i < nfds; i++)
839 		if (fd_getfile(fdv[i]) != NULL)
840 			(void)fd_close(fdv[i]);
841 }
842 
843 void
844 free_control_mbuf(struct lwp *l, struct mbuf *control, struct mbuf *uncopied)
845 {
846 	struct mbuf *next;
847 	struct cmsghdr *cmsg;
848 	bool do_free_rights = false;
849 
850 	while (control != NULL) {
851 		cmsg = mtod(control, struct cmsghdr *);
852 		if (control == uncopied)
853 			do_free_rights = true;
854 		if (do_free_rights && cmsg->cmsg_level == SOL_SOCKET
855 		    && cmsg->cmsg_type == SCM_RIGHTS)
856 			free_rights(control);
857 		next = control->m_next;
858 		m_free(control);
859 		control = next;
860 	}
861 }
862 
863 /* Copy socket control/CMSG data to user buffer, frees the mbuf */
864 int
865 copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
866 {
867 	int i, len, error = 0;
868 	struct cmsghdr *cmsg;
869 	struct mbuf *m;
870 	char *q;
871 
872 	len = mp->msg_controllen;
873 	if (len <= 0 || control == 0) {
874 		mp->msg_controllen = 0;
875 		free_control_mbuf(l, control, control);
876 		return 0;
877 	}
878 
879 	q = (char *)mp->msg_control;
880 
881 	for (m = control; m != NULL; ) {
882 		cmsg = mtod(m, struct cmsghdr *);
883 		i = m->m_len;
884 		if (len < i) {
885 			mp->msg_flags |= MSG_CTRUNC;
886 			if (cmsg->cmsg_level == SOL_SOCKET
887 			    && cmsg->cmsg_type == SCM_RIGHTS)
888 				/* Do not truncate me ... */
889 				break;
890 			i = len;
891 		}
892 		error = copyout(mtod(m, void *), q, i);
893 		ktrkuser(mbuftypes[MT_CONTROL], cmsg, cmsg->cmsg_len);
894 		if (error != 0) {
895 			/* We must free all the SCM_RIGHTS */
896 			m = control;
897 			break;
898 		}
899 		m = m->m_next;
900 		if (m)
901 			i = ALIGN(i);
902 		q += i;
903 		len -= i;
904 		if (len <= 0)
905 			break;
906 	}
907 
908 	free_control_mbuf(l, control, m);
909 
910 	mp->msg_controllen = q - (char *)mp->msg_control;
911 	return error;
912 }
913 
914 int
915 do_sys_recvmsg_so(struct lwp *l, int s, struct socket *so, struct msghdr *mp,
916     struct mbuf **from, struct mbuf **control, register_t *retsize)
917 {
918 	struct iovec	aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
919 	struct uio	auio;
920 	size_t		len, iovsz;
921 	int		i, error;
922 
923 	ktrkuser("msghdr", mp, sizeof(*mp));
924 
925 	*from = NULL;
926 	if (control != NULL)
927 		*control = NULL;
928 
929 	iovsz = mp->msg_iovlen * sizeof(struct iovec);
930 
931 	if (mp->msg_flags & MSG_IOVUSRSPACE) {
932 		if ((unsigned int)mp->msg_iovlen > UIO_SMALLIOV) {
933 			if ((unsigned int)mp->msg_iovlen > IOV_MAX) {
934 				error = EMSGSIZE;
935 				goto out;
936 			}
937 			iov = kmem_alloc(iovsz, KM_SLEEP);
938 		}
939 		if (mp->msg_iovlen != 0) {
940 			error = copyin(mp->msg_iov, iov, iovsz);
941 			if (error)
942 				goto out;
943 		}
944 		auio.uio_iov = iov;
945 	} else
946 		auio.uio_iov = mp->msg_iov;
947 	auio.uio_iovcnt = mp->msg_iovlen;
948 	auio.uio_rw = UIO_READ;
949 	auio.uio_offset = 0;			/* XXX */
950 	auio.uio_resid = 0;
951 	KASSERT(l == curlwp);
952 	auio.uio_vmspace = l->l_proc->p_vmspace;
953 
954 	tiov = auio.uio_iov;
955 	for (i = 0; i < auio.uio_iovcnt; i++, tiov++) {
956 		/*
957 		 * Reads return ssize_t because -1 is returned on error.
958 		 * Therefore we must restrict the length to SSIZE_MAX to
959 		 * avoid garbage return values.
960 		 */
961 		auio.uio_resid += tiov->iov_len;
962 		if (tiov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
963 			error = EINVAL;
964 			goto out;
965 		}
966 	}
967 
968 	if (ktrpoint(KTR_GENIO) && iovsz > 0) {
969 		ktriov = kmem_alloc(iovsz, KM_SLEEP);
970 		memcpy(ktriov, auio.uio_iov, iovsz);
971 	}
972 
973 	len = auio.uio_resid;
974 	mp->msg_flags &= MSG_USERFLAGS;
975 	error = (*so->so_receive)(so, from, &auio, NULL, control,
976 	    &mp->msg_flags);
977 	KASSERT(*from == NULL || (*from)->m_next == NULL);
978 	len -= auio.uio_resid;
979 	*retsize = len;
980 	if (error != 0 && len != 0
981 	    && (error == ERESTART || error == EINTR || error == EWOULDBLOCK))
982 		/* Some data transferred */
983 		error = 0;
984 
985 	if (ktriov != NULL) {
986 		ktrgeniov(s, UIO_READ, ktriov, len, error);
987 		kmem_free(ktriov, iovsz);
988 	}
989 
990 	if (error != 0) {
991 		m_freem(*from);
992 		*from = NULL;
993 		if (control != NULL) {
994 			free_control_mbuf(l, *control, *control);
995 			*control = NULL;
996 		}
997 	}
998  out:
999 	if (iov != aiov)
1000 		kmem_free(iov, iovsz);
1001 	return error;
1002 }
1003 
1004 
1005 int
1006 do_sys_recvmsg(struct lwp *l, int s, struct msghdr *mp,
1007     struct mbuf **from, struct mbuf **control, register_t *retsize)
1008 {
1009 	int error;
1010 	struct socket *so;
1011 
1012 	if ((error = fd_getsock(s, &so)) != 0)
1013 		return error;
1014 	error = do_sys_recvmsg_so(l, s, so, mp, from, control, retsize);
1015 	fd_putfile(s);
1016 	return error;
1017 }
1018 
1019 int
1020 sys_recvmmsg(struct lwp *l, const struct sys_recvmmsg_args *uap,
1021     register_t *retval)
1022 {
1023 	/* {
1024 		syscallarg(int)			s;
1025 		syscallarg(struct mmsghdr *)	mmsg;
1026 		syscallarg(unsigned int)	vlen;
1027 		syscallarg(unsigned int)	flags;
1028 		syscallarg(struct timespec *)	timeout;
1029 	} */
1030 	struct mmsghdr mmsg;
1031 	struct socket *so;
1032 	struct msghdr *msg = &mmsg.msg_hdr;
1033 	int error, s;
1034 	struct mbuf *from, *control;
1035 	struct timespec ts, now;
1036 	unsigned int vlen, flags, dg;
1037 
1038 	if (SCARG(uap, timeout)) {
1039 		if ((error = copyin(SCARG(uap, timeout), &ts, sizeof(ts))) != 0)
1040 			return error;
1041 		getnanotime(&now);
1042 		timespecadd(&now, &ts, &ts);
1043 	}
1044 
1045 	s = SCARG(uap, s);
1046 	if ((error = fd_getsock(s, &so)) != 0)
1047 		return error;
1048 
1049 	vlen = SCARG(uap, vlen);
1050 	if (vlen > 1024)
1051 		vlen = 1024;
1052 
1053 	from = NULL;
1054 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1055 
1056 	for (dg = 0; dg < vlen;) {
1057 		error = copyin(SCARG(uap, mmsg) + dg, &mmsg, sizeof(mmsg));
1058 		if (error)
1059 			break;
1060 
1061 		msg->msg_flags = flags & ~MSG_WAITFORONE;
1062 
1063 		if (from != NULL) {
1064 			m_free(from);
1065 			from = NULL;
1066 		}
1067 
1068 		error = do_sys_recvmsg_so(l, s, so, msg, &from,
1069 		    msg->msg_control != NULL ? &control : NULL, retval);
1070 		if (error) {
1071 			if (error == EAGAIN && dg > 0)
1072 				error = 0;
1073 			break;
1074 		}
1075 
1076 		if (msg->msg_control != NULL)
1077 			error = copyout_msg_control(l, msg, control);
1078 		if (error)
1079 			break;
1080 
1081 		error = copyout_sockname(msg->msg_name, &msg->msg_namelen, 0,
1082 		    from);
1083 		if (error)
1084 			break;
1085 
1086 		ktrkuser("msghdr", msg, sizeof *msg);
1087 		mmsg.msg_len = *retval;
1088 
1089 		error = copyout(&mmsg, SCARG(uap, mmsg) + dg, sizeof(mmsg));
1090 		if (error)
1091 			break;
1092 
1093 		dg++;
1094 		if (msg->msg_flags & MSG_OOB)
1095 			break;
1096 
1097 		if (SCARG(uap, timeout)) {
1098 			getnanotime(&now);
1099 			timespecsub(&now, &ts, &now);
1100 			if (now.tv_sec > 0)
1101 				break;
1102 		}
1103 
1104 		if (flags & MSG_WAITFORONE)
1105 			flags |= MSG_DONTWAIT;
1106 
1107 	}
1108 
1109 	if (from != NULL)
1110 		m_free(from);
1111 
1112 	*retval = dg;
1113 	if (error)
1114 		so->so_error = error;
1115 
1116 	fd_putfile(s);
1117 
1118 	/*
1119 	 * If we succeeded at least once, return 0, hopefully so->so_error
1120 	 * will catch it next time.
1121 	 */
1122 	if (dg)
1123 		return 0;
1124 
1125 	return error;
1126 }
1127 
1128 int
1129 sys_shutdown(struct lwp *l, const struct sys_shutdown_args *uap,
1130     register_t *retval)
1131 {
1132 	/* {
1133 		syscallarg(int)	s;
1134 		syscallarg(int)	how;
1135 	} */
1136 	struct socket	*so;
1137 	int		error;
1138 
1139 	if ((error = fd_getsock(SCARG(uap, s), &so)) != 0)
1140 		return error;
1141 	solock(so);
1142 	error = soshutdown(so, SCARG(uap, how));
1143 	sounlock(so);
1144 	fd_putfile(SCARG(uap, s));
1145 	return error;
1146 }
1147 
1148 int
1149 sys_setsockopt(struct lwp *l, const struct sys_setsockopt_args *uap,
1150     register_t *retval)
1151 {
1152 	/* {
1153 		syscallarg(int)			s;
1154 		syscallarg(int)			level;
1155 		syscallarg(int)			name;
1156 		syscallarg(const void *)	val;
1157 		syscallarg(unsigned int)	valsize;
1158 	} */
1159 	struct sockopt	sopt;
1160 	struct socket	*so;
1161 	file_t		*fp;
1162 	int		error;
1163 	unsigned int	len;
1164 
1165 	len = SCARG(uap, valsize);
1166 	if (len > 0 && SCARG(uap, val) == NULL)
1167 		return EINVAL;
1168 
1169 	if (len > MCLBYTES)
1170 		return EINVAL;
1171 
1172 	if ((error = fd_getsock1(SCARG(uap, s), &so, &fp)) != 0)
1173 		return (error);
1174 
1175 	sockopt_init(&sopt, SCARG(uap, level), SCARG(uap, name), len);
1176 
1177 	if (len > 0) {
1178 		error = copyin(SCARG(uap, val), sopt.sopt_data, len);
1179 		if (error)
1180 			goto out;
1181 	}
1182 
1183 	error = sosetopt(so, &sopt);
1184 	if (so->so_options & SO_NOSIGPIPE)
1185 		atomic_or_uint(&fp->f_flag, FNOSIGPIPE);
1186 	else
1187 		atomic_and_uint(&fp->f_flag, ~FNOSIGPIPE);
1188 
1189  out:
1190 	sockopt_destroy(&sopt);
1191 	fd_putfile(SCARG(uap, s));
1192 	return error;
1193 }
1194 
1195 int
1196 sys_getsockopt(struct lwp *l, const struct sys_getsockopt_args *uap,
1197     register_t *retval)
1198 {
1199 	/* {
1200 		syscallarg(int)			s;
1201 		syscallarg(int)			level;
1202 		syscallarg(int)			name;
1203 		syscallarg(void *)		val;
1204 		syscallarg(unsigned int *)	avalsize;
1205 	} */
1206 	struct sockopt	sopt;
1207 	struct socket	*so;
1208 	file_t		*fp;
1209 	unsigned int	valsize, len;
1210 	int		error;
1211 
1212 	if (SCARG(uap, val) != NULL) {
1213 		error = copyin(SCARG(uap, avalsize), &valsize, sizeof(valsize));
1214 		if (error)
1215 			return error;
1216 	} else
1217 		valsize = 0;
1218 
1219 	if ((error = fd_getsock1(SCARG(uap, s), &so, &fp)) != 0)
1220 		return (error);
1221 
1222 	if (valsize > MCLBYTES)
1223 		return EINVAL;
1224 
1225 	sockopt_init(&sopt, SCARG(uap, level), SCARG(uap, name), valsize);
1226 
1227 	if (fp->f_flag & FNOSIGPIPE)
1228 		so->so_options |= SO_NOSIGPIPE;
1229 	else
1230 		so->so_options &= ~SO_NOSIGPIPE;
1231 	error = sogetopt(so, &sopt);
1232 	if (error)
1233 		goto out;
1234 
1235 	if (valsize > 0) {
1236 		len = min(valsize, sopt.sopt_retsize);
1237 		error = copyout(sopt.sopt_data, SCARG(uap, val), len);
1238 		if (error)
1239 			goto out;
1240 
1241 		error = copyout(&len, SCARG(uap, avalsize), sizeof(len));
1242 		if (error)
1243 			goto out;
1244 	}
1245 
1246  out:
1247 	sockopt_destroy(&sopt);
1248 	fd_putfile(SCARG(uap, s));
1249 	return error;
1250 }
1251 
1252 #ifdef PIPE_SOCKETPAIR
1253 
1254 int
1255 pipe1(struct lwp *l, int *fildes, int flags)
1256 {
1257 	file_t		*rf, *wf;
1258 	struct socket	*rso, *wso;
1259 	int		fd, error;
1260 	proc_t		*p;
1261 
1262 	if (flags & ~(O_CLOEXEC|O_NONBLOCK|O_NOSIGPIPE))
1263 		return EINVAL;
1264 	p = curproc;
1265 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, l, NULL)) != 0)
1266 		return error;
1267 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, l, rso)) != 0)
1268 		goto free1;
1269 	/* remember this socket pair implements a pipe */
1270 	wso->so_state |= SS_ISAPIPE;
1271 	rso->so_state |= SS_ISAPIPE;
1272 	if ((error = fd_allocfile(&rf, &fd)) != 0)
1273 		goto free2;
1274 	fildes[0] = fd;
1275 	rf->f_flag = FREAD | flags;
1276 	rf->f_type = DTYPE_SOCKET;
1277 	rf->f_ops = &socketops;
1278 	rf->f_socket = rso;
1279 	if ((error = fd_allocfile(&wf, &fd)) != 0)
1280 		goto free3;
1281 	wf->f_flag = FWRITE | flags;
1282 	wf->f_type = DTYPE_SOCKET;
1283 	wf->f_ops = &socketops;
1284 	wf->f_socket = wso;
1285 	fildes[1] = fd;
1286 	solock(wso);
1287 	error = unp_connect2(wso, rso);
1288 	sounlock(wso);
1289 	if (error != 0)
1290 		goto free4;
1291 	fd_affix(p, wf, fildes[1]);
1292 	fd_affix(p, rf, fildes[0]);
1293 	return (0);
1294  free4:
1295 	fd_abort(p, wf, fildes[1]);
1296  free3:
1297 	fd_abort(p, rf, fildes[0]);
1298  free2:
1299 	(void)soclose(wso);
1300  free1:
1301 	(void)soclose(rso);
1302 	return error;
1303 }
1304 #endif /* PIPE_SOCKETPAIR */
1305 
1306 /*
1307  * Get peer socket name.
1308  */
1309 int
1310 do_sys_getpeername(int fd, struct sockaddr *nam)
1311 {
1312 	struct socket	*so;
1313 	int		error;
1314 
1315 	if ((error = fd_getsock(fd, &so)) != 0)
1316 		return error;
1317 
1318 	solock(so);
1319 	if ((so->so_state & SS_ISCONNECTED) == 0)
1320 		error = ENOTCONN;
1321 	else {
1322 		error = (*so->so_proto->pr_usrreqs->pr_peeraddr)(so, nam);
1323 	}
1324 	sounlock(so);
1325 	fd_putfile(fd);
1326 	return error;
1327 }
1328 
1329 /*
1330  * Get local socket name.
1331  */
1332 int
1333 do_sys_getsockname(int fd, struct sockaddr *nam)
1334 {
1335 	struct socket	*so;
1336 	int		error;
1337 
1338 	if ((error = fd_getsock(fd, &so)) != 0)
1339 		return error;
1340 
1341 	solock(so);
1342 	error = (*so->so_proto->pr_usrreqs->pr_sockaddr)(so, nam);
1343 	sounlock(so);
1344 	fd_putfile(fd);
1345 	return error;
1346 }
1347 
1348 int
1349 copyout_sockname_sb(struct sockaddr *asa, unsigned int *alen, int flags,
1350     struct sockaddr_big *addr)
1351 {
1352 	unsigned int len;
1353 	int error;
1354 
1355 	if (asa == NULL)
1356 		/* Assume application not interested */
1357 		return 0;
1358 
1359 	if (flags & MSG_LENUSRSPACE) {
1360 		error = copyin(alen, &len, sizeof(len));
1361 		if (error)
1362 			return error;
1363 	} else
1364 		len = *alen;
1365 
1366 	if (addr == NULL) {
1367 		len = 0;
1368 		error = 0;
1369 	} else {
1370 		if (len > addr->sb_len)
1371 			len = addr->sb_len;
1372 		/* XXX addr isn't an mbuf... */
1373 		ktrkuser(mbuftypes[MT_SONAME], addr, len);
1374 		error = copyout(addr, asa, len);
1375 	}
1376 
1377 	if (error == 0) {
1378 		if (flags & MSG_LENUSRSPACE)
1379 			error = copyout(&len, alen, sizeof(len));
1380 		else
1381 			*alen = len;
1382 	}
1383 
1384 	return error;
1385 }
1386 
1387 int
1388 copyout_sockname(struct sockaddr *asa, unsigned int *alen, int flags,
1389     struct mbuf *addr)
1390 {
1391 	int len;
1392 	int error;
1393 
1394 	if (asa == NULL)
1395 		/* Assume application not interested */
1396 		return 0;
1397 
1398 	if (flags & MSG_LENUSRSPACE) {
1399 		error = copyin(alen, &len, sizeof(len));
1400 		if (error)
1401 			return error;
1402 	} else
1403 		len = *alen;
1404 	if (len < 0)
1405 		return EINVAL;
1406 
1407 	if (addr == NULL) {
1408 		len = 0;
1409 		error = 0;
1410 	} else {
1411 		if (len > addr->m_len)
1412 			len = addr->m_len;
1413 		/* Maybe this ought to copy a chain ? */
1414 		ktrkuser(mbuftypes[MT_SONAME], mtod(addr, void *), len);
1415 		error = copyout(mtod(addr, void *), asa, len);
1416 	}
1417 
1418 	if (error == 0) {
1419 		if (flags & MSG_LENUSRSPACE)
1420 			error = copyout(&len, alen, sizeof(len));
1421 		else
1422 			*alen = len;
1423 	}
1424 
1425 	return error;
1426 }
1427 
1428 /*
1429  * Get socket name.
1430  */
1431 int
1432 sys_getsockname(struct lwp *l, const struct sys_getsockname_args *uap,
1433     register_t *retval)
1434 {
1435 	/* {
1436 		syscallarg(int)			fdes;
1437 		syscallarg(struct sockaddr *)	asa;
1438 		syscallarg(unsigned int *)	alen;
1439 	} */
1440 	struct sockaddr_big sbig;
1441 	int		    error;
1442 
1443 	sbig.sb_len = UCHAR_MAX;
1444 	error = do_sys_getsockname(SCARG(uap, fdes), (struct sockaddr *)&sbig);
1445 	if (error != 0)
1446 		return error;
1447 
1448 	error = copyout_sockname_sb(SCARG(uap, asa), SCARG(uap, alen),
1449 	    MSG_LENUSRSPACE, &sbig);
1450 	return error;
1451 }
1452 
1453 /*
1454  * Get name of peer for connected socket.
1455  */
1456 int
1457 sys_getpeername(struct lwp *l, const struct sys_getpeername_args *uap,
1458     register_t *retval)
1459 {
1460 	/* {
1461 		syscallarg(int)			fdes;
1462 		syscallarg(struct sockaddr *)	asa;
1463 		syscallarg(unsigned int *)	alen;
1464 	} */
1465 	struct sockaddr_big sbig;
1466 	int		    error;
1467 
1468 	sbig.sb_len = UCHAR_MAX;
1469 	error = do_sys_getpeername(SCARG(uap, fdes), (struct sockaddr *)&sbig);
1470 	if (error != 0)
1471 		return error;
1472 
1473 	error = copyout_sockname_sb(SCARG(uap, asa), SCARG(uap, alen),
1474 	    MSG_LENUSRSPACE, &sbig);
1475 	return error;
1476 }
1477 
1478 static int
1479 sockargs_sb(struct sockaddr_big *sb, const void *name, socklen_t buflen)
1480 {
1481 	int error;
1482 
1483 	/*
1484 	 * We can't allow socket names > UCHAR_MAX in length, since that
1485 	 * will overflow sb_len. Further no reasonable buflen is <=
1486 	 * offsetof(sockaddr_big, sb_data) since it shall be at least
1487 	 * the size of the preamble sb_len and sb_family members.
1488 	 */
1489 	if (buflen > UCHAR_MAX ||
1490 	    buflen <= offsetof(struct sockaddr_big, sb_data))
1491 		return EINVAL;
1492 
1493 	error = copyin(name, (void *)sb, buflen);
1494 	if (error)
1495 		return error;
1496 
1497 	ktrkuser(mbuftypes[MT_SONAME], sb, buflen);
1498 #if BYTE_ORDER != BIG_ENDIAN
1499 	/*
1500 	 * 4.3BSD compat thing - need to stay, since bind(2),
1501 	 * connect(2), sendto(2) were not versioned for COMPAT_43.
1502 	 */
1503 	if (sb->sb_family == 0 && sb->sb_len < AF_MAX)
1504 		sb->sb_family = sb->sb_len;
1505 #endif
1506 	sb->sb_len = buflen;
1507 	return 0;
1508 }
1509 
1510 /*
1511  * XXX In a perfect world, we wouldn't pass around socket control
1512  * XXX arguments in mbufs, and this could go away.
1513  */
1514 int
1515 sockargs(struct mbuf **mp, const void *bf, size_t buflen, enum uio_seg seg,
1516     int type)
1517 {
1518 	struct mbuf	*m;
1519 	int		error;
1520 
1521 	/*
1522 	 * We can't allow socket names > UCHAR_MAX in length, since that
1523 	 * will overflow sa_len.  Control data more than a page size in
1524 	 * length is just too much.
1525 	 */
1526 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
1527 		return EINVAL;
1528 
1529 	/*
1530 	 * length must greater than sizeof(sa_family) + sizeof(sa_len)
1531 	 */
1532 	if (type == MT_SONAME && buflen <= 2)
1533 		return EINVAL;
1534 
1535 	/* Allocate an mbuf to hold the arguments. */
1536 	m = m_get(M_WAIT, type);
1537 	/* can't claim.  don't who to assign it to. */
1538 	if (buflen > MLEN) {
1539 		/*
1540 		 * Won't fit into a regular mbuf, so we allocate just
1541 		 * enough external storage to hold the argument.
1542 		 */
1543 		MEXTMALLOC(m, buflen, M_WAITOK);
1544 	}
1545 	m->m_len = buflen;
1546 	if (seg == UIO_USERSPACE) {
1547 		error = copyin(bf, mtod(m, void *), buflen);
1548 		if (error) {
1549 			(void)m_free(m);
1550 			return error;
1551 		}
1552 	} else {
1553 		memcpy(mtod(m, void *), bf, buflen);
1554 	}
1555 	*mp = m;
1556 	switch (type) {
1557 	case MT_SONAME:
1558 		ktrkuser(mbuftypes[type], mtod(m, void *), buflen);
1559 
1560 		struct sockaddr *sa = mtod(m, struct sockaddr *);
1561 #if BYTE_ORDER != BIG_ENDIAN
1562 		/*
1563 		 * 4.3BSD compat thing - need to stay, since bind(2),
1564 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
1565 		 */
1566 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1567 			sa->sa_family = sa->sa_len;
1568 #endif
1569 		sa->sa_len = buflen;
1570 		return 0;
1571 	case MT_CONTROL:
1572 		if (!KTRPOINT(curproc, KTR_USER))
1573 			return 0;
1574 
1575 		struct msghdr mhdr;
1576 		mhdr.msg_control = mtod(m, void *);
1577 		mhdr.msg_controllen = buflen;
1578 		for (struct cmsghdr *cmsg = CMSG_FIRSTHDR(&mhdr); cmsg;
1579 		    cmsg = CMSG_NXTHDR(&mhdr, cmsg)) {
1580 			ktrkuser(mbuftypes[type], cmsg, cmsg->cmsg_len);
1581 		}
1582 		return 0;
1583 	default:
1584 		return EINVAL;
1585 	}
1586 }
1587