xref: /netbsd-src/sys/kern/uipc_syscalls.c (revision e89934bbf778a6d6d6894877c4da59d0c7835b0f)
1 /*	$NetBSD: uipc_syscalls.c,v 1.186 2017/02/03 16:06:45 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 1.186 2017/02/03 16:06:45 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 
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 	    NULL, 0, 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 	    NULL, 0, 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, const void *kthdr, size_t ktsize,
534     register_t *retsize)
535 {
536 
537 	struct iovec	aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
538 	struct sockaddr *sa = NULL;
539 	struct mbuf	*to, *control;
540 	struct uio	auio;
541 	size_t		len, iovsz;
542 	int		i, error;
543 
544 	if (__predict_false(kthdr == NULL && ktsize == 0)) {
545 		kthdr = mp;
546 		ktsize = sizeof(*mp);
547 	}
548 	if (__predict_true(kthdr != NULL))
549 		ktrkuser("msghdr", kthdr, ktsize);
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 		    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, 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
664 do_sys_sendmsg(struct lwp *l, int s, struct msghdr *mp, int flags,
665     const void *kthdr, size_t ktsize, 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, kthdr, ktsize,
680 	    retsize);
681 	/* msg_name and msg_control freed */
682 	fd_putfile(s);
683 	return error;
684 }
685 
686 int
687 sys_recvfrom(struct lwp *l, const struct sys_recvfrom_args *uap,
688     register_t *retval)
689 {
690 	/* {
691 		syscallarg(int)			s;
692 		syscallarg(void *)		buf;
693 		syscallarg(size_t)		len;
694 		syscallarg(int)			flags;
695 		syscallarg(struct sockaddr *)	from;
696 		syscallarg(unsigned int *)	fromlenaddr;
697 	} */
698 	struct msghdr	msg;
699 	struct iovec	aiov;
700 	int		error;
701 	struct mbuf	*from;
702 
703 	msg.msg_name = NULL;
704 	msg.msg_iov = &aiov;
705 	msg.msg_iovlen = 1;
706 	aiov.iov_base = SCARG(uap, buf);
707 	aiov.iov_len = SCARG(uap, len);
708 	msg.msg_control = NULL;
709 	msg.msg_flags = SCARG(uap, flags) & MSG_USERFLAGS;
710 
711 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, NULL, 0, &from,
712 	    NULL, retval);
713 	if (error != 0)
714 		return error;
715 
716 	error = copyout_sockname(SCARG(uap, from), SCARG(uap, fromlenaddr),
717 	    MSG_LENUSRSPACE, from);
718 	if (from != NULL)
719 		m_free(from);
720 	return error;
721 }
722 
723 int
724 sys_recvmsg(struct lwp *l, const struct sys_recvmsg_args *uap,
725     register_t *retval)
726 {
727 	/* {
728 		syscallarg(int)			s;
729 		syscallarg(struct msghdr *)	msg;
730 		syscallarg(int)			flags;
731 	} */
732 	struct msghdr	msg;
733 	int		error;
734 	struct mbuf	*from, *control;
735 
736 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
737 	if (error)
738 		return error;
739 
740 	msg.msg_flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
741 
742 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, NULL, 0, &from,
743 	    msg.msg_control != NULL ? &control : NULL, retval);
744 	if (error != 0)
745 		return error;
746 
747 	if (msg.msg_control != NULL)
748 		error = copyout_msg_control(l, &msg, control);
749 
750 	if (error == 0)
751 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
752 			from);
753 	if (from != NULL)
754 		m_free(from);
755 	if (error == 0) {
756 		ktrkuser("msghdr", &msg, sizeof msg);
757 		error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
758 	}
759 
760 	return error;
761 }
762 
763 int
764 sys_sendmmsg(struct lwp *l, const struct sys_sendmmsg_args *uap,
765     register_t *retval)
766 {
767 	/* {
768 		syscallarg(int)			s;
769 		syscallarg(struct mmsghdr *)	mmsg;
770 		syscallarg(unsigned int)	vlen;
771 		syscallarg(unsigned int)	flags;
772 	} */
773 	struct mmsghdr mmsg;
774 	struct socket *so;
775 	file_t *fp;
776 	struct msghdr *msg = &mmsg.msg_hdr;
777 	int error, s;
778 	unsigned int vlen, flags, dg;
779 
780 	s = SCARG(uap, s);
781 	if ((error = fd_getsock1(s, &so, &fp)) != 0)
782 		return error;
783 
784 	vlen = SCARG(uap, vlen);
785 	if (vlen > 1024)
786 		vlen = 1024;
787 
788 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
789 
790 	for (dg = 0; dg < vlen;) {
791 		error = copyin(SCARG(uap, mmsg) + dg, &mmsg, sizeof(mmsg));
792 		if (error)
793 			break;
794 
795 		msg->msg_flags = flags;
796 
797 		error = do_sys_sendmsg_so(l, s, so, fp, msg, flags,
798 		    &msg, sizeof(msg), retval);
799 		if (error)
800 			break;
801 
802 		ktrkuser("msghdr", msg, sizeof *msg);
803 		mmsg.msg_len = *retval;
804 		error = copyout(&mmsg, SCARG(uap, mmsg) + dg, sizeof(mmsg));
805 		if (error)
806 			break;
807 		dg++;
808 
809 	}
810 
811 	*retval = dg;
812 	if (error)
813 		so->so_error = error;
814 
815 	fd_putfile(s);
816 
817 	/*
818 	 * If we succeeded at least once, return 0, hopefully so->so_error
819 	 * will catch it next time.
820 	 */
821 	if (dg)
822 		return 0;
823 	return error;
824 }
825 
826 /*
827  * Adjust for a truncated SCM_RIGHTS control message.
828  *  This means closing any file descriptors that aren't present
829  *  in the returned buffer.
830  *  m is the mbuf holding the (already externalized) SCM_RIGHTS message.
831  */
832 static void
833 free_rights(struct mbuf *m)
834 {
835 	struct cmsghdr *cm;
836 	int *fdv;
837 	unsigned int nfds, i;
838 
839 	KASSERT(sizeof(*cm) <= m->m_len);
840 	cm = mtod(m, struct cmsghdr *);
841 
842 	KASSERT(CMSG_ALIGN(sizeof(*cm)) <= cm->cmsg_len);
843 	KASSERT(cm->cmsg_len <= m->m_len);
844 	nfds = (cm->cmsg_len - CMSG_ALIGN(sizeof(*cm))) / sizeof(int);
845 	fdv = (int *)CMSG_DATA(cm);
846 
847 	for (i = 0; i < nfds; i++)
848 		if (fd_getfile(fdv[i]) != NULL)
849 			(void)fd_close(fdv[i]);
850 }
851 
852 void
853 free_control_mbuf(struct lwp *l, struct mbuf *control, struct mbuf *uncopied)
854 {
855 	struct mbuf *next;
856 	struct cmsghdr *cmsg;
857 	bool do_free_rights = false;
858 
859 	while (control != NULL) {
860 		cmsg = mtod(control, struct cmsghdr *);
861 		if (control == uncopied)
862 			do_free_rights = true;
863 		if (do_free_rights && cmsg->cmsg_level == SOL_SOCKET
864 		    && cmsg->cmsg_type == SCM_RIGHTS)
865 			free_rights(control);
866 		next = control->m_next;
867 		m_free(control);
868 		control = next;
869 	}
870 }
871 
872 /* Copy socket control/CMSG data to user buffer, frees the mbuf */
873 int
874 copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
875 {
876 	int i, len, error = 0;
877 	struct cmsghdr *cmsg;
878 	struct mbuf *m;
879 	char *q;
880 
881 	len = mp->msg_controllen;
882 	if (len <= 0 || control == 0) {
883 		mp->msg_controllen = 0;
884 		free_control_mbuf(l, control, control);
885 		return 0;
886 	}
887 
888 	q = (char *)mp->msg_control;
889 
890 	for (m = control; m != NULL; ) {
891 		cmsg = mtod(m, struct cmsghdr *);
892 		i = m->m_len;
893 		if (len < i) {
894 			mp->msg_flags |= MSG_CTRUNC;
895 			if (cmsg->cmsg_level == SOL_SOCKET
896 			    && cmsg->cmsg_type == SCM_RIGHTS)
897 				/* Do not truncate me ... */
898 				break;
899 			i = len;
900 		}
901 		error = copyout(mtod(m, void *), q, i);
902 		ktrkuser("msgcontrol", mtod(m, void *), i);
903 		if (error != 0) {
904 			/* We must free all the SCM_RIGHTS */
905 			m = control;
906 			break;
907 		}
908 		m = m->m_next;
909 		if (m)
910 			i = ALIGN(i);
911 		q += i;
912 		len -= i;
913 		if (len <= 0)
914 			break;
915 	}
916 
917 	free_control_mbuf(l, control, m);
918 
919 	mp->msg_controllen = q - (char *)mp->msg_control;
920 	return error;
921 }
922 
923 int
924 do_sys_recvmsg_so(struct lwp *l, int s, struct socket *so, struct msghdr *mp,
925     const void *ktrhdr, size_t ktsize, struct mbuf **from,
926     struct mbuf **control, register_t *retsize)
927 {
928 	struct iovec	aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
929 	struct uio	auio;
930 	size_t		len, iovsz;
931 	int		i, error;
932 
933 	if (__predict_false(ktrhdr == NULL && ktsize == 0)) {
934 		ktrhdr = mp;
935 		ktsize = sizeof *mp;
936 	}
937 	if (__predict_true(ktrhdr != NULL))
938 		ktrkuser("msghdr", ktrhdr, ktsize);
939 
940 	*from = NULL;
941 	if (control != NULL)
942 		*control = NULL;
943 
944 	iovsz = mp->msg_iovlen * sizeof(struct iovec);
945 
946 	if (mp->msg_flags & MSG_IOVUSRSPACE) {
947 		if ((unsigned int)mp->msg_iovlen > UIO_SMALLIOV) {
948 			if ((unsigned int)mp->msg_iovlen > IOV_MAX) {
949 				error = EMSGSIZE;
950 				goto out;
951 			}
952 			iov = kmem_alloc(iovsz, KM_SLEEP);
953 		}
954 		if (mp->msg_iovlen != 0) {
955 			error = copyin(mp->msg_iov, iov, iovsz);
956 			if (error)
957 				goto out;
958 		}
959 		auio.uio_iov = iov;
960 	} else
961 		auio.uio_iov = mp->msg_iov;
962 	auio.uio_iovcnt = mp->msg_iovlen;
963 	auio.uio_rw = UIO_READ;
964 	auio.uio_offset = 0;			/* XXX */
965 	auio.uio_resid = 0;
966 	KASSERT(l == curlwp);
967 	auio.uio_vmspace = l->l_proc->p_vmspace;
968 
969 	tiov = auio.uio_iov;
970 	for (i = 0; i < auio.uio_iovcnt; i++, tiov++) {
971 		/*
972 		 * Reads return ssize_t because -1 is returned on error.
973 		 * Therefore we must restrict the length to SSIZE_MAX to
974 		 * avoid garbage return values.
975 		 */
976 		auio.uio_resid += tiov->iov_len;
977 		if (tiov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
978 			error = EINVAL;
979 			goto out;
980 		}
981 	}
982 
983 	if (ktrpoint(KTR_GENIO) && iovsz > 0) {
984 		ktriov = kmem_alloc(iovsz, KM_SLEEP);
985 		memcpy(ktriov, auio.uio_iov, iovsz);
986 	}
987 
988 	len = auio.uio_resid;
989 	mp->msg_flags &= MSG_USERFLAGS;
990 	error = (*so->so_receive)(so, from, &auio, NULL, control,
991 	    &mp->msg_flags);
992 	len -= auio.uio_resid;
993 	*retsize = len;
994 	if (error != 0 && len != 0
995 	    && (error == ERESTART || error == EINTR || error == EWOULDBLOCK))
996 		/* Some data transferred */
997 		error = 0;
998 
999 	if (ktriov != NULL) {
1000 		ktrgeniov(s, UIO_READ, ktriov, len, error);
1001 		kmem_free(ktriov, iovsz);
1002 	}
1003 
1004 	if (error != 0) {
1005 		m_freem(*from);
1006 		*from = NULL;
1007 		if (control != NULL) {
1008 			free_control_mbuf(l, *control, *control);
1009 			*control = NULL;
1010 		}
1011 	}
1012  out:
1013 	if (iov != aiov)
1014 		kmem_free(iov, iovsz);
1015 	return error;
1016 }
1017 
1018 
1019 int
1020 do_sys_recvmsg(struct lwp *l, int s, struct msghdr *mp,
1021     const void *ktrhdr, size_t ktrsize,
1022     struct mbuf **from, struct mbuf **control, register_t *retsize)
1023 {
1024 	int error;
1025 	struct socket *so;
1026 
1027 	if ((error = fd_getsock(s, &so)) != 0)
1028 		return error;
1029 	error = do_sys_recvmsg_so(l, s, so, mp, ktrhdr, ktrsize, from,
1030 	    control, retsize);
1031 	fd_putfile(s);
1032 	return error;
1033 }
1034 
1035 int
1036 sys_recvmmsg(struct lwp *l, const struct sys_recvmmsg_args *uap,
1037     register_t *retval)
1038 {
1039 	/* {
1040 		syscallarg(int)			s;
1041 		syscallarg(struct mmsghdr *)	mmsg;
1042 		syscallarg(unsigned int)	vlen;
1043 		syscallarg(unsigned int)	flags;
1044 		syscallarg(struct timespec *)	timeout;
1045 	} */
1046 	struct mmsghdr mmsg;
1047 	struct socket *so;
1048 	struct msghdr *msg = &mmsg.msg_hdr;
1049 	int error, s;
1050 	struct mbuf *from, *control;
1051 	struct timespec ts, now;
1052 	unsigned int vlen, flags, dg;
1053 
1054 	if (SCARG(uap, timeout)) {
1055 		if ((error = copyin(SCARG(uap, timeout), &ts, sizeof(ts))) != 0)
1056 			return error;
1057 		getnanotime(&now);
1058 		timespecadd(&now, &ts, &ts);
1059 	}
1060 
1061 	s = SCARG(uap, s);
1062 	if ((error = fd_getsock(s, &so)) != 0)
1063 		return error;
1064 
1065 	vlen = SCARG(uap, vlen);
1066 	if (vlen > 1024)
1067 		vlen = 1024;
1068 
1069 	from = NULL;
1070 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1071 
1072 	for (dg = 0; dg < vlen;) {
1073 		error = copyin(SCARG(uap, mmsg) + dg, &mmsg, sizeof(mmsg));
1074 		if (error)
1075 			break;
1076 
1077 		msg->msg_flags = flags & ~MSG_WAITFORONE;
1078 
1079 		if (from != NULL) {
1080 			m_free(from);
1081 			from = NULL;
1082 		}
1083 
1084 		error = do_sys_recvmsg_so(l, s, so, msg, NULL, 0, &from,
1085 		    msg->msg_control != NULL ? &control : NULL, retval);
1086 		if (error) {
1087 			if (error == EAGAIN && dg > 0)
1088 				error = 0;
1089 			break;
1090 		}
1091 
1092 		if (msg->msg_control != NULL)
1093 			error = copyout_msg_control(l, msg, control);
1094 		if (error)
1095 			break;
1096 
1097 		error = copyout_sockname(msg->msg_name, &msg->msg_namelen, 0,
1098 		    from);
1099 		if (error)
1100 			break;
1101 
1102 		ktrkuser("msghdr", msg, sizeof *msg);
1103 		mmsg.msg_len = *retval;
1104 
1105 		error = copyout(&mmsg, SCARG(uap, mmsg) + dg, sizeof(mmsg));
1106 		if (error)
1107 			break;
1108 
1109 		dg++;
1110 		if (msg->msg_flags & MSG_OOB)
1111 			break;
1112 
1113 		if (SCARG(uap, timeout)) {
1114 			getnanotime(&now);
1115 			timespecsub(&now, &ts, &now);
1116 			if (now.tv_sec > 0)
1117 				break;
1118 		}
1119 
1120 		if (flags & MSG_WAITFORONE)
1121 			flags |= MSG_DONTWAIT;
1122 
1123 	}
1124 
1125 	if (from != NULL)
1126 		m_free(from);
1127 
1128 	*retval = dg;
1129 	if (error)
1130 		so->so_error = error;
1131 
1132 	fd_putfile(s);
1133 
1134 	/*
1135 	 * If we succeeded at least once, return 0, hopefully so->so_error
1136 	 * will catch it next time.
1137 	 */
1138 	if (dg)
1139 		return 0;
1140 
1141 	return error;
1142 }
1143 
1144 int
1145 sys_shutdown(struct lwp *l, const struct sys_shutdown_args *uap,
1146     register_t *retval)
1147 {
1148 	/* {
1149 		syscallarg(int)	s;
1150 		syscallarg(int)	how;
1151 	} */
1152 	struct socket	*so;
1153 	int		error;
1154 
1155 	if ((error = fd_getsock(SCARG(uap, s), &so)) != 0)
1156 		return error;
1157 	solock(so);
1158 	error = soshutdown(so, SCARG(uap, how));
1159 	sounlock(so);
1160 	fd_putfile(SCARG(uap, s));
1161 	return error;
1162 }
1163 
1164 int
1165 sys_setsockopt(struct lwp *l, const struct sys_setsockopt_args *uap,
1166     register_t *retval)
1167 {
1168 	/* {
1169 		syscallarg(int)			s;
1170 		syscallarg(int)			level;
1171 		syscallarg(int)			name;
1172 		syscallarg(const void *)	val;
1173 		syscallarg(unsigned int)	valsize;
1174 	} */
1175 	struct sockopt	sopt;
1176 	struct socket	*so;
1177 	file_t		*fp;
1178 	int		error;
1179 	unsigned int	len;
1180 
1181 	len = SCARG(uap, valsize);
1182 	if (len > 0 && SCARG(uap, val) == NULL)
1183 		return EINVAL;
1184 
1185 	if (len > MCLBYTES)
1186 		return EINVAL;
1187 
1188 	if ((error = fd_getsock1(SCARG(uap, s), &so, &fp)) != 0)
1189 		return (error);
1190 
1191 	sockopt_init(&sopt, SCARG(uap, level), SCARG(uap, name), len);
1192 
1193 	if (len > 0) {
1194 		error = copyin(SCARG(uap, val), sopt.sopt_data, len);
1195 		if (error)
1196 			goto out;
1197 	}
1198 
1199 	error = sosetopt(so, &sopt);
1200 	if (so->so_options & SO_NOSIGPIPE)
1201 		atomic_or_uint(&fp->f_flag, FNOSIGPIPE);
1202 	else
1203 		atomic_and_uint(&fp->f_flag, ~FNOSIGPIPE);
1204 
1205  out:
1206 	sockopt_destroy(&sopt);
1207 	fd_putfile(SCARG(uap, s));
1208 	return error;
1209 }
1210 
1211 int
1212 sys_getsockopt(struct lwp *l, const struct sys_getsockopt_args *uap,
1213     register_t *retval)
1214 {
1215 	/* {
1216 		syscallarg(int)			s;
1217 		syscallarg(int)			level;
1218 		syscallarg(int)			name;
1219 		syscallarg(void *)		val;
1220 		syscallarg(unsigned int *)	avalsize;
1221 	} */
1222 	struct sockopt	sopt;
1223 	struct socket	*so;
1224 	file_t		*fp;
1225 	unsigned int	valsize, len;
1226 	int		error;
1227 
1228 	if (SCARG(uap, val) != NULL) {
1229 		error = copyin(SCARG(uap, avalsize), &valsize, sizeof(valsize));
1230 		if (error)
1231 			return error;
1232 	} else
1233 		valsize = 0;
1234 
1235 	if ((error = fd_getsock1(SCARG(uap, s), &so, &fp)) != 0)
1236 		return (error);
1237 
1238 	sockopt_init(&sopt, SCARG(uap, level), SCARG(uap, name), 0);
1239 
1240 	if (fp->f_flag & FNOSIGPIPE)
1241 		so->so_options |= SO_NOSIGPIPE;
1242 	else
1243 		so->so_options &= ~SO_NOSIGPIPE;
1244 	error = sogetopt(so, &sopt);
1245 	if (error)
1246 		goto out;
1247 
1248 	if (valsize > 0) {
1249 		len = min(valsize, sopt.sopt_size);
1250 		error = copyout(sopt.sopt_data, SCARG(uap, val), len);
1251 		if (error)
1252 			goto out;
1253 
1254 		error = copyout(&len, SCARG(uap, avalsize), sizeof(len));
1255 		if (error)
1256 			goto out;
1257 	}
1258 
1259  out:
1260 	sockopt_destroy(&sopt);
1261 	fd_putfile(SCARG(uap, s));
1262 	return error;
1263 }
1264 
1265 #ifdef PIPE_SOCKETPAIR
1266 
1267 int
1268 pipe1(struct lwp *l, register_t *retval, int flags)
1269 {
1270 	file_t		*rf, *wf;
1271 	struct socket	*rso, *wso;
1272 	int		fd, error;
1273 	proc_t		*p;
1274 
1275 	if (flags & ~(O_CLOEXEC|O_NONBLOCK|O_NOSIGPIPE))
1276 		return EINVAL;
1277 	p = curproc;
1278 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, l, NULL)) != 0)
1279 		return error;
1280 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, l, rso)) != 0)
1281 		goto free1;
1282 	/* remember this socket pair implements a pipe */
1283 	wso->so_state |= SS_ISAPIPE;
1284 	rso->so_state |= SS_ISAPIPE;
1285 	if ((error = fd_allocfile(&rf, &fd)) != 0)
1286 		goto free2;
1287 	retval[0] = fd;
1288 	rf->f_flag = FREAD | flags;
1289 	rf->f_type = DTYPE_SOCKET;
1290 	rf->f_ops = &socketops;
1291 	rf->f_socket = rso;
1292 	if ((error = fd_allocfile(&wf, &fd)) != 0)
1293 		goto free3;
1294 	wf->f_flag = FWRITE | flags;
1295 	wf->f_type = DTYPE_SOCKET;
1296 	wf->f_ops = &socketops;
1297 	wf->f_socket = wso;
1298 	retval[1] = fd;
1299 	solock(wso);
1300 	error = unp_connect2(wso, rso);
1301 	sounlock(wso);
1302 	if (error != 0)
1303 		goto free4;
1304 	fd_affix(p, wf, (int)retval[1]);
1305 	fd_affix(p, rf, (int)retval[0]);
1306 	return (0);
1307  free4:
1308 	fd_abort(p, wf, (int)retval[1]);
1309  free3:
1310 	fd_abort(p, rf, (int)retval[0]);
1311  free2:
1312 	(void)soclose(wso);
1313  free1:
1314 	(void)soclose(rso);
1315 	return error;
1316 }
1317 #endif /* PIPE_SOCKETPAIR */
1318 
1319 /*
1320  * Get peer socket name.
1321  */
1322 int
1323 do_sys_getpeername(int fd, struct sockaddr *nam)
1324 {
1325 	struct socket	*so;
1326 	int		error;
1327 
1328 	if ((error = fd_getsock(fd, &so)) != 0)
1329 		return error;
1330 
1331 	solock(so);
1332 	if ((so->so_state & SS_ISCONNECTED) == 0)
1333 		error = ENOTCONN;
1334 	else {
1335 		error = (*so->so_proto->pr_usrreqs->pr_peeraddr)(so, nam);
1336 	}
1337 	sounlock(so);
1338 	fd_putfile(fd);
1339 	return error;
1340 }
1341 
1342 /*
1343  * Get local socket name.
1344  */
1345 int
1346 do_sys_getsockname(int fd, struct sockaddr *nam)
1347 {
1348 	struct socket	*so;
1349 	int		error;
1350 
1351 	if ((error = fd_getsock(fd, &so)) != 0)
1352 		return error;
1353 
1354 	solock(so);
1355 	error = (*so->so_proto->pr_usrreqs->pr_sockaddr)(so, nam);
1356 	sounlock(so);
1357 	fd_putfile(fd);
1358 	return error;
1359 }
1360 
1361 int
1362 copyout_sockname_sb(struct sockaddr *asa, unsigned int *alen, int flags,
1363     struct sockaddr_big *addr)
1364 {
1365 	int len;
1366 	int error;
1367 
1368 	if (asa == NULL)
1369 		/* Assume application not interested */
1370 		return 0;
1371 
1372 	if (flags & MSG_LENUSRSPACE) {
1373 		error = copyin(alen, &len, sizeof(len));
1374 		if (error)
1375 			return error;
1376 	} else
1377 		len = *alen;
1378 	if (len < 0)
1379 		return EINVAL;
1380 
1381 	if (addr == NULL) {
1382 		len = 0;
1383 		error = 0;
1384 	} else {
1385 		if (len > addr->sb_len)
1386 			len = addr->sb_len;
1387 		/* XXX addr isn't an mbuf... */
1388 		ktrkuser(mbuftypes[MT_SONAME], addr, len);
1389 		error = copyout(addr, asa, len);
1390 	}
1391 
1392 	if (error == 0) {
1393 		if (flags & MSG_LENUSRSPACE)
1394 			error = copyout(&len, alen, sizeof(len));
1395 		else
1396 			*alen = len;
1397 	}
1398 
1399 	return error;
1400 }
1401 
1402 int
1403 copyout_sockname(struct sockaddr *asa, unsigned int *alen, int flags,
1404     struct mbuf *addr)
1405 {
1406 	int len;
1407 	int error;
1408 
1409 	if (asa == NULL)
1410 		/* Assume application not interested */
1411 		return 0;
1412 
1413 	if (flags & MSG_LENUSRSPACE) {
1414 		error = copyin(alen, &len, sizeof(len));
1415 		if (error)
1416 			return error;
1417 	} else
1418 		len = *alen;
1419 	if (len < 0)
1420 		return EINVAL;
1421 
1422 	if (addr == NULL) {
1423 		len = 0;
1424 		error = 0;
1425 	} else {
1426 		if (len > addr->m_len)
1427 			len = addr->m_len;
1428 		/* Maybe this ought to copy a chain ? */
1429 		ktrkuser(mbuftypes[MT_SONAME], mtod(addr, void *), len);
1430 		error = copyout(mtod(addr, void *), 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 /*
1444  * Get socket name.
1445  */
1446 int
1447 sys_getsockname(struct lwp *l, const struct sys_getsockname_args *uap,
1448     register_t *retval)
1449 {
1450 	/* {
1451 		syscallarg(int)			fdes;
1452 		syscallarg(struct sockaddr *)	asa;
1453 		syscallarg(unsigned int *)	alen;
1454 	} */
1455 	struct sockaddr_big sbig;
1456 	int		    error;
1457 
1458 	sbig.sb_len = UCHAR_MAX;
1459 	error = do_sys_getsockname(SCARG(uap, fdes), (struct sockaddr *)&sbig);
1460 	if (error != 0)
1461 		return error;
1462 
1463 	error = copyout_sockname_sb(SCARG(uap, asa), SCARG(uap, alen),
1464 	    MSG_LENUSRSPACE, &sbig);
1465 	return error;
1466 }
1467 
1468 /*
1469  * Get name of peer for connected socket.
1470  */
1471 int
1472 sys_getpeername(struct lwp *l, const struct sys_getpeername_args *uap,
1473     register_t *retval)
1474 {
1475 	/* {
1476 		syscallarg(int)			fdes;
1477 		syscallarg(struct sockaddr *)	asa;
1478 		syscallarg(unsigned int *)	alen;
1479 	} */
1480 	struct sockaddr_big sbig;
1481 	int		    error;
1482 
1483 	sbig.sb_len = UCHAR_MAX;
1484 	error = do_sys_getpeername(SCARG(uap, fdes), (struct sockaddr *)&sbig);
1485 	if (error != 0)
1486 		return error;
1487 
1488 	error = copyout_sockname_sb(SCARG(uap, asa), SCARG(uap, alen),
1489 	    MSG_LENUSRSPACE, &sbig);
1490 	return error;
1491 }
1492 
1493 static int
1494 sockargs_sb(struct sockaddr_big *sb, const void *name, socklen_t buflen)
1495 {
1496 	int error;
1497 
1498 	/*
1499 	 * We can't allow socket names > UCHAR_MAX in length, since that
1500 	 * will overflow sb_len. Further no reasonable buflen is <=
1501 	 * offsetof(sockaddr_big, sb_data) since it shall be at least
1502 	 * the size of the preamble sb_len and sb_family members.
1503 	 */
1504 	if (buflen > UCHAR_MAX ||
1505 	    buflen <= offsetof(struct sockaddr_big, sb_data))
1506 		return EINVAL;
1507 
1508 	error = copyin(name, (void *)sb, buflen);
1509 	if (error)
1510 		return error;
1511 
1512 	ktrkuser(mbuftypes[MT_SONAME], sb, buflen);
1513 #if BYTE_ORDER != BIG_ENDIAN
1514 	/*
1515 	 * 4.3BSD compat thing - need to stay, since bind(2),
1516 	 * connect(2), sendto(2) were not versioned for COMPAT_43.
1517 	 */
1518 	if (sb->sb_family == 0 && sb->sb_len < AF_MAX)
1519 		sb->sb_family = sb->sb_len;
1520 #endif
1521 	sb->sb_len = buflen;
1522 	return 0;
1523 }
1524 
1525 /*
1526  * XXX In a perfect world, we wouldn't pass around socket control
1527  * XXX arguments in mbufs, and this could go away.
1528  */
1529 int
1530 sockargs(struct mbuf **mp, const void *bf, size_t buflen, int type)
1531 {
1532 	struct sockaddr	*sa;
1533 	struct mbuf	*m;
1534 	int		error;
1535 
1536 	/*
1537 	 * We can't allow socket names > UCHAR_MAX in length, since that
1538 	 * will overflow sa_len.  Control data more than a page size in
1539 	 * length is just too much.
1540 	 */
1541 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
1542 		return EINVAL;
1543 
1544 	/*
1545 	 * length must greater than sizeof(sa_family) + sizeof(sa_len)
1546 	 */
1547 	if (type == MT_SONAME && buflen <= 2)
1548 		return EINVAL;
1549 
1550 	/* Allocate an mbuf to hold the arguments. */
1551 	m = m_get(M_WAIT, type);
1552 	/* can't claim.  don't who to assign it to. */
1553 	if (buflen > MLEN) {
1554 		/*
1555 		 * Won't fit into a regular mbuf, so we allocate just
1556 		 * enough external storage to hold the argument.
1557 		 */
1558 		MEXTMALLOC(m, buflen, M_WAITOK);
1559 	}
1560 	m->m_len = buflen;
1561 	error = copyin(bf, mtod(m, void *), buflen);
1562 	if (error) {
1563 		(void)m_free(m);
1564 		return error;
1565 	}
1566 	ktrkuser(mbuftypes[type], mtod(m, void *), buflen);
1567 	*mp = m;
1568 	if (type == MT_SONAME) {
1569 		sa = mtod(m, struct sockaddr *);
1570 #if BYTE_ORDER != BIG_ENDIAN
1571 		/*
1572 		 * 4.3BSD compat thing - need to stay, since bind(2),
1573 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
1574 		 */
1575 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1576 			sa->sa_family = sa->sa_len;
1577 #endif
1578 		sa->sa_len = buflen;
1579 	}
1580 	return 0;
1581 }
1582