xref: /freebsd/sys/kern/uipc_syscalls.c (revision f0adf7f5cdd241db2f2c817683191a6ef64a4e95)
1 /*
2  * Copyright (c) 1982, 1986, 1989, 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * sendfile(2) and related extensions:
6  * Copyright (c) 1998, David Greenman. All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)uipc_syscalls.c	8.4 (Berkeley) 2/21/94
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #include "opt_compat.h"
39 #include "opt_ktrace.h"
40 #include "opt_mac.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/lock.h>
46 #include <sys/mac.h>
47 #include <sys/mutex.h>
48 #include <sys/sysproto.h>
49 #include <sys/malloc.h>
50 #include <sys/filedesc.h>
51 #include <sys/event.h>
52 #include <sys/proc.h>
53 #include <sys/fcntl.h>
54 #include <sys/file.h>
55 #include <sys/filio.h>
56 #include <sys/mount.h>
57 #include <sys/mbuf.h>
58 #include <sys/protosw.h>
59 #include <sys/sf_buf.h>
60 #include <sys/socket.h>
61 #include <sys/socketvar.h>
62 #include <sys/signalvar.h>
63 #include <sys/syscallsubr.h>
64 #include <sys/sysctl.h>
65 #include <sys/uio.h>
66 #include <sys/vnode.h>
67 #ifdef KTRACE
68 #include <sys/ktrace.h>
69 #endif
70 
71 #include <vm/vm.h>
72 #include <vm/vm_object.h>
73 #include <vm/vm_page.h>
74 #include <vm/vm_pageout.h>
75 #include <vm/vm_kern.h>
76 #include <vm/vm_extern.h>
77 
78 static int sendit(struct thread *td, int s, struct msghdr *mp, int flags);
79 static int recvit(struct thread *td, int s, struct msghdr *mp, void *namelenp);
80 
81 static int accept1(struct thread *td, struct accept_args *uap, int compat);
82 static int do_sendfile(struct thread *td, struct sendfile_args *uap, int compat);
83 static int getsockname1(struct thread *td, struct getsockname_args *uap,
84 			int compat);
85 static int getpeername1(struct thread *td, struct getpeername_args *uap,
86 			int compat);
87 
88 /*
89  * NSFBUFS-related variables and associated sysctls
90  */
91 int nsfbufs;
92 int nsfbufspeak;
93 int nsfbufsused;
94 
95 SYSCTL_DECL(_kern_ipc);
96 SYSCTL_INT(_kern_ipc, OID_AUTO, nsfbufs, CTLFLAG_RDTUN, &nsfbufs, 0,
97     "Maximum number of sendfile(2) sf_bufs available");
98 SYSCTL_INT(_kern_ipc, OID_AUTO, nsfbufspeak, CTLFLAG_RD, &nsfbufspeak, 0,
99     "Number of sendfile(2) sf_bufs at peak usage");
100 SYSCTL_INT(_kern_ipc, OID_AUTO, nsfbufsused, CTLFLAG_RD, &nsfbufsused, 0,
101     "Number of sendfile(2) sf_bufs in use");
102 
103 /*
104  * System call interface to the socket abstraction.
105  */
106 #if defined(COMPAT_43)
107 #define COMPAT_OLDSOCK
108 #endif
109 
110 /*
111  * MPSAFE
112  */
113 int
114 socket(td, uap)
115 	struct thread *td;
116 	register struct socket_args /* {
117 		int	domain;
118 		int	type;
119 		int	protocol;
120 	} */ *uap;
121 {
122 	struct filedesc *fdp;
123 	struct socket *so;
124 	struct file *fp;
125 	int fd, error;
126 
127 	fdp = td->td_proc->p_fd;
128 	error = falloc(td, &fp, &fd);
129 	if (error)
130 		return (error);
131 	/* An extra reference on `fp' has been held for us by falloc(). */
132 	NET_LOCK_GIANT();
133 	error = socreate(uap->domain, &so, uap->type, uap->protocol,
134 	    td->td_ucred, td);
135 	NET_UNLOCK_GIANT();
136 	FILEDESC_LOCK(fdp);
137 	if (error) {
138 		if (fdp->fd_ofiles[fd] == fp) {
139 			fdp->fd_ofiles[fd] = NULL;
140 			fdunused(fdp, fd);
141 			FILEDESC_UNLOCK(fdp);
142 			fdrop(fp, td);
143 		} else {
144 			FILEDESC_UNLOCK(fdp);
145 		}
146 	} else {
147 		fp->f_data = so;	/* already has ref count */
148 		fp->f_flag = FREAD|FWRITE;
149 		fp->f_ops = &socketops;
150 		fp->f_type = DTYPE_SOCKET;
151 		FILEDESC_UNLOCK(fdp);
152 		td->td_retval[0] = fd;
153 	}
154 	fdrop(fp, td);
155 	return (error);
156 }
157 
158 /*
159  * MPSAFE
160  */
161 /* ARGSUSED */
162 int
163 bind(td, uap)
164 	struct thread *td;
165 	register struct bind_args /* {
166 		int	s;
167 		caddr_t	name;
168 		int	namelen;
169 	} */ *uap;
170 {
171 	struct sockaddr *sa;
172 	int error;
173 
174 	if ((error = getsockaddr(&sa, uap->name, uap->namelen)) != 0)
175 		return (error);
176 
177 	return (kern_bind(td, uap->s, sa));
178 }
179 
180 int
181 kern_bind(td, fd, sa)
182 	struct thread *td;
183 	int fd;
184 	struct sockaddr *sa;
185 {
186 	struct socket *so;
187 	int error;
188 
189 	NET_LOCK_GIANT();
190 	if ((error = fgetsock(td, fd, &so, NULL)) != 0)
191 		goto done2;
192 #ifdef MAC
193 	SOCK_LOCK(so);
194 	error = mac_check_socket_bind(td->td_ucred, so, sa);
195 	SOCK_UNLOCK(so);
196 	if (error)
197 		goto done1;
198 #endif
199 	error = sobind(so, sa, td);
200 #ifdef MAC
201 done1:
202 #endif
203 	fputsock(so);
204 done2:
205 	NET_UNLOCK_GIANT();
206 	FREE(sa, M_SONAME);
207 	return (error);
208 }
209 
210 /*
211  * MPSAFE
212  */
213 /* ARGSUSED */
214 int
215 listen(td, uap)
216 	struct thread *td;
217 	register struct listen_args /* {
218 		int	s;
219 		int	backlog;
220 	} */ *uap;
221 {
222 	struct socket *so;
223 	int error;
224 
225 	NET_LOCK_GIANT();
226 	if ((error = fgetsock(td, uap->s, &so, NULL)) == 0) {
227 #ifdef MAC
228 		SOCK_LOCK(so);
229 		error = mac_check_socket_listen(td->td_ucred, so);
230 		SOCK_UNLOCK(so);
231 		if (error)
232 			goto done;
233 #endif
234 		error = solisten(so, uap->backlog, td);
235 #ifdef MAC
236 done:
237 #endif
238 		fputsock(so);
239 	}
240 	NET_UNLOCK_GIANT();
241 	return(error);
242 }
243 
244 /*
245  * accept1()
246  * MPSAFE
247  */
248 static int
249 accept1(td, uap, compat)
250 	struct thread *td;
251 	register struct accept_args /* {
252 		int	s;
253 		struct sockaddr	* __restrict name;
254 		socklen_t	* __restrict anamelen;
255 	} */ *uap;
256 	int compat;
257 {
258 	struct filedesc *fdp;
259 	struct file *nfp = NULL;
260 	struct sockaddr *sa = NULL;
261 	socklen_t namelen;
262 	int error;
263 	struct socket *head, *so;
264 	int fd;
265 	u_int fflag;
266 	pid_t pgid;
267 	int tmp;
268 
269 	fdp = td->td_proc->p_fd;
270 	if (uap->name) {
271 		error = copyin(uap->anamelen, &namelen, sizeof (namelen));
272 		if(error)
273 			return (error);
274 		if (namelen < 0)
275 			return (EINVAL);
276 	}
277 	NET_LOCK_GIANT();
278 	error = fgetsock(td, uap->s, &head, &fflag);
279 	if (error)
280 		goto done2;
281 	if ((head->so_options & SO_ACCEPTCONN) == 0) {
282 		error = EINVAL;
283 		goto done;
284 	}
285 	error = falloc(td, &nfp, &fd);
286 	if (error)
287 		goto done;
288 	ACCEPT_LOCK();
289 	if ((head->so_state & SS_NBIO) && TAILQ_EMPTY(&head->so_comp)) {
290 		ACCEPT_UNLOCK();
291 		error = EWOULDBLOCK;
292 		goto noconnection;
293 	}
294 	while (TAILQ_EMPTY(&head->so_comp) && head->so_error == 0) {
295 		if (head->so_rcv.sb_state & SBS_CANTRCVMORE) {
296 			head->so_error = ECONNABORTED;
297 			break;
298 		}
299 		error = msleep(&head->so_timeo, &accept_mtx, PSOCK | PCATCH,
300 		    "accept", 0);
301 		if (error) {
302 			ACCEPT_UNLOCK();
303 			goto noconnection;
304 		}
305 	}
306 	if (head->so_error) {
307 		error = head->so_error;
308 		head->so_error = 0;
309 		ACCEPT_UNLOCK();
310 		goto noconnection;
311 	}
312 	so = TAILQ_FIRST(&head->so_comp);
313 	KASSERT(!(so->so_qstate & SQ_INCOMP), ("accept1: so SQ_INCOMP"));
314 	KASSERT(so->so_qstate & SQ_COMP, ("accept1: so not SQ_COMP"));
315 
316 	/*
317 	 * Before changing the flags on the socket, we have to bump the
318 	 * reference count.  Otherwise, if the protocol calls sofree(),
319 	 * the socket will be released due to a zero refcount.
320 	 */
321 	SOCK_LOCK(so);
322 	soref(so);			/* file descriptor reference */
323 	SOCK_UNLOCK(so);
324 
325 	TAILQ_REMOVE(&head->so_comp, so, so_list);
326 	head->so_qlen--;
327 	so->so_state |= (head->so_state & SS_NBIO);
328 	so->so_qstate &= ~SQ_COMP;
329 	so->so_head = NULL;
330 
331 	ACCEPT_UNLOCK();
332 
333 	/* An extra reference on `nfp' has been held for us by falloc(). */
334 	td->td_retval[0] = fd;
335 
336 	/* connection has been removed from the listen queue */
337 	KNOTE(&head->so_rcv.sb_sel.si_note, 0);
338 
339 	pgid = fgetown(&head->so_sigio);
340 	if (pgid != 0)
341 		fsetown(pgid, &so->so_sigio);
342 
343 	FILE_LOCK(nfp);
344 	nfp->f_data = so;	/* nfp has ref count from falloc */
345 	nfp->f_flag = fflag;
346 	nfp->f_ops = &socketops;
347 	nfp->f_type = DTYPE_SOCKET;
348 	FILE_UNLOCK(nfp);
349 	/* Sync socket nonblocking/async state with file flags */
350 	tmp = fflag & FNONBLOCK;
351 	(void) fo_ioctl(nfp, FIONBIO, &tmp, td->td_ucred, td);
352 	tmp = fflag & FASYNC;
353 	(void) fo_ioctl(nfp, FIOASYNC, &tmp, td->td_ucred, td);
354 	sa = 0;
355 	error = soaccept(so, &sa);
356 	if (error) {
357 		/*
358 		 * return a namelen of zero for older code which might
359 		 * ignore the return value from accept.
360 		 */
361 		if (uap->name != NULL) {
362 			namelen = 0;
363 			(void) copyout(&namelen,
364 			    uap->anamelen, sizeof(*uap->anamelen));
365 		}
366 		goto noconnection;
367 	}
368 	if (sa == NULL) {
369 		namelen = 0;
370 		if (uap->name)
371 			goto gotnoname;
372 		error = 0;
373 		goto done;
374 	}
375 	if (uap->name) {
376 		/* check sa_len before it is destroyed */
377 		if (namelen > sa->sa_len)
378 			namelen = sa->sa_len;
379 #ifdef COMPAT_OLDSOCK
380 		if (compat)
381 			((struct osockaddr *)sa)->sa_family =
382 			    sa->sa_family;
383 #endif
384 		error = copyout(sa, uap->name, (u_int)namelen);
385 		if (!error)
386 gotnoname:
387 			error = copyout(&namelen,
388 			    uap->anamelen, sizeof (*uap->anamelen));
389 	}
390 noconnection:
391 	if (sa)
392 		FREE(sa, M_SONAME);
393 
394 	/*
395 	 * close the new descriptor, assuming someone hasn't ripped it
396 	 * out from under us.
397 	 */
398 	if (error) {
399 		FILEDESC_LOCK(fdp);
400 		if (fdp->fd_ofiles[fd] == nfp) {
401 			fdp->fd_ofiles[fd] = NULL;
402 			fdunused(fdp, fd);
403 			FILEDESC_UNLOCK(fdp);
404 			fdrop(nfp, td);
405 		} else {
406 			FILEDESC_UNLOCK(fdp);
407 		}
408 	}
409 
410 	/*
411 	 * Release explicitly held references before returning.
412 	 */
413 done:
414 	if (nfp != NULL)
415 		fdrop(nfp, td);
416 	fputsock(head);
417 done2:
418 	NET_UNLOCK_GIANT();
419 	return (error);
420 }
421 
422 /*
423  * MPSAFE (accept1() is MPSAFE)
424  */
425 int
426 accept(td, uap)
427 	struct thread *td;
428 	struct accept_args *uap;
429 {
430 
431 	return (accept1(td, uap, 0));
432 }
433 
434 #ifdef COMPAT_OLDSOCK
435 /*
436  * MPSAFE (accept1() is MPSAFE)
437  */
438 int
439 oaccept(td, uap)
440 	struct thread *td;
441 	struct accept_args *uap;
442 {
443 
444 	return (accept1(td, uap, 1));
445 }
446 #endif /* COMPAT_OLDSOCK */
447 
448 /*
449  * MPSAFE
450  */
451 /* ARGSUSED */
452 int
453 connect(td, uap)
454 	struct thread *td;
455 	register struct connect_args /* {
456 		int	s;
457 		caddr_t	name;
458 		int	namelen;
459 	} */ *uap;
460 {
461 	struct sockaddr *sa;
462 	int error;
463 
464 	error = getsockaddr(&sa, uap->name, uap->namelen);
465 	if (error)
466 		return (error);
467 
468 	return (kern_connect(td, uap->s, sa));
469 }
470 
471 
472 int
473 kern_connect(td, fd, sa)
474 	struct thread *td;
475 	int fd;
476 	struct sockaddr *sa;
477 {
478 	struct socket *so;
479 	int error, s;
480 	int interrupted = 0;
481 
482 	NET_LOCK_GIANT();
483 	if ((error = fgetsock(td, fd, &so, NULL)) != 0)
484 		goto done2;
485 	if (so->so_state & SS_ISCONNECTING) {
486 		error = EALREADY;
487 		goto done1;
488 	}
489 #ifdef MAC
490 	SOCK_LOCK(so);
491 	error = mac_check_socket_connect(td->td_ucred, so, sa);
492 	SOCK_UNLOCK(so);
493 	if (error)
494 		goto bad;
495 #endif
496 	error = soconnect(so, sa, td);
497 	if (error)
498 		goto bad;
499 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
500 		error = EINPROGRESS;
501 		goto done1;
502 	}
503 	s = splnet();
504 	SOCK_LOCK(so);
505 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
506 		error = msleep(&so->so_timeo, SOCK_MTX(so), PSOCK | PCATCH,
507 		    "connec", 0);
508 		if (error) {
509 			if (error == EINTR || error == ERESTART)
510 				interrupted = 1;
511 			break;
512 		}
513 	}
514 	if (error == 0) {
515 		error = so->so_error;
516 		so->so_error = 0;
517 	}
518 	SOCK_UNLOCK(so);
519 	splx(s);
520 bad:
521 	if (!interrupted)
522 		so->so_state &= ~SS_ISCONNECTING;
523 	if (error == ERESTART)
524 		error = EINTR;
525 done1:
526 	fputsock(so);
527 done2:
528 	NET_UNLOCK_GIANT();
529 	FREE(sa, M_SONAME);
530 	return (error);
531 }
532 
533 /*
534  * MPSAFE
535  */
536 int
537 socketpair(td, uap)
538 	struct thread *td;
539 	register struct socketpair_args /* {
540 		int	domain;
541 		int	type;
542 		int	protocol;
543 		int	*rsv;
544 	} */ *uap;
545 {
546 	register struct filedesc *fdp = td->td_proc->p_fd;
547 	struct file *fp1, *fp2;
548 	struct socket *so1, *so2;
549 	int fd, error, sv[2];
550 
551 	NET_LOCK_GIANT();
552 	error = socreate(uap->domain, &so1, uap->type, uap->protocol,
553 	    td->td_ucred, td);
554 	if (error)
555 		goto done2;
556 	error = socreate(uap->domain, &so2, uap->type, uap->protocol,
557 	    td->td_ucred, td);
558 	if (error)
559 		goto free1;
560 	/* On success extra reference to `fp1' and 'fp2' is set by falloc. */
561 	error = falloc(td, &fp1, &fd);
562 	if (error)
563 		goto free2;
564 	sv[0] = fd;
565 	fp1->f_data = so1;	/* so1 already has ref count */
566 	error = falloc(td, &fp2, &fd);
567 	if (error)
568 		goto free3;
569 	fp2->f_data = so2;	/* so2 already has ref count */
570 	sv[1] = fd;
571 	error = soconnect2(so1, so2);
572 	if (error)
573 		goto free4;
574 	if (uap->type == SOCK_DGRAM) {
575 		/*
576 		 * Datagram socket connection is asymmetric.
577 		 */
578 		 error = soconnect2(so2, so1);
579 		 if (error)
580 			goto free4;
581 	}
582 	FILE_LOCK(fp1);
583 	fp1->f_flag = FREAD|FWRITE;
584 	fp1->f_ops = &socketops;
585 	fp1->f_type = DTYPE_SOCKET;
586 	FILE_UNLOCK(fp1);
587 	FILE_LOCK(fp2);
588 	fp2->f_flag = FREAD|FWRITE;
589 	fp2->f_ops = &socketops;
590 	fp2->f_type = DTYPE_SOCKET;
591 	FILE_UNLOCK(fp2);
592 	error = copyout(sv, uap->rsv, 2 * sizeof (int));
593 	fdrop(fp1, td);
594 	fdrop(fp2, td);
595 	goto done2;
596 free4:
597 	FILEDESC_LOCK(fdp);
598 	if (fdp->fd_ofiles[sv[1]] == fp2) {
599 		fdp->fd_ofiles[sv[1]] = NULL;
600 		fdunused(fdp, sv[1]);
601 		FILEDESC_UNLOCK(fdp);
602 		fdrop(fp2, td);
603 	} else {
604 		FILEDESC_UNLOCK(fdp);
605 	}
606 	fdrop(fp2, td);
607 free3:
608 	FILEDESC_LOCK(fdp);
609 	if (fdp->fd_ofiles[sv[0]] == fp1) {
610 		fdp->fd_ofiles[sv[0]] = NULL;
611 		fdunused(fdp, sv[0]);
612 		FILEDESC_UNLOCK(fdp);
613 		fdrop(fp1, td);
614 	} else {
615 		FILEDESC_UNLOCK(fdp);
616 	}
617 	fdrop(fp1, td);
618 free2:
619 	(void)soclose(so2);
620 free1:
621 	(void)soclose(so1);
622 done2:
623 	NET_UNLOCK_GIANT();
624 	return (error);
625 }
626 
627 static int
628 sendit(td, s, mp, flags)
629 	register struct thread *td;
630 	int s;
631 	register struct msghdr *mp;
632 	int flags;
633 {
634 	struct mbuf *control;
635 	struct sockaddr *to;
636 	int error;
637 
638 	if (mp->msg_name != NULL) {
639 		error = getsockaddr(&to, mp->msg_name, mp->msg_namelen);
640 		if (error) {
641 			to = NULL;
642 			goto bad;
643 		}
644 		mp->msg_name = to;
645 	} else {
646 		to = NULL;
647 	}
648 
649 	if (mp->msg_control) {
650 		if (mp->msg_controllen < sizeof(struct cmsghdr)
651 #ifdef COMPAT_OLDSOCK
652 		    && mp->msg_flags != MSG_COMPAT
653 #endif
654 		) {
655 			error = EINVAL;
656 			goto bad;
657 		}
658 		error = sockargs(&control, mp->msg_control,
659 		    mp->msg_controllen, MT_CONTROL);
660 		if (error)
661 			goto bad;
662 #ifdef COMPAT_OLDSOCK
663 		if (mp->msg_flags == MSG_COMPAT) {
664 			register struct cmsghdr *cm;
665 
666 			M_PREPEND(control, sizeof(*cm), M_TRYWAIT);
667 			if (control == 0) {
668 				error = ENOBUFS;
669 				goto bad;
670 			} else {
671 				cm = mtod(control, struct cmsghdr *);
672 				cm->cmsg_len = control->m_len;
673 				cm->cmsg_level = SOL_SOCKET;
674 				cm->cmsg_type = SCM_RIGHTS;
675 			}
676 		}
677 #endif
678 	} else {
679 		control = NULL;
680 	}
681 
682 	error = kern_sendit(td, s, mp, flags, control);
683 
684 bad:
685 	if (to)
686 		FREE(to, M_SONAME);
687 	return (error);
688 }
689 
690 int
691 kern_sendit(td, s, mp, flags, control)
692 	struct thread *td;
693 	int s;
694 	struct msghdr *mp;
695 	int flags;
696 	struct mbuf *control;
697 {
698 	struct uio auio;
699 	struct iovec *iov;
700 	struct socket *so;
701 	int i;
702 	int len, error;
703 #ifdef KTRACE
704 	struct uio *ktruio = NULL;
705 #endif
706 
707 	NET_LOCK_GIANT();
708 	if ((error = fgetsock(td, s, &so, NULL)) != 0)
709 		goto bad2;
710 
711 #ifdef MAC
712 	SOCK_LOCK(so);
713 	error = mac_check_socket_send(td->td_ucred, so);
714 	SOCK_UNLOCK(so);
715 	if (error)
716 		goto bad;
717 #endif
718 
719 	auio.uio_iov = mp->msg_iov;
720 	auio.uio_iovcnt = mp->msg_iovlen;
721 	auio.uio_segflg = UIO_USERSPACE;
722 	auio.uio_rw = UIO_WRITE;
723 	auio.uio_td = td;
724 	auio.uio_offset = 0;			/* XXX */
725 	auio.uio_resid = 0;
726 	iov = mp->msg_iov;
727 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
728 		if ((auio.uio_resid += iov->iov_len) < 0) {
729 			error = EINVAL;
730 			goto bad;
731 		}
732 	}
733 #ifdef KTRACE
734 	if (KTRPOINT(td, KTR_GENIO))
735 		ktruio = cloneuio(&auio);
736 #endif
737 	len = auio.uio_resid;
738 	error = so->so_proto->pr_usrreqs->pru_sosend(so, mp->msg_name, &auio,
739 	    0, control, flags, td);
740 	if (error) {
741 		if (auio.uio_resid != len && (error == ERESTART ||
742 		    error == EINTR || error == EWOULDBLOCK))
743 			error = 0;
744 		/* Generation of SIGPIPE can be controlled per socket */
745 		if (error == EPIPE && !(so->so_options & SO_NOSIGPIPE)) {
746 			PROC_LOCK(td->td_proc);
747 			psignal(td->td_proc, SIGPIPE);
748 			PROC_UNLOCK(td->td_proc);
749 		}
750 	}
751 	if (error == 0)
752 		td->td_retval[0] = len - auio.uio_resid;
753 #ifdef KTRACE
754 	if (ktruio != NULL) {
755 		ktruio->uio_resid = td->td_retval[0];
756 		ktrgenio(s, UIO_WRITE, ktruio, error);
757 	}
758 #endif
759 bad:
760 	fputsock(so);
761 bad2:
762 	NET_UNLOCK_GIANT();
763 	return (error);
764 }
765 
766 /*
767  * MPSAFE
768  */
769 int
770 sendto(td, uap)
771 	struct thread *td;
772 	register struct sendto_args /* {
773 		int	s;
774 		caddr_t	buf;
775 		size_t	len;
776 		int	flags;
777 		caddr_t	to;
778 		int	tolen;
779 	} */ *uap;
780 {
781 	struct msghdr msg;
782 	struct iovec aiov;
783 	int error;
784 
785 	msg.msg_name = uap->to;
786 	msg.msg_namelen = uap->tolen;
787 	msg.msg_iov = &aiov;
788 	msg.msg_iovlen = 1;
789 	msg.msg_control = 0;
790 #ifdef COMPAT_OLDSOCK
791 	msg.msg_flags = 0;
792 #endif
793 	aiov.iov_base = uap->buf;
794 	aiov.iov_len = uap->len;
795 	error = sendit(td, uap->s, &msg, uap->flags);
796 	return (error);
797 }
798 
799 #ifdef COMPAT_OLDSOCK
800 /*
801  * MPSAFE
802  */
803 int
804 osend(td, uap)
805 	struct thread *td;
806 	register struct osend_args /* {
807 		int	s;
808 		caddr_t	buf;
809 		int	len;
810 		int	flags;
811 	} */ *uap;
812 {
813 	struct msghdr msg;
814 	struct iovec aiov;
815 	int error;
816 
817 	msg.msg_name = 0;
818 	msg.msg_namelen = 0;
819 	msg.msg_iov = &aiov;
820 	msg.msg_iovlen = 1;
821 	aiov.iov_base = uap->buf;
822 	aiov.iov_len = uap->len;
823 	msg.msg_control = 0;
824 	msg.msg_flags = 0;
825 	error = sendit(td, uap->s, &msg, uap->flags);
826 	return (error);
827 }
828 
829 /*
830  * MPSAFE
831  */
832 int
833 osendmsg(td, uap)
834 	struct thread *td;
835 	struct osendmsg_args /* {
836 		int	s;
837 		caddr_t	msg;
838 		int	flags;
839 	} */ *uap;
840 {
841 	struct msghdr msg;
842 	struct iovec *iov;
843 	int error;
844 
845 	error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
846 	if (error)
847 		return (error);
848 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
849 	if (error)
850 		return (error);
851 	msg.msg_iov = iov;
852 	msg.msg_flags = MSG_COMPAT;
853 	error = sendit(td, uap->s, &msg, uap->flags);
854 	free(iov, M_IOV);
855 	return (error);
856 }
857 #endif
858 
859 /*
860  * MPSAFE
861  */
862 int
863 sendmsg(td, uap)
864 	struct thread *td;
865 	struct sendmsg_args /* {
866 		int	s;
867 		caddr_t	msg;
868 		int	flags;
869 	} */ *uap;
870 {
871 	struct msghdr msg;
872 	struct iovec *iov;
873 	int error;
874 
875 	error = copyin(uap->msg, &msg, sizeof (msg));
876 	if (error)
877 		return (error);
878 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
879 	if (error)
880 		return (error);
881 	msg.msg_iov = iov;
882 #ifdef COMPAT_OLDSOCK
883 	msg.msg_flags = 0;
884 #endif
885 	error = sendit(td, uap->s, &msg, uap->flags);
886 	free(iov, M_IOV);
887 	return (error);
888 }
889 
890 static int
891 recvit(td, s, mp, namelenp)
892 	struct thread *td;
893 	int s;
894 	struct msghdr *mp;
895 	void *namelenp;
896 {
897 	struct uio auio;
898 	struct iovec *iov;
899 	int i;
900 	socklen_t len;
901 	int error;
902 	struct mbuf *m, *control = 0;
903 	caddr_t ctlbuf;
904 	struct socket *so;
905 	struct sockaddr *fromsa = 0;
906 #ifdef KTRACE
907 	struct uio *ktruio = NULL;
908 #endif
909 
910 	NET_LOCK_GIANT();
911 	if ((error = fgetsock(td, s, &so, NULL)) != 0) {
912 		NET_UNLOCK_GIANT();
913 		return (error);
914 	}
915 
916 #ifdef MAC
917 	SOCK_LOCK(so);
918 	error = mac_check_socket_receive(td->td_ucred, so);
919 	SOCK_UNLOCK(so);
920 	if (error) {
921 		fputsock(so);
922 		NET_UNLOCK_GIANT();
923 		return (error);
924 	}
925 #endif
926 
927 	auio.uio_iov = mp->msg_iov;
928 	auio.uio_iovcnt = mp->msg_iovlen;
929 	auio.uio_segflg = UIO_USERSPACE;
930 	auio.uio_rw = UIO_READ;
931 	auio.uio_td = td;
932 	auio.uio_offset = 0;			/* XXX */
933 	auio.uio_resid = 0;
934 	iov = mp->msg_iov;
935 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
936 		if ((auio.uio_resid += iov->iov_len) < 0) {
937 			fputsock(so);
938 			NET_UNLOCK_GIANT();
939 			return (EINVAL);
940 		}
941 	}
942 #ifdef KTRACE
943 	if (KTRPOINT(td, KTR_GENIO))
944 		ktruio = cloneuio(&auio);
945 #endif
946 	len = auio.uio_resid;
947 	error = so->so_proto->pr_usrreqs->pru_soreceive(so, &fromsa, &auio,
948 	    (struct mbuf **)0, mp->msg_control ? &control : (struct mbuf **)0,
949 	    &mp->msg_flags);
950 	if (error) {
951 		if (auio.uio_resid != (int)len && (error == ERESTART ||
952 		    error == EINTR || error == EWOULDBLOCK))
953 			error = 0;
954 	}
955 #ifdef KTRACE
956 	if (ktruio != NULL) {
957 		ktruio->uio_resid = (int)len - auio.uio_resid;
958 		ktrgenio(s, UIO_READ, ktruio, error);
959 	}
960 #endif
961 	if (error)
962 		goto out;
963 	td->td_retval[0] = (int)len - auio.uio_resid;
964 	if (mp->msg_name) {
965 		len = mp->msg_namelen;
966 		if (len <= 0 || fromsa == 0)
967 			len = 0;
968 		else {
969 			/* save sa_len before it is destroyed by MSG_COMPAT */
970 			len = MIN(len, fromsa->sa_len);
971 #ifdef COMPAT_OLDSOCK
972 			if (mp->msg_flags & MSG_COMPAT)
973 				((struct osockaddr *)fromsa)->sa_family =
974 				    fromsa->sa_family;
975 #endif
976 			error = copyout(fromsa, mp->msg_name, (unsigned)len);
977 			if (error)
978 				goto out;
979 		}
980 		mp->msg_namelen = len;
981 		if (namelenp &&
982 		    (error = copyout(&len, namelenp, sizeof (socklen_t)))) {
983 #ifdef COMPAT_OLDSOCK
984 			if (mp->msg_flags & MSG_COMPAT)
985 				error = 0;	/* old recvfrom didn't check */
986 			else
987 #endif
988 			goto out;
989 		}
990 	}
991 	if (mp->msg_control) {
992 #ifdef COMPAT_OLDSOCK
993 		/*
994 		 * We assume that old recvmsg calls won't receive access
995 		 * rights and other control info, esp. as control info
996 		 * is always optional and those options didn't exist in 4.3.
997 		 * If we receive rights, trim the cmsghdr; anything else
998 		 * is tossed.
999 		 */
1000 		if (control && mp->msg_flags & MSG_COMPAT) {
1001 			if (mtod(control, struct cmsghdr *)->cmsg_level !=
1002 			    SOL_SOCKET ||
1003 			    mtod(control, struct cmsghdr *)->cmsg_type !=
1004 			    SCM_RIGHTS) {
1005 				mp->msg_controllen = 0;
1006 				goto out;
1007 			}
1008 			control->m_len -= sizeof (struct cmsghdr);
1009 			control->m_data += sizeof (struct cmsghdr);
1010 		}
1011 #endif
1012 		len = mp->msg_controllen;
1013 		m = control;
1014 		mp->msg_controllen = 0;
1015 		ctlbuf = mp->msg_control;
1016 
1017 		while (m && len > 0) {
1018 			unsigned int tocopy;
1019 
1020 			if (len >= m->m_len)
1021 				tocopy = m->m_len;
1022 			else {
1023 				mp->msg_flags |= MSG_CTRUNC;
1024 				tocopy = len;
1025 			}
1026 
1027 			if ((error = copyout(mtod(m, caddr_t),
1028 					ctlbuf, tocopy)) != 0)
1029 				goto out;
1030 
1031 			ctlbuf += tocopy;
1032 			len -= tocopy;
1033 			m = m->m_next;
1034 		}
1035 		mp->msg_controllen = ctlbuf - (caddr_t)mp->msg_control;
1036 	}
1037 out:
1038 	fputsock(so);
1039 	NET_UNLOCK_GIANT();
1040 	if (fromsa)
1041 		FREE(fromsa, M_SONAME);
1042 	if (control)
1043 		m_freem(control);
1044 	return (error);
1045 }
1046 
1047 /*
1048  * MPSAFE
1049  */
1050 int
1051 recvfrom(td, uap)
1052 	struct thread *td;
1053 	register struct recvfrom_args /* {
1054 		int	s;
1055 		caddr_t	buf;
1056 		size_t	len;
1057 		int	flags;
1058 		struct sockaddr * __restrict	from;
1059 		socklen_t * __restrict fromlenaddr;
1060 	} */ *uap;
1061 {
1062 	struct msghdr msg;
1063 	struct iovec aiov;
1064 	int error;
1065 
1066 	if (uap->fromlenaddr) {
1067 		error = copyin(uap->fromlenaddr,
1068 		    &msg.msg_namelen, sizeof (msg.msg_namelen));
1069 		if (error)
1070 			goto done2;
1071 	} else {
1072 		msg.msg_namelen = 0;
1073 	}
1074 	msg.msg_name = uap->from;
1075 	msg.msg_iov = &aiov;
1076 	msg.msg_iovlen = 1;
1077 	aiov.iov_base = uap->buf;
1078 	aiov.iov_len = uap->len;
1079 	msg.msg_control = 0;
1080 	msg.msg_flags = uap->flags;
1081 	error = recvit(td, uap->s, &msg, uap->fromlenaddr);
1082 done2:
1083 	return(error);
1084 }
1085 
1086 #ifdef COMPAT_OLDSOCK
1087 /*
1088  * MPSAFE
1089  */
1090 int
1091 orecvfrom(td, uap)
1092 	struct thread *td;
1093 	struct recvfrom_args *uap;
1094 {
1095 
1096 	uap->flags |= MSG_COMPAT;
1097 	return (recvfrom(td, uap));
1098 }
1099 #endif
1100 
1101 
1102 #ifdef COMPAT_OLDSOCK
1103 /*
1104  * MPSAFE
1105  */
1106 int
1107 orecv(td, uap)
1108 	struct thread *td;
1109 	register struct orecv_args /* {
1110 		int	s;
1111 		caddr_t	buf;
1112 		int	len;
1113 		int	flags;
1114 	} */ *uap;
1115 {
1116 	struct msghdr msg;
1117 	struct iovec aiov;
1118 	int error;
1119 
1120 	msg.msg_name = 0;
1121 	msg.msg_namelen = 0;
1122 	msg.msg_iov = &aiov;
1123 	msg.msg_iovlen = 1;
1124 	aiov.iov_base = uap->buf;
1125 	aiov.iov_len = uap->len;
1126 	msg.msg_control = 0;
1127 	msg.msg_flags = uap->flags;
1128 	error = recvit(td, uap->s, &msg, NULL);
1129 	return (error);
1130 }
1131 
1132 /*
1133  * Old recvmsg.  This code takes advantage of the fact that the old msghdr
1134  * overlays the new one, missing only the flags, and with the (old) access
1135  * rights where the control fields are now.
1136  *
1137  * MPSAFE
1138  */
1139 int
1140 orecvmsg(td, uap)
1141 	struct thread *td;
1142 	struct orecvmsg_args /* {
1143 		int	s;
1144 		struct	omsghdr *msg;
1145 		int	flags;
1146 	} */ *uap;
1147 {
1148 	struct msghdr msg;
1149 	struct iovec *iov;
1150 	int error;
1151 
1152 	error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
1153 	if (error)
1154 		return (error);
1155 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1156 	if (error)
1157 		return (error);
1158 	msg.msg_flags = uap->flags | MSG_COMPAT;
1159 	msg.msg_iov = iov;
1160 	error = recvit(td, uap->s, &msg, &uap->msg->msg_namelen);
1161 	if (msg.msg_controllen && error == 0)
1162 		error = copyout(&msg.msg_controllen,
1163 		    &uap->msg->msg_accrightslen, sizeof (int));
1164 	free(iov, M_IOV);
1165 	return (error);
1166 }
1167 #endif
1168 
1169 /*
1170  * MPSAFE
1171  */
1172 int
1173 recvmsg(td, uap)
1174 	struct thread *td;
1175 	struct recvmsg_args /* {
1176 		int	s;
1177 		struct	msghdr *msg;
1178 		int	flags;
1179 	} */ *uap;
1180 {
1181 	struct msghdr msg;
1182 	struct iovec *uiov, *iov;
1183 	int error;
1184 
1185 	error = copyin(uap->msg, &msg, sizeof (msg));
1186 	if (error)
1187 		return (error);
1188 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1189 	if (error)
1190 		return (error);
1191 	msg.msg_flags = uap->flags;
1192 #ifdef COMPAT_OLDSOCK
1193 	msg.msg_flags &= ~MSG_COMPAT;
1194 #endif
1195 	uiov = msg.msg_iov;
1196 	msg.msg_iov = iov;
1197 	error = recvit(td, uap->s, &msg, NULL);
1198 	if (error == 0) {
1199 		msg.msg_iov = uiov;
1200 		error = copyout(&msg, uap->msg, sizeof(msg));
1201 	}
1202 	free(iov, M_IOV);
1203 	return (error);
1204 }
1205 
1206 /*
1207  * MPSAFE
1208  */
1209 /* ARGSUSED */
1210 int
1211 shutdown(td, uap)
1212 	struct thread *td;
1213 	register struct shutdown_args /* {
1214 		int	s;
1215 		int	how;
1216 	} */ *uap;
1217 {
1218 	struct socket *so;
1219 	int error;
1220 
1221 	NET_LOCK_GIANT();
1222 	if ((error = fgetsock(td, uap->s, &so, NULL)) == 0) {
1223 		error = soshutdown(so, uap->how);
1224 		fputsock(so);
1225 	}
1226 	NET_UNLOCK_GIANT();
1227 	return(error);
1228 }
1229 
1230 /*
1231  * MPSAFE
1232  */
1233 /* ARGSUSED */
1234 int
1235 setsockopt(td, uap)
1236 	struct thread *td;
1237 	register struct setsockopt_args /* {
1238 		int	s;
1239 		int	level;
1240 		int	name;
1241 		caddr_t	val;
1242 		int	valsize;
1243 	} */ *uap;
1244 {
1245 
1246 	return (kern_setsockopt(td, uap->s, uap->level, uap->name,
1247 	    uap->val, UIO_USERSPACE, uap->valsize));
1248 }
1249 
1250 int
1251 kern_setsockopt(td, s, level, name, val, valseg, valsize)
1252 	struct thread *td;
1253 	int s;
1254 	int level;
1255 	int name;
1256 	void *val;
1257 	enum uio_seg valseg;
1258 	socklen_t valsize;
1259 {
1260 	int error;
1261 	struct socket *so;
1262 	struct sockopt sopt;
1263 
1264 	if (val == NULL && valsize != 0)
1265 		return (EFAULT);
1266 	if (valsize < 0)
1267 		return (EINVAL);
1268 
1269 	sopt.sopt_dir = SOPT_SET;
1270 	sopt.sopt_level = level;
1271 	sopt.sopt_name = name;
1272 	sopt.sopt_val = val;
1273 	sopt.sopt_valsize = valsize;
1274 	switch (valseg) {
1275 	case UIO_USERSPACE:
1276 		sopt.sopt_td = td;
1277 		break;
1278 	case UIO_SYSSPACE:
1279 		sopt.sopt_td = NULL;
1280 		break;
1281 	default:
1282 		panic("kern_setsockopt called with bad valseg");
1283 	}
1284 
1285 	NET_LOCK_GIANT();
1286 	if ((error = fgetsock(td, s, &so, NULL)) == 0) {
1287 		error = sosetopt(so, &sopt);
1288 		fputsock(so);
1289 	}
1290 	NET_UNLOCK_GIANT();
1291 	return(error);
1292 }
1293 
1294 /*
1295  * MPSAFE
1296  */
1297 /* ARGSUSED */
1298 int
1299 getsockopt(td, uap)
1300 	struct thread *td;
1301 	register struct getsockopt_args /* {
1302 		int	s;
1303 		int	level;
1304 		int	name;
1305 		void * __restrict	val;
1306 		socklen_t * __restrict avalsize;
1307 	} */ *uap;
1308 {
1309 	socklen_t valsize;
1310 	int	error;
1311 
1312 	if (uap->val) {
1313 		error = copyin(uap->avalsize, &valsize, sizeof (valsize));
1314 		if (error)
1315 			return (error);
1316 	}
1317 
1318 	error = kern_getsockopt(td, uap->s, uap->level, uap->name,
1319 	    uap->val, UIO_USERSPACE, &valsize);
1320 
1321 	if (error == 0)
1322 		error = copyout(&valsize, uap->avalsize, sizeof (valsize));
1323 	return (error);
1324 }
1325 
1326 /*
1327  * Kernel version of getsockopt.
1328  * optval can be a userland or userspace. optlen is always a kernel pointer.
1329  */
1330 int
1331 kern_getsockopt(td, s, level, name, val, valseg, valsize)
1332 	struct thread *td;
1333 	int s;
1334 	int level;
1335 	int name;
1336 	void *val;
1337 	enum uio_seg valseg;
1338 	socklen_t *valsize;
1339 {
1340 	int error;
1341 	struct  socket *so;
1342 	struct	sockopt sopt;
1343 
1344 	if (val == NULL)
1345 		*valsize = 0;
1346 	if (*valsize < 0)
1347 		return (EINVAL);
1348 
1349 	sopt.sopt_dir = SOPT_GET;
1350 	sopt.sopt_level = level;
1351 	sopt.sopt_name = name;
1352 	sopt.sopt_val = val;
1353 	sopt.sopt_valsize = (size_t)*valsize; /* checked non-negative above */
1354 	switch (valseg) {
1355 	case UIO_USERSPACE:
1356 		sopt.sopt_td = td;
1357 		break;
1358 	case UIO_SYSSPACE:
1359 		sopt.sopt_td = NULL;
1360 		break;
1361 	default:
1362 		panic("kern_getsockopt called with bad valseg");
1363 	}
1364 
1365 	NET_LOCK_GIANT();
1366 	if ((error = fgetsock(td, s, &so, NULL)) == 0) {
1367 		error = sogetopt(so, &sopt);
1368 		*valsize = sopt.sopt_valsize;
1369 		fputsock(so);
1370 	}
1371 	NET_UNLOCK_GIANT();
1372 	return (error);
1373 }
1374 
1375 /*
1376  * getsockname1() - Get socket name.
1377  *
1378  * MPSAFE
1379  */
1380 /* ARGSUSED */
1381 static int
1382 getsockname1(td, uap, compat)
1383 	struct thread *td;
1384 	register struct getsockname_args /* {
1385 		int	fdes;
1386 		struct sockaddr * __restrict asa;
1387 		socklen_t * __restrict alen;
1388 	} */ *uap;
1389 	int compat;
1390 {
1391 	struct socket *so;
1392 	struct sockaddr *sa;
1393 	socklen_t len;
1394 	int error;
1395 
1396 	NET_LOCK_GIANT();
1397 	if ((error = fgetsock(td, uap->fdes, &so, NULL)) != 0)
1398 		goto done2;
1399 	error = copyin(uap->alen, &len, sizeof (len));
1400 	if (error)
1401 		goto done1;
1402 	if (len < 0) {
1403 		error = EINVAL;
1404 		goto done1;
1405 	}
1406 	sa = 0;
1407 	error = (*so->so_proto->pr_usrreqs->pru_sockaddr)(so, &sa);
1408 	if (error)
1409 		goto bad;
1410 	if (sa == 0) {
1411 		len = 0;
1412 		goto gotnothing;
1413 	}
1414 
1415 	len = MIN(len, sa->sa_len);
1416 #ifdef COMPAT_OLDSOCK
1417 	if (compat)
1418 		((struct osockaddr *)sa)->sa_family = sa->sa_family;
1419 #endif
1420 	error = copyout(sa, uap->asa, (u_int)len);
1421 	if (error == 0)
1422 gotnothing:
1423 		error = copyout(&len, uap->alen, sizeof (len));
1424 bad:
1425 	if (sa)
1426 		FREE(sa, M_SONAME);
1427 done1:
1428 	fputsock(so);
1429 done2:
1430 	NET_UNLOCK_GIANT();
1431 	return (error);
1432 }
1433 
1434 /*
1435  * MPSAFE
1436  */
1437 int
1438 getsockname(td, uap)
1439 	struct thread *td;
1440 	struct getsockname_args *uap;
1441 {
1442 
1443 	return (getsockname1(td, uap, 0));
1444 }
1445 
1446 #ifdef COMPAT_OLDSOCK
1447 /*
1448  * MPSAFE
1449  */
1450 int
1451 ogetsockname(td, uap)
1452 	struct thread *td;
1453 	struct getsockname_args *uap;
1454 {
1455 
1456 	return (getsockname1(td, uap, 1));
1457 }
1458 #endif /* COMPAT_OLDSOCK */
1459 
1460 /*
1461  * getpeername1() - Get name of peer for connected socket.
1462  *
1463  * MPSAFE
1464  */
1465 /* ARGSUSED */
1466 static int
1467 getpeername1(td, uap, compat)
1468 	struct thread *td;
1469 	register struct getpeername_args /* {
1470 		int	fdes;
1471 		struct sockaddr * __restrict	asa;
1472 		socklen_t * __restrict	alen;
1473 	} */ *uap;
1474 	int compat;
1475 {
1476 	struct socket *so;
1477 	struct sockaddr *sa;
1478 	socklen_t len;
1479 	int error;
1480 
1481 	NET_LOCK_GIANT();
1482 	if ((error = fgetsock(td, uap->fdes, &so, NULL)) != 0)
1483 		goto done2;
1484 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1485 		error = ENOTCONN;
1486 		goto done1;
1487 	}
1488 	error = copyin(uap->alen, &len, sizeof (len));
1489 	if (error)
1490 		goto done1;
1491 	if (len < 0) {
1492 		error = EINVAL;
1493 		goto done1;
1494 	}
1495 	sa = 0;
1496 	error = (*so->so_proto->pr_usrreqs->pru_peeraddr)(so, &sa);
1497 	if (error)
1498 		goto bad;
1499 	if (sa == 0) {
1500 		len = 0;
1501 		goto gotnothing;
1502 	}
1503 	len = MIN(len, sa->sa_len);
1504 #ifdef COMPAT_OLDSOCK
1505 	if (compat)
1506 		((struct osockaddr *)sa)->sa_family =
1507 		    sa->sa_family;
1508 #endif
1509 	error = copyout(sa, uap->asa, (u_int)len);
1510 	if (error)
1511 		goto bad;
1512 gotnothing:
1513 	error = copyout(&len, uap->alen, sizeof (len));
1514 bad:
1515 	if (sa)
1516 		FREE(sa, M_SONAME);
1517 done1:
1518 	fputsock(so);
1519 done2:
1520 	NET_UNLOCK_GIANT();
1521 	return (error);
1522 }
1523 
1524 /*
1525  * MPSAFE
1526  */
1527 int
1528 getpeername(td, uap)
1529 	struct thread *td;
1530 	struct getpeername_args *uap;
1531 {
1532 
1533 	return (getpeername1(td, uap, 0));
1534 }
1535 
1536 #ifdef COMPAT_OLDSOCK
1537 /*
1538  * MPSAFE
1539  */
1540 int
1541 ogetpeername(td, uap)
1542 	struct thread *td;
1543 	struct ogetpeername_args *uap;
1544 {
1545 
1546 	/* XXX uap should have type `getpeername_args *' to begin with. */
1547 	return (getpeername1(td, (struct getpeername_args *)uap, 1));
1548 }
1549 #endif /* COMPAT_OLDSOCK */
1550 
1551 int
1552 sockargs(mp, buf, buflen, type)
1553 	struct mbuf **mp;
1554 	caddr_t buf;
1555 	int buflen, type;
1556 {
1557 	register struct sockaddr *sa;
1558 	register struct mbuf *m;
1559 	int error;
1560 
1561 	if ((u_int)buflen > MLEN) {
1562 #ifdef COMPAT_OLDSOCK
1563 		if (type == MT_SONAME && (u_int)buflen <= 112)
1564 			buflen = MLEN;		/* unix domain compat. hack */
1565 		else
1566 #endif
1567 			if ((u_int)buflen > MCLBYTES)
1568 				return (EINVAL);
1569 	}
1570 	m = m_get(M_TRYWAIT, type);
1571 	if (m == NULL)
1572 		return (ENOBUFS);
1573 	if ((u_int)buflen > MLEN) {
1574 		MCLGET(m, M_TRYWAIT);
1575 		if ((m->m_flags & M_EXT) == 0) {
1576 			m_free(m);
1577 			return (ENOBUFS);
1578 		}
1579 	}
1580 	m->m_len = buflen;
1581 	error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
1582 	if (error)
1583 		(void) m_free(m);
1584 	else {
1585 		*mp = m;
1586 		if (type == MT_SONAME) {
1587 			sa = mtod(m, struct sockaddr *);
1588 
1589 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
1590 			if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1591 				sa->sa_family = sa->sa_len;
1592 #endif
1593 			sa->sa_len = buflen;
1594 		}
1595 	}
1596 	return (error);
1597 }
1598 
1599 int
1600 getsockaddr(namp, uaddr, len)
1601 	struct sockaddr **namp;
1602 	caddr_t uaddr;
1603 	size_t len;
1604 {
1605 	struct sockaddr *sa;
1606 	int error;
1607 
1608 	if (len > SOCK_MAXADDRLEN)
1609 		return (ENAMETOOLONG);
1610 	if (len < offsetof(struct sockaddr, sa_data[0]))
1611 		return (EINVAL);
1612 	MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1613 	error = copyin(uaddr, sa, len);
1614 	if (error) {
1615 		FREE(sa, M_SONAME);
1616 	} else {
1617 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
1618 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1619 			sa->sa_family = sa->sa_len;
1620 #endif
1621 		sa->sa_len = len;
1622 		*namp = sa;
1623 	}
1624 	return (error);
1625 }
1626 
1627 /*
1628  * Detach mapped page and release resources back to the system.
1629  */
1630 void
1631 sf_buf_mext(void *addr, void *args)
1632 {
1633 	vm_page_t m;
1634 
1635 	m = sf_buf_page(args);
1636 	sf_buf_free(args);
1637 	vm_page_lock_queues();
1638 	vm_page_unwire(m, 0);
1639 	/*
1640 	 * Check for the object going away on us. This can
1641 	 * happen since we don't hold a reference to it.
1642 	 * If so, we're responsible for freeing the page.
1643 	 */
1644 	if (m->wire_count == 0 && m->object == NULL)
1645 		vm_page_free(m);
1646 	vm_page_unlock_queues();
1647 }
1648 
1649 /*
1650  * sendfile(2)
1651  *
1652  * MPSAFE
1653  *
1654  * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1655  *	 struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1656  *
1657  * Send a file specified by 'fd' and starting at 'offset' to a socket
1658  * specified by 's'. Send only 'nbytes' of the file or until EOF if
1659  * nbytes == 0. Optionally add a header and/or trailer to the socket
1660  * output. If specified, write the total number of bytes sent into *sbytes.
1661  *
1662  */
1663 int
1664 sendfile(struct thread *td, struct sendfile_args *uap)
1665 {
1666 
1667 	return (do_sendfile(td, uap, 0));
1668 }
1669 
1670 #ifdef COMPAT_FREEBSD4
1671 int
1672 freebsd4_sendfile(struct thread *td, struct freebsd4_sendfile_args *uap)
1673 {
1674 	struct sendfile_args args;
1675 
1676 	args.fd = uap->fd;
1677 	args.s = uap->s;
1678 	args.offset = uap->offset;
1679 	args.nbytes = uap->nbytes;
1680 	args.hdtr = uap->hdtr;
1681 	args.sbytes = uap->sbytes;
1682 	args.flags = uap->flags;
1683 
1684 	return (do_sendfile(td, &args, 1));
1685 }
1686 #endif /* COMPAT_FREEBSD4 */
1687 
1688 static int
1689 do_sendfile(struct thread *td, struct sendfile_args *uap, int compat)
1690 {
1691 	struct vnode *vp;
1692 	struct vm_object *obj;
1693 	struct socket *so = NULL;
1694 	struct mbuf *m, *m_header = NULL;
1695 	struct sf_buf *sf;
1696 	struct vm_page *pg;
1697 	struct writev_args nuap;
1698 	struct sf_hdtr hdtr;
1699 	struct uio *hdr_uio = NULL;
1700 	off_t off, xfsize, hdtr_size, sbytes = 0;
1701 	int error, headersize = 0, headersent = 0;
1702 
1703 	mtx_lock(&Giant);
1704 
1705 	hdtr_size = 0;
1706 
1707 	/*
1708 	 * The descriptor must be a regular file and have a backing VM object.
1709 	 */
1710 	if ((error = fgetvp_read(td, uap->fd, &vp)) != 0)
1711 		goto done;
1712 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1713 	if (vp->v_type != VREG || VOP_GETVOBJECT(vp, &obj) != 0) {
1714 		error = EINVAL;
1715 		VOP_UNLOCK(vp, 0, td);
1716 		goto done;
1717 	}
1718 	VOP_UNLOCK(vp, 0, td);
1719 	if ((error = fgetsock(td, uap->s, &so, NULL)) != 0)
1720 		goto done;
1721 	if (so->so_type != SOCK_STREAM) {
1722 		error = EINVAL;
1723 		goto done;
1724 	}
1725 	if ((so->so_state & SS_ISCONNECTED) == 0) {
1726 		error = ENOTCONN;
1727 		goto done;
1728 	}
1729 	if (uap->offset < 0) {
1730 		error = EINVAL;
1731 		goto done;
1732 	}
1733 
1734 #ifdef MAC
1735 	SOCK_LOCK(so);
1736 	error = mac_check_socket_send(td->td_ucred, so);
1737 	SOCK_UNLOCK(so);
1738 	if (error)
1739 		goto done;
1740 #endif
1741 
1742 	/*
1743 	 * If specified, get the pointer to the sf_hdtr struct for
1744 	 * any headers/trailers.
1745 	 */
1746 	if (uap->hdtr != NULL) {
1747 		error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1748 		if (error)
1749 			goto done;
1750 		/*
1751 		 * Send any headers.
1752 		 */
1753 		if (hdtr.headers != NULL) {
1754 			error = copyinuio(hdtr.headers, hdtr.hdr_cnt, &hdr_uio);
1755 			if (error)
1756 				goto done;
1757 			hdr_uio->uio_td = td;
1758 			hdr_uio->uio_rw = UIO_WRITE;
1759 			if (hdr_uio->uio_resid > 0) {
1760 				m_header = m_uiotombuf(hdr_uio, M_DONTWAIT, 0);
1761 				if (m_header == NULL)
1762 					goto done;
1763 				headersize = m_header->m_pkthdr.len;
1764 				if (compat)
1765 					sbytes += headersize;
1766 			}
1767 		}
1768 	}
1769 
1770 	/*
1771 	 * Protect against multiple writers to the socket.
1772 	 */
1773 	SOCKBUF_LOCK(&so->so_snd);
1774 	(void) sblock(&so->so_snd, M_WAITOK);
1775 	SOCKBUF_UNLOCK(&so->so_snd);
1776 
1777 	/*
1778 	 * Loop through the pages in the file, starting with the requested
1779 	 * offset. Get a file page (do I/O if necessary), map the file page
1780 	 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1781 	 * it on the socket.
1782 	 */
1783 	for (off = uap->offset; ; off += xfsize, sbytes += xfsize) {
1784 		vm_pindex_t pindex;
1785 		vm_offset_t pgoff;
1786 
1787 		pindex = OFF_TO_IDX(off);
1788 		VM_OBJECT_LOCK(obj);
1789 retry_lookup:
1790 		/*
1791 		 * Calculate the amount to transfer. Not to exceed a page,
1792 		 * the EOF, or the passed in nbytes.
1793 		 */
1794 		xfsize = obj->un_pager.vnp.vnp_size - off;
1795 		VM_OBJECT_UNLOCK(obj);
1796 		if (xfsize > PAGE_SIZE)
1797 			xfsize = PAGE_SIZE;
1798 		pgoff = (vm_offset_t)(off & PAGE_MASK);
1799 		if (PAGE_SIZE - pgoff < xfsize)
1800 			xfsize = PAGE_SIZE - pgoff;
1801 		if (uap->nbytes && xfsize > (uap->nbytes - sbytes))
1802 			xfsize = uap->nbytes - sbytes;
1803 		if (xfsize <= 0) {
1804 			if (m_header != NULL) {
1805 				m = m_header;
1806 				m_header = NULL;
1807 				goto retry_space;
1808 			} else
1809 				break;
1810 		}
1811 		/*
1812 		 * Optimize the non-blocking case by looking at the socket space
1813 		 * before going to the extra work of constituting the sf_buf.
1814 		 */
1815 		SOCKBUF_LOCK(&so->so_snd);
1816 		if ((so->so_state & SS_NBIO) && sbspace(&so->so_snd) <= 0) {
1817 			if (so->so_snd.sb_state & SBS_CANTSENDMORE)
1818 				error = EPIPE;
1819 			else
1820 				error = EAGAIN;
1821 			sbunlock(&so->so_snd);
1822 			SOCKBUF_UNLOCK(&so->so_snd);
1823 			goto done;
1824 		}
1825 		SOCKBUF_UNLOCK(&so->so_snd);
1826 		VM_OBJECT_LOCK(obj);
1827 		/*
1828 		 * Attempt to look up the page.
1829 		 *
1830 		 *	Allocate if not found
1831 		 *
1832 		 *	Wait and loop if busy.
1833 		 */
1834 		pg = vm_page_lookup(obj, pindex);
1835 
1836 		if (pg == NULL) {
1837 			pg = vm_page_alloc(obj, pindex,
1838 			    VM_ALLOC_NORMAL | VM_ALLOC_WIRED);
1839 			if (pg == NULL) {
1840 				VM_OBJECT_UNLOCK(obj);
1841 				VM_WAIT;
1842 				VM_OBJECT_LOCK(obj);
1843 				goto retry_lookup;
1844 			}
1845 			vm_page_lock_queues();
1846 			vm_page_wakeup(pg);
1847 		} else {
1848 			vm_page_lock_queues();
1849 			if (vm_page_sleep_if_busy(pg, TRUE, "sfpbsy"))
1850 				goto retry_lookup;
1851 			/*
1852 			 * Wire the page so it does not get ripped out from
1853 			 * under us.
1854 			 */
1855 			vm_page_wire(pg);
1856 		}
1857 
1858 		/*
1859 		 * If page is not valid for what we need, initiate I/O
1860 		 */
1861 
1862 		if (pg->valid && vm_page_is_valid(pg, pgoff, xfsize)) {
1863 			VM_OBJECT_UNLOCK(obj);
1864 		} else if (uap->flags & SF_NODISKIO) {
1865 			error = EBUSY;
1866 		} else {
1867 			int bsize, resid;
1868 
1869 			/*
1870 			 * Ensure that our page is still around when the I/O
1871 			 * completes.
1872 			 */
1873 			vm_page_io_start(pg);
1874 			vm_page_unlock_queues();
1875 			VM_OBJECT_UNLOCK(obj);
1876 
1877 			/*
1878 			 * Get the page from backing store.
1879 			 */
1880 			bsize = vp->v_mount->mnt_stat.f_iosize;
1881 			vn_lock(vp, LK_SHARED | LK_NOPAUSE | LK_RETRY, td);
1882 			/*
1883 			 * XXXMAC: Because we don't have fp->f_cred here,
1884 			 * we pass in NOCRED.  This is probably wrong, but
1885 			 * is consistent with our original implementation.
1886 			 */
1887 			error = vn_rdwr(UIO_READ, vp, NULL, MAXBSIZE,
1888 			    trunc_page(off), UIO_NOCOPY, IO_NODELOCKED |
1889 			    IO_VMIO | ((MAXBSIZE / bsize) << IO_SEQSHIFT),
1890 			    td->td_ucred, NOCRED, &resid, td);
1891 			VOP_UNLOCK(vp, 0, td);
1892 			if (error)
1893 				VM_OBJECT_LOCK(obj);
1894 			vm_page_lock_queues();
1895 			vm_page_io_finish(pg);
1896 			mbstat.sf_iocnt++;
1897 		}
1898 
1899 		if (error) {
1900 			vm_page_unwire(pg, 0);
1901 			/*
1902 			 * See if anyone else might know about this page.
1903 			 * If not and it is not valid, then free it.
1904 			 */
1905 			if (pg->wire_count == 0 && pg->valid == 0 &&
1906 			    pg->busy == 0 && !(pg->flags & PG_BUSY) &&
1907 			    pg->hold_count == 0) {
1908 				vm_page_busy(pg);
1909 				vm_page_free(pg);
1910 			}
1911 			vm_page_unlock_queues();
1912 			VM_OBJECT_UNLOCK(obj);
1913 			SOCKBUF_LOCK(&so->so_snd);
1914 			sbunlock(&so->so_snd);
1915 			SOCKBUF_UNLOCK(&so->so_snd);
1916 			goto done;
1917 		}
1918 		vm_page_unlock_queues();
1919 
1920 		/*
1921 		 * Get a sendfile buf. We usually wait as long as necessary,
1922 		 * but this wait can be interrupted.
1923 		 */
1924 		if ((sf = sf_buf_alloc(pg, PCATCH)) == NULL) {
1925 			mbstat.sf_allocfail++;
1926 			vm_page_lock_queues();
1927 			vm_page_unwire(pg, 0);
1928 			if (pg->wire_count == 0 && pg->object == NULL)
1929 				vm_page_free(pg);
1930 			vm_page_unlock_queues();
1931 			SOCKBUF_LOCK(&so->so_snd);
1932 			sbunlock(&so->so_snd);
1933 			SOCKBUF_UNLOCK(&so->so_snd);
1934 			error = EINTR;
1935 			goto done;
1936 		}
1937 
1938 		/*
1939 		 * Get an mbuf header and set it up as having external storage.
1940 		 */
1941 		if (m_header)
1942 			MGET(m, M_TRYWAIT, MT_DATA);
1943 		else
1944 			MGETHDR(m, M_TRYWAIT, MT_DATA);
1945 		if (m == NULL) {
1946 			error = ENOBUFS;
1947 			sf_buf_mext((void *)sf_buf_kva(sf), sf);
1948 			SOCKBUF_LOCK(&so->so_snd);
1949 			sbunlock(&so->so_snd);
1950 			SOCKBUF_UNLOCK(&so->so_snd);
1951 			goto done;
1952 		}
1953 		/*
1954 		 * Setup external storage for mbuf.
1955 		 */
1956 		MEXTADD(m, sf_buf_kva(sf), PAGE_SIZE, sf_buf_mext, sf, M_RDONLY,
1957 		    EXT_SFBUF);
1958 		m->m_data = (char *)sf_buf_kva(sf) + pgoff;
1959 		m->m_pkthdr.len = m->m_len = xfsize;
1960 
1961 		if (m_header) {
1962 			m_cat(m_header, m);
1963 			m = m_header;
1964 			m_header = NULL;
1965 			m_fixhdr(m);
1966 		}
1967 
1968 		/*
1969 		 * Add the buffer to the socket buffer chain.
1970 		 */
1971 		SOCKBUF_LOCK(&so->so_snd);
1972 retry_space:
1973 		/*
1974 		 * Make sure that the socket is still able to take more data.
1975 		 * CANTSENDMORE being true usually means that the connection
1976 		 * was closed. so_error is true when an error was sensed after
1977 		 * a previous send.
1978 		 * The state is checked after the page mapping and buffer
1979 		 * allocation above since those operations may block and make
1980 		 * any socket checks stale. From this point forward, nothing
1981 		 * blocks before the pru_send (or more accurately, any blocking
1982 		 * results in a loop back to here to re-check).
1983 		 */
1984 		SOCKBUF_LOCK_ASSERT(&so->so_snd);
1985 		if ((so->so_snd.sb_state & SBS_CANTSENDMORE) || so->so_error) {
1986 			if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
1987 				error = EPIPE;
1988 			} else {
1989 				error = so->so_error;
1990 				so->so_error = 0;
1991 			}
1992 			m_freem(m);
1993 			sbunlock(&so->so_snd);
1994 			SOCKBUF_UNLOCK(&so->so_snd);
1995 			goto done;
1996 		}
1997 		/*
1998 		 * Wait for socket space to become available. We do this just
1999 		 * after checking the connection state above in order to avoid
2000 		 * a race condition with sbwait().
2001 		 */
2002 		if (sbspace(&so->so_snd) < so->so_snd.sb_lowat) {
2003 			if (so->so_state & SS_NBIO) {
2004 				m_freem(m);
2005 				sbunlock(&so->so_snd);
2006 				SOCKBUF_UNLOCK(&so->so_snd);
2007 				error = EAGAIN;
2008 				goto done;
2009 			}
2010 			error = sbwait(&so->so_snd);
2011 			/*
2012 			 * An error from sbwait usually indicates that we've
2013 			 * been interrupted by a signal. If we've sent anything
2014 			 * then return bytes sent, otherwise return the error.
2015 			 */
2016 			if (error) {
2017 				m_freem(m);
2018 				sbunlock(&so->so_snd);
2019 				SOCKBUF_UNLOCK(&so->so_snd);
2020 				goto done;
2021 			}
2022 			goto retry_space;
2023 		}
2024 		SOCKBUF_UNLOCK(&so->so_snd);
2025 		error = (*so->so_proto->pr_usrreqs->pru_send)(so, 0, m, 0, 0, td);
2026 		if (error) {
2027 			SOCKBUF_LOCK(&so->so_snd);
2028 			sbunlock(&so->so_snd);
2029 			SOCKBUF_UNLOCK(&so->so_snd);
2030 			goto done;
2031 		}
2032 		headersent = 1;
2033 	}
2034 	SOCKBUF_LOCK(&so->so_snd);
2035 	sbunlock(&so->so_snd);
2036 	SOCKBUF_UNLOCK(&so->so_snd);
2037 
2038 	/*
2039 	 * Send trailers. Wimp out and use writev(2).
2040 	 */
2041 	if (uap->hdtr != NULL && hdtr.trailers != NULL) {
2042 			nuap.fd = uap->s;
2043 			nuap.iovp = hdtr.trailers;
2044 			nuap.iovcnt = hdtr.trl_cnt;
2045 			error = writev(td, &nuap);
2046 			if (error)
2047 				goto done;
2048 			if (compat)
2049 				sbytes += td->td_retval[0];
2050 			else
2051 				hdtr_size += td->td_retval[0];
2052 	}
2053 
2054 done:
2055 	if (headersent) {
2056 		if (!compat)
2057 			hdtr_size += headersize;
2058 	} else {
2059 		if (compat)
2060 			sbytes -= headersize;
2061 	}
2062 	/*
2063 	 * If there was no error we have to clear td->td_retval[0]
2064 	 * because it may have been set by writev.
2065 	 */
2066 	if (error == 0) {
2067 		td->td_retval[0] = 0;
2068 	}
2069 	if (uap->sbytes != NULL) {
2070 		if (!compat)
2071 			sbytes += hdtr_size;
2072 		copyout(&sbytes, uap->sbytes, sizeof(off_t));
2073 	}
2074 	if (vp)
2075 		vrele(vp);
2076 	if (so)
2077 		fputsock(so);
2078 	if (hdr_uio != NULL)
2079 		free(hdr_uio, M_IOV);
2080 	if (m_header)
2081 		m_freem(m_header);
2082 
2083 	mtx_unlock(&Giant);
2084 
2085 	if (error == ERESTART)
2086 		error = EINTR;
2087 
2088 	return (error);
2089 }
2090