xref: /freebsd/sys/compat/linux/linux_socket.c (revision 884a2a699669ec61e2366e3e358342dbc94be24a)
1 /*-
2  * Copyright (c) 1995 S�ren Schmidt
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer
10  *    in this position and unchanged.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. The name of the author may not be used to endorse or promote products
15  *    derived from this software without specific prior written permission
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 /* XXX we use functions that might not exist. */
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 
36 #include <sys/param.h>
37 #include <sys/proc.h>
38 #include <sys/systm.h>
39 #include <sys/sysproto.h>
40 #include <sys/fcntl.h>
41 #include <sys/file.h>
42 #include <sys/limits.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mutex.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/syscallsubr.h>
50 #include <sys/uio.h>
51 #include <sys/syslog.h>
52 #include <sys/un.h>
53 
54 #include <net/if.h>
55 #include <netinet/in.h>
56 #include <netinet/in_systm.h>
57 #include <netinet/ip.h>
58 #ifdef INET6
59 #include <netinet/ip6.h>
60 #include <netinet6/ip6_var.h>
61 #include <netinet6/in6_var.h>
62 #endif
63 
64 #ifdef COMPAT_LINUX32
65 #include <machine/../linux32/linux.h>
66 #include <machine/../linux32/linux32_proto.h>
67 #else
68 #include <machine/../linux/linux.h>
69 #include <machine/../linux/linux_proto.h>
70 #endif
71 #include <compat/linux/linux_socket.h>
72 #include <compat/linux/linux_util.h>
73 
74 static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
75     struct malloc_type *);
76 static int linux_to_bsd_domain(int);
77 
78 /*
79  * Reads a linux sockaddr and does any necessary translation.
80  * Linux sockaddrs don't have a length field, only a family.
81  */
82 static int
83 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
84 {
85 	int osalen = len;
86 
87 	return (do_sa_get(sap, osa, &osalen, M_SONAME));
88 }
89 
90 /*
91  * Copy the osockaddr structure pointed to by osa to kernel, adjust
92  * family and convert to sockaddr.
93  */
94 static int
95 do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
96     struct malloc_type *mtype)
97 {
98 	int error=0, bdom;
99 	struct sockaddr *sa;
100 	struct osockaddr *kosa;
101 	int alloclen;
102 #ifdef INET6
103 	int oldv6size;
104 	struct sockaddr_in6 *sin6;
105 #endif
106 
107 	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
108 		return (EINVAL);
109 
110 	alloclen = *osalen;
111 #ifdef INET6
112 	oldv6size = 0;
113 	/*
114 	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
115 	 * if it's a v4-mapped address, so reserve the proper space
116 	 * for it.
117 	 */
118 	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
119 		alloclen = sizeof (struct sockaddr_in6);
120 		oldv6size = 1;
121 	}
122 #endif
123 
124 	kosa = malloc(alloclen, mtype, M_WAITOK);
125 
126 	if ((error = copyin(osa, kosa, *osalen)))
127 		goto out;
128 
129 	bdom = linux_to_bsd_domain(kosa->sa_family);
130 	if (bdom == -1) {
131 		error = EAFNOSUPPORT;
132 		goto out;
133 	}
134 
135 #ifdef INET6
136 	/*
137 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
138 	 * which lacks the scope id compared with RFC2553 one. If we detect
139 	 * the situation, reject the address and write a message to system log.
140 	 *
141 	 * Still accept addresses for which the scope id is not used.
142 	 */
143 	if (oldv6size && bdom == AF_INET6) {
144 		sin6 = (struct sockaddr_in6 *)kosa;
145 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
146 		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
147 		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
148 		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
149 		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
150 		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
151 			sin6->sin6_scope_id = 0;
152 		} else {
153 			log(LOG_DEBUG,
154 			    "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
155 			error = EINVAL;
156 			goto out;
157 		}
158 	} else
159 #endif
160 	if (bdom == AF_INET) {
161 		alloclen = sizeof(struct sockaddr_in);
162 		if (*osalen < alloclen) {
163 			error = EINVAL;
164 			goto out;
165 		}
166 	}
167 
168 	sa = (struct sockaddr *) kosa;
169 	sa->sa_family = bdom;
170 	sa->sa_len = alloclen;
171 
172 	*sap = sa;
173 	*osalen = alloclen;
174 	return (0);
175 
176 out:
177 	free(kosa, mtype);
178 	return (error);
179 }
180 
181 static int
182 linux_to_bsd_domain(int domain)
183 {
184 
185 	switch (domain) {
186 	case LINUX_AF_UNSPEC:
187 		return (AF_UNSPEC);
188 	case LINUX_AF_UNIX:
189 		return (AF_LOCAL);
190 	case LINUX_AF_INET:
191 		return (AF_INET);
192 	case LINUX_AF_INET6:
193 		return (AF_INET6);
194 	case LINUX_AF_AX25:
195 		return (AF_CCITT);
196 	case LINUX_AF_IPX:
197 		return (AF_IPX);
198 	case LINUX_AF_APPLETALK:
199 		return (AF_APPLETALK);
200 	}
201 	return (-1);
202 }
203 
204 static int
205 bsd_to_linux_domain(int domain)
206 {
207 
208 	switch (domain) {
209 	case AF_UNSPEC:
210 		return (LINUX_AF_UNSPEC);
211 	case AF_LOCAL:
212 		return (LINUX_AF_UNIX);
213 	case AF_INET:
214 		return (LINUX_AF_INET);
215 	case AF_INET6:
216 		return (LINUX_AF_INET6);
217 	case AF_CCITT:
218 		return (LINUX_AF_AX25);
219 	case AF_IPX:
220 		return (LINUX_AF_IPX);
221 	case AF_APPLETALK:
222 		return (LINUX_AF_APPLETALK);
223 	}
224 	return (-1);
225 }
226 
227 static int
228 linux_to_bsd_sockopt_level(int level)
229 {
230 
231 	switch (level) {
232 	case LINUX_SOL_SOCKET:
233 		return (SOL_SOCKET);
234 	}
235 	return (level);
236 }
237 
238 static int
239 bsd_to_linux_sockopt_level(int level)
240 {
241 
242 	switch (level) {
243 	case SOL_SOCKET:
244 		return (LINUX_SOL_SOCKET);
245 	}
246 	return (level);
247 }
248 
249 static int
250 linux_to_bsd_ip_sockopt(int opt)
251 {
252 
253 	switch (opt) {
254 	case LINUX_IP_TOS:
255 		return (IP_TOS);
256 	case LINUX_IP_TTL:
257 		return (IP_TTL);
258 	case LINUX_IP_OPTIONS:
259 		return (IP_OPTIONS);
260 	case LINUX_IP_MULTICAST_IF:
261 		return (IP_MULTICAST_IF);
262 	case LINUX_IP_MULTICAST_TTL:
263 		return (IP_MULTICAST_TTL);
264 	case LINUX_IP_MULTICAST_LOOP:
265 		return (IP_MULTICAST_LOOP);
266 	case LINUX_IP_ADD_MEMBERSHIP:
267 		return (IP_ADD_MEMBERSHIP);
268 	case LINUX_IP_DROP_MEMBERSHIP:
269 		return (IP_DROP_MEMBERSHIP);
270 	case LINUX_IP_HDRINCL:
271 		return (IP_HDRINCL);
272 	}
273 	return (-1);
274 }
275 
276 static int
277 linux_to_bsd_so_sockopt(int opt)
278 {
279 
280 	switch (opt) {
281 	case LINUX_SO_DEBUG:
282 		return (SO_DEBUG);
283 	case LINUX_SO_REUSEADDR:
284 		return (SO_REUSEADDR);
285 	case LINUX_SO_TYPE:
286 		return (SO_TYPE);
287 	case LINUX_SO_ERROR:
288 		return (SO_ERROR);
289 	case LINUX_SO_DONTROUTE:
290 		return (SO_DONTROUTE);
291 	case LINUX_SO_BROADCAST:
292 		return (SO_BROADCAST);
293 	case LINUX_SO_SNDBUF:
294 		return (SO_SNDBUF);
295 	case LINUX_SO_RCVBUF:
296 		return (SO_RCVBUF);
297 	case LINUX_SO_KEEPALIVE:
298 		return (SO_KEEPALIVE);
299 	case LINUX_SO_OOBINLINE:
300 		return (SO_OOBINLINE);
301 	case LINUX_SO_LINGER:
302 		return (SO_LINGER);
303 	case LINUX_SO_PEERCRED:
304 		return (LOCAL_PEERCRED);
305 	case LINUX_SO_RCVLOWAT:
306 		return (SO_RCVLOWAT);
307 	case LINUX_SO_SNDLOWAT:
308 		return (SO_SNDLOWAT);
309 	case LINUX_SO_RCVTIMEO:
310 		return (SO_RCVTIMEO);
311 	case LINUX_SO_SNDTIMEO:
312 		return (SO_SNDTIMEO);
313 	case LINUX_SO_TIMESTAMP:
314 		return (SO_TIMESTAMP);
315 	case LINUX_SO_ACCEPTCONN:
316 		return (SO_ACCEPTCONN);
317 	}
318 	return (-1);
319 }
320 
321 static int
322 linux_to_bsd_msg_flags(int flags)
323 {
324 	int ret_flags = 0;
325 
326 	if (flags & LINUX_MSG_OOB)
327 		ret_flags |= MSG_OOB;
328 	if (flags & LINUX_MSG_PEEK)
329 		ret_flags |= MSG_PEEK;
330 	if (flags & LINUX_MSG_DONTROUTE)
331 		ret_flags |= MSG_DONTROUTE;
332 	if (flags & LINUX_MSG_CTRUNC)
333 		ret_flags |= MSG_CTRUNC;
334 	if (flags & LINUX_MSG_TRUNC)
335 		ret_flags |= MSG_TRUNC;
336 	if (flags & LINUX_MSG_DONTWAIT)
337 		ret_flags |= MSG_DONTWAIT;
338 	if (flags & LINUX_MSG_EOR)
339 		ret_flags |= MSG_EOR;
340 	if (flags & LINUX_MSG_WAITALL)
341 		ret_flags |= MSG_WAITALL;
342 	if (flags & LINUX_MSG_NOSIGNAL)
343 		ret_flags |= MSG_NOSIGNAL;
344 #if 0 /* not handled */
345 	if (flags & LINUX_MSG_PROXY)
346 		;
347 	if (flags & LINUX_MSG_FIN)
348 		;
349 	if (flags & LINUX_MSG_SYN)
350 		;
351 	if (flags & LINUX_MSG_CONFIRM)
352 		;
353 	if (flags & LINUX_MSG_RST)
354 		;
355 	if (flags & LINUX_MSG_ERRQUEUE)
356 		;
357 #endif
358 	return ret_flags;
359 }
360 
361 /*
362 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
363 * native syscall will fault.  Thus, we don't really need to check the
364 * return values for these functions.
365 */
366 
367 static int
368 bsd_to_linux_sockaddr(struct sockaddr *arg)
369 {
370 	struct sockaddr sa;
371 	size_t sa_len = sizeof(struct sockaddr);
372 	int error;
373 
374 	if ((error = copyin(arg, &sa, sa_len)))
375 		return (error);
376 
377 	*(u_short *)&sa = sa.sa_family;
378 
379 	error = copyout(&sa, arg, sa_len);
380 
381 	return (error);
382 }
383 
384 static int
385 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
386 {
387 	struct sockaddr sa;
388 	size_t sa_len = sizeof(struct sockaddr);
389 	int error;
390 
391 	if ((error = copyin(arg, &sa, sa_len)))
392 		return (error);
393 
394 	sa.sa_family = *(sa_family_t *)&sa;
395 	sa.sa_len = len;
396 
397 	error = copyout(&sa, arg, sa_len);
398 
399 	return (error);
400 }
401 
402 
403 static int
404 linux_sa_put(struct osockaddr *osa)
405 {
406 	struct osockaddr sa;
407 	int error, bdom;
408 
409 	/*
410 	 * Only read/write the osockaddr family part, the rest is
411 	 * not changed.
412 	 */
413 	error = copyin(osa, &sa, sizeof(sa.sa_family));
414 	if (error)
415 		return (error);
416 
417 	bdom = bsd_to_linux_domain(sa.sa_family);
418 	if (bdom == -1)
419 		return (EINVAL);
420 
421 	sa.sa_family = bdom;
422 	error = copyout(&sa, osa, sizeof(sa.sa_family));
423 	if (error)
424 		return (error);
425 
426 	return (0);
427 }
428 
429 static int
430 linux_to_bsd_cmsg_type(int cmsg_type)
431 {
432 
433 	switch (cmsg_type) {
434 	case LINUX_SCM_RIGHTS:
435 		return (SCM_RIGHTS);
436 	case LINUX_SCM_CREDENTIALS:
437 		return (SCM_CREDS);
438 	}
439 	return (-1);
440 }
441 
442 static int
443 bsd_to_linux_cmsg_type(int cmsg_type)
444 {
445 
446 	switch (cmsg_type) {
447 	case SCM_RIGHTS:
448 		return (LINUX_SCM_RIGHTS);
449 	case SCM_CREDS:
450 		return (LINUX_SCM_CREDENTIALS);
451 	}
452 	return (-1);
453 }
454 
455 static int
456 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
457 {
458 	if (lhdr->msg_controllen > INT_MAX)
459 		return (ENOBUFS);
460 
461 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
462 	bhdr->msg_namelen	= lhdr->msg_namelen;
463 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
464 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
465 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
466 
467 	/*
468 	 * msg_controllen is skipped since BSD and LINUX control messages
469 	 * are potentially different sizes (e.g. the cred structure used
470 	 * by SCM_CREDS is different between the two operating system).
471 	 *
472 	 * The caller can set it (if necessary) after converting all the
473 	 * control messages.
474 	 */
475 
476 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
477 	return (0);
478 }
479 
480 static int
481 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
482 {
483 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
484 	lhdr->msg_namelen	= bhdr->msg_namelen;
485 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
486 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
487 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
488 
489 	/*
490 	 * msg_controllen is skipped since BSD and LINUX control messages
491 	 * are potentially different sizes (e.g. the cred structure used
492 	 * by SCM_CREDS is different between the two operating system).
493 	 *
494 	 * The caller can set it (if necessary) after converting all the
495 	 * control messages.
496 	 */
497 
498 	/* msg_flags skipped */
499 	return (0);
500 }
501 
502 static int
503 linux_set_socket_flags(struct thread *td, int s, int flags)
504 {
505 	int error;
506 
507 	if (flags & LINUX_SOCK_NONBLOCK) {
508 		error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
509 		if (error)
510 			return (error);
511 	}
512 	if (flags & LINUX_SOCK_CLOEXEC) {
513 		error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
514 		if (error)
515 			return (error);
516 	}
517 	return (0);
518 }
519 
520 static int
521 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
522     struct mbuf *control, enum uio_seg segflg)
523 {
524 	struct sockaddr *to;
525 	int error;
526 
527 	if (mp->msg_name != NULL) {
528 		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
529 		if (error)
530 			return (error);
531 		mp->msg_name = to;
532 	} else
533 		to = NULL;
534 
535 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
536 	    segflg);
537 
538 	if (to)
539 		free(to, M_SONAME);
540 	return (error);
541 }
542 
543 /* Return 0 if IP_HDRINCL is set for the given socket. */
544 static int
545 linux_check_hdrincl(struct thread *td, int s)
546 {
547 	int error, optval, size_val;
548 
549 	size_val = sizeof(optval);
550 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
551 	    &optval, UIO_SYSSPACE, &size_val);
552 	if (error)
553 		return (error);
554 
555 	return (optval == 0);
556 }
557 
558 struct linux_sendto_args {
559 	int s;
560 	l_uintptr_t msg;
561 	int len;
562 	int flags;
563 	l_uintptr_t to;
564 	int tolen;
565 };
566 
567 /*
568  * Updated sendto() when IP_HDRINCL is set:
569  * tweak endian-dependent fields in the IP packet.
570  */
571 static int
572 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
573 {
574 /*
575  * linux_ip_copysize defines how many bytes we should copy
576  * from the beginning of the IP packet before we customize it for BSD.
577  * It should include all the fields we modify (ip_len and ip_off).
578  */
579 #define linux_ip_copysize	8
580 
581 	struct ip *packet;
582 	struct msghdr msg;
583 	struct iovec aiov[1];
584 	int error;
585 
586 	/* Check that the packet isn't too big or too small. */
587 	if (linux_args->len < linux_ip_copysize ||
588 	    linux_args->len > IP_MAXPACKET)
589 		return (EINVAL);
590 
591 	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
592 
593 	/* Make kernel copy of the packet to be sent */
594 	if ((error = copyin(PTRIN(linux_args->msg), packet,
595 	    linux_args->len)))
596 		goto goout;
597 
598 	/* Convert fields from Linux to BSD raw IP socket format */
599 	packet->ip_len = linux_args->len;
600 	packet->ip_off = ntohs(packet->ip_off);
601 
602 	/* Prepare the msghdr and iovec structures describing the new packet */
603 	msg.msg_name = PTRIN(linux_args->to);
604 	msg.msg_namelen = linux_args->tolen;
605 	msg.msg_iov = aiov;
606 	msg.msg_iovlen = 1;
607 	msg.msg_control = NULL;
608 	msg.msg_flags = 0;
609 	aiov[0].iov_base = (char *)packet;
610 	aiov[0].iov_len = linux_args->len;
611 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
612 	    NULL, UIO_SYSSPACE);
613 goout:
614 	free(packet, M_TEMP);
615 	return (error);
616 }
617 
618 struct linux_socket_args {
619 	int domain;
620 	int type;
621 	int protocol;
622 };
623 
624 static int
625 linux_socket(struct thread *td, struct linux_socket_args *args)
626 {
627 	struct socket_args /* {
628 		int domain;
629 		int type;
630 		int protocol;
631 	} */ bsd_args;
632 	int retval_socket, socket_flags;
633 
634 	bsd_args.protocol = args->protocol;
635 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
636 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
637 		return (EINVAL);
638 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
639 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
640 		return (EINVAL);
641 	bsd_args.domain = linux_to_bsd_domain(args->domain);
642 	if (bsd_args.domain == -1)
643 		return (EAFNOSUPPORT);
644 
645 	retval_socket = socket(td, &bsd_args);
646 	if (retval_socket)
647 		return (retval_socket);
648 
649 	retval_socket = linux_set_socket_flags(td, td->td_retval[0],
650 	    socket_flags);
651 	if (retval_socket) {
652 		(void)kern_close(td, td->td_retval[0]);
653 		goto out;
654 	}
655 
656 	if (bsd_args.type == SOCK_RAW
657 	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
658 	    && bsd_args.domain == PF_INET) {
659 		/* It's a raw IP socket: set the IP_HDRINCL option. */
660 		int hdrincl;
661 
662 		hdrincl = 1;
663 		/* We ignore any error returned by kern_setsockopt() */
664 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
665 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
666 	}
667 #ifdef INET6
668 	/*
669 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
670 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
671 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
672 	 * sysctl value.
673 	 */
674 	if (bsd_args.domain == PF_INET6) {
675 		int v6only;
676 
677 		v6only = 0;
678 		/* We ignore any error returned by setsockopt() */
679 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
680 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
681 	}
682 #endif
683 
684 out:
685 	return (retval_socket);
686 }
687 
688 struct linux_bind_args {
689 	int s;
690 	l_uintptr_t name;
691 	int namelen;
692 };
693 
694 static int
695 linux_bind(struct thread *td, struct linux_bind_args *args)
696 {
697 	struct sockaddr *sa;
698 	int error;
699 
700 	error = linux_getsockaddr(&sa, PTRIN(args->name),
701 	    args->namelen);
702 	if (error)
703 		return (error);
704 
705 	error = kern_bind(td, args->s, sa);
706 	free(sa, M_SONAME);
707 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
708 	   	return (EINVAL);
709 	return (error);
710 }
711 
712 struct linux_connect_args {
713 	int s;
714 	l_uintptr_t name;
715 	int namelen;
716 };
717 int linux_connect(struct thread *, struct linux_connect_args *);
718 
719 int
720 linux_connect(struct thread *td, struct linux_connect_args *args)
721 {
722 	struct socket *so;
723 	struct sockaddr *sa;
724 	u_int fflag;
725 	int error;
726 
727 	error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
728 	    args->namelen);
729 	if (error)
730 		return (error);
731 
732 	error = kern_connect(td, args->s, sa);
733 	free(sa, M_SONAME);
734 	if (error != EISCONN)
735 		return (error);
736 
737 	/*
738 	 * Linux doesn't return EISCONN the first time it occurs,
739 	 * when on a non-blocking socket. Instead it returns the
740 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
741 	 *
742 	 * XXXRW: Instead of using fgetsock(), check that it is a
743 	 * socket and use the file descriptor reference instead of
744 	 * creating a new one.
745 	 */
746 	error = fgetsock(td, args->s, &so, &fflag);
747 	if (error == 0) {
748 		error = EISCONN;
749 		if (fflag & FNONBLOCK) {
750 			SOCK_LOCK(so);
751 			if (so->so_emuldata == 0)
752 				error = so->so_error;
753 			so->so_emuldata = (void *)1;
754 			SOCK_UNLOCK(so);
755 		}
756 		fputsock(so);
757 	}
758 	return (error);
759 }
760 
761 struct linux_listen_args {
762 	int s;
763 	int backlog;
764 };
765 
766 static int
767 linux_listen(struct thread *td, struct linux_listen_args *args)
768 {
769 	struct listen_args /* {
770 		int s;
771 		int backlog;
772 	} */ bsd_args;
773 
774 	bsd_args.s = args->s;
775 	bsd_args.backlog = args->backlog;
776 	return (listen(td, &bsd_args));
777 }
778 
779 static int
780 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
781     l_uintptr_t namelen, int flags)
782 {
783 	struct accept_args /* {
784 		int	s;
785 		struct sockaddr * __restrict name;
786 		socklen_t * __restrict anamelen;
787 	} */ bsd_args;
788 	int error;
789 
790 	if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
791 		return (EINVAL);
792 
793 	bsd_args.s = s;
794 	/* XXX: */
795 	bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
796 	bsd_args.anamelen = PTRIN(namelen);/* XXX */
797 	error = accept(td, &bsd_args);
798 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
799 	if (error) {
800 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
801 			return (EINVAL);
802 		return (error);
803 	}
804 
805 	/*
806 	 * linux appears not to copy flags from the parent socket to the
807 	 * accepted one, so we must clear the flags in the new descriptor
808 	 * and apply the requested flags.
809 	 */
810 	error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
811 	if (error)
812 		goto out;
813 	error = linux_set_socket_flags(td, td->td_retval[0], flags);
814 	if (error)
815 		goto out;
816 	if (addr)
817 		error = linux_sa_put(PTRIN(addr));
818 
819 out:
820 	if (error) {
821 		(void)kern_close(td, td->td_retval[0]);
822 		td->td_retval[0] = 0;
823 	}
824 	return (error);
825 }
826 
827 struct linux_accept_args {
828 	int s;
829 	l_uintptr_t addr;
830 	l_uintptr_t namelen;
831 };
832 
833 static int
834 linux_accept(struct thread *td, struct linux_accept_args *args)
835 {
836 
837 	return (linux_accept_common(td, args->s, args->addr,
838 	    args->namelen, 0));
839 }
840 
841 struct linux_accept4_args {
842 	int s;
843 	l_uintptr_t addr;
844 	l_uintptr_t namelen;
845 	int flags;
846 };
847 
848 static int
849 linux_accept4(struct thread *td, struct linux_accept4_args *args)
850 {
851 
852 	return (linux_accept_common(td, args->s, args->addr,
853 	    args->namelen, args->flags));
854 }
855 
856 struct linux_getsockname_args {
857 	int s;
858 	l_uintptr_t addr;
859 	l_uintptr_t namelen;
860 };
861 
862 static int
863 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
864 {
865 	struct getsockname_args /* {
866 		int	fdes;
867 		struct sockaddr * __restrict asa;
868 		socklen_t * __restrict alen;
869 	} */ bsd_args;
870 	int error;
871 
872 	bsd_args.fdes = args->s;
873 	/* XXX: */
874 	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
875 	bsd_args.alen = PTRIN(args->namelen);	/* XXX */
876 	error = getsockname(td, &bsd_args);
877 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
878 	if (error)
879 		return (error);
880 	error = linux_sa_put(PTRIN(args->addr));
881 	if (error)
882 		return (error);
883 	return (0);
884 }
885 
886 struct linux_getpeername_args {
887 	int s;
888 	l_uintptr_t addr;
889 	l_uintptr_t namelen;
890 };
891 
892 static int
893 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
894 {
895 	struct getpeername_args /* {
896 		int fdes;
897 		caddr_t asa;
898 		int *alen;
899 	} */ bsd_args;
900 	int error;
901 
902 	bsd_args.fdes = args->s;
903 	bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
904 	bsd_args.alen = (int *)PTRIN(args->namelen);
905 	error = getpeername(td, &bsd_args);
906 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
907 	if (error)
908 		return (error);
909 	error = linux_sa_put(PTRIN(args->addr));
910 	if (error)
911 		return (error);
912 	return (0);
913 }
914 
915 struct linux_socketpair_args {
916 	int domain;
917 	int type;
918 	int protocol;
919 	l_uintptr_t rsv;
920 };
921 
922 static int
923 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
924 {
925 	struct socketpair_args /* {
926 		int domain;
927 		int type;
928 		int protocol;
929 		int *rsv;
930 	} */ bsd_args;
931 	int error, socket_flags;
932 	int sv[2];
933 
934 	bsd_args.domain = linux_to_bsd_domain(args->domain);
935 	if (bsd_args.domain != PF_LOCAL)
936 		return (EAFNOSUPPORT);
937 
938 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
939 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
940 		return (EINVAL);
941 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
942 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
943 		return (EINVAL);
944 
945 	if (args->protocol != 0 && args->protocol != PF_UNIX)
946 
947 		/*
948 		 * Use of PF_UNIX as protocol argument is not right,
949 		 * but Linux does it.
950 		 * Do not map PF_UNIX as its Linux value is identical
951 		 * to FreeBSD one.
952 		 */
953 		return (EPROTONOSUPPORT);
954 	else
955 		bsd_args.protocol = 0;
956 	bsd_args.rsv = (int *)PTRIN(args->rsv);
957 	error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
958 	    bsd_args.protocol, sv);
959 	if (error)
960 		return (error);
961 	error = linux_set_socket_flags(td, sv[0], socket_flags);
962 	if (error)
963 		goto out;
964 	error = linux_set_socket_flags(td, sv[1], socket_flags);
965 	if (error)
966 		goto out;
967 
968 	error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
969 
970 out:
971 	if (error) {
972 		(void)kern_close(td, sv[0]);
973 		(void)kern_close(td, sv[1]);
974 	}
975 	return (error);
976 }
977 
978 struct linux_send_args {
979 	int s;
980 	l_uintptr_t msg;
981 	int len;
982 	int flags;
983 };
984 
985 static int
986 linux_send(struct thread *td, struct linux_send_args *args)
987 {
988 	struct sendto_args /* {
989 		int s;
990 		caddr_t buf;
991 		int len;
992 		int flags;
993 		caddr_t to;
994 		int tolen;
995 	} */ bsd_args;
996 
997 	bsd_args.s = args->s;
998 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
999 	bsd_args.len = args->len;
1000 	bsd_args.flags = args->flags;
1001 	bsd_args.to = NULL;
1002 	bsd_args.tolen = 0;
1003 	return sendto(td, &bsd_args);
1004 }
1005 
1006 struct linux_recv_args {
1007 	int s;
1008 	l_uintptr_t msg;
1009 	int len;
1010 	int flags;
1011 };
1012 
1013 static int
1014 linux_recv(struct thread *td, struct linux_recv_args *args)
1015 {
1016 	struct recvfrom_args /* {
1017 		int s;
1018 		caddr_t buf;
1019 		int len;
1020 		int flags;
1021 		struct sockaddr *from;
1022 		socklen_t fromlenaddr;
1023 	} */ bsd_args;
1024 
1025 	bsd_args.s = args->s;
1026 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1027 	bsd_args.len = args->len;
1028 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1029 	bsd_args.from = NULL;
1030 	bsd_args.fromlenaddr = 0;
1031 	return (recvfrom(td, &bsd_args));
1032 }
1033 
1034 static int
1035 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1036 {
1037 	struct msghdr msg;
1038 	struct iovec aiov;
1039 	int error;
1040 
1041 	if (linux_check_hdrincl(td, args->s) == 0)
1042 		/* IP_HDRINCL set, tweak the packet before sending */
1043 		return (linux_sendto_hdrincl(td, args));
1044 
1045 	msg.msg_name = PTRIN(args->to);
1046 	msg.msg_namelen = args->tolen;
1047 	msg.msg_iov = &aiov;
1048 	msg.msg_iovlen = 1;
1049 	msg.msg_control = NULL;
1050 	msg.msg_flags = 0;
1051 	aiov.iov_base = PTRIN(args->msg);
1052 	aiov.iov_len = args->len;
1053 	error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1054 	    UIO_USERSPACE);
1055 	return (error);
1056 }
1057 
1058 struct linux_recvfrom_args {
1059 	int s;
1060 	l_uintptr_t buf;
1061 	int len;
1062 	int flags;
1063 	l_uintptr_t from;
1064 	l_uintptr_t fromlen;
1065 };
1066 
1067 static int
1068 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1069 {
1070 	struct recvfrom_args /* {
1071 		int	s;
1072 		caddr_t	buf;
1073 		size_t	len;
1074 		int	flags;
1075 		struct sockaddr * __restrict from;
1076 		socklen_t * __restrict fromlenaddr;
1077 	} */ bsd_args;
1078 	size_t len;
1079 	int error;
1080 
1081 	if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1082 		return (error);
1083 
1084 	bsd_args.s = args->s;
1085 	bsd_args.buf = PTRIN(args->buf);
1086 	bsd_args.len = args->len;
1087 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1088 	/* XXX: */
1089 	bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1090 	bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1091 
1092 	linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1093 	error = recvfrom(td, &bsd_args);
1094 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1095 
1096 	if (error)
1097 		return (error);
1098 	if (args->from) {
1099 		error = linux_sa_put((struct osockaddr *)
1100 		    PTRIN(args->from));
1101 		if (error)
1102 			return (error);
1103 	}
1104 	return (0);
1105 }
1106 
1107 struct linux_sendmsg_args {
1108 	int s;
1109 	l_uintptr_t msg;
1110 	int flags;
1111 };
1112 
1113 static int
1114 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1115 {
1116 	struct cmsghdr *cmsg;
1117 	struct cmsgcred cmcred;
1118 	struct mbuf *control;
1119 	struct msghdr msg;
1120 	struct l_cmsghdr linux_cmsg;
1121 	struct l_cmsghdr *ptr_cmsg;
1122 	struct l_msghdr linux_msg;
1123 	struct iovec *iov;
1124 	socklen_t datalen;
1125 	struct sockaddr *sa;
1126 	sa_family_t sa_family;
1127 	void *data;
1128 	int error;
1129 
1130 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1131 	if (error)
1132 		return (error);
1133 
1134 	/*
1135 	 * Some Linux applications (ping) define a non-NULL control data
1136 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1137 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1138 	 * order to handle this case.  This should be checked, but allows the
1139 	 * Linux ping to work.
1140 	 */
1141 	if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1142 		linux_msg.msg_control = PTROUT(NULL);
1143 
1144 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1145 	if (error)
1146 		return (error);
1147 
1148 #ifdef COMPAT_LINUX32
1149 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1150 	    &iov, EMSGSIZE);
1151 #else
1152 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1153 #endif
1154 	if (error)
1155 		return (error);
1156 
1157 	control = NULL;
1158 	cmsg = NULL;
1159 
1160 	if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1161 		error = kern_getsockname(td, args->s, &sa, &datalen);
1162 		if (error)
1163 			goto bad;
1164 		sa_family = sa->sa_family;
1165 		free(sa, M_SONAME);
1166 
1167 		error = ENOBUFS;
1168 		cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1169 		control = m_get(M_WAIT, MT_CONTROL);
1170 		if (control == NULL)
1171 			goto bad;
1172 
1173 		do {
1174 			error = copyin(ptr_cmsg, &linux_cmsg,
1175 			    sizeof(struct l_cmsghdr));
1176 			if (error)
1177 				goto bad;
1178 
1179 			error = EINVAL;
1180 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1181 				goto bad;
1182 
1183 			/*
1184 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1185 			 * so return EINVAL in any other cmsg_type
1186 			 */
1187 			cmsg->cmsg_type =
1188 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1189 			cmsg->cmsg_level =
1190 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1191 			if (cmsg->cmsg_type == -1
1192 			    || cmsg->cmsg_level != SOL_SOCKET)
1193 				goto bad;
1194 
1195 			/*
1196 			 * Some applications (e.g. pulseaudio) attempt to
1197 			 * send ancillary data even if the underlying protocol
1198 			 * doesn't support it which is not allowed in the
1199 			 * FreeBSD system call interface.
1200 			 */
1201 			if (sa_family != AF_UNIX)
1202 				continue;
1203 
1204 			data = LINUX_CMSG_DATA(ptr_cmsg);
1205 			datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1206 
1207 			switch (cmsg->cmsg_type)
1208 			{
1209 			case SCM_RIGHTS:
1210 				break;
1211 
1212 			case SCM_CREDS:
1213 				data = &cmcred;
1214 				datalen = sizeof(cmcred);
1215 
1216 				/*
1217 				 * The lower levels will fill in the structure
1218 				 */
1219 				bzero(data, datalen);
1220 				break;
1221 			}
1222 
1223 			cmsg->cmsg_len = CMSG_LEN(datalen);
1224 
1225 			error = ENOBUFS;
1226 			if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg))
1227 				goto bad;
1228 			if (!m_append(control, datalen, (c_caddr_t) data))
1229 				goto bad;
1230 		} while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1231 
1232 		if (m_length(control, NULL) == 0) {
1233 			m_freem(control);
1234 			control = NULL;
1235 		}
1236 	}
1237 
1238 	msg.msg_iov = iov;
1239 	msg.msg_flags = 0;
1240 	error = linux_sendit(td, args->s, &msg, args->flags, control,
1241 	    UIO_USERSPACE);
1242 
1243 bad:
1244 	free(iov, M_IOV);
1245 	if (cmsg)
1246 		free(cmsg, M_TEMP);
1247 	return (error);
1248 }
1249 
1250 struct linux_recvmsg_args {
1251 	int s;
1252 	l_uintptr_t msg;
1253 	int flags;
1254 };
1255 
1256 static int
1257 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1258 {
1259 	struct cmsghdr *cm;
1260 	struct cmsgcred *cmcred;
1261 	struct msghdr msg;
1262 	struct l_cmsghdr *linux_cmsg = NULL;
1263 	struct l_ucred linux_ucred;
1264 	socklen_t datalen, outlen;
1265 	struct l_msghdr linux_msg;
1266 	struct iovec *iov, *uiov;
1267 	struct mbuf *control = NULL;
1268 	struct mbuf **controlp;
1269 	caddr_t outbuf;
1270 	void *data;
1271 	int error, i, fd, fds, *fdp;
1272 
1273 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1274 	if (error)
1275 		return (error);
1276 
1277 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1278 	if (error)
1279 		return (error);
1280 
1281 #ifdef COMPAT_LINUX32
1282 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1283 	    &iov, EMSGSIZE);
1284 #else
1285 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1286 #endif
1287 	if (error)
1288 		return (error);
1289 
1290 	if (msg.msg_name) {
1291 		error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1292 		    msg.msg_namelen);
1293 		if (error)
1294 			goto bad;
1295 	}
1296 
1297 	uiov = msg.msg_iov;
1298 	msg.msg_iov = iov;
1299 	controlp = (msg.msg_control != NULL) ? &control : NULL;
1300 	error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1301 	msg.msg_iov = uiov;
1302 	if (error)
1303 		goto bad;
1304 
1305 	error = bsd_to_linux_msghdr(&msg, &linux_msg);
1306 	if (error)
1307 		goto bad;
1308 
1309 	if (linux_msg.msg_name) {
1310 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1311 		    PTRIN(linux_msg.msg_name));
1312 		if (error)
1313 			goto bad;
1314 	}
1315 	if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1316 		error = linux_sa_put(PTRIN(linux_msg.msg_name));
1317 		if (error)
1318 			goto bad;
1319 	}
1320 
1321 	outbuf = PTRIN(linux_msg.msg_control);
1322 	outlen = 0;
1323 
1324 	if (control) {
1325 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1326 
1327 		msg.msg_control = mtod(control, struct cmsghdr *);
1328 		msg.msg_controllen = control->m_len;
1329 
1330 		cm = CMSG_FIRSTHDR(&msg);
1331 
1332 		while (cm != NULL) {
1333 			linux_cmsg->cmsg_type =
1334 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1335 			linux_cmsg->cmsg_level =
1336 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1337 			if (linux_cmsg->cmsg_type == -1
1338 			    || cm->cmsg_level != SOL_SOCKET)
1339 			{
1340 				error = EINVAL;
1341 				goto bad;
1342 			}
1343 
1344 			data = CMSG_DATA(cm);
1345 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1346 
1347 			switch (cm->cmsg_type)
1348 			{
1349 			case SCM_RIGHTS:
1350 				if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1351 					fds = datalen / sizeof(int);
1352 					fdp = data;
1353 					for (i = 0; i < fds; i++) {
1354 						fd = *fdp++;
1355 						(void)kern_fcntl(td, fd,
1356 						    F_SETFD, FD_CLOEXEC);
1357 					}
1358 				}
1359 				break;
1360 
1361 			case SCM_CREDS:
1362 				/*
1363 				 * Currently LOCAL_CREDS is never in
1364 				 * effect for Linux so no need to worry
1365 				 * about sockcred
1366 				 */
1367 				if (datalen != sizeof (*cmcred)) {
1368 					error = EMSGSIZE;
1369 					goto bad;
1370 				}
1371 				cmcred = (struct cmsgcred *)data;
1372 				bzero(&linux_ucred, sizeof(linux_ucred));
1373 				linux_ucred.pid = cmcred->cmcred_pid;
1374 				linux_ucred.uid = cmcred->cmcred_uid;
1375 				linux_ucred.gid = cmcred->cmcred_gid;
1376 				data = &linux_ucred;
1377 				datalen = sizeof(linux_ucred);
1378 				break;
1379 			}
1380 
1381 			if (outlen + LINUX_CMSG_LEN(datalen) >
1382 			    linux_msg.msg_controllen) {
1383 				if (outlen == 0) {
1384 					error = EMSGSIZE;
1385 					goto bad;
1386 				} else {
1387 					linux_msg.msg_flags |=
1388 					    LINUX_MSG_CTRUNC;
1389 					goto out;
1390 				}
1391 			}
1392 
1393 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1394 
1395 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1396 			if (error)
1397 				goto bad;
1398 			outbuf += L_CMSG_HDRSZ;
1399 
1400 			error = copyout(data, outbuf, datalen);
1401 			if (error)
1402 				goto bad;
1403 
1404 			outbuf += LINUX_CMSG_ALIGN(datalen);
1405 			outlen += LINUX_CMSG_LEN(datalen);
1406 
1407 			cm = CMSG_NXTHDR(&msg, cm);
1408 		}
1409 	}
1410 
1411 out:
1412 	linux_msg.msg_controllen = outlen;
1413 	error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1414 
1415 bad:
1416 	free(iov, M_IOV);
1417 	if (control != NULL)
1418 		m_freem(control);
1419 	if (linux_cmsg != NULL)
1420 		free(linux_cmsg, M_TEMP);
1421 
1422 	return (error);
1423 }
1424 
1425 struct linux_shutdown_args {
1426 	int s;
1427 	int how;
1428 };
1429 
1430 static int
1431 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1432 {
1433 	struct shutdown_args /* {
1434 		int s;
1435 		int how;
1436 	} */ bsd_args;
1437 
1438 	bsd_args.s = args->s;
1439 	bsd_args.how = args->how;
1440 	return (shutdown(td, &bsd_args));
1441 }
1442 
1443 struct linux_setsockopt_args {
1444 	int s;
1445 	int level;
1446 	int optname;
1447 	l_uintptr_t optval;
1448 	int optlen;
1449 };
1450 
1451 static int
1452 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1453 {
1454 	struct setsockopt_args /* {
1455 		int s;
1456 		int level;
1457 		int name;
1458 		caddr_t val;
1459 		int valsize;
1460 	} */ bsd_args;
1461 	l_timeval linux_tv;
1462 	struct timeval tv;
1463 	int error, name;
1464 
1465 	bsd_args.s = args->s;
1466 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1467 	switch (bsd_args.level) {
1468 	case SOL_SOCKET:
1469 		name = linux_to_bsd_so_sockopt(args->optname);
1470 		switch (name) {
1471 		case SO_RCVTIMEO:
1472 			/* FALLTHROUGH */
1473 		case SO_SNDTIMEO:
1474 			error = copyin(PTRIN(args->optval), &linux_tv,
1475 			    sizeof(linux_tv));
1476 			if (error)
1477 				return (error);
1478 			tv.tv_sec = linux_tv.tv_sec;
1479 			tv.tv_usec = linux_tv.tv_usec;
1480 			return (kern_setsockopt(td, args->s, bsd_args.level,
1481 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1482 			/* NOTREACHED */
1483 			break;
1484 		default:
1485 			break;
1486 		}
1487 		break;
1488 	case IPPROTO_IP:
1489 		name = linux_to_bsd_ip_sockopt(args->optname);
1490 		break;
1491 	case IPPROTO_TCP:
1492 		/* Linux TCP option values match BSD's */
1493 		name = args->optname;
1494 		break;
1495 	default:
1496 		name = -1;
1497 		break;
1498 	}
1499 	if (name == -1)
1500 		return (ENOPROTOOPT);
1501 
1502 	bsd_args.name = name;
1503 	bsd_args.val = PTRIN(args->optval);
1504 	bsd_args.valsize = args->optlen;
1505 
1506 	if (name == IPV6_NEXTHOP) {
1507 		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1508 			bsd_args.valsize);
1509 		error = setsockopt(td, &bsd_args);
1510 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1511 	} else
1512 		error = setsockopt(td, &bsd_args);
1513 
1514 	return (error);
1515 }
1516 
1517 struct linux_getsockopt_args {
1518 	int s;
1519 	int level;
1520 	int optname;
1521 	l_uintptr_t optval;
1522 	l_uintptr_t optlen;
1523 };
1524 
1525 static int
1526 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1527 {
1528 	struct getsockopt_args /* {
1529 		int s;
1530 		int level;
1531 		int name;
1532 		caddr_t val;
1533 		int *avalsize;
1534 	} */ bsd_args;
1535 	l_timeval linux_tv;
1536 	struct timeval tv;
1537 	socklen_t tv_len, xulen;
1538 	struct xucred xu;
1539 	struct l_ucred lxu;
1540 	int error, name;
1541 
1542 	bsd_args.s = args->s;
1543 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1544 	switch (bsd_args.level) {
1545 	case SOL_SOCKET:
1546 		name = linux_to_bsd_so_sockopt(args->optname);
1547 		switch (name) {
1548 		case SO_RCVTIMEO:
1549 			/* FALLTHROUGH */
1550 		case SO_SNDTIMEO:
1551 			tv_len = sizeof(tv);
1552 			error = kern_getsockopt(td, args->s, bsd_args.level,
1553 			    name, &tv, UIO_SYSSPACE, &tv_len);
1554 			if (error)
1555 				return (error);
1556 			linux_tv.tv_sec = tv.tv_sec;
1557 			linux_tv.tv_usec = tv.tv_usec;
1558 			return (copyout(&linux_tv, PTRIN(args->optval),
1559 			    sizeof(linux_tv)));
1560 			/* NOTREACHED */
1561 			break;
1562 		case LOCAL_PEERCRED:
1563 			if (args->optlen != sizeof(lxu))
1564 				return (EINVAL);
1565 			xulen = sizeof(xu);
1566 			error = kern_getsockopt(td, args->s, bsd_args.level,
1567 			    name, &xu, UIO_SYSSPACE, &xulen);
1568 			if (error)
1569 				return (error);
1570 			/*
1571 			 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1572 			 */
1573 			lxu.pid = 0;
1574 			lxu.uid = xu.cr_uid;
1575 			lxu.gid = xu.cr_gid;
1576 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1577 			/* NOTREACHED */
1578 			break;
1579 		default:
1580 			break;
1581 		}
1582 		break;
1583 	case IPPROTO_IP:
1584 		name = linux_to_bsd_ip_sockopt(args->optname);
1585 		break;
1586 	case IPPROTO_TCP:
1587 		/* Linux TCP option values match BSD's */
1588 		name = args->optname;
1589 		break;
1590 	default:
1591 		name = -1;
1592 		break;
1593 	}
1594 	if (name == -1)
1595 		return (EINVAL);
1596 
1597 	bsd_args.name = name;
1598 	bsd_args.val = PTRIN(args->optval);
1599 	bsd_args.avalsize = PTRIN(args->optlen);
1600 
1601 	if (name == IPV6_NEXTHOP) {
1602 		error = getsockopt(td, &bsd_args);
1603 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1604 	} else
1605 		error = getsockopt(td, &bsd_args);
1606 
1607 	return (error);
1608 }
1609 
1610 /* Argument list sizes for linux_socketcall */
1611 
1612 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1613 
1614 static const unsigned char lxs_args[] = {
1615 	LINUX_AL(0) /* unused*/,	LINUX_AL(3) /* socket */,
1616 	LINUX_AL(3) /* bind */,		LINUX_AL(3) /* connect */,
1617 	LINUX_AL(2) /* listen */,	LINUX_AL(3) /* accept */,
1618 	LINUX_AL(3) /* getsockname */,	LINUX_AL(3) /* getpeername */,
1619 	LINUX_AL(4) /* socketpair */,	LINUX_AL(4) /* send */,
1620 	LINUX_AL(4) /* recv */,		LINUX_AL(6) /* sendto */,
1621 	LINUX_AL(6) /* recvfrom */,	LINUX_AL(2) /* shutdown */,
1622 	LINUX_AL(5) /* setsockopt */,	LINUX_AL(5) /* getsockopt */,
1623 	LINUX_AL(3) /* sendmsg */,	LINUX_AL(3) /* recvmsg */,
1624 	LINUX_AL(4) /* accept4 */
1625 };
1626 
1627 #define	LINUX_AL_SIZE	sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1628 
1629 int
1630 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1631 {
1632 	l_ulong a[6];
1633 	void *arg;
1634 	int error;
1635 
1636 	if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1637 		return (EINVAL);
1638 	error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1639 	if (error)
1640 		return (error);
1641 
1642 	arg = a;
1643 	switch (args->what) {
1644 	case LINUX_SOCKET:
1645 		return (linux_socket(td, arg));
1646 	case LINUX_BIND:
1647 		return (linux_bind(td, arg));
1648 	case LINUX_CONNECT:
1649 		return (linux_connect(td, arg));
1650 	case LINUX_LISTEN:
1651 		return (linux_listen(td, arg));
1652 	case LINUX_ACCEPT:
1653 		return (linux_accept(td, arg));
1654 	case LINUX_GETSOCKNAME:
1655 		return (linux_getsockname(td, arg));
1656 	case LINUX_GETPEERNAME:
1657 		return (linux_getpeername(td, arg));
1658 	case LINUX_SOCKETPAIR:
1659 		return (linux_socketpair(td, arg));
1660 	case LINUX_SEND:
1661 		return (linux_send(td, arg));
1662 	case LINUX_RECV:
1663 		return (linux_recv(td, arg));
1664 	case LINUX_SENDTO:
1665 		return (linux_sendto(td, arg));
1666 	case LINUX_RECVFROM:
1667 		return (linux_recvfrom(td, arg));
1668 	case LINUX_SHUTDOWN:
1669 		return (linux_shutdown(td, arg));
1670 	case LINUX_SETSOCKOPT:
1671 		return (linux_setsockopt(td, arg));
1672 	case LINUX_GETSOCKOPT:
1673 		return (linux_getsockopt(td, arg));
1674 	case LINUX_SENDMSG:
1675 		return (linux_sendmsg(td, arg));
1676 	case LINUX_RECVMSG:
1677 		return (linux_recvmsg(td, arg));
1678 	case LINUX_ACCEPT4:
1679 		return (linux_accept4(td, arg));
1680 	}
1681 
1682 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1683 	return (ENOSYS);
1684 }
1685