xref: /freebsd/sys/compat/linux/linux_socket.c (revision 5ae59dec60e3815b621ae87f74a377cf3449ca55)
1 /*-
2  * Copyright (c) 1995 Søren 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/capsicum.h>
41 #include <sys/fcntl.h>
42 #include <sys/file.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mutex.h>
47 #include <sys/mbuf.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/syscallsubr.h>
51 #include <sys/uio.h>
52 #include <sys/syslog.h>
53 #include <sys/un.h>
54 
55 #include <net/if.h>
56 #include <net/vnet.h>
57 #include <netinet/in.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/ip.h>
60 #include <netinet/tcp.h>
61 #ifdef INET6
62 #include <netinet/ip6.h>
63 #include <netinet6/ip6_var.h>
64 #endif
65 
66 #ifdef COMPAT_LINUX32
67 #include <machine/../linux32/linux.h>
68 #include <machine/../linux32/linux32_proto.h>
69 #else
70 #include <machine/../linux/linux.h>
71 #include <machine/../linux/linux_proto.h>
72 #endif
73 #include <compat/linux/linux_socket.h>
74 #include <compat/linux/linux_util.h>
75 
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  * Copy the osockaddr structure pointed to by osa to kernel, adjust
82  * family and convert to sockaddr.
83  */
84 static int
85 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
86 {
87 	struct sockaddr *sa;
88 	struct osockaddr *kosa;
89 #ifdef INET6
90 	struct sockaddr_in6 *sin6;
91 	int oldv6size;
92 #endif
93 	char *name;
94 	int bdom, error, hdrlen, namelen;
95 
96 	if (salen < 2 || salen > UCHAR_MAX || !osa)
97 		return (EINVAL);
98 
99 #ifdef INET6
100 	oldv6size = 0;
101 	/*
102 	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
103 	 * if it's a v4-mapped address, so reserve the proper space
104 	 * for it.
105 	 */
106 	if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
107 		salen += sizeof(uint32_t);
108 		oldv6size = 1;
109 	}
110 #endif
111 
112 	kosa = malloc(salen, M_SONAME, M_WAITOK);
113 
114 	if ((error = copyin(osa, kosa, salen)))
115 		goto out;
116 
117 	bdom = linux_to_bsd_domain(kosa->sa_family);
118 	if (bdom == -1) {
119 		error = EAFNOSUPPORT;
120 		goto out;
121 	}
122 
123 #ifdef INET6
124 	/*
125 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
126 	 * which lacks the scope id compared with RFC2553 one. If we detect
127 	 * the situation, reject the address and write a message to system log.
128 	 *
129 	 * Still accept addresses for which the scope id is not used.
130 	 */
131 	if (oldv6size) {
132 		if (bdom == AF_INET6) {
133 			sin6 = (struct sockaddr_in6 *)kosa;
134 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
135 			    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
136 			     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
137 			     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
138 			     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
139 			     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
140 				sin6->sin6_scope_id = 0;
141 			} else {
142 				log(LOG_DEBUG,
143 				    "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
144 				error = EINVAL;
145 				goto out;
146 			}
147 		} else
148 			salen -= sizeof(uint32_t);
149 	}
150 #endif
151 	if (bdom == AF_INET) {
152 		if (salen < sizeof(struct sockaddr_in)) {
153 			error = EINVAL;
154 			goto out;
155 		}
156 		salen = sizeof(struct sockaddr_in);
157 	}
158 
159 	if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
160 		hdrlen = offsetof(struct sockaddr_un, sun_path);
161 		name = ((struct sockaddr_un *)kosa)->sun_path;
162 		if (*name == '\0') {
163 			/*
164 		 	 * Linux abstract namespace starts with a NULL byte.
165 			 * XXX We do not support abstract namespace yet.
166 			 */
167 			namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
168 		} else
169 			namelen = strnlen(name, salen - hdrlen);
170 		salen = hdrlen + namelen;
171 		if (salen > sizeof(struct sockaddr_un)) {
172 			error = ENAMETOOLONG;
173 			goto out;
174 		}
175 	}
176 
177 	sa = (struct sockaddr *)kosa;
178 	sa->sa_family = bdom;
179 	sa->sa_len = salen;
180 
181 	*sap = sa;
182 	return (0);
183 
184 out:
185 	free(kosa, M_SONAME);
186 	return (error);
187 }
188 
189 static int
190 linux_to_bsd_domain(int domain)
191 {
192 
193 	switch (domain) {
194 	case LINUX_AF_UNSPEC:
195 		return (AF_UNSPEC);
196 	case LINUX_AF_UNIX:
197 		return (AF_LOCAL);
198 	case LINUX_AF_INET:
199 		return (AF_INET);
200 	case LINUX_AF_INET6:
201 		return (AF_INET6);
202 	case LINUX_AF_AX25:
203 		return (AF_CCITT);
204 	case LINUX_AF_IPX:
205 		return (AF_IPX);
206 	case LINUX_AF_APPLETALK:
207 		return (AF_APPLETALK);
208 	}
209 	return (-1);
210 }
211 
212 static int
213 bsd_to_linux_domain(int domain)
214 {
215 
216 	switch (domain) {
217 	case AF_UNSPEC:
218 		return (LINUX_AF_UNSPEC);
219 	case AF_LOCAL:
220 		return (LINUX_AF_UNIX);
221 	case AF_INET:
222 		return (LINUX_AF_INET);
223 	case AF_INET6:
224 		return (LINUX_AF_INET6);
225 	case AF_CCITT:
226 		return (LINUX_AF_AX25);
227 	case AF_IPX:
228 		return (LINUX_AF_IPX);
229 	case AF_APPLETALK:
230 		return (LINUX_AF_APPLETALK);
231 	}
232 	return (-1);
233 }
234 
235 static int
236 linux_to_bsd_sockopt_level(int level)
237 {
238 
239 	switch (level) {
240 	case LINUX_SOL_SOCKET:
241 		return (SOL_SOCKET);
242 	}
243 	return (level);
244 }
245 
246 static int
247 bsd_to_linux_sockopt_level(int level)
248 {
249 
250 	switch (level) {
251 	case SOL_SOCKET:
252 		return (LINUX_SOL_SOCKET);
253 	}
254 	return (level);
255 }
256 
257 static int
258 linux_to_bsd_ip_sockopt(int opt)
259 {
260 
261 	switch (opt) {
262 	case LINUX_IP_TOS:
263 		return (IP_TOS);
264 	case LINUX_IP_TTL:
265 		return (IP_TTL);
266 	case LINUX_IP_OPTIONS:
267 		return (IP_OPTIONS);
268 	case LINUX_IP_MULTICAST_IF:
269 		return (IP_MULTICAST_IF);
270 	case LINUX_IP_MULTICAST_TTL:
271 		return (IP_MULTICAST_TTL);
272 	case LINUX_IP_MULTICAST_LOOP:
273 		return (IP_MULTICAST_LOOP);
274 	case LINUX_IP_ADD_MEMBERSHIP:
275 		return (IP_ADD_MEMBERSHIP);
276 	case LINUX_IP_DROP_MEMBERSHIP:
277 		return (IP_DROP_MEMBERSHIP);
278 	case LINUX_IP_HDRINCL:
279 		return (IP_HDRINCL);
280 	}
281 	return (-1);
282 }
283 
284 static int
285 linux_to_bsd_so_sockopt(int opt)
286 {
287 
288 	switch (opt) {
289 	case LINUX_SO_DEBUG:
290 		return (SO_DEBUG);
291 	case LINUX_SO_REUSEADDR:
292 		return (SO_REUSEADDR);
293 	case LINUX_SO_TYPE:
294 		return (SO_TYPE);
295 	case LINUX_SO_ERROR:
296 		return (SO_ERROR);
297 	case LINUX_SO_DONTROUTE:
298 		return (SO_DONTROUTE);
299 	case LINUX_SO_BROADCAST:
300 		return (SO_BROADCAST);
301 	case LINUX_SO_SNDBUF:
302 		return (SO_SNDBUF);
303 	case LINUX_SO_RCVBUF:
304 		return (SO_RCVBUF);
305 	case LINUX_SO_KEEPALIVE:
306 		return (SO_KEEPALIVE);
307 	case LINUX_SO_OOBINLINE:
308 		return (SO_OOBINLINE);
309 	case LINUX_SO_LINGER:
310 		return (SO_LINGER);
311 	case LINUX_SO_PEERCRED:
312 		return (LOCAL_PEERCRED);
313 	case LINUX_SO_RCVLOWAT:
314 		return (SO_RCVLOWAT);
315 	case LINUX_SO_SNDLOWAT:
316 		return (SO_SNDLOWAT);
317 	case LINUX_SO_RCVTIMEO:
318 		return (SO_RCVTIMEO);
319 	case LINUX_SO_SNDTIMEO:
320 		return (SO_SNDTIMEO);
321 	case LINUX_SO_TIMESTAMP:
322 		return (SO_TIMESTAMP);
323 	case LINUX_SO_ACCEPTCONN:
324 		return (SO_ACCEPTCONN);
325 	}
326 	return (-1);
327 }
328 
329 static int
330 linux_to_bsd_tcp_sockopt(int opt)
331 {
332 
333 	switch (opt) {
334 	case LINUX_TCP_NODELAY:
335 		return (TCP_NODELAY);
336 	case LINUX_TCP_MAXSEG:
337 		return (TCP_MAXSEG);
338 	case LINUX_TCP_KEEPIDLE:
339 		return (TCP_KEEPIDLE);
340 	case LINUX_TCP_KEEPINTVL:
341 		return (TCP_KEEPINTVL);
342 	case LINUX_TCP_KEEPCNT:
343 		return (TCP_KEEPCNT);
344 	case LINUX_TCP_MD5SIG:
345 		return (TCP_MD5SIG);
346 	}
347 	return (-1);
348 }
349 
350 static int
351 linux_to_bsd_msg_flags(int flags)
352 {
353 	int ret_flags = 0;
354 
355 	if (flags & LINUX_MSG_OOB)
356 		ret_flags |= MSG_OOB;
357 	if (flags & LINUX_MSG_PEEK)
358 		ret_flags |= MSG_PEEK;
359 	if (flags & LINUX_MSG_DONTROUTE)
360 		ret_flags |= MSG_DONTROUTE;
361 	if (flags & LINUX_MSG_CTRUNC)
362 		ret_flags |= MSG_CTRUNC;
363 	if (flags & LINUX_MSG_TRUNC)
364 		ret_flags |= MSG_TRUNC;
365 	if (flags & LINUX_MSG_DONTWAIT)
366 		ret_flags |= MSG_DONTWAIT;
367 	if (flags & LINUX_MSG_EOR)
368 		ret_flags |= MSG_EOR;
369 	if (flags & LINUX_MSG_WAITALL)
370 		ret_flags |= MSG_WAITALL;
371 	if (flags & LINUX_MSG_NOSIGNAL)
372 		ret_flags |= MSG_NOSIGNAL;
373 #if 0 /* not handled */
374 	if (flags & LINUX_MSG_PROXY)
375 		;
376 	if (flags & LINUX_MSG_FIN)
377 		;
378 	if (flags & LINUX_MSG_SYN)
379 		;
380 	if (flags & LINUX_MSG_CONFIRM)
381 		;
382 	if (flags & LINUX_MSG_RST)
383 		;
384 	if (flags & LINUX_MSG_ERRQUEUE)
385 		;
386 #endif
387 	return ret_flags;
388 }
389 
390 /*
391 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
392 * native syscall will fault.  Thus, we don't really need to check the
393 * return values for these functions.
394 */
395 
396 static int
397 bsd_to_linux_sockaddr(struct sockaddr *arg)
398 {
399 	struct sockaddr sa;
400 	size_t sa_len = sizeof(struct sockaddr);
401 	int error;
402 
403 	if ((error = copyin(arg, &sa, sa_len)))
404 		return (error);
405 
406 	*(u_short *)&sa = sa.sa_family;
407 
408 	error = copyout(&sa, arg, sa_len);
409 
410 	return (error);
411 }
412 
413 static int
414 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
415 {
416 	struct sockaddr sa;
417 	size_t sa_len = sizeof(struct sockaddr);
418 	int error;
419 
420 	if ((error = copyin(arg, &sa, sa_len)))
421 		return (error);
422 
423 	sa.sa_family = *(sa_family_t *)&sa;
424 	sa.sa_len = len;
425 
426 	error = copyout(&sa, arg, sa_len);
427 
428 	return (error);
429 }
430 
431 
432 static int
433 linux_sa_put(struct osockaddr *osa)
434 {
435 	struct osockaddr sa;
436 	int error, bdom;
437 
438 	/*
439 	 * Only read/write the osockaddr family part, the rest is
440 	 * not changed.
441 	 */
442 	error = copyin(osa, &sa, sizeof(sa.sa_family));
443 	if (error)
444 		return (error);
445 
446 	bdom = bsd_to_linux_domain(sa.sa_family);
447 	if (bdom == -1)
448 		return (EINVAL);
449 
450 	sa.sa_family = bdom;
451 	error = copyout(&sa, osa, sizeof(sa.sa_family));
452 	if (error)
453 		return (error);
454 
455 	return (0);
456 }
457 
458 static int
459 linux_to_bsd_cmsg_type(int cmsg_type)
460 {
461 
462 	switch (cmsg_type) {
463 	case LINUX_SCM_RIGHTS:
464 		return (SCM_RIGHTS);
465 	case LINUX_SCM_CREDENTIALS:
466 		return (SCM_CREDS);
467 	}
468 	return (-1);
469 }
470 
471 static int
472 bsd_to_linux_cmsg_type(int cmsg_type)
473 {
474 
475 	switch (cmsg_type) {
476 	case SCM_RIGHTS:
477 		return (LINUX_SCM_RIGHTS);
478 	case SCM_CREDS:
479 		return (LINUX_SCM_CREDENTIALS);
480 	}
481 	return (-1);
482 }
483 
484 static int
485 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
486 {
487 	if (lhdr->msg_controllen > INT_MAX)
488 		return (ENOBUFS);
489 
490 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
491 	bhdr->msg_namelen	= lhdr->msg_namelen;
492 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
493 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
494 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
495 
496 	/*
497 	 * msg_controllen is skipped since BSD and LINUX control messages
498 	 * are potentially different sizes (e.g. the cred structure used
499 	 * by SCM_CREDS is different between the two operating system).
500 	 *
501 	 * The caller can set it (if necessary) after converting all the
502 	 * control messages.
503 	 */
504 
505 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
506 	return (0);
507 }
508 
509 static int
510 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
511 {
512 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
513 	lhdr->msg_namelen	= bhdr->msg_namelen;
514 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
515 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
516 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
517 
518 	/*
519 	 * msg_controllen is skipped since BSD and LINUX control messages
520 	 * are potentially different sizes (e.g. the cred structure used
521 	 * by SCM_CREDS is different between the two operating system).
522 	 *
523 	 * The caller can set it (if necessary) after converting all the
524 	 * control messages.
525 	 */
526 
527 	/* msg_flags skipped */
528 	return (0);
529 }
530 
531 static int
532 linux_set_socket_flags(struct thread *td, int s, int flags)
533 {
534 	int error;
535 
536 	if (flags & LINUX_SOCK_NONBLOCK) {
537 		error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
538 		if (error)
539 			return (error);
540 	}
541 	if (flags & LINUX_SOCK_CLOEXEC) {
542 		error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
543 		if (error)
544 			return (error);
545 	}
546 	return (0);
547 }
548 
549 static int
550 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
551     struct mbuf *control, enum uio_seg segflg)
552 {
553 	struct sockaddr *to;
554 	int error;
555 
556 	if (mp->msg_name != NULL) {
557 		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
558 		if (error)
559 			return (error);
560 		mp->msg_name = to;
561 	} else
562 		to = NULL;
563 
564 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
565 	    segflg);
566 
567 	if (to)
568 		free(to, M_SONAME);
569 	return (error);
570 }
571 
572 /* Return 0 if IP_HDRINCL is set for the given socket. */
573 static int
574 linux_check_hdrincl(struct thread *td, int s)
575 {
576 	int error, optval, size_val;
577 
578 	size_val = sizeof(optval);
579 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
580 	    &optval, UIO_SYSSPACE, &size_val);
581 	if (error)
582 		return (error);
583 
584 	return (optval == 0);
585 }
586 
587 struct linux_sendto_args {
588 	int s;
589 	l_uintptr_t msg;
590 	int len;
591 	int flags;
592 	l_uintptr_t to;
593 	int tolen;
594 };
595 
596 /*
597  * Updated sendto() when IP_HDRINCL is set:
598  * tweak endian-dependent fields in the IP packet.
599  */
600 static int
601 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
602 {
603 /*
604  * linux_ip_copysize defines how many bytes we should copy
605  * from the beginning of the IP packet before we customize it for BSD.
606  * It should include all the fields we modify (ip_len and ip_off).
607  */
608 #define linux_ip_copysize	8
609 
610 	struct ip *packet;
611 	struct msghdr msg;
612 	struct iovec aiov[1];
613 	int error;
614 
615 	/* Check that the packet isn't too big or too small. */
616 	if (linux_args->len < linux_ip_copysize ||
617 	    linux_args->len > IP_MAXPACKET)
618 		return (EINVAL);
619 
620 	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
621 
622 	/* Make kernel copy of the packet to be sent */
623 	if ((error = copyin(PTRIN(linux_args->msg), packet,
624 	    linux_args->len)))
625 		goto goout;
626 
627 	/* Convert fields from Linux to BSD raw IP socket format */
628 	packet->ip_len = linux_args->len;
629 	packet->ip_off = ntohs(packet->ip_off);
630 
631 	/* Prepare the msghdr and iovec structures describing the new packet */
632 	msg.msg_name = PTRIN(linux_args->to);
633 	msg.msg_namelen = linux_args->tolen;
634 	msg.msg_iov = aiov;
635 	msg.msg_iovlen = 1;
636 	msg.msg_control = NULL;
637 	msg.msg_flags = 0;
638 	aiov[0].iov_base = (char *)packet;
639 	aiov[0].iov_len = linux_args->len;
640 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
641 	    NULL, UIO_SYSSPACE);
642 goout:
643 	free(packet, M_TEMP);
644 	return (error);
645 }
646 
647 struct linux_socket_args {
648 	int domain;
649 	int type;
650 	int protocol;
651 };
652 
653 static int
654 linux_socket(struct thread *td, struct linux_socket_args *args)
655 {
656 	struct socket_args /* {
657 		int domain;
658 		int type;
659 		int protocol;
660 	} */ bsd_args;
661 	int retval_socket, socket_flags;
662 
663 	bsd_args.protocol = args->protocol;
664 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
665 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
666 		return (EINVAL);
667 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
668 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
669 		return (EINVAL);
670 	bsd_args.domain = linux_to_bsd_domain(args->domain);
671 	if (bsd_args.domain == -1)
672 		return (EAFNOSUPPORT);
673 
674 	retval_socket = sys_socket(td, &bsd_args);
675 	if (retval_socket)
676 		return (retval_socket);
677 
678 	retval_socket = linux_set_socket_flags(td, td->td_retval[0],
679 	    socket_flags);
680 	if (retval_socket) {
681 		(void)kern_close(td, td->td_retval[0]);
682 		goto out;
683 	}
684 
685 	if (bsd_args.type == SOCK_RAW
686 	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
687 	    && bsd_args.domain == PF_INET) {
688 		/* It's a raw IP socket: set the IP_HDRINCL option. */
689 		int hdrincl;
690 
691 		hdrincl = 1;
692 		/* We ignore any error returned by kern_setsockopt() */
693 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
694 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
695 	}
696 #ifdef INET6
697 	/*
698 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
699 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
700 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
701 	 * sysctl value.
702 	 */
703 	if (bsd_args.domain == PF_INET6) {
704 		int v6only;
705 
706 		v6only = 0;
707 		/* We ignore any error returned by setsockopt() */
708 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
709 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
710 	}
711 #endif
712 
713 out:
714 	return (retval_socket);
715 }
716 
717 struct linux_bind_args {
718 	int s;
719 	l_uintptr_t name;
720 	int namelen;
721 };
722 
723 static int
724 linux_bind(struct thread *td, struct linux_bind_args *args)
725 {
726 	struct sockaddr *sa;
727 	int error;
728 
729 	error = linux_getsockaddr(&sa, PTRIN(args->name),
730 	    args->namelen);
731 	if (error)
732 		return (error);
733 
734 	error = kern_bind(td, args->s, sa);
735 	free(sa, M_SONAME);
736 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
737 	   	return (EINVAL);
738 	return (error);
739 }
740 
741 struct linux_connect_args {
742 	int s;
743 	l_uintptr_t name;
744 	int namelen;
745 };
746 int linux_connect(struct thread *, struct linux_connect_args *);
747 
748 int
749 linux_connect(struct thread *td, struct linux_connect_args *args)
750 {
751 	cap_rights_t rights;
752 	struct socket *so;
753 	struct sockaddr *sa;
754 	u_int fflag;
755 	int error;
756 
757 	error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
758 	    args->namelen);
759 	if (error)
760 		return (error);
761 
762 	error = kern_connect(td, args->s, sa);
763 	free(sa, M_SONAME);
764 	if (error != EISCONN)
765 		return (error);
766 
767 	/*
768 	 * Linux doesn't return EISCONN the first time it occurs,
769 	 * when on a non-blocking socket. Instead it returns the
770 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
771 	 *
772 	 * XXXRW: Instead of using fgetsock(), check that it is a
773 	 * socket and use the file descriptor reference instead of
774 	 * creating a new one.
775 	 */
776 	error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
777 	    &so, &fflag);
778 	if (error == 0) {
779 		error = EISCONN;
780 		if (fflag & FNONBLOCK) {
781 			SOCK_LOCK(so);
782 			if (so->so_emuldata == 0)
783 				error = so->so_error;
784 			so->so_emuldata = (void *)1;
785 			SOCK_UNLOCK(so);
786 		}
787 		fputsock(so);
788 	}
789 	return (error);
790 }
791 
792 struct linux_listen_args {
793 	int s;
794 	int backlog;
795 };
796 
797 static int
798 linux_listen(struct thread *td, struct linux_listen_args *args)
799 {
800 	struct listen_args /* {
801 		int s;
802 		int backlog;
803 	} */ bsd_args;
804 
805 	bsd_args.s = args->s;
806 	bsd_args.backlog = args->backlog;
807 	return (sys_listen(td, &bsd_args));
808 }
809 
810 static int
811 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
812     l_uintptr_t namelen, int flags)
813 {
814 	struct accept_args /* {
815 		int	s;
816 		struct sockaddr * __restrict name;
817 		socklen_t * __restrict anamelen;
818 	} */ bsd_args;
819 	int error;
820 
821 	if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
822 		return (EINVAL);
823 
824 	bsd_args.s = s;
825 	/* XXX: */
826 	bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
827 	bsd_args.anamelen = PTRIN(namelen);/* XXX */
828 	error = sys_accept(td, &bsd_args);
829 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
830 	if (error) {
831 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
832 			return (EINVAL);
833 		return (error);
834 	}
835 
836 	/*
837 	 * linux appears not to copy flags from the parent socket to the
838 	 * accepted one, so we must clear the flags in the new descriptor
839 	 * and apply the requested flags.
840 	 */
841 	error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
842 	if (error)
843 		goto out;
844 	error = linux_set_socket_flags(td, td->td_retval[0], flags);
845 	if (error)
846 		goto out;
847 	if (addr)
848 		error = linux_sa_put(PTRIN(addr));
849 
850 out:
851 	if (error) {
852 		(void)kern_close(td, td->td_retval[0]);
853 		td->td_retval[0] = 0;
854 	}
855 	return (error);
856 }
857 
858 struct linux_accept_args {
859 	int s;
860 	l_uintptr_t addr;
861 	l_uintptr_t namelen;
862 };
863 
864 static int
865 linux_accept(struct thread *td, struct linux_accept_args *args)
866 {
867 
868 	return (linux_accept_common(td, args->s, args->addr,
869 	    args->namelen, 0));
870 }
871 
872 struct linux_accept4_args {
873 	int s;
874 	l_uintptr_t addr;
875 	l_uintptr_t namelen;
876 	int flags;
877 };
878 
879 static int
880 linux_accept4(struct thread *td, struct linux_accept4_args *args)
881 {
882 
883 	return (linux_accept_common(td, args->s, args->addr,
884 	    args->namelen, args->flags));
885 }
886 
887 struct linux_getsockname_args {
888 	int s;
889 	l_uintptr_t addr;
890 	l_uintptr_t namelen;
891 };
892 
893 static int
894 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
895 {
896 	struct getsockname_args /* {
897 		int	fdes;
898 		struct sockaddr * __restrict asa;
899 		socklen_t * __restrict alen;
900 	} */ bsd_args;
901 	int error;
902 
903 	bsd_args.fdes = args->s;
904 	/* XXX: */
905 	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
906 	bsd_args.alen = PTRIN(args->namelen);	/* XXX */
907 	error = sys_getsockname(td, &bsd_args);
908 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
909 	if (error)
910 		return (error);
911 	error = linux_sa_put(PTRIN(args->addr));
912 	if (error)
913 		return (error);
914 	return (0);
915 }
916 
917 struct linux_getpeername_args {
918 	int s;
919 	l_uintptr_t addr;
920 	l_uintptr_t namelen;
921 };
922 
923 static int
924 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
925 {
926 	struct getpeername_args /* {
927 		int fdes;
928 		caddr_t asa;
929 		int *alen;
930 	} */ bsd_args;
931 	int error;
932 
933 	bsd_args.fdes = args->s;
934 	bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
935 	bsd_args.alen = (int *)PTRIN(args->namelen);
936 	error = sys_getpeername(td, &bsd_args);
937 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
938 	if (error)
939 		return (error);
940 	error = linux_sa_put(PTRIN(args->addr));
941 	if (error)
942 		return (error);
943 	return (0);
944 }
945 
946 struct linux_socketpair_args {
947 	int domain;
948 	int type;
949 	int protocol;
950 	l_uintptr_t rsv;
951 };
952 
953 static int
954 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
955 {
956 	struct socketpair_args /* {
957 		int domain;
958 		int type;
959 		int protocol;
960 		int *rsv;
961 	} */ bsd_args;
962 	int error, socket_flags;
963 	int sv[2];
964 
965 	bsd_args.domain = linux_to_bsd_domain(args->domain);
966 	if (bsd_args.domain != PF_LOCAL)
967 		return (EAFNOSUPPORT);
968 
969 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
970 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
971 		return (EINVAL);
972 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
973 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
974 		return (EINVAL);
975 
976 	if (args->protocol != 0 && args->protocol != PF_UNIX)
977 
978 		/*
979 		 * Use of PF_UNIX as protocol argument is not right,
980 		 * but Linux does it.
981 		 * Do not map PF_UNIX as its Linux value is identical
982 		 * to FreeBSD one.
983 		 */
984 		return (EPROTONOSUPPORT);
985 	else
986 		bsd_args.protocol = 0;
987 	bsd_args.rsv = (int *)PTRIN(args->rsv);
988 	error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
989 	    bsd_args.protocol, sv);
990 	if (error)
991 		return (error);
992 	error = linux_set_socket_flags(td, sv[0], socket_flags);
993 	if (error)
994 		goto out;
995 	error = linux_set_socket_flags(td, sv[1], socket_flags);
996 	if (error)
997 		goto out;
998 
999 	error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
1000 
1001 out:
1002 	if (error) {
1003 		(void)kern_close(td, sv[0]);
1004 		(void)kern_close(td, sv[1]);
1005 	}
1006 	return (error);
1007 }
1008 
1009 struct linux_send_args {
1010 	int s;
1011 	l_uintptr_t msg;
1012 	int len;
1013 	int flags;
1014 };
1015 
1016 static int
1017 linux_send(struct thread *td, struct linux_send_args *args)
1018 {
1019 	struct sendto_args /* {
1020 		int s;
1021 		caddr_t buf;
1022 		int len;
1023 		int flags;
1024 		caddr_t to;
1025 		int tolen;
1026 	} */ bsd_args;
1027 
1028 	bsd_args.s = args->s;
1029 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1030 	bsd_args.len = args->len;
1031 	bsd_args.flags = args->flags;
1032 	bsd_args.to = NULL;
1033 	bsd_args.tolen = 0;
1034 	return sys_sendto(td, &bsd_args);
1035 }
1036 
1037 struct linux_recv_args {
1038 	int s;
1039 	l_uintptr_t msg;
1040 	int len;
1041 	int flags;
1042 };
1043 
1044 static int
1045 linux_recv(struct thread *td, struct linux_recv_args *args)
1046 {
1047 	struct recvfrom_args /* {
1048 		int s;
1049 		caddr_t buf;
1050 		int len;
1051 		int flags;
1052 		struct sockaddr *from;
1053 		socklen_t fromlenaddr;
1054 	} */ bsd_args;
1055 
1056 	bsd_args.s = args->s;
1057 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1058 	bsd_args.len = args->len;
1059 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1060 	bsd_args.from = NULL;
1061 	bsd_args.fromlenaddr = 0;
1062 	return (sys_recvfrom(td, &bsd_args));
1063 }
1064 
1065 static int
1066 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1067 {
1068 	struct msghdr msg;
1069 	struct iovec aiov;
1070 	int error;
1071 
1072 	if (linux_check_hdrincl(td, args->s) == 0)
1073 		/* IP_HDRINCL set, tweak the packet before sending */
1074 		return (linux_sendto_hdrincl(td, args));
1075 
1076 	msg.msg_name = PTRIN(args->to);
1077 	msg.msg_namelen = args->tolen;
1078 	msg.msg_iov = &aiov;
1079 	msg.msg_iovlen = 1;
1080 	msg.msg_control = NULL;
1081 	msg.msg_flags = 0;
1082 	aiov.iov_base = PTRIN(args->msg);
1083 	aiov.iov_len = args->len;
1084 	error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1085 	    UIO_USERSPACE);
1086 	return (error);
1087 }
1088 
1089 struct linux_recvfrom_args {
1090 	int s;
1091 	l_uintptr_t buf;
1092 	int len;
1093 	int flags;
1094 	l_uintptr_t from;
1095 	l_uintptr_t fromlen;
1096 };
1097 
1098 static int
1099 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1100 {
1101 	struct recvfrom_args /* {
1102 		int	s;
1103 		caddr_t	buf;
1104 		size_t	len;
1105 		int	flags;
1106 		struct sockaddr * __restrict from;
1107 		socklen_t * __restrict fromlenaddr;
1108 	} */ bsd_args;
1109 	size_t len;
1110 	int error;
1111 
1112 	if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1113 		return (error);
1114 
1115 	bsd_args.s = args->s;
1116 	bsd_args.buf = PTRIN(args->buf);
1117 	bsd_args.len = args->len;
1118 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1119 	/* XXX: */
1120 	bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1121 	bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1122 
1123 	linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1124 	error = sys_recvfrom(td, &bsd_args);
1125 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1126 
1127 	if (error)
1128 		return (error);
1129 	if (args->from) {
1130 		error = linux_sa_put((struct osockaddr *)
1131 		    PTRIN(args->from));
1132 		if (error)
1133 			return (error);
1134 	}
1135 	return (0);
1136 }
1137 
1138 struct linux_sendmsg_args {
1139 	int s;
1140 	l_uintptr_t msg;
1141 	int flags;
1142 };
1143 
1144 static int
1145 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1146 {
1147 	struct cmsghdr *cmsg;
1148 	struct cmsgcred cmcred;
1149 	struct mbuf *control;
1150 	struct msghdr msg;
1151 	struct l_cmsghdr linux_cmsg;
1152 	struct l_cmsghdr *ptr_cmsg;
1153 	struct l_msghdr linux_msg;
1154 	struct iovec *iov;
1155 	socklen_t datalen;
1156 	struct sockaddr *sa;
1157 	sa_family_t sa_family;
1158 	void *data;
1159 	int error;
1160 
1161 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1162 	if (error)
1163 		return (error);
1164 
1165 	/*
1166 	 * Some Linux applications (ping) define a non-NULL control data
1167 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1168 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1169 	 * order to handle this case.  This should be checked, but allows the
1170 	 * Linux ping to work.
1171 	 */
1172 	if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1173 		linux_msg.msg_control = PTROUT(NULL);
1174 
1175 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1176 	if (error)
1177 		return (error);
1178 
1179 #ifdef COMPAT_LINUX32
1180 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1181 	    &iov, EMSGSIZE);
1182 #else
1183 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1184 #endif
1185 	if (error)
1186 		return (error);
1187 
1188 	control = NULL;
1189 	cmsg = NULL;
1190 
1191 	if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1192 		error = kern_getsockname(td, args->s, &sa, &datalen);
1193 		if (error)
1194 			goto bad;
1195 		sa_family = sa->sa_family;
1196 		free(sa, M_SONAME);
1197 
1198 		error = ENOBUFS;
1199 		cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1200 		control = m_get(M_WAITOK, MT_CONTROL);
1201 		if (control == NULL)
1202 			goto bad;
1203 
1204 		do {
1205 			error = copyin(ptr_cmsg, &linux_cmsg,
1206 			    sizeof(struct l_cmsghdr));
1207 			if (error)
1208 				goto bad;
1209 
1210 			error = EINVAL;
1211 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1212 				goto bad;
1213 
1214 			/*
1215 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1216 			 * so return EINVAL in any other cmsg_type
1217 			 */
1218 			cmsg->cmsg_type =
1219 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1220 			cmsg->cmsg_level =
1221 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1222 			if (cmsg->cmsg_type == -1
1223 			    || cmsg->cmsg_level != SOL_SOCKET)
1224 				goto bad;
1225 
1226 			/*
1227 			 * Some applications (e.g. pulseaudio) attempt to
1228 			 * send ancillary data even if the underlying protocol
1229 			 * doesn't support it which is not allowed in the
1230 			 * FreeBSD system call interface.
1231 			 */
1232 			if (sa_family != AF_UNIX)
1233 				continue;
1234 
1235 			data = LINUX_CMSG_DATA(ptr_cmsg);
1236 			datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1237 
1238 			switch (cmsg->cmsg_type)
1239 			{
1240 			case SCM_RIGHTS:
1241 				break;
1242 
1243 			case SCM_CREDS:
1244 				data = &cmcred;
1245 				datalen = sizeof(cmcred);
1246 
1247 				/*
1248 				 * The lower levels will fill in the structure
1249 				 */
1250 				bzero(data, datalen);
1251 				break;
1252 			}
1253 
1254 			cmsg->cmsg_len = CMSG_LEN(datalen);
1255 
1256 			error = ENOBUFS;
1257 			if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1258 				goto bad;
1259 			if (!m_append(control, datalen, (c_caddr_t)data))
1260 				goto bad;
1261 		} while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1262 
1263 		if (m_length(control, NULL) == 0) {
1264 			m_freem(control);
1265 			control = NULL;
1266 		}
1267 	}
1268 
1269 	msg.msg_iov = iov;
1270 	msg.msg_flags = 0;
1271 	error = linux_sendit(td, args->s, &msg, args->flags, control,
1272 	    UIO_USERSPACE);
1273 
1274 bad:
1275 	free(iov, M_IOV);
1276 	if (cmsg)
1277 		free(cmsg, M_TEMP);
1278 	return (error);
1279 }
1280 
1281 struct linux_recvmsg_args {
1282 	int s;
1283 	l_uintptr_t msg;
1284 	int flags;
1285 };
1286 
1287 static int
1288 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1289 {
1290 	struct cmsghdr *cm;
1291 	struct cmsgcred *cmcred;
1292 	struct msghdr msg;
1293 	struct l_cmsghdr *linux_cmsg = NULL;
1294 	struct l_ucred linux_ucred;
1295 	socklen_t datalen, outlen;
1296 	struct l_msghdr linux_msg;
1297 	struct iovec *iov, *uiov;
1298 	struct mbuf *control = NULL;
1299 	struct mbuf **controlp;
1300 	caddr_t outbuf;
1301 	void *data;
1302 	int error, i, fd, fds, *fdp;
1303 
1304 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1305 	if (error)
1306 		return (error);
1307 
1308 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1309 	if (error)
1310 		return (error);
1311 
1312 #ifdef COMPAT_LINUX32
1313 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1314 	    &iov, EMSGSIZE);
1315 #else
1316 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1317 #endif
1318 	if (error)
1319 		return (error);
1320 
1321 	if (msg.msg_name) {
1322 		error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1323 		    msg.msg_namelen);
1324 		if (error)
1325 			goto bad;
1326 	}
1327 
1328 	uiov = msg.msg_iov;
1329 	msg.msg_iov = iov;
1330 	controlp = (msg.msg_control != NULL) ? &control : NULL;
1331 	error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1332 	msg.msg_iov = uiov;
1333 	if (error)
1334 		goto bad;
1335 
1336 	error = bsd_to_linux_msghdr(&msg, &linux_msg);
1337 	if (error)
1338 		goto bad;
1339 
1340 	if (linux_msg.msg_name) {
1341 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1342 		    PTRIN(linux_msg.msg_name));
1343 		if (error)
1344 			goto bad;
1345 	}
1346 	if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1347 		error = linux_sa_put(PTRIN(linux_msg.msg_name));
1348 		if (error)
1349 			goto bad;
1350 	}
1351 
1352 	outbuf = PTRIN(linux_msg.msg_control);
1353 	outlen = 0;
1354 
1355 	if (control) {
1356 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1357 
1358 		msg.msg_control = mtod(control, struct cmsghdr *);
1359 		msg.msg_controllen = control->m_len;
1360 
1361 		cm = CMSG_FIRSTHDR(&msg);
1362 
1363 		while (cm != NULL) {
1364 			linux_cmsg->cmsg_type =
1365 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1366 			linux_cmsg->cmsg_level =
1367 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1368 			if (linux_cmsg->cmsg_type == -1
1369 			    || cm->cmsg_level != SOL_SOCKET)
1370 			{
1371 				error = EINVAL;
1372 				goto bad;
1373 			}
1374 
1375 			data = CMSG_DATA(cm);
1376 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1377 
1378 			switch (cm->cmsg_type)
1379 			{
1380 			case SCM_RIGHTS:
1381 				if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1382 					fds = datalen / sizeof(int);
1383 					fdp = data;
1384 					for (i = 0; i < fds; i++) {
1385 						fd = *fdp++;
1386 						(void)kern_fcntl(td, fd,
1387 						    F_SETFD, FD_CLOEXEC);
1388 					}
1389 				}
1390 				break;
1391 
1392 			case SCM_CREDS:
1393 				/*
1394 				 * Currently LOCAL_CREDS is never in
1395 				 * effect for Linux so no need to worry
1396 				 * about sockcred
1397 				 */
1398 				if (datalen != sizeof(*cmcred)) {
1399 					error = EMSGSIZE;
1400 					goto bad;
1401 				}
1402 				cmcred = (struct cmsgcred *)data;
1403 				bzero(&linux_ucred, sizeof(linux_ucred));
1404 				linux_ucred.pid = cmcred->cmcred_pid;
1405 				linux_ucred.uid = cmcred->cmcred_uid;
1406 				linux_ucred.gid = cmcred->cmcred_gid;
1407 				data = &linux_ucred;
1408 				datalen = sizeof(linux_ucred);
1409 				break;
1410 			}
1411 
1412 			if (outlen + LINUX_CMSG_LEN(datalen) >
1413 			    linux_msg.msg_controllen) {
1414 				if (outlen == 0) {
1415 					error = EMSGSIZE;
1416 					goto bad;
1417 				} else {
1418 					linux_msg.msg_flags |=
1419 					    LINUX_MSG_CTRUNC;
1420 					goto out;
1421 				}
1422 			}
1423 
1424 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1425 
1426 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1427 			if (error)
1428 				goto bad;
1429 			outbuf += L_CMSG_HDRSZ;
1430 
1431 			error = copyout(data, outbuf, datalen);
1432 			if (error)
1433 				goto bad;
1434 
1435 			outbuf += LINUX_CMSG_ALIGN(datalen);
1436 			outlen += LINUX_CMSG_LEN(datalen);
1437 
1438 			cm = CMSG_NXTHDR(&msg, cm);
1439 		}
1440 	}
1441 
1442 out:
1443 	linux_msg.msg_controllen = outlen;
1444 	error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1445 
1446 bad:
1447 	free(iov, M_IOV);
1448 	m_freem(control);
1449 	free(linux_cmsg, M_TEMP);
1450 
1451 	return (error);
1452 }
1453 
1454 struct linux_shutdown_args {
1455 	int s;
1456 	int how;
1457 };
1458 
1459 static int
1460 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1461 {
1462 	struct shutdown_args /* {
1463 		int s;
1464 		int how;
1465 	} */ bsd_args;
1466 
1467 	bsd_args.s = args->s;
1468 	bsd_args.how = args->how;
1469 	return (sys_shutdown(td, &bsd_args));
1470 }
1471 
1472 struct linux_setsockopt_args {
1473 	int s;
1474 	int level;
1475 	int optname;
1476 	l_uintptr_t optval;
1477 	int optlen;
1478 };
1479 
1480 static int
1481 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1482 {
1483 	struct setsockopt_args /* {
1484 		int s;
1485 		int level;
1486 		int name;
1487 		caddr_t val;
1488 		int valsize;
1489 	} */ bsd_args;
1490 	l_timeval linux_tv;
1491 	struct timeval tv;
1492 	int error, name;
1493 
1494 	bsd_args.s = args->s;
1495 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1496 	switch (bsd_args.level) {
1497 	case SOL_SOCKET:
1498 		name = linux_to_bsd_so_sockopt(args->optname);
1499 		switch (name) {
1500 		case SO_RCVTIMEO:
1501 			/* FALLTHROUGH */
1502 		case SO_SNDTIMEO:
1503 			error = copyin(PTRIN(args->optval), &linux_tv,
1504 			    sizeof(linux_tv));
1505 			if (error)
1506 				return (error);
1507 			tv.tv_sec = linux_tv.tv_sec;
1508 			tv.tv_usec = linux_tv.tv_usec;
1509 			return (kern_setsockopt(td, args->s, bsd_args.level,
1510 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1511 			/* NOTREACHED */
1512 			break;
1513 		default:
1514 			break;
1515 		}
1516 		break;
1517 	case IPPROTO_IP:
1518 		name = linux_to_bsd_ip_sockopt(args->optname);
1519 		break;
1520 	case IPPROTO_TCP:
1521 		name = linux_to_bsd_tcp_sockopt(args->optname);
1522 		break;
1523 	default:
1524 		name = -1;
1525 		break;
1526 	}
1527 	if (name == -1)
1528 		return (ENOPROTOOPT);
1529 
1530 	bsd_args.name = name;
1531 	bsd_args.val = PTRIN(args->optval);
1532 	bsd_args.valsize = args->optlen;
1533 
1534 	if (name == IPV6_NEXTHOP) {
1535 		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1536 			bsd_args.valsize);
1537 		error = sys_setsockopt(td, &bsd_args);
1538 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1539 	} else
1540 		error = sys_setsockopt(td, &bsd_args);
1541 
1542 	return (error);
1543 }
1544 
1545 struct linux_getsockopt_args {
1546 	int s;
1547 	int level;
1548 	int optname;
1549 	l_uintptr_t optval;
1550 	l_uintptr_t optlen;
1551 };
1552 
1553 static int
1554 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1555 {
1556 	struct getsockopt_args /* {
1557 		int s;
1558 		int level;
1559 		int name;
1560 		caddr_t val;
1561 		int *avalsize;
1562 	} */ bsd_args;
1563 	l_timeval linux_tv;
1564 	struct timeval tv;
1565 	socklen_t tv_len, xulen;
1566 	struct xucred xu;
1567 	struct l_ucred lxu;
1568 	int error, name;
1569 
1570 	bsd_args.s = args->s;
1571 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1572 	switch (bsd_args.level) {
1573 	case SOL_SOCKET:
1574 		name = linux_to_bsd_so_sockopt(args->optname);
1575 		switch (name) {
1576 		case SO_RCVTIMEO:
1577 			/* FALLTHROUGH */
1578 		case SO_SNDTIMEO:
1579 			tv_len = sizeof(tv);
1580 			error = kern_getsockopt(td, args->s, bsd_args.level,
1581 			    name, &tv, UIO_SYSSPACE, &tv_len);
1582 			if (error)
1583 				return (error);
1584 			linux_tv.tv_sec = tv.tv_sec;
1585 			linux_tv.tv_usec = tv.tv_usec;
1586 			return (copyout(&linux_tv, PTRIN(args->optval),
1587 			    sizeof(linux_tv)));
1588 			/* NOTREACHED */
1589 			break;
1590 		case LOCAL_PEERCRED:
1591 			if (args->optlen != sizeof(lxu))
1592 				return (EINVAL);
1593 			xulen = sizeof(xu);
1594 			error = kern_getsockopt(td, args->s, bsd_args.level,
1595 			    name, &xu, UIO_SYSSPACE, &xulen);
1596 			if (error)
1597 				return (error);
1598 			/*
1599 			 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1600 			 */
1601 			lxu.pid = 0;
1602 			lxu.uid = xu.cr_uid;
1603 			lxu.gid = xu.cr_gid;
1604 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1605 			/* NOTREACHED */
1606 			break;
1607 		default:
1608 			break;
1609 		}
1610 		break;
1611 	case IPPROTO_IP:
1612 		name = linux_to_bsd_ip_sockopt(args->optname);
1613 		break;
1614 	case IPPROTO_TCP:
1615 		name = linux_to_bsd_tcp_sockopt(args->optname);
1616 		break;
1617 	default:
1618 		name = -1;
1619 		break;
1620 	}
1621 	if (name == -1)
1622 		return (EINVAL);
1623 
1624 	bsd_args.name = name;
1625 	bsd_args.val = PTRIN(args->optval);
1626 	bsd_args.avalsize = PTRIN(args->optlen);
1627 
1628 	if (name == IPV6_NEXTHOP) {
1629 		error = sys_getsockopt(td, &bsd_args);
1630 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1631 	} else
1632 		error = sys_getsockopt(td, &bsd_args);
1633 
1634 	return (error);
1635 }
1636 
1637 /* Argument list sizes for linux_socketcall */
1638 
1639 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1640 
1641 static const unsigned char lxs_args[] = {
1642 	LINUX_AL(0) /* unused*/,	LINUX_AL(3) /* socket */,
1643 	LINUX_AL(3) /* bind */,		LINUX_AL(3) /* connect */,
1644 	LINUX_AL(2) /* listen */,	LINUX_AL(3) /* accept */,
1645 	LINUX_AL(3) /* getsockname */,	LINUX_AL(3) /* getpeername */,
1646 	LINUX_AL(4) /* socketpair */,	LINUX_AL(4) /* send */,
1647 	LINUX_AL(4) /* recv */,		LINUX_AL(6) /* sendto */,
1648 	LINUX_AL(6) /* recvfrom */,	LINUX_AL(2) /* shutdown */,
1649 	LINUX_AL(5) /* setsockopt */,	LINUX_AL(5) /* getsockopt */,
1650 	LINUX_AL(3) /* sendmsg */,	LINUX_AL(3) /* recvmsg */,
1651 	LINUX_AL(4) /* accept4 */
1652 };
1653 
1654 #define	LINUX_AL_SIZE	sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1655 
1656 int
1657 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1658 {
1659 	l_ulong a[6];
1660 	void *arg;
1661 	int error;
1662 
1663 	if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1664 		return (EINVAL);
1665 	error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1666 	if (error)
1667 		return (error);
1668 
1669 	arg = a;
1670 	switch (args->what) {
1671 	case LINUX_SOCKET:
1672 		return (linux_socket(td, arg));
1673 	case LINUX_BIND:
1674 		return (linux_bind(td, arg));
1675 	case LINUX_CONNECT:
1676 		return (linux_connect(td, arg));
1677 	case LINUX_LISTEN:
1678 		return (linux_listen(td, arg));
1679 	case LINUX_ACCEPT:
1680 		return (linux_accept(td, arg));
1681 	case LINUX_GETSOCKNAME:
1682 		return (linux_getsockname(td, arg));
1683 	case LINUX_GETPEERNAME:
1684 		return (linux_getpeername(td, arg));
1685 	case LINUX_SOCKETPAIR:
1686 		return (linux_socketpair(td, arg));
1687 	case LINUX_SEND:
1688 		return (linux_send(td, arg));
1689 	case LINUX_RECV:
1690 		return (linux_recv(td, arg));
1691 	case LINUX_SENDTO:
1692 		return (linux_sendto(td, arg));
1693 	case LINUX_RECVFROM:
1694 		return (linux_recvfrom(td, arg));
1695 	case LINUX_SHUTDOWN:
1696 		return (linux_shutdown(td, arg));
1697 	case LINUX_SETSOCKOPT:
1698 		return (linux_setsockopt(td, arg));
1699 	case LINUX_GETSOCKOPT:
1700 		return (linux_getsockopt(td, arg));
1701 	case LINUX_SENDMSG:
1702 		return (linux_sendmsg(td, arg));
1703 	case LINUX_RECVMSG:
1704 		return (linux_recvmsg(td, arg));
1705 	case LINUX_ACCEPT4:
1706 		return (linux_accept4(td, arg));
1707 	}
1708 
1709 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1710 	return (ENOSYS);
1711 }
1712