xref: /freebsd/sys/compat/linux/linux_socket.c (revision ec273ebf3b6aed5fba8c56b6ece5ad8693a48ea7)
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;
577 	socklen_t size_val;
578 
579 	size_val = sizeof(optval);
580 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
581 	    &optval, UIO_SYSSPACE, &size_val);
582 	if (error)
583 		return (error);
584 
585 	return (optval == 0);
586 }
587 
588 struct linux_sendto_args {
589 	int s;
590 	l_uintptr_t msg;
591 	int len;
592 	int flags;
593 	l_uintptr_t to;
594 	int tolen;
595 };
596 
597 /*
598  * Updated sendto() when IP_HDRINCL is set:
599  * tweak endian-dependent fields in the IP packet.
600  */
601 static int
602 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
603 {
604 /*
605  * linux_ip_copysize defines how many bytes we should copy
606  * from the beginning of the IP packet before we customize it for BSD.
607  * It should include all the fields we modify (ip_len and ip_off).
608  */
609 #define linux_ip_copysize	8
610 
611 	struct ip *packet;
612 	struct msghdr msg;
613 	struct iovec aiov[1];
614 	int error;
615 
616 	/* Check that the packet isn't too big or too small. */
617 	if (linux_args->len < linux_ip_copysize ||
618 	    linux_args->len > IP_MAXPACKET)
619 		return (EINVAL);
620 
621 	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
622 
623 	/* Make kernel copy of the packet to be sent */
624 	if ((error = copyin(PTRIN(linux_args->msg), packet,
625 	    linux_args->len)))
626 		goto goout;
627 
628 	/* Convert fields from Linux to BSD raw IP socket format */
629 	packet->ip_len = linux_args->len;
630 	packet->ip_off = ntohs(packet->ip_off);
631 
632 	/* Prepare the msghdr and iovec structures describing the new packet */
633 	msg.msg_name = PTRIN(linux_args->to);
634 	msg.msg_namelen = linux_args->tolen;
635 	msg.msg_iov = aiov;
636 	msg.msg_iovlen = 1;
637 	msg.msg_control = NULL;
638 	msg.msg_flags = 0;
639 	aiov[0].iov_base = (char *)packet;
640 	aiov[0].iov_len = linux_args->len;
641 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
642 	    NULL, UIO_SYSSPACE);
643 goout:
644 	free(packet, M_TEMP);
645 	return (error);
646 }
647 
648 struct linux_socket_args {
649 	int domain;
650 	int type;
651 	int protocol;
652 };
653 
654 static int
655 linux_socket(struct thread *td, struct linux_socket_args *args)
656 {
657 	struct socket_args /* {
658 		int domain;
659 		int type;
660 		int protocol;
661 	} */ bsd_args;
662 	int retval_socket, socket_flags;
663 
664 	bsd_args.protocol = args->protocol;
665 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
666 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
667 		return (EINVAL);
668 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
669 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
670 		return (EINVAL);
671 	bsd_args.domain = linux_to_bsd_domain(args->domain);
672 	if (bsd_args.domain == -1)
673 		return (EAFNOSUPPORT);
674 
675 	retval_socket = sys_socket(td, &bsd_args);
676 	if (retval_socket)
677 		return (retval_socket);
678 
679 	retval_socket = linux_set_socket_flags(td, td->td_retval[0],
680 	    socket_flags);
681 	if (retval_socket) {
682 		(void)kern_close(td, td->td_retval[0]);
683 		goto out;
684 	}
685 
686 	if (bsd_args.type == SOCK_RAW
687 	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
688 	    && bsd_args.domain == PF_INET) {
689 		/* It's a raw IP socket: set the IP_HDRINCL option. */
690 		int hdrincl;
691 
692 		hdrincl = 1;
693 		/* We ignore any error returned by kern_setsockopt() */
694 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
695 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
696 	}
697 #ifdef INET6
698 	/*
699 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
700 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
701 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
702 	 * sysctl value.
703 	 */
704 	if (bsd_args.domain == PF_INET6) {
705 		int v6only;
706 
707 		v6only = 0;
708 		/* We ignore any error returned by setsockopt() */
709 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
710 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
711 	}
712 #endif
713 
714 out:
715 	return (retval_socket);
716 }
717 
718 struct linux_bind_args {
719 	int s;
720 	l_uintptr_t name;
721 	int namelen;
722 };
723 
724 static int
725 linux_bind(struct thread *td, struct linux_bind_args *args)
726 {
727 	struct sockaddr *sa;
728 	int error;
729 
730 	error = linux_getsockaddr(&sa, PTRIN(args->name),
731 	    args->namelen);
732 	if (error)
733 		return (error);
734 
735 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
736 	free(sa, M_SONAME);
737 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
738 	   	return (EINVAL);
739 	return (error);
740 }
741 
742 struct linux_connect_args {
743 	int s;
744 	l_uintptr_t name;
745 	int namelen;
746 };
747 int linux_connect(struct thread *, struct linux_connect_args *);
748 
749 int
750 linux_connect(struct thread *td, struct linux_connect_args *args)
751 {
752 	cap_rights_t rights;
753 	struct socket *so;
754 	struct sockaddr *sa;
755 	u_int fflag;
756 	int error;
757 
758 	error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
759 	    args->namelen);
760 	if (error)
761 		return (error);
762 
763 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
764 	free(sa, M_SONAME);
765 	if (error != EISCONN)
766 		return (error);
767 
768 	/*
769 	 * Linux doesn't return EISCONN the first time it occurs,
770 	 * when on a non-blocking socket. Instead it returns the
771 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
772 	 *
773 	 * XXXRW: Instead of using fgetsock(), check that it is a
774 	 * socket and use the file descriptor reference instead of
775 	 * creating a new one.
776 	 */
777 	error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
778 	    &so, &fflag);
779 	if (error == 0) {
780 		error = EISCONN;
781 		if (fflag & FNONBLOCK) {
782 			SOCK_LOCK(so);
783 			if (so->so_emuldata == 0)
784 				error = so->so_error;
785 			so->so_emuldata = (void *)1;
786 			SOCK_UNLOCK(so);
787 		}
788 		fputsock(so);
789 	}
790 	return (error);
791 }
792 
793 struct linux_listen_args {
794 	int s;
795 	int backlog;
796 };
797 
798 static int
799 linux_listen(struct thread *td, struct linux_listen_args *args)
800 {
801 	struct listen_args /* {
802 		int s;
803 		int backlog;
804 	} */ bsd_args;
805 
806 	bsd_args.s = args->s;
807 	bsd_args.backlog = args->backlog;
808 	return (sys_listen(td, &bsd_args));
809 }
810 
811 static int
812 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
813     l_uintptr_t namelen, int flags)
814 {
815 	struct accept_args /* {
816 		int	s;
817 		struct sockaddr * __restrict name;
818 		socklen_t * __restrict anamelen;
819 	} */ bsd_args;
820 	int error;
821 
822 	if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
823 		return (EINVAL);
824 
825 	bsd_args.s = s;
826 	/* XXX: */
827 	bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
828 	bsd_args.anamelen = PTRIN(namelen);/* XXX */
829 	error = sys_accept(td, &bsd_args);
830 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
831 	if (error) {
832 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
833 			return (EINVAL);
834 		return (error);
835 	}
836 
837 	/*
838 	 * linux appears not to copy flags from the parent socket to the
839 	 * accepted one, so we must clear the flags in the new descriptor
840 	 * and apply the requested flags.
841 	 */
842 	error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
843 	if (error)
844 		goto out;
845 	error = linux_set_socket_flags(td, td->td_retval[0], flags);
846 	if (error)
847 		goto out;
848 	if (addr)
849 		error = linux_sa_put(PTRIN(addr));
850 
851 out:
852 	if (error) {
853 		(void)kern_close(td, td->td_retval[0]);
854 		td->td_retval[0] = 0;
855 	}
856 	return (error);
857 }
858 
859 struct linux_accept_args {
860 	int s;
861 	l_uintptr_t addr;
862 	l_uintptr_t namelen;
863 };
864 
865 static int
866 linux_accept(struct thread *td, struct linux_accept_args *args)
867 {
868 
869 	return (linux_accept_common(td, args->s, args->addr,
870 	    args->namelen, 0));
871 }
872 
873 struct linux_accept4_args {
874 	int s;
875 	l_uintptr_t addr;
876 	l_uintptr_t namelen;
877 	int flags;
878 };
879 
880 static int
881 linux_accept4(struct thread *td, struct linux_accept4_args *args)
882 {
883 
884 	return (linux_accept_common(td, args->s, args->addr,
885 	    args->namelen, args->flags));
886 }
887 
888 struct linux_getsockname_args {
889 	int s;
890 	l_uintptr_t addr;
891 	l_uintptr_t namelen;
892 };
893 
894 static int
895 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
896 {
897 	struct getsockname_args /* {
898 		int	fdes;
899 		struct sockaddr * __restrict asa;
900 		socklen_t * __restrict alen;
901 	} */ bsd_args;
902 	int error;
903 
904 	bsd_args.fdes = args->s;
905 	/* XXX: */
906 	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
907 	bsd_args.alen = PTRIN(args->namelen);	/* XXX */
908 	error = sys_getsockname(td, &bsd_args);
909 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
910 	if (error)
911 		return (error);
912 	error = linux_sa_put(PTRIN(args->addr));
913 	if (error)
914 		return (error);
915 	return (0);
916 }
917 
918 struct linux_getpeername_args {
919 	int s;
920 	l_uintptr_t addr;
921 	l_uintptr_t namelen;
922 };
923 
924 static int
925 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
926 {
927 	struct getpeername_args /* {
928 		int fdes;
929 		caddr_t asa;
930 		int *alen;
931 	} */ bsd_args;
932 	int error;
933 
934 	bsd_args.fdes = args->s;
935 	bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
936 	bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
937 	error = sys_getpeername(td, &bsd_args);
938 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
939 	if (error)
940 		return (error);
941 	error = linux_sa_put(PTRIN(args->addr));
942 	if (error)
943 		return (error);
944 	return (0);
945 }
946 
947 struct linux_socketpair_args {
948 	int domain;
949 	int type;
950 	int protocol;
951 	l_uintptr_t rsv;
952 };
953 
954 static int
955 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
956 {
957 	struct socketpair_args /* {
958 		int domain;
959 		int type;
960 		int protocol;
961 		int *rsv;
962 	} */ bsd_args;
963 	int error, socket_flags;
964 	int sv[2];
965 
966 	bsd_args.domain = linux_to_bsd_domain(args->domain);
967 	if (bsd_args.domain != PF_LOCAL)
968 		return (EAFNOSUPPORT);
969 
970 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
971 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
972 		return (EINVAL);
973 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
974 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
975 		return (EINVAL);
976 
977 	if (args->protocol != 0 && args->protocol != PF_UNIX)
978 
979 		/*
980 		 * Use of PF_UNIX as protocol argument is not right,
981 		 * but Linux does it.
982 		 * Do not map PF_UNIX as its Linux value is identical
983 		 * to FreeBSD one.
984 		 */
985 		return (EPROTONOSUPPORT);
986 	else
987 		bsd_args.protocol = 0;
988 	bsd_args.rsv = (int *)PTRIN(args->rsv);
989 	error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
990 	    bsd_args.protocol, sv);
991 	if (error)
992 		return (error);
993 	error = linux_set_socket_flags(td, sv[0], socket_flags);
994 	if (error)
995 		goto out;
996 	error = linux_set_socket_flags(td, sv[1], socket_flags);
997 	if (error)
998 		goto out;
999 
1000 	error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
1001 
1002 out:
1003 	if (error) {
1004 		(void)kern_close(td, sv[0]);
1005 		(void)kern_close(td, sv[1]);
1006 	}
1007 	return (error);
1008 }
1009 
1010 struct linux_send_args {
1011 	int s;
1012 	l_uintptr_t msg;
1013 	int len;
1014 	int flags;
1015 };
1016 
1017 static int
1018 linux_send(struct thread *td, struct linux_send_args *args)
1019 {
1020 	struct sendto_args /* {
1021 		int s;
1022 		caddr_t buf;
1023 		int len;
1024 		int flags;
1025 		caddr_t to;
1026 		int tolen;
1027 	} */ bsd_args;
1028 
1029 	bsd_args.s = args->s;
1030 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1031 	bsd_args.len = args->len;
1032 	bsd_args.flags = args->flags;
1033 	bsd_args.to = NULL;
1034 	bsd_args.tolen = 0;
1035 	return sys_sendto(td, &bsd_args);
1036 }
1037 
1038 struct linux_recv_args {
1039 	int s;
1040 	l_uintptr_t msg;
1041 	int len;
1042 	int flags;
1043 };
1044 
1045 static int
1046 linux_recv(struct thread *td, struct linux_recv_args *args)
1047 {
1048 	struct recvfrom_args /* {
1049 		int s;
1050 		caddr_t buf;
1051 		int len;
1052 		int flags;
1053 		struct sockaddr *from;
1054 		socklen_t fromlenaddr;
1055 	} */ bsd_args;
1056 
1057 	bsd_args.s = args->s;
1058 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1059 	bsd_args.len = args->len;
1060 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1061 	bsd_args.from = NULL;
1062 	bsd_args.fromlenaddr = 0;
1063 	return (sys_recvfrom(td, &bsd_args));
1064 }
1065 
1066 static int
1067 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1068 {
1069 	struct msghdr msg;
1070 	struct iovec aiov;
1071 	int error;
1072 
1073 	if (linux_check_hdrincl(td, args->s) == 0)
1074 		/* IP_HDRINCL set, tweak the packet before sending */
1075 		return (linux_sendto_hdrincl(td, args));
1076 
1077 	msg.msg_name = PTRIN(args->to);
1078 	msg.msg_namelen = args->tolen;
1079 	msg.msg_iov = &aiov;
1080 	msg.msg_iovlen = 1;
1081 	msg.msg_control = NULL;
1082 	msg.msg_flags = 0;
1083 	aiov.iov_base = PTRIN(args->msg);
1084 	aiov.iov_len = args->len;
1085 	error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1086 	    UIO_USERSPACE);
1087 	return (error);
1088 }
1089 
1090 struct linux_recvfrom_args {
1091 	int s;
1092 	l_uintptr_t buf;
1093 	int len;
1094 	int flags;
1095 	l_uintptr_t from;
1096 	l_uintptr_t fromlen;
1097 };
1098 
1099 static int
1100 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1101 {
1102 	struct recvfrom_args /* {
1103 		int	s;
1104 		caddr_t	buf;
1105 		size_t	len;
1106 		int	flags;
1107 		struct sockaddr * __restrict from;
1108 		socklen_t * __restrict fromlenaddr;
1109 	} */ bsd_args;
1110 	size_t len;
1111 	int error;
1112 
1113 	if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1114 		return (error);
1115 
1116 	bsd_args.s = args->s;
1117 	bsd_args.buf = PTRIN(args->buf);
1118 	bsd_args.len = args->len;
1119 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1120 	/* XXX: */
1121 	bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1122 	bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1123 
1124 	linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1125 	error = sys_recvfrom(td, &bsd_args);
1126 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1127 
1128 	if (error)
1129 		return (error);
1130 	if (args->from) {
1131 		error = linux_sa_put((struct osockaddr *)
1132 		    PTRIN(args->from));
1133 		if (error)
1134 			return (error);
1135 	}
1136 	return (0);
1137 }
1138 
1139 struct linux_sendmsg_args {
1140 	int s;
1141 	l_uintptr_t msg;
1142 	int flags;
1143 };
1144 
1145 static int
1146 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1147 {
1148 	struct cmsghdr *cmsg;
1149 	struct cmsgcred cmcred;
1150 	struct mbuf *control;
1151 	struct msghdr msg;
1152 	struct l_cmsghdr linux_cmsg;
1153 	struct l_cmsghdr *ptr_cmsg;
1154 	struct l_msghdr linux_msg;
1155 	struct iovec *iov;
1156 	socklen_t datalen;
1157 	struct sockaddr *sa;
1158 	sa_family_t sa_family;
1159 	void *data;
1160 	int error;
1161 
1162 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1163 	if (error)
1164 		return (error);
1165 
1166 	/*
1167 	 * Some Linux applications (ping) define a non-NULL control data
1168 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1169 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1170 	 * order to handle this case.  This should be checked, but allows the
1171 	 * Linux ping to work.
1172 	 */
1173 	if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1174 		linux_msg.msg_control = PTROUT(NULL);
1175 
1176 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1177 	if (error)
1178 		return (error);
1179 
1180 #ifdef COMPAT_LINUX32
1181 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1182 	    &iov, EMSGSIZE);
1183 #else
1184 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1185 #endif
1186 	if (error)
1187 		return (error);
1188 
1189 	control = NULL;
1190 	cmsg = NULL;
1191 
1192 	if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1193 		error = kern_getsockname(td, args->s, &sa, &datalen);
1194 		if (error)
1195 			goto bad;
1196 		sa_family = sa->sa_family;
1197 		free(sa, M_SONAME);
1198 
1199 		error = ENOBUFS;
1200 		cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1201 		control = m_get(M_WAITOK, MT_CONTROL);
1202 		if (control == NULL)
1203 			goto bad;
1204 
1205 		do {
1206 			error = copyin(ptr_cmsg, &linux_cmsg,
1207 			    sizeof(struct l_cmsghdr));
1208 			if (error)
1209 				goto bad;
1210 
1211 			error = EINVAL;
1212 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1213 				goto bad;
1214 
1215 			/*
1216 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1217 			 * so return EINVAL in any other cmsg_type
1218 			 */
1219 			cmsg->cmsg_type =
1220 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1221 			cmsg->cmsg_level =
1222 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1223 			if (cmsg->cmsg_type == -1
1224 			    || cmsg->cmsg_level != SOL_SOCKET)
1225 				goto bad;
1226 
1227 			/*
1228 			 * Some applications (e.g. pulseaudio) attempt to
1229 			 * send ancillary data even if the underlying protocol
1230 			 * doesn't support it which is not allowed in the
1231 			 * FreeBSD system call interface.
1232 			 */
1233 			if (sa_family != AF_UNIX)
1234 				continue;
1235 
1236 			data = LINUX_CMSG_DATA(ptr_cmsg);
1237 			datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1238 
1239 			switch (cmsg->cmsg_type)
1240 			{
1241 			case SCM_RIGHTS:
1242 				break;
1243 
1244 			case SCM_CREDS:
1245 				data = &cmcred;
1246 				datalen = sizeof(cmcred);
1247 
1248 				/*
1249 				 * The lower levels will fill in the structure
1250 				 */
1251 				bzero(data, datalen);
1252 				break;
1253 			}
1254 
1255 			cmsg->cmsg_len = CMSG_LEN(datalen);
1256 
1257 			error = ENOBUFS;
1258 			if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1259 				goto bad;
1260 			if (!m_append(control, datalen, (c_caddr_t)data))
1261 				goto bad;
1262 		} while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1263 
1264 		if (m_length(control, NULL) == 0) {
1265 			m_freem(control);
1266 			control = NULL;
1267 		}
1268 	}
1269 
1270 	msg.msg_iov = iov;
1271 	msg.msg_flags = 0;
1272 	error = linux_sendit(td, args->s, &msg, args->flags, control,
1273 	    UIO_USERSPACE);
1274 
1275 bad:
1276 	free(iov, M_IOV);
1277 	if (cmsg)
1278 		free(cmsg, M_TEMP);
1279 	return (error);
1280 }
1281 
1282 struct linux_recvmsg_args {
1283 	int s;
1284 	l_uintptr_t msg;
1285 	int flags;
1286 };
1287 
1288 static int
1289 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1290 {
1291 	struct cmsghdr *cm;
1292 	struct cmsgcred *cmcred;
1293 	struct msghdr msg;
1294 	struct l_cmsghdr *linux_cmsg = NULL;
1295 	struct l_ucred linux_ucred;
1296 	socklen_t datalen, outlen;
1297 	struct l_msghdr linux_msg;
1298 	struct iovec *iov, *uiov;
1299 	struct mbuf *control = NULL;
1300 	struct mbuf **controlp;
1301 	caddr_t outbuf;
1302 	void *data;
1303 	int error, i, fd, fds, *fdp;
1304 
1305 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1306 	if (error)
1307 		return (error);
1308 
1309 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1310 	if (error)
1311 		return (error);
1312 
1313 #ifdef COMPAT_LINUX32
1314 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1315 	    &iov, EMSGSIZE);
1316 #else
1317 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1318 #endif
1319 	if (error)
1320 		return (error);
1321 
1322 	if (msg.msg_name) {
1323 		error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1324 		    msg.msg_namelen);
1325 		if (error)
1326 			goto bad;
1327 	}
1328 
1329 	uiov = msg.msg_iov;
1330 	msg.msg_iov = iov;
1331 	controlp = (msg.msg_control != NULL) ? &control : NULL;
1332 	error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1333 	msg.msg_iov = uiov;
1334 	if (error)
1335 		goto bad;
1336 
1337 	error = bsd_to_linux_msghdr(&msg, &linux_msg);
1338 	if (error)
1339 		goto bad;
1340 
1341 	if (linux_msg.msg_name) {
1342 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1343 		    PTRIN(linux_msg.msg_name));
1344 		if (error)
1345 			goto bad;
1346 	}
1347 	if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1348 		error = linux_sa_put(PTRIN(linux_msg.msg_name));
1349 		if (error)
1350 			goto bad;
1351 	}
1352 
1353 	outbuf = PTRIN(linux_msg.msg_control);
1354 	outlen = 0;
1355 
1356 	if (control) {
1357 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1358 
1359 		msg.msg_control = mtod(control, struct cmsghdr *);
1360 		msg.msg_controllen = control->m_len;
1361 
1362 		cm = CMSG_FIRSTHDR(&msg);
1363 
1364 		while (cm != NULL) {
1365 			linux_cmsg->cmsg_type =
1366 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1367 			linux_cmsg->cmsg_level =
1368 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1369 			if (linux_cmsg->cmsg_type == -1
1370 			    || cm->cmsg_level != SOL_SOCKET)
1371 			{
1372 				error = EINVAL;
1373 				goto bad;
1374 			}
1375 
1376 			data = CMSG_DATA(cm);
1377 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1378 
1379 			switch (cm->cmsg_type)
1380 			{
1381 			case SCM_RIGHTS:
1382 				if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1383 					fds = datalen / sizeof(int);
1384 					fdp = data;
1385 					for (i = 0; i < fds; i++) {
1386 						fd = *fdp++;
1387 						(void)kern_fcntl(td, fd,
1388 						    F_SETFD, FD_CLOEXEC);
1389 					}
1390 				}
1391 				break;
1392 
1393 			case SCM_CREDS:
1394 				/*
1395 				 * Currently LOCAL_CREDS is never in
1396 				 * effect for Linux so no need to worry
1397 				 * about sockcred
1398 				 */
1399 				if (datalen != sizeof(*cmcred)) {
1400 					error = EMSGSIZE;
1401 					goto bad;
1402 				}
1403 				cmcred = (struct cmsgcred *)data;
1404 				bzero(&linux_ucred, sizeof(linux_ucred));
1405 				linux_ucred.pid = cmcred->cmcred_pid;
1406 				linux_ucred.uid = cmcred->cmcred_uid;
1407 				linux_ucred.gid = cmcred->cmcred_gid;
1408 				data = &linux_ucred;
1409 				datalen = sizeof(linux_ucred);
1410 				break;
1411 			}
1412 
1413 			if (outlen + LINUX_CMSG_LEN(datalen) >
1414 			    linux_msg.msg_controllen) {
1415 				if (outlen == 0) {
1416 					error = EMSGSIZE;
1417 					goto bad;
1418 				} else {
1419 					linux_msg.msg_flags |=
1420 					    LINUX_MSG_CTRUNC;
1421 					goto out;
1422 				}
1423 			}
1424 
1425 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1426 
1427 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1428 			if (error)
1429 				goto bad;
1430 			outbuf += L_CMSG_HDRSZ;
1431 
1432 			error = copyout(data, outbuf, datalen);
1433 			if (error)
1434 				goto bad;
1435 
1436 			outbuf += LINUX_CMSG_ALIGN(datalen);
1437 			outlen += LINUX_CMSG_LEN(datalen);
1438 
1439 			cm = CMSG_NXTHDR(&msg, cm);
1440 		}
1441 	}
1442 
1443 out:
1444 	linux_msg.msg_controllen = outlen;
1445 	error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1446 
1447 bad:
1448 	free(iov, M_IOV);
1449 	m_freem(control);
1450 	free(linux_cmsg, M_TEMP);
1451 
1452 	return (error);
1453 }
1454 
1455 struct linux_shutdown_args {
1456 	int s;
1457 	int how;
1458 };
1459 
1460 static int
1461 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1462 {
1463 	struct shutdown_args /* {
1464 		int s;
1465 		int how;
1466 	} */ bsd_args;
1467 
1468 	bsd_args.s = args->s;
1469 	bsd_args.how = args->how;
1470 	return (sys_shutdown(td, &bsd_args));
1471 }
1472 
1473 struct linux_setsockopt_args {
1474 	int s;
1475 	int level;
1476 	int optname;
1477 	l_uintptr_t optval;
1478 	int optlen;
1479 };
1480 
1481 static int
1482 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1483 {
1484 	struct setsockopt_args /* {
1485 		int s;
1486 		int level;
1487 		int name;
1488 		caddr_t val;
1489 		int valsize;
1490 	} */ bsd_args;
1491 	l_timeval linux_tv;
1492 	struct timeval tv;
1493 	int error, name;
1494 
1495 	bsd_args.s = args->s;
1496 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1497 	switch (bsd_args.level) {
1498 	case SOL_SOCKET:
1499 		name = linux_to_bsd_so_sockopt(args->optname);
1500 		switch (name) {
1501 		case SO_RCVTIMEO:
1502 			/* FALLTHROUGH */
1503 		case SO_SNDTIMEO:
1504 			error = copyin(PTRIN(args->optval), &linux_tv,
1505 			    sizeof(linux_tv));
1506 			if (error)
1507 				return (error);
1508 			tv.tv_sec = linux_tv.tv_sec;
1509 			tv.tv_usec = linux_tv.tv_usec;
1510 			return (kern_setsockopt(td, args->s, bsd_args.level,
1511 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1512 			/* NOTREACHED */
1513 			break;
1514 		default:
1515 			break;
1516 		}
1517 		break;
1518 	case IPPROTO_IP:
1519 		name = linux_to_bsd_ip_sockopt(args->optname);
1520 		break;
1521 	case IPPROTO_TCP:
1522 		name = linux_to_bsd_tcp_sockopt(args->optname);
1523 		break;
1524 	default:
1525 		name = -1;
1526 		break;
1527 	}
1528 	if (name == -1)
1529 		return (ENOPROTOOPT);
1530 
1531 	bsd_args.name = name;
1532 	bsd_args.val = PTRIN(args->optval);
1533 	bsd_args.valsize = args->optlen;
1534 
1535 	if (name == IPV6_NEXTHOP) {
1536 		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1537 			bsd_args.valsize);
1538 		error = sys_setsockopt(td, &bsd_args);
1539 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1540 	} else
1541 		error = sys_setsockopt(td, &bsd_args);
1542 
1543 	return (error);
1544 }
1545 
1546 struct linux_getsockopt_args {
1547 	int s;
1548 	int level;
1549 	int optname;
1550 	l_uintptr_t optval;
1551 	l_uintptr_t optlen;
1552 };
1553 
1554 static int
1555 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1556 {
1557 	struct getsockopt_args /* {
1558 		int s;
1559 		int level;
1560 		int name;
1561 		caddr_t val;
1562 		int *avalsize;
1563 	} */ bsd_args;
1564 	l_timeval linux_tv;
1565 	struct timeval tv;
1566 	socklen_t tv_len, xulen;
1567 	struct xucred xu;
1568 	struct l_ucred lxu;
1569 	int error, name;
1570 
1571 	bsd_args.s = args->s;
1572 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1573 	switch (bsd_args.level) {
1574 	case SOL_SOCKET:
1575 		name = linux_to_bsd_so_sockopt(args->optname);
1576 		switch (name) {
1577 		case SO_RCVTIMEO:
1578 			/* FALLTHROUGH */
1579 		case SO_SNDTIMEO:
1580 			tv_len = sizeof(tv);
1581 			error = kern_getsockopt(td, args->s, bsd_args.level,
1582 			    name, &tv, UIO_SYSSPACE, &tv_len);
1583 			if (error)
1584 				return (error);
1585 			linux_tv.tv_sec = tv.tv_sec;
1586 			linux_tv.tv_usec = tv.tv_usec;
1587 			return (copyout(&linux_tv, PTRIN(args->optval),
1588 			    sizeof(linux_tv)));
1589 			/* NOTREACHED */
1590 			break;
1591 		case LOCAL_PEERCRED:
1592 			if (args->optlen != sizeof(lxu))
1593 				return (EINVAL);
1594 			xulen = sizeof(xu);
1595 			error = kern_getsockopt(td, args->s, bsd_args.level,
1596 			    name, &xu, UIO_SYSSPACE, &xulen);
1597 			if (error)
1598 				return (error);
1599 			/*
1600 			 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1601 			 */
1602 			lxu.pid = 0;
1603 			lxu.uid = xu.cr_uid;
1604 			lxu.gid = xu.cr_gid;
1605 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1606 			/* NOTREACHED */
1607 			break;
1608 		default:
1609 			break;
1610 		}
1611 		break;
1612 	case IPPROTO_IP:
1613 		name = linux_to_bsd_ip_sockopt(args->optname);
1614 		break;
1615 	case IPPROTO_TCP:
1616 		name = linux_to_bsd_tcp_sockopt(args->optname);
1617 		break;
1618 	default:
1619 		name = -1;
1620 		break;
1621 	}
1622 	if (name == -1)
1623 		return (EINVAL);
1624 
1625 	bsd_args.name = name;
1626 	bsd_args.val = PTRIN(args->optval);
1627 	bsd_args.avalsize = PTRIN(args->optlen);
1628 
1629 	if (name == IPV6_NEXTHOP) {
1630 		error = sys_getsockopt(td, &bsd_args);
1631 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1632 	} else
1633 		error = sys_getsockopt(td, &bsd_args);
1634 
1635 	return (error);
1636 }
1637 
1638 /* Argument list sizes for linux_socketcall */
1639 
1640 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1641 
1642 static const unsigned char lxs_args[] = {
1643 	LINUX_AL(0) /* unused*/,	LINUX_AL(3) /* socket */,
1644 	LINUX_AL(3) /* bind */,		LINUX_AL(3) /* connect */,
1645 	LINUX_AL(2) /* listen */,	LINUX_AL(3) /* accept */,
1646 	LINUX_AL(3) /* getsockname */,	LINUX_AL(3) /* getpeername */,
1647 	LINUX_AL(4) /* socketpair */,	LINUX_AL(4) /* send */,
1648 	LINUX_AL(4) /* recv */,		LINUX_AL(6) /* sendto */,
1649 	LINUX_AL(6) /* recvfrom */,	LINUX_AL(2) /* shutdown */,
1650 	LINUX_AL(5) /* setsockopt */,	LINUX_AL(5) /* getsockopt */,
1651 	LINUX_AL(3) /* sendmsg */,	LINUX_AL(3) /* recvmsg */,
1652 	LINUX_AL(4) /* accept4 */
1653 };
1654 
1655 #define	LINUX_AL_SIZE	sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1656 
1657 int
1658 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1659 {
1660 	l_ulong a[6];
1661 	void *arg;
1662 	int error;
1663 
1664 	if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1665 		return (EINVAL);
1666 	error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1667 	if (error)
1668 		return (error);
1669 
1670 	arg = a;
1671 	switch (args->what) {
1672 	case LINUX_SOCKET:
1673 		return (linux_socket(td, arg));
1674 	case LINUX_BIND:
1675 		return (linux_bind(td, arg));
1676 	case LINUX_CONNECT:
1677 		return (linux_connect(td, arg));
1678 	case LINUX_LISTEN:
1679 		return (linux_listen(td, arg));
1680 	case LINUX_ACCEPT:
1681 		return (linux_accept(td, arg));
1682 	case LINUX_GETSOCKNAME:
1683 		return (linux_getsockname(td, arg));
1684 	case LINUX_GETPEERNAME:
1685 		return (linux_getpeername(td, arg));
1686 	case LINUX_SOCKETPAIR:
1687 		return (linux_socketpair(td, arg));
1688 	case LINUX_SEND:
1689 		return (linux_send(td, arg));
1690 	case LINUX_RECV:
1691 		return (linux_recv(td, arg));
1692 	case LINUX_SENDTO:
1693 		return (linux_sendto(td, arg));
1694 	case LINUX_RECVFROM:
1695 		return (linux_recvfrom(td, arg));
1696 	case LINUX_SHUTDOWN:
1697 		return (linux_shutdown(td, arg));
1698 	case LINUX_SETSOCKOPT:
1699 		return (linux_setsockopt(td, arg));
1700 	case LINUX_GETSOCKOPT:
1701 		return (linux_getsockopt(td, arg));
1702 	case LINUX_SENDMSG:
1703 		return (linux_sendmsg(td, arg));
1704 	case LINUX_RECVMSG:
1705 		return (linux_recvmsg(td, arg));
1706 	case LINUX_ACCEPT4:
1707 		return (linux_accept4(td, arg));
1708 	}
1709 
1710 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1711 	return (ENOSYS);
1712 }
1713