xref: /freebsd/sys/compat/linux/linux_socket.c (revision 4ab7403bbd76f466d482cd69035091c04e409c09)
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_file.h>
74 #include <compat/linux/linux_socket.h>
75 #include <compat/linux/linux_util.h>
76 
77 static int linux_to_bsd_domain(int);
78 
79 /*
80  * Reads a linux sockaddr and does any necessary translation.
81  * Linux sockaddrs don't have a length field, only a family.
82  * Copy the osockaddr structure pointed to by osa to kernel, adjust
83  * family and convert to sockaddr.
84  */
85 static int
86 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
87 {
88 	struct sockaddr *sa;
89 	struct osockaddr *kosa;
90 #ifdef INET6
91 	struct sockaddr_in6 *sin6;
92 	int oldv6size;
93 #endif
94 	char *name;
95 	int bdom, error, hdrlen, namelen;
96 
97 	if (salen < 2 || salen > UCHAR_MAX || !osa)
98 		return (EINVAL);
99 
100 #ifdef INET6
101 	oldv6size = 0;
102 	/*
103 	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
104 	 * if it's a v4-mapped address, so reserve the proper space
105 	 * for it.
106 	 */
107 	if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
108 		salen += sizeof(uint32_t);
109 		oldv6size = 1;
110 	}
111 #endif
112 
113 	kosa = malloc(salen, M_SONAME, M_WAITOK);
114 
115 	if ((error = copyin(osa, kosa, salen)))
116 		goto out;
117 
118 	bdom = linux_to_bsd_domain(kosa->sa_family);
119 	if (bdom == -1) {
120 		error = EAFNOSUPPORT;
121 		goto out;
122 	}
123 
124 #ifdef INET6
125 	/*
126 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
127 	 * which lacks the scope id compared with RFC2553 one. If we detect
128 	 * the situation, reject the address and write a message to system log.
129 	 *
130 	 * Still accept addresses for which the scope id is not used.
131 	 */
132 	if (oldv6size) {
133 		if (bdom == AF_INET6) {
134 			sin6 = (struct sockaddr_in6 *)kosa;
135 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
136 			    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
137 			     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
138 			     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
139 			     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
140 			     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
141 				sin6->sin6_scope_id = 0;
142 			} else {
143 				log(LOG_DEBUG,
144 				    "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
145 				error = EINVAL;
146 				goto out;
147 			}
148 		} else
149 			salen -= sizeof(uint32_t);
150 	}
151 #endif
152 	if (bdom == AF_INET) {
153 		if (salen < sizeof(struct sockaddr_in)) {
154 			error = EINVAL;
155 			goto out;
156 		}
157 		salen = sizeof(struct sockaddr_in);
158 	}
159 
160 	if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
161 		hdrlen = offsetof(struct sockaddr_un, sun_path);
162 		name = ((struct sockaddr_un *)kosa)->sun_path;
163 		if (*name == '\0') {
164 			/*
165 		 	 * Linux abstract namespace starts with a NULL byte.
166 			 * XXX We do not support abstract namespace yet.
167 			 */
168 			namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
169 		} else
170 			namelen = strnlen(name, salen - hdrlen);
171 		salen = hdrlen + namelen;
172 		if (salen > sizeof(struct sockaddr_un)) {
173 			error = ENAMETOOLONG;
174 			goto out;
175 		}
176 	}
177 
178 	sa = (struct sockaddr *)kosa;
179 	sa->sa_family = bdom;
180 	sa->sa_len = salen;
181 
182 	*sap = sa;
183 	return (0);
184 
185 out:
186 	free(kosa, M_SONAME);
187 	return (error);
188 }
189 
190 static int
191 linux_to_bsd_domain(int domain)
192 {
193 
194 	switch (domain) {
195 	case LINUX_AF_UNSPEC:
196 		return (AF_UNSPEC);
197 	case LINUX_AF_UNIX:
198 		return (AF_LOCAL);
199 	case LINUX_AF_INET:
200 		return (AF_INET);
201 	case LINUX_AF_INET6:
202 		return (AF_INET6);
203 	case LINUX_AF_AX25:
204 		return (AF_CCITT);
205 	case LINUX_AF_IPX:
206 		return (AF_IPX);
207 	case LINUX_AF_APPLETALK:
208 		return (AF_APPLETALK);
209 	}
210 	return (-1);
211 }
212 
213 static int
214 bsd_to_linux_domain(int domain)
215 {
216 
217 	switch (domain) {
218 	case AF_UNSPEC:
219 		return (LINUX_AF_UNSPEC);
220 	case AF_LOCAL:
221 		return (LINUX_AF_UNIX);
222 	case AF_INET:
223 		return (LINUX_AF_INET);
224 	case AF_INET6:
225 		return (LINUX_AF_INET6);
226 	case AF_CCITT:
227 		return (LINUX_AF_AX25);
228 	case AF_IPX:
229 		return (LINUX_AF_IPX);
230 	case AF_APPLETALK:
231 		return (LINUX_AF_APPLETALK);
232 	}
233 	return (-1);
234 }
235 
236 static int
237 linux_to_bsd_sockopt_level(int level)
238 {
239 
240 	switch (level) {
241 	case LINUX_SOL_SOCKET:
242 		return (SOL_SOCKET);
243 	}
244 	return (level);
245 }
246 
247 static int
248 bsd_to_linux_sockopt_level(int level)
249 {
250 
251 	switch (level) {
252 	case SOL_SOCKET:
253 		return (LINUX_SOL_SOCKET);
254 	}
255 	return (level);
256 }
257 
258 static int
259 linux_to_bsd_ip_sockopt(int opt)
260 {
261 
262 	switch (opt) {
263 	case LINUX_IP_TOS:
264 		return (IP_TOS);
265 	case LINUX_IP_TTL:
266 		return (IP_TTL);
267 	case LINUX_IP_OPTIONS:
268 		return (IP_OPTIONS);
269 	case LINUX_IP_MULTICAST_IF:
270 		return (IP_MULTICAST_IF);
271 	case LINUX_IP_MULTICAST_TTL:
272 		return (IP_MULTICAST_TTL);
273 	case LINUX_IP_MULTICAST_LOOP:
274 		return (IP_MULTICAST_LOOP);
275 	case LINUX_IP_ADD_MEMBERSHIP:
276 		return (IP_ADD_MEMBERSHIP);
277 	case LINUX_IP_DROP_MEMBERSHIP:
278 		return (IP_DROP_MEMBERSHIP);
279 	case LINUX_IP_HDRINCL:
280 		return (IP_HDRINCL);
281 	}
282 	return (-1);
283 }
284 
285 static int
286 linux_to_bsd_so_sockopt(int opt)
287 {
288 
289 	switch (opt) {
290 	case LINUX_SO_DEBUG:
291 		return (SO_DEBUG);
292 	case LINUX_SO_REUSEADDR:
293 		return (SO_REUSEADDR);
294 	case LINUX_SO_TYPE:
295 		return (SO_TYPE);
296 	case LINUX_SO_ERROR:
297 		return (SO_ERROR);
298 	case LINUX_SO_DONTROUTE:
299 		return (SO_DONTROUTE);
300 	case LINUX_SO_BROADCAST:
301 		return (SO_BROADCAST);
302 	case LINUX_SO_SNDBUF:
303 		return (SO_SNDBUF);
304 	case LINUX_SO_RCVBUF:
305 		return (SO_RCVBUF);
306 	case LINUX_SO_KEEPALIVE:
307 		return (SO_KEEPALIVE);
308 	case LINUX_SO_OOBINLINE:
309 		return (SO_OOBINLINE);
310 	case LINUX_SO_LINGER:
311 		return (SO_LINGER);
312 	case LINUX_SO_PEERCRED:
313 		return (LOCAL_PEERCRED);
314 	case LINUX_SO_RCVLOWAT:
315 		return (SO_RCVLOWAT);
316 	case LINUX_SO_SNDLOWAT:
317 		return (SO_SNDLOWAT);
318 	case LINUX_SO_RCVTIMEO:
319 		return (SO_RCVTIMEO);
320 	case LINUX_SO_SNDTIMEO:
321 		return (SO_SNDTIMEO);
322 	case LINUX_SO_TIMESTAMP:
323 		return (SO_TIMESTAMP);
324 	case LINUX_SO_ACCEPTCONN:
325 		return (SO_ACCEPTCONN);
326 	}
327 	return (-1);
328 }
329 
330 static int
331 linux_to_bsd_tcp_sockopt(int opt)
332 {
333 
334 	switch (opt) {
335 	case LINUX_TCP_NODELAY:
336 		return (TCP_NODELAY);
337 	case LINUX_TCP_MAXSEG:
338 		return (TCP_MAXSEG);
339 	case LINUX_TCP_KEEPIDLE:
340 		return (TCP_KEEPIDLE);
341 	case LINUX_TCP_KEEPINTVL:
342 		return (TCP_KEEPINTVL);
343 	case LINUX_TCP_KEEPCNT:
344 		return (TCP_KEEPCNT);
345 	case LINUX_TCP_MD5SIG:
346 		return (TCP_MD5SIG);
347 	}
348 	return (-1);
349 }
350 
351 static int
352 linux_to_bsd_msg_flags(int flags)
353 {
354 	int ret_flags = 0;
355 
356 	if (flags & LINUX_MSG_OOB)
357 		ret_flags |= MSG_OOB;
358 	if (flags & LINUX_MSG_PEEK)
359 		ret_flags |= MSG_PEEK;
360 	if (flags & LINUX_MSG_DONTROUTE)
361 		ret_flags |= MSG_DONTROUTE;
362 	if (flags & LINUX_MSG_CTRUNC)
363 		ret_flags |= MSG_CTRUNC;
364 	if (flags & LINUX_MSG_TRUNC)
365 		ret_flags |= MSG_TRUNC;
366 	if (flags & LINUX_MSG_DONTWAIT)
367 		ret_flags |= MSG_DONTWAIT;
368 	if (flags & LINUX_MSG_EOR)
369 		ret_flags |= MSG_EOR;
370 	if (flags & LINUX_MSG_WAITALL)
371 		ret_flags |= MSG_WAITALL;
372 	if (flags & LINUX_MSG_NOSIGNAL)
373 		ret_flags |= MSG_NOSIGNAL;
374 #if 0 /* not handled */
375 	if (flags & LINUX_MSG_PROXY)
376 		;
377 	if (flags & LINUX_MSG_FIN)
378 		;
379 	if (flags & LINUX_MSG_SYN)
380 		;
381 	if (flags & LINUX_MSG_CONFIRM)
382 		;
383 	if (flags & LINUX_MSG_RST)
384 		;
385 	if (flags & LINUX_MSG_ERRQUEUE)
386 		;
387 #endif
388 	return ret_flags;
389 }
390 
391 /*
392 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
393 * native syscall will fault.  Thus, we don't really need to check the
394 * return values for these functions.
395 */
396 
397 static int
398 bsd_to_linux_sockaddr(struct sockaddr *arg)
399 {
400 	struct sockaddr sa;
401 	size_t sa_len = sizeof(struct sockaddr);
402 	int error;
403 
404 	if ((error = copyin(arg, &sa, sa_len)))
405 		return (error);
406 
407 	*(u_short *)&sa = sa.sa_family;
408 
409 	error = copyout(&sa, arg, sa_len);
410 
411 	return (error);
412 }
413 
414 static int
415 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
416 {
417 	struct sockaddr sa;
418 	size_t sa_len = sizeof(struct sockaddr);
419 	int error;
420 
421 	if ((error = copyin(arg, &sa, sa_len)))
422 		return (error);
423 
424 	sa.sa_family = *(sa_family_t *)&sa;
425 	sa.sa_len = len;
426 
427 	error = copyout(&sa, arg, sa_len);
428 
429 	return (error);
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 /*
589  * Updated sendto() when IP_HDRINCL is set:
590  * tweak endian-dependent fields in the IP packet.
591  */
592 static int
593 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
594 {
595 /*
596  * linux_ip_copysize defines how many bytes we should copy
597  * from the beginning of the IP packet before we customize it for BSD.
598  * It should include all the fields we modify (ip_len and ip_off).
599  */
600 #define linux_ip_copysize	8
601 
602 	struct ip *packet;
603 	struct msghdr msg;
604 	struct iovec aiov[1];
605 	int error;
606 
607 	/* Check that the packet isn't too big or too small. */
608 	if (linux_args->len < linux_ip_copysize ||
609 	    linux_args->len > IP_MAXPACKET)
610 		return (EINVAL);
611 
612 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
613 
614 	/* Make kernel copy of the packet to be sent */
615 	if ((error = copyin(PTRIN(linux_args->msg), packet,
616 	    linux_args->len)))
617 		goto goout;
618 
619 	/* Convert fields from Linux to BSD raw IP socket format */
620 	packet->ip_len = linux_args->len;
621 	packet->ip_off = ntohs(packet->ip_off);
622 
623 	/* Prepare the msghdr and iovec structures describing the new packet */
624 	msg.msg_name = PTRIN(linux_args->to);
625 	msg.msg_namelen = linux_args->tolen;
626 	msg.msg_iov = aiov;
627 	msg.msg_iovlen = 1;
628 	msg.msg_control = NULL;
629 	msg.msg_flags = 0;
630 	aiov[0].iov_base = (char *)packet;
631 	aiov[0].iov_len = linux_args->len;
632 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
633 	    NULL, UIO_SYSSPACE);
634 goout:
635 	free(packet, M_LINUX);
636 	return (error);
637 }
638 
639 int
640 linux_socket(struct thread *td, struct linux_socket_args *args)
641 {
642 	struct socket_args /* {
643 		int domain;
644 		int type;
645 		int protocol;
646 	} */ bsd_args;
647 	int retval_socket, socket_flags;
648 
649 	bsd_args.protocol = args->protocol;
650 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
651 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
652 		return (EINVAL);
653 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
654 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
655 		return (EINVAL);
656 	bsd_args.domain = linux_to_bsd_domain(args->domain);
657 	if (bsd_args.domain == -1)
658 		return (EAFNOSUPPORT);
659 
660 	retval_socket = sys_socket(td, &bsd_args);
661 	if (retval_socket)
662 		return (retval_socket);
663 
664 	retval_socket = linux_set_socket_flags(td, td->td_retval[0],
665 	    socket_flags);
666 	if (retval_socket) {
667 		(void)kern_close(td, td->td_retval[0]);
668 		goto out;
669 	}
670 
671 	if (bsd_args.type == SOCK_RAW
672 	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
673 	    && bsd_args.domain == PF_INET) {
674 		/* It's a raw IP socket: set the IP_HDRINCL option. */
675 		int hdrincl;
676 
677 		hdrincl = 1;
678 		/* We ignore any error returned by kern_setsockopt() */
679 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
680 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
681 	}
682 #ifdef INET6
683 	/*
684 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
685 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
686 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
687 	 * sysctl value.
688 	 */
689 	if (bsd_args.domain == PF_INET6) {
690 		int v6only;
691 
692 		v6only = 0;
693 		/* We ignore any error returned by setsockopt() */
694 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
695 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
696 	}
697 #endif
698 
699 out:
700 	return (retval_socket);
701 }
702 
703 int
704 linux_bind(struct thread *td, struct linux_bind_args *args)
705 {
706 	struct sockaddr *sa;
707 	int error;
708 
709 	error = linux_getsockaddr(&sa, PTRIN(args->name),
710 	    args->namelen);
711 	if (error)
712 		return (error);
713 
714 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
715 	free(sa, M_SONAME);
716 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
717 	   	return (EINVAL);
718 	return (error);
719 }
720 
721 int
722 linux_connect(struct thread *td, struct linux_connect_args *args)
723 {
724 	cap_rights_t rights;
725 	struct socket *so;
726 	struct sockaddr *sa;
727 	u_int fflag;
728 	int error;
729 
730 	error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
731 	    args->namelen);
732 	if (error)
733 		return (error);
734 
735 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
736 	free(sa, M_SONAME);
737 	if (error != EISCONN)
738 		return (error);
739 
740 	/*
741 	 * Linux doesn't return EISCONN the first time it occurs,
742 	 * when on a non-blocking socket. Instead it returns the
743 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
744 	 *
745 	 * XXXRW: Instead of using fgetsock(), check that it is a
746 	 * socket and use the file descriptor reference instead of
747 	 * creating a new one.
748 	 */
749 	error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
750 	    &so, &fflag);
751 	if (error == 0) {
752 		error = EISCONN;
753 		if (fflag & FNONBLOCK) {
754 			SOCK_LOCK(so);
755 			if (so->so_emuldata == 0)
756 				error = so->so_error;
757 			so->so_emuldata = (void *)1;
758 			SOCK_UNLOCK(so);
759 		}
760 		fputsock(so);
761 	}
762 	return (error);
763 }
764 
765 int
766 linux_listen(struct thread *td, struct linux_listen_args *args)
767 {
768 	struct listen_args /* {
769 		int s;
770 		int backlog;
771 	} */ bsd_args;
772 
773 	bsd_args.s = args->s;
774 	bsd_args.backlog = args->backlog;
775 	return (sys_listen(td, &bsd_args));
776 }
777 
778 static int
779 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
780     l_uintptr_t namelen, int flags)
781 {
782 	struct accept_args /* {
783 		int	s;
784 		struct sockaddr * __restrict name;
785 		socklen_t * __restrict anamelen;
786 	} */ bsd_args;
787 	int error;
788 
789 	if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
790 		return (EINVAL);
791 
792 	bsd_args.s = s;
793 	/* XXX: */
794 	bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
795 	bsd_args.anamelen = PTRIN(namelen);/* XXX */
796 	error = sys_accept(td, &bsd_args);
797 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
798 	if (error) {
799 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
800 			return (EINVAL);
801 		return (error);
802 	}
803 
804 	/*
805 	 * linux appears not to copy flags from the parent socket to the
806 	 * accepted one, so we must clear the flags in the new descriptor
807 	 * and apply the requested flags.
808 	 */
809 	error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
810 	if (error)
811 		goto out;
812 	error = linux_set_socket_flags(td, td->td_retval[0], flags);
813 	if (error)
814 		goto out;
815 	if (addr)
816 		error = linux_sa_put(PTRIN(addr));
817 
818 out:
819 	if (error) {
820 		(void)kern_close(td, td->td_retval[0]);
821 		td->td_retval[0] = 0;
822 	}
823 	return (error);
824 }
825 
826 int
827 linux_accept(struct thread *td, struct linux_accept_args *args)
828 {
829 
830 	return (linux_accept_common(td, args->s, args->addr,
831 	    args->namelen, 0));
832 }
833 
834 int
835 linux_accept4(struct thread *td, struct linux_accept4_args *args)
836 {
837 
838 	return (linux_accept_common(td, args->s, args->addr,
839 	    args->namelen, args->flags));
840 }
841 
842 int
843 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
844 {
845 	struct getsockname_args /* {
846 		int	fdes;
847 		struct sockaddr * __restrict asa;
848 		socklen_t * __restrict alen;
849 	} */ bsd_args;
850 	int error;
851 
852 	bsd_args.fdes = args->s;
853 	/* XXX: */
854 	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
855 	bsd_args.alen = PTRIN(args->namelen);	/* XXX */
856 	error = sys_getsockname(td, &bsd_args);
857 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
858 	if (error)
859 		return (error);
860 	error = linux_sa_put(PTRIN(args->addr));
861 	if (error)
862 		return (error);
863 	return (0);
864 }
865 
866 int
867 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
868 {
869 	struct getpeername_args /* {
870 		int fdes;
871 		caddr_t asa;
872 		int *alen;
873 	} */ bsd_args;
874 	int error;
875 
876 	bsd_args.fdes = args->s;
877 	bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
878 	bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
879 	error = sys_getpeername(td, &bsd_args);
880 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
881 	if (error)
882 		return (error);
883 	error = linux_sa_put(PTRIN(args->addr));
884 	if (error)
885 		return (error);
886 	return (0);
887 }
888 
889 int
890 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
891 {
892 	struct socketpair_args /* {
893 		int domain;
894 		int type;
895 		int protocol;
896 		int *rsv;
897 	} */ bsd_args;
898 	int error, socket_flags;
899 	int sv[2];
900 
901 	bsd_args.domain = linux_to_bsd_domain(args->domain);
902 	if (bsd_args.domain != PF_LOCAL)
903 		return (EAFNOSUPPORT);
904 
905 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
906 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
907 		return (EINVAL);
908 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
909 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
910 		return (EINVAL);
911 
912 	if (args->protocol != 0 && args->protocol != PF_UNIX)
913 
914 		/*
915 		 * Use of PF_UNIX as protocol argument is not right,
916 		 * but Linux does it.
917 		 * Do not map PF_UNIX as its Linux value is identical
918 		 * to FreeBSD one.
919 		 */
920 		return (EPROTONOSUPPORT);
921 	else
922 		bsd_args.protocol = 0;
923 	bsd_args.rsv = (int *)PTRIN(args->rsv);
924 	error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
925 	    bsd_args.protocol, sv);
926 	if (error)
927 		return (error);
928 	error = linux_set_socket_flags(td, sv[0], socket_flags);
929 	if (error)
930 		goto out;
931 	error = linux_set_socket_flags(td, sv[1], socket_flags);
932 	if (error)
933 		goto out;
934 
935 	error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
936 
937 out:
938 	if (error) {
939 		(void)kern_close(td, sv[0]);
940 		(void)kern_close(td, sv[1]);
941 	}
942 	return (error);
943 }
944 
945 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
946 struct linux_send_args {
947 	int s;
948 	l_uintptr_t msg;
949 	int len;
950 	int flags;
951 };
952 
953 static int
954 linux_send(struct thread *td, struct linux_send_args *args)
955 {
956 	struct sendto_args /* {
957 		int s;
958 		caddr_t buf;
959 		int len;
960 		int flags;
961 		caddr_t to;
962 		int tolen;
963 	} */ bsd_args;
964 
965 	bsd_args.s = args->s;
966 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
967 	bsd_args.len = args->len;
968 	bsd_args.flags = args->flags;
969 	bsd_args.to = NULL;
970 	bsd_args.tolen = 0;
971 	return sys_sendto(td, &bsd_args);
972 }
973 
974 struct linux_recv_args {
975 	int s;
976 	l_uintptr_t msg;
977 	int len;
978 	int flags;
979 };
980 
981 static int
982 linux_recv(struct thread *td, struct linux_recv_args *args)
983 {
984 	struct recvfrom_args /* {
985 		int s;
986 		caddr_t buf;
987 		int len;
988 		int flags;
989 		struct sockaddr *from;
990 		socklen_t fromlenaddr;
991 	} */ bsd_args;
992 
993 	bsd_args.s = args->s;
994 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
995 	bsd_args.len = args->len;
996 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
997 	bsd_args.from = NULL;
998 	bsd_args.fromlenaddr = 0;
999 	return (sys_recvfrom(td, &bsd_args));
1000 }
1001 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1002 
1003 int
1004 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1005 {
1006 	struct msghdr msg;
1007 	struct iovec aiov;
1008 	int error;
1009 
1010 	if (linux_check_hdrincl(td, args->s) == 0)
1011 		/* IP_HDRINCL set, tweak the packet before sending */
1012 		return (linux_sendto_hdrincl(td, args));
1013 
1014 	msg.msg_name = PTRIN(args->to);
1015 	msg.msg_namelen = args->tolen;
1016 	msg.msg_iov = &aiov;
1017 	msg.msg_iovlen = 1;
1018 	msg.msg_control = NULL;
1019 	msg.msg_flags = 0;
1020 	aiov.iov_base = PTRIN(args->msg);
1021 	aiov.iov_len = args->len;
1022 	error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1023 	    UIO_USERSPACE);
1024 	return (error);
1025 }
1026 
1027 int
1028 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1029 {
1030 	struct msghdr msg;
1031 	struct iovec aiov;
1032 	int error;
1033 
1034 	if (PTRIN(args->fromlen) != NULL) {
1035 		error = copyin(PTRIN(args->fromlen), &msg.msg_namelen,
1036 		    sizeof(msg.msg_namelen));
1037 		if (error != 0)
1038 			return (error);
1039 
1040 		error = linux_to_bsd_sockaddr((struct sockaddr *)PTRIN(args->from),
1041 		    msg.msg_namelen);
1042 		if (error != 0)
1043 			return (error);
1044 	} else
1045 		msg.msg_namelen = 0;
1046 
1047 	msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from);
1048 	msg.msg_iov = &aiov;
1049 	msg.msg_iovlen = 1;
1050 	aiov.iov_base = PTRIN(args->buf);
1051 	aiov.iov_len = args->len;
1052 	msg.msg_control = 0;
1053 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1054 
1055 	error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL);
1056 	if (error != 0)
1057 		return (error);
1058 
1059 	if (PTRIN(args->from) != NULL) {
1060 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1061 		    PTRIN(args->from));
1062 		if (error != 0)
1063 			return (error);
1064 
1065 		error = linux_sa_put((struct osockaddr *)
1066 		    PTRIN(args->from));
1067 	}
1068 
1069 	if (PTRIN(args->fromlen) != NULL)
1070 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1071 		    sizeof(msg.msg_namelen));
1072 
1073 	return (error);
1074 }
1075 
1076 int
1077 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1078 {
1079 	struct cmsghdr *cmsg;
1080 	struct cmsgcred cmcred;
1081 	struct mbuf *control;
1082 	struct msghdr msg;
1083 	struct l_cmsghdr linux_cmsg;
1084 	struct l_cmsghdr *ptr_cmsg;
1085 	struct l_msghdr linux_msg;
1086 	struct iovec *iov;
1087 	socklen_t datalen;
1088 	struct sockaddr *sa;
1089 	sa_family_t sa_family;
1090 	void *data;
1091 	int error;
1092 
1093 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1094 	if (error)
1095 		return (error);
1096 
1097 	/*
1098 	 * Some Linux applications (ping) define a non-NULL control data
1099 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1100 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1101 	 * order to handle this case.  This should be checked, but allows the
1102 	 * Linux ping to work.
1103 	 */
1104 	if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1105 		linux_msg.msg_control = PTROUT(NULL);
1106 
1107 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1108 	if (error)
1109 		return (error);
1110 
1111 #ifdef COMPAT_LINUX32
1112 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1113 	    &iov, EMSGSIZE);
1114 #else
1115 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1116 #endif
1117 	if (error)
1118 		return (error);
1119 
1120 	control = NULL;
1121 	cmsg = NULL;
1122 
1123 	if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1124 		error = kern_getsockname(td, args->s, &sa, &datalen);
1125 		if (error)
1126 			goto bad;
1127 		sa_family = sa->sa_family;
1128 		free(sa, M_SONAME);
1129 
1130 		error = ENOBUFS;
1131 		cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1132 		control = m_get(M_WAITOK, MT_CONTROL);
1133 		if (control == NULL)
1134 			goto bad;
1135 
1136 		do {
1137 			error = copyin(ptr_cmsg, &linux_cmsg,
1138 			    sizeof(struct l_cmsghdr));
1139 			if (error)
1140 				goto bad;
1141 
1142 			error = EINVAL;
1143 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1144 				goto bad;
1145 
1146 			/*
1147 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1148 			 * so return EINVAL in any other cmsg_type
1149 			 */
1150 			cmsg->cmsg_type =
1151 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1152 			cmsg->cmsg_level =
1153 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1154 			if (cmsg->cmsg_type == -1
1155 			    || cmsg->cmsg_level != SOL_SOCKET)
1156 				goto bad;
1157 
1158 			/*
1159 			 * Some applications (e.g. pulseaudio) attempt to
1160 			 * send ancillary data even if the underlying protocol
1161 			 * doesn't support it which is not allowed in the
1162 			 * FreeBSD system call interface.
1163 			 */
1164 			if (sa_family != AF_UNIX)
1165 				continue;
1166 
1167 			data = LINUX_CMSG_DATA(ptr_cmsg);
1168 			datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1169 
1170 			switch (cmsg->cmsg_type)
1171 			{
1172 			case SCM_RIGHTS:
1173 				break;
1174 
1175 			case SCM_CREDS:
1176 				data = &cmcred;
1177 				datalen = sizeof(cmcred);
1178 
1179 				/*
1180 				 * The lower levels will fill in the structure
1181 				 */
1182 				bzero(data, datalen);
1183 				break;
1184 			}
1185 
1186 			cmsg->cmsg_len = CMSG_LEN(datalen);
1187 
1188 			error = ENOBUFS;
1189 			if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1190 				goto bad;
1191 			if (!m_append(control, datalen, (c_caddr_t)data))
1192 				goto bad;
1193 		} while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1194 
1195 		if (m_length(control, NULL) == 0) {
1196 			m_freem(control);
1197 			control = NULL;
1198 		}
1199 	}
1200 
1201 	msg.msg_iov = iov;
1202 	msg.msg_flags = 0;
1203 	error = linux_sendit(td, args->s, &msg, args->flags, control,
1204 	    UIO_USERSPACE);
1205 
1206 bad:
1207 	free(iov, M_IOV);
1208 	if (cmsg)
1209 		free(cmsg, M_LINUX);
1210 	return (error);
1211 }
1212 
1213 int
1214 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1215 {
1216 	struct cmsghdr *cm;
1217 	struct cmsgcred *cmcred;
1218 	struct msghdr msg;
1219 	struct l_cmsghdr *linux_cmsg = NULL;
1220 	struct l_ucred linux_ucred;
1221 	socklen_t datalen, outlen;
1222 	struct l_msghdr linux_msg;
1223 	struct iovec *iov, *uiov;
1224 	struct mbuf *control = NULL;
1225 	struct mbuf **controlp;
1226 	caddr_t outbuf;
1227 	void *data;
1228 	int error, i, fd, fds, *fdp;
1229 
1230 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1231 	if (error)
1232 		return (error);
1233 
1234 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1235 	if (error)
1236 		return (error);
1237 
1238 #ifdef COMPAT_LINUX32
1239 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1240 	    &iov, EMSGSIZE);
1241 #else
1242 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1243 #endif
1244 	if (error)
1245 		return (error);
1246 
1247 	if (msg.msg_name) {
1248 		error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1249 		    msg.msg_namelen);
1250 		if (error)
1251 			goto bad;
1252 	}
1253 
1254 	uiov = msg.msg_iov;
1255 	msg.msg_iov = iov;
1256 	controlp = (msg.msg_control != NULL) ? &control : NULL;
1257 	error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1258 	msg.msg_iov = uiov;
1259 	if (error)
1260 		goto bad;
1261 
1262 	error = bsd_to_linux_msghdr(&msg, &linux_msg);
1263 	if (error)
1264 		goto bad;
1265 
1266 	if (linux_msg.msg_name) {
1267 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1268 		    PTRIN(linux_msg.msg_name));
1269 		if (error)
1270 			goto bad;
1271 	}
1272 	if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1273 		error = linux_sa_put(PTRIN(linux_msg.msg_name));
1274 		if (error)
1275 			goto bad;
1276 	}
1277 
1278 	outbuf = PTRIN(linux_msg.msg_control);
1279 	outlen = 0;
1280 
1281 	if (control) {
1282 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1283 
1284 		msg.msg_control = mtod(control, struct cmsghdr *);
1285 		msg.msg_controllen = control->m_len;
1286 
1287 		cm = CMSG_FIRSTHDR(&msg);
1288 
1289 		while (cm != NULL) {
1290 			linux_cmsg->cmsg_type =
1291 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1292 			linux_cmsg->cmsg_level =
1293 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1294 			if (linux_cmsg->cmsg_type == -1
1295 			    || cm->cmsg_level != SOL_SOCKET)
1296 			{
1297 				error = EINVAL;
1298 				goto bad;
1299 			}
1300 
1301 			data = CMSG_DATA(cm);
1302 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1303 
1304 			switch (cm->cmsg_type)
1305 			{
1306 			case SCM_RIGHTS:
1307 				if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1308 					fds = datalen / sizeof(int);
1309 					fdp = data;
1310 					for (i = 0; i < fds; i++) {
1311 						fd = *fdp++;
1312 						(void)kern_fcntl(td, fd,
1313 						    F_SETFD, FD_CLOEXEC);
1314 					}
1315 				}
1316 				break;
1317 
1318 			case SCM_CREDS:
1319 				/*
1320 				 * Currently LOCAL_CREDS is never in
1321 				 * effect for Linux so no need to worry
1322 				 * about sockcred
1323 				 */
1324 				if (datalen != sizeof(*cmcred)) {
1325 					error = EMSGSIZE;
1326 					goto bad;
1327 				}
1328 				cmcred = (struct cmsgcred *)data;
1329 				bzero(&linux_ucred, sizeof(linux_ucred));
1330 				linux_ucred.pid = cmcred->cmcred_pid;
1331 				linux_ucred.uid = cmcred->cmcred_uid;
1332 				linux_ucred.gid = cmcred->cmcred_gid;
1333 				data = &linux_ucred;
1334 				datalen = sizeof(linux_ucred);
1335 				break;
1336 			}
1337 
1338 			if (outlen + LINUX_CMSG_LEN(datalen) >
1339 			    linux_msg.msg_controllen) {
1340 				if (outlen == 0) {
1341 					error = EMSGSIZE;
1342 					goto bad;
1343 				} else {
1344 					linux_msg.msg_flags |=
1345 					    LINUX_MSG_CTRUNC;
1346 					goto out;
1347 				}
1348 			}
1349 
1350 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1351 
1352 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1353 			if (error)
1354 				goto bad;
1355 			outbuf += L_CMSG_HDRSZ;
1356 
1357 			error = copyout(data, outbuf, datalen);
1358 			if (error)
1359 				goto bad;
1360 
1361 			outbuf += LINUX_CMSG_ALIGN(datalen);
1362 			outlen += LINUX_CMSG_LEN(datalen);
1363 
1364 			cm = CMSG_NXTHDR(&msg, cm);
1365 		}
1366 	}
1367 
1368 out:
1369 	linux_msg.msg_controllen = outlen;
1370 	error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1371 
1372 bad:
1373 	free(iov, M_IOV);
1374 	m_freem(control);
1375 	free(linux_cmsg, M_LINUX);
1376 
1377 	return (error);
1378 }
1379 
1380 int
1381 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1382 {
1383 	struct shutdown_args /* {
1384 		int s;
1385 		int how;
1386 	} */ bsd_args;
1387 
1388 	bsd_args.s = args->s;
1389 	bsd_args.how = args->how;
1390 	return (sys_shutdown(td, &bsd_args));
1391 }
1392 
1393 int
1394 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1395 {
1396 	struct setsockopt_args /* {
1397 		int s;
1398 		int level;
1399 		int name;
1400 		caddr_t val;
1401 		int valsize;
1402 	} */ bsd_args;
1403 	l_timeval linux_tv;
1404 	struct timeval tv;
1405 	int error, name;
1406 
1407 	bsd_args.s = args->s;
1408 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1409 	switch (bsd_args.level) {
1410 	case SOL_SOCKET:
1411 		name = linux_to_bsd_so_sockopt(args->optname);
1412 		switch (name) {
1413 		case SO_RCVTIMEO:
1414 			/* FALLTHROUGH */
1415 		case SO_SNDTIMEO:
1416 			error = copyin(PTRIN(args->optval), &linux_tv,
1417 			    sizeof(linux_tv));
1418 			if (error)
1419 				return (error);
1420 			tv.tv_sec = linux_tv.tv_sec;
1421 			tv.tv_usec = linux_tv.tv_usec;
1422 			return (kern_setsockopt(td, args->s, bsd_args.level,
1423 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1424 			/* NOTREACHED */
1425 			break;
1426 		default:
1427 			break;
1428 		}
1429 		break;
1430 	case IPPROTO_IP:
1431 		name = linux_to_bsd_ip_sockopt(args->optname);
1432 		break;
1433 	case IPPROTO_TCP:
1434 		name = linux_to_bsd_tcp_sockopt(args->optname);
1435 		break;
1436 	default:
1437 		name = -1;
1438 		break;
1439 	}
1440 	if (name == -1)
1441 		return (ENOPROTOOPT);
1442 
1443 	bsd_args.name = name;
1444 	bsd_args.val = PTRIN(args->optval);
1445 	bsd_args.valsize = args->optlen;
1446 
1447 	if (name == IPV6_NEXTHOP) {
1448 		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1449 			bsd_args.valsize);
1450 		error = sys_setsockopt(td, &bsd_args);
1451 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1452 	} else
1453 		error = sys_setsockopt(td, &bsd_args);
1454 
1455 	return (error);
1456 }
1457 
1458 int
1459 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1460 {
1461 	struct getsockopt_args /* {
1462 		int s;
1463 		int level;
1464 		int name;
1465 		caddr_t val;
1466 		int *avalsize;
1467 	} */ bsd_args;
1468 	l_timeval linux_tv;
1469 	struct timeval tv;
1470 	socklen_t tv_len, xulen;
1471 	struct xucred xu;
1472 	struct l_ucred lxu;
1473 	int error, name;
1474 
1475 	bsd_args.s = args->s;
1476 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1477 	switch (bsd_args.level) {
1478 	case SOL_SOCKET:
1479 		name = linux_to_bsd_so_sockopt(args->optname);
1480 		switch (name) {
1481 		case SO_RCVTIMEO:
1482 			/* FALLTHROUGH */
1483 		case SO_SNDTIMEO:
1484 			tv_len = sizeof(tv);
1485 			error = kern_getsockopt(td, args->s, bsd_args.level,
1486 			    name, &tv, UIO_SYSSPACE, &tv_len);
1487 			if (error)
1488 				return (error);
1489 			linux_tv.tv_sec = tv.tv_sec;
1490 			linux_tv.tv_usec = tv.tv_usec;
1491 			return (copyout(&linux_tv, PTRIN(args->optval),
1492 			    sizeof(linux_tv)));
1493 			/* NOTREACHED */
1494 			break;
1495 		case LOCAL_PEERCRED:
1496 			if (args->optlen != sizeof(lxu))
1497 				return (EINVAL);
1498 			xulen = sizeof(xu);
1499 			error = kern_getsockopt(td, args->s, bsd_args.level,
1500 			    name, &xu, UIO_SYSSPACE, &xulen);
1501 			if (error)
1502 				return (error);
1503 			/*
1504 			 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1505 			 */
1506 			lxu.pid = 0;
1507 			lxu.uid = xu.cr_uid;
1508 			lxu.gid = xu.cr_gid;
1509 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1510 			/* NOTREACHED */
1511 			break;
1512 		default:
1513 			break;
1514 		}
1515 		break;
1516 	case IPPROTO_IP:
1517 		name = linux_to_bsd_ip_sockopt(args->optname);
1518 		break;
1519 	case IPPROTO_TCP:
1520 		name = linux_to_bsd_tcp_sockopt(args->optname);
1521 		break;
1522 	default:
1523 		name = -1;
1524 		break;
1525 	}
1526 	if (name == -1)
1527 		return (EINVAL);
1528 
1529 	bsd_args.name = name;
1530 	bsd_args.val = PTRIN(args->optval);
1531 	bsd_args.avalsize = PTRIN(args->optlen);
1532 
1533 	if (name == IPV6_NEXTHOP) {
1534 		error = sys_getsockopt(td, &bsd_args);
1535 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1536 	} else
1537 		error = sys_getsockopt(td, &bsd_args);
1538 
1539 	return (error);
1540 }
1541 
1542 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1543 
1544 /* Argument list sizes for linux_socketcall */
1545 
1546 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1547 
1548 static const unsigned char lxs_args[] = {
1549 	LINUX_AL(0) /* unused*/,	LINUX_AL(3) /* socket */,
1550 	LINUX_AL(3) /* bind */,		LINUX_AL(3) /* connect */,
1551 	LINUX_AL(2) /* listen */,	LINUX_AL(3) /* accept */,
1552 	LINUX_AL(3) /* getsockname */,	LINUX_AL(3) /* getpeername */,
1553 	LINUX_AL(4) /* socketpair */,	LINUX_AL(4) /* send */,
1554 	LINUX_AL(4) /* recv */,		LINUX_AL(6) /* sendto */,
1555 	LINUX_AL(6) /* recvfrom */,	LINUX_AL(2) /* shutdown */,
1556 	LINUX_AL(5) /* setsockopt */,	LINUX_AL(5) /* getsockopt */,
1557 	LINUX_AL(3) /* sendmsg */,	LINUX_AL(3) /* recvmsg */,
1558 	LINUX_AL(4) /* accept4 */
1559 };
1560 
1561 #define	LINUX_AL_SIZE	sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1562 
1563 int
1564 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1565 {
1566 	l_ulong a[6];
1567 	void *arg;
1568 	int error;
1569 
1570 	if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1571 		return (EINVAL);
1572 	error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1573 	if (error)
1574 		return (error);
1575 
1576 	arg = a;
1577 	switch (args->what) {
1578 	case LINUX_SOCKET:
1579 		return (linux_socket(td, arg));
1580 	case LINUX_BIND:
1581 		return (linux_bind(td, arg));
1582 	case LINUX_CONNECT:
1583 		return (linux_connect(td, arg));
1584 	case LINUX_LISTEN:
1585 		return (linux_listen(td, arg));
1586 	case LINUX_ACCEPT:
1587 		return (linux_accept(td, arg));
1588 	case LINUX_GETSOCKNAME:
1589 		return (linux_getsockname(td, arg));
1590 	case LINUX_GETPEERNAME:
1591 		return (linux_getpeername(td, arg));
1592 	case LINUX_SOCKETPAIR:
1593 		return (linux_socketpair(td, arg));
1594 	case LINUX_SEND:
1595 		return (linux_send(td, arg));
1596 	case LINUX_RECV:
1597 		return (linux_recv(td, arg));
1598 	case LINUX_SENDTO:
1599 		return (linux_sendto(td, arg));
1600 	case LINUX_RECVFROM:
1601 		return (linux_recvfrom(td, arg));
1602 	case LINUX_SHUTDOWN:
1603 		return (linux_shutdown(td, arg));
1604 	case LINUX_SETSOCKOPT:
1605 		return (linux_setsockopt(td, arg));
1606 	case LINUX_GETSOCKOPT:
1607 		return (linux_getsockopt(td, arg));
1608 	case LINUX_SENDMSG:
1609 		return (linux_sendmsg(td, arg));
1610 	case LINUX_RECVMSG:
1611 		return (linux_recvmsg(td, arg));
1612 	case LINUX_ACCEPT4:
1613 		return (linux_accept4(td, arg));
1614 	}
1615 
1616 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1617 	return (ENOSYS);
1618 }
1619 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1620