xref: /freebsd/sys/compat/linux/linux_socket.c (revision bb15ca603fa442c72dde3f3cb8b46db6970e3950)
1 /*-
2  * Copyright (c) 1995 S�ren Schmidt
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer
10  *    in this position and unchanged.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. The name of the author may not be used to endorse or promote products
15  *    derived from this software without specific prior written permission
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 /* XXX we use functions that might not exist. */
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 
36 #include <sys/param.h>
37 #include <sys/proc.h>
38 #include <sys/systm.h>
39 #include <sys/sysproto.h>
40 #include <sys/capability.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 <netinet/in.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/ip.h>
59 #ifdef INET6
60 #include <netinet/ip6.h>
61 #include <netinet6/ip6_var.h>
62 #include <netinet6/in6_var.h>
63 #endif
64 
65 #ifdef COMPAT_LINUX32
66 #include <machine/../linux32/linux.h>
67 #include <machine/../linux32/linux32_proto.h>
68 #else
69 #include <machine/../linux/linux.h>
70 #include <machine/../linux/linux_proto.h>
71 #endif
72 #include <compat/linux/linux_socket.h>
73 #include <compat/linux/linux_util.h>
74 
75 static int linux_to_bsd_domain(int);
76 
77 /*
78  * Reads a linux sockaddr and does any necessary translation.
79  * Linux sockaddrs don't have a length field, only a family.
80  * Copy the osockaddr structure pointed to by osa to kernel, adjust
81  * family and convert to sockaddr.
82  */
83 static int
84 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
85 {
86 	struct sockaddr *sa;
87 	struct osockaddr *kosa;
88 #ifdef INET6
89 	struct sockaddr_in6 *sin6;
90 	int oldv6size;
91 #endif
92 	char *name;
93 	int bdom, error, hdrlen, namelen;
94 
95 	if (salen < 2 || salen > UCHAR_MAX || !osa)
96 		return (EINVAL);
97 
98 #ifdef INET6
99 	oldv6size = 0;
100 	/*
101 	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
102 	 * if it's a v4-mapped address, so reserve the proper space
103 	 * for it.
104 	 */
105 	if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
106 		salen += sizeof(uint32_t);
107 		oldv6size = 1;
108 	}
109 #endif
110 
111 	kosa = malloc(salen, M_SONAME, M_WAITOK);
112 
113 	if ((error = copyin(osa, kosa, salen)))
114 		goto out;
115 
116 	bdom = linux_to_bsd_domain(kosa->sa_family);
117 	if (bdom == -1) {
118 		error = EAFNOSUPPORT;
119 		goto out;
120 	}
121 
122 #ifdef INET6
123 	/*
124 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
125 	 * which lacks the scope id compared with RFC2553 one. If we detect
126 	 * the situation, reject the address and write a message to system log.
127 	 *
128 	 * Still accept addresses for which the scope id is not used.
129 	 */
130 	if (oldv6size) {
131 		if (bdom == AF_INET6) {
132 			sin6 = (struct sockaddr_in6 *)kosa;
133 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
134 			    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
135 			     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
136 			     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
137 			     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
138 			     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
139 				sin6->sin6_scope_id = 0;
140 			} else {
141 				log(LOG_DEBUG,
142 				    "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
143 				error = EINVAL;
144 				goto out;
145 			}
146 		} else
147 			salen -= sizeof(uint32_t);
148 	}
149 #endif
150 	if (bdom == AF_INET) {
151 		if (salen < sizeof(struct sockaddr_in)) {
152 			error = EINVAL;
153 			goto out;
154 		}
155 		salen = sizeof(struct sockaddr_in);
156 	}
157 
158 	if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
159 		hdrlen = offsetof(struct sockaddr_un, sun_path);
160 		name = ((struct sockaddr_un *)kosa)->sun_path;
161 		if (*name == '\0') {
162 			/*
163 		 	 * Linux abstract namespace starts with a NULL byte.
164 			 * XXX We do not support abstract namespace yet.
165 			 */
166 			namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
167 		} else
168 			namelen = strnlen(name, salen - hdrlen);
169 		salen = hdrlen + namelen;
170 		if (salen > sizeof(struct sockaddr_un)) {
171 			error = ENAMETOOLONG;
172 			goto out;
173 		}
174 	}
175 
176 	sa = (struct sockaddr *)kosa;
177 	sa->sa_family = bdom;
178 	sa->sa_len = salen;
179 
180 	*sap = sa;
181 	return (0);
182 
183 out:
184 	free(kosa, M_SONAME);
185 	return (error);
186 }
187 
188 static int
189 linux_to_bsd_domain(int domain)
190 {
191 
192 	switch (domain) {
193 	case LINUX_AF_UNSPEC:
194 		return (AF_UNSPEC);
195 	case LINUX_AF_UNIX:
196 		return (AF_LOCAL);
197 	case LINUX_AF_INET:
198 		return (AF_INET);
199 	case LINUX_AF_INET6:
200 		return (AF_INET6);
201 	case LINUX_AF_AX25:
202 		return (AF_CCITT);
203 	case LINUX_AF_IPX:
204 		return (AF_IPX);
205 	case LINUX_AF_APPLETALK:
206 		return (AF_APPLETALK);
207 	}
208 	return (-1);
209 }
210 
211 static int
212 bsd_to_linux_domain(int domain)
213 {
214 
215 	switch (domain) {
216 	case AF_UNSPEC:
217 		return (LINUX_AF_UNSPEC);
218 	case AF_LOCAL:
219 		return (LINUX_AF_UNIX);
220 	case AF_INET:
221 		return (LINUX_AF_INET);
222 	case AF_INET6:
223 		return (LINUX_AF_INET6);
224 	case AF_CCITT:
225 		return (LINUX_AF_AX25);
226 	case AF_IPX:
227 		return (LINUX_AF_IPX);
228 	case AF_APPLETALK:
229 		return (LINUX_AF_APPLETALK);
230 	}
231 	return (-1);
232 }
233 
234 static int
235 linux_to_bsd_sockopt_level(int level)
236 {
237 
238 	switch (level) {
239 	case LINUX_SOL_SOCKET:
240 		return (SOL_SOCKET);
241 	}
242 	return (level);
243 }
244 
245 static int
246 bsd_to_linux_sockopt_level(int level)
247 {
248 
249 	switch (level) {
250 	case SOL_SOCKET:
251 		return (LINUX_SOL_SOCKET);
252 	}
253 	return (level);
254 }
255 
256 static int
257 linux_to_bsd_ip_sockopt(int opt)
258 {
259 
260 	switch (opt) {
261 	case LINUX_IP_TOS:
262 		return (IP_TOS);
263 	case LINUX_IP_TTL:
264 		return (IP_TTL);
265 	case LINUX_IP_OPTIONS:
266 		return (IP_OPTIONS);
267 	case LINUX_IP_MULTICAST_IF:
268 		return (IP_MULTICAST_IF);
269 	case LINUX_IP_MULTICAST_TTL:
270 		return (IP_MULTICAST_TTL);
271 	case LINUX_IP_MULTICAST_LOOP:
272 		return (IP_MULTICAST_LOOP);
273 	case LINUX_IP_ADD_MEMBERSHIP:
274 		return (IP_ADD_MEMBERSHIP);
275 	case LINUX_IP_DROP_MEMBERSHIP:
276 		return (IP_DROP_MEMBERSHIP);
277 	case LINUX_IP_HDRINCL:
278 		return (IP_HDRINCL);
279 	}
280 	return (-1);
281 }
282 
283 static int
284 linux_to_bsd_so_sockopt(int opt)
285 {
286 
287 	switch (opt) {
288 	case LINUX_SO_DEBUG:
289 		return (SO_DEBUG);
290 	case LINUX_SO_REUSEADDR:
291 		return (SO_REUSEADDR);
292 	case LINUX_SO_TYPE:
293 		return (SO_TYPE);
294 	case LINUX_SO_ERROR:
295 		return (SO_ERROR);
296 	case LINUX_SO_DONTROUTE:
297 		return (SO_DONTROUTE);
298 	case LINUX_SO_BROADCAST:
299 		return (SO_BROADCAST);
300 	case LINUX_SO_SNDBUF:
301 		return (SO_SNDBUF);
302 	case LINUX_SO_RCVBUF:
303 		return (SO_RCVBUF);
304 	case LINUX_SO_KEEPALIVE:
305 		return (SO_KEEPALIVE);
306 	case LINUX_SO_OOBINLINE:
307 		return (SO_OOBINLINE);
308 	case LINUX_SO_LINGER:
309 		return (SO_LINGER);
310 	case LINUX_SO_PEERCRED:
311 		return (LOCAL_PEERCRED);
312 	case LINUX_SO_RCVLOWAT:
313 		return (SO_RCVLOWAT);
314 	case LINUX_SO_SNDLOWAT:
315 		return (SO_SNDLOWAT);
316 	case LINUX_SO_RCVTIMEO:
317 		return (SO_RCVTIMEO);
318 	case LINUX_SO_SNDTIMEO:
319 		return (SO_SNDTIMEO);
320 	case LINUX_SO_TIMESTAMP:
321 		return (SO_TIMESTAMP);
322 	case LINUX_SO_ACCEPTCONN:
323 		return (SO_ACCEPTCONN);
324 	}
325 	return (-1);
326 }
327 
328 static int
329 linux_to_bsd_msg_flags(int flags)
330 {
331 	int ret_flags = 0;
332 
333 	if (flags & LINUX_MSG_OOB)
334 		ret_flags |= MSG_OOB;
335 	if (flags & LINUX_MSG_PEEK)
336 		ret_flags |= MSG_PEEK;
337 	if (flags & LINUX_MSG_DONTROUTE)
338 		ret_flags |= MSG_DONTROUTE;
339 	if (flags & LINUX_MSG_CTRUNC)
340 		ret_flags |= MSG_CTRUNC;
341 	if (flags & LINUX_MSG_TRUNC)
342 		ret_flags |= MSG_TRUNC;
343 	if (flags & LINUX_MSG_DONTWAIT)
344 		ret_flags |= MSG_DONTWAIT;
345 	if (flags & LINUX_MSG_EOR)
346 		ret_flags |= MSG_EOR;
347 	if (flags & LINUX_MSG_WAITALL)
348 		ret_flags |= MSG_WAITALL;
349 	if (flags & LINUX_MSG_NOSIGNAL)
350 		ret_flags |= MSG_NOSIGNAL;
351 #if 0 /* not handled */
352 	if (flags & LINUX_MSG_PROXY)
353 		;
354 	if (flags & LINUX_MSG_FIN)
355 		;
356 	if (flags & LINUX_MSG_SYN)
357 		;
358 	if (flags & LINUX_MSG_CONFIRM)
359 		;
360 	if (flags & LINUX_MSG_RST)
361 		;
362 	if (flags & LINUX_MSG_ERRQUEUE)
363 		;
364 #endif
365 	return ret_flags;
366 }
367 
368 /*
369 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
370 * native syscall will fault.  Thus, we don't really need to check the
371 * return values for these functions.
372 */
373 
374 static int
375 bsd_to_linux_sockaddr(struct sockaddr *arg)
376 {
377 	struct sockaddr sa;
378 	size_t sa_len = sizeof(struct sockaddr);
379 	int error;
380 
381 	if ((error = copyin(arg, &sa, sa_len)))
382 		return (error);
383 
384 	*(u_short *)&sa = sa.sa_family;
385 
386 	error = copyout(&sa, arg, sa_len);
387 
388 	return (error);
389 }
390 
391 static int
392 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
393 {
394 	struct sockaddr sa;
395 	size_t sa_len = sizeof(struct sockaddr);
396 	int error;
397 
398 	if ((error = copyin(arg, &sa, sa_len)))
399 		return (error);
400 
401 	sa.sa_family = *(sa_family_t *)&sa;
402 	sa.sa_len = len;
403 
404 	error = copyout(&sa, arg, sa_len);
405 
406 	return (error);
407 }
408 
409 
410 static int
411 linux_sa_put(struct osockaddr *osa)
412 {
413 	struct osockaddr sa;
414 	int error, bdom;
415 
416 	/*
417 	 * Only read/write the osockaddr family part, the rest is
418 	 * not changed.
419 	 */
420 	error = copyin(osa, &sa, sizeof(sa.sa_family));
421 	if (error)
422 		return (error);
423 
424 	bdom = bsd_to_linux_domain(sa.sa_family);
425 	if (bdom == -1)
426 		return (EINVAL);
427 
428 	sa.sa_family = bdom;
429 	error = copyout(&sa, osa, sizeof(sa.sa_family));
430 	if (error)
431 		return (error);
432 
433 	return (0);
434 }
435 
436 static int
437 linux_to_bsd_cmsg_type(int cmsg_type)
438 {
439 
440 	switch (cmsg_type) {
441 	case LINUX_SCM_RIGHTS:
442 		return (SCM_RIGHTS);
443 	case LINUX_SCM_CREDENTIALS:
444 		return (SCM_CREDS);
445 	}
446 	return (-1);
447 }
448 
449 static int
450 bsd_to_linux_cmsg_type(int cmsg_type)
451 {
452 
453 	switch (cmsg_type) {
454 	case SCM_RIGHTS:
455 		return (LINUX_SCM_RIGHTS);
456 	case SCM_CREDS:
457 		return (LINUX_SCM_CREDENTIALS);
458 	}
459 	return (-1);
460 }
461 
462 static int
463 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
464 {
465 	if (lhdr->msg_controllen > INT_MAX)
466 		return (ENOBUFS);
467 
468 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
469 	bhdr->msg_namelen	= lhdr->msg_namelen;
470 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
471 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
472 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
473 
474 	/*
475 	 * msg_controllen is skipped since BSD and LINUX control messages
476 	 * are potentially different sizes (e.g. the cred structure used
477 	 * by SCM_CREDS is different between the two operating system).
478 	 *
479 	 * The caller can set it (if necessary) after converting all the
480 	 * control messages.
481 	 */
482 
483 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
484 	return (0);
485 }
486 
487 static int
488 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
489 {
490 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
491 	lhdr->msg_namelen	= bhdr->msg_namelen;
492 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
493 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
494 	lhdr->msg_control	= PTROUT(bhdr->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 	/* msg_flags skipped */
506 	return (0);
507 }
508 
509 static int
510 linux_set_socket_flags(struct thread *td, int s, int flags)
511 {
512 	int error;
513 
514 	if (flags & LINUX_SOCK_NONBLOCK) {
515 		error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
516 		if (error)
517 			return (error);
518 	}
519 	if (flags & LINUX_SOCK_CLOEXEC) {
520 		error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
521 		if (error)
522 			return (error);
523 	}
524 	return (0);
525 }
526 
527 static int
528 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
529     struct mbuf *control, enum uio_seg segflg)
530 {
531 	struct sockaddr *to;
532 	int error;
533 
534 	if (mp->msg_name != NULL) {
535 		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
536 		if (error)
537 			return (error);
538 		mp->msg_name = to;
539 	} else
540 		to = NULL;
541 
542 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
543 	    segflg);
544 
545 	if (to)
546 		free(to, M_SONAME);
547 	return (error);
548 }
549 
550 /* Return 0 if IP_HDRINCL is set for the given socket. */
551 static int
552 linux_check_hdrincl(struct thread *td, int s)
553 {
554 	int error, optval, size_val;
555 
556 	size_val = sizeof(optval);
557 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
558 	    &optval, UIO_SYSSPACE, &size_val);
559 	if (error)
560 		return (error);
561 
562 	return (optval == 0);
563 }
564 
565 struct linux_sendto_args {
566 	int s;
567 	l_uintptr_t msg;
568 	int len;
569 	int flags;
570 	l_uintptr_t to;
571 	int tolen;
572 };
573 
574 /*
575  * Updated sendto() when IP_HDRINCL is set:
576  * tweak endian-dependent fields in the IP packet.
577  */
578 static int
579 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
580 {
581 /*
582  * linux_ip_copysize defines how many bytes we should copy
583  * from the beginning of the IP packet before we customize it for BSD.
584  * It should include all the fields we modify (ip_len and ip_off).
585  */
586 #define linux_ip_copysize	8
587 
588 	struct ip *packet;
589 	struct msghdr msg;
590 	struct iovec aiov[1];
591 	int error;
592 
593 	/* Check that the packet isn't too big or too small. */
594 	if (linux_args->len < linux_ip_copysize ||
595 	    linux_args->len > IP_MAXPACKET)
596 		return (EINVAL);
597 
598 	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
599 
600 	/* Make kernel copy of the packet to be sent */
601 	if ((error = copyin(PTRIN(linux_args->msg), packet,
602 	    linux_args->len)))
603 		goto goout;
604 
605 	/* Convert fields from Linux to BSD raw IP socket format */
606 	packet->ip_len = linux_args->len;
607 	packet->ip_off = ntohs(packet->ip_off);
608 
609 	/* Prepare the msghdr and iovec structures describing the new packet */
610 	msg.msg_name = PTRIN(linux_args->to);
611 	msg.msg_namelen = linux_args->tolen;
612 	msg.msg_iov = aiov;
613 	msg.msg_iovlen = 1;
614 	msg.msg_control = NULL;
615 	msg.msg_flags = 0;
616 	aiov[0].iov_base = (char *)packet;
617 	aiov[0].iov_len = linux_args->len;
618 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
619 	    NULL, UIO_SYSSPACE);
620 goout:
621 	free(packet, M_TEMP);
622 	return (error);
623 }
624 
625 struct linux_socket_args {
626 	int domain;
627 	int type;
628 	int protocol;
629 };
630 
631 static int
632 linux_socket(struct thread *td, struct linux_socket_args *args)
633 {
634 	struct socket_args /* {
635 		int domain;
636 		int type;
637 		int protocol;
638 	} */ bsd_args;
639 	int retval_socket, socket_flags;
640 
641 	bsd_args.protocol = args->protocol;
642 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
643 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
644 		return (EINVAL);
645 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
646 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
647 		return (EINVAL);
648 	bsd_args.domain = linux_to_bsd_domain(args->domain);
649 	if (bsd_args.domain == -1)
650 		return (EAFNOSUPPORT);
651 
652 	retval_socket = sys_socket(td, &bsd_args);
653 	if (retval_socket)
654 		return (retval_socket);
655 
656 	retval_socket = linux_set_socket_flags(td, td->td_retval[0],
657 	    socket_flags);
658 	if (retval_socket) {
659 		(void)kern_close(td, td->td_retval[0]);
660 		goto out;
661 	}
662 
663 	if (bsd_args.type == SOCK_RAW
664 	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
665 	    && bsd_args.domain == PF_INET) {
666 		/* It's a raw IP socket: set the IP_HDRINCL option. */
667 		int hdrincl;
668 
669 		hdrincl = 1;
670 		/* We ignore any error returned by kern_setsockopt() */
671 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
672 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
673 	}
674 #ifdef INET6
675 	/*
676 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
677 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
678 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
679 	 * sysctl value.
680 	 */
681 	if (bsd_args.domain == PF_INET6) {
682 		int v6only;
683 
684 		v6only = 0;
685 		/* We ignore any error returned by setsockopt() */
686 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
687 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
688 	}
689 #endif
690 
691 out:
692 	return (retval_socket);
693 }
694 
695 struct linux_bind_args {
696 	int s;
697 	l_uintptr_t name;
698 	int namelen;
699 };
700 
701 static int
702 linux_bind(struct thread *td, struct linux_bind_args *args)
703 {
704 	struct sockaddr *sa;
705 	int error;
706 
707 	error = linux_getsockaddr(&sa, PTRIN(args->name),
708 	    args->namelen);
709 	if (error)
710 		return (error);
711 
712 	error = kern_bind(td, args->s, sa);
713 	free(sa, M_SONAME);
714 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
715 	   	return (EINVAL);
716 	return (error);
717 }
718 
719 struct linux_connect_args {
720 	int s;
721 	l_uintptr_t name;
722 	int namelen;
723 };
724 int linux_connect(struct thread *, struct linux_connect_args *);
725 
726 int
727 linux_connect(struct thread *td, struct linux_connect_args *args)
728 {
729 	struct socket *so;
730 	struct sockaddr *sa;
731 	u_int fflag;
732 	int error;
733 
734 	error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
735 	    args->namelen);
736 	if (error)
737 		return (error);
738 
739 	error = kern_connect(td, args->s, sa);
740 	free(sa, M_SONAME);
741 	if (error != EISCONN)
742 		return (error);
743 
744 	/*
745 	 * Linux doesn't return EISCONN the first time it occurs,
746 	 * when on a non-blocking socket. Instead it returns the
747 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
748 	 *
749 	 * XXXRW: Instead of using fgetsock(), check that it is a
750 	 * socket and use the file descriptor reference instead of
751 	 * creating a new one.
752 	 */
753 	error = fgetsock(td, args->s, CAP_CONNECT, &so, &fflag);
754 	if (error == 0) {
755 		error = EISCONN;
756 		if (fflag & FNONBLOCK) {
757 			SOCK_LOCK(so);
758 			if (so->so_emuldata == 0)
759 				error = so->so_error;
760 			so->so_emuldata = (void *)1;
761 			SOCK_UNLOCK(so);
762 		}
763 		fputsock(so);
764 	}
765 	return (error);
766 }
767 
768 struct linux_listen_args {
769 	int s;
770 	int backlog;
771 };
772 
773 static int
774 linux_listen(struct thread *td, struct linux_listen_args *args)
775 {
776 	struct listen_args /* {
777 		int s;
778 		int backlog;
779 	} */ bsd_args;
780 
781 	bsd_args.s = args->s;
782 	bsd_args.backlog = args->backlog;
783 	return (sys_listen(td, &bsd_args));
784 }
785 
786 static int
787 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
788     l_uintptr_t namelen, int flags)
789 {
790 	struct accept_args /* {
791 		int	s;
792 		struct sockaddr * __restrict name;
793 		socklen_t * __restrict anamelen;
794 	} */ bsd_args;
795 	int error;
796 
797 	if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
798 		return (EINVAL);
799 
800 	bsd_args.s = s;
801 	/* XXX: */
802 	bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
803 	bsd_args.anamelen = PTRIN(namelen);/* XXX */
804 	error = sys_accept(td, &bsd_args);
805 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
806 	if (error) {
807 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
808 			return (EINVAL);
809 		return (error);
810 	}
811 
812 	/*
813 	 * linux appears not to copy flags from the parent socket to the
814 	 * accepted one, so we must clear the flags in the new descriptor
815 	 * and apply the requested flags.
816 	 */
817 	error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
818 	if (error)
819 		goto out;
820 	error = linux_set_socket_flags(td, td->td_retval[0], flags);
821 	if (error)
822 		goto out;
823 	if (addr)
824 		error = linux_sa_put(PTRIN(addr));
825 
826 out:
827 	if (error) {
828 		(void)kern_close(td, td->td_retval[0]);
829 		td->td_retval[0] = 0;
830 	}
831 	return (error);
832 }
833 
834 struct linux_accept_args {
835 	int s;
836 	l_uintptr_t addr;
837 	l_uintptr_t namelen;
838 };
839 
840 static int
841 linux_accept(struct thread *td, struct linux_accept_args *args)
842 {
843 
844 	return (linux_accept_common(td, args->s, args->addr,
845 	    args->namelen, 0));
846 }
847 
848 struct linux_accept4_args {
849 	int s;
850 	l_uintptr_t addr;
851 	l_uintptr_t namelen;
852 	int flags;
853 };
854 
855 static int
856 linux_accept4(struct thread *td, struct linux_accept4_args *args)
857 {
858 
859 	return (linux_accept_common(td, args->s, args->addr,
860 	    args->namelen, args->flags));
861 }
862 
863 struct linux_getsockname_args {
864 	int s;
865 	l_uintptr_t addr;
866 	l_uintptr_t namelen;
867 };
868 
869 static int
870 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
871 {
872 	struct getsockname_args /* {
873 		int	fdes;
874 		struct sockaddr * __restrict asa;
875 		socklen_t * __restrict alen;
876 	} */ bsd_args;
877 	int error;
878 
879 	bsd_args.fdes = args->s;
880 	/* XXX: */
881 	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
882 	bsd_args.alen = PTRIN(args->namelen);	/* XXX */
883 	error = sys_getsockname(td, &bsd_args);
884 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
885 	if (error)
886 		return (error);
887 	error = linux_sa_put(PTRIN(args->addr));
888 	if (error)
889 		return (error);
890 	return (0);
891 }
892 
893 struct linux_getpeername_args {
894 	int s;
895 	l_uintptr_t addr;
896 	l_uintptr_t namelen;
897 };
898 
899 static int
900 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
901 {
902 	struct getpeername_args /* {
903 		int fdes;
904 		caddr_t asa;
905 		int *alen;
906 	} */ bsd_args;
907 	int error;
908 
909 	bsd_args.fdes = args->s;
910 	bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
911 	bsd_args.alen = (int *)PTRIN(args->namelen);
912 	error = sys_getpeername(td, &bsd_args);
913 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
914 	if (error)
915 		return (error);
916 	error = linux_sa_put(PTRIN(args->addr));
917 	if (error)
918 		return (error);
919 	return (0);
920 }
921 
922 struct linux_socketpair_args {
923 	int domain;
924 	int type;
925 	int protocol;
926 	l_uintptr_t rsv;
927 };
928 
929 static int
930 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
931 {
932 	struct socketpair_args /* {
933 		int domain;
934 		int type;
935 		int protocol;
936 		int *rsv;
937 	} */ bsd_args;
938 	int error, socket_flags;
939 	int sv[2];
940 
941 	bsd_args.domain = linux_to_bsd_domain(args->domain);
942 	if (bsd_args.domain != PF_LOCAL)
943 		return (EAFNOSUPPORT);
944 
945 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
946 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
947 		return (EINVAL);
948 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
949 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
950 		return (EINVAL);
951 
952 	if (args->protocol != 0 && args->protocol != PF_UNIX)
953 
954 		/*
955 		 * Use of PF_UNIX as protocol argument is not right,
956 		 * but Linux does it.
957 		 * Do not map PF_UNIX as its Linux value is identical
958 		 * to FreeBSD one.
959 		 */
960 		return (EPROTONOSUPPORT);
961 	else
962 		bsd_args.protocol = 0;
963 	bsd_args.rsv = (int *)PTRIN(args->rsv);
964 	error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
965 	    bsd_args.protocol, sv);
966 	if (error)
967 		return (error);
968 	error = linux_set_socket_flags(td, sv[0], socket_flags);
969 	if (error)
970 		goto out;
971 	error = linux_set_socket_flags(td, sv[1], socket_flags);
972 	if (error)
973 		goto out;
974 
975 	error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
976 
977 out:
978 	if (error) {
979 		(void)kern_close(td, sv[0]);
980 		(void)kern_close(td, sv[1]);
981 	}
982 	return (error);
983 }
984 
985 struct linux_send_args {
986 	int s;
987 	l_uintptr_t msg;
988 	int len;
989 	int flags;
990 };
991 
992 static int
993 linux_send(struct thread *td, struct linux_send_args *args)
994 {
995 	struct sendto_args /* {
996 		int s;
997 		caddr_t buf;
998 		int len;
999 		int flags;
1000 		caddr_t to;
1001 		int tolen;
1002 	} */ bsd_args;
1003 
1004 	bsd_args.s = args->s;
1005 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1006 	bsd_args.len = args->len;
1007 	bsd_args.flags = args->flags;
1008 	bsd_args.to = NULL;
1009 	bsd_args.tolen = 0;
1010 	return sys_sendto(td, &bsd_args);
1011 }
1012 
1013 struct linux_recv_args {
1014 	int s;
1015 	l_uintptr_t msg;
1016 	int len;
1017 	int flags;
1018 };
1019 
1020 static int
1021 linux_recv(struct thread *td, struct linux_recv_args *args)
1022 {
1023 	struct recvfrom_args /* {
1024 		int s;
1025 		caddr_t buf;
1026 		int len;
1027 		int flags;
1028 		struct sockaddr *from;
1029 		socklen_t fromlenaddr;
1030 	} */ bsd_args;
1031 
1032 	bsd_args.s = args->s;
1033 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1034 	bsd_args.len = args->len;
1035 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1036 	bsd_args.from = NULL;
1037 	bsd_args.fromlenaddr = 0;
1038 	return (sys_recvfrom(td, &bsd_args));
1039 }
1040 
1041 static int
1042 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1043 {
1044 	struct msghdr msg;
1045 	struct iovec aiov;
1046 	int error;
1047 
1048 	if (linux_check_hdrincl(td, args->s) == 0)
1049 		/* IP_HDRINCL set, tweak the packet before sending */
1050 		return (linux_sendto_hdrincl(td, args));
1051 
1052 	msg.msg_name = PTRIN(args->to);
1053 	msg.msg_namelen = args->tolen;
1054 	msg.msg_iov = &aiov;
1055 	msg.msg_iovlen = 1;
1056 	msg.msg_control = NULL;
1057 	msg.msg_flags = 0;
1058 	aiov.iov_base = PTRIN(args->msg);
1059 	aiov.iov_len = args->len;
1060 	error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1061 	    UIO_USERSPACE);
1062 	return (error);
1063 }
1064 
1065 struct linux_recvfrom_args {
1066 	int s;
1067 	l_uintptr_t buf;
1068 	int len;
1069 	int flags;
1070 	l_uintptr_t from;
1071 	l_uintptr_t fromlen;
1072 };
1073 
1074 static int
1075 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1076 {
1077 	struct recvfrom_args /* {
1078 		int	s;
1079 		caddr_t	buf;
1080 		size_t	len;
1081 		int	flags;
1082 		struct sockaddr * __restrict from;
1083 		socklen_t * __restrict fromlenaddr;
1084 	} */ bsd_args;
1085 	size_t len;
1086 	int error;
1087 
1088 	if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1089 		return (error);
1090 
1091 	bsd_args.s = args->s;
1092 	bsd_args.buf = PTRIN(args->buf);
1093 	bsd_args.len = args->len;
1094 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1095 	/* XXX: */
1096 	bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1097 	bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1098 
1099 	linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1100 	error = sys_recvfrom(td, &bsd_args);
1101 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1102 
1103 	if (error)
1104 		return (error);
1105 	if (args->from) {
1106 		error = linux_sa_put((struct osockaddr *)
1107 		    PTRIN(args->from));
1108 		if (error)
1109 			return (error);
1110 	}
1111 	return (0);
1112 }
1113 
1114 struct linux_sendmsg_args {
1115 	int s;
1116 	l_uintptr_t msg;
1117 	int flags;
1118 };
1119 
1120 static int
1121 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1122 {
1123 	struct cmsghdr *cmsg;
1124 	struct cmsgcred cmcred;
1125 	struct mbuf *control;
1126 	struct msghdr msg;
1127 	struct l_cmsghdr linux_cmsg;
1128 	struct l_cmsghdr *ptr_cmsg;
1129 	struct l_msghdr linux_msg;
1130 	struct iovec *iov;
1131 	socklen_t datalen;
1132 	struct sockaddr *sa;
1133 	sa_family_t sa_family;
1134 	void *data;
1135 	int error;
1136 
1137 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1138 	if (error)
1139 		return (error);
1140 
1141 	/*
1142 	 * Some Linux applications (ping) define a non-NULL control data
1143 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1144 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1145 	 * order to handle this case.  This should be checked, but allows the
1146 	 * Linux ping to work.
1147 	 */
1148 	if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1149 		linux_msg.msg_control = PTROUT(NULL);
1150 
1151 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1152 	if (error)
1153 		return (error);
1154 
1155 #ifdef COMPAT_LINUX32
1156 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1157 	    &iov, EMSGSIZE);
1158 #else
1159 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1160 #endif
1161 	if (error)
1162 		return (error);
1163 
1164 	control = NULL;
1165 	cmsg = NULL;
1166 
1167 	if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1168 		error = kern_getsockname(td, args->s, &sa, &datalen);
1169 		if (error)
1170 			goto bad;
1171 		sa_family = sa->sa_family;
1172 		free(sa, M_SONAME);
1173 
1174 		error = ENOBUFS;
1175 		cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1176 		control = m_get(M_WAIT, MT_CONTROL);
1177 		if (control == NULL)
1178 			goto bad;
1179 
1180 		do {
1181 			error = copyin(ptr_cmsg, &linux_cmsg,
1182 			    sizeof(struct l_cmsghdr));
1183 			if (error)
1184 				goto bad;
1185 
1186 			error = EINVAL;
1187 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1188 				goto bad;
1189 
1190 			/*
1191 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1192 			 * so return EINVAL in any other cmsg_type
1193 			 */
1194 			cmsg->cmsg_type =
1195 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1196 			cmsg->cmsg_level =
1197 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1198 			if (cmsg->cmsg_type == -1
1199 			    || cmsg->cmsg_level != SOL_SOCKET)
1200 				goto bad;
1201 
1202 			/*
1203 			 * Some applications (e.g. pulseaudio) attempt to
1204 			 * send ancillary data even if the underlying protocol
1205 			 * doesn't support it which is not allowed in the
1206 			 * FreeBSD system call interface.
1207 			 */
1208 			if (sa_family != AF_UNIX)
1209 				continue;
1210 
1211 			data = LINUX_CMSG_DATA(ptr_cmsg);
1212 			datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1213 
1214 			switch (cmsg->cmsg_type)
1215 			{
1216 			case SCM_RIGHTS:
1217 				break;
1218 
1219 			case SCM_CREDS:
1220 				data = &cmcred;
1221 				datalen = sizeof(cmcred);
1222 
1223 				/*
1224 				 * The lower levels will fill in the structure
1225 				 */
1226 				bzero(data, datalen);
1227 				break;
1228 			}
1229 
1230 			cmsg->cmsg_len = CMSG_LEN(datalen);
1231 
1232 			error = ENOBUFS;
1233 			if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1234 				goto bad;
1235 			if (!m_append(control, datalen, (c_caddr_t)data))
1236 				goto bad;
1237 		} while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1238 
1239 		if (m_length(control, NULL) == 0) {
1240 			m_freem(control);
1241 			control = NULL;
1242 		}
1243 	}
1244 
1245 	msg.msg_iov = iov;
1246 	msg.msg_flags = 0;
1247 	error = linux_sendit(td, args->s, &msg, args->flags, control,
1248 	    UIO_USERSPACE);
1249 
1250 bad:
1251 	free(iov, M_IOV);
1252 	if (cmsg)
1253 		free(cmsg, M_TEMP);
1254 	return (error);
1255 }
1256 
1257 struct linux_recvmsg_args {
1258 	int s;
1259 	l_uintptr_t msg;
1260 	int flags;
1261 };
1262 
1263 static int
1264 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1265 {
1266 	struct cmsghdr *cm;
1267 	struct cmsgcred *cmcred;
1268 	struct msghdr msg;
1269 	struct l_cmsghdr *linux_cmsg = NULL;
1270 	struct l_ucred linux_ucred;
1271 	socklen_t datalen, outlen;
1272 	struct l_msghdr linux_msg;
1273 	struct iovec *iov, *uiov;
1274 	struct mbuf *control = NULL;
1275 	struct mbuf **controlp;
1276 	caddr_t outbuf;
1277 	void *data;
1278 	int error, i, fd, fds, *fdp;
1279 
1280 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1281 	if (error)
1282 		return (error);
1283 
1284 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1285 	if (error)
1286 		return (error);
1287 
1288 #ifdef COMPAT_LINUX32
1289 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1290 	    &iov, EMSGSIZE);
1291 #else
1292 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1293 #endif
1294 	if (error)
1295 		return (error);
1296 
1297 	if (msg.msg_name) {
1298 		error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1299 		    msg.msg_namelen);
1300 		if (error)
1301 			goto bad;
1302 	}
1303 
1304 	uiov = msg.msg_iov;
1305 	msg.msg_iov = iov;
1306 	controlp = (msg.msg_control != NULL) ? &control : NULL;
1307 	error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1308 	msg.msg_iov = uiov;
1309 	if (error)
1310 		goto bad;
1311 
1312 	error = bsd_to_linux_msghdr(&msg, &linux_msg);
1313 	if (error)
1314 		goto bad;
1315 
1316 	if (linux_msg.msg_name) {
1317 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1318 		    PTRIN(linux_msg.msg_name));
1319 		if (error)
1320 			goto bad;
1321 	}
1322 	if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1323 		error = linux_sa_put(PTRIN(linux_msg.msg_name));
1324 		if (error)
1325 			goto bad;
1326 	}
1327 
1328 	outbuf = PTRIN(linux_msg.msg_control);
1329 	outlen = 0;
1330 
1331 	if (control) {
1332 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1333 
1334 		msg.msg_control = mtod(control, struct cmsghdr *);
1335 		msg.msg_controllen = control->m_len;
1336 
1337 		cm = CMSG_FIRSTHDR(&msg);
1338 
1339 		while (cm != NULL) {
1340 			linux_cmsg->cmsg_type =
1341 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1342 			linux_cmsg->cmsg_level =
1343 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1344 			if (linux_cmsg->cmsg_type == -1
1345 			    || cm->cmsg_level != SOL_SOCKET)
1346 			{
1347 				error = EINVAL;
1348 				goto bad;
1349 			}
1350 
1351 			data = CMSG_DATA(cm);
1352 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1353 
1354 			switch (cm->cmsg_type)
1355 			{
1356 			case SCM_RIGHTS:
1357 				if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1358 					fds = datalen / sizeof(int);
1359 					fdp = data;
1360 					for (i = 0; i < fds; i++) {
1361 						fd = *fdp++;
1362 						(void)kern_fcntl(td, fd,
1363 						    F_SETFD, FD_CLOEXEC);
1364 					}
1365 				}
1366 				break;
1367 
1368 			case SCM_CREDS:
1369 				/*
1370 				 * Currently LOCAL_CREDS is never in
1371 				 * effect for Linux so no need to worry
1372 				 * about sockcred
1373 				 */
1374 				if (datalen != sizeof(*cmcred)) {
1375 					error = EMSGSIZE;
1376 					goto bad;
1377 				}
1378 				cmcred = (struct cmsgcred *)data;
1379 				bzero(&linux_ucred, sizeof(linux_ucred));
1380 				linux_ucred.pid = cmcred->cmcred_pid;
1381 				linux_ucred.uid = cmcred->cmcred_uid;
1382 				linux_ucred.gid = cmcred->cmcred_gid;
1383 				data = &linux_ucred;
1384 				datalen = sizeof(linux_ucred);
1385 				break;
1386 			}
1387 
1388 			if (outlen + LINUX_CMSG_LEN(datalen) >
1389 			    linux_msg.msg_controllen) {
1390 				if (outlen == 0) {
1391 					error = EMSGSIZE;
1392 					goto bad;
1393 				} else {
1394 					linux_msg.msg_flags |=
1395 					    LINUX_MSG_CTRUNC;
1396 					goto out;
1397 				}
1398 			}
1399 
1400 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1401 
1402 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1403 			if (error)
1404 				goto bad;
1405 			outbuf += L_CMSG_HDRSZ;
1406 
1407 			error = copyout(data, outbuf, datalen);
1408 			if (error)
1409 				goto bad;
1410 
1411 			outbuf += LINUX_CMSG_ALIGN(datalen);
1412 			outlen += LINUX_CMSG_LEN(datalen);
1413 
1414 			cm = CMSG_NXTHDR(&msg, cm);
1415 		}
1416 	}
1417 
1418 out:
1419 	linux_msg.msg_controllen = outlen;
1420 	error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1421 
1422 bad:
1423 	free(iov, M_IOV);
1424 	if (control != NULL)
1425 		m_freem(control);
1426 	if (linux_cmsg != NULL)
1427 		free(linux_cmsg, M_TEMP);
1428 
1429 	return (error);
1430 }
1431 
1432 struct linux_shutdown_args {
1433 	int s;
1434 	int how;
1435 };
1436 
1437 static int
1438 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1439 {
1440 	struct shutdown_args /* {
1441 		int s;
1442 		int how;
1443 	} */ bsd_args;
1444 
1445 	bsd_args.s = args->s;
1446 	bsd_args.how = args->how;
1447 	return (sys_shutdown(td, &bsd_args));
1448 }
1449 
1450 struct linux_setsockopt_args {
1451 	int s;
1452 	int level;
1453 	int optname;
1454 	l_uintptr_t optval;
1455 	int optlen;
1456 };
1457 
1458 static int
1459 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1460 {
1461 	struct setsockopt_args /* {
1462 		int s;
1463 		int level;
1464 		int name;
1465 		caddr_t val;
1466 		int valsize;
1467 	} */ bsd_args;
1468 	l_timeval linux_tv;
1469 	struct timeval tv;
1470 	int error, name;
1471 
1472 	bsd_args.s = args->s;
1473 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1474 	switch (bsd_args.level) {
1475 	case SOL_SOCKET:
1476 		name = linux_to_bsd_so_sockopt(args->optname);
1477 		switch (name) {
1478 		case SO_RCVTIMEO:
1479 			/* FALLTHROUGH */
1480 		case SO_SNDTIMEO:
1481 			error = copyin(PTRIN(args->optval), &linux_tv,
1482 			    sizeof(linux_tv));
1483 			if (error)
1484 				return (error);
1485 			tv.tv_sec = linux_tv.tv_sec;
1486 			tv.tv_usec = linux_tv.tv_usec;
1487 			return (kern_setsockopt(td, args->s, bsd_args.level,
1488 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1489 			/* NOTREACHED */
1490 			break;
1491 		default:
1492 			break;
1493 		}
1494 		break;
1495 	case IPPROTO_IP:
1496 		name = linux_to_bsd_ip_sockopt(args->optname);
1497 		break;
1498 	case IPPROTO_TCP:
1499 		/* Linux TCP option values match BSD's */
1500 		name = args->optname;
1501 		break;
1502 	default:
1503 		name = -1;
1504 		break;
1505 	}
1506 	if (name == -1)
1507 		return (ENOPROTOOPT);
1508 
1509 	bsd_args.name = name;
1510 	bsd_args.val = PTRIN(args->optval);
1511 	bsd_args.valsize = args->optlen;
1512 
1513 	if (name == IPV6_NEXTHOP) {
1514 		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1515 			bsd_args.valsize);
1516 		error = sys_setsockopt(td, &bsd_args);
1517 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1518 	} else
1519 		error = sys_setsockopt(td, &bsd_args);
1520 
1521 	return (error);
1522 }
1523 
1524 struct linux_getsockopt_args {
1525 	int s;
1526 	int level;
1527 	int optname;
1528 	l_uintptr_t optval;
1529 	l_uintptr_t optlen;
1530 };
1531 
1532 static int
1533 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1534 {
1535 	struct getsockopt_args /* {
1536 		int s;
1537 		int level;
1538 		int name;
1539 		caddr_t val;
1540 		int *avalsize;
1541 	} */ bsd_args;
1542 	l_timeval linux_tv;
1543 	struct timeval tv;
1544 	socklen_t tv_len, xulen;
1545 	struct xucred xu;
1546 	struct l_ucred lxu;
1547 	int error, name;
1548 
1549 	bsd_args.s = args->s;
1550 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1551 	switch (bsd_args.level) {
1552 	case SOL_SOCKET:
1553 		name = linux_to_bsd_so_sockopt(args->optname);
1554 		switch (name) {
1555 		case SO_RCVTIMEO:
1556 			/* FALLTHROUGH */
1557 		case SO_SNDTIMEO:
1558 			tv_len = sizeof(tv);
1559 			error = kern_getsockopt(td, args->s, bsd_args.level,
1560 			    name, &tv, UIO_SYSSPACE, &tv_len);
1561 			if (error)
1562 				return (error);
1563 			linux_tv.tv_sec = tv.tv_sec;
1564 			linux_tv.tv_usec = tv.tv_usec;
1565 			return (copyout(&linux_tv, PTRIN(args->optval),
1566 			    sizeof(linux_tv)));
1567 			/* NOTREACHED */
1568 			break;
1569 		case LOCAL_PEERCRED:
1570 			if (args->optlen != sizeof(lxu))
1571 				return (EINVAL);
1572 			xulen = sizeof(xu);
1573 			error = kern_getsockopt(td, args->s, bsd_args.level,
1574 			    name, &xu, UIO_SYSSPACE, &xulen);
1575 			if (error)
1576 				return (error);
1577 			/*
1578 			 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1579 			 */
1580 			lxu.pid = 0;
1581 			lxu.uid = xu.cr_uid;
1582 			lxu.gid = xu.cr_gid;
1583 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1584 			/* NOTREACHED */
1585 			break;
1586 		default:
1587 			break;
1588 		}
1589 		break;
1590 	case IPPROTO_IP:
1591 		name = linux_to_bsd_ip_sockopt(args->optname);
1592 		break;
1593 	case IPPROTO_TCP:
1594 		/* Linux TCP option values match BSD's */
1595 		name = args->optname;
1596 		break;
1597 	default:
1598 		name = -1;
1599 		break;
1600 	}
1601 	if (name == -1)
1602 		return (EINVAL);
1603 
1604 	bsd_args.name = name;
1605 	bsd_args.val = PTRIN(args->optval);
1606 	bsd_args.avalsize = PTRIN(args->optlen);
1607 
1608 	if (name == IPV6_NEXTHOP) {
1609 		error = sys_getsockopt(td, &bsd_args);
1610 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1611 	} else
1612 		error = sys_getsockopt(td, &bsd_args);
1613 
1614 	return (error);
1615 }
1616 
1617 /* Argument list sizes for linux_socketcall */
1618 
1619 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1620 
1621 static const unsigned char lxs_args[] = {
1622 	LINUX_AL(0) /* unused*/,	LINUX_AL(3) /* socket */,
1623 	LINUX_AL(3) /* bind */,		LINUX_AL(3) /* connect */,
1624 	LINUX_AL(2) /* listen */,	LINUX_AL(3) /* accept */,
1625 	LINUX_AL(3) /* getsockname */,	LINUX_AL(3) /* getpeername */,
1626 	LINUX_AL(4) /* socketpair */,	LINUX_AL(4) /* send */,
1627 	LINUX_AL(4) /* recv */,		LINUX_AL(6) /* sendto */,
1628 	LINUX_AL(6) /* recvfrom */,	LINUX_AL(2) /* shutdown */,
1629 	LINUX_AL(5) /* setsockopt */,	LINUX_AL(5) /* getsockopt */,
1630 	LINUX_AL(3) /* sendmsg */,	LINUX_AL(3) /* recvmsg */,
1631 	LINUX_AL(4) /* accept4 */
1632 };
1633 
1634 #define	LINUX_AL_SIZE	sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1635 
1636 int
1637 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1638 {
1639 	l_ulong a[6];
1640 	void *arg;
1641 	int error;
1642 
1643 	if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1644 		return (EINVAL);
1645 	error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1646 	if (error)
1647 		return (error);
1648 
1649 	arg = a;
1650 	switch (args->what) {
1651 	case LINUX_SOCKET:
1652 		return (linux_socket(td, arg));
1653 	case LINUX_BIND:
1654 		return (linux_bind(td, arg));
1655 	case LINUX_CONNECT:
1656 		return (linux_connect(td, arg));
1657 	case LINUX_LISTEN:
1658 		return (linux_listen(td, arg));
1659 	case LINUX_ACCEPT:
1660 		return (linux_accept(td, arg));
1661 	case LINUX_GETSOCKNAME:
1662 		return (linux_getsockname(td, arg));
1663 	case LINUX_GETPEERNAME:
1664 		return (linux_getpeername(td, arg));
1665 	case LINUX_SOCKETPAIR:
1666 		return (linux_socketpair(td, arg));
1667 	case LINUX_SEND:
1668 		return (linux_send(td, arg));
1669 	case LINUX_RECV:
1670 		return (linux_recv(td, arg));
1671 	case LINUX_SENDTO:
1672 		return (linux_sendto(td, arg));
1673 	case LINUX_RECVFROM:
1674 		return (linux_recvfrom(td, arg));
1675 	case LINUX_SHUTDOWN:
1676 		return (linux_shutdown(td, arg));
1677 	case LINUX_SETSOCKOPT:
1678 		return (linux_setsockopt(td, arg));
1679 	case LINUX_GETSOCKOPT:
1680 		return (linux_getsockopt(td, arg));
1681 	case LINUX_SENDMSG:
1682 		return (linux_sendmsg(td, arg));
1683 	case LINUX_RECVMSG:
1684 		return (linux_recvmsg(td, arg));
1685 	case LINUX_ACCEPT4:
1686 		return (linux_accept4(td, arg));
1687 	}
1688 
1689 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1690 	return (ENOSYS);
1691 }
1692