xref: /freebsd/sys/compat/linux/linux_socket.c (revision 874108aed99d76099ff9eb6c8d830479a504c1ad)
1 /*-
2  * Copyright (c) 1995 S�ren Schmidt
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer
10  *    in this position and unchanged.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. The name of the author may not be used to endorse or promote products
15  *    derived from this software without specific prior written permission
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 /* XXX we use functions that might not exist. */
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 
36 #include <sys/param.h>
37 #include <sys/proc.h>
38 #include <sys/systm.h>
39 #include <sys/sysproto.h>
40 #include <sys/fcntl.h>
41 #include <sys/file.h>
42 #include <sys/limits.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mutex.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/syscallsubr.h>
50 #include <sys/uio.h>
51 #include <sys/syslog.h>
52 #include <sys/un.h>
53 
54 #include <net/if.h>
55 #include <netinet/in.h>
56 #include <netinet/in_systm.h>
57 #include <netinet/ip.h>
58 #ifdef INET6
59 #include <netinet/ip6.h>
60 #include <netinet6/ip6_var.h>
61 #include <netinet6/in6_var.h>
62 #endif
63 
64 #ifdef COMPAT_LINUX32
65 #include <machine/../linux32/linux.h>
66 #include <machine/../linux32/linux32_proto.h>
67 #else
68 #include <machine/../linux/linux.h>
69 #include <machine/../linux/linux_proto.h>
70 #endif
71 #include <compat/linux/linux_socket.h>
72 #include <compat/linux/linux_util.h>
73 
74 static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
75     struct malloc_type *);
76 static int linux_to_bsd_domain(int);
77 
78 /*
79  * Reads a linux sockaddr and does any necessary translation.
80  * Linux sockaddrs don't have a length field, only a family.
81  */
82 static int
83 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
84 {
85 	int osalen = len;
86 
87 	return (do_sa_get(sap, osa, &osalen, M_SONAME));
88 }
89 
90 /*
91  * Copy the osockaddr structure pointed to by osa to kernel, adjust
92  * family and convert to sockaddr.
93  */
94 static int
95 do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
96     struct malloc_type *mtype)
97 {
98 	int error=0, bdom;
99 	struct sockaddr *sa;
100 	struct osockaddr *kosa;
101 	int alloclen;
102 #ifdef INET6
103 	int oldv6size;
104 	struct sockaddr_in6 *sin6;
105 #endif
106 
107 	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
108 		return (EINVAL);
109 
110 	alloclen = *osalen;
111 #ifdef INET6
112 	oldv6size = 0;
113 	/*
114 	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
115 	 * if it's a v4-mapped address, so reserve the proper space
116 	 * for it.
117 	 */
118 	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
119 		alloclen = sizeof (struct sockaddr_in6);
120 		oldv6size = 1;
121 	}
122 #endif
123 
124 	kosa = malloc(alloclen, mtype, M_WAITOK);
125 
126 	if ((error = copyin(osa, kosa, *osalen)))
127 		goto out;
128 
129 	bdom = linux_to_bsd_domain(kosa->sa_family);
130 	if (bdom == -1) {
131 		error = EINVAL;
132 		goto out;
133 	}
134 
135 #ifdef INET6
136 	/*
137 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
138 	 * which lacks the scope id compared with RFC2553 one. If we detect
139 	 * the situation, reject the address and write a message to system log.
140 	 *
141 	 * Still accept addresses for which the scope id is not used.
142 	 */
143 	if (oldv6size && bdom == AF_INET6) {
144 		sin6 = (struct sockaddr_in6 *)kosa;
145 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
146 		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
147 		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
148 		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
149 		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
150 		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
151 			sin6->sin6_scope_id = 0;
152 		} else {
153 			log(LOG_DEBUG,
154 			    "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
155 			error = EINVAL;
156 			goto out;
157 		}
158 	} else
159 #endif
160 	if (bdom == AF_INET)
161 		alloclen = sizeof(struct sockaddr_in);
162 
163 	sa = (struct sockaddr *) kosa;
164 	sa->sa_family = bdom;
165 	sa->sa_len = alloclen;
166 
167 	*sap = sa;
168 	*osalen = alloclen;
169 	return (0);
170 
171 out:
172 	free(kosa, mtype);
173 	return (error);
174 }
175 
176 static int
177 linux_to_bsd_domain(int domain)
178 {
179 
180 	switch (domain) {
181 	case LINUX_AF_UNSPEC:
182 		return (AF_UNSPEC);
183 	case LINUX_AF_UNIX:
184 		return (AF_LOCAL);
185 	case LINUX_AF_INET:
186 		return (AF_INET);
187 	case LINUX_AF_INET6:
188 		return (AF_INET6);
189 	case LINUX_AF_AX25:
190 		return (AF_CCITT);
191 	case LINUX_AF_IPX:
192 		return (AF_IPX);
193 	case LINUX_AF_APPLETALK:
194 		return (AF_APPLETALK);
195 	}
196 	return (-1);
197 }
198 
199 static int
200 bsd_to_linux_domain(int domain)
201 {
202 
203 	switch (domain) {
204 	case AF_UNSPEC:
205 		return (LINUX_AF_UNSPEC);
206 	case AF_LOCAL:
207 		return (LINUX_AF_UNIX);
208 	case AF_INET:
209 		return (LINUX_AF_INET);
210 	case AF_INET6:
211 		return (LINUX_AF_INET6);
212 	case AF_CCITT:
213 		return (LINUX_AF_AX25);
214 	case AF_IPX:
215 		return (LINUX_AF_IPX);
216 	case AF_APPLETALK:
217 		return (LINUX_AF_APPLETALK);
218 	}
219 	return (-1);
220 }
221 
222 static int
223 linux_to_bsd_sockopt_level(int level)
224 {
225 
226 	switch (level) {
227 	case LINUX_SOL_SOCKET:
228 		return (SOL_SOCKET);
229 	}
230 	return (level);
231 }
232 
233 static int
234 bsd_to_linux_sockopt_level(int level)
235 {
236 
237 	switch (level) {
238 	case SOL_SOCKET:
239 		return (LINUX_SOL_SOCKET);
240 	}
241 	return (level);
242 }
243 
244 static int
245 linux_to_bsd_ip_sockopt(int opt)
246 {
247 
248 	switch (opt) {
249 	case LINUX_IP_TOS:
250 		return (IP_TOS);
251 	case LINUX_IP_TTL:
252 		return (IP_TTL);
253 	case LINUX_IP_OPTIONS:
254 		return (IP_OPTIONS);
255 	case LINUX_IP_MULTICAST_IF:
256 		return (IP_MULTICAST_IF);
257 	case LINUX_IP_MULTICAST_TTL:
258 		return (IP_MULTICAST_TTL);
259 	case LINUX_IP_MULTICAST_LOOP:
260 		return (IP_MULTICAST_LOOP);
261 	case LINUX_IP_ADD_MEMBERSHIP:
262 		return (IP_ADD_MEMBERSHIP);
263 	case LINUX_IP_DROP_MEMBERSHIP:
264 		return (IP_DROP_MEMBERSHIP);
265 	case LINUX_IP_HDRINCL:
266 		return (IP_HDRINCL);
267 	}
268 	return (-1);
269 }
270 
271 static int
272 linux_to_bsd_so_sockopt(int opt)
273 {
274 
275 	switch (opt) {
276 	case LINUX_SO_DEBUG:
277 		return (SO_DEBUG);
278 	case LINUX_SO_REUSEADDR:
279 		return (SO_REUSEADDR);
280 	case LINUX_SO_TYPE:
281 		return (SO_TYPE);
282 	case LINUX_SO_ERROR:
283 		return (SO_ERROR);
284 	case LINUX_SO_DONTROUTE:
285 		return (SO_DONTROUTE);
286 	case LINUX_SO_BROADCAST:
287 		return (SO_BROADCAST);
288 	case LINUX_SO_SNDBUF:
289 		return (SO_SNDBUF);
290 	case LINUX_SO_RCVBUF:
291 		return (SO_RCVBUF);
292 	case LINUX_SO_KEEPALIVE:
293 		return (SO_KEEPALIVE);
294 	case LINUX_SO_OOBINLINE:
295 		return (SO_OOBINLINE);
296 	case LINUX_SO_LINGER:
297 		return (SO_LINGER);
298 	case LINUX_SO_PEERCRED:
299 		return (LOCAL_PEERCRED);
300 	case LINUX_SO_RCVLOWAT:
301 		return (SO_RCVLOWAT);
302 	case LINUX_SO_SNDLOWAT:
303 		return (SO_SNDLOWAT);
304 	case LINUX_SO_RCVTIMEO:
305 		return (SO_RCVTIMEO);
306 	case LINUX_SO_SNDTIMEO:
307 		return (SO_SNDTIMEO);
308 	case LINUX_SO_TIMESTAMP:
309 		return (SO_TIMESTAMP);
310 	case LINUX_SO_ACCEPTCONN:
311 		return (SO_ACCEPTCONN);
312 	}
313 	return (-1);
314 }
315 
316 static int
317 linux_to_bsd_msg_flags(int flags)
318 {
319 	int ret_flags = 0;
320 
321 	if (flags & LINUX_MSG_OOB)
322 		ret_flags |= MSG_OOB;
323 	if (flags & LINUX_MSG_PEEK)
324 		ret_flags |= MSG_PEEK;
325 	if (flags & LINUX_MSG_DONTROUTE)
326 		ret_flags |= MSG_DONTROUTE;
327 	if (flags & LINUX_MSG_CTRUNC)
328 		ret_flags |= MSG_CTRUNC;
329 	if (flags & LINUX_MSG_TRUNC)
330 		ret_flags |= MSG_TRUNC;
331 	if (flags & LINUX_MSG_DONTWAIT)
332 		ret_flags |= MSG_DONTWAIT;
333 	if (flags & LINUX_MSG_EOR)
334 		ret_flags |= MSG_EOR;
335 	if (flags & LINUX_MSG_WAITALL)
336 		ret_flags |= MSG_WAITALL;
337 	if (flags & LINUX_MSG_NOSIGNAL)
338 		ret_flags |= MSG_NOSIGNAL;
339 #if 0 /* not handled */
340 	if (flags & LINUX_MSG_PROXY)
341 		;
342 	if (flags & LINUX_MSG_FIN)
343 		;
344 	if (flags & LINUX_MSG_SYN)
345 		;
346 	if (flags & LINUX_MSG_CONFIRM)
347 		;
348 	if (flags & LINUX_MSG_RST)
349 		;
350 	if (flags & LINUX_MSG_ERRQUEUE)
351 		;
352 #endif
353 	return ret_flags;
354 }
355 
356 /*
357 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
358 * native syscall will fault.  Thus, we don't really need to check the
359 * return values for these functions.
360 */
361 
362 static int
363 bsd_to_linux_sockaddr(struct sockaddr *arg)
364 {
365 	struct sockaddr sa;
366 	size_t sa_len = sizeof(struct sockaddr);
367 	int error;
368 
369 	if ((error = copyin(arg, &sa, sa_len)))
370 		return (error);
371 
372 	*(u_short *)&sa = sa.sa_family;
373 
374 	error = copyout(&sa, arg, sa_len);
375 
376 	return (error);
377 }
378 
379 static int
380 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
381 {
382 	struct sockaddr sa;
383 	size_t sa_len = sizeof(struct sockaddr);
384 	int error;
385 
386 	if ((error = copyin(arg, &sa, sa_len)))
387 		return (error);
388 
389 	sa.sa_family = *(sa_family_t *)&sa;
390 	sa.sa_len = len;
391 
392 	error = copyout(&sa, arg, sa_len);
393 
394 	return (error);
395 }
396 
397 
398 static int
399 linux_sa_put(struct osockaddr *osa)
400 {
401 	struct osockaddr sa;
402 	int error, bdom;
403 
404 	/*
405 	 * Only read/write the osockaddr family part, the rest is
406 	 * not changed.
407 	 */
408 	error = copyin(osa, &sa, sizeof(sa.sa_family));
409 	if (error)
410 		return (error);
411 
412 	bdom = bsd_to_linux_domain(sa.sa_family);
413 	if (bdom == -1)
414 		return (EINVAL);
415 
416 	sa.sa_family = bdom;
417 	error = copyout(&sa, osa, sizeof(sa.sa_family));
418 	if (error)
419 		return (error);
420 
421 	return (0);
422 }
423 
424 static int
425 linux_to_bsd_cmsg_type(int cmsg_type)
426 {
427 
428 	switch (cmsg_type) {
429 	case LINUX_SCM_RIGHTS:
430 		return (SCM_RIGHTS);
431 	}
432 	return (-1);
433 }
434 
435 static int
436 bsd_to_linux_cmsg_type(int cmsg_type)
437 {
438 
439 	switch (cmsg_type) {
440 	case SCM_RIGHTS:
441 		return (LINUX_SCM_RIGHTS);
442 	}
443 	return (-1);
444 }
445 
446 static int
447 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
448 {
449 	if (lhdr->msg_controllen > INT_MAX)
450 		return (ENOBUFS);
451 
452 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
453 	bhdr->msg_namelen	= lhdr->msg_namelen;
454 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
455 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
456 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
457 	bhdr->msg_controllen	= lhdr->msg_controllen;
458 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
459 	return (0);
460 }
461 
462 static int
463 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
464 {
465 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
466 	lhdr->msg_namelen	= bhdr->msg_namelen;
467 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
468 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
469 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
470 	lhdr->msg_controllen	= bhdr->msg_controllen;
471 	/* msg_flags skipped */
472 	return (0);
473 }
474 
475 static int
476 linux_set_socket_flags(struct thread *td, int s, int flags)
477 {
478 	int error;
479 
480 	if (flags & LINUX_SOCK_NONBLOCK) {
481 		error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
482 		if (error)
483 			return (error);
484 	}
485 	if (flags & LINUX_SOCK_CLOEXEC) {
486 		error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
487 		if (error)
488 			return (error);
489 	}
490 	return (0);
491 }
492 
493 static int
494 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
495     struct mbuf *control, enum uio_seg segflg)
496 {
497 	struct sockaddr *to;
498 	int error;
499 
500 	if (mp->msg_name != NULL) {
501 		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
502 		if (error)
503 			return (error);
504 		mp->msg_name = to;
505 	} else
506 		to = NULL;
507 
508 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
509 	    segflg);
510 
511 	if (to)
512 		free(to, M_SONAME);
513 	return (error);
514 }
515 
516 /* Return 0 if IP_HDRINCL is set for the given socket. */
517 static int
518 linux_check_hdrincl(struct thread *td, int s)
519 {
520 	int error, optval, size_val;
521 
522 	size_val = sizeof(optval);
523 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
524 	    &optval, UIO_SYSSPACE, &size_val);
525 	if (error)
526 		return (error);
527 
528 	return (optval == 0);
529 }
530 
531 struct linux_sendto_args {
532 	int s;
533 	l_uintptr_t msg;
534 	int len;
535 	int flags;
536 	l_uintptr_t to;
537 	int tolen;
538 };
539 
540 /*
541  * Updated sendto() when IP_HDRINCL is set:
542  * tweak endian-dependent fields in the IP packet.
543  */
544 static int
545 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
546 {
547 /*
548  * linux_ip_copysize defines how many bytes we should copy
549  * from the beginning of the IP packet before we customize it for BSD.
550  * It should include all the fields we modify (ip_len and ip_off).
551  */
552 #define linux_ip_copysize	8
553 
554 	struct ip *packet;
555 	struct msghdr msg;
556 	struct iovec aiov[1];
557 	int error;
558 
559 	/* Check that the packet isn't too big or too small. */
560 	if (linux_args->len < linux_ip_copysize ||
561 	    linux_args->len > IP_MAXPACKET)
562 		return (EINVAL);
563 
564 	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
565 
566 	/* Make kernel copy of the packet to be sent */
567 	if ((error = copyin(PTRIN(linux_args->msg), packet,
568 	    linux_args->len)))
569 		goto goout;
570 
571 	/* Convert fields from Linux to BSD raw IP socket format */
572 	packet->ip_len = linux_args->len;
573 	packet->ip_off = ntohs(packet->ip_off);
574 
575 	/* Prepare the msghdr and iovec structures describing the new packet */
576 	msg.msg_name = PTRIN(linux_args->to);
577 	msg.msg_namelen = linux_args->tolen;
578 	msg.msg_iov = aiov;
579 	msg.msg_iovlen = 1;
580 	msg.msg_control = NULL;
581 	msg.msg_flags = 0;
582 	aiov[0].iov_base = (char *)packet;
583 	aiov[0].iov_len = linux_args->len;
584 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
585 	    NULL, UIO_SYSSPACE);
586 goout:
587 	free(packet, M_TEMP);
588 	return (error);
589 }
590 
591 struct linux_socket_args {
592 	int domain;
593 	int type;
594 	int protocol;
595 };
596 
597 static int
598 linux_socket(struct thread *td, struct linux_socket_args *args)
599 {
600 	struct socket_args /* {
601 		int domain;
602 		int type;
603 		int protocol;
604 	} */ bsd_args;
605 	int retval_socket, socket_flags;
606 
607 	bsd_args.protocol = args->protocol;
608 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
609 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
610 		return (EINVAL);
611 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
612 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
613 		return (EINVAL);
614 	bsd_args.domain = linux_to_bsd_domain(args->domain);
615 	if (bsd_args.domain == -1)
616 		return (EAFNOSUPPORT);
617 
618 	retval_socket = socket(td, &bsd_args);
619 	if (retval_socket)
620 		return (retval_socket);
621 
622 	retval_socket = linux_set_socket_flags(td, td->td_retval[0],
623 	    socket_flags);
624 	if (retval_socket) {
625 		(void)kern_close(td, td->td_retval[0]);
626 		goto out;
627 	}
628 
629 	if (bsd_args.type == SOCK_RAW
630 	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
631 	    && bsd_args.domain == PF_INET) {
632 		/* It's a raw IP socket: set the IP_HDRINCL option. */
633 		int hdrincl;
634 
635 		hdrincl = 1;
636 		/* We ignore any error returned by kern_setsockopt() */
637 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
638 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
639 	}
640 #ifdef INET6
641 	/*
642 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
643 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
644 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
645 	 * sysctl value.
646 	 */
647 	if (bsd_args.domain == PF_INET6) {
648 		int v6only;
649 
650 		v6only = 0;
651 		/* We ignore any error returned by setsockopt() */
652 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
653 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
654 	}
655 #endif
656 
657 out:
658 	return (retval_socket);
659 }
660 
661 struct linux_bind_args {
662 	int s;
663 	l_uintptr_t name;
664 	int namelen;
665 };
666 
667 static int
668 linux_bind(struct thread *td, struct linux_bind_args *args)
669 {
670 	struct sockaddr *sa;
671 	int error;
672 
673 	error = linux_getsockaddr(&sa, PTRIN(args->name),
674 	    args->namelen);
675 	if (error)
676 		return (error);
677 
678 	error = kern_bind(td, args->s, sa);
679 	free(sa, M_SONAME);
680 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
681 	   	return (EINVAL);
682 	return (error);
683 }
684 
685 struct linux_connect_args {
686 	int s;
687 	l_uintptr_t name;
688 	int namelen;
689 };
690 int linux_connect(struct thread *, struct linux_connect_args *);
691 
692 int
693 linux_connect(struct thread *td, struct linux_connect_args *args)
694 {
695 	struct socket *so;
696 	struct sockaddr *sa;
697 	u_int fflag;
698 	int error;
699 
700 	error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
701 	    args->namelen);
702 	if (error)
703 		return (error);
704 
705 	error = kern_connect(td, args->s, sa);
706 	free(sa, M_SONAME);
707 	if (error != EISCONN)
708 		return (error);
709 
710 	/*
711 	 * Linux doesn't return EISCONN the first time it occurs,
712 	 * when on a non-blocking socket. Instead it returns the
713 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
714 	 *
715 	 * XXXRW: Instead of using fgetsock(), check that it is a
716 	 * socket and use the file descriptor reference instead of
717 	 * creating a new one.
718 	 */
719 	error = fgetsock(td, args->s, &so, &fflag);
720 	if (error == 0) {
721 		error = EISCONN;
722 		if (fflag & FNONBLOCK) {
723 			SOCK_LOCK(so);
724 			if (so->so_emuldata == 0)
725 				error = so->so_error;
726 			so->so_emuldata = (void *)1;
727 			SOCK_UNLOCK(so);
728 		}
729 		fputsock(so);
730 	}
731 	return (error);
732 }
733 
734 struct linux_listen_args {
735 	int s;
736 	int backlog;
737 };
738 
739 static int
740 linux_listen(struct thread *td, struct linux_listen_args *args)
741 {
742 	struct listen_args /* {
743 		int s;
744 		int backlog;
745 	} */ bsd_args;
746 
747 	bsd_args.s = args->s;
748 	bsd_args.backlog = args->backlog;
749 	return (listen(td, &bsd_args));
750 }
751 
752 static int
753 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
754     l_uintptr_t namelen, int flags)
755 {
756 	struct accept_args /* {
757 		int	s;
758 		struct sockaddr * __restrict name;
759 		socklen_t * __restrict anamelen;
760 	} */ bsd_args;
761 	int error;
762 
763 	if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
764 		return (EINVAL);
765 
766 	bsd_args.s = s;
767 	/* XXX: */
768 	bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
769 	bsd_args.anamelen = PTRIN(namelen);/* XXX */
770 	error = accept(td, &bsd_args);
771 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
772 	if (error) {
773 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
774 			return (EINVAL);
775 		return (error);
776 	}
777 
778 	/*
779 	 * linux appears not to copy flags from the parent socket to the
780 	 * accepted one, so we must clear the flags in the new descriptor
781 	 * and apply the requested flags.
782 	 */
783 	error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
784 	if (error)
785 		goto out;
786 	error = linux_set_socket_flags(td, td->td_retval[0], flags);
787 	if (error)
788 		goto out;
789 	if (addr)
790 		error = linux_sa_put(PTRIN(addr));
791 
792 out:
793 	if (error) {
794 		(void)kern_close(td, td->td_retval[0]);
795 		td->td_retval[0] = 0;
796 	}
797 	return (error);
798 }
799 
800 struct linux_accept_args {
801 	int s;
802 	l_uintptr_t addr;
803 	l_uintptr_t namelen;
804 };
805 
806 static int
807 linux_accept(struct thread *td, struct linux_accept_args *args)
808 {
809 
810 	return (linux_accept_common(td, args->s, args->addr,
811 	    args->namelen, 0));
812 }
813 
814 struct linux_accept4_args {
815 	int s;
816 	l_uintptr_t addr;
817 	l_uintptr_t namelen;
818 	int flags;
819 };
820 
821 static int
822 linux_accept4(struct thread *td, struct linux_accept4_args *args)
823 {
824 
825 	return (linux_accept_common(td, args->s, args->addr,
826 	    args->namelen, args->flags));
827 }
828 
829 struct linux_getsockname_args {
830 	int s;
831 	l_uintptr_t addr;
832 	l_uintptr_t namelen;
833 };
834 
835 static int
836 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
837 {
838 	struct getsockname_args /* {
839 		int	fdes;
840 		struct sockaddr * __restrict asa;
841 		socklen_t * __restrict alen;
842 	} */ bsd_args;
843 	int error;
844 
845 	bsd_args.fdes = args->s;
846 	/* XXX: */
847 	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
848 	bsd_args.alen = PTRIN(args->namelen);	/* XXX */
849 	error = getsockname(td, &bsd_args);
850 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
851 	if (error)
852 		return (error);
853 	error = linux_sa_put(PTRIN(args->addr));
854 	if (error)
855 		return (error);
856 	return (0);
857 }
858 
859 struct linux_getpeername_args {
860 	int s;
861 	l_uintptr_t addr;
862 	l_uintptr_t namelen;
863 };
864 
865 static int
866 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
867 {
868 	struct getpeername_args /* {
869 		int fdes;
870 		caddr_t asa;
871 		int *alen;
872 	} */ bsd_args;
873 	int error;
874 
875 	bsd_args.fdes = args->s;
876 	bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
877 	bsd_args.alen = (int *)PTRIN(args->namelen);
878 	error = getpeername(td, &bsd_args);
879 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
880 	if (error)
881 		return (error);
882 	error = linux_sa_put(PTRIN(args->addr));
883 	if (error)
884 		return (error);
885 	return (0);
886 }
887 
888 struct linux_socketpair_args {
889 	int domain;
890 	int type;
891 	int protocol;
892 	l_uintptr_t rsv;
893 };
894 
895 static int
896 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
897 {
898 	struct socketpair_args /* {
899 		int domain;
900 		int type;
901 		int protocol;
902 		int *rsv;
903 	} */ bsd_args;
904 	int error, socket_flags;
905 	int sv[2];
906 
907 	bsd_args.domain = linux_to_bsd_domain(args->domain);
908 	if (bsd_args.domain != PF_LOCAL)
909 		return (EAFNOSUPPORT);
910 
911 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
912 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
913 		return (EINVAL);
914 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
915 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
916 		return (EINVAL);
917 
918 	if (args->protocol != 0 && args->protocol != PF_UNIX)
919 
920 		/*
921 		 * Use of PF_UNIX as protocol argument is not right,
922 		 * but Linux does it.
923 		 * Do not map PF_UNIX as its Linux value is identical
924 		 * to FreeBSD one.
925 		 */
926 		return (EPROTONOSUPPORT);
927 	else
928 		bsd_args.protocol = 0;
929 	bsd_args.rsv = (int *)PTRIN(args->rsv);
930 	error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
931 	    bsd_args.protocol, sv);
932 	if (error)
933 		return (error);
934 	error = linux_set_socket_flags(td, sv[0], socket_flags);
935 	if (error)
936 		goto out;
937 	error = linux_set_socket_flags(td, sv[1], socket_flags);
938 	if (error)
939 		goto out;
940 
941 	error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
942 
943 out:
944 	if (error) {
945 		(void)kern_close(td, sv[0]);
946 		(void)kern_close(td, sv[1]);
947 	}
948 	return (error);
949 }
950 
951 struct linux_send_args {
952 	int s;
953 	l_uintptr_t msg;
954 	int len;
955 	int flags;
956 };
957 
958 static int
959 linux_send(struct thread *td, struct linux_send_args *args)
960 {
961 	struct sendto_args /* {
962 		int s;
963 		caddr_t buf;
964 		int len;
965 		int flags;
966 		caddr_t to;
967 		int tolen;
968 	} */ bsd_args;
969 
970 	bsd_args.s = args->s;
971 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
972 	bsd_args.len = args->len;
973 	bsd_args.flags = args->flags;
974 	bsd_args.to = NULL;
975 	bsd_args.tolen = 0;
976 	return sendto(td, &bsd_args);
977 }
978 
979 struct linux_recv_args {
980 	int s;
981 	l_uintptr_t msg;
982 	int len;
983 	int flags;
984 };
985 
986 static int
987 linux_recv(struct thread *td, struct linux_recv_args *args)
988 {
989 	struct recvfrom_args /* {
990 		int s;
991 		caddr_t buf;
992 		int len;
993 		int flags;
994 		struct sockaddr *from;
995 		socklen_t fromlenaddr;
996 	} */ bsd_args;
997 
998 	bsd_args.s = args->s;
999 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1000 	bsd_args.len = args->len;
1001 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1002 	bsd_args.from = NULL;
1003 	bsd_args.fromlenaddr = 0;
1004 	return (recvfrom(td, &bsd_args));
1005 }
1006 
1007 static int
1008 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1009 {
1010 	struct msghdr msg;
1011 	struct iovec aiov;
1012 	int error;
1013 
1014 	if (linux_check_hdrincl(td, args->s) == 0)
1015 		/* IP_HDRINCL set, tweak the packet before sending */
1016 		return (linux_sendto_hdrincl(td, args));
1017 
1018 	msg.msg_name = PTRIN(args->to);
1019 	msg.msg_namelen = args->tolen;
1020 	msg.msg_iov = &aiov;
1021 	msg.msg_iovlen = 1;
1022 	msg.msg_control = NULL;
1023 	msg.msg_flags = 0;
1024 	aiov.iov_base = PTRIN(args->msg);
1025 	aiov.iov_len = args->len;
1026 	error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1027 	    UIO_USERSPACE);
1028 	return (error);
1029 }
1030 
1031 struct linux_recvfrom_args {
1032 	int s;
1033 	l_uintptr_t buf;
1034 	int len;
1035 	int flags;
1036 	l_uintptr_t from;
1037 	l_uintptr_t fromlen;
1038 };
1039 
1040 static int
1041 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1042 {
1043 	struct recvfrom_args /* {
1044 		int	s;
1045 		caddr_t	buf;
1046 		size_t	len;
1047 		int	flags;
1048 		struct sockaddr * __restrict from;
1049 		socklen_t * __restrict fromlenaddr;
1050 	} */ bsd_args;
1051 	size_t len;
1052 	int error;
1053 
1054 	if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1055 		return (error);
1056 
1057 	bsd_args.s = args->s;
1058 	bsd_args.buf = PTRIN(args->buf);
1059 	bsd_args.len = args->len;
1060 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1061 	/* XXX: */
1062 	bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1063 	bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1064 
1065 	linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1066 	error = recvfrom(td, &bsd_args);
1067 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1068 
1069 	if (error)
1070 		return (error);
1071 	if (args->from) {
1072 		error = linux_sa_put((struct osockaddr *)
1073 		    PTRIN(args->from));
1074 		if (error)
1075 			return (error);
1076 	}
1077 	return (0);
1078 }
1079 
1080 struct linux_sendmsg_args {
1081 	int s;
1082 	l_uintptr_t msg;
1083 	int flags;
1084 };
1085 
1086 static int
1087 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1088 {
1089 	struct cmsghdr *cmsg;
1090 	struct mbuf *control;
1091 	struct msghdr msg;
1092 	struct l_cmsghdr linux_cmsg;
1093 	struct l_cmsghdr *ptr_cmsg;
1094 	struct l_msghdr linux_msg;
1095 	struct iovec *iov;
1096 	socklen_t datalen;
1097 	void *data;
1098 	int error;
1099 
1100 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1101 	if (error)
1102 		return (error);
1103 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1104 	if (error)
1105 		return (error);
1106 
1107 	/*
1108 	 * Some Linux applications (ping) define a non-NULL control data
1109 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1110 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1111 	 * order to handle this case.  This should be checked, but allows the
1112 	 * Linux ping to work.
1113 	 */
1114 	if (msg.msg_control != NULL && msg.msg_controllen == 0)
1115 		msg.msg_control = NULL;
1116 
1117 #ifdef COMPAT_LINUX32
1118 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1119 	    &iov, EMSGSIZE);
1120 #else
1121 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1122 #endif
1123 	if (error)
1124 		return (error);
1125 
1126 	if (msg.msg_control != NULL) {
1127 		error = ENOBUFS;
1128 		cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1129 		control = m_get(M_WAIT, MT_CONTROL);
1130 		if (control == NULL)
1131 			goto bad;
1132 		ptr_cmsg = LINUX_CMSG_FIRSTHDR(&msg);
1133 
1134 		do {
1135 			error = copyin(ptr_cmsg, &linux_cmsg,
1136 			    sizeof(struct l_cmsghdr));
1137 			if (error)
1138 				goto bad;
1139 
1140 			error = EINVAL;
1141 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1142 				goto bad;
1143 
1144 			/*
1145 			 * Now we support only SCM_RIGHTS, so return EINVAL
1146 			 * in any other cmsg_type
1147 			 */
1148 			if ((cmsg->cmsg_type =
1149 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type)) == -1)
1150 				goto bad;
1151 			cmsg->cmsg_level =
1152 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1153 
1154 			datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1155 			cmsg->cmsg_len = CMSG_LEN(datalen);
1156 			data = LINUX_CMSG_DATA(ptr_cmsg);
1157 
1158 			error = ENOBUFS;
1159 			if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg))
1160 				goto bad;
1161 			if (!m_append(control, datalen, (c_caddr_t) data))
1162 				goto bad;
1163 		} while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&msg, ptr_cmsg)));
1164 	} else {
1165 		control = NULL;
1166 		cmsg = NULL;
1167 	}
1168 
1169 	msg.msg_iov = iov;
1170 	msg.msg_flags = 0;
1171 	error = linux_sendit(td, args->s, &msg, args->flags, control,
1172 	    UIO_USERSPACE);
1173 
1174 bad:
1175 	free(iov, M_IOV);
1176 	if (cmsg)
1177 		free(cmsg, M_TEMP);
1178 	return (error);
1179 }
1180 
1181 struct linux_recvmsg_args {
1182 	int s;
1183 	l_uintptr_t msg;
1184 	int flags;
1185 };
1186 
1187 static int
1188 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1189 {
1190 	struct cmsghdr *cm;
1191 	struct msghdr msg;
1192 	struct l_cmsghdr *linux_cmsg = NULL;
1193 	socklen_t datalen, outlen, clen;
1194 	struct l_msghdr linux_msg;
1195 	struct iovec *iov, *uiov;
1196 	struct mbuf *control = NULL;
1197 	struct mbuf **controlp;
1198 	caddr_t outbuf;
1199 	void *data;
1200 	int error, i, fd, fds, *fdp;
1201 
1202 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1203 	if (error)
1204 		return (error);
1205 
1206 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1207 	if (error)
1208 		return (error);
1209 
1210 #ifdef COMPAT_LINUX32
1211 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1212 	    &iov, EMSGSIZE);
1213 #else
1214 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1215 #endif
1216 	if (error)
1217 		return (error);
1218 
1219 	if (msg.msg_name) {
1220 		error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1221 		    msg.msg_namelen);
1222 		if (error)
1223 			goto bad;
1224 	}
1225 
1226 	uiov = msg.msg_iov;
1227 	msg.msg_iov = iov;
1228 	controlp = (msg.msg_control != NULL) ? &control : NULL;
1229 	error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1230 	msg.msg_iov = uiov;
1231 	if (error)
1232 		goto bad;
1233 
1234 	error = bsd_to_linux_msghdr(&msg, &linux_msg);
1235 	if (error)
1236 		goto bad;
1237 
1238 	if (linux_msg.msg_name) {
1239 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1240 		    PTRIN(linux_msg.msg_name));
1241 		if (error)
1242 			goto bad;
1243 	}
1244 	if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1245 		error = linux_sa_put(PTRIN(linux_msg.msg_name));
1246 		if (error)
1247 			goto bad;
1248 	}
1249 
1250 	if (control) {
1251 
1252 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1253 		outbuf = PTRIN(linux_msg.msg_control);
1254 		cm = mtod(control, struct cmsghdr *);
1255 		outlen = 0;
1256 		clen = control->m_len;
1257 
1258 		while (cm != NULL) {
1259 
1260 			if ((linux_cmsg->cmsg_type =
1261 			    bsd_to_linux_cmsg_type(cm->cmsg_type)) == -1)
1262 			{
1263 				error = EINVAL;
1264 				goto bad;
1265 			}
1266 			data = CMSG_DATA(cm);
1267 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1268 
1269 			switch (linux_cmsg->cmsg_type)
1270 			{
1271 			case LINUX_SCM_RIGHTS:
1272 				if (outlen + LINUX_CMSG_LEN(datalen) >
1273 				    linux_msg.msg_controllen) {
1274 					if (outlen == 0) {
1275 						error = EMSGSIZE;
1276 						goto bad;
1277 					} else {
1278 						linux_msg.msg_flags |=
1279 						    LINUX_MSG_CTRUNC;
1280 						goto out;
1281 					}
1282 				}
1283 				if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1284 					fds = datalen / sizeof(int);
1285 					fdp = data;
1286 					for (i = 0; i < fds; i++) {
1287 						fd = *fdp++;
1288 						(void)kern_fcntl(td, fd,
1289 						    F_SETFD, FD_CLOEXEC);
1290 					}
1291 				}
1292 				break;
1293 			}
1294 
1295 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1296 			linux_cmsg->cmsg_level =
1297 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1298 
1299 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1300 			if (error)
1301 				goto bad;
1302 			outbuf += L_CMSG_HDRSZ;
1303 
1304 			error = copyout(data, outbuf, datalen);
1305 			if (error)
1306 				goto bad;
1307 
1308 			outbuf += LINUX_CMSG_ALIGN(datalen);
1309 			outlen += LINUX_CMSG_LEN(datalen);
1310 			linux_msg.msg_controllen = outlen;
1311 
1312 			if (CMSG_SPACE(datalen) < clen) {
1313 				clen -= CMSG_SPACE(datalen);
1314 				cm = (struct cmsghdr *)
1315 				    ((caddr_t)cm + CMSG_SPACE(datalen));
1316 			} else
1317 				cm = NULL;
1318 		}
1319 	}
1320 
1321 out:
1322 	error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1323 
1324 bad:
1325 	free(iov, M_IOV);
1326 	if (control != NULL)
1327 		m_freem(control);
1328 	if (linux_cmsg != NULL)
1329 		free(linux_cmsg, M_TEMP);
1330 
1331 	return (error);
1332 }
1333 
1334 struct linux_shutdown_args {
1335 	int s;
1336 	int how;
1337 };
1338 
1339 static int
1340 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1341 {
1342 	struct shutdown_args /* {
1343 		int s;
1344 		int how;
1345 	} */ bsd_args;
1346 
1347 	bsd_args.s = args->s;
1348 	bsd_args.how = args->how;
1349 	return (shutdown(td, &bsd_args));
1350 }
1351 
1352 struct linux_setsockopt_args {
1353 	int s;
1354 	int level;
1355 	int optname;
1356 	l_uintptr_t optval;
1357 	int optlen;
1358 };
1359 
1360 static int
1361 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1362 {
1363 	struct setsockopt_args /* {
1364 		int s;
1365 		int level;
1366 		int name;
1367 		caddr_t val;
1368 		int valsize;
1369 	} */ bsd_args;
1370 	l_timeval linux_tv;
1371 	struct timeval tv;
1372 	int error, name;
1373 
1374 	bsd_args.s = args->s;
1375 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1376 	switch (bsd_args.level) {
1377 	case SOL_SOCKET:
1378 		name = linux_to_bsd_so_sockopt(args->optname);
1379 		switch (name) {
1380 		case SO_RCVTIMEO:
1381 			/* FALLTHROUGH */
1382 		case SO_SNDTIMEO:
1383 			error = copyin(PTRIN(args->optval), &linux_tv,
1384 			    sizeof(linux_tv));
1385 			if (error)
1386 				return (error);
1387 			tv.tv_sec = linux_tv.tv_sec;
1388 			tv.tv_usec = linux_tv.tv_usec;
1389 			return (kern_setsockopt(td, args->s, bsd_args.level,
1390 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1391 			/* NOTREACHED */
1392 			break;
1393 		default:
1394 			break;
1395 		}
1396 		break;
1397 	case IPPROTO_IP:
1398 		name = linux_to_bsd_ip_sockopt(args->optname);
1399 		break;
1400 	case IPPROTO_TCP:
1401 		/* Linux TCP option values match BSD's */
1402 		name = args->optname;
1403 		break;
1404 	default:
1405 		name = -1;
1406 		break;
1407 	}
1408 	if (name == -1)
1409 		return (ENOPROTOOPT);
1410 
1411 	bsd_args.name = name;
1412 	bsd_args.val = PTRIN(args->optval);
1413 	bsd_args.valsize = args->optlen;
1414 
1415 	if (name == IPV6_NEXTHOP) {
1416 		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1417 			bsd_args.valsize);
1418 		error = setsockopt(td, &bsd_args);
1419 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1420 	} else
1421 		error = setsockopt(td, &bsd_args);
1422 
1423 	return (error);
1424 }
1425 
1426 struct linux_getsockopt_args {
1427 	int s;
1428 	int level;
1429 	int optname;
1430 	l_uintptr_t optval;
1431 	l_uintptr_t optlen;
1432 };
1433 
1434 static int
1435 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1436 {
1437 	struct getsockopt_args /* {
1438 		int s;
1439 		int level;
1440 		int name;
1441 		caddr_t val;
1442 		int *avalsize;
1443 	} */ bsd_args;
1444 	l_timeval linux_tv;
1445 	struct timeval tv;
1446 	socklen_t tv_len, xulen;
1447 	struct xucred xu;
1448 	struct l_ucred lxu;
1449 	int error, name;
1450 
1451 	bsd_args.s = args->s;
1452 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1453 	switch (bsd_args.level) {
1454 	case SOL_SOCKET:
1455 		name = linux_to_bsd_so_sockopt(args->optname);
1456 		switch (name) {
1457 		case SO_RCVTIMEO:
1458 			/* FALLTHROUGH */
1459 		case SO_SNDTIMEO:
1460 			tv_len = sizeof(tv);
1461 			error = kern_getsockopt(td, args->s, bsd_args.level,
1462 			    name, &tv, UIO_SYSSPACE, &tv_len);
1463 			if (error)
1464 				return (error);
1465 			linux_tv.tv_sec = tv.tv_sec;
1466 			linux_tv.tv_usec = tv.tv_usec;
1467 			return (copyout(&linux_tv, PTRIN(args->optval),
1468 			    sizeof(linux_tv)));
1469 			/* NOTREACHED */
1470 			break;
1471 		case LOCAL_PEERCRED:
1472 			if (args->optlen != sizeof(lxu))
1473 				return (EINVAL);
1474 			xulen = sizeof(xu);
1475 			error = kern_getsockopt(td, args->s, bsd_args.level,
1476 			    name, &xu, UIO_SYSSPACE, &xulen);
1477 			if (error)
1478 				return (error);
1479 			/*
1480 			 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1481 			 */
1482 			lxu.pid = 0;
1483 			lxu.uid = xu.cr_uid;
1484 			lxu.gid = xu.cr_gid;
1485 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1486 			/* NOTREACHED */
1487 			break;
1488 		default:
1489 			break;
1490 		}
1491 		break;
1492 	case IPPROTO_IP:
1493 		name = linux_to_bsd_ip_sockopt(args->optname);
1494 		break;
1495 	case IPPROTO_TCP:
1496 		/* Linux TCP option values match BSD's */
1497 		name = args->optname;
1498 		break;
1499 	default:
1500 		name = -1;
1501 		break;
1502 	}
1503 	if (name == -1)
1504 		return (EINVAL);
1505 
1506 	bsd_args.name = name;
1507 	bsd_args.val = PTRIN(args->optval);
1508 	bsd_args.avalsize = PTRIN(args->optlen);
1509 
1510 	if (name == IPV6_NEXTHOP) {
1511 		error = getsockopt(td, &bsd_args);
1512 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1513 	} else
1514 		error = getsockopt(td, &bsd_args);
1515 
1516 	return (error);
1517 }
1518 
1519 /* Argument list sizes for linux_socketcall */
1520 
1521 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1522 
1523 static const unsigned char lxs_args[] = {
1524 	LINUX_AL(0) /* unused*/,	LINUX_AL(3) /* socket */,
1525 	LINUX_AL(3) /* bind */,		LINUX_AL(3) /* connect */,
1526 	LINUX_AL(2) /* listen */,	LINUX_AL(3) /* accept */,
1527 	LINUX_AL(3) /* getsockname */,	LINUX_AL(3) /* getpeername */,
1528 	LINUX_AL(4) /* socketpair */,	LINUX_AL(4) /* send */,
1529 	LINUX_AL(4) /* recv */,		LINUX_AL(6) /* sendto */,
1530 	LINUX_AL(6) /* recvfrom */,	LINUX_AL(2) /* shutdown */,
1531 	LINUX_AL(5) /* setsockopt */,	LINUX_AL(5) /* getsockopt */,
1532 	LINUX_AL(3) /* sendmsg */,	LINUX_AL(3) /* recvmsg */,
1533 	LINUX_AL(4) /* accept4 */
1534 };
1535 
1536 #define	LINUX_AL_SIZE	sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1537 
1538 int
1539 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1540 {
1541 	l_ulong a[6];
1542 	void *arg;
1543 	int error;
1544 
1545 	if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1546 		return (EINVAL);
1547 	error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1548 	if (error)
1549 		return (error);
1550 
1551 	arg = a;
1552 	switch (args->what) {
1553 	case LINUX_SOCKET:
1554 		return (linux_socket(td, arg));
1555 	case LINUX_BIND:
1556 		return (linux_bind(td, arg));
1557 	case LINUX_CONNECT:
1558 		return (linux_connect(td, arg));
1559 	case LINUX_LISTEN:
1560 		return (linux_listen(td, arg));
1561 	case LINUX_ACCEPT:
1562 		return (linux_accept(td, arg));
1563 	case LINUX_GETSOCKNAME:
1564 		return (linux_getsockname(td, arg));
1565 	case LINUX_GETPEERNAME:
1566 		return (linux_getpeername(td, arg));
1567 	case LINUX_SOCKETPAIR:
1568 		return (linux_socketpair(td, arg));
1569 	case LINUX_SEND:
1570 		return (linux_send(td, arg));
1571 	case LINUX_RECV:
1572 		return (linux_recv(td, arg));
1573 	case LINUX_SENDTO:
1574 		return (linux_sendto(td, arg));
1575 	case LINUX_RECVFROM:
1576 		return (linux_recvfrom(td, arg));
1577 	case LINUX_SHUTDOWN:
1578 		return (linux_shutdown(td, arg));
1579 	case LINUX_SETSOCKOPT:
1580 		return (linux_setsockopt(td, arg));
1581 	case LINUX_GETSOCKOPT:
1582 		return (linux_getsockopt(td, arg));
1583 	case LINUX_SENDMSG:
1584 		return (linux_sendmsg(td, arg));
1585 	case LINUX_RECVMSG:
1586 		return (linux_recvmsg(td, arg));
1587 	case LINUX_ACCEPT4:
1588 		return (linux_accept4(td, arg));
1589 	}
1590 
1591 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1592 	return (ENOSYS);
1593 }
1594