xref: /freebsd/sys/compat/linux/linux_socket.c (revision 830940567b49bb0c08dfaed40418999e76616909)
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
643 	 * default and some apps depend on this. So, set V6ONLY to 0
644 	 * for Linux apps if the sysctl value is set to 1.
645 	 */
646 	if (bsd_args.domain == PF_INET6
647 #ifndef KLD_MODULE
648 	    /*
649 	     * XXX: Avoid undefined symbol error with an IPv4 only
650 	     * kernel.
651 	     */
652 	    && V_ip6_v6only
653 #endif
654 	    ) {
655 		int v6only;
656 
657 		v6only = 0;
658 		/* We ignore any error returned by setsockopt() */
659 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
660 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
661 	}
662 #endif
663 
664 out:
665 	return (retval_socket);
666 }
667 
668 struct linux_bind_args {
669 	int s;
670 	l_uintptr_t name;
671 	int namelen;
672 };
673 
674 static int
675 linux_bind(struct thread *td, struct linux_bind_args *args)
676 {
677 	struct sockaddr *sa;
678 	int error;
679 
680 	error = linux_getsockaddr(&sa, PTRIN(args->name),
681 	    args->namelen);
682 	if (error)
683 		return (error);
684 
685 	error = kern_bind(td, args->s, sa);
686 	free(sa, M_SONAME);
687 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
688 	   	return (EINVAL);
689 	return (error);
690 }
691 
692 struct linux_connect_args {
693 	int s;
694 	l_uintptr_t name;
695 	int namelen;
696 };
697 int linux_connect(struct thread *, struct linux_connect_args *);
698 
699 int
700 linux_connect(struct thread *td, struct linux_connect_args *args)
701 {
702 	struct socket *so;
703 	struct sockaddr *sa;
704 	u_int fflag;
705 	int error;
706 
707 	error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
708 	    args->namelen);
709 	if (error)
710 		return (error);
711 
712 	error = kern_connect(td, args->s, sa);
713 	free(sa, M_SONAME);
714 	if (error != EISCONN)
715 		return (error);
716 
717 	/*
718 	 * Linux doesn't return EISCONN the first time it occurs,
719 	 * when on a non-blocking socket. Instead it returns the
720 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
721 	 *
722 	 * XXXRW: Instead of using fgetsock(), check that it is a
723 	 * socket and use the file descriptor reference instead of
724 	 * creating a new one.
725 	 */
726 	error = fgetsock(td, args->s, &so, &fflag);
727 	if (error == 0) {
728 		error = EISCONN;
729 		if (fflag & FNONBLOCK) {
730 			SOCK_LOCK(so);
731 			if (so->so_emuldata == 0)
732 				error = so->so_error;
733 			so->so_emuldata = (void *)1;
734 			SOCK_UNLOCK(so);
735 		}
736 		fputsock(so);
737 	}
738 	return (error);
739 }
740 
741 struct linux_listen_args {
742 	int s;
743 	int backlog;
744 };
745 
746 static int
747 linux_listen(struct thread *td, struct linux_listen_args *args)
748 {
749 	struct listen_args /* {
750 		int s;
751 		int backlog;
752 	} */ bsd_args;
753 
754 	bsd_args.s = args->s;
755 	bsd_args.backlog = args->backlog;
756 	return (listen(td, &bsd_args));
757 }
758 
759 static int
760 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
761     l_uintptr_t namelen, int flags)
762 {
763 	struct accept_args /* {
764 		int	s;
765 		struct sockaddr * __restrict name;
766 		socklen_t * __restrict anamelen;
767 	} */ bsd_args;
768 	int error;
769 
770 	if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
771 		return (EINVAL);
772 
773 	bsd_args.s = s;
774 	/* XXX: */
775 	bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
776 	bsd_args.anamelen = PTRIN(namelen);/* XXX */
777 	error = accept(td, &bsd_args);
778 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
779 	if (error) {
780 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
781 			return (EINVAL);
782 		return (error);
783 	}
784 
785 	/*
786 	 * linux appears not to copy flags from the parent socket to the
787 	 * accepted one, so we must clear the flags in the new descriptor
788 	 * and apply the requested flags.
789 	 */
790 	error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
791 	if (error)
792 		goto out;
793 	error = linux_set_socket_flags(td, td->td_retval[0], flags);
794 	if (error)
795 		goto out;
796 	if (addr)
797 		error = linux_sa_put(PTRIN(addr));
798 
799 out:
800 	if (error) {
801 		(void)kern_close(td, td->td_retval[0]);
802 		td->td_retval[0] = 0;
803 	}
804 	return (error);
805 }
806 
807 struct linux_accept_args {
808 	int s;
809 	l_uintptr_t addr;
810 	l_uintptr_t namelen;
811 };
812 
813 static int
814 linux_accept(struct thread *td, struct linux_accept_args *args)
815 {
816 
817 	return (linux_accept_common(td, args->s, args->addr,
818 	    args->namelen, 0));
819 }
820 
821 struct linux_accept4_args {
822 	int s;
823 	l_uintptr_t addr;
824 	l_uintptr_t namelen;
825 	int flags;
826 };
827 
828 static int
829 linux_accept4(struct thread *td, struct linux_accept4_args *args)
830 {
831 
832 	return (linux_accept_common(td, args->s, args->addr,
833 	    args->namelen, args->flags));
834 }
835 
836 struct linux_getsockname_args {
837 	int s;
838 	l_uintptr_t addr;
839 	l_uintptr_t namelen;
840 };
841 
842 static 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 = 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 struct linux_getpeername_args {
867 	int s;
868 	l_uintptr_t addr;
869 	l_uintptr_t namelen;
870 };
871 
872 static int
873 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
874 {
875 	struct getpeername_args /* {
876 		int fdes;
877 		caddr_t asa;
878 		int *alen;
879 	} */ bsd_args;
880 	int error;
881 
882 	bsd_args.fdes = args->s;
883 	bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
884 	bsd_args.alen = (int *)PTRIN(args->namelen);
885 	error = getpeername(td, &bsd_args);
886 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
887 	if (error)
888 		return (error);
889 	error = linux_sa_put(PTRIN(args->addr));
890 	if (error)
891 		return (error);
892 	return (0);
893 }
894 
895 struct linux_socketpair_args {
896 	int domain;
897 	int type;
898 	int protocol;
899 	l_uintptr_t rsv;
900 };
901 
902 static int
903 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
904 {
905 	struct socketpair_args /* {
906 		int domain;
907 		int type;
908 		int protocol;
909 		int *rsv;
910 	} */ bsd_args;
911 	int error, socket_flags;
912 	int sv[2];
913 
914 	bsd_args.domain = linux_to_bsd_domain(args->domain);
915 	if (bsd_args.domain != PF_LOCAL)
916 		return (EAFNOSUPPORT);
917 
918 	socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
919 	if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
920 		return (EINVAL);
921 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
922 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
923 		return (EINVAL);
924 
925 	if (args->protocol != 0 && args->protocol != PF_UNIX)
926 
927 		/*
928 		 * Use of PF_UNIX as protocol argument is not right,
929 		 * but Linux does it.
930 		 * Do not map PF_UNIX as its Linux value is identical
931 		 * to FreeBSD one.
932 		 */
933 		return (EPROTONOSUPPORT);
934 	else
935 		bsd_args.protocol = 0;
936 	bsd_args.rsv = (int *)PTRIN(args->rsv);
937 	error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
938 	    bsd_args.protocol, sv);
939 	if (error)
940 		return (error);
941 	error = linux_set_socket_flags(td, sv[0], socket_flags);
942 	if (error)
943 		goto out;
944 	error = linux_set_socket_flags(td, sv[1], socket_flags);
945 	if (error)
946 		goto out;
947 
948 	error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
949 
950 out:
951 	if (error) {
952 		(void)kern_close(td, sv[0]);
953 		(void)kern_close(td, sv[1]);
954 	}
955 	return (error);
956 }
957 
958 struct linux_send_args {
959 	int s;
960 	l_uintptr_t msg;
961 	int len;
962 	int flags;
963 };
964 
965 static int
966 linux_send(struct thread *td, struct linux_send_args *args)
967 {
968 	struct sendto_args /* {
969 		int s;
970 		caddr_t buf;
971 		int len;
972 		int flags;
973 		caddr_t to;
974 		int tolen;
975 	} */ bsd_args;
976 
977 	bsd_args.s = args->s;
978 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
979 	bsd_args.len = args->len;
980 	bsd_args.flags = args->flags;
981 	bsd_args.to = NULL;
982 	bsd_args.tolen = 0;
983 	return sendto(td, &bsd_args);
984 }
985 
986 struct linux_recv_args {
987 	int s;
988 	l_uintptr_t msg;
989 	int len;
990 	int flags;
991 };
992 
993 static int
994 linux_recv(struct thread *td, struct linux_recv_args *args)
995 {
996 	struct recvfrom_args /* {
997 		int s;
998 		caddr_t buf;
999 		int len;
1000 		int flags;
1001 		struct sockaddr *from;
1002 		socklen_t fromlenaddr;
1003 	} */ bsd_args;
1004 
1005 	bsd_args.s = args->s;
1006 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1007 	bsd_args.len = args->len;
1008 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1009 	bsd_args.from = NULL;
1010 	bsd_args.fromlenaddr = 0;
1011 	return (recvfrom(td, &bsd_args));
1012 }
1013 
1014 static int
1015 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1016 {
1017 	struct msghdr msg;
1018 	struct iovec aiov;
1019 	int error;
1020 
1021 	if (linux_check_hdrincl(td, args->s) == 0)
1022 		/* IP_HDRINCL set, tweak the packet before sending */
1023 		return (linux_sendto_hdrincl(td, args));
1024 
1025 	msg.msg_name = PTRIN(args->to);
1026 	msg.msg_namelen = args->tolen;
1027 	msg.msg_iov = &aiov;
1028 	msg.msg_iovlen = 1;
1029 	msg.msg_control = NULL;
1030 	msg.msg_flags = 0;
1031 	aiov.iov_base = PTRIN(args->msg);
1032 	aiov.iov_len = args->len;
1033 	error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1034 	    UIO_USERSPACE);
1035 	return (error);
1036 }
1037 
1038 struct linux_recvfrom_args {
1039 	int s;
1040 	l_uintptr_t buf;
1041 	int len;
1042 	int flags;
1043 	l_uintptr_t from;
1044 	l_uintptr_t fromlen;
1045 };
1046 
1047 static int
1048 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1049 {
1050 	struct recvfrom_args /* {
1051 		int	s;
1052 		caddr_t	buf;
1053 		size_t	len;
1054 		int	flags;
1055 		struct sockaddr * __restrict from;
1056 		socklen_t * __restrict fromlenaddr;
1057 	} */ bsd_args;
1058 	size_t len;
1059 	int error;
1060 
1061 	if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1062 		return (error);
1063 
1064 	bsd_args.s = args->s;
1065 	bsd_args.buf = PTRIN(args->buf);
1066 	bsd_args.len = args->len;
1067 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1068 	/* XXX: */
1069 	bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1070 	bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1071 
1072 	linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1073 	error = recvfrom(td, &bsd_args);
1074 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1075 
1076 	if (error)
1077 		return (error);
1078 	if (args->from) {
1079 		error = linux_sa_put((struct osockaddr *)
1080 		    PTRIN(args->from));
1081 		if (error)
1082 			return (error);
1083 	}
1084 	return (0);
1085 }
1086 
1087 struct linux_sendmsg_args {
1088 	int s;
1089 	l_uintptr_t msg;
1090 	int flags;
1091 };
1092 
1093 static int
1094 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1095 {
1096 	struct cmsghdr *cmsg;
1097 	struct mbuf *control;
1098 	struct msghdr msg;
1099 	struct l_cmsghdr linux_cmsg;
1100 	struct l_cmsghdr *ptr_cmsg;
1101 	struct l_msghdr linux_msg;
1102 	struct iovec *iov;
1103 	socklen_t datalen;
1104 	void *data;
1105 	int error;
1106 
1107 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1108 	if (error)
1109 		return (error);
1110 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1111 	if (error)
1112 		return (error);
1113 
1114 	/*
1115 	 * Some Linux applications (ping) define a non-NULL control data
1116 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1117 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1118 	 * order to handle this case.  This should be checked, but allows the
1119 	 * Linux ping to work.
1120 	 */
1121 	if (msg.msg_control != NULL && msg.msg_controllen == 0)
1122 		msg.msg_control = NULL;
1123 
1124 #ifdef COMPAT_LINUX32
1125 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1126 	    &iov, EMSGSIZE);
1127 #else
1128 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1129 #endif
1130 	if (error)
1131 		return (error);
1132 
1133 	if (msg.msg_control != NULL) {
1134 		error = ENOBUFS;
1135 		cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1136 		control = m_get(M_WAIT, MT_CONTROL);
1137 		if (control == NULL)
1138 			goto bad;
1139 		ptr_cmsg = LINUX_CMSG_FIRSTHDR(&msg);
1140 
1141 		do {
1142 			error = copyin(ptr_cmsg, &linux_cmsg,
1143 			    sizeof(struct l_cmsghdr));
1144 			if (error)
1145 				goto bad;
1146 
1147 			error = EINVAL;
1148 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1149 				goto bad;
1150 
1151 			/*
1152 			 * Now we support only SCM_RIGHTS, so return EINVAL
1153 			 * in any other cmsg_type
1154 			 */
1155 			if ((cmsg->cmsg_type =
1156 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type)) == -1)
1157 				goto bad;
1158 			cmsg->cmsg_level =
1159 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1160 
1161 			datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1162 			cmsg->cmsg_len = CMSG_LEN(datalen);
1163 			data = LINUX_CMSG_DATA(ptr_cmsg);
1164 
1165 			error = ENOBUFS;
1166 			if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg))
1167 				goto bad;
1168 			if (!m_append(control, datalen, (c_caddr_t) data))
1169 				goto bad;
1170 		} while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&msg, ptr_cmsg)));
1171 	} else {
1172 		control = NULL;
1173 		cmsg = NULL;
1174 	}
1175 
1176 	msg.msg_iov = iov;
1177 	msg.msg_flags = 0;
1178 	error = linux_sendit(td, args->s, &msg, args->flags, control,
1179 	    UIO_USERSPACE);
1180 
1181 bad:
1182 	free(iov, M_IOV);
1183 	if (cmsg)
1184 		free(cmsg, M_TEMP);
1185 	return (error);
1186 }
1187 
1188 struct linux_recvmsg_args {
1189 	int s;
1190 	l_uintptr_t msg;
1191 	int flags;
1192 };
1193 
1194 static int
1195 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1196 {
1197 	struct cmsghdr *cm;
1198 	struct msghdr msg;
1199 	struct l_cmsghdr *linux_cmsg = NULL;
1200 	socklen_t datalen, outlen, clen;
1201 	struct l_msghdr linux_msg;
1202 	struct iovec *iov, *uiov;
1203 	struct mbuf *control = NULL;
1204 	struct mbuf **controlp;
1205 	caddr_t outbuf;
1206 	void *data;
1207 	int error, i, fd, fds, *fdp;
1208 
1209 	error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1210 	if (error)
1211 		return (error);
1212 
1213 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1214 	if (error)
1215 		return (error);
1216 
1217 #ifdef COMPAT_LINUX32
1218 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1219 	    &iov, EMSGSIZE);
1220 #else
1221 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1222 #endif
1223 	if (error)
1224 		return (error);
1225 
1226 	if (msg.msg_name) {
1227 		error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1228 		    msg.msg_namelen);
1229 		if (error)
1230 			goto bad;
1231 	}
1232 
1233 	uiov = msg.msg_iov;
1234 	msg.msg_iov = iov;
1235 	controlp = (msg.msg_control != NULL) ? &control : NULL;
1236 	error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1237 	msg.msg_iov = uiov;
1238 	if (error)
1239 		goto bad;
1240 
1241 	error = bsd_to_linux_msghdr(&msg, &linux_msg);
1242 	if (error)
1243 		goto bad;
1244 
1245 	if (linux_msg.msg_name) {
1246 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1247 		    PTRIN(linux_msg.msg_name));
1248 		if (error)
1249 			goto bad;
1250 	}
1251 	if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1252 		error = linux_sa_put(PTRIN(linux_msg.msg_name));
1253 		if (error)
1254 			goto bad;
1255 	}
1256 
1257 	if (control) {
1258 
1259 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1260 		outbuf = PTRIN(linux_msg.msg_control);
1261 		cm = mtod(control, struct cmsghdr *);
1262 		outlen = 0;
1263 		clen = control->m_len;
1264 
1265 		while (cm != NULL) {
1266 
1267 			if ((linux_cmsg->cmsg_type =
1268 			    bsd_to_linux_cmsg_type(cm->cmsg_type)) == -1)
1269 			{
1270 				error = EINVAL;
1271 				goto bad;
1272 			}
1273 			data = CMSG_DATA(cm);
1274 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1275 
1276 			switch (linux_cmsg->cmsg_type)
1277 			{
1278 			case LINUX_SCM_RIGHTS:
1279 				if (outlen + LINUX_CMSG_LEN(datalen) >
1280 				    linux_msg.msg_controllen) {
1281 					if (outlen == 0) {
1282 						error = EMSGSIZE;
1283 						goto bad;
1284 					} else {
1285 						linux_msg.msg_flags |=
1286 						    LINUX_MSG_CTRUNC;
1287 						goto out;
1288 					}
1289 				}
1290 				if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1291 					fds = datalen / sizeof(int);
1292 					fdp = data;
1293 					for (i = 0; i < fds; i++) {
1294 						fd = *fdp++;
1295 						(void)kern_fcntl(td, fd,
1296 						    F_SETFD, FD_CLOEXEC);
1297 					}
1298 				}
1299 				break;
1300 			}
1301 
1302 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1303 			linux_cmsg->cmsg_level =
1304 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1305 
1306 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1307 			if (error)
1308 				goto bad;
1309 			outbuf += L_CMSG_HDRSZ;
1310 
1311 			error = copyout(data, outbuf, datalen);
1312 			if (error)
1313 				goto bad;
1314 
1315 			outbuf += LINUX_CMSG_ALIGN(datalen);
1316 			outlen += LINUX_CMSG_LEN(datalen);
1317 			linux_msg.msg_controllen = outlen;
1318 
1319 			if (CMSG_SPACE(datalen) < clen) {
1320 				clen -= CMSG_SPACE(datalen);
1321 				cm = (struct cmsghdr *)
1322 				    ((caddr_t)cm + CMSG_SPACE(datalen));
1323 			} else
1324 				cm = NULL;
1325 		}
1326 	}
1327 
1328 out:
1329 	error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1330 
1331 bad:
1332 	free(iov, M_IOV);
1333 	if (control != NULL)
1334 		m_freem(control);
1335 	if (linux_cmsg != NULL)
1336 		free(linux_cmsg, M_TEMP);
1337 
1338 	return (error);
1339 }
1340 
1341 struct linux_shutdown_args {
1342 	int s;
1343 	int how;
1344 };
1345 
1346 static int
1347 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1348 {
1349 	struct shutdown_args /* {
1350 		int s;
1351 		int how;
1352 	} */ bsd_args;
1353 
1354 	bsd_args.s = args->s;
1355 	bsd_args.how = args->how;
1356 	return (shutdown(td, &bsd_args));
1357 }
1358 
1359 struct linux_setsockopt_args {
1360 	int s;
1361 	int level;
1362 	int optname;
1363 	l_uintptr_t optval;
1364 	int optlen;
1365 };
1366 
1367 static int
1368 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1369 {
1370 	struct setsockopt_args /* {
1371 		int s;
1372 		int level;
1373 		int name;
1374 		caddr_t val;
1375 		int valsize;
1376 	} */ bsd_args;
1377 	l_timeval linux_tv;
1378 	struct timeval tv;
1379 	int error, name;
1380 
1381 	bsd_args.s = args->s;
1382 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1383 	switch (bsd_args.level) {
1384 	case SOL_SOCKET:
1385 		name = linux_to_bsd_so_sockopt(args->optname);
1386 		switch (name) {
1387 		case SO_RCVTIMEO:
1388 			/* FALLTHROUGH */
1389 		case SO_SNDTIMEO:
1390 			error = copyin(PTRIN(args->optval), &linux_tv,
1391 			    sizeof(linux_tv));
1392 			if (error)
1393 				return (error);
1394 			tv.tv_sec = linux_tv.tv_sec;
1395 			tv.tv_usec = linux_tv.tv_usec;
1396 			return (kern_setsockopt(td, args->s, bsd_args.level,
1397 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1398 			/* NOTREACHED */
1399 			break;
1400 		default:
1401 			break;
1402 		}
1403 		break;
1404 	case IPPROTO_IP:
1405 		name = linux_to_bsd_ip_sockopt(args->optname);
1406 		break;
1407 	case IPPROTO_TCP:
1408 		/* Linux TCP option values match BSD's */
1409 		name = args->optname;
1410 		break;
1411 	default:
1412 		name = -1;
1413 		break;
1414 	}
1415 	if (name == -1)
1416 		return (ENOPROTOOPT);
1417 
1418 	bsd_args.name = name;
1419 	bsd_args.val = PTRIN(args->optval);
1420 	bsd_args.valsize = args->optlen;
1421 
1422 	if (name == IPV6_NEXTHOP) {
1423 		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1424 			bsd_args.valsize);
1425 		error = setsockopt(td, &bsd_args);
1426 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1427 	} else
1428 		error = setsockopt(td, &bsd_args);
1429 
1430 	return (error);
1431 }
1432 
1433 struct linux_getsockopt_args {
1434 	int s;
1435 	int level;
1436 	int optname;
1437 	l_uintptr_t optval;
1438 	l_uintptr_t optlen;
1439 };
1440 
1441 static int
1442 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1443 {
1444 	struct getsockopt_args /* {
1445 		int s;
1446 		int level;
1447 		int name;
1448 		caddr_t val;
1449 		int *avalsize;
1450 	} */ bsd_args;
1451 	l_timeval linux_tv;
1452 	struct timeval tv;
1453 	socklen_t tv_len, xulen;
1454 	struct xucred xu;
1455 	struct l_ucred lxu;
1456 	int error, name;
1457 
1458 	bsd_args.s = args->s;
1459 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1460 	switch (bsd_args.level) {
1461 	case SOL_SOCKET:
1462 		name = linux_to_bsd_so_sockopt(args->optname);
1463 		switch (name) {
1464 		case SO_RCVTIMEO:
1465 			/* FALLTHROUGH */
1466 		case SO_SNDTIMEO:
1467 			tv_len = sizeof(tv);
1468 			error = kern_getsockopt(td, args->s, bsd_args.level,
1469 			    name, &tv, UIO_SYSSPACE, &tv_len);
1470 			if (error)
1471 				return (error);
1472 			linux_tv.tv_sec = tv.tv_sec;
1473 			linux_tv.tv_usec = tv.tv_usec;
1474 			return (copyout(&linux_tv, PTRIN(args->optval),
1475 			    sizeof(linux_tv)));
1476 			/* NOTREACHED */
1477 			break;
1478 		case LOCAL_PEERCRED:
1479 			if (args->optlen != sizeof(lxu))
1480 				return (EINVAL);
1481 			xulen = sizeof(xu);
1482 			error = kern_getsockopt(td, args->s, bsd_args.level,
1483 			    name, &xu, UIO_SYSSPACE, &xulen);
1484 			if (error)
1485 				return (error);
1486 			/*
1487 			 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1488 			 */
1489 			lxu.pid = 0;
1490 			lxu.uid = xu.cr_uid;
1491 			lxu.gid = xu.cr_gid;
1492 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1493 			/* NOTREACHED */
1494 			break;
1495 		default:
1496 			break;
1497 		}
1498 		break;
1499 	case IPPROTO_IP:
1500 		name = linux_to_bsd_ip_sockopt(args->optname);
1501 		break;
1502 	case IPPROTO_TCP:
1503 		/* Linux TCP option values match BSD's */
1504 		name = args->optname;
1505 		break;
1506 	default:
1507 		name = -1;
1508 		break;
1509 	}
1510 	if (name == -1)
1511 		return (EINVAL);
1512 
1513 	bsd_args.name = name;
1514 	bsd_args.val = PTRIN(args->optval);
1515 	bsd_args.avalsize = PTRIN(args->optlen);
1516 
1517 	if (name == IPV6_NEXTHOP) {
1518 		error = getsockopt(td, &bsd_args);
1519 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1520 	} else
1521 		error = getsockopt(td, &bsd_args);
1522 
1523 	return (error);
1524 }
1525 
1526 /* Argument list sizes for linux_socketcall */
1527 
1528 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1529 
1530 static const unsigned char lxs_args[] = {
1531 	LINUX_AL(0) /* unused*/,	LINUX_AL(3) /* socket */,
1532 	LINUX_AL(3) /* bind */,		LINUX_AL(3) /* connect */,
1533 	LINUX_AL(2) /* listen */,	LINUX_AL(3) /* accept */,
1534 	LINUX_AL(3) /* getsockname */,	LINUX_AL(3) /* getpeername */,
1535 	LINUX_AL(4) /* socketpair */,	LINUX_AL(4) /* send */,
1536 	LINUX_AL(4) /* recv */,		LINUX_AL(6) /* sendto */,
1537 	LINUX_AL(6) /* recvfrom */,	LINUX_AL(2) /* shutdown */,
1538 	LINUX_AL(5) /* setsockopt */,	LINUX_AL(5) /* getsockopt */,
1539 	LINUX_AL(3) /* sendmsg */,	LINUX_AL(3) /* recvmsg */,
1540 	LINUX_AL(4) /* accept4 */
1541 };
1542 
1543 #define	LINUX_AL_SIZE	sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1544 
1545 int
1546 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1547 {
1548 	l_ulong a[6];
1549 	void *arg;
1550 	int error;
1551 
1552 	if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1553 		return (EINVAL);
1554 	error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1555 	if (error)
1556 		return (error);
1557 
1558 	arg = a;
1559 	switch (args->what) {
1560 	case LINUX_SOCKET:
1561 		return (linux_socket(td, arg));
1562 	case LINUX_BIND:
1563 		return (linux_bind(td, arg));
1564 	case LINUX_CONNECT:
1565 		return (linux_connect(td, arg));
1566 	case LINUX_LISTEN:
1567 		return (linux_listen(td, arg));
1568 	case LINUX_ACCEPT:
1569 		return (linux_accept(td, arg));
1570 	case LINUX_GETSOCKNAME:
1571 		return (linux_getsockname(td, arg));
1572 	case LINUX_GETPEERNAME:
1573 		return (linux_getpeername(td, arg));
1574 	case LINUX_SOCKETPAIR:
1575 		return (linux_socketpair(td, arg));
1576 	case LINUX_SEND:
1577 		return (linux_send(td, arg));
1578 	case LINUX_RECV:
1579 		return (linux_recv(td, arg));
1580 	case LINUX_SENDTO:
1581 		return (linux_sendto(td, arg));
1582 	case LINUX_RECVFROM:
1583 		return (linux_recvfrom(td, arg));
1584 	case LINUX_SHUTDOWN:
1585 		return (linux_shutdown(td, arg));
1586 	case LINUX_SETSOCKOPT:
1587 		return (linux_setsockopt(td, arg));
1588 	case LINUX_GETSOCKOPT:
1589 		return (linux_getsockopt(td, arg));
1590 	case LINUX_SENDMSG:
1591 		return (linux_sendmsg(td, arg));
1592 	case LINUX_RECVMSG:
1593 		return (linux_recvmsg(td, arg));
1594 	case LINUX_ACCEPT4:
1595 		return (linux_accept4(td, arg));
1596 	}
1597 
1598 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1599 	return (ENOSYS);
1600 }
1601