xref: /freebsd/sys/compat/linux/linux_socket.c (revision 9336e0699bda8a301cd2bfa37106b6ec5e32012e)
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 <netinet/in.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/ip.h>
57 #ifdef INET6
58 #include <netinet/ip6.h>
59 #include <netinet6/ip6_var.h>
60 #endif
61 
62 #ifdef COMPAT_LINUX32
63 #include <machine/../linux32/linux.h>
64 #include <machine/../linux32/linux32_proto.h>
65 #else
66 #include <machine/../linux/linux.h>
67 #include <machine/../linux/linux_proto.h>
68 #endif
69 #include <compat/linux/linux_socket.h>
70 #include <compat/linux/linux_util.h>
71 
72 static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
73     struct malloc_type *);
74 static int linux_to_bsd_domain(int);
75 
76 /*
77  * Reads a linux sockaddr and does any necessary translation.
78  * Linux sockaddrs don't have a length field, only a family.
79  */
80 static int
81 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
82 {
83 	int osalen = len;
84 
85 	return (do_sa_get(sap, osa, &osalen, M_SONAME));
86 }
87 
88 /*
89  * Copy the osockaddr structure pointed to by osa to kernel, adjust
90  * family and convert to sockaddr.
91  */
92 static int
93 do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
94     struct malloc_type *mtype)
95 {
96 	int error=0, bdom;
97 	struct sockaddr *sa;
98 	struct osockaddr *kosa;
99 	int alloclen;
100 #ifdef INET6
101 	int oldv6size;
102 	struct sockaddr_in6 *sin6;
103 #endif
104 
105 	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
106 		return (EINVAL);
107 
108 	alloclen = *osalen;
109 #ifdef INET6
110 	oldv6size = 0;
111 	/*
112 	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
113 	 * if it's a v4-mapped address, so reserve the proper space
114 	 * for it.
115 	 */
116 	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
117 		alloclen = sizeof (struct sockaddr_in6);
118 		oldv6size = 1;
119 	}
120 #endif
121 
122 	MALLOC(kosa, struct osockaddr *, alloclen, mtype, M_WAITOK);
123 
124 	if ((error = copyin(osa, kosa, *osalen)))
125 		goto out;
126 
127 	bdom = linux_to_bsd_domain(kosa->sa_family);
128 	if (bdom == -1) {
129 		error = EINVAL;
130 		goto out;
131 	}
132 
133 #ifdef INET6
134 	/*
135 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
136 	 * which lacks the scope id compared with RFC2553 one. If we detect
137 	 * the situation, reject the address and write a message to system log.
138 	 *
139 	 * Still accept addresses for which the scope id is not used.
140 	 */
141 	if (oldv6size && bdom == AF_INET6) {
142 		sin6 = (struct sockaddr_in6 *)kosa;
143 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
144 		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
145 		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
146 		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
147 		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
148 		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
149 			sin6->sin6_scope_id = 0;
150 		} else {
151 			log(LOG_DEBUG,
152 			    "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
153 			error = EINVAL;
154 			goto out;
155 		}
156 	} else
157 #endif
158 	if (bdom == AF_INET)
159 		alloclen = sizeof(struct sockaddr_in);
160 
161 	sa = (struct sockaddr *) kosa;
162 	sa->sa_family = bdom;
163 	sa->sa_len = alloclen;
164 
165 	*sap = sa;
166 	*osalen = alloclen;
167 	return (0);
168 
169 out:
170 	FREE(kosa, mtype);
171 	return (error);
172 }
173 
174 static int
175 linux_to_bsd_domain(int domain)
176 {
177 
178 	switch (domain) {
179 	case LINUX_AF_UNSPEC:
180 		return (AF_UNSPEC);
181 	case LINUX_AF_UNIX:
182 		return (AF_LOCAL);
183 	case LINUX_AF_INET:
184 		return (AF_INET);
185 	case LINUX_AF_INET6:
186 		return (AF_INET6);
187 	case LINUX_AF_AX25:
188 		return (AF_CCITT);
189 	case LINUX_AF_IPX:
190 		return (AF_IPX);
191 	case LINUX_AF_APPLETALK:
192 		return (AF_APPLETALK);
193 	}
194 	return (-1);
195 }
196 
197 static int
198 bsd_to_linux_domain(int domain)
199 {
200 
201 	switch (domain) {
202 	case AF_UNSPEC:
203 		return (LINUX_AF_UNSPEC);
204 	case AF_LOCAL:
205 		return (LINUX_AF_UNIX);
206 	case AF_INET:
207 		return (LINUX_AF_INET);
208 	case AF_INET6:
209 		return (LINUX_AF_INET6);
210 	case AF_CCITT:
211 		return (LINUX_AF_AX25);
212 	case AF_IPX:
213 		return (LINUX_AF_IPX);
214 	case AF_APPLETALK:
215 		return (LINUX_AF_APPLETALK);
216 	}
217 	return (-1);
218 }
219 
220 static int
221 linux_to_bsd_sockopt_level(int level)
222 {
223 
224 	switch (level) {
225 	case LINUX_SOL_SOCKET:
226 		return (SOL_SOCKET);
227 	}
228 	return (level);
229 }
230 
231 static int
232 bsd_to_linux_sockopt_level(int level)
233 {
234 
235 	switch (level) {
236 	case SOL_SOCKET:
237 		return (LINUX_SOL_SOCKET);
238 	}
239 	return (level);
240 }
241 
242 static int
243 linux_to_bsd_ip_sockopt(int opt)
244 {
245 
246 	switch (opt) {
247 	case LINUX_IP_TOS:
248 		return (IP_TOS);
249 	case LINUX_IP_TTL:
250 		return (IP_TTL);
251 	case LINUX_IP_OPTIONS:
252 		return (IP_OPTIONS);
253 	case LINUX_IP_MULTICAST_IF:
254 		return (IP_MULTICAST_IF);
255 	case LINUX_IP_MULTICAST_TTL:
256 		return (IP_MULTICAST_TTL);
257 	case LINUX_IP_MULTICAST_LOOP:
258 		return (IP_MULTICAST_LOOP);
259 	case LINUX_IP_ADD_MEMBERSHIP:
260 		return (IP_ADD_MEMBERSHIP);
261 	case LINUX_IP_DROP_MEMBERSHIP:
262 		return (IP_DROP_MEMBERSHIP);
263 	case LINUX_IP_HDRINCL:
264 		return (IP_HDRINCL);
265 	}
266 	return (-1);
267 }
268 
269 static int
270 linux_to_bsd_so_sockopt(int opt)
271 {
272 
273 	switch (opt) {
274 	case LINUX_SO_DEBUG:
275 		return (SO_DEBUG);
276 	case LINUX_SO_REUSEADDR:
277 		return (SO_REUSEADDR);
278 	case LINUX_SO_TYPE:
279 		return (SO_TYPE);
280 	case LINUX_SO_ERROR:
281 		return (SO_ERROR);
282 	case LINUX_SO_DONTROUTE:
283 		return (SO_DONTROUTE);
284 	case LINUX_SO_BROADCAST:
285 		return (SO_BROADCAST);
286 	case LINUX_SO_SNDBUF:
287 		return (SO_SNDBUF);
288 	case LINUX_SO_RCVBUF:
289 		return (SO_RCVBUF);
290 	case LINUX_SO_KEEPALIVE:
291 		return (SO_KEEPALIVE);
292 	case LINUX_SO_OOBINLINE:
293 		return (SO_OOBINLINE);
294 	case LINUX_SO_LINGER:
295 		return (SO_LINGER);
296 	case LINUX_SO_PEERCRED:
297 		return (LOCAL_PEERCRED);
298 	case LINUX_SO_RCVLOWAT:
299 		return (SO_RCVLOWAT);
300 	case LINUX_SO_SNDLOWAT:
301 		return (SO_SNDLOWAT);
302 	case LINUX_SO_RCVTIMEO:
303 		return (SO_RCVTIMEO);
304 	case LINUX_SO_SNDTIMEO:
305 		return (SO_SNDTIMEO);
306 	case LINUX_SO_TIMESTAMP:
307 		return (SO_TIMESTAMP);
308 	case LINUX_SO_ACCEPTCONN:
309 		return (SO_ACCEPTCONN);
310 	}
311 	return (-1);
312 }
313 
314 static int
315 linux_to_bsd_msg_flags(int flags)
316 {
317 	int ret_flags = 0;
318 
319 	if (flags & LINUX_MSG_OOB)
320 		ret_flags |= MSG_OOB;
321 	if (flags & LINUX_MSG_PEEK)
322 		ret_flags |= MSG_PEEK;
323 	if (flags & LINUX_MSG_DONTROUTE)
324 		ret_flags |= MSG_DONTROUTE;
325 	if (flags & LINUX_MSG_CTRUNC)
326 		ret_flags |= MSG_CTRUNC;
327 	if (flags & LINUX_MSG_TRUNC)
328 		ret_flags |= MSG_TRUNC;
329 	if (flags & LINUX_MSG_DONTWAIT)
330 		ret_flags |= MSG_DONTWAIT;
331 	if (flags & LINUX_MSG_EOR)
332 		ret_flags |= MSG_EOR;
333 	if (flags & LINUX_MSG_WAITALL)
334 		ret_flags |= MSG_WAITALL;
335 	if (flags & LINUX_MSG_NOSIGNAL)
336 		ret_flags |= MSG_NOSIGNAL;
337 #if 0 /* not handled */
338 	if (flags & LINUX_MSG_PROXY)
339 		;
340 	if (flags & LINUX_MSG_FIN)
341 		;
342 	if (flags & LINUX_MSG_SYN)
343 		;
344 	if (flags & LINUX_MSG_CONFIRM)
345 		;
346 	if (flags & LINUX_MSG_RST)
347 		;
348 	if (flags & LINUX_MSG_ERRQUEUE)
349 		;
350 #endif
351 	return ret_flags;
352 }
353 
354 /*
355 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
356 * native syscall will fault.  Thus, we don't really need to check the
357 * return values for these functions.
358 */
359 
360 static int
361 bsd_to_linux_sockaddr(struct sockaddr *arg)
362 {
363 	struct sockaddr sa;
364 	size_t sa_len = sizeof(struct sockaddr);
365 	int error;
366 
367 	if ((error = copyin(arg, &sa, sa_len)))
368 		return (error);
369 
370 	*(u_short *)&sa = sa.sa_family;
371 
372 	error = copyout(&sa, arg, sa_len);
373 
374 	return (error);
375 }
376 
377 static int
378 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
379 {
380 	struct sockaddr sa;
381 	size_t sa_len = sizeof(struct sockaddr);
382 	int error;
383 
384 	if ((error = copyin(arg, &sa, sa_len)))
385 		return (error);
386 
387 	sa.sa_family = *(sa_family_t *)&sa;
388 	sa.sa_len = len;
389 
390 	error = copyout(&sa, arg, sa_len);
391 
392 	return (error);
393 }
394 
395 
396 static int
397 linux_sa_put(struct osockaddr *osa)
398 {
399 	struct osockaddr sa;
400 	int error, bdom;
401 
402 	/*
403 	 * Only read/write the osockaddr family part, the rest is
404 	 * not changed.
405 	 */
406 	error = copyin(osa, &sa, sizeof(sa.sa_family));
407 	if (error)
408 		return (error);
409 
410 	bdom = bsd_to_linux_domain(sa.sa_family);
411 	if (bdom == -1)
412 		return (EINVAL);
413 
414 	sa.sa_family = bdom;
415 	error = copyout(&sa, osa, sizeof(sa.sa_family));
416 	if (error)
417 		return (error);
418 
419 	return (0);
420 }
421 
422 static int
423 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
424     enum uio_seg segflg)
425 {
426 	struct mbuf *control;
427 	struct sockaddr *to;
428 	int error;
429 
430 	if (mp->msg_name != NULL) {
431 		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
432 		if (error)
433 			return (error);
434 		mp->msg_name = to;
435 	} else
436 		to = NULL;
437 
438 	if (mp->msg_control != NULL) {
439 		struct cmsghdr *cmsg;
440 
441 		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
442 			error = EINVAL;
443 			goto bad;
444 		}
445 		error = sockargs(&control, mp->msg_control,
446 		    mp->msg_controllen, MT_CONTROL);
447 		if (error)
448 			goto bad;
449 
450 		cmsg = mtod(control, struct cmsghdr *);
451 		cmsg->cmsg_level = linux_to_bsd_sockopt_level(cmsg->cmsg_level);
452 	} else
453 		control = NULL;
454 
455 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
456 	    segflg);
457 
458 bad:
459 	if (to)
460 		FREE(to, M_SONAME);
461 	return (error);
462 }
463 
464 /* Return 0 if IP_HDRINCL is set for the given socket. */
465 static int
466 linux_check_hdrincl(struct thread *td, int s)
467 {
468 	int error, optval, size_val;
469 
470 	size_val = sizeof(optval);
471 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
472 	    &optval, UIO_SYSSPACE, &size_val);
473 	if (error)
474 		return (error);
475 
476 	return (optval == 0);
477 }
478 
479 struct linux_sendto_args {
480 	int s;
481 	l_uintptr_t msg;
482 	int len;
483 	int flags;
484 	l_uintptr_t to;
485 	int tolen;
486 };
487 
488 /*
489  * Updated sendto() when IP_HDRINCL is set:
490  * tweak endian-dependent fields in the IP packet.
491  */
492 static int
493 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
494 {
495 /*
496  * linux_ip_copysize defines how many bytes we should copy
497  * from the beginning of the IP packet before we customize it for BSD.
498  * It should include all the fields we modify (ip_len and ip_off).
499  */
500 #define linux_ip_copysize	8
501 
502 	struct ip *packet;
503 	struct msghdr msg;
504 	struct iovec aiov[1];
505 	int error;
506 
507 	/* Check that the packet isn't too big or too small. */
508 	if (linux_args->len < linux_ip_copysize ||
509 	    linux_args->len > IP_MAXPACKET)
510 		return (EINVAL);
511 
512 	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
513 
514 	/* Make kernel copy of the packet to be sent */
515 	if ((error = copyin(PTRIN(linux_args->msg), packet,
516 	    linux_args->len)))
517 		goto goout;
518 
519 	/* Convert fields from Linux to BSD raw IP socket format */
520 	packet->ip_len = linux_args->len;
521 	packet->ip_off = ntohs(packet->ip_off);
522 
523 	/* Prepare the msghdr and iovec structures describing the new packet */
524 	msg.msg_name = PTRIN(linux_args->to);
525 	msg.msg_namelen = linux_args->tolen;
526 	msg.msg_iov = aiov;
527 	msg.msg_iovlen = 1;
528 	msg.msg_control = NULL;
529 	msg.msg_flags = 0;
530 	aiov[0].iov_base = (char *)packet;
531 	aiov[0].iov_len = linux_args->len;
532 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
533 	    UIO_SYSSPACE);
534 goout:
535 	free(packet, M_TEMP);
536 	return (error);
537 }
538 
539 struct linux_socket_args {
540 	int domain;
541 	int type;
542 	int protocol;
543 };
544 
545 static int
546 linux_socket(struct thread *td, struct linux_socket_args *args)
547 {
548 	struct linux_socket_args linux_args;
549 	struct socket_args /* {
550 		int domain;
551 		int type;
552 		int protocol;
553 	} */ bsd_args;
554 	int error;
555 	int retval_socket;
556 
557 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
558 		return (error);
559 
560 	bsd_args.protocol = linux_args.protocol;
561 	bsd_args.type = linux_args.type;
562 	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
563 	if (bsd_args.domain == -1)
564 		return (EINVAL);
565 
566 	retval_socket = socket(td, &bsd_args);
567 	if (bsd_args.type == SOCK_RAW
568 	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
569 	    && bsd_args.domain == AF_INET
570 	    && retval_socket >= 0) {
571 		/* It's a raw IP socket: set the IP_HDRINCL option. */
572 		int hdrincl;
573 
574 		hdrincl = 1;
575 		/* We ignore any error returned by kern_setsockopt() */
576 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
577 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
578 	}
579 #ifdef INET6
580 	/*
581 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
582 	 * default and some apps depend on this. So, set V6ONLY to 0
583 	 * for Linux apps if the sysctl value is set to 1.
584 	 */
585 	if (bsd_args.domain == PF_INET6 && retval_socket >= 0
586 #ifndef KLD_MODULE
587 	    /*
588 	     * XXX: Avoid undefined symbol error with an IPv4 only
589 	     * kernel.
590 	     */
591 	    && ip6_v6only
592 #endif
593 	    ) {
594 		int v6only;
595 
596 		v6only = 0;
597 		/* We ignore any error returned by setsockopt() */
598 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
599 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
600 	}
601 #endif
602 
603 	return (retval_socket);
604 }
605 
606 struct linux_bind_args {
607 	int s;
608 	l_uintptr_t name;
609 	int namelen;
610 };
611 
612 static int
613 linux_bind(struct thread *td, struct linux_bind_args *args)
614 {
615 	struct linux_bind_args linux_args;
616 	struct sockaddr *sa;
617 	int error;
618 
619 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
620 		return (error);
621 
622 	error = linux_getsockaddr(&sa, PTRIN(linux_args.name),
623 	    linux_args.namelen);
624 	if (error)
625 		return (error);
626 
627 	error = kern_bind(td, linux_args.s, sa);
628 	free(sa, M_SONAME);
629 	if (error == EADDRNOTAVAIL && linux_args.namelen != sizeof(struct sockaddr_in))
630 	   	return (EINVAL);
631 	return (error);
632 }
633 
634 struct linux_connect_args {
635 	int s;
636 	l_uintptr_t name;
637 	int namelen;
638 };
639 int linux_connect(struct thread *, struct linux_connect_args *);
640 
641 int
642 linux_connect(struct thread *td, struct linux_connect_args *args)
643 {
644 	struct linux_connect_args linux_args;
645 	struct socket *so;
646 	struct sockaddr *sa;
647 	u_int fflag;
648 	int error;
649 
650 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
651 		return (error);
652 
653 	error = linux_getsockaddr(&sa,
654 	    (struct osockaddr *)PTRIN(linux_args.name),
655 	    linux_args.namelen);
656 	if (error)
657 		return (error);
658 
659 	error = kern_connect(td, linux_args.s, sa);
660 	free(sa, M_SONAME);
661 	if (error != EISCONN)
662 		return (error);
663 
664 	/*
665 	 * Linux doesn't return EISCONN the first time it occurs,
666 	 * when on a non-blocking socket. Instead it returns the
667 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
668 	 *
669 	 * XXXRW: Instead of using fgetsock(), check that it is a
670 	 * socket and use the file descriptor reference instead of
671 	 * creating a new one.
672 	 */
673 	error = fgetsock(td, linux_args.s, &so, &fflag);
674 	if (error == 0) {
675 		error = EISCONN;
676 		if (fflag & FNONBLOCK) {
677 			SOCK_LOCK(so);
678 			if (so->so_emuldata == 0)
679 				error = so->so_error;
680 			so->so_emuldata = (void *)1;
681 			SOCK_UNLOCK(so);
682 		}
683 		fputsock(so);
684 	}
685 	return (error);
686 }
687 
688 struct linux_listen_args {
689 	int s;
690 	int backlog;
691 };
692 
693 static int
694 linux_listen(struct thread *td, struct linux_listen_args *args)
695 {
696 	struct linux_listen_args linux_args;
697 	struct listen_args /* {
698 		int s;
699 		int backlog;
700 	} */ bsd_args;
701 	int error;
702 
703 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
704 		return (error);
705 
706 	bsd_args.s = linux_args.s;
707 	bsd_args.backlog = linux_args.backlog;
708 	return (listen(td, &bsd_args));
709 }
710 
711 struct linux_accept_args {
712 	int s;
713 	l_uintptr_t addr;
714 	l_uintptr_t namelen;
715 };
716 
717 static int
718 linux_accept(struct thread *td, struct linux_accept_args *args)
719 {
720 	struct linux_accept_args linux_args;
721 	struct accept_args /* {
722 		int	s;
723 		struct sockaddr * __restrict name;
724 		socklen_t * __restrict anamelen;
725 	} */ bsd_args;
726 	int error, fd;
727 
728 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
729 		return (error);
730 
731 	bsd_args.s = linux_args.s;
732 	/* XXX: */
733 	bsd_args.name = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
734 	bsd_args.anamelen = PTRIN(linux_args.namelen);/* XXX */
735 	error = accept(td, &bsd_args);
736 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
737 	if (error) {
738 	   	if (error == EFAULT && linux_args.namelen != sizeof(struct sockaddr_in))
739 		   	return (EINVAL);
740 		return (error);
741 	}
742 	if (linux_args.addr) {
743 		error = linux_sa_put(PTRIN(linux_args.addr));
744 		if (error) {
745 			(void)kern_close(td, td->td_retval[0]);
746 			return (error);
747 		}
748 	}
749 
750 	/*
751 	 * linux appears not to copy flags from the parent socket to the
752 	 * accepted one, so we must clear the flags in the new descriptor.
753 	 * Ignore any errors, because we already have an open fd.
754 	 */
755 	fd = td->td_retval[0];
756 	(void)kern_fcntl(td, fd, F_SETFL, 0);
757 	td->td_retval[0] = fd;
758 	return (0);
759 }
760 
761 struct linux_getsockname_args {
762 	int s;
763 	l_uintptr_t addr;
764 	l_uintptr_t namelen;
765 };
766 
767 static int
768 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
769 {
770 	struct linux_getsockname_args linux_args;
771 	struct getsockname_args /* {
772 		int	fdes;
773 		struct sockaddr * __restrict asa;
774 		socklen_t * __restrict alen;
775 	} */ bsd_args;
776 	int error;
777 
778 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
779 		return (error);
780 
781 	bsd_args.fdes = linux_args.s;
782 	/* XXX: */
783 	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
784 	bsd_args.alen = PTRIN(linux_args.namelen);	/* XXX */
785 	error = getsockname(td, &bsd_args);
786 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
787 	if (error)
788 		return (error);
789 	error = linux_sa_put(PTRIN(linux_args.addr));
790 	if (error)
791 		return (error);
792 	return (0);
793 }
794 
795 struct linux_getpeername_args {
796 	int s;
797 	l_uintptr_t addr;
798 	l_uintptr_t namelen;
799 };
800 
801 static int
802 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
803 {
804 	struct linux_getpeername_args linux_args;
805 	struct getpeername_args /* {
806 		int fdes;
807 		caddr_t asa;
808 		int *alen;
809 	} */ bsd_args;
810 	int error;
811 
812 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
813 		return (error);
814 
815 	bsd_args.fdes = linux_args.s;
816 	bsd_args.asa = (struct sockaddr *)PTRIN(linux_args.addr);
817 	bsd_args.alen = (int *)PTRIN(linux_args.namelen);
818 	error = getpeername(td, &bsd_args);
819 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
820 	if (error)
821 		return (error);
822 	error = linux_sa_put(PTRIN(linux_args.addr));
823 	if (error)
824 		return (error);
825 	return (0);
826 }
827 
828 struct linux_socketpair_args {
829 	int domain;
830 	int type;
831 	int protocol;
832 	l_uintptr_t rsv;
833 };
834 
835 static int
836 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
837 {
838 	struct linux_socketpair_args linux_args;
839 	struct socketpair_args /* {
840 		int domain;
841 		int type;
842 		int protocol;
843 		int *rsv;
844 	} */ bsd_args;
845 	int error;
846 
847 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
848 		return (error);
849 
850 	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
851 	if (bsd_args.domain == -1)
852 		return (EINVAL);
853 
854 	bsd_args.type = linux_args.type;
855 	bsd_args.protocol = linux_args.protocol;
856 	bsd_args.rsv = (int *)PTRIN(linux_args.rsv);
857 	return (socketpair(td, &bsd_args));
858 }
859 
860 struct linux_send_args {
861 	int s;
862 	l_uintptr_t msg;
863 	int len;
864 	int flags;
865 };
866 
867 static int
868 linux_send(struct thread *td, struct linux_send_args *args)
869 {
870 	struct linux_send_args linux_args;
871 	struct sendto_args /* {
872 		int s;
873 		caddr_t buf;
874 		int len;
875 		int flags;
876 		caddr_t to;
877 		int tolen;
878 	} */ bsd_args;
879 	int error;
880 
881 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
882 		return (error);
883 
884 	bsd_args.s = linux_args.s;
885 	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
886 	bsd_args.len = linux_args.len;
887 	bsd_args.flags = linux_args.flags;
888 	bsd_args.to = NULL;
889 	bsd_args.tolen = 0;
890 	return sendto(td, &bsd_args);
891 }
892 
893 struct linux_recv_args {
894 	int s;
895 	l_uintptr_t msg;
896 	int len;
897 	int flags;
898 };
899 
900 static int
901 linux_recv(struct thread *td, struct linux_recv_args *args)
902 {
903 	struct linux_recv_args linux_args;
904 	struct recvfrom_args /* {
905 		int s;
906 		caddr_t buf;
907 		int len;
908 		int flags;
909 		struct sockaddr *from;
910 		socklen_t fromlenaddr;
911 	} */ bsd_args;
912 	int error;
913 
914 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
915 		return (error);
916 
917 	bsd_args.s = linux_args.s;
918 	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
919 	bsd_args.len = linux_args.len;
920 	bsd_args.flags = linux_args.flags;
921 	bsd_args.from = NULL;
922 	bsd_args.fromlenaddr = 0;
923 	return (recvfrom(td, &bsd_args));
924 }
925 
926 static int
927 linux_sendto(struct thread *td, struct linux_sendto_args *args)
928 {
929 	struct linux_sendto_args linux_args;
930 	struct msghdr msg;
931 	struct iovec aiov;
932 	int error;
933 
934 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
935 		return (error);
936 
937 	if (linux_check_hdrincl(td, linux_args.s) == 0)
938 		/* IP_HDRINCL set, tweak the packet before sending */
939 		return (linux_sendto_hdrincl(td, &linux_args));
940 
941 	msg.msg_name = PTRIN(linux_args.to);
942 	msg.msg_namelen = linux_args.tolen;
943 	msg.msg_iov = &aiov;
944 	msg.msg_iovlen = 1;
945 	msg.msg_control = NULL;
946 	msg.msg_flags = 0;
947 	aiov.iov_base = PTRIN(linux_args.msg);
948 	aiov.iov_len = linux_args.len;
949 	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
950 	    UIO_USERSPACE);
951 	return (error);
952 }
953 
954 struct linux_recvfrom_args {
955 	int s;
956 	l_uintptr_t buf;
957 	int len;
958 	int flags;
959 	l_uintptr_t from;
960 	l_uintptr_t fromlen;
961 };
962 
963 static int
964 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
965 {
966 	struct linux_recvfrom_args linux_args;
967 	struct recvfrom_args /* {
968 		int	s;
969 		caddr_t	buf;
970 		size_t	len;
971 		int	flags;
972 		struct sockaddr * __restrict from;
973 		socklen_t * __restrict fromlenaddr;
974 	} */ bsd_args;
975 	size_t len;
976 	int error;
977 
978 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
979 		return (error);
980 
981 	if ((error = copyin(PTRIN(linux_args.fromlen), &len, sizeof(size_t))))
982 		return (error);
983 
984 	bsd_args.s = linux_args.s;
985 	bsd_args.buf = PTRIN(linux_args.buf);
986 	bsd_args.len = linux_args.len;
987 	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
988 	/* XXX: */
989 	bsd_args.from = (struct sockaddr * __restrict)PTRIN(linux_args.from);
990 	bsd_args.fromlenaddr = PTRIN(linux_args.fromlen);/* XXX */
991 
992 	linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
993 	error = recvfrom(td, &bsd_args);
994 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
995 
996 	if (error)
997 		return (error);
998 	if (linux_args.from) {
999 		error = linux_sa_put((struct osockaddr *)
1000 		    PTRIN(linux_args.from));
1001 		if (error)
1002 			return (error);
1003 	}
1004 	return (0);
1005 }
1006 
1007 struct linux_sendmsg_args {
1008 	int s;
1009 	l_uintptr_t msg;
1010 	int flags;
1011 };
1012 
1013 static int
1014 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1015 {
1016 	struct linux_sendmsg_args linux_args;
1017 	struct msghdr msg;
1018 	struct iovec *iov;
1019 	int error;
1020 
1021 	/* XXXTJR sendmsg is broken on amd64 */
1022 
1023 	error = copyin(args, &linux_args, sizeof(linux_args));
1024 	if (error)
1025 		return (error);
1026 	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
1027 	if (error)
1028 		return (error);
1029 
1030 	/*
1031 	 * Some Linux applications (ping) define a non-NULL control data
1032 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1033 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1034 	 * order to handle this case.  This should be checked, but allows the
1035 	 * Linux ping to work.
1036 	 */
1037 	if (msg.msg_control != NULL && msg.msg_controllen == 0)
1038 		msg.msg_control = NULL;
1039 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1040 	if (error)
1041 		return (error);
1042 	msg.msg_iov = iov;
1043 	msg.msg_flags = 0;
1044 	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
1045 	    UIO_USERSPACE);
1046 	free(iov, M_IOV);
1047 	return (error);
1048 }
1049 
1050 struct linux_recvmsg_args {
1051 	int s;
1052 	l_uintptr_t msg;
1053 	int flags;
1054 };
1055 
1056 static int
1057 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1058 {
1059 	struct linux_recvmsg_args linux_args;
1060 	struct recvmsg_args /* {
1061 		int	s;
1062 		struct	msghdr *msg;
1063 		int	flags;
1064 	} */ bsd_args;
1065 	struct msghdr msg;
1066 	struct cmsghdr *cmsg;
1067 	int error;
1068 
1069 	/* XXXTJR recvmsg is broken on amd64 */
1070 
1071 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1072 		return (error);
1073 
1074 	if ((error = copyin(PTRIN(args->msg), &msg, sizeof (msg))))
1075 		return (error);
1076 
1077 	bsd_args.s = linux_args.s;
1078 	bsd_args.msg = PTRIN(linux_args.msg);
1079 	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
1080 	if (msg.msg_name) {
1081 	   	linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1082 		      msg.msg_namelen);
1083 		error = recvmsg(td, &bsd_args);
1084 		bsd_to_linux_sockaddr((struct sockaddr *)msg.msg_name);
1085 	} else
1086 	   	error = recvmsg(td, &bsd_args);
1087 	if (error)
1088 		return (error);
1089 
1090 	if (bsd_args.msg->msg_control != NULL &&
1091 	    bsd_args.msg->msg_controllen > 0) {
1092 		cmsg = (struct cmsghdr*)bsd_args.msg->msg_control;
1093 		cmsg->cmsg_level = bsd_to_linux_sockopt_level(cmsg->cmsg_level);
1094 	}
1095 
1096 	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
1097 	if (error)
1098 		return (error);
1099 	if (msg.msg_name && msg.msg_namelen > 2)
1100 		error = linux_sa_put(msg.msg_name);
1101 	return (error);
1102 }
1103 
1104 struct linux_shutdown_args {
1105 	int s;
1106 	int how;
1107 };
1108 
1109 static int
1110 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1111 {
1112 	struct linux_shutdown_args linux_args;
1113 	struct shutdown_args /* {
1114 		int s;
1115 		int how;
1116 	} */ bsd_args;
1117 	int error;
1118 
1119 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1120 		return (error);
1121 
1122 	bsd_args.s = linux_args.s;
1123 	bsd_args.how = linux_args.how;
1124 	return (shutdown(td, &bsd_args));
1125 }
1126 
1127 struct linux_setsockopt_args {
1128 	int s;
1129 	int level;
1130 	int optname;
1131 	l_uintptr_t optval;
1132 	int optlen;
1133 };
1134 
1135 static int
1136 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1137 {
1138 	struct linux_setsockopt_args linux_args;
1139 	struct setsockopt_args /* {
1140 		int s;
1141 		int level;
1142 		int name;
1143 		caddr_t val;
1144 		int valsize;
1145 	} */ bsd_args;
1146 	int error, name;
1147 
1148 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1149 		return (error);
1150 
1151 	bsd_args.s = linux_args.s;
1152 	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1153 	switch (bsd_args.level) {
1154 	case SOL_SOCKET:
1155 		name = linux_to_bsd_so_sockopt(linux_args.optname);
1156 		break;
1157 	case IPPROTO_IP:
1158 		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1159 		break;
1160 	case IPPROTO_TCP:
1161 		/* Linux TCP option values match BSD's */
1162 		name = linux_args.optname;
1163 		break;
1164 	default:
1165 		name = -1;
1166 		break;
1167 	}
1168 	if (name == -1)
1169 		return (ENOPROTOOPT);
1170 
1171 	bsd_args.name = name;
1172 	bsd_args.val = PTRIN(linux_args.optval);
1173 	bsd_args.valsize = linux_args.optlen;
1174 
1175 	if (name == IPV6_NEXTHOP) {
1176 		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1177 			bsd_args.valsize);
1178 		error = setsockopt(td, &bsd_args);
1179 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1180 	} else
1181 		error = setsockopt(td, &bsd_args);
1182 
1183 	return (error);
1184 }
1185 
1186 struct linux_getsockopt_args {
1187 	int s;
1188 	int level;
1189 	int optname;
1190 	l_uintptr_t optval;
1191 	l_uintptr_t optlen;
1192 };
1193 
1194 static int
1195 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1196 {
1197 	struct linux_getsockopt_args linux_args;
1198 	struct getsockopt_args /* {
1199 		int s;
1200 		int level;
1201 		int name;
1202 		caddr_t val;
1203 		int *avalsize;
1204 	} */ bsd_args;
1205 	int error, name;
1206 
1207 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1208 		return (error);
1209 
1210 	bsd_args.s = linux_args.s;
1211 	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1212 	switch (bsd_args.level) {
1213 	case SOL_SOCKET:
1214 		name = linux_to_bsd_so_sockopt(linux_args.optname);
1215 		break;
1216 	case IPPROTO_IP:
1217 		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1218 		break;
1219 	case IPPROTO_TCP:
1220 		/* Linux TCP option values match BSD's */
1221 		name = linux_args.optname;
1222 		break;
1223 	default:
1224 		name = -1;
1225 		break;
1226 	}
1227 	if (name == -1)
1228 		return (EINVAL);
1229 
1230 	bsd_args.name = name;
1231 	bsd_args.val = PTRIN(linux_args.optval);
1232 	bsd_args.avalsize = PTRIN(linux_args.optlen);
1233 
1234 	if (name == IPV6_NEXTHOP) {
1235 		error = getsockopt(td, &bsd_args);
1236 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1237 	} else
1238 		error = getsockopt(td, &bsd_args);
1239 
1240 	return (error);
1241 }
1242 
1243 int
1244 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1245 {
1246 	void *arg = (void *)(intptr_t)args->args;
1247 
1248 	switch (args->what) {
1249 	case LINUX_SOCKET:
1250 		return (linux_socket(td, arg));
1251 	case LINUX_BIND:
1252 		return (linux_bind(td, arg));
1253 	case LINUX_CONNECT:
1254 		return (linux_connect(td, arg));
1255 	case LINUX_LISTEN:
1256 		return (linux_listen(td, arg));
1257 	case LINUX_ACCEPT:
1258 		return (linux_accept(td, arg));
1259 	case LINUX_GETSOCKNAME:
1260 		return (linux_getsockname(td, arg));
1261 	case LINUX_GETPEERNAME:
1262 		return (linux_getpeername(td, arg));
1263 	case LINUX_SOCKETPAIR:
1264 		return (linux_socketpair(td, arg));
1265 	case LINUX_SEND:
1266 		return (linux_send(td, arg));
1267 	case LINUX_RECV:
1268 		return (linux_recv(td, arg));
1269 	case LINUX_SENDTO:
1270 		return (linux_sendto(td, arg));
1271 	case LINUX_RECVFROM:
1272 		return (linux_recvfrom(td, arg));
1273 	case LINUX_SHUTDOWN:
1274 		return (linux_shutdown(td, arg));
1275 	case LINUX_SETSOCKOPT:
1276 		return (linux_setsockopt(td, arg));
1277 	case LINUX_GETSOCKOPT:
1278 		return (linux_getsockopt(td, arg));
1279 	case LINUX_SENDMSG:
1280 		return (linux_sendmsg(td, arg));
1281 	case LINUX_RECVMSG:
1282 		return (linux_recvmsg(td, arg));
1283 	}
1284 
1285 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1286 	return (ENOSYS);
1287 }
1288