xref: /freebsd/sys/compat/linux/linux_socket.c (revision 3416500aef140042c64bc149cb1ec6620483bc44)
1 /*-
2  * Copyright (c) 1995 Søren Schmidt
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer
10  *    in this position and unchanged.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. The name of the author may not be used to endorse or promote products
15  *    derived from this software without specific prior written permission
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 /* XXX we use functions that might not exist. */
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 
36 #include <sys/param.h>
37 #include <sys/proc.h>
38 #include <sys/systm.h>
39 #include <sys/sysproto.h>
40 #include <sys/capsicum.h>
41 #include <sys/fcntl.h>
42 #include <sys/file.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mutex.h>
47 #include <sys/mbuf.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/syscallsubr.h>
51 #include <sys/uio.h>
52 #include <sys/syslog.h>
53 #include <sys/un.h>
54 
55 #include <net/if.h>
56 #include <net/vnet.h>
57 #include <netinet/in.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/ip.h>
60 #include <netinet/tcp.h>
61 #ifdef INET6
62 #include <netinet/ip6.h>
63 #include <netinet6/ip6_var.h>
64 #endif
65 
66 #ifdef COMPAT_LINUX32
67 #include <machine/../linux32/linux.h>
68 #include <machine/../linux32/linux32_proto.h>
69 #else
70 #include <machine/../linux/linux.h>
71 #include <machine/../linux/linux_proto.h>
72 #endif
73 #include <compat/linux/linux_file.h>
74 #include <compat/linux/linux_socket.h>
75 #include <compat/linux/linux_timer.h>
76 #include <compat/linux/linux_util.h>
77 
78 static int linux_to_bsd_domain(int);
79 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
80 					l_uint);
81 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
82 					l_uint, struct msghdr *);
83 static int linux_set_socket_flags(int, int *);
84 
85 /*
86  * Reads a linux sockaddr and does any necessary translation.
87  * Linux sockaddrs don't have a length field, only a family.
88  * Copy the osockaddr structure pointed to by osa to kernel, adjust
89  * family and convert to sockaddr.
90  */
91 static int
92 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
93 {
94 	struct sockaddr *sa;
95 	struct osockaddr *kosa;
96 #ifdef INET6
97 	struct sockaddr_in6 *sin6;
98 	int oldv6size;
99 #endif
100 	char *name;
101 	int bdom, error, hdrlen, namelen;
102 
103 	if (salen < 2 || salen > UCHAR_MAX || !osa)
104 		return (EINVAL);
105 
106 #ifdef INET6
107 	oldv6size = 0;
108 	/*
109 	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
110 	 * if it's a v4-mapped address, so reserve the proper space
111 	 * for it.
112 	 */
113 	if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
114 		salen += sizeof(uint32_t);
115 		oldv6size = 1;
116 	}
117 #endif
118 
119 	kosa = malloc(salen, M_SONAME, M_WAITOK);
120 
121 	if ((error = copyin(osa, kosa, salen)))
122 		goto out;
123 
124 	bdom = linux_to_bsd_domain(kosa->sa_family);
125 	if (bdom == -1) {
126 		error = EAFNOSUPPORT;
127 		goto out;
128 	}
129 
130 #ifdef INET6
131 	/*
132 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
133 	 * which lacks the scope id compared with RFC2553 one. If we detect
134 	 * the situation, reject the address and write a message to system log.
135 	 *
136 	 * Still accept addresses for which the scope id is not used.
137 	 */
138 	if (oldv6size) {
139 		if (bdom == AF_INET6) {
140 			sin6 = (struct sockaddr_in6 *)kosa;
141 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
142 			    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
143 			     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
144 			     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
145 			     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
146 			     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
147 				sin6->sin6_scope_id = 0;
148 			} else {
149 				log(LOG_DEBUG,
150 				    "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
151 				error = EINVAL;
152 				goto out;
153 			}
154 		} else
155 			salen -= sizeof(uint32_t);
156 	}
157 #endif
158 	if (bdom == AF_INET) {
159 		if (salen < sizeof(struct sockaddr_in)) {
160 			error = EINVAL;
161 			goto out;
162 		}
163 		salen = sizeof(struct sockaddr_in);
164 	}
165 
166 	if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
167 		hdrlen = offsetof(struct sockaddr_un, sun_path);
168 		name = ((struct sockaddr_un *)kosa)->sun_path;
169 		if (*name == '\0') {
170 			/*
171 		 	 * Linux abstract namespace starts with a NULL byte.
172 			 * XXX We do not support abstract namespace yet.
173 			 */
174 			namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
175 		} else
176 			namelen = strnlen(name, salen - hdrlen);
177 		salen = hdrlen + namelen;
178 		if (salen > sizeof(struct sockaddr_un)) {
179 			error = ENAMETOOLONG;
180 			goto out;
181 		}
182 	}
183 
184 	sa = (struct sockaddr *)kosa;
185 	sa->sa_family = bdom;
186 	sa->sa_len = salen;
187 
188 	*sap = sa;
189 	return (0);
190 
191 out:
192 	free(kosa, M_SONAME);
193 	return (error);
194 }
195 
196 static int
197 linux_to_bsd_domain(int domain)
198 {
199 
200 	switch (domain) {
201 	case LINUX_AF_UNSPEC:
202 		return (AF_UNSPEC);
203 	case LINUX_AF_UNIX:
204 		return (AF_LOCAL);
205 	case LINUX_AF_INET:
206 		return (AF_INET);
207 	case LINUX_AF_INET6:
208 		return (AF_INET6);
209 	case LINUX_AF_AX25:
210 		return (AF_CCITT);
211 	case LINUX_AF_IPX:
212 		return (AF_IPX);
213 	case LINUX_AF_APPLETALK:
214 		return (AF_APPLETALK);
215 	}
216 	return (-1);
217 }
218 
219 static int
220 bsd_to_linux_domain(int domain)
221 {
222 
223 	switch (domain) {
224 	case AF_UNSPEC:
225 		return (LINUX_AF_UNSPEC);
226 	case AF_LOCAL:
227 		return (LINUX_AF_UNIX);
228 	case AF_INET:
229 		return (LINUX_AF_INET);
230 	case AF_INET6:
231 		return (LINUX_AF_INET6);
232 	case AF_CCITT:
233 		return (LINUX_AF_AX25);
234 	case AF_IPX:
235 		return (LINUX_AF_IPX);
236 	case AF_APPLETALK:
237 		return (LINUX_AF_APPLETALK);
238 	}
239 	return (-1);
240 }
241 
242 static int
243 linux_to_bsd_sockopt_level(int level)
244 {
245 
246 	switch (level) {
247 	case LINUX_SOL_SOCKET:
248 		return (SOL_SOCKET);
249 	}
250 	return (level);
251 }
252 
253 static int
254 bsd_to_linux_sockopt_level(int level)
255 {
256 
257 	switch (level) {
258 	case SOL_SOCKET:
259 		return (LINUX_SOL_SOCKET);
260 	}
261 	return (level);
262 }
263 
264 static int
265 linux_to_bsd_ip_sockopt(int opt)
266 {
267 
268 	switch (opt) {
269 	case LINUX_IP_TOS:
270 		return (IP_TOS);
271 	case LINUX_IP_TTL:
272 		return (IP_TTL);
273 	case LINUX_IP_OPTIONS:
274 		return (IP_OPTIONS);
275 	case LINUX_IP_MULTICAST_IF:
276 		return (IP_MULTICAST_IF);
277 	case LINUX_IP_MULTICAST_TTL:
278 		return (IP_MULTICAST_TTL);
279 	case LINUX_IP_MULTICAST_LOOP:
280 		return (IP_MULTICAST_LOOP);
281 	case LINUX_IP_ADD_MEMBERSHIP:
282 		return (IP_ADD_MEMBERSHIP);
283 	case LINUX_IP_DROP_MEMBERSHIP:
284 		return (IP_DROP_MEMBERSHIP);
285 	case LINUX_IP_HDRINCL:
286 		return (IP_HDRINCL);
287 	}
288 	return (-1);
289 }
290 
291 static int
292 linux_to_bsd_ip6_sockopt(int opt)
293 {
294 
295 	switch (opt) {
296 	case LINUX_IPV6_NEXTHOP:
297 		return (IPV6_NEXTHOP);
298 	case LINUX_IPV6_UNICAST_HOPS:
299 		return (IPV6_UNICAST_HOPS);
300 	case LINUX_IPV6_MULTICAST_IF:
301 		return (IPV6_MULTICAST_IF);
302 	case LINUX_IPV6_MULTICAST_HOPS:
303 		return (IPV6_MULTICAST_HOPS);
304 	case LINUX_IPV6_MULTICAST_LOOP:
305 		return (IPV6_MULTICAST_LOOP);
306 	case LINUX_IPV6_ADD_MEMBERSHIP:
307 		return (IPV6_JOIN_GROUP);
308 	case LINUX_IPV6_DROP_MEMBERSHIP:
309 		return (IPV6_LEAVE_GROUP);
310 	case LINUX_IPV6_V6ONLY:
311 		return (IPV6_V6ONLY);
312 	case LINUX_IPV6_DONTFRAG:
313 		return (IPV6_DONTFRAG);
314 #if 0
315 	case LINUX_IPV6_CHECKSUM:
316 		return (IPV6_CHECKSUM);
317 	case LINUX_IPV6_RECVPKTINFO:
318 		return (IPV6_RECVPKTINFO);
319 	case LINUX_IPV6_PKTINFO:
320 		return (IPV6_PKTINFO);
321 	case LINUX_IPV6_RECVHOPLIMIT:
322 		return (IPV6_RECVHOPLIMIT);
323 	case LINUX_IPV6_HOPLIMIT:
324 		return (IPV6_HOPLIMIT);
325 	case LINUX_IPV6_RECVHOPOPTS:
326 		return (IPV6_RECVHOPOPTS);
327 	case LINUX_IPV6_HOPOPTS:
328 		return (IPV6_HOPOPTS);
329 	case LINUX_IPV6_RTHDRDSTOPTS:
330 		return (IPV6_RTHDRDSTOPTS);
331 	case LINUX_IPV6_RECVRTHDR:
332 		return (IPV6_RECVRTHDR);
333 	case LINUX_IPV6_RTHDR:
334 		return (IPV6_RTHDR);
335 	case LINUX_IPV6_RECVDSTOPTS:
336 		return (IPV6_RECVDSTOPTS);
337 	case LINUX_IPV6_DSTOPTS:
338 		return (IPV6_DSTOPTS);
339 	case LINUX_IPV6_RECVPATHMTU:
340 		return (IPV6_RECVPATHMTU);
341 	case LINUX_IPV6_PATHMTU:
342 		return (IPV6_PATHMTU);
343 #endif
344 	}
345 	return (-1);
346 }
347 
348 static int
349 linux_to_bsd_so_sockopt(int opt)
350 {
351 
352 	switch (opt) {
353 	case LINUX_SO_DEBUG:
354 		return (SO_DEBUG);
355 	case LINUX_SO_REUSEADDR:
356 		return (SO_REUSEADDR);
357 	case LINUX_SO_TYPE:
358 		return (SO_TYPE);
359 	case LINUX_SO_ERROR:
360 		return (SO_ERROR);
361 	case LINUX_SO_DONTROUTE:
362 		return (SO_DONTROUTE);
363 	case LINUX_SO_BROADCAST:
364 		return (SO_BROADCAST);
365 	case LINUX_SO_SNDBUF:
366 		return (SO_SNDBUF);
367 	case LINUX_SO_RCVBUF:
368 		return (SO_RCVBUF);
369 	case LINUX_SO_KEEPALIVE:
370 		return (SO_KEEPALIVE);
371 	case LINUX_SO_OOBINLINE:
372 		return (SO_OOBINLINE);
373 	case LINUX_SO_LINGER:
374 		return (SO_LINGER);
375 	case LINUX_SO_PEERCRED:
376 		return (LOCAL_PEERCRED);
377 	case LINUX_SO_RCVLOWAT:
378 		return (SO_RCVLOWAT);
379 	case LINUX_SO_SNDLOWAT:
380 		return (SO_SNDLOWAT);
381 	case LINUX_SO_RCVTIMEO:
382 		return (SO_RCVTIMEO);
383 	case LINUX_SO_SNDTIMEO:
384 		return (SO_SNDTIMEO);
385 	case LINUX_SO_TIMESTAMP:
386 		return (SO_TIMESTAMP);
387 	case LINUX_SO_ACCEPTCONN:
388 		return (SO_ACCEPTCONN);
389 	}
390 	return (-1);
391 }
392 
393 static int
394 linux_to_bsd_tcp_sockopt(int opt)
395 {
396 
397 	switch (opt) {
398 	case LINUX_TCP_NODELAY:
399 		return (TCP_NODELAY);
400 	case LINUX_TCP_MAXSEG:
401 		return (TCP_MAXSEG);
402 	case LINUX_TCP_KEEPIDLE:
403 		return (TCP_KEEPIDLE);
404 	case LINUX_TCP_KEEPINTVL:
405 		return (TCP_KEEPINTVL);
406 	case LINUX_TCP_KEEPCNT:
407 		return (TCP_KEEPCNT);
408 	case LINUX_TCP_MD5SIG:
409 		return (TCP_MD5SIG);
410 	}
411 	return (-1);
412 }
413 
414 static int
415 linux_to_bsd_msg_flags(int flags)
416 {
417 	int ret_flags = 0;
418 
419 	if (flags & LINUX_MSG_OOB)
420 		ret_flags |= MSG_OOB;
421 	if (flags & LINUX_MSG_PEEK)
422 		ret_flags |= MSG_PEEK;
423 	if (flags & LINUX_MSG_DONTROUTE)
424 		ret_flags |= MSG_DONTROUTE;
425 	if (flags & LINUX_MSG_CTRUNC)
426 		ret_flags |= MSG_CTRUNC;
427 	if (flags & LINUX_MSG_TRUNC)
428 		ret_flags |= MSG_TRUNC;
429 	if (flags & LINUX_MSG_DONTWAIT)
430 		ret_flags |= MSG_DONTWAIT;
431 	if (flags & LINUX_MSG_EOR)
432 		ret_flags |= MSG_EOR;
433 	if (flags & LINUX_MSG_WAITALL)
434 		ret_flags |= MSG_WAITALL;
435 	if (flags & LINUX_MSG_NOSIGNAL)
436 		ret_flags |= MSG_NOSIGNAL;
437 #if 0 /* not handled */
438 	if (flags & LINUX_MSG_PROXY)
439 		;
440 	if (flags & LINUX_MSG_FIN)
441 		;
442 	if (flags & LINUX_MSG_SYN)
443 		;
444 	if (flags & LINUX_MSG_CONFIRM)
445 		;
446 	if (flags & LINUX_MSG_RST)
447 		;
448 	if (flags & LINUX_MSG_ERRQUEUE)
449 		;
450 #endif
451 	return (ret_flags);
452 }
453 
454 /*
455 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
456 * native syscall will fault.  Thus, we don't really need to check the
457 * return values for these functions.
458 */
459 
460 static int
461 bsd_to_linux_sockaddr(struct sockaddr *arg)
462 {
463 	struct sockaddr sa;
464 	size_t sa_len = sizeof(struct sockaddr);
465 	int error, bdom;
466 
467 	if ((error = copyin(arg, &sa, sa_len)))
468 		return (error);
469 
470 	bdom = bsd_to_linux_domain(sa.sa_family);
471 	if (bdom == -1)
472 		return (EAFNOSUPPORT);
473 
474 	*(u_short *)&sa = bdom;
475 	return (copyout(&sa, arg, sa_len));
476 }
477 
478 static int
479 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
480 {
481 	struct sockaddr sa;
482 	size_t sa_len = sizeof(struct sockaddr);
483 	int error, bdom;
484 
485 	if ((error = copyin(arg, &sa, sa_len)))
486 		return (error);
487 
488 	bdom = linux_to_bsd_domain(*(sa_family_t *)&sa);
489 	if (bdom == -1)
490 		return (EAFNOSUPPORT);
491 
492 	sa.sa_family = bdom;
493 	sa.sa_len = len;
494 	return (copyout(&sa, arg, sa_len));
495 }
496 
497 static int
498 linux_sa_put(struct osockaddr *osa)
499 {
500 	struct osockaddr sa;
501 	int error, bdom;
502 
503 	/*
504 	 * Only read/write the osockaddr family part, the rest is
505 	 * not changed.
506 	 */
507 	error = copyin(osa, &sa, sizeof(sa.sa_family));
508 	if (error != 0)
509 		return (error);
510 
511 	bdom = bsd_to_linux_domain(sa.sa_family);
512 	if (bdom == -1)
513 		return (EINVAL);
514 
515 	sa.sa_family = bdom;
516 	return (copyout(&sa, osa, sizeof(sa.sa_family)));
517 }
518 
519 static int
520 linux_to_bsd_cmsg_type(int cmsg_type)
521 {
522 
523 	switch (cmsg_type) {
524 	case LINUX_SCM_RIGHTS:
525 		return (SCM_RIGHTS);
526 	case LINUX_SCM_CREDENTIALS:
527 		return (SCM_CREDS);
528 	}
529 	return (-1);
530 }
531 
532 static int
533 bsd_to_linux_cmsg_type(int cmsg_type)
534 {
535 
536 	switch (cmsg_type) {
537 	case SCM_RIGHTS:
538 		return (LINUX_SCM_RIGHTS);
539 	case SCM_CREDS:
540 		return (LINUX_SCM_CREDENTIALS);
541 	case SCM_TIMESTAMP:
542 		return (LINUX_SCM_TIMESTAMP);
543 	}
544 	return (-1);
545 }
546 
547 static int
548 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
549 {
550 	if (lhdr->msg_controllen > INT_MAX)
551 		return (ENOBUFS);
552 
553 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
554 	bhdr->msg_namelen	= lhdr->msg_namelen;
555 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
556 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
557 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
558 
559 	/*
560 	 * msg_controllen is skipped since BSD and LINUX control messages
561 	 * are potentially different sizes (e.g. the cred structure used
562 	 * by SCM_CREDS is different between the two operating system).
563 	 *
564 	 * The caller can set it (if necessary) after converting all the
565 	 * control messages.
566 	 */
567 
568 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
569 	return (0);
570 }
571 
572 static int
573 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
574 {
575 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
576 	lhdr->msg_namelen	= bhdr->msg_namelen;
577 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
578 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
579 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
580 
581 	/*
582 	 * msg_controllen is skipped since BSD and LINUX control messages
583 	 * are potentially different sizes (e.g. the cred structure used
584 	 * by SCM_CREDS is different between the two operating system).
585 	 *
586 	 * The caller can set it (if necessary) after converting all the
587 	 * control messages.
588 	 */
589 
590 	/* msg_flags skipped */
591 	return (0);
592 }
593 
594 static int
595 linux_set_socket_flags(int lflags, int *flags)
596 {
597 
598 	if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
599 		return (EINVAL);
600 	if (lflags & LINUX_SOCK_NONBLOCK)
601 		*flags |= SOCK_NONBLOCK;
602 	if (lflags & LINUX_SOCK_CLOEXEC)
603 		*flags |= SOCK_CLOEXEC;
604 	return (0);
605 }
606 
607 static int
608 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
609     struct mbuf *control, enum uio_seg segflg)
610 {
611 	struct sockaddr *to;
612 	int error;
613 
614 	if (mp->msg_name != NULL) {
615 		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
616 		if (error != 0)
617 			return (error);
618 		mp->msg_name = to;
619 	} else
620 		to = NULL;
621 
622 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
623 	    segflg);
624 
625 	if (to)
626 		free(to, M_SONAME);
627 	return (error);
628 }
629 
630 /* Return 0 if IP_HDRINCL is set for the given socket. */
631 static int
632 linux_check_hdrincl(struct thread *td, int s)
633 {
634 	int error, optval;
635 	socklen_t size_val;
636 
637 	size_val = sizeof(optval);
638 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
639 	    &optval, UIO_SYSSPACE, &size_val);
640 	if (error != 0)
641 		return (error);
642 
643 	return (optval == 0);
644 }
645 
646 /*
647  * Updated sendto() when IP_HDRINCL is set:
648  * tweak endian-dependent fields in the IP packet.
649  */
650 static int
651 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
652 {
653 /*
654  * linux_ip_copysize defines how many bytes we should copy
655  * from the beginning of the IP packet before we customize it for BSD.
656  * It should include all the fields we modify (ip_len and ip_off).
657  */
658 #define linux_ip_copysize	8
659 
660 	struct ip *packet;
661 	struct msghdr msg;
662 	struct iovec aiov[1];
663 	int error;
664 
665 	/* Check that the packet isn't too big or too small. */
666 	if (linux_args->len < linux_ip_copysize ||
667 	    linux_args->len > IP_MAXPACKET)
668 		return (EINVAL);
669 
670 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
671 
672 	/* Make kernel copy of the packet to be sent */
673 	if ((error = copyin(PTRIN(linux_args->msg), packet,
674 	    linux_args->len)))
675 		goto goout;
676 
677 	/* Convert fields from Linux to BSD raw IP socket format */
678 	packet->ip_len = linux_args->len;
679 	packet->ip_off = ntohs(packet->ip_off);
680 
681 	/* Prepare the msghdr and iovec structures describing the new packet */
682 	msg.msg_name = PTRIN(linux_args->to);
683 	msg.msg_namelen = linux_args->tolen;
684 	msg.msg_iov = aiov;
685 	msg.msg_iovlen = 1;
686 	msg.msg_control = NULL;
687 	msg.msg_flags = 0;
688 	aiov[0].iov_base = (char *)packet;
689 	aiov[0].iov_len = linux_args->len;
690 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
691 	    NULL, UIO_SYSSPACE);
692 goout:
693 	free(packet, M_LINUX);
694 	return (error);
695 }
696 
697 int
698 linux_socket(struct thread *td, struct linux_socket_args *args)
699 {
700 	int domain, retval_socket, type;
701 
702 	type = args->type & LINUX_SOCK_TYPE_MASK;
703 	if (type < 0 || type > LINUX_SOCK_MAX)
704 		return (EINVAL);
705 	retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
706 		&type);
707 	if (retval_socket != 0)
708 		return (retval_socket);
709 	domain = linux_to_bsd_domain(args->domain);
710 	if (domain == -1)
711 		return (EAFNOSUPPORT);
712 
713 	retval_socket = kern_socket(td, domain, type, args->protocol);
714 	if (retval_socket)
715 		return (retval_socket);
716 
717 	if (type == SOCK_RAW
718 	    && (args->protocol == IPPROTO_RAW || args->protocol == 0)
719 	    && domain == PF_INET) {
720 		/* It's a raw IP socket: set the IP_HDRINCL option. */
721 		int hdrincl;
722 
723 		hdrincl = 1;
724 		/* We ignore any error returned by kern_setsockopt() */
725 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
726 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
727 	}
728 #ifdef INET6
729 	/*
730 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
731 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
732 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
733 	 * sysctl value.
734 	 */
735 	if (domain == PF_INET6) {
736 		int v6only;
737 
738 		v6only = 0;
739 		/* We ignore any error returned by setsockopt() */
740 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
741 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
742 	}
743 #endif
744 
745 	return (retval_socket);
746 }
747 
748 int
749 linux_bind(struct thread *td, struct linux_bind_args *args)
750 {
751 	struct sockaddr *sa;
752 	int error;
753 
754 	error = linux_getsockaddr(&sa, PTRIN(args->name),
755 	    args->namelen);
756 	if (error != 0)
757 		return (error);
758 
759 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
760 	free(sa, M_SONAME);
761 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
762 	   	return (EINVAL);
763 	return (error);
764 }
765 
766 int
767 linux_connect(struct thread *td, struct linux_connect_args *args)
768 {
769 	cap_rights_t rights;
770 	struct socket *so;
771 	struct sockaddr *sa;
772 	struct file *fp;
773 	u_int fflag;
774 	int error;
775 
776 	error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
777 	    args->namelen);
778 	if (error != 0)
779 		return (error);
780 
781 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
782 	free(sa, M_SONAME);
783 	if (error != EISCONN)
784 		return (error);
785 
786 	/*
787 	 * Linux doesn't return EISCONN the first time it occurs,
788 	 * when on a non-blocking socket. Instead it returns the
789 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
790 	 */
791 	error = getsock_cap(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
792 	    &fp, &fflag, NULL);
793 	if (error != 0)
794 		return (error);
795 
796 	error = EISCONN;
797 	so = fp->f_data;
798 	if (fflag & FNONBLOCK) {
799 		SOCK_LOCK(so);
800 		if (so->so_emuldata == 0)
801 			error = so->so_error;
802 		so->so_emuldata = (void *)1;
803 		SOCK_UNLOCK(so);
804 	}
805 	fdrop(fp, td);
806 
807 	return (error);
808 }
809 
810 int
811 linux_listen(struct thread *td, struct linux_listen_args *args)
812 {
813 
814 	return (kern_listen(td, args->s, args->backlog));
815 }
816 
817 static int
818 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
819     l_uintptr_t namelen, int flags)
820 {
821 	struct accept4_args /* {
822 		int	s;
823 		struct sockaddr * __restrict name;
824 		socklen_t * __restrict anamelen;
825 		int	flags;
826 	} */ bsd_args;
827 	cap_rights_t rights;
828 	struct socket *so;
829 	struct file *fp;
830 	int error, error1;
831 
832 	bsd_args.s = s;
833 	bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
834 	bsd_args.anamelen = PTRIN(namelen);
835 	bsd_args.flags = 0;
836 	error = linux_set_socket_flags(flags, &bsd_args.flags);
837 	if (error != 0)
838 		return (error);
839 	error = sys_accept4(td, &bsd_args);
840 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
841 	if (error != 0) {
842 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
843 			return (EINVAL);
844 		if (error == EINVAL) {
845 			error1 = getsock_cap(td, s,
846 			    cap_rights_init(&rights, CAP_ACCEPT), &fp, NULL, NULL);
847 			if (error1 != 0)
848 				return (error1);
849 			so = fp->f_data;
850 			if (so->so_type == SOCK_DGRAM) {
851 				fdrop(fp, td);
852 				return (EOPNOTSUPP);
853 			}
854 			fdrop(fp, td);
855 		}
856 		return (error);
857 	}
858 	if (addr)
859 		error = linux_sa_put(PTRIN(addr));
860 	if (error != 0) {
861 		(void)kern_close(td, td->td_retval[0]);
862 		td->td_retval[0] = 0;
863 	}
864 	return (error);
865 }
866 
867 int
868 linux_accept(struct thread *td, struct linux_accept_args *args)
869 {
870 
871 	return (linux_accept_common(td, args->s, args->addr,
872 	    args->namelen, 0));
873 }
874 
875 int
876 linux_accept4(struct thread *td, struct linux_accept4_args *args)
877 {
878 
879 	return (linux_accept_common(td, args->s, args->addr,
880 	    args->namelen, args->flags));
881 }
882 
883 int
884 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
885 {
886 	struct getsockname_args /* {
887 		int	fdes;
888 		struct sockaddr * __restrict asa;
889 		socklen_t * __restrict alen;
890 	} */ bsd_args;
891 	int error;
892 
893 	bsd_args.fdes = args->s;
894 	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
895 	bsd_args.alen = PTRIN(args->namelen);
896 	error = sys_getsockname(td, &bsd_args);
897 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
898 	if (error != 0)
899 		return (error);
900 	return (linux_sa_put(PTRIN(args->addr)));
901 }
902 
903 int
904 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
905 {
906 	struct getpeername_args /* {
907 		int fdes;
908 		caddr_t asa;
909 		int *alen;
910 	} */ bsd_args;
911 	int error;
912 
913 	bsd_args.fdes = args->s;
914 	bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
915 	bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
916 	error = sys_getpeername(td, &bsd_args);
917 	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
918 	if (error != 0)
919 		return (error);
920 	return (linux_sa_put(PTRIN(args->addr)));
921 }
922 
923 int
924 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
925 {
926 	struct socketpair_args /* {
927 		int domain;
928 		int type;
929 		int protocol;
930 		int *rsv;
931 	} */ bsd_args;
932 	int error;
933 
934 	bsd_args.domain = linux_to_bsd_domain(args->domain);
935 	if (bsd_args.domain != PF_LOCAL)
936 		return (EAFNOSUPPORT);
937 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
938 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
939 		return (EINVAL);
940 	error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
941 		&bsd_args.type);
942 	if (error != 0)
943 		return (error);
944 	if (args->protocol != 0 && args->protocol != PF_UNIX)
945 
946 		/*
947 		 * Use of PF_UNIX as protocol argument is not right,
948 		 * but Linux does it.
949 		 * Do not map PF_UNIX as its Linux value is identical
950 		 * to FreeBSD one.
951 		 */
952 		return (EPROTONOSUPPORT);
953 	else
954 		bsd_args.protocol = 0;
955 	bsd_args.rsv = (int *)PTRIN(args->rsv);
956 	return (sys_socketpair(td, &bsd_args));
957 }
958 
959 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
960 struct linux_send_args {
961 	register_t s;
962 	register_t msg;
963 	register_t len;
964 	register_t flags;
965 };
966 
967 static int
968 linux_send(struct thread *td, struct linux_send_args *args)
969 {
970 	struct sendto_args /* {
971 		int s;
972 		caddr_t buf;
973 		int len;
974 		int flags;
975 		caddr_t to;
976 		int tolen;
977 	} */ bsd_args;
978 
979 	bsd_args.s = args->s;
980 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
981 	bsd_args.len = args->len;
982 	bsd_args.flags = args->flags;
983 	bsd_args.to = NULL;
984 	bsd_args.tolen = 0;
985 	return (sys_sendto(td, &bsd_args));
986 }
987 
988 struct linux_recv_args {
989 	register_t s;
990 	register_t msg;
991 	register_t len;
992 	register_t flags;
993 };
994 
995 static int
996 linux_recv(struct thread *td, struct linux_recv_args *args)
997 {
998 	struct recvfrom_args /* {
999 		int s;
1000 		caddr_t buf;
1001 		int len;
1002 		int flags;
1003 		struct sockaddr *from;
1004 		socklen_t fromlenaddr;
1005 	} */ bsd_args;
1006 
1007 	bsd_args.s = args->s;
1008 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1009 	bsd_args.len = args->len;
1010 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1011 	bsd_args.from = NULL;
1012 	bsd_args.fromlenaddr = 0;
1013 	return (sys_recvfrom(td, &bsd_args));
1014 }
1015 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1016 
1017 int
1018 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1019 {
1020 	struct msghdr msg;
1021 	struct iovec aiov;
1022 
1023 	if (linux_check_hdrincl(td, args->s) == 0)
1024 		/* IP_HDRINCL set, tweak the packet before sending */
1025 		return (linux_sendto_hdrincl(td, args));
1026 
1027 	msg.msg_name = PTRIN(args->to);
1028 	msg.msg_namelen = args->tolen;
1029 	msg.msg_iov = &aiov;
1030 	msg.msg_iovlen = 1;
1031 	msg.msg_control = NULL;
1032 	msg.msg_flags = 0;
1033 	aiov.iov_base = PTRIN(args->msg);
1034 	aiov.iov_len = args->len;
1035 	return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1036 	    UIO_USERSPACE));
1037 }
1038 
1039 int
1040 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1041 {
1042 	struct msghdr msg;
1043 	struct iovec aiov;
1044 	int error, fromlen;
1045 
1046 	if (PTRIN(args->fromlen) != NULL) {
1047 		error = copyin(PTRIN(args->fromlen), &fromlen,
1048 		    sizeof(fromlen));
1049 		if (error != 0)
1050 			return (error);
1051 		if (fromlen < 0)
1052 			return (EINVAL);
1053 		msg.msg_namelen = fromlen;
1054 	} else
1055 		msg.msg_namelen = 0;
1056 
1057 	msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from);
1058 	msg.msg_iov = &aiov;
1059 	msg.msg_iovlen = 1;
1060 	aiov.iov_base = PTRIN(args->buf);
1061 	aiov.iov_len = args->len;
1062 	msg.msg_control = 0;
1063 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1064 
1065 	error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL);
1066 	if (error != 0)
1067 		return (error);
1068 
1069 	if (PTRIN(args->from) != NULL) {
1070 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1071 		    PTRIN(args->from));
1072 		if (error != 0)
1073 			return (error);
1074 
1075 		error = linux_sa_put((struct osockaddr *)
1076 		    PTRIN(args->from));
1077 	}
1078 
1079 	if (PTRIN(args->fromlen) != NULL)
1080 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1081 		    sizeof(msg.msg_namelen));
1082 
1083 	return (error);
1084 }
1085 
1086 static int
1087 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1088     l_uint flags)
1089 {
1090 	struct cmsghdr *cmsg;
1091 	struct cmsgcred cmcred;
1092 	struct mbuf *control;
1093 	struct msghdr msg;
1094 	struct l_cmsghdr linux_cmsg;
1095 	struct l_cmsghdr *ptr_cmsg;
1096 	struct l_msghdr linux_msg;
1097 	struct iovec *iov;
1098 	socklen_t datalen;
1099 	struct sockaddr *sa;
1100 	sa_family_t sa_family;
1101 	void *data;
1102 	int error;
1103 
1104 	error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1105 	if (error != 0)
1106 		return (error);
1107 
1108 	/*
1109 	 * Some Linux applications (ping) define a non-NULL control data
1110 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1111 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1112 	 * order to handle this case.  This should be checked, but allows the
1113 	 * Linux ping to work.
1114 	 */
1115 	if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1116 		linux_msg.msg_control = PTROUT(NULL);
1117 
1118 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
1119 	if (error != 0)
1120 		return (error);
1121 
1122 #ifdef COMPAT_LINUX32
1123 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1124 	    &iov, EMSGSIZE);
1125 #else
1126 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1127 #endif
1128 	if (error != 0)
1129 		return (error);
1130 
1131 	control = NULL;
1132 	cmsg = NULL;
1133 
1134 	if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1135 		error = kern_getsockname(td, s, &sa, &datalen);
1136 		if (error != 0)
1137 			goto bad;
1138 		sa_family = sa->sa_family;
1139 		free(sa, M_SONAME);
1140 
1141 		error = ENOBUFS;
1142 		cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO);
1143 		control = m_get(M_WAITOK, MT_CONTROL);
1144 
1145 		do {
1146 			error = copyin(ptr_cmsg, &linux_cmsg,
1147 			    sizeof(struct l_cmsghdr));
1148 			if (error != 0)
1149 				goto bad;
1150 
1151 			error = EINVAL;
1152 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1153 				goto bad;
1154 
1155 			/*
1156 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1157 			 * so return EINVAL in any other cmsg_type
1158 			 */
1159 			cmsg->cmsg_type =
1160 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1161 			cmsg->cmsg_level =
1162 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1163 			if (cmsg->cmsg_type == -1
1164 			    || cmsg->cmsg_level != SOL_SOCKET)
1165 				goto bad;
1166 
1167 			/*
1168 			 * Some applications (e.g. pulseaudio) attempt to
1169 			 * send ancillary data even if the underlying protocol
1170 			 * doesn't support it which is not allowed in the
1171 			 * FreeBSD system call interface.
1172 			 */
1173 			if (sa_family != AF_UNIX)
1174 				continue;
1175 
1176 			data = LINUX_CMSG_DATA(ptr_cmsg);
1177 			datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1178 
1179 			switch (cmsg->cmsg_type)
1180 			{
1181 			case SCM_RIGHTS:
1182 				break;
1183 
1184 			case SCM_CREDS:
1185 				data = &cmcred;
1186 				datalen = sizeof(cmcred);
1187 
1188 				/*
1189 				 * The lower levels will fill in the structure
1190 				 */
1191 				bzero(data, datalen);
1192 				break;
1193 			}
1194 
1195 			cmsg->cmsg_len = CMSG_LEN(datalen);
1196 
1197 			error = ENOBUFS;
1198 			if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1199 				goto bad;
1200 			if (!m_append(control, datalen, (c_caddr_t)data))
1201 				goto bad;
1202 		} while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1203 
1204 		if (m_length(control, NULL) == 0) {
1205 			m_freem(control);
1206 			control = NULL;
1207 		}
1208 	}
1209 
1210 	msg.msg_iov = iov;
1211 	msg.msg_flags = 0;
1212 	error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1213 	control = NULL;
1214 
1215 bad:
1216 	m_freem(control);
1217 	free(iov, M_IOV);
1218 	if (cmsg)
1219 		free(cmsg, M_LINUX);
1220 	return (error);
1221 }
1222 
1223 int
1224 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1225 {
1226 
1227 	return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1228 	    args->flags));
1229 }
1230 
1231 int
1232 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1233 {
1234 	struct l_mmsghdr *msg;
1235 	l_uint retval;
1236 	int error, datagrams;
1237 
1238 	if (args->vlen > UIO_MAXIOV)
1239 		args->vlen = UIO_MAXIOV;
1240 
1241 	msg = PTRIN(args->msg);
1242 	datagrams = 0;
1243 	while (datagrams < args->vlen) {
1244 		error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1245 		    args->flags);
1246 		if (error != 0)
1247 			break;
1248 
1249 		retval = td->td_retval[0];
1250 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1251 		if (error != 0)
1252 			break;
1253 		++msg;
1254 		++datagrams;
1255 	}
1256 	if (error == 0)
1257 		td->td_retval[0] = datagrams;
1258 	return (error);
1259 }
1260 
1261 static int
1262 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1263     l_uint flags, struct msghdr *msg)
1264 {
1265 	struct cmsghdr *cm;
1266 	struct cmsgcred *cmcred;
1267 	struct l_cmsghdr *linux_cmsg = NULL;
1268 	struct l_ucred linux_ucred;
1269 	socklen_t datalen, outlen;
1270 	struct l_msghdr linux_msg;
1271 	struct iovec *iov, *uiov;
1272 	struct mbuf *control = NULL;
1273 	struct mbuf **controlp;
1274 	struct timeval *ftmvl;
1275 	l_timeval ltmvl;
1276 	caddr_t outbuf;
1277 	void *data;
1278 	int error, i, fd, fds, *fdp;
1279 
1280 	error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1281 	if (error != 0)
1282 		return (error);
1283 
1284 	error = linux_to_bsd_msghdr(msg, &linux_msg);
1285 	if (error != 0)
1286 		return (error);
1287 
1288 #ifdef COMPAT_LINUX32
1289 	error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1290 	    &iov, EMSGSIZE);
1291 #else
1292 	error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1293 #endif
1294 	if (error != 0)
1295 		return (error);
1296 
1297 	if (msg->msg_name) {
1298 		error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name,
1299 		    msg->msg_namelen);
1300 		if (error != 0)
1301 			goto bad;
1302 	}
1303 
1304 	uiov = msg->msg_iov;
1305 	msg->msg_iov = iov;
1306 	controlp = (msg->msg_control != NULL) ? &control : NULL;
1307 	error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp);
1308 	msg->msg_iov = uiov;
1309 	if (error != 0)
1310 		goto bad;
1311 
1312 	error = bsd_to_linux_msghdr(msg, &linux_msg);
1313 	if (error != 0)
1314 		goto bad;
1315 
1316 	if (linux_msg.msg_name) {
1317 		error = bsd_to_linux_sockaddr((struct sockaddr *)
1318 		    PTRIN(linux_msg.msg_name));
1319 		if (error != 0)
1320 			goto bad;
1321 	}
1322 	if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1323 		error = linux_sa_put(PTRIN(linux_msg.msg_name));
1324 		if (error != 0)
1325 			goto bad;
1326 	}
1327 
1328 	outbuf = PTRIN(linux_msg.msg_control);
1329 	outlen = 0;
1330 
1331 	if (control) {
1332 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1333 
1334 		msg->msg_control = mtod(control, struct cmsghdr *);
1335 		msg->msg_controllen = control->m_len;
1336 
1337 		cm = CMSG_FIRSTHDR(msg);
1338 
1339 		while (cm != NULL) {
1340 			linux_cmsg->cmsg_type =
1341 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1342 			linux_cmsg->cmsg_level =
1343 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1344 			if (linux_cmsg->cmsg_type == -1
1345 			    || cm->cmsg_level != SOL_SOCKET)
1346 			{
1347 				error = EINVAL;
1348 				goto bad;
1349 			}
1350 
1351 			data = CMSG_DATA(cm);
1352 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1353 
1354 			switch (cm->cmsg_type)
1355 			{
1356 			case SCM_RIGHTS:
1357 				if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1358 					fds = datalen / sizeof(int);
1359 					fdp = data;
1360 					for (i = 0; i < fds; i++) {
1361 						fd = *fdp++;
1362 						(void)kern_fcntl(td, fd,
1363 						    F_SETFD, FD_CLOEXEC);
1364 					}
1365 				}
1366 				break;
1367 
1368 			case SCM_CREDS:
1369 				/*
1370 				 * Currently LOCAL_CREDS is never in
1371 				 * effect for Linux so no need to worry
1372 				 * about sockcred
1373 				 */
1374 				if (datalen != sizeof(*cmcred)) {
1375 					error = EMSGSIZE;
1376 					goto bad;
1377 				}
1378 				cmcred = (struct cmsgcred *)data;
1379 				bzero(&linux_ucred, sizeof(linux_ucred));
1380 				linux_ucred.pid = cmcred->cmcred_pid;
1381 				linux_ucred.uid = cmcred->cmcred_uid;
1382 				linux_ucred.gid = cmcred->cmcred_gid;
1383 				data = &linux_ucred;
1384 				datalen = sizeof(linux_ucred);
1385 				break;
1386 
1387 			case SCM_TIMESTAMP:
1388 				if (datalen != sizeof(struct timeval)) {
1389 					error = EMSGSIZE;
1390 					goto bad;
1391 				}
1392 				ftmvl = (struct timeval *)data;
1393 				ltmvl.tv_sec = ftmvl->tv_sec;
1394 				ltmvl.tv_usec = ftmvl->tv_usec;
1395 				data = &ltmvl;
1396 				datalen = sizeof(ltmvl);
1397 				break;
1398 			}
1399 
1400 			if (outlen + LINUX_CMSG_LEN(datalen) >
1401 			    linux_msg.msg_controllen) {
1402 				if (outlen == 0) {
1403 					error = EMSGSIZE;
1404 					goto bad;
1405 				} else {
1406 					linux_msg.msg_flags |=
1407 					    LINUX_MSG_CTRUNC;
1408 					goto out;
1409 				}
1410 			}
1411 
1412 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1413 
1414 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1415 			if (error != 0)
1416 				goto bad;
1417 			outbuf += L_CMSG_HDRSZ;
1418 
1419 			error = copyout(data, outbuf, datalen);
1420 			if (error != 0)
1421 				goto bad;
1422 
1423 			outbuf += LINUX_CMSG_ALIGN(datalen);
1424 			outlen += LINUX_CMSG_LEN(datalen);
1425 
1426 			cm = CMSG_NXTHDR(msg, cm);
1427 		}
1428 	}
1429 
1430 out:
1431 	linux_msg.msg_controllen = outlen;
1432 	error = copyout(&linux_msg, msghdr, sizeof(linux_msg));
1433 
1434 bad:
1435 	free(iov, M_IOV);
1436 	m_freem(control);
1437 	free(linux_cmsg, M_LINUX);
1438 
1439 	return (error);
1440 }
1441 
1442 int
1443 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1444 {
1445 	struct msghdr bsd_msg;
1446 
1447 	return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1448 	    args->flags, &bsd_msg));
1449 }
1450 
1451 int
1452 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1453 {
1454 	struct l_mmsghdr *msg;
1455 	struct msghdr bsd_msg;
1456 	struct l_timespec lts;
1457 	struct timespec ts, tts;
1458 	l_uint retval;
1459 	int error, datagrams;
1460 
1461 	if (args->timeout) {
1462 		error = copyin(args->timeout, &lts, sizeof(struct l_timespec));
1463 		if (error != 0)
1464 			return (error);
1465 		error = linux_to_native_timespec(&ts, &lts);
1466 		if (error != 0)
1467 			return (error);
1468 		getnanotime(&tts);
1469 		timespecadd(&tts, &ts);
1470 	}
1471 
1472 	msg = PTRIN(args->msg);
1473 	datagrams = 0;
1474 	while (datagrams < args->vlen) {
1475 		error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1476 		    args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1477 		if (error != 0)
1478 			break;
1479 
1480 		retval = td->td_retval[0];
1481 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1482 		if (error != 0)
1483 			break;
1484 		++msg;
1485 		++datagrams;
1486 
1487 		/*
1488 		 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1489 		 */
1490 		if (args->flags & LINUX_MSG_WAITFORONE)
1491 			args->flags |= LINUX_MSG_DONTWAIT;
1492 
1493 		/*
1494 		 * See BUGS section of recvmmsg(2).
1495 		 */
1496 		if (args->timeout) {
1497 			getnanotime(&ts);
1498 			timespecsub(&ts, &tts);
1499 			if (!timespecisset(&ts) || ts.tv_sec > 0)
1500 				break;
1501 		}
1502 		/* Out of band data, return right away. */
1503 		if (bsd_msg.msg_flags & MSG_OOB)
1504 			break;
1505 	}
1506 	if (error == 0)
1507 		td->td_retval[0] = datagrams;
1508 	return (error);
1509 }
1510 
1511 int
1512 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1513 {
1514 
1515 	return (kern_shutdown(td, args->s, args->how));
1516 }
1517 
1518 int
1519 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1520 {
1521 	struct setsockopt_args /* {
1522 		int s;
1523 		int level;
1524 		int name;
1525 		caddr_t val;
1526 		int valsize;
1527 	} */ bsd_args;
1528 	l_timeval linux_tv;
1529 	struct timeval tv;
1530 	int error, name;
1531 
1532 	bsd_args.s = args->s;
1533 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1534 	switch (bsd_args.level) {
1535 	case SOL_SOCKET:
1536 		name = linux_to_bsd_so_sockopt(args->optname);
1537 		switch (name) {
1538 		case SO_RCVTIMEO:
1539 			/* FALLTHROUGH */
1540 		case SO_SNDTIMEO:
1541 			error = copyin(PTRIN(args->optval), &linux_tv,
1542 			    sizeof(linux_tv));
1543 			if (error != 0)
1544 				return (error);
1545 			tv.tv_sec = linux_tv.tv_sec;
1546 			tv.tv_usec = linux_tv.tv_usec;
1547 			return (kern_setsockopt(td, args->s, bsd_args.level,
1548 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1549 			/* NOTREACHED */
1550 		default:
1551 			break;
1552 		}
1553 		break;
1554 	case IPPROTO_IP:
1555 		name = linux_to_bsd_ip_sockopt(args->optname);
1556 		break;
1557 	case IPPROTO_IPV6:
1558 		name = linux_to_bsd_ip6_sockopt(args->optname);
1559 		break;
1560 	case IPPROTO_TCP:
1561 		name = linux_to_bsd_tcp_sockopt(args->optname);
1562 		break;
1563 	default:
1564 		name = -1;
1565 		break;
1566 	}
1567 	if (name == -1)
1568 		return (ENOPROTOOPT);
1569 
1570 	bsd_args.name = name;
1571 	bsd_args.val = PTRIN(args->optval);
1572 	bsd_args.valsize = args->optlen;
1573 
1574 	if (name == IPV6_NEXTHOP) {
1575 		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1576 			bsd_args.valsize);
1577 		error = sys_setsockopt(td, &bsd_args);
1578 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1579 	} else
1580 		error = sys_setsockopt(td, &bsd_args);
1581 
1582 	return (error);
1583 }
1584 
1585 int
1586 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1587 {
1588 	struct getsockopt_args /* {
1589 		int s;
1590 		int level;
1591 		int name;
1592 		caddr_t val;
1593 		int *avalsize;
1594 	} */ bsd_args;
1595 	l_timeval linux_tv;
1596 	struct timeval tv;
1597 	socklen_t tv_len, xulen, len;
1598 	struct xucred xu;
1599 	struct l_ucred lxu;
1600 	int error, name, newval;
1601 
1602 	bsd_args.s = args->s;
1603 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1604 	switch (bsd_args.level) {
1605 	case SOL_SOCKET:
1606 		name = linux_to_bsd_so_sockopt(args->optname);
1607 		switch (name) {
1608 		case SO_RCVTIMEO:
1609 			/* FALLTHROUGH */
1610 		case SO_SNDTIMEO:
1611 			tv_len = sizeof(tv);
1612 			error = kern_getsockopt(td, args->s, bsd_args.level,
1613 			    name, &tv, UIO_SYSSPACE, &tv_len);
1614 			if (error != 0)
1615 				return (error);
1616 			linux_tv.tv_sec = tv.tv_sec;
1617 			linux_tv.tv_usec = tv.tv_usec;
1618 			return (copyout(&linux_tv, PTRIN(args->optval),
1619 			    sizeof(linux_tv)));
1620 			/* NOTREACHED */
1621 		case LOCAL_PEERCRED:
1622 			if (args->optlen < sizeof(lxu))
1623 				return (EINVAL);
1624 			xulen = sizeof(xu);
1625 			error = kern_getsockopt(td, args->s, bsd_args.level,
1626 			    name, &xu, UIO_SYSSPACE, &xulen);
1627 			if (error != 0)
1628 				return (error);
1629 			/*
1630 			 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1631 			 */
1632 			lxu.pid = 0;
1633 			lxu.uid = xu.cr_uid;
1634 			lxu.gid = xu.cr_gid;
1635 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1636 			/* NOTREACHED */
1637 		case SO_ERROR:
1638 			len = sizeof(newval);
1639 			error = kern_getsockopt(td, args->s, bsd_args.level,
1640 			    name, &newval, UIO_SYSSPACE, &len);
1641 			if (error != 0)
1642 				return (error);
1643 			newval = -SV_ABI_ERRNO(td->td_proc, newval);
1644 			return (copyout(&newval, PTRIN(args->optval), len));
1645 			/* NOTREACHED */
1646 		default:
1647 			break;
1648 		}
1649 		break;
1650 	case IPPROTO_IP:
1651 		name = linux_to_bsd_ip_sockopt(args->optname);
1652 		break;
1653 	case IPPROTO_IPV6:
1654 		name = linux_to_bsd_ip6_sockopt(args->optname);
1655 		break;
1656 	case IPPROTO_TCP:
1657 		name = linux_to_bsd_tcp_sockopt(args->optname);
1658 		break;
1659 	default:
1660 		name = -1;
1661 		break;
1662 	}
1663 	if (name == -1)
1664 		return (EINVAL);
1665 
1666 	bsd_args.name = name;
1667 	bsd_args.val = PTRIN(args->optval);
1668 	bsd_args.avalsize = PTRIN(args->optlen);
1669 
1670 	if (name == IPV6_NEXTHOP) {
1671 		error = sys_getsockopt(td, &bsd_args);
1672 		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1673 	} else
1674 		error = sys_getsockopt(td, &bsd_args);
1675 
1676 	return (error);
1677 }
1678 
1679 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1680 
1681 /* Argument list sizes for linux_socketcall */
1682 static const unsigned char lxs_args_cnt[] = {
1683 	0 /* unused*/,		3 /* socket */,
1684 	3 /* bind */,		3 /* connect */,
1685 	2 /* listen */,		3 /* accept */,
1686 	3 /* getsockname */,	3 /* getpeername */,
1687 	4 /* socketpair */,	4 /* send */,
1688 	4 /* recv */,		6 /* sendto */,
1689 	6 /* recvfrom */,	2 /* shutdown */,
1690 	5 /* setsockopt */,	5 /* getsockopt */,
1691 	3 /* sendmsg */,	3 /* recvmsg */,
1692 	4 /* accept4 */,	5 /* recvmmsg */,
1693 	4 /* sendmmsg */
1694 };
1695 #define	LINUX_ARGS_CNT		(nitems(lxs_args_cnt) - 1)
1696 #define	LINUX_ARG_SIZE(x)	(lxs_args_cnt[x] * sizeof(l_ulong))
1697 
1698 int
1699 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1700 {
1701 	l_ulong a[6];
1702 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1703 	register_t l_args[6];
1704 #endif
1705 	void *arg;
1706 	int error;
1707 
1708 	if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
1709 		return (EINVAL);
1710 	error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
1711 	if (error != 0)
1712 		return (error);
1713 
1714 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1715 	for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
1716 		l_args[i] = a[i];
1717 	arg = l_args;
1718 #else
1719 	arg = a;
1720 #endif
1721 	switch (args->what) {
1722 	case LINUX_SOCKET:
1723 		return (linux_socket(td, arg));
1724 	case LINUX_BIND:
1725 		return (linux_bind(td, arg));
1726 	case LINUX_CONNECT:
1727 		return (linux_connect(td, arg));
1728 	case LINUX_LISTEN:
1729 		return (linux_listen(td, arg));
1730 	case LINUX_ACCEPT:
1731 		return (linux_accept(td, arg));
1732 	case LINUX_GETSOCKNAME:
1733 		return (linux_getsockname(td, arg));
1734 	case LINUX_GETPEERNAME:
1735 		return (linux_getpeername(td, arg));
1736 	case LINUX_SOCKETPAIR:
1737 		return (linux_socketpair(td, arg));
1738 	case LINUX_SEND:
1739 		return (linux_send(td, arg));
1740 	case LINUX_RECV:
1741 		return (linux_recv(td, arg));
1742 	case LINUX_SENDTO:
1743 		return (linux_sendto(td, arg));
1744 	case LINUX_RECVFROM:
1745 		return (linux_recvfrom(td, arg));
1746 	case LINUX_SHUTDOWN:
1747 		return (linux_shutdown(td, arg));
1748 	case LINUX_SETSOCKOPT:
1749 		return (linux_setsockopt(td, arg));
1750 	case LINUX_GETSOCKOPT:
1751 		return (linux_getsockopt(td, arg));
1752 	case LINUX_SENDMSG:
1753 		return (linux_sendmsg(td, arg));
1754 	case LINUX_RECVMSG:
1755 		return (linux_recvmsg(td, arg));
1756 	case LINUX_ACCEPT4:
1757 		return (linux_accept4(td, arg));
1758 	case LINUX_RECVMMSG:
1759 		return (linux_recvmmsg(td, arg));
1760 	case LINUX_SENDMMSG:
1761 		return (linux_sendmmsg(td, arg));
1762 	}
1763 
1764 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1765 	return (ENOSYS);
1766 }
1767 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1768