xref: /freebsd/sys/compat/linux/linux_socket.c (revision b9f654b163bce26de79705e77b872427c9f2afa1)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 1995 Søren Schmidt
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * 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_common.h>
74 #include <compat/linux/linux_file.h>
75 #include <compat/linux/linux_socket.h>
76 #include <compat/linux/linux_timer.h>
77 #include <compat/linux/linux_util.h>
78 
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 static int
87 linux_to_bsd_sockopt_level(int level)
88 {
89 
90 	switch (level) {
91 	case LINUX_SOL_SOCKET:
92 		return (SOL_SOCKET);
93 	}
94 	return (level);
95 }
96 
97 static int
98 bsd_to_linux_sockopt_level(int level)
99 {
100 
101 	switch (level) {
102 	case SOL_SOCKET:
103 		return (LINUX_SOL_SOCKET);
104 	}
105 	return (level);
106 }
107 
108 static int
109 linux_to_bsd_ip_sockopt(int opt)
110 {
111 
112 	switch (opt) {
113 	case LINUX_IP_TOS:
114 		return (IP_TOS);
115 	case LINUX_IP_TTL:
116 		return (IP_TTL);
117 	case LINUX_IP_OPTIONS:
118 		return (IP_OPTIONS);
119 	case LINUX_IP_MULTICAST_IF:
120 		return (IP_MULTICAST_IF);
121 	case LINUX_IP_MULTICAST_TTL:
122 		return (IP_MULTICAST_TTL);
123 	case LINUX_IP_MULTICAST_LOOP:
124 		return (IP_MULTICAST_LOOP);
125 	case LINUX_IP_ADD_MEMBERSHIP:
126 		return (IP_ADD_MEMBERSHIP);
127 	case LINUX_IP_DROP_MEMBERSHIP:
128 		return (IP_DROP_MEMBERSHIP);
129 	case LINUX_IP_HDRINCL:
130 		return (IP_HDRINCL);
131 	}
132 	return (-1);
133 }
134 
135 static int
136 linux_to_bsd_ip6_sockopt(int opt)
137 {
138 
139 	switch (opt) {
140 	case LINUX_IPV6_NEXTHOP:
141 		return (IPV6_NEXTHOP);
142 	case LINUX_IPV6_UNICAST_HOPS:
143 		return (IPV6_UNICAST_HOPS);
144 	case LINUX_IPV6_MULTICAST_IF:
145 		return (IPV6_MULTICAST_IF);
146 	case LINUX_IPV6_MULTICAST_HOPS:
147 		return (IPV6_MULTICAST_HOPS);
148 	case LINUX_IPV6_MULTICAST_LOOP:
149 		return (IPV6_MULTICAST_LOOP);
150 	case LINUX_IPV6_ADD_MEMBERSHIP:
151 		return (IPV6_JOIN_GROUP);
152 	case LINUX_IPV6_DROP_MEMBERSHIP:
153 		return (IPV6_LEAVE_GROUP);
154 	case LINUX_IPV6_V6ONLY:
155 		return (IPV6_V6ONLY);
156 	case LINUX_IPV6_DONTFRAG:
157 		return (IPV6_DONTFRAG);
158 #if 0
159 	case LINUX_IPV6_CHECKSUM:
160 		return (IPV6_CHECKSUM);
161 	case LINUX_IPV6_RECVPKTINFO:
162 		return (IPV6_RECVPKTINFO);
163 	case LINUX_IPV6_PKTINFO:
164 		return (IPV6_PKTINFO);
165 	case LINUX_IPV6_RECVHOPLIMIT:
166 		return (IPV6_RECVHOPLIMIT);
167 	case LINUX_IPV6_HOPLIMIT:
168 		return (IPV6_HOPLIMIT);
169 	case LINUX_IPV6_RECVHOPOPTS:
170 		return (IPV6_RECVHOPOPTS);
171 	case LINUX_IPV6_HOPOPTS:
172 		return (IPV6_HOPOPTS);
173 	case LINUX_IPV6_RTHDRDSTOPTS:
174 		return (IPV6_RTHDRDSTOPTS);
175 	case LINUX_IPV6_RECVRTHDR:
176 		return (IPV6_RECVRTHDR);
177 	case LINUX_IPV6_RTHDR:
178 		return (IPV6_RTHDR);
179 	case LINUX_IPV6_RECVDSTOPTS:
180 		return (IPV6_RECVDSTOPTS);
181 	case LINUX_IPV6_DSTOPTS:
182 		return (IPV6_DSTOPTS);
183 	case LINUX_IPV6_RECVPATHMTU:
184 		return (IPV6_RECVPATHMTU);
185 	case LINUX_IPV6_PATHMTU:
186 		return (IPV6_PATHMTU);
187 #endif
188 	}
189 	return (-1);
190 }
191 
192 static int
193 linux_to_bsd_so_sockopt(int opt)
194 {
195 
196 	switch (opt) {
197 	case LINUX_SO_DEBUG:
198 		return (SO_DEBUG);
199 	case LINUX_SO_REUSEADDR:
200 		return (SO_REUSEADDR);
201 	case LINUX_SO_TYPE:
202 		return (SO_TYPE);
203 	case LINUX_SO_ERROR:
204 		return (SO_ERROR);
205 	case LINUX_SO_DONTROUTE:
206 		return (SO_DONTROUTE);
207 	case LINUX_SO_BROADCAST:
208 		return (SO_BROADCAST);
209 	case LINUX_SO_SNDBUF:
210 		return (SO_SNDBUF);
211 	case LINUX_SO_RCVBUF:
212 		return (SO_RCVBUF);
213 	case LINUX_SO_KEEPALIVE:
214 		return (SO_KEEPALIVE);
215 	case LINUX_SO_OOBINLINE:
216 		return (SO_OOBINLINE);
217 	case LINUX_SO_LINGER:
218 		return (SO_LINGER);
219 	case LINUX_SO_PEERCRED:
220 		return (LOCAL_PEERCRED);
221 	case LINUX_SO_RCVLOWAT:
222 		return (SO_RCVLOWAT);
223 	case LINUX_SO_SNDLOWAT:
224 		return (SO_SNDLOWAT);
225 	case LINUX_SO_RCVTIMEO:
226 		return (SO_RCVTIMEO);
227 	case LINUX_SO_SNDTIMEO:
228 		return (SO_SNDTIMEO);
229 	case LINUX_SO_TIMESTAMP:
230 		return (SO_TIMESTAMP);
231 	case LINUX_SO_ACCEPTCONN:
232 		return (SO_ACCEPTCONN);
233 	}
234 	return (-1);
235 }
236 
237 static int
238 linux_to_bsd_tcp_sockopt(int opt)
239 {
240 
241 	switch (opt) {
242 	case LINUX_TCP_NODELAY:
243 		return (TCP_NODELAY);
244 	case LINUX_TCP_MAXSEG:
245 		return (TCP_MAXSEG);
246 	case LINUX_TCP_KEEPIDLE:
247 		return (TCP_KEEPIDLE);
248 	case LINUX_TCP_KEEPINTVL:
249 		return (TCP_KEEPINTVL);
250 	case LINUX_TCP_KEEPCNT:
251 		return (TCP_KEEPCNT);
252 	case LINUX_TCP_MD5SIG:
253 		return (TCP_MD5SIG);
254 	}
255 	return (-1);
256 }
257 
258 static int
259 linux_to_bsd_msg_flags(int flags)
260 {
261 	int ret_flags = 0;
262 
263 	if (flags & LINUX_MSG_OOB)
264 		ret_flags |= MSG_OOB;
265 	if (flags & LINUX_MSG_PEEK)
266 		ret_flags |= MSG_PEEK;
267 	if (flags & LINUX_MSG_DONTROUTE)
268 		ret_flags |= MSG_DONTROUTE;
269 	if (flags & LINUX_MSG_CTRUNC)
270 		ret_flags |= MSG_CTRUNC;
271 	if (flags & LINUX_MSG_TRUNC)
272 		ret_flags |= MSG_TRUNC;
273 	if (flags & LINUX_MSG_DONTWAIT)
274 		ret_flags |= MSG_DONTWAIT;
275 	if (flags & LINUX_MSG_EOR)
276 		ret_flags |= MSG_EOR;
277 	if (flags & LINUX_MSG_WAITALL)
278 		ret_flags |= MSG_WAITALL;
279 	if (flags & LINUX_MSG_NOSIGNAL)
280 		ret_flags |= MSG_NOSIGNAL;
281 #if 0 /* not handled */
282 	if (flags & LINUX_MSG_PROXY)
283 		;
284 	if (flags & LINUX_MSG_FIN)
285 		;
286 	if (flags & LINUX_MSG_SYN)
287 		;
288 	if (flags & LINUX_MSG_CONFIRM)
289 		;
290 	if (flags & LINUX_MSG_RST)
291 		;
292 	if (flags & LINUX_MSG_ERRQUEUE)
293 		;
294 #endif
295 	return (ret_flags);
296 }
297 
298 static int
299 linux_to_bsd_cmsg_type(int cmsg_type)
300 {
301 
302 	switch (cmsg_type) {
303 	case LINUX_SCM_RIGHTS:
304 		return (SCM_RIGHTS);
305 	case LINUX_SCM_CREDENTIALS:
306 		return (SCM_CREDS);
307 	}
308 	return (-1);
309 }
310 
311 static int
312 bsd_to_linux_cmsg_type(int cmsg_type)
313 {
314 
315 	switch (cmsg_type) {
316 	case SCM_RIGHTS:
317 		return (LINUX_SCM_RIGHTS);
318 	case SCM_CREDS:
319 		return (LINUX_SCM_CREDENTIALS);
320 	case SCM_TIMESTAMP:
321 		return (LINUX_SCM_TIMESTAMP);
322 	}
323 	return (-1);
324 }
325 
326 static int
327 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
328 {
329 	if (lhdr->msg_controllen > INT_MAX)
330 		return (ENOBUFS);
331 
332 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
333 	bhdr->msg_namelen	= lhdr->msg_namelen;
334 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
335 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
336 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
337 
338 	/*
339 	 * msg_controllen is skipped since BSD and LINUX control messages
340 	 * are potentially different sizes (e.g. the cred structure used
341 	 * by SCM_CREDS is different between the two operating system).
342 	 *
343 	 * The caller can set it (if necessary) after converting all the
344 	 * control messages.
345 	 */
346 
347 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
348 	return (0);
349 }
350 
351 static int
352 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
353 {
354 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
355 	lhdr->msg_namelen	= bhdr->msg_namelen;
356 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
357 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
358 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
359 
360 	/*
361 	 * msg_controllen is skipped since BSD and LINUX control messages
362 	 * are potentially different sizes (e.g. the cred structure used
363 	 * by SCM_CREDS is different between the two operating system).
364 	 *
365 	 * The caller can set it (if necessary) after converting all the
366 	 * control messages.
367 	 */
368 
369 	/* msg_flags skipped */
370 	return (0);
371 }
372 
373 static int
374 linux_set_socket_flags(int lflags, int *flags)
375 {
376 
377 	if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
378 		return (EINVAL);
379 	if (lflags & LINUX_SOCK_NONBLOCK)
380 		*flags |= SOCK_NONBLOCK;
381 	if (lflags & LINUX_SOCK_CLOEXEC)
382 		*flags |= SOCK_CLOEXEC;
383 	return (0);
384 }
385 
386 static int
387 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
388     struct mbuf *control, enum uio_seg segflg)
389 {
390 	struct sockaddr *to;
391 	int error, len;
392 
393 	if (mp->msg_name != NULL) {
394 		len = mp->msg_namelen;
395 		error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len);
396 		if (error != 0)
397 			return (error);
398 		mp->msg_name = to;
399 	} else
400 		to = NULL;
401 
402 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
403 	    segflg);
404 
405 	if (to)
406 		free(to, M_SONAME);
407 	return (error);
408 }
409 
410 /* Return 0 if IP_HDRINCL is set for the given socket. */
411 static int
412 linux_check_hdrincl(struct thread *td, int s)
413 {
414 	int error, optval;
415 	socklen_t size_val;
416 
417 	size_val = sizeof(optval);
418 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
419 	    &optval, UIO_SYSSPACE, &size_val);
420 	if (error != 0)
421 		return (error);
422 
423 	return (optval == 0);
424 }
425 
426 /*
427  * Updated sendto() when IP_HDRINCL is set:
428  * tweak endian-dependent fields in the IP packet.
429  */
430 static int
431 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
432 {
433 /*
434  * linux_ip_copysize defines how many bytes we should copy
435  * from the beginning of the IP packet before we customize it for BSD.
436  * It should include all the fields we modify (ip_len and ip_off).
437  */
438 #define linux_ip_copysize	8
439 
440 	struct ip *packet;
441 	struct msghdr msg;
442 	struct iovec aiov[1];
443 	int error;
444 
445 	/* Check that the packet isn't too big or too small. */
446 	if (linux_args->len < linux_ip_copysize ||
447 	    linux_args->len > IP_MAXPACKET)
448 		return (EINVAL);
449 
450 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
451 
452 	/* Make kernel copy of the packet to be sent */
453 	if ((error = copyin(PTRIN(linux_args->msg), packet,
454 	    linux_args->len)))
455 		goto goout;
456 
457 	/* Convert fields from Linux to BSD raw IP socket format */
458 	packet->ip_len = linux_args->len;
459 	packet->ip_off = ntohs(packet->ip_off);
460 
461 	/* Prepare the msghdr and iovec structures describing the new packet */
462 	msg.msg_name = PTRIN(linux_args->to);
463 	msg.msg_namelen = linux_args->tolen;
464 	msg.msg_iov = aiov;
465 	msg.msg_iovlen = 1;
466 	msg.msg_control = NULL;
467 	msg.msg_flags = 0;
468 	aiov[0].iov_base = (char *)packet;
469 	aiov[0].iov_len = linux_args->len;
470 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
471 	    NULL, UIO_SYSSPACE);
472 goout:
473 	free(packet, M_LINUX);
474 	return (error);
475 }
476 
477 int
478 linux_socket(struct thread *td, struct linux_socket_args *args)
479 {
480 	int domain, retval_socket, type;
481 
482 	type = args->type & LINUX_SOCK_TYPE_MASK;
483 	if (type < 0 || type > LINUX_SOCK_MAX)
484 		return (EINVAL);
485 	retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
486 		&type);
487 	if (retval_socket != 0)
488 		return (retval_socket);
489 	domain = linux_to_bsd_domain(args->domain);
490 	if (domain == -1)
491 		return (EAFNOSUPPORT);
492 
493 	retval_socket = kern_socket(td, domain, type, args->protocol);
494 	if (retval_socket)
495 		return (retval_socket);
496 
497 	if (type == SOCK_RAW
498 	    && (args->protocol == IPPROTO_RAW || args->protocol == 0)
499 	    && domain == PF_INET) {
500 		/* It's a raw IP socket: set the IP_HDRINCL option. */
501 		int hdrincl;
502 
503 		hdrincl = 1;
504 		/* We ignore any error returned by kern_setsockopt() */
505 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
506 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
507 	}
508 #ifdef INET6
509 	/*
510 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
511 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
512 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
513 	 * sysctl value.
514 	 */
515 	if (domain == PF_INET6) {
516 		int v6only;
517 
518 		v6only = 0;
519 		/* We ignore any error returned by setsockopt() */
520 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
521 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
522 	}
523 #endif
524 
525 	return (retval_socket);
526 }
527 
528 int
529 linux_bind(struct thread *td, struct linux_bind_args *args)
530 {
531 	struct sockaddr *sa;
532 	int error;
533 
534 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
535 	    &args->namelen);
536 	if (error != 0)
537 		return (error);
538 
539 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
540 	free(sa, M_SONAME);
541 
542 	/* XXX */
543 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
544 		return (EINVAL);
545 	return (error);
546 }
547 
548 int
549 linux_connect(struct thread *td, struct linux_connect_args *args)
550 {
551 	struct socket *so;
552 	struct sockaddr *sa;
553 	struct file *fp;
554 	u_int fflag;
555 	int error;
556 
557 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
558 	    &args->namelen);
559 	if (error != 0)
560 		return (error);
561 
562 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
563 	free(sa, M_SONAME);
564 	if (error != EISCONN)
565 		return (error);
566 
567 	/*
568 	 * Linux doesn't return EISCONN the first time it occurs,
569 	 * when on a non-blocking socket. Instead it returns the
570 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
571 	 */
572 	error = getsock_cap(td, args->s, &cap_connect_rights,
573 	    &fp, &fflag, NULL);
574 	if (error != 0)
575 		return (error);
576 
577 	error = EISCONN;
578 	so = fp->f_data;
579 	if (fflag & FNONBLOCK) {
580 		SOCK_LOCK(so);
581 		if (so->so_emuldata == 0)
582 			error = so->so_error;
583 		so->so_emuldata = (void *)1;
584 		SOCK_UNLOCK(so);
585 	}
586 	fdrop(fp, td);
587 
588 	return (error);
589 }
590 
591 int
592 linux_listen(struct thread *td, struct linux_listen_args *args)
593 {
594 
595 	return (kern_listen(td, args->s, args->backlog));
596 }
597 
598 static int
599 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
600     l_uintptr_t namelen, int flags)
601 {
602 	struct l_sockaddr *lsa;
603 	struct sockaddr *sa;
604 	struct file *fp;
605 	int bflags, len;
606 	struct socket *so;
607 	int error, error1;
608 
609 	bflags = 0;
610 	error = linux_set_socket_flags(flags, &bflags);
611 	if (error != 0)
612 		return (error);
613 
614 	sa = NULL;
615 	if (PTRIN(addr) == NULL) {
616 		len = 0;
617 		error = kern_accept4(td, s, NULL, NULL, bflags, NULL);
618 	} else {
619 		error = copyin(PTRIN(namelen), &len, sizeof(len));
620 		if (error != 0)
621 			return (error);
622 		if (len < 0)
623 			return (EINVAL);
624 		error = kern_accept4(td, s, &sa, &len, bflags, &fp);
625 		if (error == 0)
626 			fdrop(fp, td);
627 	}
628 
629 	if (error != 0) {
630 		/*
631 		 * XXX. This is wrong, different sockaddr structures
632 		 * have different sizes.
633 		 */
634 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
635 		{
636 			error = EINVAL;
637 			goto out;
638 		}
639 		if (error == EINVAL) {
640 			error1 = getsock_cap(td, s, &cap_accept_rights, &fp, NULL, NULL);
641 			if (error1 != 0) {
642 				error = error1;
643 				goto out;
644 			}
645 			so = fp->f_data;
646 			if (so->so_type == SOCK_DGRAM)
647 				error = EOPNOTSUPP;
648 			fdrop(fp, td);
649 		}
650 		goto out;
651 	}
652 
653 	if (len != 0 && error == 0) {
654 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
655 		if (error == 0)
656 			error = copyout(lsa, PTRIN(addr), len);
657 		free(lsa, M_SONAME);
658 	}
659 
660 	free(sa, M_SONAME);
661 
662 out:
663 	if (error != 0) {
664 		(void)kern_close(td, td->td_retval[0]);
665 		td->td_retval[0] = 0;
666 	}
667 	return (error);
668 }
669 
670 int
671 linux_accept(struct thread *td, struct linux_accept_args *args)
672 {
673 
674 	return (linux_accept_common(td, args->s, args->addr,
675 	    args->namelen, 0));
676 }
677 
678 int
679 linux_accept4(struct thread *td, struct linux_accept4_args *args)
680 {
681 
682 	return (linux_accept_common(td, args->s, args->addr,
683 	    args->namelen, args->flags));
684 }
685 
686 int
687 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
688 {
689 	struct l_sockaddr *lsa;
690 	struct sockaddr *sa;
691 	int len, error;
692 
693 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
694 	if (error != 0)
695 		return (error);
696 
697 	error = kern_getsockname(td, args->s, &sa, &len);
698 	if (error != 0)
699 		return (error);
700 
701 	if (len != 0) {
702 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
703 		if (error == 0)
704 			error = copyout(lsa, PTRIN(args->addr),
705 			    len);
706 		free(lsa, M_SONAME);
707 	}
708 
709 	free(sa, M_SONAME);
710 	if (error == 0)
711 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
712 	return (error);
713 }
714 
715 int
716 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
717 {
718 	struct l_sockaddr *lsa;
719 	struct sockaddr *sa;
720 	int len, error;
721 
722 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
723 	if (error != 0)
724 		return (error);
725 	if (len < 0)
726 		return (EINVAL);
727 
728 	error = kern_getpeername(td, args->s, &sa, &len);
729 	if (error != 0)
730 		return (error);
731 
732 	if (len != 0) {
733 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
734 		if (error == 0)
735 			error = copyout(lsa, PTRIN(args->addr),
736 			    len);
737 		free(lsa, M_SONAME);
738 	}
739 
740 	free(sa, M_SONAME);
741 	if (error == 0)
742 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
743 	return (error);
744 }
745 
746 int
747 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
748 {
749 	struct socketpair_args /* {
750 		int domain;
751 		int type;
752 		int protocol;
753 		int *rsv;
754 	} */ bsd_args;
755 	int error;
756 
757 	bsd_args.domain = linux_to_bsd_domain(args->domain);
758 	if (bsd_args.domain != PF_LOCAL)
759 		return (EAFNOSUPPORT);
760 	bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
761 	if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
762 		return (EINVAL);
763 	error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
764 		&bsd_args.type);
765 	if (error != 0)
766 		return (error);
767 	if (args->protocol != 0 && args->protocol != PF_UNIX)
768 
769 		/*
770 		 * Use of PF_UNIX as protocol argument is not right,
771 		 * but Linux does it.
772 		 * Do not map PF_UNIX as its Linux value is identical
773 		 * to FreeBSD one.
774 		 */
775 		return (EPROTONOSUPPORT);
776 	else
777 		bsd_args.protocol = 0;
778 	bsd_args.rsv = (int *)PTRIN(args->rsv);
779 	return (sys_socketpair(td, &bsd_args));
780 }
781 
782 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
783 struct linux_send_args {
784 	register_t s;
785 	register_t msg;
786 	register_t len;
787 	register_t flags;
788 };
789 
790 static int
791 linux_send(struct thread *td, struct linux_send_args *args)
792 {
793 	struct sendto_args /* {
794 		int s;
795 		caddr_t buf;
796 		int len;
797 		int flags;
798 		caddr_t to;
799 		int tolen;
800 	} */ bsd_args;
801 	struct file *fp;
802 	int error, fflag;
803 
804 	bsd_args.s = args->s;
805 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
806 	bsd_args.len = args->len;
807 	bsd_args.flags = args->flags;
808 	bsd_args.to = NULL;
809 	bsd_args.tolen = 0;
810 	error = sys_sendto(td, &bsd_args);
811 	if (error == ENOTCONN) {
812 		/*
813 		 * Linux doesn't return ENOTCONN for non-blocking sockets.
814 		 * Instead it returns the EAGAIN.
815 		 */
816 		error = getsock_cap(td, args->s, &cap_send_rights, &fp,
817 		    &fflag, NULL);
818 		if (error == 0) {
819 			if (fflag & FNONBLOCK)
820 				error = EAGAIN;
821 			fdrop(fp, td);
822 		}
823 	}
824 	return (error);
825 }
826 
827 struct linux_recv_args {
828 	register_t s;
829 	register_t msg;
830 	register_t len;
831 	register_t flags;
832 };
833 
834 static int
835 linux_recv(struct thread *td, struct linux_recv_args *args)
836 {
837 	struct recvfrom_args /* {
838 		int s;
839 		caddr_t buf;
840 		int len;
841 		int flags;
842 		struct sockaddr *from;
843 		socklen_t fromlenaddr;
844 	} */ bsd_args;
845 
846 	bsd_args.s = args->s;
847 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
848 	bsd_args.len = args->len;
849 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
850 	bsd_args.from = NULL;
851 	bsd_args.fromlenaddr = 0;
852 	return (sys_recvfrom(td, &bsd_args));
853 }
854 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
855 
856 int
857 linux_sendto(struct thread *td, struct linux_sendto_args *args)
858 {
859 	struct msghdr msg;
860 	struct iovec aiov;
861 
862 	if (linux_check_hdrincl(td, args->s) == 0)
863 		/* IP_HDRINCL set, tweak the packet before sending */
864 		return (linux_sendto_hdrincl(td, args));
865 
866 	msg.msg_name = PTRIN(args->to);
867 	msg.msg_namelen = args->tolen;
868 	msg.msg_iov = &aiov;
869 	msg.msg_iovlen = 1;
870 	msg.msg_control = NULL;
871 	msg.msg_flags = 0;
872 	aiov.iov_base = PTRIN(args->msg);
873 	aiov.iov_len = args->len;
874 	return (linux_sendit(td, args->s, &msg, args->flags, NULL,
875 	    UIO_USERSPACE));
876 }
877 
878 int
879 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
880 {
881 	struct l_sockaddr *lsa;
882 	struct sockaddr *sa;
883 	struct msghdr msg;
884 	struct iovec aiov;
885 	int error, fromlen;
886 
887 	if (PTRIN(args->fromlen) != NULL) {
888 		error = copyin(PTRIN(args->fromlen), &fromlen,
889 		    sizeof(fromlen));
890 		if (error != 0)
891 			return (error);
892 		if (fromlen < 0)
893 			return (EINVAL);
894 		sa = malloc(fromlen, M_SONAME, M_WAITOK);
895 	} else {
896 		fromlen = 0;
897 		sa = NULL;
898 	}
899 
900 	msg.msg_name = sa;
901 	msg.msg_namelen = fromlen;
902 	msg.msg_iov = &aiov;
903 	msg.msg_iovlen = 1;
904 	aiov.iov_base = PTRIN(args->buf);
905 	aiov.iov_len = args->len;
906 	msg.msg_control = 0;
907 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
908 
909 	error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL);
910 	if (error != 0)
911 		goto out;
912 
913 	if (PTRIN(args->from) != NULL) {
914 		error = bsd_to_linux_sockaddr(sa, &lsa, msg.msg_namelen);
915 		if (error == 0)
916 			error = copyout(lsa, PTRIN(args->from),
917 			    msg.msg_namelen);
918 		free(lsa, M_SONAME);
919 	}
920 
921 	if (error == 0 && PTRIN(args->fromlen) != NULL)
922 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
923 		    sizeof(msg.msg_namelen));
924 out:
925 	free(sa, M_SONAME);
926 	return (error);
927 }
928 
929 static int
930 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
931     l_uint flags)
932 {
933 	struct cmsghdr *cmsg;
934 	struct mbuf *control;
935 	struct msghdr msg;
936 	struct l_cmsghdr linux_cmsg;
937 	struct l_cmsghdr *ptr_cmsg;
938 	struct l_msghdr linux_msg;
939 	struct iovec *iov;
940 	socklen_t datalen;
941 	struct sockaddr *sa;
942 	struct socket *so;
943 	sa_family_t sa_family;
944 	struct file *fp;
945 	void *data;
946 	l_size_t len;
947 	l_size_t clen;
948 	int error, fflag;
949 
950 	error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
951 	if (error != 0)
952 		return (error);
953 
954 	/*
955 	 * Some Linux applications (ping) define a non-NULL control data
956 	 * pointer, but a msg_controllen of 0, which is not allowed in the
957 	 * FreeBSD system call interface.  NULL the msg_control pointer in
958 	 * order to handle this case.  This should be checked, but allows the
959 	 * Linux ping to work.
960 	 */
961 	if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
962 		linux_msg.msg_control = PTROUT(NULL);
963 
964 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
965 	if (error != 0)
966 		return (error);
967 
968 #ifdef COMPAT_LINUX32
969 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
970 	    &iov, EMSGSIZE);
971 #else
972 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
973 #endif
974 	if (error != 0)
975 		return (error);
976 
977 	control = NULL;
978 
979 	error = kern_getsockname(td, s, &sa, &datalen);
980 	if (error != 0)
981 		goto bad;
982 	sa_family = sa->sa_family;
983 	free(sa, M_SONAME);
984 
985 	if (flags & LINUX_MSG_OOB) {
986 		error = EOPNOTSUPP;
987 		if (sa_family == AF_UNIX)
988 			goto bad;
989 
990 		error = getsock_cap(td, s, &cap_send_rights, &fp,
991 		    &fflag, NULL);
992 		if (error != 0)
993 			goto bad;
994 		so = fp->f_data;
995 		if (so->so_type != SOCK_STREAM)
996 			error = EOPNOTSUPP;
997 		fdrop(fp, td);
998 		if (error != 0)
999 			goto bad;
1000 	}
1001 
1002 	if (linux_msg.msg_controllen >= sizeof(struct l_cmsghdr)) {
1003 
1004 		error = ENOBUFS;
1005 		control = m_get(M_WAITOK, MT_CONTROL);
1006 		MCLGET(control, M_WAITOK);
1007 		data = mtod(control, void *);
1008 		datalen = 0;
1009 
1010 		ptr_cmsg = PTRIN(linux_msg.msg_control);
1011 		clen = linux_msg.msg_controllen;
1012 		do {
1013 			error = copyin(ptr_cmsg, &linux_cmsg,
1014 			    sizeof(struct l_cmsghdr));
1015 			if (error != 0)
1016 				goto bad;
1017 
1018 			error = EINVAL;
1019 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1020 			    linux_cmsg.cmsg_len > clen)
1021 				goto bad;
1022 
1023 			if (datalen + CMSG_HDRSZ > MCLBYTES)
1024 				goto bad;
1025 
1026 			/*
1027 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1028 			 * so return EINVAL in any other cmsg_type
1029 			 */
1030 			cmsg = data;
1031 			cmsg->cmsg_type =
1032 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1033 			cmsg->cmsg_level =
1034 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1035 			if (cmsg->cmsg_type == -1
1036 			    || cmsg->cmsg_level != SOL_SOCKET)
1037 				goto bad;
1038 
1039 			/*
1040 			 * Some applications (e.g. pulseaudio) attempt to
1041 			 * send ancillary data even if the underlying protocol
1042 			 * doesn't support it which is not allowed in the
1043 			 * FreeBSD system call interface.
1044 			 */
1045 			if (sa_family != AF_UNIX)
1046 				continue;
1047 
1048 			if (cmsg->cmsg_type == SCM_CREDS) {
1049 				len = sizeof(struct cmsgcred);
1050 				if (datalen + CMSG_SPACE(len) > MCLBYTES)
1051 					goto bad;
1052 
1053 				/*
1054 				 * The lower levels will fill in the structure
1055 				 */
1056 				memset(CMSG_DATA(data), 0, len);
1057 			} else {
1058 				len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1059 				if (datalen + CMSG_SPACE(len) < datalen ||
1060 				    datalen + CMSG_SPACE(len) > MCLBYTES)
1061 					goto bad;
1062 
1063 				error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1064 				    CMSG_DATA(data), len);
1065 				if (error != 0)
1066 					goto bad;
1067 			}
1068 
1069 			cmsg->cmsg_len = CMSG_LEN(len);
1070 			data = (char *)data + CMSG_SPACE(len);
1071 			datalen += CMSG_SPACE(len);
1072 
1073 			if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1074 				break;
1075 
1076 			clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1077 			ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1078 			    LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1079 		} while(clen >= sizeof(struct l_cmsghdr));
1080 
1081 		control->m_len = datalen;
1082 		if (datalen == 0) {
1083 			m_freem(control);
1084 			control = NULL;
1085 		}
1086 	}
1087 
1088 	msg.msg_iov = iov;
1089 	msg.msg_flags = 0;
1090 	error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1091 	control = NULL;
1092 
1093 bad:
1094 	m_freem(control);
1095 	free(iov, M_IOV);
1096 	return (error);
1097 }
1098 
1099 int
1100 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1101 {
1102 
1103 	return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1104 	    args->flags));
1105 }
1106 
1107 int
1108 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1109 {
1110 	struct l_mmsghdr *msg;
1111 	l_uint retval;
1112 	int error, datagrams;
1113 
1114 	if (args->vlen > UIO_MAXIOV)
1115 		args->vlen = UIO_MAXIOV;
1116 
1117 	msg = PTRIN(args->msg);
1118 	datagrams = 0;
1119 	while (datagrams < args->vlen) {
1120 		error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1121 		    args->flags);
1122 		if (error != 0)
1123 			break;
1124 
1125 		retval = td->td_retval[0];
1126 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1127 		if (error != 0)
1128 			break;
1129 		++msg;
1130 		++datagrams;
1131 	}
1132 	if (error == 0)
1133 		td->td_retval[0] = datagrams;
1134 	return (error);
1135 }
1136 
1137 static int
1138 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1139     l_uint flags, struct msghdr *msg)
1140 {
1141 	struct cmsghdr *cm;
1142 	struct cmsgcred *cmcred;
1143 	struct l_cmsghdr *linux_cmsg = NULL;
1144 	struct l_ucred linux_ucred;
1145 	socklen_t datalen, maxlen, outlen;
1146 	struct l_msghdr linux_msg;
1147 	struct iovec *iov, *uiov;
1148 	struct mbuf *control = NULL;
1149 	struct mbuf **controlp;
1150 	struct timeval *ftmvl;
1151 	struct l_sockaddr *lsa;
1152 	struct sockaddr *sa;
1153 	l_timeval ltmvl;
1154 	caddr_t outbuf;
1155 	void *data;
1156 	int error, i, fd, fds, *fdp;
1157 
1158 	error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1159 	if (error != 0)
1160 		return (error);
1161 
1162 	error = linux_to_bsd_msghdr(msg, &linux_msg);
1163 	if (error != 0)
1164 		return (error);
1165 
1166 #ifdef COMPAT_LINUX32
1167 	error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1168 	    &iov, EMSGSIZE);
1169 #else
1170 	error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1171 #endif
1172 	if (error != 0)
1173 		return (error);
1174 
1175 	if (msg->msg_name) {
1176 		sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK);
1177 		msg->msg_name = sa;
1178 	} else
1179 		sa = NULL;
1180 
1181 	uiov = msg->msg_iov;
1182 	msg->msg_iov = iov;
1183 	controlp = (msg->msg_control != NULL) ? &control : NULL;
1184 	error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp);
1185 	msg->msg_iov = uiov;
1186 	if (error != 0)
1187 		goto bad;
1188 
1189 	if (msg->msg_name) {
1190 		msg->msg_name = PTRIN(linux_msg.msg_name);
1191 		error = bsd_to_linux_sockaddr(sa, &lsa, msg->msg_namelen);
1192 		if (error == 0)
1193 			error = copyout(lsa, PTRIN(msg->msg_name),
1194 			    msg->msg_namelen);
1195 		free(lsa, M_SONAME);
1196 		if (error != 0)
1197 			goto bad;
1198 	}
1199 
1200 	error = bsd_to_linux_msghdr(msg, &linux_msg);
1201 	if (error != 0)
1202 		goto bad;
1203 
1204 	maxlen = linux_msg.msg_controllen;
1205 	linux_msg.msg_controllen = 0;
1206 	if (control) {
1207 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1208 
1209 		msg->msg_control = mtod(control, struct cmsghdr *);
1210 		msg->msg_controllen = control->m_len;
1211 
1212 		cm = CMSG_FIRSTHDR(msg);
1213 		outbuf = PTRIN(linux_msg.msg_control);
1214 		outlen = 0;
1215 		while (cm != NULL) {
1216 			linux_cmsg->cmsg_type =
1217 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1218 			linux_cmsg->cmsg_level =
1219 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1220 			if (linux_cmsg->cmsg_type == -1 ||
1221 			    cm->cmsg_level != SOL_SOCKET) {
1222 				error = EINVAL;
1223 				goto bad;
1224 			}
1225 
1226 			data = CMSG_DATA(cm);
1227 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1228 
1229 			switch (cm->cmsg_type) {
1230 			case SCM_RIGHTS:
1231 				if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1232 					fds = datalen / sizeof(int);
1233 					fdp = data;
1234 					for (i = 0; i < fds; i++) {
1235 						fd = *fdp++;
1236 						(void)kern_fcntl(td, fd,
1237 						    F_SETFD, FD_CLOEXEC);
1238 					}
1239 				}
1240 				break;
1241 
1242 			case SCM_CREDS:
1243 				/*
1244 				 * Currently LOCAL_CREDS is never in
1245 				 * effect for Linux so no need to worry
1246 				 * about sockcred
1247 				 */
1248 				if (datalen != sizeof(*cmcred)) {
1249 					error = EMSGSIZE;
1250 					goto bad;
1251 				}
1252 				cmcred = (struct cmsgcred *)data;
1253 				bzero(&linux_ucred, sizeof(linux_ucred));
1254 				linux_ucred.pid = cmcred->cmcred_pid;
1255 				linux_ucred.uid = cmcred->cmcred_uid;
1256 				linux_ucred.gid = cmcred->cmcred_gid;
1257 				data = &linux_ucred;
1258 				datalen = sizeof(linux_ucred);
1259 				break;
1260 
1261 			case SCM_TIMESTAMP:
1262 				if (datalen != sizeof(struct timeval)) {
1263 					error = EMSGSIZE;
1264 					goto bad;
1265 				}
1266 				ftmvl = (struct timeval *)data;
1267 				ltmvl.tv_sec = ftmvl->tv_sec;
1268 				ltmvl.tv_usec = ftmvl->tv_usec;
1269 				data = &ltmvl;
1270 				datalen = sizeof(ltmvl);
1271 				break;
1272 			}
1273 
1274 			if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1275 				if (outlen == 0) {
1276 					error = EMSGSIZE;
1277 					goto bad;
1278 				} else {
1279 					linux_msg.msg_flags |= LINUX_MSG_CTRUNC;
1280 					m_dispose_extcontrolm(control);
1281 					goto out;
1282 				}
1283 			}
1284 
1285 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1286 
1287 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1288 			if (error != 0)
1289 				goto bad;
1290 			outbuf += L_CMSG_HDRSZ;
1291 
1292 			error = copyout(data, outbuf, datalen);
1293 			if (error != 0)
1294 				goto bad;
1295 
1296 			outbuf += LINUX_CMSG_ALIGN(datalen);
1297 			outlen += LINUX_CMSG_LEN(datalen);
1298 
1299 			cm = CMSG_NXTHDR(msg, cm);
1300 		}
1301 		linux_msg.msg_controllen = outlen;
1302 	}
1303 
1304 out:
1305 	error = copyout(&linux_msg, msghdr, sizeof(linux_msg));
1306 
1307 bad:
1308 	if (control != NULL) {
1309 		if (error != 0)
1310 			m_dispose_extcontrolm(control);
1311 		m_freem(control);
1312 	}
1313 	free(iov, M_IOV);
1314 	free(linux_cmsg, M_LINUX);
1315 	free(sa, M_SONAME);
1316 
1317 	return (error);
1318 }
1319 
1320 int
1321 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1322 {
1323 	struct msghdr bsd_msg;
1324 
1325 	return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1326 	    args->flags, &bsd_msg));
1327 }
1328 
1329 int
1330 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1331 {
1332 	struct l_mmsghdr *msg;
1333 	struct msghdr bsd_msg;
1334 	struct l_timespec lts;
1335 	struct timespec ts, tts;
1336 	l_uint retval;
1337 	int error, datagrams;
1338 
1339 	if (args->timeout) {
1340 		error = copyin(args->timeout, &lts, sizeof(struct l_timespec));
1341 		if (error != 0)
1342 			return (error);
1343 		error = linux_to_native_timespec(&ts, &lts);
1344 		if (error != 0)
1345 			return (error);
1346 		getnanotime(&tts);
1347 		timespecadd(&tts, &ts, &tts);
1348 	}
1349 
1350 	msg = PTRIN(args->msg);
1351 	datagrams = 0;
1352 	while (datagrams < args->vlen) {
1353 		error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1354 		    args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1355 		if (error != 0)
1356 			break;
1357 
1358 		retval = td->td_retval[0];
1359 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1360 		if (error != 0)
1361 			break;
1362 		++msg;
1363 		++datagrams;
1364 
1365 		/*
1366 		 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1367 		 */
1368 		if (args->flags & LINUX_MSG_WAITFORONE)
1369 			args->flags |= LINUX_MSG_DONTWAIT;
1370 
1371 		/*
1372 		 * See BUGS section of recvmmsg(2).
1373 		 */
1374 		if (args->timeout) {
1375 			getnanotime(&ts);
1376 			timespecsub(&ts, &tts, &ts);
1377 			if (!timespecisset(&ts) || ts.tv_sec > 0)
1378 				break;
1379 		}
1380 		/* Out of band data, return right away. */
1381 		if (bsd_msg.msg_flags & MSG_OOB)
1382 			break;
1383 	}
1384 	if (error == 0)
1385 		td->td_retval[0] = datagrams;
1386 	return (error);
1387 }
1388 
1389 int
1390 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1391 {
1392 
1393 	return (kern_shutdown(td, args->s, args->how));
1394 }
1395 
1396 int
1397 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1398 {
1399 	struct setsockopt_args /* {
1400 		int s;
1401 		int level;
1402 		int name;
1403 		const void *val;
1404 		int valsize;
1405 	} */ bsd_args;
1406 	l_timeval linux_tv;
1407 	struct sockaddr *sa;
1408 	struct timeval tv;
1409 	socklen_t len;
1410 	int error, name;
1411 
1412 	bsd_args.s = args->s;
1413 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1414 	switch (bsd_args.level) {
1415 	case SOL_SOCKET:
1416 		name = linux_to_bsd_so_sockopt(args->optname);
1417 		switch (name) {
1418 		case SO_RCVTIMEO:
1419 			/* FALLTHROUGH */
1420 		case SO_SNDTIMEO:
1421 			error = copyin(PTRIN(args->optval), &linux_tv,
1422 			    sizeof(linux_tv));
1423 			if (error != 0)
1424 				return (error);
1425 			tv.tv_sec = linux_tv.tv_sec;
1426 			tv.tv_usec = linux_tv.tv_usec;
1427 			return (kern_setsockopt(td, args->s, bsd_args.level,
1428 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1429 			/* NOTREACHED */
1430 		default:
1431 			break;
1432 		}
1433 		break;
1434 	case IPPROTO_IP:
1435 		name = linux_to_bsd_ip_sockopt(args->optname);
1436 		break;
1437 	case IPPROTO_IPV6:
1438 		name = linux_to_bsd_ip6_sockopt(args->optname);
1439 		break;
1440 	case IPPROTO_TCP:
1441 		name = linux_to_bsd_tcp_sockopt(args->optname);
1442 		break;
1443 	default:
1444 		name = -1;
1445 		break;
1446 	}
1447 	if (name == -1)
1448 		return (ENOPROTOOPT);
1449 
1450 
1451 	if (name == IPV6_NEXTHOP) {
1452 
1453 		len = args->optlen;
1454 		error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len);
1455 		if (error != 0)
1456 			return (error);
1457 
1458 		error = kern_setsockopt(td, args->s, bsd_args.level,
1459 		    name, sa, UIO_SYSSPACE, len);
1460 		free(sa, M_SONAME);
1461 	} else {
1462 		bsd_args.name = name;
1463 		bsd_args.val = PTRIN(args->optval);
1464 		bsd_args.valsize = args->optlen;
1465 		error = sys_setsockopt(td, &bsd_args);
1466 	}
1467 
1468 	return (error);
1469 }
1470 
1471 int
1472 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1473 {
1474 	struct getsockopt_args /* {
1475 		int s;
1476 		int level;
1477 		int name;
1478 		caddr_t val;
1479 		int *avalsize;
1480 	} */ bsd_args;
1481 	l_timeval linux_tv;
1482 	struct timeval tv;
1483 	socklen_t tv_len, xulen, len;
1484 	struct l_sockaddr *lsa;
1485 	struct sockaddr *sa;
1486 	struct xucred xu;
1487 	struct l_ucred lxu;
1488 	int error, name, newval;
1489 
1490 	bsd_args.s = args->s;
1491 	bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1492 	switch (bsd_args.level) {
1493 	case SOL_SOCKET:
1494 		name = linux_to_bsd_so_sockopt(args->optname);
1495 		switch (name) {
1496 		case SO_RCVTIMEO:
1497 			/* FALLTHROUGH */
1498 		case SO_SNDTIMEO:
1499 			tv_len = sizeof(tv);
1500 			error = kern_getsockopt(td, args->s, bsd_args.level,
1501 			    name, &tv, UIO_SYSSPACE, &tv_len);
1502 			if (error != 0)
1503 				return (error);
1504 			linux_tv.tv_sec = tv.tv_sec;
1505 			linux_tv.tv_usec = tv.tv_usec;
1506 			return (copyout(&linux_tv, PTRIN(args->optval),
1507 			    sizeof(linux_tv)));
1508 			/* NOTREACHED */
1509 		case LOCAL_PEERCRED:
1510 			if (args->optlen < sizeof(lxu))
1511 				return (EINVAL);
1512 			/*
1513 			 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
1514 			 * but by the Unix socket's level 0.
1515 			 */
1516 			bsd_args.level = 0;
1517 			xulen = sizeof(xu);
1518 			error = kern_getsockopt(td, args->s, bsd_args.level,
1519 			    name, &xu, UIO_SYSSPACE, &xulen);
1520 			if (error != 0)
1521 				return (error);
1522 			lxu.pid = xu.cr_pid;
1523 			lxu.uid = xu.cr_uid;
1524 			lxu.gid = xu.cr_gid;
1525 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1526 			/* NOTREACHED */
1527 		case SO_ERROR:
1528 			len = sizeof(newval);
1529 			error = kern_getsockopt(td, args->s, bsd_args.level,
1530 			    name, &newval, UIO_SYSSPACE, &len);
1531 			if (error != 0)
1532 				return (error);
1533 			newval = -SV_ABI_ERRNO(td->td_proc, newval);
1534 			return (copyout(&newval, PTRIN(args->optval), len));
1535 			/* NOTREACHED */
1536 		default:
1537 			break;
1538 		}
1539 		break;
1540 	case IPPROTO_IP:
1541 		name = linux_to_bsd_ip_sockopt(args->optname);
1542 		break;
1543 	case IPPROTO_IPV6:
1544 		name = linux_to_bsd_ip6_sockopt(args->optname);
1545 		break;
1546 	case IPPROTO_TCP:
1547 		name = linux_to_bsd_tcp_sockopt(args->optname);
1548 		break;
1549 	default:
1550 		name = -1;
1551 		break;
1552 	}
1553 	if (name == -1)
1554 		return (EINVAL);
1555 
1556 	bsd_args.name = name;
1557 	bsd_args.avalsize = PTRIN(args->optlen);
1558 
1559 	if (name == IPV6_NEXTHOP) {
1560 		error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1561                 if (error != 0)
1562                         return (error);
1563 		sa = malloc(len, M_SONAME, M_WAITOK);
1564 
1565 		error = kern_getsockopt(td, args->s, bsd_args.level,
1566 		    name, sa, UIO_SYSSPACE, &len);
1567 		if (error != 0)
1568 			goto out;
1569 
1570 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
1571 		if (error == 0)
1572 			error = copyout(lsa, PTRIN(args->optval), len);
1573 		free(lsa, M_SONAME);
1574 		if (error == 0)
1575 			error = copyout(&len, PTRIN(args->optlen),
1576 			    sizeof(len));
1577 out:
1578 		free(sa, M_SONAME);
1579 	} else {
1580 		bsd_args.val = PTRIN(args->optval);
1581 		error = sys_getsockopt(td, &bsd_args);
1582 	}
1583 
1584 	return (error);
1585 }
1586 
1587 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1588 
1589 /* Argument list sizes for linux_socketcall */
1590 static const unsigned char lxs_args_cnt[] = {
1591 	0 /* unused*/,		3 /* socket */,
1592 	3 /* bind */,		3 /* connect */,
1593 	2 /* listen */,		3 /* accept */,
1594 	3 /* getsockname */,	3 /* getpeername */,
1595 	4 /* socketpair */,	4 /* send */,
1596 	4 /* recv */,		6 /* sendto */,
1597 	6 /* recvfrom */,	2 /* shutdown */,
1598 	5 /* setsockopt */,	5 /* getsockopt */,
1599 	3 /* sendmsg */,	3 /* recvmsg */,
1600 	4 /* accept4 */,	5 /* recvmmsg */,
1601 	4 /* sendmmsg */
1602 };
1603 #define	LINUX_ARGS_CNT		(nitems(lxs_args_cnt) - 1)
1604 #define	LINUX_ARG_SIZE(x)	(lxs_args_cnt[x] * sizeof(l_ulong))
1605 
1606 int
1607 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1608 {
1609 	l_ulong a[6];
1610 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1611 	register_t l_args[6];
1612 #endif
1613 	void *arg;
1614 	int error;
1615 
1616 	if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
1617 		return (EINVAL);
1618 	error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
1619 	if (error != 0)
1620 		return (error);
1621 
1622 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1623 	for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
1624 		l_args[i] = a[i];
1625 	arg = l_args;
1626 #else
1627 	arg = a;
1628 #endif
1629 	switch (args->what) {
1630 	case LINUX_SOCKET:
1631 		return (linux_socket(td, arg));
1632 	case LINUX_BIND:
1633 		return (linux_bind(td, arg));
1634 	case LINUX_CONNECT:
1635 		return (linux_connect(td, arg));
1636 	case LINUX_LISTEN:
1637 		return (linux_listen(td, arg));
1638 	case LINUX_ACCEPT:
1639 		return (linux_accept(td, arg));
1640 	case LINUX_GETSOCKNAME:
1641 		return (linux_getsockname(td, arg));
1642 	case LINUX_GETPEERNAME:
1643 		return (linux_getpeername(td, arg));
1644 	case LINUX_SOCKETPAIR:
1645 		return (linux_socketpair(td, arg));
1646 	case LINUX_SEND:
1647 		return (linux_send(td, arg));
1648 	case LINUX_RECV:
1649 		return (linux_recv(td, arg));
1650 	case LINUX_SENDTO:
1651 		return (linux_sendto(td, arg));
1652 	case LINUX_RECVFROM:
1653 		return (linux_recvfrom(td, arg));
1654 	case LINUX_SHUTDOWN:
1655 		return (linux_shutdown(td, arg));
1656 	case LINUX_SETSOCKOPT:
1657 		return (linux_setsockopt(td, arg));
1658 	case LINUX_GETSOCKOPT:
1659 		return (linux_getsockopt(td, arg));
1660 	case LINUX_SENDMSG:
1661 		return (linux_sendmsg(td, arg));
1662 	case LINUX_RECVMSG:
1663 		return (linux_recvmsg(td, arg));
1664 	case LINUX_ACCEPT4:
1665 		return (linux_accept4(td, arg));
1666 	case LINUX_RECVMMSG:
1667 		return (linux_recvmmsg(td, arg));
1668 	case LINUX_SENDMMSG:
1669 		return (linux_sendmmsg(td, arg));
1670 	}
1671 
1672 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1673 	return (ENOSYS);
1674 }
1675 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1676