xref: /freebsd/sys/compat/linux/linux_socket.c (revision 99282790b7d01ec3c4072621d46a0d7302517ad4)
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/stat.h>
53 #include <sys/syslog.h>
54 #include <sys/un.h>
55 #include <sys/unistd.h>
56 
57 #include <security/audit/audit.h>
58 
59 #include <net/if.h>
60 #include <net/vnet.h>
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/ip.h>
64 #include <netinet/tcp.h>
65 #ifdef INET6
66 #include <netinet/ip6.h>
67 #include <netinet6/ip6_var.h>
68 #endif
69 
70 #ifdef COMPAT_LINUX32
71 #include <machine/../linux32/linux.h>
72 #include <machine/../linux32/linux32_proto.h>
73 #else
74 #include <machine/../linux/linux.h>
75 #include <machine/../linux/linux_proto.h>
76 #endif
77 #include <compat/linux/linux_common.h>
78 #include <compat/linux/linux_file.h>
79 #include <compat/linux/linux_mib.h>
80 #include <compat/linux/linux_socket.h>
81 #include <compat/linux/linux_timer.h>
82 #include <compat/linux/linux_util.h>
83 
84 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
85 					l_uint);
86 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
87 					l_uint, struct msghdr *);
88 static int linux_set_socket_flags(int, int *);
89 
90 
91 static int
92 linux_to_bsd_sockopt_level(int level)
93 {
94 
95 	if (level == LINUX_SOL_SOCKET)
96 		return (SOL_SOCKET);
97 	/* Remaining values are RFC-defined protocol numbers. */
98 	return (level);
99 }
100 
101 static int
102 bsd_to_linux_sockopt_level(int level)
103 {
104 
105 	if (level == SOL_SOCKET)
106 		return (LINUX_SOL_SOCKET);
107 	return (level);
108 }
109 
110 static int
111 linux_to_bsd_ip_sockopt(int opt)
112 {
113 
114 	switch (opt) {
115 	case LINUX_IP_TOS:
116 		return (IP_TOS);
117 	case LINUX_IP_TTL:
118 		return (IP_TTL);
119 	case LINUX_IP_OPTIONS:
120 		return (IP_OPTIONS);
121 	case LINUX_IP_MULTICAST_IF:
122 		return (IP_MULTICAST_IF);
123 	case LINUX_IP_MULTICAST_TTL:
124 		return (IP_MULTICAST_TTL);
125 	case LINUX_IP_MULTICAST_LOOP:
126 		return (IP_MULTICAST_LOOP);
127 	case LINUX_IP_ADD_MEMBERSHIP:
128 		return (IP_ADD_MEMBERSHIP);
129 	case LINUX_IP_DROP_MEMBERSHIP:
130 		return (IP_DROP_MEMBERSHIP);
131 	case LINUX_IP_HDRINCL:
132 		return (IP_HDRINCL);
133 	}
134 	return (-1);
135 }
136 
137 static int
138 linux_to_bsd_ip6_sockopt(int opt)
139 {
140 
141 	switch (opt) {
142 	case LINUX_IPV6_NEXTHOP:
143 		return (IPV6_NEXTHOP);
144 	case LINUX_IPV6_UNICAST_HOPS:
145 		return (IPV6_UNICAST_HOPS);
146 	case LINUX_IPV6_MULTICAST_IF:
147 		return (IPV6_MULTICAST_IF);
148 	case LINUX_IPV6_MULTICAST_HOPS:
149 		return (IPV6_MULTICAST_HOPS);
150 	case LINUX_IPV6_MULTICAST_LOOP:
151 		return (IPV6_MULTICAST_LOOP);
152 	case LINUX_IPV6_ADD_MEMBERSHIP:
153 		return (IPV6_JOIN_GROUP);
154 	case LINUX_IPV6_DROP_MEMBERSHIP:
155 		return (IPV6_LEAVE_GROUP);
156 	case LINUX_IPV6_V6ONLY:
157 		return (IPV6_V6ONLY);
158 	case LINUX_IPV6_DONTFRAG:
159 		return (IPV6_DONTFRAG);
160 #if 0
161 	case LINUX_IPV6_CHECKSUM:
162 		return (IPV6_CHECKSUM);
163 	case LINUX_IPV6_RECVPKTINFO:
164 		return (IPV6_RECVPKTINFO);
165 	case LINUX_IPV6_PKTINFO:
166 		return (IPV6_PKTINFO);
167 	case LINUX_IPV6_RECVHOPLIMIT:
168 		return (IPV6_RECVHOPLIMIT);
169 	case LINUX_IPV6_HOPLIMIT:
170 		return (IPV6_HOPLIMIT);
171 	case LINUX_IPV6_RECVHOPOPTS:
172 		return (IPV6_RECVHOPOPTS);
173 	case LINUX_IPV6_HOPOPTS:
174 		return (IPV6_HOPOPTS);
175 	case LINUX_IPV6_RTHDRDSTOPTS:
176 		return (IPV6_RTHDRDSTOPTS);
177 	case LINUX_IPV6_RECVRTHDR:
178 		return (IPV6_RECVRTHDR);
179 	case LINUX_IPV6_RTHDR:
180 		return (IPV6_RTHDR);
181 	case LINUX_IPV6_RECVDSTOPTS:
182 		return (IPV6_RECVDSTOPTS);
183 	case LINUX_IPV6_DSTOPTS:
184 		return (IPV6_DSTOPTS);
185 	case LINUX_IPV6_RECVPATHMTU:
186 		return (IPV6_RECVPATHMTU);
187 	case LINUX_IPV6_PATHMTU:
188 		return (IPV6_PATHMTU);
189 #endif
190 	}
191 	return (-1);
192 }
193 
194 static int
195 linux_to_bsd_so_sockopt(int opt)
196 {
197 
198 	switch (opt) {
199 	case LINUX_SO_DEBUG:
200 		return (SO_DEBUG);
201 	case LINUX_SO_REUSEADDR:
202 		return (SO_REUSEADDR);
203 	case LINUX_SO_TYPE:
204 		return (SO_TYPE);
205 	case LINUX_SO_ERROR:
206 		return (SO_ERROR);
207 	case LINUX_SO_DONTROUTE:
208 		return (SO_DONTROUTE);
209 	case LINUX_SO_BROADCAST:
210 		return (SO_BROADCAST);
211 	case LINUX_SO_SNDBUF:
212 	case LINUX_SO_SNDBUFFORCE:
213 		return (SO_SNDBUF);
214 	case LINUX_SO_RCVBUF:
215 	case LINUX_SO_RCVBUFFORCE:
216 		return (SO_RCVBUF);
217 	case LINUX_SO_KEEPALIVE:
218 		return (SO_KEEPALIVE);
219 	case LINUX_SO_OOBINLINE:
220 		return (SO_OOBINLINE);
221 	case LINUX_SO_LINGER:
222 		return (SO_LINGER);
223 	case LINUX_SO_REUSEPORT:
224 		return (SO_REUSEPORT_LB);
225 	case LINUX_SO_PEERCRED:
226 		return (LOCAL_PEERCRED);
227 	case LINUX_SO_RCVLOWAT:
228 		return (SO_RCVLOWAT);
229 	case LINUX_SO_SNDLOWAT:
230 		return (SO_SNDLOWAT);
231 	case LINUX_SO_RCVTIMEO:
232 		return (SO_RCVTIMEO);
233 	case LINUX_SO_SNDTIMEO:
234 		return (SO_SNDTIMEO);
235 	case LINUX_SO_TIMESTAMP:
236 		return (SO_TIMESTAMP);
237 	case LINUX_SO_ACCEPTCONN:
238 		return (SO_ACCEPTCONN);
239 	}
240 	return (-1);
241 }
242 
243 static int
244 linux_to_bsd_tcp_sockopt(int opt)
245 {
246 
247 	switch (opt) {
248 	case LINUX_TCP_NODELAY:
249 		return (TCP_NODELAY);
250 	case LINUX_TCP_MAXSEG:
251 		return (TCP_MAXSEG);
252 	case LINUX_TCP_CORK:
253 		return (TCP_NOPUSH);
254 	case LINUX_TCP_KEEPIDLE:
255 		return (TCP_KEEPIDLE);
256 	case LINUX_TCP_KEEPINTVL:
257 		return (TCP_KEEPINTVL);
258 	case LINUX_TCP_KEEPCNT:
259 		return (TCP_KEEPCNT);
260 	case LINUX_TCP_MD5SIG:
261 		return (TCP_MD5SIG);
262 	}
263 	return (-1);
264 }
265 
266 static int
267 linux_to_bsd_msg_flags(int flags)
268 {
269 	int ret_flags = 0;
270 
271 	if (flags & LINUX_MSG_OOB)
272 		ret_flags |= MSG_OOB;
273 	if (flags & LINUX_MSG_PEEK)
274 		ret_flags |= MSG_PEEK;
275 	if (flags & LINUX_MSG_DONTROUTE)
276 		ret_flags |= MSG_DONTROUTE;
277 	if (flags & LINUX_MSG_CTRUNC)
278 		ret_flags |= MSG_CTRUNC;
279 	if (flags & LINUX_MSG_TRUNC)
280 		ret_flags |= MSG_TRUNC;
281 	if (flags & LINUX_MSG_DONTWAIT)
282 		ret_flags |= MSG_DONTWAIT;
283 	if (flags & LINUX_MSG_EOR)
284 		ret_flags |= MSG_EOR;
285 	if (flags & LINUX_MSG_WAITALL)
286 		ret_flags |= MSG_WAITALL;
287 	if (flags & LINUX_MSG_NOSIGNAL)
288 		ret_flags |= MSG_NOSIGNAL;
289 #if 0 /* not handled */
290 	if (flags & LINUX_MSG_PROXY)
291 		;
292 	if (flags & LINUX_MSG_FIN)
293 		;
294 	if (flags & LINUX_MSG_SYN)
295 		;
296 	if (flags & LINUX_MSG_CONFIRM)
297 		;
298 	if (flags & LINUX_MSG_RST)
299 		;
300 	if (flags & LINUX_MSG_ERRQUEUE)
301 		;
302 #endif
303 	return (ret_flags);
304 }
305 
306 static int
307 linux_to_bsd_cmsg_type(int cmsg_type)
308 {
309 
310 	switch (cmsg_type) {
311 	case LINUX_SCM_RIGHTS:
312 		return (SCM_RIGHTS);
313 	case LINUX_SCM_CREDENTIALS:
314 		return (SCM_CREDS);
315 	}
316 	return (-1);
317 }
318 
319 static int
320 bsd_to_linux_cmsg_type(int cmsg_type)
321 {
322 
323 	switch (cmsg_type) {
324 	case SCM_RIGHTS:
325 		return (LINUX_SCM_RIGHTS);
326 	case SCM_CREDS:
327 		return (LINUX_SCM_CREDENTIALS);
328 	case SCM_TIMESTAMP:
329 		return (LINUX_SCM_TIMESTAMP);
330 	}
331 	return (-1);
332 }
333 
334 static int
335 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
336 {
337 	if (lhdr->msg_controllen > INT_MAX)
338 		return (ENOBUFS);
339 
340 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
341 	bhdr->msg_namelen	= lhdr->msg_namelen;
342 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
343 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
344 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
345 
346 	/*
347 	 * msg_controllen is skipped since BSD and LINUX control messages
348 	 * are potentially different sizes (e.g. the cred structure used
349 	 * by SCM_CREDS is different between the two operating system).
350 	 *
351 	 * The caller can set it (if necessary) after converting all the
352 	 * control messages.
353 	 */
354 
355 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
356 	return (0);
357 }
358 
359 static int
360 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
361 {
362 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
363 	lhdr->msg_namelen	= bhdr->msg_namelen;
364 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
365 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
366 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
367 
368 	/*
369 	 * msg_controllen is skipped since BSD and LINUX control messages
370 	 * are potentially different sizes (e.g. the cred structure used
371 	 * by SCM_CREDS is different between the two operating system).
372 	 *
373 	 * The caller can set it (if necessary) after converting all the
374 	 * control messages.
375 	 */
376 
377 	/* msg_flags skipped */
378 	return (0);
379 }
380 
381 static int
382 linux_set_socket_flags(int lflags, int *flags)
383 {
384 
385 	if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
386 		return (EINVAL);
387 	if (lflags & LINUX_SOCK_NONBLOCK)
388 		*flags |= SOCK_NONBLOCK;
389 	if (lflags & LINUX_SOCK_CLOEXEC)
390 		*flags |= SOCK_CLOEXEC;
391 	return (0);
392 }
393 
394 static int
395 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
396     struct mbuf *control, enum uio_seg segflg)
397 {
398 	struct sockaddr *to;
399 	int error, len;
400 
401 	if (mp->msg_name != NULL) {
402 		len = mp->msg_namelen;
403 		error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len);
404 		if (error != 0)
405 			return (error);
406 		mp->msg_name = to;
407 	} else
408 		to = NULL;
409 
410 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
411 	    segflg);
412 
413 	if (to)
414 		free(to, M_SONAME);
415 	return (error);
416 }
417 
418 /* Return 0 if IP_HDRINCL is set for the given socket. */
419 static int
420 linux_check_hdrincl(struct thread *td, int s)
421 {
422 	int error, optval;
423 	socklen_t size_val;
424 
425 	size_val = sizeof(optval);
426 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
427 	    &optval, UIO_SYSSPACE, &size_val);
428 	if (error != 0)
429 		return (error);
430 
431 	return (optval == 0);
432 }
433 
434 /*
435  * Updated sendto() when IP_HDRINCL is set:
436  * tweak endian-dependent fields in the IP packet.
437  */
438 static int
439 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
440 {
441 /*
442  * linux_ip_copysize defines how many bytes we should copy
443  * from the beginning of the IP packet before we customize it for BSD.
444  * It should include all the fields we modify (ip_len and ip_off).
445  */
446 #define linux_ip_copysize	8
447 
448 	struct ip *packet;
449 	struct msghdr msg;
450 	struct iovec aiov[1];
451 	int error;
452 
453 	/* Check that the packet isn't too big or too small. */
454 	if (linux_args->len < linux_ip_copysize ||
455 	    linux_args->len > IP_MAXPACKET)
456 		return (EINVAL);
457 
458 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
459 
460 	/* Make kernel copy of the packet to be sent */
461 	if ((error = copyin(PTRIN(linux_args->msg), packet,
462 	    linux_args->len)))
463 		goto goout;
464 
465 	/* Convert fields from Linux to BSD raw IP socket format */
466 	packet->ip_len = linux_args->len;
467 	packet->ip_off = ntohs(packet->ip_off);
468 
469 	/* Prepare the msghdr and iovec structures describing the new packet */
470 	msg.msg_name = PTRIN(linux_args->to);
471 	msg.msg_namelen = linux_args->tolen;
472 	msg.msg_iov = aiov;
473 	msg.msg_iovlen = 1;
474 	msg.msg_control = NULL;
475 	msg.msg_flags = 0;
476 	aiov[0].iov_base = (char *)packet;
477 	aiov[0].iov_len = linux_args->len;
478 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
479 	    NULL, UIO_SYSSPACE);
480 goout:
481 	free(packet, M_LINUX);
482 	return (error);
483 }
484 
485 int
486 linux_socket(struct thread *td, struct linux_socket_args *args)
487 {
488 	int domain, retval_socket, type;
489 
490 	type = args->type & LINUX_SOCK_TYPE_MASK;
491 	if (type < 0 || type > LINUX_SOCK_MAX)
492 		return (EINVAL);
493 	retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
494 		&type);
495 	if (retval_socket != 0)
496 		return (retval_socket);
497 	domain = linux_to_bsd_domain(args->domain);
498 	if (domain == -1)
499 		return (EAFNOSUPPORT);
500 
501 	retval_socket = kern_socket(td, domain, type, args->protocol);
502 	if (retval_socket)
503 		return (retval_socket);
504 
505 	if (type == SOCK_RAW
506 	    && (args->protocol == IPPROTO_RAW || args->protocol == 0)
507 	    && domain == PF_INET) {
508 		/* It's a raw IP socket: set the IP_HDRINCL option. */
509 		int hdrincl;
510 
511 		hdrincl = 1;
512 		/* We ignore any error returned by kern_setsockopt() */
513 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
514 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
515 	}
516 #ifdef INET6
517 	/*
518 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
519 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
520 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
521 	 * sysctl value.
522 	 */
523 	if (domain == PF_INET6) {
524 		int v6only;
525 
526 		v6only = 0;
527 		/* We ignore any error returned by setsockopt() */
528 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
529 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
530 	}
531 #endif
532 
533 	return (retval_socket);
534 }
535 
536 int
537 linux_bind(struct thread *td, struct linux_bind_args *args)
538 {
539 	struct sockaddr *sa;
540 	int error;
541 
542 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
543 	    &args->namelen);
544 	if (error != 0)
545 		return (error);
546 
547 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
548 	free(sa, M_SONAME);
549 
550 	/* XXX */
551 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
552 		return (EINVAL);
553 	return (error);
554 }
555 
556 int
557 linux_connect(struct thread *td, struct linux_connect_args *args)
558 {
559 	struct socket *so;
560 	struct sockaddr *sa;
561 	struct file *fp;
562 	u_int fflag;
563 	int error;
564 
565 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
566 	    &args->namelen);
567 	if (error != 0)
568 		return (error);
569 
570 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
571 	free(sa, M_SONAME);
572 	if (error != EISCONN)
573 		return (error);
574 
575 	/*
576 	 * Linux doesn't return EISCONN the first time it occurs,
577 	 * when on a non-blocking socket. Instead it returns the
578 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
579 	 */
580 	error = getsock_cap(td, args->s, &cap_connect_rights,
581 	    &fp, &fflag, NULL);
582 	if (error != 0)
583 		return (error);
584 
585 	error = EISCONN;
586 	so = fp->f_data;
587 	if (fflag & FNONBLOCK) {
588 		SOCK_LOCK(so);
589 		if (so->so_emuldata == 0)
590 			error = so->so_error;
591 		so->so_emuldata = (void *)1;
592 		SOCK_UNLOCK(so);
593 	}
594 	fdrop(fp, td);
595 
596 	return (error);
597 }
598 
599 int
600 linux_listen(struct thread *td, struct linux_listen_args *args)
601 {
602 
603 	return (kern_listen(td, args->s, args->backlog));
604 }
605 
606 static int
607 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
608     l_uintptr_t namelen, int flags)
609 {
610 	struct l_sockaddr *lsa;
611 	struct sockaddr *sa;
612 	struct file *fp;
613 	int bflags, len;
614 	struct socket *so;
615 	int error, error1;
616 
617 	bflags = 0;
618 	error = linux_set_socket_flags(flags, &bflags);
619 	if (error != 0)
620 		return (error);
621 
622 	sa = NULL;
623 	if (PTRIN(addr) == NULL) {
624 		len = 0;
625 		error = kern_accept4(td, s, NULL, NULL, bflags, NULL);
626 	} else {
627 		error = copyin(PTRIN(namelen), &len, sizeof(len));
628 		if (error != 0)
629 			return (error);
630 		if (len < 0)
631 			return (EINVAL);
632 		error = kern_accept4(td, s, &sa, &len, bflags, &fp);
633 		if (error == 0)
634 			fdrop(fp, td);
635 	}
636 
637 	if (error != 0) {
638 		/*
639 		 * XXX. This is wrong, different sockaddr structures
640 		 * have different sizes.
641 		 */
642 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
643 		{
644 			error = EINVAL;
645 			goto out;
646 		}
647 		if (error == EINVAL) {
648 			error1 = getsock_cap(td, s, &cap_accept_rights, &fp, NULL, NULL);
649 			if (error1 != 0) {
650 				error = error1;
651 				goto out;
652 			}
653 			so = fp->f_data;
654 			if (so->so_type == SOCK_DGRAM)
655 				error = EOPNOTSUPP;
656 			fdrop(fp, td);
657 		}
658 		goto out;
659 	}
660 
661 	if (len != 0 && error == 0) {
662 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
663 		if (error == 0)
664 			error = copyout(lsa, PTRIN(addr), len);
665 		free(lsa, M_SONAME);
666 	}
667 
668 	free(sa, M_SONAME);
669 
670 out:
671 	if (error != 0) {
672 		(void)kern_close(td, td->td_retval[0]);
673 		td->td_retval[0] = 0;
674 	}
675 	return (error);
676 }
677 
678 int
679 linux_accept(struct thread *td, struct linux_accept_args *args)
680 {
681 
682 	return (linux_accept_common(td, args->s, args->addr,
683 	    args->namelen, 0));
684 }
685 
686 int
687 linux_accept4(struct thread *td, struct linux_accept4_args *args)
688 {
689 
690 	return (linux_accept_common(td, args->s, args->addr,
691 	    args->namelen, args->flags));
692 }
693 
694 int
695 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
696 {
697 	struct l_sockaddr *lsa;
698 	struct sockaddr *sa;
699 	int len, error;
700 
701 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
702 	if (error != 0)
703 		return (error);
704 
705 	error = kern_getsockname(td, args->s, &sa, &len);
706 	if (error != 0)
707 		return (error);
708 
709 	if (len != 0) {
710 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
711 		if (error == 0)
712 			error = copyout(lsa, PTRIN(args->addr),
713 			    len);
714 		free(lsa, M_SONAME);
715 	}
716 
717 	free(sa, M_SONAME);
718 	if (error == 0)
719 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
720 	return (error);
721 }
722 
723 int
724 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
725 {
726 	struct l_sockaddr *lsa;
727 	struct sockaddr *sa;
728 	int len, error;
729 
730 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
731 	if (error != 0)
732 		return (error);
733 	if (len < 0)
734 		return (EINVAL);
735 
736 	error = kern_getpeername(td, args->s, &sa, &len);
737 	if (error != 0)
738 		return (error);
739 
740 	if (len != 0) {
741 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
742 		if (error == 0)
743 			error = copyout(lsa, PTRIN(args->addr),
744 			    len);
745 		free(lsa, M_SONAME);
746 	}
747 
748 	free(sa, M_SONAME);
749 	if (error == 0)
750 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
751 	return (error);
752 }
753 
754 int
755 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
756 {
757 	int domain, error, sv[2], type;
758 
759 	domain = linux_to_bsd_domain(args->domain);
760 	if (domain != PF_LOCAL)
761 		return (EAFNOSUPPORT);
762 	type = args->type & LINUX_SOCK_TYPE_MASK;
763 	if (type < 0 || type > LINUX_SOCK_MAX)
764 		return (EINVAL);
765 	error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
766 	    &type);
767 	if (error != 0)
768 		return (error);
769 	if (args->protocol != 0 && args->protocol != PF_UNIX) {
770 
771 		/*
772 		 * Use of PF_UNIX as protocol argument is not right,
773 		 * but Linux does it.
774 		 * Do not map PF_UNIX as its Linux value is identical
775 		 * to FreeBSD one.
776 		 */
777 		return (EPROTONOSUPPORT);
778 	}
779 	error = kern_socketpair(td, domain, type, 0, sv);
780 	if (error != 0)
781                 return (error);
782         error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int));
783         if (error != 0) {
784                 (void)kern_close(td, sv[0]);
785                 (void)kern_close(td, sv[1]);
786         }
787 	return (error);
788 }
789 
790 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
791 struct linux_send_args {
792 	register_t s;
793 	register_t msg;
794 	register_t len;
795 	register_t flags;
796 };
797 
798 static int
799 linux_send(struct thread *td, struct linux_send_args *args)
800 {
801 	struct sendto_args /* {
802 		int s;
803 		caddr_t buf;
804 		int len;
805 		int flags;
806 		caddr_t to;
807 		int tolen;
808 	} */ bsd_args;
809 	struct file *fp;
810 	int error, fflag;
811 
812 	bsd_args.s = args->s;
813 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
814 	bsd_args.len = args->len;
815 	bsd_args.flags = args->flags;
816 	bsd_args.to = NULL;
817 	bsd_args.tolen = 0;
818 	error = sys_sendto(td, &bsd_args);
819 	if (error == ENOTCONN) {
820 		/*
821 		 * Linux doesn't return ENOTCONN for non-blocking sockets.
822 		 * Instead it returns the EAGAIN.
823 		 */
824 		error = getsock_cap(td, args->s, &cap_send_rights, &fp,
825 		    &fflag, NULL);
826 		if (error == 0) {
827 			if (fflag & FNONBLOCK)
828 				error = EAGAIN;
829 			fdrop(fp, td);
830 		}
831 	}
832 	return (error);
833 }
834 
835 struct linux_recv_args {
836 	register_t s;
837 	register_t msg;
838 	register_t len;
839 	register_t flags;
840 };
841 
842 static int
843 linux_recv(struct thread *td, struct linux_recv_args *args)
844 {
845 	struct recvfrom_args /* {
846 		int s;
847 		caddr_t buf;
848 		int len;
849 		int flags;
850 		struct sockaddr *from;
851 		socklen_t fromlenaddr;
852 	} */ bsd_args;
853 
854 	bsd_args.s = args->s;
855 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
856 	bsd_args.len = args->len;
857 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
858 	bsd_args.from = NULL;
859 	bsd_args.fromlenaddr = 0;
860 	return (sys_recvfrom(td, &bsd_args));
861 }
862 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
863 
864 int
865 linux_sendto(struct thread *td, struct linux_sendto_args *args)
866 {
867 	struct msghdr msg;
868 	struct iovec aiov;
869 
870 	if (linux_check_hdrincl(td, args->s) == 0)
871 		/* IP_HDRINCL set, tweak the packet before sending */
872 		return (linux_sendto_hdrincl(td, args));
873 
874 	msg.msg_name = PTRIN(args->to);
875 	msg.msg_namelen = args->tolen;
876 	msg.msg_iov = &aiov;
877 	msg.msg_iovlen = 1;
878 	msg.msg_control = NULL;
879 	msg.msg_flags = 0;
880 	aiov.iov_base = PTRIN(args->msg);
881 	aiov.iov_len = args->len;
882 	return (linux_sendit(td, args->s, &msg, args->flags, NULL,
883 	    UIO_USERSPACE));
884 }
885 
886 int
887 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
888 {
889 	struct l_sockaddr *lsa;
890 	struct sockaddr *sa;
891 	struct msghdr msg;
892 	struct iovec aiov;
893 	int error, fromlen;
894 
895 	if (PTRIN(args->fromlen) != NULL) {
896 		error = copyin(PTRIN(args->fromlen), &fromlen,
897 		    sizeof(fromlen));
898 		if (error != 0)
899 			return (error);
900 		if (fromlen < 0)
901 			return (EINVAL);
902 		sa = malloc(fromlen, M_SONAME, M_WAITOK);
903 	} else {
904 		fromlen = 0;
905 		sa = NULL;
906 	}
907 
908 	msg.msg_name = sa;
909 	msg.msg_namelen = fromlen;
910 	msg.msg_iov = &aiov;
911 	msg.msg_iovlen = 1;
912 	aiov.iov_base = PTRIN(args->buf);
913 	aiov.iov_len = args->len;
914 	msg.msg_control = 0;
915 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
916 
917 	error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL);
918 	if (error != 0)
919 		goto out;
920 
921 	if (PTRIN(args->from) != NULL) {
922 		error = bsd_to_linux_sockaddr(sa, &lsa, msg.msg_namelen);
923 		if (error == 0)
924 			error = copyout(lsa, PTRIN(args->from),
925 			    msg.msg_namelen);
926 		free(lsa, M_SONAME);
927 	}
928 
929 	if (error == 0 && PTRIN(args->fromlen) != NULL)
930 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
931 		    sizeof(msg.msg_namelen));
932 out:
933 	free(sa, M_SONAME);
934 	return (error);
935 }
936 
937 static int
938 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
939     l_uint flags)
940 {
941 	struct cmsghdr *cmsg;
942 	struct mbuf *control;
943 	struct msghdr msg;
944 	struct l_cmsghdr linux_cmsg;
945 	struct l_cmsghdr *ptr_cmsg;
946 	struct l_msghdr linux_msghdr;
947 	struct iovec *iov;
948 	socklen_t datalen;
949 	struct sockaddr *sa;
950 	struct socket *so;
951 	sa_family_t sa_family;
952 	struct file *fp;
953 	void *data;
954 	l_size_t len;
955 	l_size_t clen;
956 	int error, fflag;
957 
958 	error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
959 	if (error != 0)
960 		return (error);
961 
962 	/*
963 	 * Some Linux applications (ping) define a non-NULL control data
964 	 * pointer, but a msg_controllen of 0, which is not allowed in the
965 	 * FreeBSD system call interface.  NULL the msg_control pointer in
966 	 * order to handle this case.  This should be checked, but allows the
967 	 * Linux ping to work.
968 	 */
969 	if (PTRIN(linux_msghdr.msg_control) != NULL &&
970 	    linux_msghdr.msg_controllen == 0)
971 		linux_msghdr.msg_control = PTROUT(NULL);
972 
973 	error = linux_to_bsd_msghdr(&msg, &linux_msghdr);
974 	if (error != 0)
975 		return (error);
976 
977 #ifdef COMPAT_LINUX32
978 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
979 	    &iov, EMSGSIZE);
980 #else
981 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
982 #endif
983 	if (error != 0)
984 		return (error);
985 
986 	control = NULL;
987 
988 	error = kern_getsockname(td, s, &sa, &datalen);
989 	if (error != 0)
990 		goto bad;
991 	sa_family = sa->sa_family;
992 	free(sa, M_SONAME);
993 
994 	if (flags & LINUX_MSG_OOB) {
995 		error = EOPNOTSUPP;
996 		if (sa_family == AF_UNIX)
997 			goto bad;
998 
999 		error = getsock_cap(td, s, &cap_send_rights, &fp,
1000 		    &fflag, NULL);
1001 		if (error != 0)
1002 			goto bad;
1003 		so = fp->f_data;
1004 		if (so->so_type != SOCK_STREAM)
1005 			error = EOPNOTSUPP;
1006 		fdrop(fp, td);
1007 		if (error != 0)
1008 			goto bad;
1009 	}
1010 
1011 	if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) {
1012 
1013 		error = ENOBUFS;
1014 		control = m_get(M_WAITOK, MT_CONTROL);
1015 		MCLGET(control, M_WAITOK);
1016 		data = mtod(control, void *);
1017 		datalen = 0;
1018 
1019 		ptr_cmsg = PTRIN(linux_msghdr.msg_control);
1020 		clen = linux_msghdr.msg_controllen;
1021 		do {
1022 			error = copyin(ptr_cmsg, &linux_cmsg,
1023 			    sizeof(struct l_cmsghdr));
1024 			if (error != 0)
1025 				goto bad;
1026 
1027 			error = EINVAL;
1028 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1029 			    linux_cmsg.cmsg_len > clen)
1030 				goto bad;
1031 
1032 			if (datalen + CMSG_HDRSZ > MCLBYTES)
1033 				goto bad;
1034 
1035 			/*
1036 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1037 			 * so return EINVAL in any other cmsg_type
1038 			 */
1039 			cmsg = data;
1040 			cmsg->cmsg_type =
1041 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1042 			cmsg->cmsg_level =
1043 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1044 			if (cmsg->cmsg_type == -1
1045 			    || cmsg->cmsg_level != SOL_SOCKET) {
1046 				linux_msg(curthread,
1047 				    "unsupported sendmsg cmsg level %d type %d",
1048 				    linux_cmsg.cmsg_level, linux_cmsg.cmsg_type);
1049 				goto bad;
1050 			}
1051 
1052 			/*
1053 			 * Some applications (e.g. pulseaudio) attempt to
1054 			 * send ancillary data even if the underlying protocol
1055 			 * doesn't support it which is not allowed in the
1056 			 * FreeBSD system call interface.
1057 			 */
1058 			if (sa_family != AF_UNIX)
1059 				continue;
1060 
1061 			if (cmsg->cmsg_type == SCM_CREDS) {
1062 				len = sizeof(struct cmsgcred);
1063 				if (datalen + CMSG_SPACE(len) > MCLBYTES)
1064 					goto bad;
1065 
1066 				/*
1067 				 * The lower levels will fill in the structure
1068 				 */
1069 				memset(CMSG_DATA(data), 0, len);
1070 			} else {
1071 				len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1072 				if (datalen + CMSG_SPACE(len) < datalen ||
1073 				    datalen + CMSG_SPACE(len) > MCLBYTES)
1074 					goto bad;
1075 
1076 				error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1077 				    CMSG_DATA(data), len);
1078 				if (error != 0)
1079 					goto bad;
1080 			}
1081 
1082 			cmsg->cmsg_len = CMSG_LEN(len);
1083 			data = (char *)data + CMSG_SPACE(len);
1084 			datalen += CMSG_SPACE(len);
1085 
1086 			if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1087 				break;
1088 
1089 			clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1090 			ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1091 			    LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1092 		} while(clen >= sizeof(struct l_cmsghdr));
1093 
1094 		control->m_len = datalen;
1095 		if (datalen == 0) {
1096 			m_freem(control);
1097 			control = NULL;
1098 		}
1099 	}
1100 
1101 	msg.msg_iov = iov;
1102 	msg.msg_flags = 0;
1103 	error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1104 	control = NULL;
1105 
1106 bad:
1107 	m_freem(control);
1108 	free(iov, M_IOV);
1109 	return (error);
1110 }
1111 
1112 int
1113 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1114 {
1115 
1116 	return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1117 	    args->flags));
1118 }
1119 
1120 int
1121 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1122 {
1123 	struct l_mmsghdr *msg;
1124 	l_uint retval;
1125 	int error, datagrams;
1126 
1127 	if (args->vlen > UIO_MAXIOV)
1128 		args->vlen = UIO_MAXIOV;
1129 
1130 	msg = PTRIN(args->msg);
1131 	datagrams = 0;
1132 	while (datagrams < args->vlen) {
1133 		error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1134 		    args->flags);
1135 		if (error != 0)
1136 			break;
1137 
1138 		retval = td->td_retval[0];
1139 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1140 		if (error != 0)
1141 			break;
1142 		++msg;
1143 		++datagrams;
1144 	}
1145 	if (error == 0)
1146 		td->td_retval[0] = datagrams;
1147 	return (error);
1148 }
1149 
1150 static int
1151 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1152     l_uint flags, struct msghdr *msg)
1153 {
1154 	struct cmsghdr *cm;
1155 	struct cmsgcred *cmcred;
1156 	struct l_cmsghdr *linux_cmsg = NULL;
1157 	struct l_ucred linux_ucred;
1158 	socklen_t datalen, maxlen, outlen;
1159 	struct l_msghdr linux_msghdr;
1160 	struct iovec *iov, *uiov;
1161 	struct mbuf *control = NULL;
1162 	struct mbuf **controlp;
1163 	struct timeval *ftmvl;
1164 	struct l_sockaddr *lsa;
1165 	struct sockaddr *sa;
1166 	l_timeval ltmvl;
1167 	caddr_t outbuf;
1168 	void *data;
1169 	int error, i, fd, fds, *fdp;
1170 
1171 	error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
1172 	if (error != 0)
1173 		return (error);
1174 
1175 	error = linux_to_bsd_msghdr(msg, &linux_msghdr);
1176 	if (error != 0)
1177 		return (error);
1178 
1179 #ifdef COMPAT_LINUX32
1180 	error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1181 	    &iov, EMSGSIZE);
1182 #else
1183 	error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1184 #endif
1185 	if (error != 0)
1186 		return (error);
1187 
1188 	if (msg->msg_name) {
1189 		sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK);
1190 		msg->msg_name = sa;
1191 	} else
1192 		sa = NULL;
1193 
1194 	uiov = msg->msg_iov;
1195 	msg->msg_iov = iov;
1196 	controlp = (msg->msg_control != NULL) ? &control : NULL;
1197 	error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp);
1198 	msg->msg_iov = uiov;
1199 	if (error != 0)
1200 		goto bad;
1201 
1202 	if (msg->msg_name) {
1203 		msg->msg_name = PTRIN(linux_msghdr.msg_name);
1204 		error = bsd_to_linux_sockaddr(sa, &lsa, msg->msg_namelen);
1205 		if (error == 0)
1206 			error = copyout(lsa, PTRIN(msg->msg_name),
1207 			    msg->msg_namelen);
1208 		free(lsa, M_SONAME);
1209 		if (error != 0)
1210 			goto bad;
1211 	}
1212 
1213 	error = bsd_to_linux_msghdr(msg, &linux_msghdr);
1214 	if (error != 0)
1215 		goto bad;
1216 
1217 	maxlen = linux_msghdr.msg_controllen;
1218 	linux_msghdr.msg_controllen = 0;
1219 	if (control) {
1220 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1221 
1222 		msg->msg_control = mtod(control, struct cmsghdr *);
1223 		msg->msg_controllen = control->m_len;
1224 
1225 		cm = CMSG_FIRSTHDR(msg);
1226 		outbuf = PTRIN(linux_msghdr.msg_control);
1227 		outlen = 0;
1228 		while (cm != NULL) {
1229 			linux_cmsg->cmsg_type =
1230 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1231 			linux_cmsg->cmsg_level =
1232 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1233 			if (linux_cmsg->cmsg_type == -1 ||
1234 			    cm->cmsg_level != SOL_SOCKET) {
1235 				linux_msg(curthread,
1236 				    "unsupported recvmsg cmsg level %d type %d",
1237 				    cm->cmsg_level, cm->cmsg_type);
1238 				error = EINVAL;
1239 				goto bad;
1240 			}
1241 
1242 			data = CMSG_DATA(cm);
1243 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1244 
1245 			switch (cm->cmsg_type) {
1246 			case SCM_RIGHTS:
1247 				if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1248 					fds = datalen / sizeof(int);
1249 					fdp = data;
1250 					for (i = 0; i < fds; i++) {
1251 						fd = *fdp++;
1252 						(void)kern_fcntl(td, fd,
1253 						    F_SETFD, FD_CLOEXEC);
1254 					}
1255 				}
1256 				break;
1257 
1258 			case SCM_CREDS:
1259 				/*
1260 				 * Currently LOCAL_CREDS is never in
1261 				 * effect for Linux so no need to worry
1262 				 * about sockcred
1263 				 */
1264 				if (datalen != sizeof(*cmcred)) {
1265 					error = EMSGSIZE;
1266 					goto bad;
1267 				}
1268 				cmcred = (struct cmsgcred *)data;
1269 				bzero(&linux_ucred, sizeof(linux_ucred));
1270 				linux_ucred.pid = cmcred->cmcred_pid;
1271 				linux_ucred.uid = cmcred->cmcred_uid;
1272 				linux_ucred.gid = cmcred->cmcred_gid;
1273 				data = &linux_ucred;
1274 				datalen = sizeof(linux_ucred);
1275 				break;
1276 
1277 			case SCM_TIMESTAMP:
1278 				if (datalen != sizeof(struct timeval)) {
1279 					error = EMSGSIZE;
1280 					goto bad;
1281 				}
1282 				ftmvl = (struct timeval *)data;
1283 				ltmvl.tv_sec = ftmvl->tv_sec;
1284 				ltmvl.tv_usec = ftmvl->tv_usec;
1285 				data = &ltmvl;
1286 				datalen = sizeof(ltmvl);
1287 				break;
1288 			}
1289 
1290 			if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1291 				if (outlen == 0) {
1292 					error = EMSGSIZE;
1293 					goto bad;
1294 				} else {
1295 					linux_msghdr.msg_flags |= LINUX_MSG_CTRUNC;
1296 					m_dispose_extcontrolm(control);
1297 					goto out;
1298 				}
1299 			}
1300 
1301 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1302 
1303 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1304 			if (error != 0)
1305 				goto bad;
1306 			outbuf += L_CMSG_HDRSZ;
1307 
1308 			error = copyout(data, outbuf, datalen);
1309 			if (error != 0)
1310 				goto bad;
1311 
1312 			outbuf += LINUX_CMSG_ALIGN(datalen);
1313 			outlen += LINUX_CMSG_LEN(datalen);
1314 
1315 			cm = CMSG_NXTHDR(msg, cm);
1316 		}
1317 		linux_msghdr.msg_controllen = outlen;
1318 	}
1319 
1320 out:
1321 	error = copyout(&linux_msghdr, msghdr, sizeof(linux_msghdr));
1322 
1323 bad:
1324 	if (control != NULL) {
1325 		if (error != 0)
1326 			m_dispose_extcontrolm(control);
1327 		m_freem(control);
1328 	}
1329 	free(iov, M_IOV);
1330 	free(linux_cmsg, M_LINUX);
1331 	free(sa, M_SONAME);
1332 
1333 	return (error);
1334 }
1335 
1336 int
1337 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1338 {
1339 	struct msghdr bsd_msg;
1340 
1341 	return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1342 	    args->flags, &bsd_msg));
1343 }
1344 
1345 int
1346 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1347 {
1348 	struct l_mmsghdr *msg;
1349 	struct msghdr bsd_msg;
1350 	struct l_timespec lts;
1351 	struct timespec ts, tts;
1352 	l_uint retval;
1353 	int error, datagrams;
1354 
1355 	if (args->timeout) {
1356 		error = copyin(args->timeout, &lts, sizeof(struct l_timespec));
1357 		if (error != 0)
1358 			return (error);
1359 		error = linux_to_native_timespec(&ts, &lts);
1360 		if (error != 0)
1361 			return (error);
1362 		getnanotime(&tts);
1363 		timespecadd(&tts, &ts, &tts);
1364 	}
1365 
1366 	msg = PTRIN(args->msg);
1367 	datagrams = 0;
1368 	while (datagrams < args->vlen) {
1369 		error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1370 		    args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1371 		if (error != 0)
1372 			break;
1373 
1374 		retval = td->td_retval[0];
1375 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1376 		if (error != 0)
1377 			break;
1378 		++msg;
1379 		++datagrams;
1380 
1381 		/*
1382 		 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1383 		 */
1384 		if (args->flags & LINUX_MSG_WAITFORONE)
1385 			args->flags |= LINUX_MSG_DONTWAIT;
1386 
1387 		/*
1388 		 * See BUGS section of recvmmsg(2).
1389 		 */
1390 		if (args->timeout) {
1391 			getnanotime(&ts);
1392 			timespecsub(&ts, &tts, &ts);
1393 			if (!timespecisset(&ts) || ts.tv_sec > 0)
1394 				break;
1395 		}
1396 		/* Out of band data, return right away. */
1397 		if (bsd_msg.msg_flags & MSG_OOB)
1398 			break;
1399 	}
1400 	if (error == 0)
1401 		td->td_retval[0] = datagrams;
1402 	return (error);
1403 }
1404 
1405 int
1406 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1407 {
1408 
1409 	return (kern_shutdown(td, args->s, args->how));
1410 }
1411 
1412 int
1413 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1414 {
1415 	l_timeval linux_tv;
1416 	struct sockaddr *sa;
1417 	struct timeval tv;
1418 	socklen_t len;
1419 	int error, level, name;
1420 
1421 	level = linux_to_bsd_sockopt_level(args->level);
1422 	switch (level) {
1423 	case SOL_SOCKET:
1424 		name = linux_to_bsd_so_sockopt(args->optname);
1425 		switch (name) {
1426 		case SO_RCVTIMEO:
1427 			/* FALLTHROUGH */
1428 		case SO_SNDTIMEO:
1429 			error = copyin(PTRIN(args->optval), &linux_tv,
1430 			    sizeof(linux_tv));
1431 			if (error != 0)
1432 				return (error);
1433 			tv.tv_sec = linux_tv.tv_sec;
1434 			tv.tv_usec = linux_tv.tv_usec;
1435 			return (kern_setsockopt(td, args->s, level,
1436 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1437 			/* NOTREACHED */
1438 		default:
1439 			break;
1440 		}
1441 		break;
1442 	case IPPROTO_IP:
1443 		if (args->optname == LINUX_IP_RECVERR &&
1444 		    linux_ignore_ip_recverr) {
1445 			/*
1446 			 * XXX: This is a hack to unbreak DNS resolution
1447 			 *	with glibc 2.30 and above.
1448 			 */
1449 			return (0);
1450 		}
1451 		name = linux_to_bsd_ip_sockopt(args->optname);
1452 		break;
1453 	case IPPROTO_IPV6:
1454 		name = linux_to_bsd_ip6_sockopt(args->optname);
1455 		break;
1456 	case IPPROTO_TCP:
1457 		name = linux_to_bsd_tcp_sockopt(args->optname);
1458 		break;
1459 	default:
1460 		name = -1;
1461 		break;
1462 	}
1463 	if (name == -1) {
1464 		linux_msg(curthread,
1465 		    "unsupported setsockopt level %d optname %d",
1466 		    args->level, args->optname);
1467 		return (ENOPROTOOPT);
1468 	}
1469 
1470 	if (name == IPV6_NEXTHOP) {
1471 		len = args->optlen;
1472 		error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len);
1473 		if (error != 0)
1474 			return (error);
1475 
1476 		error = kern_setsockopt(td, args->s, level,
1477 		    name, sa, UIO_SYSSPACE, len);
1478 		free(sa, M_SONAME);
1479 	} else {
1480 		error = kern_setsockopt(td, args->s, level,
1481 		    name, PTRIN(args->optval), UIO_USERSPACE, args->optlen);
1482 	}
1483 
1484 	return (error);
1485 }
1486 
1487 int
1488 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1489 {
1490 	l_timeval linux_tv;
1491 	struct timeval tv;
1492 	socklen_t tv_len, xulen, len;
1493 	struct l_sockaddr *lsa;
1494 	struct sockaddr *sa;
1495 	struct xucred xu;
1496 	struct l_ucred lxu;
1497 	int error, level, name, newval;
1498 
1499 	level = linux_to_bsd_sockopt_level(args->level);
1500 	switch (level) {
1501 	case SOL_SOCKET:
1502 		name = linux_to_bsd_so_sockopt(args->optname);
1503 		switch (name) {
1504 		case SO_RCVTIMEO:
1505 			/* FALLTHROUGH */
1506 		case SO_SNDTIMEO:
1507 			tv_len = sizeof(tv);
1508 			error = kern_getsockopt(td, args->s, level,
1509 			    name, &tv, UIO_SYSSPACE, &tv_len);
1510 			if (error != 0)
1511 				return (error);
1512 			linux_tv.tv_sec = tv.tv_sec;
1513 			linux_tv.tv_usec = tv.tv_usec;
1514 			return (copyout(&linux_tv, PTRIN(args->optval),
1515 			    sizeof(linux_tv)));
1516 			/* NOTREACHED */
1517 		case LOCAL_PEERCRED:
1518 			if (args->optlen < sizeof(lxu))
1519 				return (EINVAL);
1520 			/*
1521 			 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
1522 			 * but by the Unix socket's level 0.
1523 			 */
1524 			level = 0;
1525 			xulen = sizeof(xu);
1526 			error = kern_getsockopt(td, args->s, level,
1527 			    name, &xu, UIO_SYSSPACE, &xulen);
1528 			if (error != 0)
1529 				return (error);
1530 			lxu.pid = xu.cr_pid;
1531 			lxu.uid = xu.cr_uid;
1532 			lxu.gid = xu.cr_gid;
1533 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1534 			/* NOTREACHED */
1535 		case SO_ERROR:
1536 			len = sizeof(newval);
1537 			error = kern_getsockopt(td, args->s, level,
1538 			    name, &newval, UIO_SYSSPACE, &len);
1539 			if (error != 0)
1540 				return (error);
1541 			newval = -SV_ABI_ERRNO(td->td_proc, newval);
1542 			return (copyout(&newval, PTRIN(args->optval), len));
1543 			/* NOTREACHED */
1544 		default:
1545 			break;
1546 		}
1547 		break;
1548 	case IPPROTO_IP:
1549 		name = linux_to_bsd_ip_sockopt(args->optname);
1550 		break;
1551 	case IPPROTO_IPV6:
1552 		name = linux_to_bsd_ip6_sockopt(args->optname);
1553 		break;
1554 	case IPPROTO_TCP:
1555 		name = linux_to_bsd_tcp_sockopt(args->optname);
1556 		break;
1557 	default:
1558 		name = -1;
1559 		break;
1560 	}
1561 	if (name == -1) {
1562 		linux_msg(curthread,
1563 		    "unsupported getsockopt level %d optname %d",
1564 		    args->level, args->optname);
1565 		return (EINVAL);
1566 	}
1567 
1568 	if (name == IPV6_NEXTHOP) {
1569 		error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1570                 if (error != 0)
1571                         return (error);
1572 		sa = malloc(len, M_SONAME, M_WAITOK);
1573 
1574 		error = kern_getsockopt(td, args->s, level,
1575 		    name, sa, UIO_SYSSPACE, &len);
1576 		if (error != 0)
1577 			goto out;
1578 
1579 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
1580 		if (error == 0)
1581 			error = copyout(lsa, PTRIN(args->optval), len);
1582 		free(lsa, M_SONAME);
1583 		if (error == 0)
1584 			error = copyout(&len, PTRIN(args->optlen),
1585 			    sizeof(len));
1586 out:
1587 		free(sa, M_SONAME);
1588 	} else {
1589 		if (args->optval) {
1590 			error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1591 			if (error != 0)
1592 				return (error);
1593 		}
1594 		error = kern_getsockopt(td, args->s, level,
1595 		    name, PTRIN(args->optval), UIO_USERSPACE, &len);
1596 		if (error == 0)
1597 			error = copyout(&len, PTRIN(args->optlen),
1598 			    sizeof(len));
1599 	}
1600 
1601 	return (error);
1602 }
1603 
1604 static int
1605 linux_sendfile_common(struct thread *td, l_int out, l_int in,
1606     l_loff_t *offset, l_size_t count)
1607 {
1608 	off_t bytes_read;
1609 	int error;
1610 	l_loff_t current_offset;
1611 	struct file *fp;
1612 
1613 	AUDIT_ARG_FD(in);
1614 	error = fget_read(td, in, &cap_pread_rights, &fp);
1615 	if (error != 0)
1616 		return (error);
1617 
1618 	if (offset != NULL) {
1619 		current_offset = *offset;
1620 	} else {
1621 		error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1622 		    fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE;
1623 		if (error != 0)
1624 			goto drop;
1625 		current_offset = td->td_uretoff.tdu_off;
1626 	}
1627 
1628 	bytes_read = 0;
1629 
1630 	/* Linux cannot have 0 count. */
1631 	if (count <= 0 || current_offset < 0) {
1632 		error = EINVAL;
1633 		goto drop;
1634 	}
1635 
1636 	error = fo_sendfile(fp, out, NULL, NULL, current_offset, count,
1637 	    &bytes_read, 0, td);
1638 	if (error != 0)
1639 		goto drop;
1640 	current_offset += bytes_read;
1641 
1642 	if (offset != NULL) {
1643 		*offset = current_offset;
1644 	} else {
1645 		error = fo_seek(fp, current_offset, SEEK_SET, td);
1646 		if (error != 0)
1647 			goto drop;
1648 	}
1649 
1650 	td->td_retval[0] = (ssize_t)bytes_read;
1651 drop:
1652 	fdrop(fp, td);
1653 	return (error);
1654 }
1655 
1656 int
1657 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg)
1658 {
1659 	/*
1660 	 * Differences between FreeBSD and Linux sendfile:
1661 	 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to
1662 	 *   mean send the whole file.)  In linux_sendfile given fds are still
1663 	 *   checked for validity when the count is 0.
1664 	 * - Linux can send to any fd whereas FreeBSD only supports sockets.
1665 	 *   The same restriction follows for linux_sendfile.
1666 	 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr.
1667 	 * - Linux takes an offset pointer and updates it to the read location.
1668 	 *   FreeBSD takes in an offset and a 'bytes read' parameter which is
1669 	 *   only filled if it isn't NULL.  We use this parameter to update the
1670 	 *   offset pointer if it exists.
1671 	 * - Linux sendfile returns bytes read on success while FreeBSD
1672 	 *   returns 0.  We use the 'bytes read' parameter to get this value.
1673 	 */
1674 
1675 	l_loff_t offset64;
1676 	l_long offset;
1677 	int ret;
1678 	int error;
1679 
1680 	if (arg->offset != NULL) {
1681 		error = copyin(arg->offset, &offset, sizeof(offset));
1682 		if (error != 0)
1683 			return (error);
1684 		offset64 = (l_loff_t)offset;
1685 	}
1686 
1687 	ret = linux_sendfile_common(td, arg->out, arg->in,
1688 	    arg->offset != NULL ? &offset64 : NULL, arg->count);
1689 
1690 	if (arg->offset != NULL) {
1691 #if defined(__i386__) || defined(__arm__) || \
1692     (defined(__amd64__) && defined(COMPAT_LINUX32))
1693 		if (offset64 > INT32_MAX)
1694 			return (EOVERFLOW);
1695 #endif
1696 		offset = (l_long)offset64;
1697 		error = copyout(&offset, arg->offset, sizeof(offset));
1698 		if (error != 0)
1699 			return (error);
1700 	}
1701 
1702 	return (ret);
1703 }
1704 
1705 #if defined(__i386__) || defined(__arm__) || \
1706     (defined(__amd64__) && defined(COMPAT_LINUX32))
1707 
1708 int
1709 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg)
1710 {
1711 	l_loff_t offset;
1712 	int ret;
1713 	int error;
1714 
1715 	if (arg->offset != NULL) {
1716 		error = copyin(arg->offset, &offset, sizeof(offset));
1717 		if (error != 0)
1718 			return (error);
1719 	}
1720 
1721 	ret = linux_sendfile_common(td, arg->out, arg->in,
1722 		arg->offset != NULL ? &offset : NULL, arg->count);
1723 
1724 	if (arg->offset != NULL) {
1725 		error = copyout(&offset, arg->offset, sizeof(offset));
1726 		if (error != 0)
1727 			return (error);
1728 	}
1729 
1730 	return (ret);
1731 }
1732 
1733 /* Argument list sizes for linux_socketcall */
1734 static const unsigned char lxs_args_cnt[] = {
1735 	0 /* unused*/,		3 /* socket */,
1736 	3 /* bind */,		3 /* connect */,
1737 	2 /* listen */,		3 /* accept */,
1738 	3 /* getsockname */,	3 /* getpeername */,
1739 	4 /* socketpair */,	4 /* send */,
1740 	4 /* recv */,		6 /* sendto */,
1741 	6 /* recvfrom */,	2 /* shutdown */,
1742 	5 /* setsockopt */,	5 /* getsockopt */,
1743 	3 /* sendmsg */,	3 /* recvmsg */,
1744 	4 /* accept4 */,	5 /* recvmmsg */,
1745 	4 /* sendmmsg */,	4 /* sendfile */
1746 };
1747 #define	LINUX_ARGS_CNT		(nitems(lxs_args_cnt) - 1)
1748 #define	LINUX_ARG_SIZE(x)	(lxs_args_cnt[x] * sizeof(l_ulong))
1749 
1750 int
1751 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1752 {
1753 	l_ulong a[6];
1754 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1755 	register_t l_args[6];
1756 #endif
1757 	void *arg;
1758 	int error;
1759 
1760 	if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
1761 		return (EINVAL);
1762 	error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
1763 	if (error != 0)
1764 		return (error);
1765 
1766 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1767 	for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
1768 		l_args[i] = a[i];
1769 	arg = l_args;
1770 #else
1771 	arg = a;
1772 #endif
1773 	switch (args->what) {
1774 	case LINUX_SOCKET:
1775 		return (linux_socket(td, arg));
1776 	case LINUX_BIND:
1777 		return (linux_bind(td, arg));
1778 	case LINUX_CONNECT:
1779 		return (linux_connect(td, arg));
1780 	case LINUX_LISTEN:
1781 		return (linux_listen(td, arg));
1782 	case LINUX_ACCEPT:
1783 		return (linux_accept(td, arg));
1784 	case LINUX_GETSOCKNAME:
1785 		return (linux_getsockname(td, arg));
1786 	case LINUX_GETPEERNAME:
1787 		return (linux_getpeername(td, arg));
1788 	case LINUX_SOCKETPAIR:
1789 		return (linux_socketpair(td, arg));
1790 	case LINUX_SEND:
1791 		return (linux_send(td, arg));
1792 	case LINUX_RECV:
1793 		return (linux_recv(td, arg));
1794 	case LINUX_SENDTO:
1795 		return (linux_sendto(td, arg));
1796 	case LINUX_RECVFROM:
1797 		return (linux_recvfrom(td, arg));
1798 	case LINUX_SHUTDOWN:
1799 		return (linux_shutdown(td, arg));
1800 	case LINUX_SETSOCKOPT:
1801 		return (linux_setsockopt(td, arg));
1802 	case LINUX_GETSOCKOPT:
1803 		return (linux_getsockopt(td, arg));
1804 	case LINUX_SENDMSG:
1805 		return (linux_sendmsg(td, arg));
1806 	case LINUX_RECVMSG:
1807 		return (linux_recvmsg(td, arg));
1808 	case LINUX_ACCEPT4:
1809 		return (linux_accept4(td, arg));
1810 	case LINUX_RECVMMSG:
1811 		return (linux_recvmmsg(td, arg));
1812 	case LINUX_SENDMMSG:
1813 		return (linux_sendmmsg(td, arg));
1814 	case LINUX_SENDFILE:
1815 		return (linux_sendfile(td, arg));
1816 	}
1817 
1818 	linux_msg(td, "socket type %d not implemented", args->what);
1819 	return (ENOSYS);
1820 }
1821 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */
1822