xref: /freebsd/sys/compat/linux/linux_socket.c (revision 25fb30bd9abc492359ad1f66901a06cb8cd08370)
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/filedesc.h>
44 #include <sys/limits.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mutex.h>
48 #include <sys/mbuf.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/syscallsubr.h>
52 #include <sys/uio.h>
53 #include <sys/stat.h>
54 #include <sys/syslog.h>
55 #include <sys/un.h>
56 #include <sys/unistd.h>
57 
58 #include <security/audit/audit.h>
59 
60 #include <net/if.h>
61 #include <net/vnet.h>
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/ip.h>
65 #include <netinet/tcp.h>
66 #ifdef INET6
67 #include <netinet/ip6.h>
68 #include <netinet6/ip6_var.h>
69 #endif
70 
71 #ifdef COMPAT_LINUX32
72 #include <machine/../linux32/linux.h>
73 #include <machine/../linux32/linux32_proto.h>
74 #else
75 #include <machine/../linux/linux.h>
76 #include <machine/../linux/linux_proto.h>
77 #endif
78 #include <compat/linux/linux_common.h>
79 #include <compat/linux/linux_file.h>
80 #include <compat/linux/linux_mib.h>
81 #include <compat/linux/linux_socket.h>
82 #include <compat/linux/linux_timer.h>
83 #include <compat/linux/linux_util.h>
84 
85 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
86 					l_uint);
87 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
88 					l_uint, struct msghdr *);
89 static int linux_set_socket_flags(int, int *);
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 	case LINUX_SO_PROTOCOL:
240 		return (SO_PROTOCOL);
241 	}
242 	return (-1);
243 }
244 
245 static int
246 linux_to_bsd_tcp_sockopt(int opt)
247 {
248 
249 	switch (opt) {
250 	case LINUX_TCP_NODELAY:
251 		return (TCP_NODELAY);
252 	case LINUX_TCP_MAXSEG:
253 		return (TCP_MAXSEG);
254 	case LINUX_TCP_CORK:
255 		return (TCP_NOPUSH);
256 	case LINUX_TCP_KEEPIDLE:
257 		return (TCP_KEEPIDLE);
258 	case LINUX_TCP_KEEPINTVL:
259 		return (TCP_KEEPINTVL);
260 	case LINUX_TCP_KEEPCNT:
261 		return (TCP_KEEPCNT);
262 	case LINUX_TCP_MD5SIG:
263 		return (TCP_MD5SIG);
264 	}
265 	return (-1);
266 }
267 
268 static int
269 linux_to_bsd_msg_flags(int flags)
270 {
271 	int ret_flags = 0;
272 
273 	if (flags & LINUX_MSG_OOB)
274 		ret_flags |= MSG_OOB;
275 	if (flags & LINUX_MSG_PEEK)
276 		ret_flags |= MSG_PEEK;
277 	if (flags & LINUX_MSG_DONTROUTE)
278 		ret_flags |= MSG_DONTROUTE;
279 	if (flags & LINUX_MSG_CTRUNC)
280 		ret_flags |= MSG_CTRUNC;
281 	if (flags & LINUX_MSG_TRUNC)
282 		ret_flags |= MSG_TRUNC;
283 	if (flags & LINUX_MSG_DONTWAIT)
284 		ret_flags |= MSG_DONTWAIT;
285 	if (flags & LINUX_MSG_EOR)
286 		ret_flags |= MSG_EOR;
287 	if (flags & LINUX_MSG_WAITALL)
288 		ret_flags |= MSG_WAITALL;
289 	if (flags & LINUX_MSG_NOSIGNAL)
290 		ret_flags |= MSG_NOSIGNAL;
291 #if 0 /* not handled */
292 	if (flags & LINUX_MSG_PROXY)
293 		;
294 	if (flags & LINUX_MSG_FIN)
295 		;
296 	if (flags & LINUX_MSG_SYN)
297 		;
298 	if (flags & LINUX_MSG_CONFIRM)
299 		;
300 	if (flags & LINUX_MSG_RST)
301 		;
302 	if (flags & LINUX_MSG_ERRQUEUE)
303 		;
304 #endif
305 	return (ret_flags);
306 }
307 
308 static int
309 linux_to_bsd_cmsg_type(int cmsg_type)
310 {
311 
312 	switch (cmsg_type) {
313 	case LINUX_SCM_RIGHTS:
314 		return (SCM_RIGHTS);
315 	case LINUX_SCM_CREDENTIALS:
316 		return (SCM_CREDS);
317 	}
318 	return (-1);
319 }
320 
321 static int
322 bsd_to_linux_cmsg_type(int cmsg_type)
323 {
324 
325 	switch (cmsg_type) {
326 	case SCM_RIGHTS:
327 		return (LINUX_SCM_RIGHTS);
328 	case SCM_CREDS:
329 		return (LINUX_SCM_CREDENTIALS);
330 	case SCM_TIMESTAMP:
331 		return (LINUX_SCM_TIMESTAMP);
332 	}
333 	return (-1);
334 }
335 
336 static int
337 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
338 {
339 	if (lhdr->msg_controllen > INT_MAX)
340 		return (ENOBUFS);
341 
342 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
343 	bhdr->msg_namelen	= lhdr->msg_namelen;
344 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
345 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
346 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
347 
348 	/*
349 	 * msg_controllen is skipped since BSD and LINUX control messages
350 	 * are potentially different sizes (e.g. the cred structure used
351 	 * by SCM_CREDS is different between the two operating system).
352 	 *
353 	 * The caller can set it (if necessary) after converting all the
354 	 * control messages.
355 	 */
356 
357 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
358 	return (0);
359 }
360 
361 static int
362 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
363 {
364 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
365 	lhdr->msg_namelen	= bhdr->msg_namelen;
366 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
367 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
368 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
369 
370 	/*
371 	 * msg_controllen is skipped since BSD and LINUX control messages
372 	 * are potentially different sizes (e.g. the cred structure used
373 	 * by SCM_CREDS is different between the two operating system).
374 	 *
375 	 * The caller can set it (if necessary) after converting all the
376 	 * control messages.
377 	 */
378 
379 	/* msg_flags skipped */
380 	return (0);
381 }
382 
383 static int
384 linux_set_socket_flags(int lflags, int *flags)
385 {
386 
387 	if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
388 		return (EINVAL);
389 	if (lflags & LINUX_SOCK_NONBLOCK)
390 		*flags |= SOCK_NONBLOCK;
391 	if (lflags & LINUX_SOCK_CLOEXEC)
392 		*flags |= SOCK_CLOEXEC;
393 	return (0);
394 }
395 
396 static int
397 linux_copyout_sockaddr(const struct sockaddr *sa, void *uaddr, size_t len)
398 {
399 	struct l_sockaddr *lsa;
400 	int error;
401 
402 	error = bsd_to_linux_sockaddr(sa, &lsa, len);
403 	if (error != 0)
404 		return (error);
405 
406 	error = copyout(lsa, uaddr, len);
407 	free(lsa, M_SONAME);
408 
409 	return (error);
410 }
411 
412 static int
413 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
414     struct mbuf *control, enum uio_seg segflg)
415 {
416 	struct sockaddr *to;
417 	int error, len;
418 
419 	if (mp->msg_name != NULL) {
420 		len = mp->msg_namelen;
421 		error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len);
422 		if (error != 0)
423 			return (error);
424 		mp->msg_name = to;
425 	} else
426 		to = NULL;
427 
428 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
429 	    segflg);
430 
431 	if (to)
432 		free(to, M_SONAME);
433 	return (error);
434 }
435 
436 /* Return 0 if IP_HDRINCL is set for the given socket. */
437 static int
438 linux_check_hdrincl(struct thread *td, int s)
439 {
440 	int error, optval;
441 	socklen_t size_val;
442 
443 	size_val = sizeof(optval);
444 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
445 	    &optval, UIO_SYSSPACE, &size_val);
446 	if (error != 0)
447 		return (error);
448 
449 	return (optval == 0);
450 }
451 
452 /*
453  * Updated sendto() when IP_HDRINCL is set:
454  * tweak endian-dependent fields in the IP packet.
455  */
456 static int
457 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
458 {
459 /*
460  * linux_ip_copysize defines how many bytes we should copy
461  * from the beginning of the IP packet before we customize it for BSD.
462  * It should include all the fields we modify (ip_len and ip_off).
463  */
464 #define linux_ip_copysize	8
465 
466 	struct ip *packet;
467 	struct msghdr msg;
468 	struct iovec aiov[1];
469 	int error;
470 
471 	/* Check that the packet isn't too big or too small. */
472 	if (linux_args->len < linux_ip_copysize ||
473 	    linux_args->len > IP_MAXPACKET)
474 		return (EINVAL);
475 
476 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
477 
478 	/* Make kernel copy of the packet to be sent */
479 	if ((error = copyin(PTRIN(linux_args->msg), packet,
480 	    linux_args->len)))
481 		goto goout;
482 
483 	/* Convert fields from Linux to BSD raw IP socket format */
484 	packet->ip_len = linux_args->len;
485 	packet->ip_off = ntohs(packet->ip_off);
486 
487 	/* Prepare the msghdr and iovec structures describing the new packet */
488 	msg.msg_name = PTRIN(linux_args->to);
489 	msg.msg_namelen = linux_args->tolen;
490 	msg.msg_iov = aiov;
491 	msg.msg_iovlen = 1;
492 	msg.msg_control = NULL;
493 	msg.msg_flags = 0;
494 	aiov[0].iov_base = (char *)packet;
495 	aiov[0].iov_len = linux_args->len;
496 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
497 	    NULL, UIO_SYSSPACE);
498 goout:
499 	free(packet, M_LINUX);
500 	return (error);
501 }
502 
503 int
504 linux_socket(struct thread *td, struct linux_socket_args *args)
505 {
506 	int domain, retval_socket, type;
507 
508 	type = args->type & LINUX_SOCK_TYPE_MASK;
509 	if (type < 0 || type > LINUX_SOCK_MAX)
510 		return (EINVAL);
511 	retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
512 		&type);
513 	if (retval_socket != 0)
514 		return (retval_socket);
515 	domain = linux_to_bsd_domain(args->domain);
516 	if (domain == -1) {
517 		linux_msg(curthread, "unsupported socket domain %d, type %d, protocol %d",
518 		    args->domain, args->type & LINUX_SOCK_TYPE_MASK, args->protocol);
519 		return (EAFNOSUPPORT);
520 	}
521 
522 	retval_socket = kern_socket(td, domain, type, args->protocol);
523 	if (retval_socket)
524 		return (retval_socket);
525 
526 	if (type == SOCK_RAW
527 	    && (args->protocol == IPPROTO_RAW || args->protocol == 0)
528 	    && domain == PF_INET) {
529 		/* It's a raw IP socket: set the IP_HDRINCL option. */
530 		int hdrincl;
531 
532 		hdrincl = 1;
533 		/* We ignore any error returned by kern_setsockopt() */
534 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
535 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
536 	}
537 #ifdef INET6
538 	/*
539 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
540 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
541 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
542 	 * sysctl value.
543 	 */
544 	if (domain == PF_INET6) {
545 		int v6only;
546 
547 		v6only = 0;
548 		/* We ignore any error returned by setsockopt() */
549 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
550 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
551 	}
552 #endif
553 
554 	return (retval_socket);
555 }
556 
557 int
558 linux_bind(struct thread *td, struct linux_bind_args *args)
559 {
560 	struct sockaddr *sa;
561 	int error;
562 
563 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
564 	    &args->namelen);
565 	if (error != 0)
566 		return (error);
567 
568 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
569 	free(sa, M_SONAME);
570 
571 	/* XXX */
572 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
573 		return (EINVAL);
574 	return (error);
575 }
576 
577 int
578 linux_connect(struct thread *td, struct linux_connect_args *args)
579 {
580 	struct socket *so;
581 	struct sockaddr *sa;
582 	struct file *fp;
583 	u_int fflag;
584 	int error;
585 
586 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
587 	    &args->namelen);
588 	if (error != 0)
589 		return (error);
590 
591 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
592 	free(sa, M_SONAME);
593 	if (error != EISCONN)
594 		return (error);
595 
596 	/*
597 	 * Linux doesn't return EISCONN the first time it occurs,
598 	 * when on a non-blocking socket. Instead it returns the
599 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
600 	 */
601 	error = getsock_cap(td, args->s, &cap_connect_rights,
602 	    &fp, &fflag, NULL);
603 	if (error != 0)
604 		return (error);
605 
606 	error = EISCONN;
607 	so = fp->f_data;
608 	if (fflag & FNONBLOCK) {
609 		SOCK_LOCK(so);
610 		if (so->so_emuldata == 0)
611 			error = so->so_error;
612 		so->so_emuldata = (void *)1;
613 		SOCK_UNLOCK(so);
614 	}
615 	fdrop(fp, td);
616 
617 	return (error);
618 }
619 
620 int
621 linux_listen(struct thread *td, struct linux_listen_args *args)
622 {
623 
624 	return (kern_listen(td, args->s, args->backlog));
625 }
626 
627 static int
628 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
629     l_uintptr_t namelen, int flags)
630 {
631 	struct sockaddr *sa;
632 	struct file *fp, *fp1;
633 	int bflags, len;
634 	struct socket *so;
635 	int error, error1;
636 
637 	bflags = 0;
638 	fp = NULL;
639 	sa = NULL;
640 
641 	error = linux_set_socket_flags(flags, &bflags);
642 	if (error != 0)
643 		return (error);
644 
645 	if (PTRIN(addr) == NULL) {
646 		len = 0;
647 		error = kern_accept4(td, s, NULL, NULL, bflags, NULL);
648 	} else {
649 		error = copyin(PTRIN(namelen), &len, sizeof(len));
650 		if (error != 0)
651 			return (error);
652 		if (len < 0)
653 			return (EINVAL);
654 		error = kern_accept4(td, s, &sa, &len, bflags, &fp);
655 	}
656 
657 	/*
658 	 * Translate errno values into ones used by Linux.
659 	 */
660 	if (error != 0) {
661 		/*
662 		 * XXX. This is wrong, different sockaddr structures
663 		 * have different sizes.
664 		 */
665 		switch (error) {
666 		case EFAULT:
667 			if (namelen != sizeof(struct sockaddr_in))
668 				error = EINVAL;
669 			break;
670 		case EINVAL:
671 			error1 = getsock_cap(td, s, &cap_accept_rights, &fp1, NULL, NULL);
672 			if (error1 != 0) {
673 				error = error1;
674 				break;
675 			}
676 			so = fp1->f_data;
677 			if (so->so_type == SOCK_DGRAM)
678 				error = EOPNOTSUPP;
679 			fdrop(fp1, td);
680 			break;
681 		}
682 		return (error);
683 	}
684 
685 	if (len != 0) {
686 		error = linux_copyout_sockaddr(sa, PTRIN(addr), len);
687 
688 		/*
689 		 * XXX: We should also copyout the len, shouldn't we?
690 		 */
691 
692 		if (error != 0) {
693 			fdclose(td, fp, td->td_retval[0]);
694 			td->td_retval[0] = 0;
695 		}
696 	}
697 	if (fp != NULL)
698 		fdrop(fp, td);
699 	free(sa, M_SONAME);
700 	return (error);
701 }
702 
703 int
704 linux_accept(struct thread *td, struct linux_accept_args *args)
705 {
706 
707 	return (linux_accept_common(td, args->s, args->addr,
708 	    args->namelen, 0));
709 }
710 
711 int
712 linux_accept4(struct thread *td, struct linux_accept4_args *args)
713 {
714 
715 	return (linux_accept_common(td, args->s, args->addr,
716 	    args->namelen, args->flags));
717 }
718 
719 int
720 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
721 {
722 	struct sockaddr *sa;
723 	int len, error;
724 
725 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
726 	if (error != 0)
727 		return (error);
728 
729 	error = kern_getsockname(td, args->s, &sa, &len);
730 	if (error != 0)
731 		return (error);
732 
733 	if (len != 0)
734 		error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
735 
736 	free(sa, M_SONAME);
737 	if (error == 0)
738 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
739 	return (error);
740 }
741 
742 int
743 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
744 {
745 	struct sockaddr *sa;
746 	int len, error;
747 
748 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
749 	if (error != 0)
750 		return (error);
751 	if (len < 0)
752 		return (EINVAL);
753 
754 	error = kern_getpeername(td, args->s, &sa, &len);
755 	if (error != 0)
756 		return (error);
757 
758 	if (len != 0)
759 		error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
760 
761 	free(sa, M_SONAME);
762 	if (error == 0)
763 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
764 	return (error);
765 }
766 
767 int
768 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
769 {
770 	int domain, error, sv[2], type;
771 
772 	domain = linux_to_bsd_domain(args->domain);
773 	if (domain != PF_LOCAL)
774 		return (EAFNOSUPPORT);
775 	type = args->type & LINUX_SOCK_TYPE_MASK;
776 	if (type < 0 || type > LINUX_SOCK_MAX)
777 		return (EINVAL);
778 	error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
779 	    &type);
780 	if (error != 0)
781 		return (error);
782 	if (args->protocol != 0 && args->protocol != PF_UNIX) {
783 		/*
784 		 * Use of PF_UNIX as protocol argument is not right,
785 		 * but Linux does it.
786 		 * Do not map PF_UNIX as its Linux value is identical
787 		 * to FreeBSD one.
788 		 */
789 		return (EPROTONOSUPPORT);
790 	}
791 	error = kern_socketpair(td, domain, type, 0, sv);
792 	if (error != 0)
793                 return (error);
794         error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int));
795         if (error != 0) {
796                 (void)kern_close(td, sv[0]);
797                 (void)kern_close(td, sv[1]);
798         }
799 	return (error);
800 }
801 
802 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
803 struct linux_send_args {
804 	register_t s;
805 	register_t msg;
806 	register_t len;
807 	register_t flags;
808 };
809 
810 static int
811 linux_send(struct thread *td, struct linux_send_args *args)
812 {
813 	struct sendto_args /* {
814 		int s;
815 		caddr_t buf;
816 		int len;
817 		int flags;
818 		caddr_t to;
819 		int tolen;
820 	} */ bsd_args;
821 	struct file *fp;
822 	int error, fflag;
823 
824 	bsd_args.s = args->s;
825 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
826 	bsd_args.len = args->len;
827 	bsd_args.flags = args->flags;
828 	bsd_args.to = NULL;
829 	bsd_args.tolen = 0;
830 	error = sys_sendto(td, &bsd_args);
831 	if (error == ENOTCONN) {
832 		/*
833 		 * Linux doesn't return ENOTCONN for non-blocking sockets.
834 		 * Instead it returns the EAGAIN.
835 		 */
836 		error = getsock_cap(td, args->s, &cap_send_rights, &fp,
837 		    &fflag, NULL);
838 		if (error == 0) {
839 			if (fflag & FNONBLOCK)
840 				error = EAGAIN;
841 			fdrop(fp, td);
842 		}
843 	}
844 	return (error);
845 }
846 
847 struct linux_recv_args {
848 	register_t s;
849 	register_t msg;
850 	register_t len;
851 	register_t flags;
852 };
853 
854 static int
855 linux_recv(struct thread *td, struct linux_recv_args *args)
856 {
857 	struct recvfrom_args /* {
858 		int s;
859 		caddr_t buf;
860 		int len;
861 		int flags;
862 		struct sockaddr *from;
863 		socklen_t fromlenaddr;
864 	} */ bsd_args;
865 
866 	bsd_args.s = args->s;
867 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
868 	bsd_args.len = args->len;
869 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
870 	bsd_args.from = NULL;
871 	bsd_args.fromlenaddr = 0;
872 	return (sys_recvfrom(td, &bsd_args));
873 }
874 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
875 
876 int
877 linux_sendto(struct thread *td, struct linux_sendto_args *args)
878 {
879 	struct msghdr msg;
880 	struct iovec aiov;
881 
882 	if (linux_check_hdrincl(td, args->s) == 0)
883 		/* IP_HDRINCL set, tweak the packet before sending */
884 		return (linux_sendto_hdrincl(td, args));
885 
886 	msg.msg_name = PTRIN(args->to);
887 	msg.msg_namelen = args->tolen;
888 	msg.msg_iov = &aiov;
889 	msg.msg_iovlen = 1;
890 	msg.msg_control = NULL;
891 	msg.msg_flags = 0;
892 	aiov.iov_base = PTRIN(args->msg);
893 	aiov.iov_len = args->len;
894 	return (linux_sendit(td, args->s, &msg, args->flags, NULL,
895 	    UIO_USERSPACE));
896 }
897 
898 int
899 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
900 {
901 	struct sockaddr *sa;
902 	struct msghdr msg;
903 	struct iovec aiov;
904 	int error, fromlen;
905 
906 	if (PTRIN(args->fromlen) != NULL) {
907 		error = copyin(PTRIN(args->fromlen), &fromlen,
908 		    sizeof(fromlen));
909 		if (error != 0)
910 			return (error);
911 		if (fromlen < 0)
912 			return (EINVAL);
913 		sa = malloc(fromlen, M_SONAME, M_WAITOK);
914 	} else {
915 		fromlen = 0;
916 		sa = NULL;
917 	}
918 
919 	msg.msg_name = sa;
920 	msg.msg_namelen = fromlen;
921 	msg.msg_iov = &aiov;
922 	msg.msg_iovlen = 1;
923 	aiov.iov_base = PTRIN(args->buf);
924 	aiov.iov_len = args->len;
925 	msg.msg_control = 0;
926 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
927 
928 	error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL);
929 	if (error != 0)
930 		goto out;
931 
932 	if (PTRIN(args->from) != NULL)
933 		error = linux_copyout_sockaddr(sa, PTRIN(args->from), msg.msg_namelen);
934 
935 	if (error == 0 && PTRIN(args->fromlen) != NULL)
936 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
937 		    sizeof(msg.msg_namelen));
938 out:
939 	free(sa, M_SONAME);
940 	return (error);
941 }
942 
943 static int
944 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
945     l_uint flags)
946 {
947 	struct cmsghdr *cmsg;
948 	struct mbuf *control;
949 	struct msghdr msg;
950 	struct l_cmsghdr linux_cmsg;
951 	struct l_cmsghdr *ptr_cmsg;
952 	struct l_msghdr linux_msghdr;
953 	struct iovec *iov;
954 	socklen_t datalen;
955 	struct sockaddr *sa;
956 	struct socket *so;
957 	sa_family_t sa_family;
958 	struct file *fp;
959 	void *data;
960 	l_size_t len;
961 	l_size_t clen;
962 	int error, fflag;
963 
964 	error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
965 	if (error != 0)
966 		return (error);
967 
968 	/*
969 	 * Some Linux applications (ping) define a non-NULL control data
970 	 * pointer, but a msg_controllen of 0, which is not allowed in the
971 	 * FreeBSD system call interface.  NULL the msg_control pointer in
972 	 * order to handle this case.  This should be checked, but allows the
973 	 * Linux ping to work.
974 	 */
975 	if (PTRIN(linux_msghdr.msg_control) != NULL &&
976 	    linux_msghdr.msg_controllen == 0)
977 		linux_msghdr.msg_control = PTROUT(NULL);
978 
979 	error = linux_to_bsd_msghdr(&msg, &linux_msghdr);
980 	if (error != 0)
981 		return (error);
982 
983 #ifdef COMPAT_LINUX32
984 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
985 	    &iov, EMSGSIZE);
986 #else
987 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
988 #endif
989 	if (error != 0)
990 		return (error);
991 
992 	control = NULL;
993 
994 	error = kern_getsockname(td, s, &sa, &datalen);
995 	if (error != 0)
996 		goto bad;
997 	sa_family = sa->sa_family;
998 	free(sa, M_SONAME);
999 
1000 	if (flags & LINUX_MSG_OOB) {
1001 		error = EOPNOTSUPP;
1002 		if (sa_family == AF_UNIX)
1003 			goto bad;
1004 
1005 		error = getsock_cap(td, s, &cap_send_rights, &fp,
1006 		    &fflag, NULL);
1007 		if (error != 0)
1008 			goto bad;
1009 		so = fp->f_data;
1010 		if (so->so_type != SOCK_STREAM)
1011 			error = EOPNOTSUPP;
1012 		fdrop(fp, td);
1013 		if (error != 0)
1014 			goto bad;
1015 	}
1016 
1017 	if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) {
1018 		error = ENOBUFS;
1019 		control = m_get(M_WAITOK, MT_CONTROL);
1020 		MCLGET(control, M_WAITOK);
1021 		data = mtod(control, void *);
1022 		datalen = 0;
1023 
1024 		ptr_cmsg = PTRIN(linux_msghdr.msg_control);
1025 		clen = linux_msghdr.msg_controllen;
1026 		do {
1027 			error = copyin(ptr_cmsg, &linux_cmsg,
1028 			    sizeof(struct l_cmsghdr));
1029 			if (error != 0)
1030 				goto bad;
1031 
1032 			error = EINVAL;
1033 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1034 			    linux_cmsg.cmsg_len > clen)
1035 				goto bad;
1036 
1037 			if (datalen + CMSG_HDRSZ > MCLBYTES)
1038 				goto bad;
1039 
1040 			/*
1041 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1042 			 * so return EINVAL in any other cmsg_type
1043 			 */
1044 			cmsg = data;
1045 			cmsg->cmsg_type =
1046 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1047 			cmsg->cmsg_level =
1048 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1049 			if (cmsg->cmsg_type == -1
1050 			    || cmsg->cmsg_level != SOL_SOCKET) {
1051 				linux_msg(curthread,
1052 				    "unsupported sendmsg cmsg level %d type %d",
1053 				    linux_cmsg.cmsg_level, linux_cmsg.cmsg_type);
1054 				goto bad;
1055 			}
1056 
1057 			/*
1058 			 * Some applications (e.g. pulseaudio) attempt to
1059 			 * send ancillary data even if the underlying protocol
1060 			 * doesn't support it which is not allowed in the
1061 			 * FreeBSD system call interface.
1062 			 */
1063 			if (sa_family != AF_UNIX)
1064 				goto next;
1065 
1066 			if (cmsg->cmsg_type == SCM_CREDS) {
1067 				len = sizeof(struct cmsgcred);
1068 				if (datalen + CMSG_SPACE(len) > MCLBYTES)
1069 					goto bad;
1070 
1071 				/*
1072 				 * The lower levels will fill in the structure
1073 				 */
1074 				memset(CMSG_DATA(data), 0, len);
1075 			} else {
1076 				len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1077 				if (datalen + CMSG_SPACE(len) < datalen ||
1078 				    datalen + CMSG_SPACE(len) > MCLBYTES)
1079 					goto bad;
1080 
1081 				error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1082 				    CMSG_DATA(data), len);
1083 				if (error != 0)
1084 					goto bad;
1085 			}
1086 
1087 			cmsg->cmsg_len = CMSG_LEN(len);
1088 			data = (char *)data + CMSG_SPACE(len);
1089 			datalen += CMSG_SPACE(len);
1090 
1091 next:
1092 			if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1093 				break;
1094 
1095 			clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1096 			ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1097 			    LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1098 		} while(clen >= sizeof(struct l_cmsghdr));
1099 
1100 		control->m_len = datalen;
1101 		if (datalen == 0) {
1102 			m_freem(control);
1103 			control = NULL;
1104 		}
1105 	}
1106 
1107 	msg.msg_iov = iov;
1108 	msg.msg_flags = 0;
1109 	error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1110 	control = NULL;
1111 
1112 bad:
1113 	m_freem(control);
1114 	free(iov, M_IOV);
1115 	return (error);
1116 }
1117 
1118 int
1119 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1120 {
1121 
1122 	return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1123 	    args->flags));
1124 }
1125 
1126 int
1127 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1128 {
1129 	struct l_mmsghdr *msg;
1130 	l_uint retval;
1131 	int error, datagrams;
1132 
1133 	if (args->vlen > UIO_MAXIOV)
1134 		args->vlen = UIO_MAXIOV;
1135 
1136 	msg = PTRIN(args->msg);
1137 	datagrams = 0;
1138 	while (datagrams < args->vlen) {
1139 		error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1140 		    args->flags);
1141 		if (error != 0)
1142 			break;
1143 
1144 		retval = td->td_retval[0];
1145 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1146 		if (error != 0)
1147 			break;
1148 		++msg;
1149 		++datagrams;
1150 	}
1151 	if (error == 0)
1152 		td->td_retval[0] = datagrams;
1153 	return (error);
1154 }
1155 
1156 static int
1157 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1158     l_uint flags, struct msghdr *msg)
1159 {
1160 	struct cmsghdr *cm;
1161 	struct cmsgcred *cmcred;
1162 	struct l_cmsghdr *linux_cmsg = NULL;
1163 	struct l_ucred linux_ucred;
1164 	socklen_t datalen, maxlen, outlen;
1165 	struct l_msghdr linux_msghdr;
1166 	struct iovec *iov, *uiov;
1167 	struct mbuf *control = NULL;
1168 	struct mbuf **controlp;
1169 	struct timeval *ftmvl;
1170 	struct sockaddr *sa;
1171 	l_timeval ltmvl;
1172 	caddr_t outbuf;
1173 	void *data;
1174 	int error, i, fd, fds, *fdp;
1175 
1176 	error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
1177 	if (error != 0)
1178 		return (error);
1179 
1180 	error = linux_to_bsd_msghdr(msg, &linux_msghdr);
1181 	if (error != 0)
1182 		return (error);
1183 
1184 #ifdef COMPAT_LINUX32
1185 	error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1186 	    &iov, EMSGSIZE);
1187 #else
1188 	error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1189 #endif
1190 	if (error != 0)
1191 		return (error);
1192 
1193 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1194 		msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN);
1195 		sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK);
1196 		msg->msg_name = sa;
1197 	} else {
1198 		sa = NULL;
1199 		msg->msg_name = NULL;
1200 	}
1201 
1202 	uiov = msg->msg_iov;
1203 	msg->msg_iov = iov;
1204 	controlp = (msg->msg_control != NULL) ? &control : NULL;
1205 	error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp);
1206 	msg->msg_iov = uiov;
1207 	if (error != 0)
1208 		goto bad;
1209 
1210 	/*
1211 	 * Note that kern_recvit() updates msg->msg_namelen.
1212 	 */
1213 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1214 		msg->msg_name = PTRIN(linux_msghdr.msg_name);
1215 		error = linux_copyout_sockaddr(sa,
1216 		    PTRIN(msg->msg_name), msg->msg_namelen);
1217 		if (error != 0)
1218 			goto bad;
1219 	}
1220 
1221 	error = bsd_to_linux_msghdr(msg, &linux_msghdr);
1222 	if (error != 0)
1223 		goto bad;
1224 
1225 	maxlen = linux_msghdr.msg_controllen;
1226 	linux_msghdr.msg_controllen = 0;
1227 	if (control) {
1228 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1229 
1230 		msg->msg_control = mtod(control, struct cmsghdr *);
1231 		msg->msg_controllen = control->m_len;
1232 
1233 		cm = CMSG_FIRSTHDR(msg);
1234 		outbuf = PTRIN(linux_msghdr.msg_control);
1235 		outlen = 0;
1236 		while (cm != NULL) {
1237 			linux_cmsg->cmsg_type =
1238 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1239 			linux_cmsg->cmsg_level =
1240 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1241 			if (linux_cmsg->cmsg_type == -1 ||
1242 			    cm->cmsg_level != SOL_SOCKET) {
1243 				linux_msg(curthread,
1244 				    "unsupported recvmsg cmsg level %d type %d",
1245 				    cm->cmsg_level, cm->cmsg_type);
1246 				error = EINVAL;
1247 				goto bad;
1248 			}
1249 
1250 			data = CMSG_DATA(cm);
1251 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1252 
1253 			switch (cm->cmsg_type) {
1254 			case SCM_RIGHTS:
1255 				if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1256 					fds = datalen / sizeof(int);
1257 					fdp = data;
1258 					for (i = 0; i < fds; i++) {
1259 						fd = *fdp++;
1260 						(void)kern_fcntl(td, fd,
1261 						    F_SETFD, FD_CLOEXEC);
1262 					}
1263 				}
1264 				break;
1265 
1266 			case SCM_CREDS:
1267 				/*
1268 				 * Currently LOCAL_CREDS is never in
1269 				 * effect for Linux so no need to worry
1270 				 * about sockcred
1271 				 */
1272 				if (datalen != sizeof(*cmcred)) {
1273 					error = EMSGSIZE;
1274 					goto bad;
1275 				}
1276 				cmcred = (struct cmsgcred *)data;
1277 				bzero(&linux_ucred, sizeof(linux_ucred));
1278 				linux_ucred.pid = cmcred->cmcred_pid;
1279 				linux_ucred.uid = cmcred->cmcred_uid;
1280 				linux_ucred.gid = cmcred->cmcred_gid;
1281 				data = &linux_ucred;
1282 				datalen = sizeof(linux_ucred);
1283 				break;
1284 
1285 			case SCM_TIMESTAMP:
1286 				if (datalen != sizeof(struct timeval)) {
1287 					error = EMSGSIZE;
1288 					goto bad;
1289 				}
1290 				ftmvl = (struct timeval *)data;
1291 				ltmvl.tv_sec = ftmvl->tv_sec;
1292 				ltmvl.tv_usec = ftmvl->tv_usec;
1293 				data = &ltmvl;
1294 				datalen = sizeof(ltmvl);
1295 				break;
1296 			}
1297 
1298 			if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1299 				if (outlen == 0) {
1300 					error = EMSGSIZE;
1301 					goto bad;
1302 				} else {
1303 					linux_msghdr.msg_flags |= LINUX_MSG_CTRUNC;
1304 					m_dispose_extcontrolm(control);
1305 					goto out;
1306 				}
1307 			}
1308 
1309 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1310 
1311 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1312 			if (error != 0)
1313 				goto bad;
1314 			outbuf += L_CMSG_HDRSZ;
1315 
1316 			error = copyout(data, outbuf, datalen);
1317 			if (error != 0)
1318 				goto bad;
1319 
1320 			outbuf += LINUX_CMSG_ALIGN(datalen);
1321 			outlen += LINUX_CMSG_LEN(datalen);
1322 
1323 			cm = CMSG_NXTHDR(msg, cm);
1324 		}
1325 		linux_msghdr.msg_controllen = outlen;
1326 	}
1327 
1328 out:
1329 	error = copyout(&linux_msghdr, msghdr, sizeof(linux_msghdr));
1330 
1331 bad:
1332 	if (control != NULL) {
1333 		if (error != 0)
1334 			m_dispose_extcontrolm(control);
1335 		m_freem(control);
1336 	}
1337 	free(iov, M_IOV);
1338 	free(linux_cmsg, M_LINUX);
1339 	free(sa, M_SONAME);
1340 
1341 	return (error);
1342 }
1343 
1344 int
1345 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1346 {
1347 	struct msghdr bsd_msg;
1348 
1349 	return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1350 	    args->flags, &bsd_msg));
1351 }
1352 
1353 int
1354 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1355 {
1356 	struct l_mmsghdr *msg;
1357 	struct msghdr bsd_msg;
1358 	struct l_timespec lts;
1359 	struct timespec ts, tts;
1360 	l_uint retval;
1361 	int error, datagrams;
1362 
1363 	if (args->timeout) {
1364 		error = copyin(args->timeout, &lts, sizeof(struct l_timespec));
1365 		if (error != 0)
1366 			return (error);
1367 		error = linux_to_native_timespec(&ts, &lts);
1368 		if (error != 0)
1369 			return (error);
1370 		getnanotime(&tts);
1371 		timespecadd(&tts, &ts, &tts);
1372 	}
1373 
1374 	msg = PTRIN(args->msg);
1375 	datagrams = 0;
1376 	while (datagrams < args->vlen) {
1377 		error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1378 		    args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1379 		if (error != 0)
1380 			break;
1381 
1382 		retval = td->td_retval[0];
1383 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1384 		if (error != 0)
1385 			break;
1386 		++msg;
1387 		++datagrams;
1388 
1389 		/*
1390 		 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1391 		 */
1392 		if (args->flags & LINUX_MSG_WAITFORONE)
1393 			args->flags |= LINUX_MSG_DONTWAIT;
1394 
1395 		/*
1396 		 * See BUGS section of recvmmsg(2).
1397 		 */
1398 		if (args->timeout) {
1399 			getnanotime(&ts);
1400 			timespecsub(&ts, &tts, &ts);
1401 			if (!timespecisset(&ts) || ts.tv_sec > 0)
1402 				break;
1403 		}
1404 		/* Out of band data, return right away. */
1405 		if (bsd_msg.msg_flags & MSG_OOB)
1406 			break;
1407 	}
1408 	if (error == 0)
1409 		td->td_retval[0] = datagrams;
1410 	return (error);
1411 }
1412 
1413 int
1414 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1415 {
1416 
1417 	return (kern_shutdown(td, args->s, args->how));
1418 }
1419 
1420 int
1421 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1422 {
1423 	l_timeval linux_tv;
1424 	struct sockaddr *sa;
1425 	struct timeval tv;
1426 	socklen_t len;
1427 	int error, level, name;
1428 
1429 	level = linux_to_bsd_sockopt_level(args->level);
1430 	switch (level) {
1431 	case SOL_SOCKET:
1432 		name = linux_to_bsd_so_sockopt(args->optname);
1433 		switch (name) {
1434 		case SO_RCVTIMEO:
1435 			/* FALLTHROUGH */
1436 		case SO_SNDTIMEO:
1437 			error = copyin(PTRIN(args->optval), &linux_tv,
1438 			    sizeof(linux_tv));
1439 			if (error != 0)
1440 				return (error);
1441 			tv.tv_sec = linux_tv.tv_sec;
1442 			tv.tv_usec = linux_tv.tv_usec;
1443 			return (kern_setsockopt(td, args->s, level,
1444 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1445 			/* NOTREACHED */
1446 		default:
1447 			break;
1448 		}
1449 		break;
1450 	case IPPROTO_IP:
1451 		if (args->optname == LINUX_IP_RECVERR &&
1452 		    linux_ignore_ip_recverr) {
1453 			/*
1454 			 * XXX: This is a hack to unbreak DNS resolution
1455 			 *	with glibc 2.30 and above.
1456 			 */
1457 			return (0);
1458 		}
1459 		name = linux_to_bsd_ip_sockopt(args->optname);
1460 		break;
1461 	case IPPROTO_IPV6:
1462 		name = linux_to_bsd_ip6_sockopt(args->optname);
1463 		break;
1464 	case IPPROTO_TCP:
1465 		name = linux_to_bsd_tcp_sockopt(args->optname);
1466 		break;
1467 	default:
1468 		name = -1;
1469 		break;
1470 	}
1471 	if (name == -1) {
1472 		linux_msg(curthread,
1473 		    "unsupported setsockopt level %d optname %d",
1474 		    args->level, args->optname);
1475 		return (ENOPROTOOPT);
1476 	}
1477 
1478 	if (name == IPV6_NEXTHOP) {
1479 		len = args->optlen;
1480 		error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len);
1481 		if (error != 0)
1482 			return (error);
1483 
1484 		error = kern_setsockopt(td, args->s, level,
1485 		    name, sa, UIO_SYSSPACE, len);
1486 		free(sa, M_SONAME);
1487 	} else {
1488 		error = kern_setsockopt(td, args->s, level,
1489 		    name, PTRIN(args->optval), UIO_USERSPACE, args->optlen);
1490 	}
1491 
1492 	return (error);
1493 }
1494 
1495 int
1496 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1497 {
1498 	l_timeval linux_tv;
1499 	struct timeval tv;
1500 	socklen_t tv_len, xulen, len;
1501 	struct sockaddr *sa;
1502 	struct xucred xu;
1503 	struct l_ucred lxu;
1504 	int error, level, name, newval;
1505 
1506 	level = linux_to_bsd_sockopt_level(args->level);
1507 	switch (level) {
1508 	case SOL_SOCKET:
1509 		name = linux_to_bsd_so_sockopt(args->optname);
1510 		switch (name) {
1511 		case SO_RCVTIMEO:
1512 			/* FALLTHROUGH */
1513 		case SO_SNDTIMEO:
1514 			tv_len = sizeof(tv);
1515 			error = kern_getsockopt(td, args->s, level,
1516 			    name, &tv, UIO_SYSSPACE, &tv_len);
1517 			if (error != 0)
1518 				return (error);
1519 			linux_tv.tv_sec = tv.tv_sec;
1520 			linux_tv.tv_usec = tv.tv_usec;
1521 			return (copyout(&linux_tv, PTRIN(args->optval),
1522 			    sizeof(linux_tv)));
1523 			/* NOTREACHED */
1524 		case LOCAL_PEERCRED:
1525 			if (args->optlen < sizeof(lxu))
1526 				return (EINVAL);
1527 			/*
1528 			 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
1529 			 * but by the Unix socket's level 0.
1530 			 */
1531 			level = 0;
1532 			xulen = sizeof(xu);
1533 			error = kern_getsockopt(td, args->s, level,
1534 			    name, &xu, UIO_SYSSPACE, &xulen);
1535 			if (error != 0)
1536 				return (error);
1537 			lxu.pid = xu.cr_pid;
1538 			lxu.uid = xu.cr_uid;
1539 			lxu.gid = xu.cr_gid;
1540 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1541 			/* NOTREACHED */
1542 		case SO_ERROR:
1543 			len = sizeof(newval);
1544 			error = kern_getsockopt(td, args->s, level,
1545 			    name, &newval, UIO_SYSSPACE, &len);
1546 			if (error != 0)
1547 				return (error);
1548 			newval = -linux_to_bsd_errno(newval);
1549 			return (copyout(&newval, PTRIN(args->optval), len));
1550 			/* NOTREACHED */
1551 		default:
1552 			break;
1553 		}
1554 		break;
1555 	case IPPROTO_IP:
1556 		name = linux_to_bsd_ip_sockopt(args->optname);
1557 		break;
1558 	case IPPROTO_IPV6:
1559 		name = linux_to_bsd_ip6_sockopt(args->optname);
1560 		break;
1561 	case IPPROTO_TCP:
1562 		name = linux_to_bsd_tcp_sockopt(args->optname);
1563 		break;
1564 	default:
1565 		name = -1;
1566 		break;
1567 	}
1568 	if (name == -1) {
1569 		linux_msg(curthread,
1570 		    "unsupported getsockopt level %d optname %d",
1571 		    args->level, args->optname);
1572 		return (EINVAL);
1573 	}
1574 
1575 	if (name == IPV6_NEXTHOP) {
1576 		error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1577                 if (error != 0)
1578                         return (error);
1579 		sa = malloc(len, M_SONAME, M_WAITOK);
1580 
1581 		error = kern_getsockopt(td, args->s, level,
1582 		    name, sa, UIO_SYSSPACE, &len);
1583 		if (error != 0)
1584 			goto out;
1585 
1586 		error = linux_copyout_sockaddr(sa, PTRIN(args->optval), len);
1587 		if (error == 0)
1588 			error = copyout(&len, PTRIN(args->optlen),
1589 			    sizeof(len));
1590 out:
1591 		free(sa, M_SONAME);
1592 	} else {
1593 		if (args->optval) {
1594 			error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1595 			if (error != 0)
1596 				return (error);
1597 		}
1598 		error = kern_getsockopt(td, args->s, level,
1599 		    name, PTRIN(args->optval), UIO_USERSPACE, &len);
1600 		if (error == 0)
1601 			error = copyout(&len, PTRIN(args->optlen),
1602 			    sizeof(len));
1603 	}
1604 
1605 	return (error);
1606 }
1607 
1608 static int
1609 linux_sendfile_common(struct thread *td, l_int out, l_int in,
1610     l_loff_t *offset, l_size_t count)
1611 {
1612 	off_t bytes_read;
1613 	int error;
1614 	l_loff_t current_offset;
1615 	struct file *fp;
1616 
1617 	AUDIT_ARG_FD(in);
1618 	error = fget_read(td, in, &cap_pread_rights, &fp);
1619 	if (error != 0)
1620 		return (error);
1621 
1622 	if (offset != NULL) {
1623 		current_offset = *offset;
1624 	} else {
1625 		error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1626 		    fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE;
1627 		if (error != 0)
1628 			goto drop;
1629 		current_offset = td->td_uretoff.tdu_off;
1630 	}
1631 
1632 	bytes_read = 0;
1633 
1634 	/* Linux cannot have 0 count. */
1635 	if (count <= 0 || current_offset < 0) {
1636 		error = EINVAL;
1637 		goto drop;
1638 	}
1639 
1640 	error = fo_sendfile(fp, out, NULL, NULL, current_offset, count,
1641 	    &bytes_read, 0, td);
1642 	if (error != 0)
1643 		goto drop;
1644 	current_offset += bytes_read;
1645 
1646 	if (offset != NULL) {
1647 		*offset = current_offset;
1648 	} else {
1649 		error = fo_seek(fp, current_offset, SEEK_SET, td);
1650 		if (error != 0)
1651 			goto drop;
1652 	}
1653 
1654 	td->td_retval[0] = (ssize_t)bytes_read;
1655 drop:
1656 	fdrop(fp, td);
1657 	return (error);
1658 }
1659 
1660 int
1661 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg)
1662 {
1663 	/*
1664 	 * Differences between FreeBSD and Linux sendfile:
1665 	 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to
1666 	 *   mean send the whole file.)  In linux_sendfile given fds are still
1667 	 *   checked for validity when the count is 0.
1668 	 * - Linux can send to any fd whereas FreeBSD only supports sockets.
1669 	 *   The same restriction follows for linux_sendfile.
1670 	 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr.
1671 	 * - Linux takes an offset pointer and updates it to the read location.
1672 	 *   FreeBSD takes in an offset and a 'bytes read' parameter which is
1673 	 *   only filled if it isn't NULL.  We use this parameter to update the
1674 	 *   offset pointer if it exists.
1675 	 * - Linux sendfile returns bytes read on success while FreeBSD
1676 	 *   returns 0.  We use the 'bytes read' parameter to get this value.
1677 	 */
1678 
1679 	l_loff_t offset64;
1680 	l_long offset;
1681 	int ret;
1682 	int error;
1683 
1684 	if (arg->offset != NULL) {
1685 		error = copyin(arg->offset, &offset, sizeof(offset));
1686 		if (error != 0)
1687 			return (error);
1688 		offset64 = (l_loff_t)offset;
1689 	}
1690 
1691 	ret = linux_sendfile_common(td, arg->out, arg->in,
1692 	    arg->offset != NULL ? &offset64 : NULL, arg->count);
1693 
1694 	if (arg->offset != NULL) {
1695 #if defined(__i386__) || defined(__arm__) || \
1696     (defined(__amd64__) && defined(COMPAT_LINUX32))
1697 		if (offset64 > INT32_MAX)
1698 			return (EOVERFLOW);
1699 #endif
1700 		offset = (l_long)offset64;
1701 		error = copyout(&offset, arg->offset, sizeof(offset));
1702 		if (error != 0)
1703 			return (error);
1704 	}
1705 
1706 	return (ret);
1707 }
1708 
1709 #if defined(__i386__) || defined(__arm__) || \
1710     (defined(__amd64__) && defined(COMPAT_LINUX32))
1711 
1712 int
1713 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg)
1714 {
1715 	l_loff_t offset;
1716 	int ret;
1717 	int error;
1718 
1719 	if (arg->offset != NULL) {
1720 		error = copyin(arg->offset, &offset, sizeof(offset));
1721 		if (error != 0)
1722 			return (error);
1723 	}
1724 
1725 	ret = linux_sendfile_common(td, arg->out, arg->in,
1726 		arg->offset != NULL ? &offset : NULL, arg->count);
1727 
1728 	if (arg->offset != NULL) {
1729 		error = copyout(&offset, arg->offset, sizeof(offset));
1730 		if (error != 0)
1731 			return (error);
1732 	}
1733 
1734 	return (ret);
1735 }
1736 
1737 /* Argument list sizes for linux_socketcall */
1738 static const unsigned char lxs_args_cnt[] = {
1739 	0 /* unused*/,		3 /* socket */,
1740 	3 /* bind */,		3 /* connect */,
1741 	2 /* listen */,		3 /* accept */,
1742 	3 /* getsockname */,	3 /* getpeername */,
1743 	4 /* socketpair */,	4 /* send */,
1744 	4 /* recv */,		6 /* sendto */,
1745 	6 /* recvfrom */,	2 /* shutdown */,
1746 	5 /* setsockopt */,	5 /* getsockopt */,
1747 	3 /* sendmsg */,	3 /* recvmsg */,
1748 	4 /* accept4 */,	5 /* recvmmsg */,
1749 	4 /* sendmmsg */,	4 /* sendfile */
1750 };
1751 #define	LINUX_ARGS_CNT		(nitems(lxs_args_cnt) - 1)
1752 #define	LINUX_ARG_SIZE(x)	(lxs_args_cnt[x] * sizeof(l_ulong))
1753 
1754 int
1755 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1756 {
1757 	l_ulong a[6];
1758 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1759 	register_t l_args[6];
1760 #endif
1761 	void *arg;
1762 	int error;
1763 
1764 	if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
1765 		return (EINVAL);
1766 	error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
1767 	if (error != 0)
1768 		return (error);
1769 
1770 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1771 	for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
1772 		l_args[i] = a[i];
1773 	arg = l_args;
1774 #else
1775 	arg = a;
1776 #endif
1777 	switch (args->what) {
1778 	case LINUX_SOCKET:
1779 		return (linux_socket(td, arg));
1780 	case LINUX_BIND:
1781 		return (linux_bind(td, arg));
1782 	case LINUX_CONNECT:
1783 		return (linux_connect(td, arg));
1784 	case LINUX_LISTEN:
1785 		return (linux_listen(td, arg));
1786 	case LINUX_ACCEPT:
1787 		return (linux_accept(td, arg));
1788 	case LINUX_GETSOCKNAME:
1789 		return (linux_getsockname(td, arg));
1790 	case LINUX_GETPEERNAME:
1791 		return (linux_getpeername(td, arg));
1792 	case LINUX_SOCKETPAIR:
1793 		return (linux_socketpair(td, arg));
1794 	case LINUX_SEND:
1795 		return (linux_send(td, arg));
1796 	case LINUX_RECV:
1797 		return (linux_recv(td, arg));
1798 	case LINUX_SENDTO:
1799 		return (linux_sendto(td, arg));
1800 	case LINUX_RECVFROM:
1801 		return (linux_recvfrom(td, arg));
1802 	case LINUX_SHUTDOWN:
1803 		return (linux_shutdown(td, arg));
1804 	case LINUX_SETSOCKOPT:
1805 		return (linux_setsockopt(td, arg));
1806 	case LINUX_GETSOCKOPT:
1807 		return (linux_getsockopt(td, arg));
1808 	case LINUX_SENDMSG:
1809 		return (linux_sendmsg(td, arg));
1810 	case LINUX_RECVMSG:
1811 		return (linux_recvmsg(td, arg));
1812 	case LINUX_ACCEPT4:
1813 		return (linux_accept4(td, arg));
1814 	case LINUX_RECVMMSG:
1815 		return (linux_recvmmsg(td, arg));
1816 	case LINUX_SENDMMSG:
1817 		return (linux_sendmmsg(td, arg));
1818 	case LINUX_SENDFILE:
1819 		return (linux_sendfile(td, arg));
1820 	}
1821 
1822 	linux_msg(td, "socket type %d not implemented", args->what);
1823 	return (ENOSYS);
1824 }
1825 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */
1826