xref: /freebsd/sys/compat/linux/linux_socket.c (revision 580744621f33383027108364dcadad718df46ffe)
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 	switch (level) {
96 	case LINUX_SOL_SOCKET:
97 		return (SOL_SOCKET);
98 	}
99 	return (level);
100 }
101 
102 static int
103 bsd_to_linux_sockopt_level(int level)
104 {
105 
106 	switch (level) {
107 	case SOL_SOCKET:
108 		return (LINUX_SOL_SOCKET);
109 	}
110 	return (level);
111 }
112 
113 static int
114 linux_to_bsd_ip_sockopt(int opt)
115 {
116 
117 	switch (opt) {
118 	case LINUX_IP_TOS:
119 		return (IP_TOS);
120 	case LINUX_IP_TTL:
121 		return (IP_TTL);
122 	case LINUX_IP_OPTIONS:
123 		return (IP_OPTIONS);
124 	case LINUX_IP_MULTICAST_IF:
125 		return (IP_MULTICAST_IF);
126 	case LINUX_IP_MULTICAST_TTL:
127 		return (IP_MULTICAST_TTL);
128 	case LINUX_IP_MULTICAST_LOOP:
129 		return (IP_MULTICAST_LOOP);
130 	case LINUX_IP_ADD_MEMBERSHIP:
131 		return (IP_ADD_MEMBERSHIP);
132 	case LINUX_IP_DROP_MEMBERSHIP:
133 		return (IP_DROP_MEMBERSHIP);
134 	case LINUX_IP_HDRINCL:
135 		return (IP_HDRINCL);
136 	}
137 	return (-1);
138 }
139 
140 static int
141 linux_to_bsd_ip6_sockopt(int opt)
142 {
143 
144 	switch (opt) {
145 	case LINUX_IPV6_NEXTHOP:
146 		return (IPV6_NEXTHOP);
147 	case LINUX_IPV6_UNICAST_HOPS:
148 		return (IPV6_UNICAST_HOPS);
149 	case LINUX_IPV6_MULTICAST_IF:
150 		return (IPV6_MULTICAST_IF);
151 	case LINUX_IPV6_MULTICAST_HOPS:
152 		return (IPV6_MULTICAST_HOPS);
153 	case LINUX_IPV6_MULTICAST_LOOP:
154 		return (IPV6_MULTICAST_LOOP);
155 	case LINUX_IPV6_ADD_MEMBERSHIP:
156 		return (IPV6_JOIN_GROUP);
157 	case LINUX_IPV6_DROP_MEMBERSHIP:
158 		return (IPV6_LEAVE_GROUP);
159 	case LINUX_IPV6_V6ONLY:
160 		return (IPV6_V6ONLY);
161 	case LINUX_IPV6_DONTFRAG:
162 		return (IPV6_DONTFRAG);
163 #if 0
164 	case LINUX_IPV6_CHECKSUM:
165 		return (IPV6_CHECKSUM);
166 	case LINUX_IPV6_RECVPKTINFO:
167 		return (IPV6_RECVPKTINFO);
168 	case LINUX_IPV6_PKTINFO:
169 		return (IPV6_PKTINFO);
170 	case LINUX_IPV6_RECVHOPLIMIT:
171 		return (IPV6_RECVHOPLIMIT);
172 	case LINUX_IPV6_HOPLIMIT:
173 		return (IPV6_HOPLIMIT);
174 	case LINUX_IPV6_RECVHOPOPTS:
175 		return (IPV6_RECVHOPOPTS);
176 	case LINUX_IPV6_HOPOPTS:
177 		return (IPV6_HOPOPTS);
178 	case LINUX_IPV6_RTHDRDSTOPTS:
179 		return (IPV6_RTHDRDSTOPTS);
180 	case LINUX_IPV6_RECVRTHDR:
181 		return (IPV6_RECVRTHDR);
182 	case LINUX_IPV6_RTHDR:
183 		return (IPV6_RTHDR);
184 	case LINUX_IPV6_RECVDSTOPTS:
185 		return (IPV6_RECVDSTOPTS);
186 	case LINUX_IPV6_DSTOPTS:
187 		return (IPV6_DSTOPTS);
188 	case LINUX_IPV6_RECVPATHMTU:
189 		return (IPV6_RECVPATHMTU);
190 	case LINUX_IPV6_PATHMTU:
191 		return (IPV6_PATHMTU);
192 #endif
193 	}
194 	return (-1);
195 }
196 
197 static int
198 linux_to_bsd_so_sockopt(int opt)
199 {
200 
201 	switch (opt) {
202 	case LINUX_SO_DEBUG:
203 		return (SO_DEBUG);
204 	case LINUX_SO_REUSEADDR:
205 		return (SO_REUSEADDR);
206 	case LINUX_SO_TYPE:
207 		return (SO_TYPE);
208 	case LINUX_SO_ERROR:
209 		return (SO_ERROR);
210 	case LINUX_SO_DONTROUTE:
211 		return (SO_DONTROUTE);
212 	case LINUX_SO_BROADCAST:
213 		return (SO_BROADCAST);
214 	case LINUX_SO_SNDBUF:
215 		return (SO_SNDBUF);
216 	case LINUX_SO_RCVBUF:
217 		return (SO_RCVBUF);
218 	case LINUX_SO_KEEPALIVE:
219 		return (SO_KEEPALIVE);
220 	case LINUX_SO_OOBINLINE:
221 		return (SO_OOBINLINE);
222 	case LINUX_SO_LINGER:
223 		return (SO_LINGER);
224 	case LINUX_SO_PEERCRED:
225 		return (LOCAL_PEERCRED);
226 	case LINUX_SO_RCVLOWAT:
227 		return (SO_RCVLOWAT);
228 	case LINUX_SO_SNDLOWAT:
229 		return (SO_SNDLOWAT);
230 	case LINUX_SO_RCVTIMEO:
231 		return (SO_RCVTIMEO);
232 	case LINUX_SO_SNDTIMEO:
233 		return (SO_SNDTIMEO);
234 	case LINUX_SO_TIMESTAMP:
235 		return (SO_TIMESTAMP);
236 	case LINUX_SO_ACCEPTCONN:
237 		return (SO_ACCEPTCONN);
238 	}
239 	return (-1);
240 }
241 
242 static int
243 linux_to_bsd_tcp_sockopt(int opt)
244 {
245 
246 	switch (opt) {
247 	case LINUX_TCP_NODELAY:
248 		return (TCP_NODELAY);
249 	case LINUX_TCP_MAXSEG:
250 		return (TCP_MAXSEG);
251 	case LINUX_TCP_CORK:
252 		return (TCP_NOPUSH);
253 	case LINUX_TCP_KEEPIDLE:
254 		return (TCP_KEEPIDLE);
255 	case LINUX_TCP_KEEPINTVL:
256 		return (TCP_KEEPINTVL);
257 	case LINUX_TCP_KEEPCNT:
258 		return (TCP_KEEPCNT);
259 	case LINUX_TCP_MD5SIG:
260 		return (TCP_MD5SIG);
261 	}
262 	return (-1);
263 }
264 
265 static int
266 linux_to_bsd_msg_flags(int flags)
267 {
268 	int ret_flags = 0;
269 
270 	if (flags & LINUX_MSG_OOB)
271 		ret_flags |= MSG_OOB;
272 	if (flags & LINUX_MSG_PEEK)
273 		ret_flags |= MSG_PEEK;
274 	if (flags & LINUX_MSG_DONTROUTE)
275 		ret_flags |= MSG_DONTROUTE;
276 	if (flags & LINUX_MSG_CTRUNC)
277 		ret_flags |= MSG_CTRUNC;
278 	if (flags & LINUX_MSG_TRUNC)
279 		ret_flags |= MSG_TRUNC;
280 	if (flags & LINUX_MSG_DONTWAIT)
281 		ret_flags |= MSG_DONTWAIT;
282 	if (flags & LINUX_MSG_EOR)
283 		ret_flags |= MSG_EOR;
284 	if (flags & LINUX_MSG_WAITALL)
285 		ret_flags |= MSG_WAITALL;
286 	if (flags & LINUX_MSG_NOSIGNAL)
287 		ret_flags |= MSG_NOSIGNAL;
288 #if 0 /* not handled */
289 	if (flags & LINUX_MSG_PROXY)
290 		;
291 	if (flags & LINUX_MSG_FIN)
292 		;
293 	if (flags & LINUX_MSG_SYN)
294 		;
295 	if (flags & LINUX_MSG_CONFIRM)
296 		;
297 	if (flags & LINUX_MSG_RST)
298 		;
299 	if (flags & LINUX_MSG_ERRQUEUE)
300 		;
301 #endif
302 	return (ret_flags);
303 }
304 
305 static int
306 linux_to_bsd_cmsg_type(int cmsg_type)
307 {
308 
309 	switch (cmsg_type) {
310 	case LINUX_SCM_RIGHTS:
311 		return (SCM_RIGHTS);
312 	case LINUX_SCM_CREDENTIALS:
313 		return (SCM_CREDS);
314 	}
315 	return (-1);
316 }
317 
318 static int
319 bsd_to_linux_cmsg_type(int cmsg_type)
320 {
321 
322 	switch (cmsg_type) {
323 	case SCM_RIGHTS:
324 		return (LINUX_SCM_RIGHTS);
325 	case SCM_CREDS:
326 		return (LINUX_SCM_CREDENTIALS);
327 	case SCM_TIMESTAMP:
328 		return (LINUX_SCM_TIMESTAMP);
329 	}
330 	return (-1);
331 }
332 
333 static int
334 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
335 {
336 	if (lhdr->msg_controllen > INT_MAX)
337 		return (ENOBUFS);
338 
339 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
340 	bhdr->msg_namelen	= lhdr->msg_namelen;
341 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
342 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
343 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
344 
345 	/*
346 	 * msg_controllen is skipped since BSD and LINUX control messages
347 	 * are potentially different sizes (e.g. the cred structure used
348 	 * by SCM_CREDS is different between the two operating system).
349 	 *
350 	 * The caller can set it (if necessary) after converting all the
351 	 * control messages.
352 	 */
353 
354 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
355 	return (0);
356 }
357 
358 static int
359 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
360 {
361 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
362 	lhdr->msg_namelen	= bhdr->msg_namelen;
363 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
364 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
365 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
366 
367 	/*
368 	 * msg_controllen is skipped since BSD and LINUX control messages
369 	 * are potentially different sizes (e.g. the cred structure used
370 	 * by SCM_CREDS is different between the two operating system).
371 	 *
372 	 * The caller can set it (if necessary) after converting all the
373 	 * control messages.
374 	 */
375 
376 	/* msg_flags skipped */
377 	return (0);
378 }
379 
380 static int
381 linux_set_socket_flags(int lflags, int *flags)
382 {
383 
384 	if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
385 		return (EINVAL);
386 	if (lflags & LINUX_SOCK_NONBLOCK)
387 		*flags |= SOCK_NONBLOCK;
388 	if (lflags & LINUX_SOCK_CLOEXEC)
389 		*flags |= SOCK_CLOEXEC;
390 	return (0);
391 }
392 
393 static int
394 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
395     struct mbuf *control, enum uio_seg segflg)
396 {
397 	struct sockaddr *to;
398 	int error, len;
399 
400 	if (mp->msg_name != NULL) {
401 		len = mp->msg_namelen;
402 		error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len);
403 		if (error != 0)
404 			return (error);
405 		mp->msg_name = to;
406 	} else
407 		to = NULL;
408 
409 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
410 	    segflg);
411 
412 	if (to)
413 		free(to, M_SONAME);
414 	return (error);
415 }
416 
417 /* Return 0 if IP_HDRINCL is set for the given socket. */
418 static int
419 linux_check_hdrincl(struct thread *td, int s)
420 {
421 	int error, optval;
422 	socklen_t size_val;
423 
424 	size_val = sizeof(optval);
425 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
426 	    &optval, UIO_SYSSPACE, &size_val);
427 	if (error != 0)
428 		return (error);
429 
430 	return (optval == 0);
431 }
432 
433 /*
434  * Updated sendto() when IP_HDRINCL is set:
435  * tweak endian-dependent fields in the IP packet.
436  */
437 static int
438 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
439 {
440 /*
441  * linux_ip_copysize defines how many bytes we should copy
442  * from the beginning of the IP packet before we customize it for BSD.
443  * It should include all the fields we modify (ip_len and ip_off).
444  */
445 #define linux_ip_copysize	8
446 
447 	struct ip *packet;
448 	struct msghdr msg;
449 	struct iovec aiov[1];
450 	int error;
451 
452 	/* Check that the packet isn't too big or too small. */
453 	if (linux_args->len < linux_ip_copysize ||
454 	    linux_args->len > IP_MAXPACKET)
455 		return (EINVAL);
456 
457 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
458 
459 	/* Make kernel copy of the packet to be sent */
460 	if ((error = copyin(PTRIN(linux_args->msg), packet,
461 	    linux_args->len)))
462 		goto goout;
463 
464 	/* Convert fields from Linux to BSD raw IP socket format */
465 	packet->ip_len = linux_args->len;
466 	packet->ip_off = ntohs(packet->ip_off);
467 
468 	/* Prepare the msghdr and iovec structures describing the new packet */
469 	msg.msg_name = PTRIN(linux_args->to);
470 	msg.msg_namelen = linux_args->tolen;
471 	msg.msg_iov = aiov;
472 	msg.msg_iovlen = 1;
473 	msg.msg_control = NULL;
474 	msg.msg_flags = 0;
475 	aiov[0].iov_base = (char *)packet;
476 	aiov[0].iov_len = linux_args->len;
477 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
478 	    NULL, UIO_SYSSPACE);
479 goout:
480 	free(packet, M_LINUX);
481 	return (error);
482 }
483 
484 int
485 linux_socket(struct thread *td, struct linux_socket_args *args)
486 {
487 	int domain, retval_socket, type;
488 
489 	type = args->type & LINUX_SOCK_TYPE_MASK;
490 	if (type < 0 || type > LINUX_SOCK_MAX)
491 		return (EINVAL);
492 	retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
493 		&type);
494 	if (retval_socket != 0)
495 		return (retval_socket);
496 	domain = linux_to_bsd_domain(args->domain);
497 	if (domain == -1)
498 		return (EAFNOSUPPORT);
499 
500 	retval_socket = kern_socket(td, domain, type, args->protocol);
501 	if (retval_socket)
502 		return (retval_socket);
503 
504 	if (type == SOCK_RAW
505 	    && (args->protocol == IPPROTO_RAW || args->protocol == 0)
506 	    && domain == PF_INET) {
507 		/* It's a raw IP socket: set the IP_HDRINCL option. */
508 		int hdrincl;
509 
510 		hdrincl = 1;
511 		/* We ignore any error returned by kern_setsockopt() */
512 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
513 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
514 	}
515 #ifdef INET6
516 	/*
517 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
518 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
519 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
520 	 * sysctl value.
521 	 */
522 	if (domain == PF_INET6) {
523 		int v6only;
524 
525 		v6only = 0;
526 		/* We ignore any error returned by setsockopt() */
527 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
528 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
529 	}
530 #endif
531 
532 	return (retval_socket);
533 }
534 
535 int
536 linux_bind(struct thread *td, struct linux_bind_args *args)
537 {
538 	struct sockaddr *sa;
539 	int error;
540 
541 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
542 	    &args->namelen);
543 	if (error != 0)
544 		return (error);
545 
546 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
547 	free(sa, M_SONAME);
548 
549 	/* XXX */
550 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
551 		return (EINVAL);
552 	return (error);
553 }
554 
555 int
556 linux_connect(struct thread *td, struct linux_connect_args *args)
557 {
558 	struct socket *so;
559 	struct sockaddr *sa;
560 	struct file *fp;
561 	u_int fflag;
562 	int error;
563 
564 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
565 	    &args->namelen);
566 	if (error != 0)
567 		return (error);
568 
569 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
570 	free(sa, M_SONAME);
571 	if (error != EISCONN)
572 		return (error);
573 
574 	/*
575 	 * Linux doesn't return EISCONN the first time it occurs,
576 	 * when on a non-blocking socket. Instead it returns the
577 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
578 	 */
579 	error = getsock_cap(td, args->s, &cap_connect_rights,
580 	    &fp, &fflag, NULL);
581 	if (error != 0)
582 		return (error);
583 
584 	error = EISCONN;
585 	so = fp->f_data;
586 	if (fflag & FNONBLOCK) {
587 		SOCK_LOCK(so);
588 		if (so->so_emuldata == 0)
589 			error = so->so_error;
590 		so->so_emuldata = (void *)1;
591 		SOCK_UNLOCK(so);
592 	}
593 	fdrop(fp, td);
594 
595 	return (error);
596 }
597 
598 int
599 linux_listen(struct thread *td, struct linux_listen_args *args)
600 {
601 
602 	return (kern_listen(td, args->s, args->backlog));
603 }
604 
605 static int
606 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
607     l_uintptr_t namelen, int flags)
608 {
609 	struct l_sockaddr *lsa;
610 	struct sockaddr *sa;
611 	struct file *fp;
612 	int bflags, len;
613 	struct socket *so;
614 	int error, error1;
615 
616 	bflags = 0;
617 	error = linux_set_socket_flags(flags, &bflags);
618 	if (error != 0)
619 		return (error);
620 
621 	sa = NULL;
622 	if (PTRIN(addr) == NULL) {
623 		len = 0;
624 		error = kern_accept4(td, s, NULL, NULL, bflags, NULL);
625 	} else {
626 		error = copyin(PTRIN(namelen), &len, sizeof(len));
627 		if (error != 0)
628 			return (error);
629 		if (len < 0)
630 			return (EINVAL);
631 		error = kern_accept4(td, s, &sa, &len, bflags, &fp);
632 		if (error == 0)
633 			fdrop(fp, td);
634 	}
635 
636 	if (error != 0) {
637 		/*
638 		 * XXX. This is wrong, different sockaddr structures
639 		 * have different sizes.
640 		 */
641 		if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
642 		{
643 			error = EINVAL;
644 			goto out;
645 		}
646 		if (error == EINVAL) {
647 			error1 = getsock_cap(td, s, &cap_accept_rights, &fp, NULL, NULL);
648 			if (error1 != 0) {
649 				error = error1;
650 				goto out;
651 			}
652 			so = fp->f_data;
653 			if (so->so_type == SOCK_DGRAM)
654 				error = EOPNOTSUPP;
655 			fdrop(fp, td);
656 		}
657 		goto out;
658 	}
659 
660 	if (len != 0 && error == 0) {
661 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
662 		if (error == 0)
663 			error = copyout(lsa, PTRIN(addr), len);
664 		free(lsa, M_SONAME);
665 	}
666 
667 	free(sa, M_SONAME);
668 
669 out:
670 	if (error != 0) {
671 		(void)kern_close(td, td->td_retval[0]);
672 		td->td_retval[0] = 0;
673 	}
674 	return (error);
675 }
676 
677 int
678 linux_accept(struct thread *td, struct linux_accept_args *args)
679 {
680 
681 	return (linux_accept_common(td, args->s, args->addr,
682 	    args->namelen, 0));
683 }
684 
685 int
686 linux_accept4(struct thread *td, struct linux_accept4_args *args)
687 {
688 
689 	return (linux_accept_common(td, args->s, args->addr,
690 	    args->namelen, args->flags));
691 }
692 
693 int
694 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
695 {
696 	struct l_sockaddr *lsa;
697 	struct sockaddr *sa;
698 	int len, error;
699 
700 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
701 	if (error != 0)
702 		return (error);
703 
704 	error = kern_getsockname(td, args->s, &sa, &len);
705 	if (error != 0)
706 		return (error);
707 
708 	if (len != 0) {
709 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
710 		if (error == 0)
711 			error = copyout(lsa, PTRIN(args->addr),
712 			    len);
713 		free(lsa, M_SONAME);
714 	}
715 
716 	free(sa, M_SONAME);
717 	if (error == 0)
718 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
719 	return (error);
720 }
721 
722 int
723 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
724 {
725 	struct l_sockaddr *lsa;
726 	struct sockaddr *sa;
727 	int len, error;
728 
729 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
730 	if (error != 0)
731 		return (error);
732 	if (len < 0)
733 		return (EINVAL);
734 
735 	error = kern_getpeername(td, args->s, &sa, &len);
736 	if (error != 0)
737 		return (error);
738 
739 	if (len != 0) {
740 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
741 		if (error == 0)
742 			error = copyout(lsa, PTRIN(args->addr),
743 			    len);
744 		free(lsa, M_SONAME);
745 	}
746 
747 	free(sa, M_SONAME);
748 	if (error == 0)
749 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
750 	return (error);
751 }
752 
753 int
754 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
755 {
756 	int domain, error, sv[2], type;
757 
758 	domain = linux_to_bsd_domain(args->domain);
759 	if (domain != PF_LOCAL)
760 		return (EAFNOSUPPORT);
761 	type = args->type & LINUX_SOCK_TYPE_MASK;
762 	if (type < 0 || type > LINUX_SOCK_MAX)
763 		return (EINVAL);
764 	error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
765 	    &type);
766 	if (error != 0)
767 		return (error);
768 	if (args->protocol != 0 && args->protocol != PF_UNIX) {
769 
770 		/*
771 		 * Use of PF_UNIX as protocol argument is not right,
772 		 * but Linux does it.
773 		 * Do not map PF_UNIX as its Linux value is identical
774 		 * to FreeBSD one.
775 		 */
776 		return (EPROTONOSUPPORT);
777 	}
778 	error = kern_socketpair(td, domain, type, 0, sv);
779 	if (error != 0)
780                 return (error);
781         error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int));
782         if (error != 0) {
783                 (void)kern_close(td, sv[0]);
784                 (void)kern_close(td, sv[1]);
785         }
786 	return (error);
787 }
788 
789 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
790 struct linux_send_args {
791 	register_t s;
792 	register_t msg;
793 	register_t len;
794 	register_t flags;
795 };
796 
797 static int
798 linux_send(struct thread *td, struct linux_send_args *args)
799 {
800 	struct sendto_args /* {
801 		int s;
802 		caddr_t buf;
803 		int len;
804 		int flags;
805 		caddr_t to;
806 		int tolen;
807 	} */ bsd_args;
808 	struct file *fp;
809 	int error, fflag;
810 
811 	bsd_args.s = args->s;
812 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
813 	bsd_args.len = args->len;
814 	bsd_args.flags = args->flags;
815 	bsd_args.to = NULL;
816 	bsd_args.tolen = 0;
817 	error = sys_sendto(td, &bsd_args);
818 	if (error == ENOTCONN) {
819 		/*
820 		 * Linux doesn't return ENOTCONN for non-blocking sockets.
821 		 * Instead it returns the EAGAIN.
822 		 */
823 		error = getsock_cap(td, args->s, &cap_send_rights, &fp,
824 		    &fflag, NULL);
825 		if (error == 0) {
826 			if (fflag & FNONBLOCK)
827 				error = EAGAIN;
828 			fdrop(fp, td);
829 		}
830 	}
831 	return (error);
832 }
833 
834 struct linux_recv_args {
835 	register_t s;
836 	register_t msg;
837 	register_t len;
838 	register_t flags;
839 };
840 
841 static int
842 linux_recv(struct thread *td, struct linux_recv_args *args)
843 {
844 	struct recvfrom_args /* {
845 		int s;
846 		caddr_t buf;
847 		int len;
848 		int flags;
849 		struct sockaddr *from;
850 		socklen_t fromlenaddr;
851 	} */ bsd_args;
852 
853 	bsd_args.s = args->s;
854 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
855 	bsd_args.len = args->len;
856 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
857 	bsd_args.from = NULL;
858 	bsd_args.fromlenaddr = 0;
859 	return (sys_recvfrom(td, &bsd_args));
860 }
861 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
862 
863 int
864 linux_sendto(struct thread *td, struct linux_sendto_args *args)
865 {
866 	struct msghdr msg;
867 	struct iovec aiov;
868 
869 	if (linux_check_hdrincl(td, args->s) == 0)
870 		/* IP_HDRINCL set, tweak the packet before sending */
871 		return (linux_sendto_hdrincl(td, args));
872 
873 	msg.msg_name = PTRIN(args->to);
874 	msg.msg_namelen = args->tolen;
875 	msg.msg_iov = &aiov;
876 	msg.msg_iovlen = 1;
877 	msg.msg_control = NULL;
878 	msg.msg_flags = 0;
879 	aiov.iov_base = PTRIN(args->msg);
880 	aiov.iov_len = args->len;
881 	return (linux_sendit(td, args->s, &msg, args->flags, NULL,
882 	    UIO_USERSPACE));
883 }
884 
885 int
886 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
887 {
888 	struct l_sockaddr *lsa;
889 	struct sockaddr *sa;
890 	struct msghdr msg;
891 	struct iovec aiov;
892 	int error, fromlen;
893 
894 	if (PTRIN(args->fromlen) != NULL) {
895 		error = copyin(PTRIN(args->fromlen), &fromlen,
896 		    sizeof(fromlen));
897 		if (error != 0)
898 			return (error);
899 		if (fromlen < 0)
900 			return (EINVAL);
901 		sa = malloc(fromlen, M_SONAME, M_WAITOK);
902 	} else {
903 		fromlen = 0;
904 		sa = NULL;
905 	}
906 
907 	msg.msg_name = sa;
908 	msg.msg_namelen = fromlen;
909 	msg.msg_iov = &aiov;
910 	msg.msg_iovlen = 1;
911 	aiov.iov_base = PTRIN(args->buf);
912 	aiov.iov_len = args->len;
913 	msg.msg_control = 0;
914 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
915 
916 	error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL);
917 	if (error != 0)
918 		goto out;
919 
920 	if (PTRIN(args->from) != NULL) {
921 		error = bsd_to_linux_sockaddr(sa, &lsa, msg.msg_namelen);
922 		if (error == 0)
923 			error = copyout(lsa, PTRIN(args->from),
924 			    msg.msg_namelen);
925 		free(lsa, M_SONAME);
926 	}
927 
928 	if (error == 0 && PTRIN(args->fromlen) != NULL)
929 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
930 		    sizeof(msg.msg_namelen));
931 out:
932 	free(sa, M_SONAME);
933 	return (error);
934 }
935 
936 static int
937 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
938     l_uint flags)
939 {
940 	struct cmsghdr *cmsg;
941 	struct mbuf *control;
942 	struct msghdr msg;
943 	struct l_cmsghdr linux_cmsg;
944 	struct l_cmsghdr *ptr_cmsg;
945 	struct l_msghdr linux_msg;
946 	struct iovec *iov;
947 	socklen_t datalen;
948 	struct sockaddr *sa;
949 	struct socket *so;
950 	sa_family_t sa_family;
951 	struct file *fp;
952 	void *data;
953 	l_size_t len;
954 	l_size_t clen;
955 	int error, fflag;
956 
957 	error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
958 	if (error != 0)
959 		return (error);
960 
961 	/*
962 	 * Some Linux applications (ping) define a non-NULL control data
963 	 * pointer, but a msg_controllen of 0, which is not allowed in the
964 	 * FreeBSD system call interface.  NULL the msg_control pointer in
965 	 * order to handle this case.  This should be checked, but allows the
966 	 * Linux ping to work.
967 	 */
968 	if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
969 		linux_msg.msg_control = PTROUT(NULL);
970 
971 	error = linux_to_bsd_msghdr(&msg, &linux_msg);
972 	if (error != 0)
973 		return (error);
974 
975 #ifdef COMPAT_LINUX32
976 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
977 	    &iov, EMSGSIZE);
978 #else
979 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
980 #endif
981 	if (error != 0)
982 		return (error);
983 
984 	control = NULL;
985 
986 	error = kern_getsockname(td, s, &sa, &datalen);
987 	if (error != 0)
988 		goto bad;
989 	sa_family = sa->sa_family;
990 	free(sa, M_SONAME);
991 
992 	if (flags & LINUX_MSG_OOB) {
993 		error = EOPNOTSUPP;
994 		if (sa_family == AF_UNIX)
995 			goto bad;
996 
997 		error = getsock_cap(td, s, &cap_send_rights, &fp,
998 		    &fflag, NULL);
999 		if (error != 0)
1000 			goto bad;
1001 		so = fp->f_data;
1002 		if (so->so_type != SOCK_STREAM)
1003 			error = EOPNOTSUPP;
1004 		fdrop(fp, td);
1005 		if (error != 0)
1006 			goto bad;
1007 	}
1008 
1009 	if (linux_msg.msg_controllen >= sizeof(struct l_cmsghdr)) {
1010 
1011 		error = ENOBUFS;
1012 		control = m_get(M_WAITOK, MT_CONTROL);
1013 		MCLGET(control, M_WAITOK);
1014 		data = mtod(control, void *);
1015 		datalen = 0;
1016 
1017 		ptr_cmsg = PTRIN(linux_msg.msg_control);
1018 		clen = linux_msg.msg_controllen;
1019 		do {
1020 			error = copyin(ptr_cmsg, &linux_cmsg,
1021 			    sizeof(struct l_cmsghdr));
1022 			if (error != 0)
1023 				goto bad;
1024 
1025 			error = EINVAL;
1026 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1027 			    linux_cmsg.cmsg_len > clen)
1028 				goto bad;
1029 
1030 			if (datalen + CMSG_HDRSZ > MCLBYTES)
1031 				goto bad;
1032 
1033 			/*
1034 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1035 			 * so return EINVAL in any other cmsg_type
1036 			 */
1037 			cmsg = data;
1038 			cmsg->cmsg_type =
1039 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1040 			cmsg->cmsg_level =
1041 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1042 			if (cmsg->cmsg_type == -1
1043 			    || cmsg->cmsg_level != SOL_SOCKET)
1044 				goto bad;
1045 
1046 			/*
1047 			 * Some applications (e.g. pulseaudio) attempt to
1048 			 * send ancillary data even if the underlying protocol
1049 			 * doesn't support it which is not allowed in the
1050 			 * FreeBSD system call interface.
1051 			 */
1052 			if (sa_family != AF_UNIX)
1053 				continue;
1054 
1055 			if (cmsg->cmsg_type == SCM_CREDS) {
1056 				len = sizeof(struct cmsgcred);
1057 				if (datalen + CMSG_SPACE(len) > MCLBYTES)
1058 					goto bad;
1059 
1060 				/*
1061 				 * The lower levels will fill in the structure
1062 				 */
1063 				memset(CMSG_DATA(data), 0, len);
1064 			} else {
1065 				len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1066 				if (datalen + CMSG_SPACE(len) < datalen ||
1067 				    datalen + CMSG_SPACE(len) > MCLBYTES)
1068 					goto bad;
1069 
1070 				error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1071 				    CMSG_DATA(data), len);
1072 				if (error != 0)
1073 					goto bad;
1074 			}
1075 
1076 			cmsg->cmsg_len = CMSG_LEN(len);
1077 			data = (char *)data + CMSG_SPACE(len);
1078 			datalen += CMSG_SPACE(len);
1079 
1080 			if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1081 				break;
1082 
1083 			clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1084 			ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1085 			    LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1086 		} while(clen >= sizeof(struct l_cmsghdr));
1087 
1088 		control->m_len = datalen;
1089 		if (datalen == 0) {
1090 			m_freem(control);
1091 			control = NULL;
1092 		}
1093 	}
1094 
1095 	msg.msg_iov = iov;
1096 	msg.msg_flags = 0;
1097 	error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1098 	control = NULL;
1099 
1100 bad:
1101 	m_freem(control);
1102 	free(iov, M_IOV);
1103 	return (error);
1104 }
1105 
1106 int
1107 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1108 {
1109 
1110 	return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1111 	    args->flags));
1112 }
1113 
1114 int
1115 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1116 {
1117 	struct l_mmsghdr *msg;
1118 	l_uint retval;
1119 	int error, datagrams;
1120 
1121 	if (args->vlen > UIO_MAXIOV)
1122 		args->vlen = UIO_MAXIOV;
1123 
1124 	msg = PTRIN(args->msg);
1125 	datagrams = 0;
1126 	while (datagrams < args->vlen) {
1127 		error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1128 		    args->flags);
1129 		if (error != 0)
1130 			break;
1131 
1132 		retval = td->td_retval[0];
1133 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1134 		if (error != 0)
1135 			break;
1136 		++msg;
1137 		++datagrams;
1138 	}
1139 	if (error == 0)
1140 		td->td_retval[0] = datagrams;
1141 	return (error);
1142 }
1143 
1144 static int
1145 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1146     l_uint flags, struct msghdr *msg)
1147 {
1148 	struct cmsghdr *cm;
1149 	struct cmsgcred *cmcred;
1150 	struct l_cmsghdr *linux_cmsg = NULL;
1151 	struct l_ucred linux_ucred;
1152 	socklen_t datalen, maxlen, outlen;
1153 	struct l_msghdr linux_msg;
1154 	struct iovec *iov, *uiov;
1155 	struct mbuf *control = NULL;
1156 	struct mbuf **controlp;
1157 	struct timeval *ftmvl;
1158 	struct l_sockaddr *lsa;
1159 	struct sockaddr *sa;
1160 	l_timeval ltmvl;
1161 	caddr_t outbuf;
1162 	void *data;
1163 	int error, i, fd, fds, *fdp;
1164 
1165 	error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1166 	if (error != 0)
1167 		return (error);
1168 
1169 	error = linux_to_bsd_msghdr(msg, &linux_msg);
1170 	if (error != 0)
1171 		return (error);
1172 
1173 #ifdef COMPAT_LINUX32
1174 	error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1175 	    &iov, EMSGSIZE);
1176 #else
1177 	error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1178 #endif
1179 	if (error != 0)
1180 		return (error);
1181 
1182 	if (msg->msg_name) {
1183 		sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK);
1184 		msg->msg_name = sa;
1185 	} else
1186 		sa = NULL;
1187 
1188 	uiov = msg->msg_iov;
1189 	msg->msg_iov = iov;
1190 	controlp = (msg->msg_control != NULL) ? &control : NULL;
1191 	error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp);
1192 	msg->msg_iov = uiov;
1193 	if (error != 0)
1194 		goto bad;
1195 
1196 	if (msg->msg_name) {
1197 		msg->msg_name = PTRIN(linux_msg.msg_name);
1198 		error = bsd_to_linux_sockaddr(sa, &lsa, msg->msg_namelen);
1199 		if (error == 0)
1200 			error = copyout(lsa, PTRIN(msg->msg_name),
1201 			    msg->msg_namelen);
1202 		free(lsa, M_SONAME);
1203 		if (error != 0)
1204 			goto bad;
1205 	}
1206 
1207 	error = bsd_to_linux_msghdr(msg, &linux_msg);
1208 	if (error != 0)
1209 		goto bad;
1210 
1211 	maxlen = linux_msg.msg_controllen;
1212 	linux_msg.msg_controllen = 0;
1213 	if (control) {
1214 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1215 
1216 		msg->msg_control = mtod(control, struct cmsghdr *);
1217 		msg->msg_controllen = control->m_len;
1218 
1219 		cm = CMSG_FIRSTHDR(msg);
1220 		outbuf = PTRIN(linux_msg.msg_control);
1221 		outlen = 0;
1222 		while (cm != NULL) {
1223 			linux_cmsg->cmsg_type =
1224 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1225 			linux_cmsg->cmsg_level =
1226 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1227 			if (linux_cmsg->cmsg_type == -1 ||
1228 			    cm->cmsg_level != SOL_SOCKET) {
1229 				error = EINVAL;
1230 				goto bad;
1231 			}
1232 
1233 			data = CMSG_DATA(cm);
1234 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1235 
1236 			switch (cm->cmsg_type) {
1237 			case SCM_RIGHTS:
1238 				if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1239 					fds = datalen / sizeof(int);
1240 					fdp = data;
1241 					for (i = 0; i < fds; i++) {
1242 						fd = *fdp++;
1243 						(void)kern_fcntl(td, fd,
1244 						    F_SETFD, FD_CLOEXEC);
1245 					}
1246 				}
1247 				break;
1248 
1249 			case SCM_CREDS:
1250 				/*
1251 				 * Currently LOCAL_CREDS is never in
1252 				 * effect for Linux so no need to worry
1253 				 * about sockcred
1254 				 */
1255 				if (datalen != sizeof(*cmcred)) {
1256 					error = EMSGSIZE;
1257 					goto bad;
1258 				}
1259 				cmcred = (struct cmsgcred *)data;
1260 				bzero(&linux_ucred, sizeof(linux_ucred));
1261 				linux_ucred.pid = cmcred->cmcred_pid;
1262 				linux_ucred.uid = cmcred->cmcred_uid;
1263 				linux_ucred.gid = cmcred->cmcred_gid;
1264 				data = &linux_ucred;
1265 				datalen = sizeof(linux_ucred);
1266 				break;
1267 
1268 			case SCM_TIMESTAMP:
1269 				if (datalen != sizeof(struct timeval)) {
1270 					error = EMSGSIZE;
1271 					goto bad;
1272 				}
1273 				ftmvl = (struct timeval *)data;
1274 				ltmvl.tv_sec = ftmvl->tv_sec;
1275 				ltmvl.tv_usec = ftmvl->tv_usec;
1276 				data = &ltmvl;
1277 				datalen = sizeof(ltmvl);
1278 				break;
1279 			}
1280 
1281 			if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1282 				if (outlen == 0) {
1283 					error = EMSGSIZE;
1284 					goto bad;
1285 				} else {
1286 					linux_msg.msg_flags |= LINUX_MSG_CTRUNC;
1287 					m_dispose_extcontrolm(control);
1288 					goto out;
1289 				}
1290 			}
1291 
1292 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1293 
1294 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1295 			if (error != 0)
1296 				goto bad;
1297 			outbuf += L_CMSG_HDRSZ;
1298 
1299 			error = copyout(data, outbuf, datalen);
1300 			if (error != 0)
1301 				goto bad;
1302 
1303 			outbuf += LINUX_CMSG_ALIGN(datalen);
1304 			outlen += LINUX_CMSG_LEN(datalen);
1305 
1306 			cm = CMSG_NXTHDR(msg, cm);
1307 		}
1308 		linux_msg.msg_controllen = outlen;
1309 	}
1310 
1311 out:
1312 	error = copyout(&linux_msg, msghdr, sizeof(linux_msg));
1313 
1314 bad:
1315 	if (control != NULL) {
1316 		if (error != 0)
1317 			m_dispose_extcontrolm(control);
1318 		m_freem(control);
1319 	}
1320 	free(iov, M_IOV);
1321 	free(linux_cmsg, M_LINUX);
1322 	free(sa, M_SONAME);
1323 
1324 	return (error);
1325 }
1326 
1327 int
1328 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1329 {
1330 	struct msghdr bsd_msg;
1331 
1332 	return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1333 	    args->flags, &bsd_msg));
1334 }
1335 
1336 int
1337 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1338 {
1339 	struct l_mmsghdr *msg;
1340 	struct msghdr bsd_msg;
1341 	struct l_timespec lts;
1342 	struct timespec ts, tts;
1343 	l_uint retval;
1344 	int error, datagrams;
1345 
1346 	if (args->timeout) {
1347 		error = copyin(args->timeout, &lts, sizeof(struct l_timespec));
1348 		if (error != 0)
1349 			return (error);
1350 		error = linux_to_native_timespec(&ts, &lts);
1351 		if (error != 0)
1352 			return (error);
1353 		getnanotime(&tts);
1354 		timespecadd(&tts, &ts, &tts);
1355 	}
1356 
1357 	msg = PTRIN(args->msg);
1358 	datagrams = 0;
1359 	while (datagrams < args->vlen) {
1360 		error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1361 		    args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1362 		if (error != 0)
1363 			break;
1364 
1365 		retval = td->td_retval[0];
1366 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1367 		if (error != 0)
1368 			break;
1369 		++msg;
1370 		++datagrams;
1371 
1372 		/*
1373 		 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1374 		 */
1375 		if (args->flags & LINUX_MSG_WAITFORONE)
1376 			args->flags |= LINUX_MSG_DONTWAIT;
1377 
1378 		/*
1379 		 * See BUGS section of recvmmsg(2).
1380 		 */
1381 		if (args->timeout) {
1382 			getnanotime(&ts);
1383 			timespecsub(&ts, &tts, &ts);
1384 			if (!timespecisset(&ts) || ts.tv_sec > 0)
1385 				break;
1386 		}
1387 		/* Out of band data, return right away. */
1388 		if (bsd_msg.msg_flags & MSG_OOB)
1389 			break;
1390 	}
1391 	if (error == 0)
1392 		td->td_retval[0] = datagrams;
1393 	return (error);
1394 }
1395 
1396 int
1397 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1398 {
1399 
1400 	return (kern_shutdown(td, args->s, args->how));
1401 }
1402 
1403 int
1404 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1405 {
1406 	l_timeval linux_tv;
1407 	struct sockaddr *sa;
1408 	struct timeval tv;
1409 	socklen_t len;
1410 	int error, level, name;
1411 
1412 	level = linux_to_bsd_sockopt_level(args->level);
1413 	switch (level) {
1414 	case SOL_SOCKET:
1415 		name = linux_to_bsd_so_sockopt(args->optname);
1416 		switch (name) {
1417 		case SO_RCVTIMEO:
1418 			/* FALLTHROUGH */
1419 		case SO_SNDTIMEO:
1420 			error = copyin(PTRIN(args->optval), &linux_tv,
1421 			    sizeof(linux_tv));
1422 			if (error != 0)
1423 				return (error);
1424 			tv.tv_sec = linux_tv.tv_sec;
1425 			tv.tv_usec = linux_tv.tv_usec;
1426 			return (kern_setsockopt(td, args->s, level,
1427 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1428 			/* NOTREACHED */
1429 		default:
1430 			break;
1431 		}
1432 		break;
1433 	case IPPROTO_IP:
1434 		if (args->optname == LINUX_IP_RECVERR &&
1435 		    linux_ignore_ip_recverr) {
1436 			/*
1437 			 * XXX: This is a hack to unbreak DNS resolution
1438 			 *	with glibc 2.30 and above.
1439 			 */
1440 			return (0);
1441 		}
1442 		name = linux_to_bsd_ip_sockopt(args->optname);
1443 		break;
1444 	case IPPROTO_IPV6:
1445 		name = linux_to_bsd_ip6_sockopt(args->optname);
1446 		break;
1447 	case IPPROTO_TCP:
1448 		name = linux_to_bsd_tcp_sockopt(args->optname);
1449 		break;
1450 	default:
1451 		name = -1;
1452 		break;
1453 	}
1454 	if (name == -1) {
1455 		linux_msg(curthread,
1456 		    "unsupported setsockopt level %d optname %d",
1457 		    args->level, args->optname);
1458 		return (ENOPROTOOPT);
1459 	}
1460 
1461 	if (name == IPV6_NEXTHOP) {
1462 		len = args->optlen;
1463 		error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len);
1464 		if (error != 0)
1465 			return (error);
1466 
1467 		error = kern_setsockopt(td, args->s, level,
1468 		    name, sa, UIO_SYSSPACE, len);
1469 		free(sa, M_SONAME);
1470 	} else {
1471 		error = kern_setsockopt(td, args->s, level,
1472 		    name, PTRIN(args->optval), UIO_USERSPACE, args->optlen);
1473 	}
1474 
1475 	return (error);
1476 }
1477 
1478 int
1479 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1480 {
1481 	l_timeval linux_tv;
1482 	struct timeval tv;
1483 	socklen_t tv_len, xulen, len;
1484 	struct l_sockaddr *lsa;
1485 	struct sockaddr *sa;
1486 	struct xucred xu;
1487 	struct l_ucred lxu;
1488 	int error, level, name, newval;
1489 
1490 	level = linux_to_bsd_sockopt_level(args->level);
1491 	switch (level) {
1492 	case SOL_SOCKET:
1493 		name = linux_to_bsd_so_sockopt(args->optname);
1494 		switch (name) {
1495 		case SO_RCVTIMEO:
1496 			/* FALLTHROUGH */
1497 		case SO_SNDTIMEO:
1498 			tv_len = sizeof(tv);
1499 			error = kern_getsockopt(td, args->s, level,
1500 			    name, &tv, UIO_SYSSPACE, &tv_len);
1501 			if (error != 0)
1502 				return (error);
1503 			linux_tv.tv_sec = tv.tv_sec;
1504 			linux_tv.tv_usec = tv.tv_usec;
1505 			return (copyout(&linux_tv, PTRIN(args->optval),
1506 			    sizeof(linux_tv)));
1507 			/* NOTREACHED */
1508 		case LOCAL_PEERCRED:
1509 			if (args->optlen < sizeof(lxu))
1510 				return (EINVAL);
1511 			/*
1512 			 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
1513 			 * but by the Unix socket's level 0.
1514 			 */
1515 			level = 0;
1516 			xulen = sizeof(xu);
1517 			error = kern_getsockopt(td, args->s, level,
1518 			    name, &xu, UIO_SYSSPACE, &xulen);
1519 			if (error != 0)
1520 				return (error);
1521 			lxu.pid = xu.cr_pid;
1522 			lxu.uid = xu.cr_uid;
1523 			lxu.gid = xu.cr_gid;
1524 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1525 			/* NOTREACHED */
1526 		case SO_ERROR:
1527 			len = sizeof(newval);
1528 			error = kern_getsockopt(td, args->s, level,
1529 			    name, &newval, UIO_SYSSPACE, &len);
1530 			if (error != 0)
1531 				return (error);
1532 			newval = -SV_ABI_ERRNO(td->td_proc, newval);
1533 			return (copyout(&newval, PTRIN(args->optval), len));
1534 			/* NOTREACHED */
1535 		default:
1536 			break;
1537 		}
1538 		break;
1539 	case IPPROTO_IP:
1540 		name = linux_to_bsd_ip_sockopt(args->optname);
1541 		break;
1542 	case IPPROTO_IPV6:
1543 		name = linux_to_bsd_ip6_sockopt(args->optname);
1544 		break;
1545 	case IPPROTO_TCP:
1546 		name = linux_to_bsd_tcp_sockopt(args->optname);
1547 		break;
1548 	default:
1549 		name = -1;
1550 		break;
1551 	}
1552 	if (name == -1) {
1553 		linux_msg(curthread,
1554 		    "unsupported getsockopt level %d optname %d",
1555 		    args->level, args->optname);
1556 		return (EINVAL);
1557 	}
1558 
1559 	if (name == IPV6_NEXTHOP) {
1560 		error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1561                 if (error != 0)
1562                         return (error);
1563 		sa = malloc(len, M_SONAME, M_WAITOK);
1564 
1565 		error = kern_getsockopt(td, args->s, level,
1566 		    name, sa, UIO_SYSSPACE, &len);
1567 		if (error != 0)
1568 			goto out;
1569 
1570 		error = bsd_to_linux_sockaddr(sa, &lsa, len);
1571 		if (error == 0)
1572 			error = copyout(lsa, PTRIN(args->optval), len);
1573 		free(lsa, M_SONAME);
1574 		if (error == 0)
1575 			error = copyout(&len, PTRIN(args->optlen),
1576 			    sizeof(len));
1577 out:
1578 		free(sa, M_SONAME);
1579 	} else {
1580 		if (args->optval) {
1581 			error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1582 			if (error != 0)
1583 				return (error);
1584 		}
1585 		error = kern_getsockopt(td, args->s, level,
1586 		    name, PTRIN(args->optval), UIO_USERSPACE, &len);
1587 		if (error == 0)
1588 			error = copyout(&len, PTRIN(args->optlen),
1589 			    sizeof(len));
1590 	}
1591 
1592 	return (error);
1593 }
1594 
1595 static int
1596 linux_sendfile_common(struct thread *td, l_int out, l_int in,
1597     l_loff_t *offset, l_size_t count)
1598 {
1599 	off_t bytes_read;
1600 	int error;
1601 	l_loff_t current_offset;
1602 	struct file *fp;
1603 
1604 	AUDIT_ARG_FD(in);
1605 	error = fget_read(td, in, &cap_pread_rights, &fp);
1606 	if (error != 0)
1607 		return (error);
1608 
1609 	if (offset != NULL) {
1610 		current_offset = *offset;
1611 	} else {
1612 		error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1613 		    fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE;
1614 		if (error != 0)
1615 			goto drop;
1616 		current_offset = td->td_uretoff.tdu_off;
1617 	}
1618 
1619 	bytes_read = 0;
1620 
1621 	/* Linux cannot have 0 count. */
1622 	if (count <= 0 || current_offset < 0) {
1623 		error = EINVAL;
1624 		goto drop;
1625 	}
1626 
1627 	error = fo_sendfile(fp, out, NULL, NULL, current_offset, count,
1628 	    &bytes_read, 0, td);
1629 	if (error != 0)
1630 		goto drop;
1631 	current_offset += bytes_read;
1632 
1633 	if (offset != NULL) {
1634 		*offset = current_offset;
1635 	} else {
1636 		error = fo_seek(fp, current_offset, SEEK_SET, td);
1637 		if (error != 0)
1638 			goto drop;
1639 	}
1640 
1641 	td->td_retval[0] = (ssize_t)bytes_read;
1642 drop:
1643 	fdrop(fp, td);
1644 	return (error);
1645 }
1646 
1647 int
1648 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg)
1649 {
1650 	/*
1651 	 * Differences between FreeBSD and Linux sendfile:
1652 	 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to
1653 	 *   mean send the whole file.)  In linux_sendfile given fds are still
1654 	 *   checked for validity when the count is 0.
1655 	 * - Linux can send to any fd whereas FreeBSD only supports sockets.
1656 	 *   The same restriction follows for linux_sendfile.
1657 	 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr.
1658 	 * - Linux takes an offset pointer and updates it to the read location.
1659 	 *   FreeBSD takes in an offset and a 'bytes read' parameter which is
1660 	 *   only filled if it isn't NULL.  We use this parameter to update the
1661 	 *   offset pointer if it exists.
1662 	 * - Linux sendfile returns bytes read on success while FreeBSD
1663 	 *   returns 0.  We use the 'bytes read' parameter to get this value.
1664 	 */
1665 
1666 	l_loff_t offset64;
1667 	l_long offset;
1668 	int ret;
1669 	int error;
1670 
1671 	if (arg->offset != NULL) {
1672 		error = copyin(arg->offset, &offset, sizeof(offset));
1673 		if (error != 0)
1674 			return (error);
1675 		offset64 = (l_loff_t)offset;
1676 	}
1677 
1678 	ret = linux_sendfile_common(td, arg->out, arg->in,
1679 	    arg->offset != NULL ? &offset64 : NULL, arg->count);
1680 
1681 	if (arg->offset != NULL) {
1682 #if defined(__i386__) || defined(__arm__) || \
1683     (defined(__amd64__) && defined(COMPAT_LINUX32))
1684 		if (offset64 > INT32_MAX)
1685 			return (EOVERFLOW);
1686 #endif
1687 		offset = (l_long)offset64;
1688 		error = copyout(&offset, arg->offset, sizeof(offset));
1689 		if (error != 0)
1690 			return (error);
1691 	}
1692 
1693 	return (ret);
1694 }
1695 
1696 #if defined(__i386__) || defined(__arm__) || \
1697     (defined(__amd64__) && defined(COMPAT_LINUX32))
1698 
1699 int
1700 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg)
1701 {
1702 	l_loff_t offset;
1703 	int ret;
1704 	int error;
1705 
1706 	if (arg->offset != NULL) {
1707 		error = copyin(arg->offset, &offset, sizeof(offset));
1708 		if (error != 0)
1709 			return (error);
1710 	}
1711 
1712 	ret = linux_sendfile_common(td, arg->out, arg->in,
1713 		arg->offset != NULL ? &offset : NULL, arg->count);
1714 
1715 	if (arg->offset != NULL) {
1716 		error = copyout(&offset, arg->offset, sizeof(offset));
1717 		if (error != 0)
1718 			return (error);
1719 	}
1720 
1721 	return (ret);
1722 }
1723 
1724 /* Argument list sizes for linux_socketcall */
1725 static const unsigned char lxs_args_cnt[] = {
1726 	0 /* unused*/,		3 /* socket */,
1727 	3 /* bind */,		3 /* connect */,
1728 	2 /* listen */,		3 /* accept */,
1729 	3 /* getsockname */,	3 /* getpeername */,
1730 	4 /* socketpair */,	4 /* send */,
1731 	4 /* recv */,		6 /* sendto */,
1732 	6 /* recvfrom */,	2 /* shutdown */,
1733 	5 /* setsockopt */,	5 /* getsockopt */,
1734 	3 /* sendmsg */,	3 /* recvmsg */,
1735 	4 /* accept4 */,	5 /* recvmmsg */,
1736 	4 /* sendmmsg */,	4 /* sendfile */
1737 };
1738 #define	LINUX_ARGS_CNT		(nitems(lxs_args_cnt) - 1)
1739 #define	LINUX_ARG_SIZE(x)	(lxs_args_cnt[x] * sizeof(l_ulong))
1740 
1741 int
1742 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1743 {
1744 	l_ulong a[6];
1745 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1746 	register_t l_args[6];
1747 #endif
1748 	void *arg;
1749 	int error;
1750 
1751 	if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
1752 		return (EINVAL);
1753 	error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
1754 	if (error != 0)
1755 		return (error);
1756 
1757 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1758 	for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
1759 		l_args[i] = a[i];
1760 	arg = l_args;
1761 #else
1762 	arg = a;
1763 #endif
1764 	switch (args->what) {
1765 	case LINUX_SOCKET:
1766 		return (linux_socket(td, arg));
1767 	case LINUX_BIND:
1768 		return (linux_bind(td, arg));
1769 	case LINUX_CONNECT:
1770 		return (linux_connect(td, arg));
1771 	case LINUX_LISTEN:
1772 		return (linux_listen(td, arg));
1773 	case LINUX_ACCEPT:
1774 		return (linux_accept(td, arg));
1775 	case LINUX_GETSOCKNAME:
1776 		return (linux_getsockname(td, arg));
1777 	case LINUX_GETPEERNAME:
1778 		return (linux_getpeername(td, arg));
1779 	case LINUX_SOCKETPAIR:
1780 		return (linux_socketpair(td, arg));
1781 	case LINUX_SEND:
1782 		return (linux_send(td, arg));
1783 	case LINUX_RECV:
1784 		return (linux_recv(td, arg));
1785 	case LINUX_SENDTO:
1786 		return (linux_sendto(td, arg));
1787 	case LINUX_RECVFROM:
1788 		return (linux_recvfrom(td, arg));
1789 	case LINUX_SHUTDOWN:
1790 		return (linux_shutdown(td, arg));
1791 	case LINUX_SETSOCKOPT:
1792 		return (linux_setsockopt(td, arg));
1793 	case LINUX_GETSOCKOPT:
1794 		return (linux_getsockopt(td, arg));
1795 	case LINUX_SENDMSG:
1796 		return (linux_sendmsg(td, arg));
1797 	case LINUX_RECVMSG:
1798 		return (linux_recvmsg(td, arg));
1799 	case LINUX_ACCEPT4:
1800 		return (linux_accept4(td, arg));
1801 	case LINUX_RECVMMSG:
1802 		return (linux_recvmmsg(td, arg));
1803 	case LINUX_SENDMMSG:
1804 		return (linux_sendmmsg(td, arg));
1805 	case LINUX_SENDFILE:
1806 		return (linux_sendfile(td, arg));
1807 	}
1808 
1809 	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1810 	return (ENOSYS);
1811 }
1812 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */
1813