xref: /freebsd/sys/compat/linux/linux_socket.c (revision 81ad6265)
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_emul.h>
80 #include <compat/linux/linux_file.h>
81 #include <compat/linux/linux_mib.h>
82 #include <compat/linux/linux_socket.h>
83 #include <compat/linux/linux_timer.h>
84 #include <compat/linux/linux_util.h>
85 
86 #define	SECURITY_CONTEXT_STRING	"unconfined"
87 
88 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
89 					l_uint);
90 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
91 					l_uint, struct msghdr *);
92 static int linux_set_socket_flags(int, int *);
93 
94 #define	SOL_NETLINK	270
95 
96 static int
97 linux_to_bsd_sockopt_level(int level)
98 {
99 
100 	if (level == LINUX_SOL_SOCKET)
101 		return (SOL_SOCKET);
102 	/* Remaining values are RFC-defined protocol numbers. */
103 	return (level);
104 }
105 
106 static int
107 bsd_to_linux_sockopt_level(int level)
108 {
109 
110 	if (level == SOL_SOCKET)
111 		return (LINUX_SOL_SOCKET);
112 	return (level);
113 }
114 
115 static int
116 linux_to_bsd_ip_sockopt(int opt)
117 {
118 
119 	switch (opt) {
120 	/* known and translated sockopts */
121 	case LINUX_IP_TOS:
122 		return (IP_TOS);
123 	case LINUX_IP_TTL:
124 		return (IP_TTL);
125 	case LINUX_IP_HDRINCL:
126 		return (IP_HDRINCL);
127 	case LINUX_IP_OPTIONS:
128 		return (IP_OPTIONS);
129 	case LINUX_IP_RECVOPTS:
130 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVOPTS");
131 		return (IP_RECVOPTS);
132 	case LINUX_IP_RETOPTS:
133 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_REETOPTS");
134 		return (IP_RETOPTS);
135 	case LINUX_IP_RECVTTL:
136 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTTL");
137 		return (IP_RECVTTL);
138 	case LINUX_IP_RECVTOS:
139 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTOS");
140 		return (IP_RECVTOS);
141 	case LINUX_IP_FREEBIND:
142 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_FREEBIND");
143 		return (IP_BINDANY);
144 	case LINUX_IP_IPSEC_POLICY:
145 		/* we have this option, but not documented in ip(4) manpage */
146 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_IPSEC_POLICY");
147 		return (IP_IPSEC_POLICY);
148 	case LINUX_IP_MINTTL:
149 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MINTTL");
150 		return (IP_MINTTL);
151 	case LINUX_IP_MULTICAST_IF:
152 		return (IP_MULTICAST_IF);
153 	case LINUX_IP_MULTICAST_TTL:
154 		return (IP_MULTICAST_TTL);
155 	case LINUX_IP_MULTICAST_LOOP:
156 		return (IP_MULTICAST_LOOP);
157 	case LINUX_IP_ADD_MEMBERSHIP:
158 		return (IP_ADD_MEMBERSHIP);
159 	case LINUX_IP_DROP_MEMBERSHIP:
160 		return (IP_DROP_MEMBERSHIP);
161 	case LINUX_IP_UNBLOCK_SOURCE:
162 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_UNBLOCK_SOURCE");
163 		return (IP_UNBLOCK_SOURCE);
164 	case LINUX_IP_BLOCK_SOURCE:
165 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_BLOCK_SOURCE");
166 		return (IP_BLOCK_SOURCE);
167 	case LINUX_IP_ADD_SOURCE_MEMBERSHIP:
168 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_ADD_SOURCE_MEMBERSHIP");
169 		return (IP_ADD_SOURCE_MEMBERSHIP);
170 	case LINUX_IP_DROP_SOURCE_MEMBERSHIP:
171 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_DROP_SOURCE_MEMBERSHIP");
172 		return (IP_DROP_SOURCE_MEMBERSHIP);
173 	case LINUX_MCAST_JOIN_GROUP:
174 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_GROUP");
175 		return (MCAST_JOIN_GROUP);
176 	case LINUX_MCAST_LEAVE_GROUP:
177 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_GROUP");
178 		return (MCAST_LEAVE_GROUP);
179 	case LINUX_MCAST_JOIN_SOURCE_GROUP:
180 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_SOURCE_GROUP");
181 		return (MCAST_JOIN_SOURCE_GROUP);
182 	case LINUX_MCAST_LEAVE_SOURCE_GROUP:
183 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_SOURCE_GROUP");
184 		return (MCAST_LEAVE_SOURCE_GROUP);
185 	case LINUX_IP_RECVORIGDSTADDR:
186 		return (IP_RECVORIGDSTADDR);
187 
188 	/* known but not implemented sockopts */
189 	case LINUX_IP_ROUTER_ALERT:
190 		LINUX_RATELIMIT_MSG_OPT1(
191 		    "unsupported IPv4 socket option IP_ROUTER_ALERT (%d), you can not do user-space routing from linux programs",
192 		    opt);
193 		return (-2);
194 	case LINUX_IP_PKTINFO:
195 		LINUX_RATELIMIT_MSG_OPT1(
196 		    "unsupported IPv4 socket option IP_PKTINFO (%d), you can not get extended packet info for datagram sockets in linux programs",
197 		    opt);
198 		return (-2);
199 	case LINUX_IP_PKTOPTIONS:
200 		LINUX_RATELIMIT_MSG_OPT1(
201 		    "unsupported IPv4 socket option IP_PKTOPTIONS (%d)",
202 		    opt);
203 		return (-2);
204 	case LINUX_IP_MTU_DISCOVER:
205 		LINUX_RATELIMIT_MSG_OPT1(
206 		    "unsupported IPv4 socket option IP_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery",
207 		    opt);
208 		return (-2);
209 	case LINUX_IP_RECVERR:
210 		/* needed by steam */
211 		LINUX_RATELIMIT_MSG_OPT1(
212 		    "unsupported IPv4 socket option IP_RECVERR (%d), you can not get extended reliability info in linux programs",
213 		    opt);
214 		return (-2);
215 	case LINUX_IP_MTU:
216 		LINUX_RATELIMIT_MSG_OPT1(
217 		    "unsupported IPv4 socket option IP_MTU (%d), your linux program can not control the MTU on this socket",
218 		    opt);
219 		return (-2);
220 	case LINUX_IP_XFRM_POLICY:
221 		LINUX_RATELIMIT_MSG_OPT1(
222 		    "unsupported IPv4 socket option IP_XFRM_POLICY (%d)",
223 		    opt);
224 		return (-2);
225 	case LINUX_IP_PASSSEC:
226 		/* needed by steam */
227 		LINUX_RATELIMIT_MSG_OPT1(
228 		    "unsupported IPv4 socket option IP_PASSSEC (%d), you can not get IPSEC related credential information associated with this socket in linux programs -- if you do not use IPSEC, you can ignore this",
229 		    opt);
230 		return (-2);
231 	case LINUX_IP_TRANSPARENT:
232 		/* IP_BINDANY or more? */
233 		LINUX_RATELIMIT_MSG_OPT1(
234 		    "unsupported IPv4 socket option IP_TRANSPARENT (%d), you can not enable transparent proxying in linux programs -- note, IP_FREEBIND is supported, no idea if the FreeBSD IP_BINDANY is equivalent to the Linux IP_TRANSPARENT or not, any info is welcome",
235 		    opt);
236 		return (-2);
237 	case LINUX_IP_NODEFRAG:
238 		LINUX_RATELIMIT_MSG_OPT1(
239 		    "unsupported IPv4 socket option IP_NODEFRAG (%d)",
240 		    opt);
241 		return (-2);
242 	case LINUX_IP_CHECKSUM:
243 		LINUX_RATELIMIT_MSG_OPT1(
244 		    "unsupported IPv4 socket option IP_CHECKSUM (%d)",
245 		    opt);
246 		return (-2);
247 	case LINUX_IP_BIND_ADDRESS_NO_PORT:
248 		LINUX_RATELIMIT_MSG_OPT1(
249 		    "unsupported IPv4 socket option IP_BIND_ADDRESS_NO_PORT (%d)",
250 		    opt);
251 		return (-2);
252 	case LINUX_IP_RECVFRAGSIZE:
253 		LINUX_RATELIMIT_MSG_OPT1(
254 		    "unsupported IPv4 socket option IP_RECVFRAGSIZE (%d)",
255 		    opt);
256 		return (-2);
257 	case LINUX_MCAST_MSFILTER:
258 		LINUX_RATELIMIT_MSG_OPT1(
259 		    "unsupported IPv4 socket option IP_MCAST_MSFILTER (%d)",
260 		    opt);
261 		return (-2);
262 	case LINUX_IP_MULTICAST_ALL:
263 		LINUX_RATELIMIT_MSG_OPT1(
264 		    "unsupported IPv4 socket option IP_MULTICAST_ALL (%d), your linux program will not see all multicast groups joined by the entire system, only those the program joined itself on this socket",
265 		    opt);
266 		return (-2);
267 	case LINUX_IP_UNICAST_IF:
268 		LINUX_RATELIMIT_MSG_OPT1(
269 		    "unsupported IPv4 socket option IP_UNICAST_IF (%d)",
270 		    opt);
271 		return (-2);
272 
273 	/* unknown sockopts */
274 	default:
275 		return (-1);
276 	}
277 }
278 
279 static int
280 linux_to_bsd_ip6_sockopt(int opt)
281 {
282 
283 	switch (opt) {
284 	/* known and translated sockopts */
285 	case LINUX_IPV6_2292PKTINFO:
286 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTINFO");
287 		return (IPV6_2292PKTINFO);
288 	case LINUX_IPV6_2292HOPOPTS:
289 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPOPTS");
290 		return (IPV6_2292HOPOPTS);
291 	case LINUX_IPV6_2292DSTOPTS:
292 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292DSTOPTS");
293 		return (IPV6_2292DSTOPTS);
294 	case LINUX_IPV6_2292RTHDR:
295 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292RTHDR");
296 		return (IPV6_2292RTHDR);
297 	case LINUX_IPV6_2292PKTOPTIONS:
298 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTOPTIONS");
299 		return (IPV6_2292PKTOPTIONS);
300 	case LINUX_IPV6_CHECKSUM:
301 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_CHECKSUM");
302 		return (IPV6_CHECKSUM);
303 	case LINUX_IPV6_2292HOPLIMIT:
304 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPLIMIT");
305 		return (IPV6_2292HOPLIMIT);
306 	case LINUX_IPV6_NEXTHOP:
307 		return (IPV6_NEXTHOP);
308 	case LINUX_IPV6_UNICAST_HOPS:
309 		return (IPV6_UNICAST_HOPS);
310 	case LINUX_IPV6_MULTICAST_IF:
311 		return (IPV6_MULTICAST_IF);
312 	case LINUX_IPV6_MULTICAST_HOPS:
313 		return (IPV6_MULTICAST_HOPS);
314 	case LINUX_IPV6_MULTICAST_LOOP:
315 		return (IPV6_MULTICAST_LOOP);
316 	case LINUX_IPV6_ADD_MEMBERSHIP:
317 		return (IPV6_JOIN_GROUP);
318 	case LINUX_IPV6_DROP_MEMBERSHIP:
319 		return (IPV6_LEAVE_GROUP);
320 	case LINUX_IPV6_V6ONLY:
321 		return (IPV6_V6ONLY);
322 	case LINUX_IPV6_IPSEC_POLICY:
323 		/* we have this option, but not documented in ip6(4) manpage */
324 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_IPSEC_POLICY");
325 		return (IPV6_IPSEC_POLICY);
326 	case LINUX_MCAST_JOIN_GROUP:
327 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_JOIN_GROUP");
328 		return (IPV6_JOIN_GROUP);
329 	case LINUX_MCAST_LEAVE_GROUP:
330 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_LEAVE_GROUP");
331 		return (IPV6_LEAVE_GROUP);
332 	case LINUX_IPV6_RECVPKTINFO:
333 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPKTINFO");
334 		return (IPV6_RECVPKTINFO);
335 	case LINUX_IPV6_PKTINFO:
336 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PKTINFO");
337 		return (IPV6_PKTINFO);
338 	case LINUX_IPV6_RECVHOPLIMIT:
339 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPLIMIT");
340 		return (IPV6_RECVHOPLIMIT);
341 	case LINUX_IPV6_HOPLIMIT:
342 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPLIMIT");
343 		return (IPV6_HOPLIMIT);
344 	case LINUX_IPV6_RECVHOPOPTS:
345 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPOPTS");
346 		return (IPV6_RECVHOPOPTS);
347 	case LINUX_IPV6_HOPOPTS:
348 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPOPTS");
349 		return (IPV6_HOPOPTS);
350 	case LINUX_IPV6_RTHDRDSTOPTS:
351 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDRDSTOPTS");
352 		return (IPV6_RTHDRDSTOPTS);
353 	case LINUX_IPV6_RECVRTHDR:
354 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVRTHDR");
355 		return (IPV6_RECVRTHDR);
356 	case LINUX_IPV6_RTHDR:
357 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDR");
358 		return (IPV6_RTHDR);
359 	case LINUX_IPV6_RECVDSTOPTS:
360 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVDSTOPTS");
361 		return (IPV6_RECVDSTOPTS);
362 	case LINUX_IPV6_DSTOPTS:
363 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_DSTOPTS");
364 		return (IPV6_DSTOPTS);
365 	case LINUX_IPV6_RECVPATHMTU:
366 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPATHMTU");
367 		return (IPV6_RECVPATHMTU);
368 	case LINUX_IPV6_PATHMTU:
369 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PATHMTU");
370 		return (IPV6_PATHMTU);
371 	case LINUX_IPV6_DONTFRAG:
372 		return (IPV6_DONTFRAG);
373 	case LINUX_IPV6_AUTOFLOWLABEL:
374 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_AUTOFLOWLABEL");
375 		return (IPV6_AUTOFLOWLABEL);
376 	case LINUX_IPV6_ORIGDSTADDR:
377 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_ORIGDSTADDR");
378 		return (IPV6_ORIGDSTADDR);
379 	case LINUX_IPV6_FREEBIND:
380 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_FREEBIND");
381 		return (IPV6_BINDANY);
382 
383 	/* known but not implemented sockopts */
384 	case LINUX_IPV6_ADDRFORM:
385 		LINUX_RATELIMIT_MSG_OPT1(
386 		    "unsupported IPv6 socket option IPV6_ADDRFORM (%d), you linux program can not convert the socket to IPv4",
387 		    opt);
388 		return (-2);
389 	case LINUX_IPV6_AUTHHDR:
390 		LINUX_RATELIMIT_MSG_OPT1(
391 		    "unsupported IPv6 socket option IPV6_AUTHHDR (%d), your linux program can not get the authentication header info of IPv6 packets",
392 		    opt);
393 		return (-2);
394 	case LINUX_IPV6_FLOWINFO:
395 		LINUX_RATELIMIT_MSG_OPT1(
396 		    "unsupported IPv6 socket option IPV6_FLOWINFO (%d), your linux program can not get the flowid of IPv6 packets",
397 		    opt);
398 		return (-2);
399 	case LINUX_IPV6_ROUTER_ALERT:
400 		LINUX_RATELIMIT_MSG_OPT1(
401 		    "unsupported IPv6 socket option IPV6_ROUTER_ALERT (%d), you can not do user-space routing from linux programs",
402 		    opt);
403 		return (-2);
404 	case LINUX_IPV6_MTU_DISCOVER:
405 		LINUX_RATELIMIT_MSG_OPT1(
406 		    "unsupported IPv6 socket option IPV6_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery",
407 		    opt);
408 		return (-2);
409 	case LINUX_IPV6_MTU:
410 		LINUX_RATELIMIT_MSG_OPT1(
411 		    "unsupported IPv6 socket option IPV6_MTU (%d), your linux program can not control the MTU on this socket",
412 		    opt);
413 		return (-2);
414 	case LINUX_IPV6_JOIN_ANYCAST:
415 		LINUX_RATELIMIT_MSG_OPT1(
416 		    "unsupported IPv6 socket option IPV6_JOIN_ANYCAST (%d)",
417 		    opt);
418 		return (-2);
419 	case LINUX_IPV6_LEAVE_ANYCAST:
420 		LINUX_RATELIMIT_MSG_OPT1(
421 		    "unsupported IPv6 socket option IPV6_LEAVE_ANYCAST (%d)",
422 		    opt);
423 		return (-2);
424 	case LINUX_IPV6_MULTICAST_ALL:
425 		LINUX_RATELIMIT_MSG_OPT1(
426 		    "unsupported IPv6 socket option IPV6_MULTICAST_ALL (%d)",
427 		    opt);
428 		return (-2);
429 	case LINUX_IPV6_ROUTER_ALERT_ISOLATE:
430 		LINUX_RATELIMIT_MSG_OPT1(
431 		    "unsupported IPv6 socket option IPV6_ROUTER_ALERT_ISOLATE (%d)",
432 		    opt);
433 		return (-2);
434 	case LINUX_IPV6_FLOWLABEL_MGR:
435 		LINUX_RATELIMIT_MSG_OPT1(
436 		    "unsupported IPv6 socket option IPV6_FLOWLABEL_MGR (%d)",
437 		    opt);
438 		return (-2);
439 	case LINUX_IPV6_FLOWINFO_SEND:
440 		LINUX_RATELIMIT_MSG_OPT1(
441 		    "unsupported IPv6 socket option IPV6_FLOWINFO_SEND (%d)",
442 		    opt);
443 		return (-2);
444 	case LINUX_IPV6_XFRM_POLICY:
445 		LINUX_RATELIMIT_MSG_OPT1(
446 		    "unsupported IPv6 socket option IPV6_XFRM_POLICY (%d)",
447 		    opt);
448 		return (-2);
449 	case LINUX_IPV6_HDRINCL:
450 		LINUX_RATELIMIT_MSG_OPT1(
451 		    "unsupported IPv6 socket option IPV6_HDRINCL (%d)",
452 		    opt);
453 		return (-2);
454 	case LINUX_MCAST_BLOCK_SOURCE:
455 		LINUX_RATELIMIT_MSG_OPT1(
456 		    "unsupported IPv6 socket option MCAST_BLOCK_SOURCE (%d), your linux program may see more multicast stuff than it wants",
457 		    opt);
458 		return (-2);
459 	case LINUX_MCAST_UNBLOCK_SOURCE:
460 		LINUX_RATELIMIT_MSG_OPT1(
461 		    "unsupported IPv6 socket option MCAST_UNBLOCK_SOURCE (%d), your linux program may not see all the multicast stuff it wants",
462 		    opt);
463 		return (-2);
464 	case LINUX_MCAST_JOIN_SOURCE_GROUP:
465 		LINUX_RATELIMIT_MSG_OPT1(
466 		    "unsupported IPv6 socket option MCAST_JOIN_SOURCE_GROUP (%d), your linux program is not able to join a multicast source group",
467 		    opt);
468 		return (-2);
469 	case LINUX_MCAST_LEAVE_SOURCE_GROUP:
470 		LINUX_RATELIMIT_MSG_OPT1(
471 		    "unsupported IPv6 socket option MCAST_LEAVE_SOURCE_GROUP (%d), your linux program is not able to leave a multicast source group -- but it was also not able to join one, so no issue",
472 		    opt);
473 		return (-2);
474 	case LINUX_MCAST_MSFILTER:
475 		LINUX_RATELIMIT_MSG_OPT1(
476 		    "unsupported IPv6 socket option MCAST_MSFILTER (%d), your linux program can not manipulate the multicast filter, it may see more multicast data than it wants to see",
477 		    opt);
478 		return (-2);
479 	case LINUX_IPV6_ADDR_PREFERENCES:
480 		LINUX_RATELIMIT_MSG_OPT1(
481 		    "unsupported IPv6 socket option IPV6_ADDR_PREFERENCES (%d)",
482 		    opt);
483 		return (-2);
484 	case LINUX_IPV6_MINHOPCOUNT:
485 		LINUX_RATELIMIT_MSG_OPT1(
486 		    "unsupported IPv6 socket option IPV6_MINHOPCOUNT (%d)",
487 		    opt);
488 		return (-2);
489 	case LINUX_IPV6_TRANSPARENT:
490 		/* IP_BINDANY or more? */
491 		LINUX_RATELIMIT_MSG_OPT1(
492 		    "unsupported IPv6 socket option IPV6_TRANSPARENT (%d), you can not enable transparent proxying in linux programs -- note, IP_FREEBIND is supported, no idea if the FreeBSD IP_BINDANY is equivalent to the Linux IP_TRANSPARENT or not, any info is welcome",
493 		    opt);
494 		return (-2);
495 	case LINUX_IPV6_UNICAST_IF:
496 		LINUX_RATELIMIT_MSG_OPT1(
497 		    "unsupported IPv6 socket option IPV6_UNICAST_IF (%d)",
498 		    opt);
499 		return (-2);
500 	case LINUX_IPV6_RECVFRAGSIZE:
501 		LINUX_RATELIMIT_MSG_OPT1(
502 		    "unsupported IPv6 socket option IPV6_RECVFRAGSIZE (%d)",
503 		    opt);
504 		return (-2);
505 
506 	/* unknown sockopts */
507 	default:
508 		return (-1);
509 	}
510 }
511 
512 static int
513 linux_to_bsd_so_sockopt(int opt)
514 {
515 
516 	switch (opt) {
517 	case LINUX_SO_DEBUG:
518 		return (SO_DEBUG);
519 	case LINUX_SO_REUSEADDR:
520 		return (SO_REUSEADDR);
521 	case LINUX_SO_TYPE:
522 		return (SO_TYPE);
523 	case LINUX_SO_ERROR:
524 		return (SO_ERROR);
525 	case LINUX_SO_DONTROUTE:
526 		return (SO_DONTROUTE);
527 	case LINUX_SO_BROADCAST:
528 		return (SO_BROADCAST);
529 	case LINUX_SO_SNDBUF:
530 	case LINUX_SO_SNDBUFFORCE:
531 		return (SO_SNDBUF);
532 	case LINUX_SO_RCVBUF:
533 	case LINUX_SO_RCVBUFFORCE:
534 		return (SO_RCVBUF);
535 	case LINUX_SO_KEEPALIVE:
536 		return (SO_KEEPALIVE);
537 	case LINUX_SO_OOBINLINE:
538 		return (SO_OOBINLINE);
539 	case LINUX_SO_LINGER:
540 		return (SO_LINGER);
541 	case LINUX_SO_REUSEPORT:
542 		return (SO_REUSEPORT_LB);
543 	case LINUX_SO_PASSCRED:
544 		return (LOCAL_CREDS_PERSISTENT);
545 	case LINUX_SO_PEERCRED:
546 		return (LOCAL_PEERCRED);
547 	case LINUX_SO_RCVLOWAT:
548 		return (SO_RCVLOWAT);
549 	case LINUX_SO_SNDLOWAT:
550 		return (SO_SNDLOWAT);
551 	case LINUX_SO_RCVTIMEO:
552 		return (SO_RCVTIMEO);
553 	case LINUX_SO_SNDTIMEO:
554 		return (SO_SNDTIMEO);
555 	case LINUX_SO_TIMESTAMPO:
556 	case LINUX_SO_TIMESTAMPN:
557 		return (SO_TIMESTAMP);
558 	case LINUX_SO_TIMESTAMPNSO:
559 	case LINUX_SO_TIMESTAMPNSN:
560 		return (SO_BINTIME);
561 	case LINUX_SO_ACCEPTCONN:
562 		return (SO_ACCEPTCONN);
563 	case LINUX_SO_PROTOCOL:
564 		return (SO_PROTOCOL);
565 	case LINUX_SO_DOMAIN:
566 		return (SO_DOMAIN);
567 	}
568 	return (-1);
569 }
570 
571 static int
572 linux_to_bsd_tcp_sockopt(int opt)
573 {
574 
575 	switch (opt) {
576 	case LINUX_TCP_NODELAY:
577 		return (TCP_NODELAY);
578 	case LINUX_TCP_MAXSEG:
579 		return (TCP_MAXSEG);
580 	case LINUX_TCP_CORK:
581 		return (TCP_NOPUSH);
582 	case LINUX_TCP_KEEPIDLE:
583 		return (TCP_KEEPIDLE);
584 	case LINUX_TCP_KEEPINTVL:
585 		return (TCP_KEEPINTVL);
586 	case LINUX_TCP_KEEPCNT:
587 		return (TCP_KEEPCNT);
588 	case LINUX_TCP_INFO:
589 		LINUX_RATELIMIT_MSG_OPT1(
590 		    "unsupported TCP socket option TCP_INFO (%d)", opt);
591 		return (-2);
592 	case LINUX_TCP_MD5SIG:
593 		return (TCP_MD5SIG);
594 	}
595 	return (-1);
596 }
597 
598 static int
599 linux_to_bsd_msg_flags(int flags)
600 {
601 	int ret_flags = 0;
602 
603 	if (flags & LINUX_MSG_OOB)
604 		ret_flags |= MSG_OOB;
605 	if (flags & LINUX_MSG_PEEK)
606 		ret_flags |= MSG_PEEK;
607 	if (flags & LINUX_MSG_DONTROUTE)
608 		ret_flags |= MSG_DONTROUTE;
609 	if (flags & LINUX_MSG_CTRUNC)
610 		ret_flags |= MSG_CTRUNC;
611 	if (flags & LINUX_MSG_TRUNC)
612 		ret_flags |= MSG_TRUNC;
613 	if (flags & LINUX_MSG_DONTWAIT)
614 		ret_flags |= MSG_DONTWAIT;
615 	if (flags & LINUX_MSG_EOR)
616 		ret_flags |= MSG_EOR;
617 	if (flags & LINUX_MSG_WAITALL)
618 		ret_flags |= MSG_WAITALL;
619 	if (flags & LINUX_MSG_NOSIGNAL)
620 		ret_flags |= MSG_NOSIGNAL;
621 	if (flags & LINUX_MSG_PROXY)
622 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_PROXY (%d) not handled",
623 		    LINUX_MSG_PROXY);
624 	if (flags & LINUX_MSG_FIN)
625 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_FIN (%d) not handled",
626 		    LINUX_MSG_FIN);
627 	if (flags & LINUX_MSG_SYN)
628 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_SYN (%d) not handled",
629 		    LINUX_MSG_SYN);
630 	if (flags & LINUX_MSG_CONFIRM)
631 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_CONFIRM (%d) not handled",
632 		    LINUX_MSG_CONFIRM);
633 	if (flags & LINUX_MSG_RST)
634 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_RST (%d) not handled",
635 		    LINUX_MSG_RST);
636 	if (flags & LINUX_MSG_ERRQUEUE)
637 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_ERRQUEUE (%d) not handled",
638 		    LINUX_MSG_ERRQUEUE);
639 	return (ret_flags);
640 }
641 
642 static int
643 linux_to_bsd_cmsg_type(int cmsg_type)
644 {
645 
646 	switch (cmsg_type) {
647 	case LINUX_SCM_RIGHTS:
648 		return (SCM_RIGHTS);
649 	case LINUX_SCM_CREDENTIALS:
650 		return (SCM_CREDS);
651 	}
652 	return (-1);
653 }
654 
655 static int
656 bsd_to_linux_ip_cmsg_type(int cmsg_type)
657 {
658 
659 	switch (cmsg_type) {
660 	case IP_RECVORIGDSTADDR:
661 		return (LINUX_IP_RECVORIGDSTADDR);
662 	}
663 	return (-1);
664 }
665 
666 static int
667 bsd_to_linux_cmsg_type(struct proc *p, int cmsg_type, int cmsg_level)
668 {
669 	struct linux_pemuldata *pem;
670 
671 	if (cmsg_level == IPPROTO_IP)
672 		return (bsd_to_linux_ip_cmsg_type(cmsg_type));
673 	if (cmsg_level != SOL_SOCKET)
674 		return (-1);
675 
676 	pem = pem_find(p);
677 
678 	switch (cmsg_type) {
679 	case SCM_RIGHTS:
680 		return (LINUX_SCM_RIGHTS);
681 	case SCM_CREDS:
682 		return (LINUX_SCM_CREDENTIALS);
683 	case SCM_CREDS2:
684 		return (LINUX_SCM_CREDENTIALS);
685 	case SCM_TIMESTAMP:
686 		return (pem->so_timestamp);
687 	case SCM_BINTIME:
688 		return (pem->so_timestampns);
689 	}
690 	return (-1);
691 }
692 
693 static int
694 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
695 {
696 	if (lhdr->msg_controllen > INT_MAX)
697 		return (ENOBUFS);
698 
699 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
700 	bhdr->msg_namelen	= lhdr->msg_namelen;
701 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
702 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
703 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
704 
705 	/*
706 	 * msg_controllen is skipped since BSD and LINUX control messages
707 	 * are potentially different sizes (e.g. the cred structure used
708 	 * by SCM_CREDS is different between the two operating system).
709 	 *
710 	 * The caller can set it (if necessary) after converting all the
711 	 * control messages.
712 	 */
713 
714 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
715 	return (0);
716 }
717 
718 static int
719 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
720 {
721 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
722 	lhdr->msg_namelen	= bhdr->msg_namelen;
723 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
724 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
725 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
726 
727 	/*
728 	 * msg_controllen is skipped since BSD and LINUX control messages
729 	 * are potentially different sizes (e.g. the cred structure used
730 	 * by SCM_CREDS is different between the two operating system).
731 	 *
732 	 * The caller can set it (if necessary) after converting all the
733 	 * control messages.
734 	 */
735 
736 	/* msg_flags skipped */
737 	return (0);
738 }
739 
740 static int
741 linux_set_socket_flags(int lflags, int *flags)
742 {
743 
744 	if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
745 		return (EINVAL);
746 	if (lflags & LINUX_SOCK_NONBLOCK)
747 		*flags |= SOCK_NONBLOCK;
748 	if (lflags & LINUX_SOCK_CLOEXEC)
749 		*flags |= SOCK_CLOEXEC;
750 	return (0);
751 }
752 
753 static int
754 linux_copyout_sockaddr(const struct sockaddr *sa, void *uaddr, size_t len)
755 {
756 	struct l_sockaddr *lsa;
757 	int error;
758 
759 	error = bsd_to_linux_sockaddr(sa, &lsa, len);
760 	if (error != 0)
761 		return (error);
762 
763 	error = copyout(lsa, uaddr, len);
764 	free(lsa, M_LINUX);
765 
766 	return (error);
767 }
768 
769 static int
770 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
771     struct mbuf *control, enum uio_seg segflg)
772 {
773 	struct sockaddr *to;
774 	int error, len;
775 
776 	if (mp->msg_name != NULL) {
777 		len = mp->msg_namelen;
778 		error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len);
779 		if (error != 0)
780 			return (error);
781 		mp->msg_name = to;
782 	} else
783 		to = NULL;
784 
785 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
786 	    segflg);
787 
788 	if (to)
789 		free(to, M_SONAME);
790 	return (error);
791 }
792 
793 /* Return 0 if IP_HDRINCL is set for the given socket. */
794 static int
795 linux_check_hdrincl(struct thread *td, int s)
796 {
797 	int error, optval;
798 	socklen_t size_val;
799 
800 	size_val = sizeof(optval);
801 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
802 	    &optval, UIO_SYSSPACE, &size_val);
803 	if (error != 0)
804 		return (error);
805 
806 	return (optval == 0);
807 }
808 
809 /*
810  * Updated sendto() when IP_HDRINCL is set:
811  * tweak endian-dependent fields in the IP packet.
812  */
813 static int
814 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
815 {
816 /*
817  * linux_ip_copysize defines how many bytes we should copy
818  * from the beginning of the IP packet before we customize it for BSD.
819  * It should include all the fields we modify (ip_len and ip_off).
820  */
821 #define linux_ip_copysize	8
822 
823 	struct ip *packet;
824 	struct msghdr msg;
825 	struct iovec aiov[1];
826 	int error;
827 
828 	/* Check that the packet isn't too big or too small. */
829 	if (linux_args->len < linux_ip_copysize ||
830 	    linux_args->len > IP_MAXPACKET)
831 		return (EINVAL);
832 
833 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
834 
835 	/* Make kernel copy of the packet to be sent */
836 	if ((error = copyin(PTRIN(linux_args->msg), packet,
837 	    linux_args->len)))
838 		goto goout;
839 
840 	/* Convert fields from Linux to BSD raw IP socket format */
841 	packet->ip_len = linux_args->len;
842 	packet->ip_off = ntohs(packet->ip_off);
843 
844 	/* Prepare the msghdr and iovec structures describing the new packet */
845 	msg.msg_name = PTRIN(linux_args->to);
846 	msg.msg_namelen = linux_args->tolen;
847 	msg.msg_iov = aiov;
848 	msg.msg_iovlen = 1;
849 	msg.msg_control = NULL;
850 	msg.msg_flags = 0;
851 	aiov[0].iov_base = (char *)packet;
852 	aiov[0].iov_len = linux_args->len;
853 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
854 	    NULL, UIO_SYSSPACE);
855 goout:
856 	free(packet, M_LINUX);
857 	return (error);
858 }
859 
860 static const char *linux_netlink_names[] = {
861 	[LINUX_NETLINK_ROUTE] = "ROUTE",
862 	[LINUX_NETLINK_SOCK_DIAG] = "SOCK_DIAG",
863 	[LINUX_NETLINK_NFLOG] = "NFLOG",
864 	[LINUX_NETLINK_SELINUX] = "SELINUX",
865 	[LINUX_NETLINK_AUDIT] = "AUDIT",
866 	[LINUX_NETLINK_FIB_LOOKUP] = "FIB_LOOKUP",
867 	[LINUX_NETLINK_NETFILTER] = "NETFILTER",
868 	[LINUX_NETLINK_KOBJECT_UEVENT] = "KOBJECT_UEVENT",
869 };
870 
871 int
872 linux_socket(struct thread *td, struct linux_socket_args *args)
873 {
874 	int domain, retval_socket, type;
875 
876 	type = args->type & LINUX_SOCK_TYPE_MASK;
877 	if (type < 0 || type > LINUX_SOCK_MAX)
878 		return (EINVAL);
879 	retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
880 		&type);
881 	if (retval_socket != 0)
882 		return (retval_socket);
883 	domain = linux_to_bsd_domain(args->domain);
884 	if (domain == -1) {
885 		/* Mask off SOCK_NONBLOCK / CLOEXEC for error messages. */
886 		type = args->type & LINUX_SOCK_TYPE_MASK;
887 		if (args->domain == LINUX_AF_NETLINK &&
888 		    args->protocol == LINUX_NETLINK_AUDIT) {
889 			; /* Do nothing, quietly. */
890 		} else if (args->domain == LINUX_AF_NETLINK) {
891 			const char *nl_name;
892 
893 			if (args->protocol >= 0 &&
894 			    args->protocol < nitems(linux_netlink_names))
895 				nl_name = linux_netlink_names[args->protocol];
896 			else
897 				nl_name = NULL;
898 			if (nl_name != NULL)
899 				linux_msg(curthread,
900 				    "unsupported socket(AF_NETLINK, %d, "
901 				    "NETLINK_%s)", type, nl_name);
902 			else
903 				linux_msg(curthread,
904 				    "unsupported socket(AF_NETLINK, %d, %d)",
905 				    type, args->protocol);
906 		} else {
907 			linux_msg(curthread, "unsupported socket domain %d, "
908 			    "type %d, protocol %d", args->domain, type,
909 			    args->protocol);
910 		}
911 		return (EAFNOSUPPORT);
912 	}
913 
914 	retval_socket = kern_socket(td, domain, type, args->protocol);
915 	if (retval_socket)
916 		return (retval_socket);
917 
918 	if (type == SOCK_RAW
919 	    && (args->protocol == IPPROTO_RAW || args->protocol == 0)
920 	    && domain == PF_INET) {
921 		/* It's a raw IP socket: set the IP_HDRINCL option. */
922 		int hdrincl;
923 
924 		hdrincl = 1;
925 		/* We ignore any error returned by kern_setsockopt() */
926 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
927 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
928 	}
929 #ifdef INET6
930 	/*
931 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
932 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
933 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
934 	 * sysctl value.
935 	 */
936 	if (domain == PF_INET6) {
937 		int v6only;
938 
939 		v6only = 0;
940 		/* We ignore any error returned by setsockopt() */
941 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
942 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
943 	}
944 #endif
945 
946 	return (retval_socket);
947 }
948 
949 int
950 linux_bind(struct thread *td, struct linux_bind_args *args)
951 {
952 	struct sockaddr *sa;
953 	int error;
954 
955 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
956 	    &args->namelen);
957 	if (error != 0)
958 		return (error);
959 
960 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
961 	free(sa, M_SONAME);
962 
963 	/* XXX */
964 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
965 		return (EINVAL);
966 	return (error);
967 }
968 
969 int
970 linux_connect(struct thread *td, struct linux_connect_args *args)
971 {
972 	struct socket *so;
973 	struct sockaddr *sa;
974 	struct file *fp;
975 	int error;
976 
977 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
978 	    &args->namelen);
979 	if (error != 0)
980 		return (error);
981 
982 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
983 	free(sa, M_SONAME);
984 	if (error != EISCONN)
985 		return (error);
986 
987 	/*
988 	 * Linux doesn't return EISCONN the first time it occurs,
989 	 * when on a non-blocking socket. Instead it returns the
990 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
991 	 */
992 	error = getsock(td, args->s, &cap_connect_rights, &fp);
993 	if (error != 0)
994 		return (error);
995 
996 	error = EISCONN;
997 	so = fp->f_data;
998 	if (atomic_load_int(&fp->f_flag) & FNONBLOCK) {
999 		SOCK_LOCK(so);
1000 		if (so->so_emuldata == 0)
1001 			error = so->so_error;
1002 		so->so_emuldata = (void *)1;
1003 		SOCK_UNLOCK(so);
1004 	}
1005 	fdrop(fp, td);
1006 
1007 	return (error);
1008 }
1009 
1010 int
1011 linux_listen(struct thread *td, struct linux_listen_args *args)
1012 {
1013 
1014 	return (kern_listen(td, args->s, args->backlog));
1015 }
1016 
1017 static int
1018 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
1019     l_uintptr_t namelen, int flags)
1020 {
1021 	struct sockaddr *sa;
1022 	struct file *fp, *fp1;
1023 	int bflags, len;
1024 	struct socket *so;
1025 	int error, error1;
1026 
1027 	bflags = 0;
1028 	fp = NULL;
1029 	sa = NULL;
1030 
1031 	error = linux_set_socket_flags(flags, &bflags);
1032 	if (error != 0)
1033 		return (error);
1034 
1035 	if (PTRIN(addr) == NULL) {
1036 		len = 0;
1037 		error = kern_accept4(td, s, NULL, NULL, bflags, NULL);
1038 	} else {
1039 		error = copyin(PTRIN(namelen), &len, sizeof(len));
1040 		if (error != 0)
1041 			return (error);
1042 		if (len < 0)
1043 			return (EINVAL);
1044 		error = kern_accept4(td, s, &sa, &len, bflags, &fp);
1045 	}
1046 
1047 	/*
1048 	 * Translate errno values into ones used by Linux.
1049 	 */
1050 	if (error != 0) {
1051 		/*
1052 		 * XXX. This is wrong, different sockaddr structures
1053 		 * have different sizes.
1054 		 */
1055 		switch (error) {
1056 		case EFAULT:
1057 			if (namelen != sizeof(struct sockaddr_in))
1058 				error = EINVAL;
1059 			break;
1060 		case EINVAL:
1061 			error1 = getsock(td, s, &cap_accept_rights, &fp1);
1062 			if (error1 != 0) {
1063 				error = error1;
1064 				break;
1065 			}
1066 			so = fp1->f_data;
1067 			if (so->so_type == SOCK_DGRAM)
1068 				error = EOPNOTSUPP;
1069 			fdrop(fp1, td);
1070 			break;
1071 		}
1072 		return (error);
1073 	}
1074 
1075 	if (len != 0) {
1076 		error = linux_copyout_sockaddr(sa, PTRIN(addr), len);
1077 		if (error == 0)
1078 			error = copyout(&len, PTRIN(namelen),
1079 			    sizeof(len));
1080 		if (error != 0) {
1081 			fdclose(td, fp, td->td_retval[0]);
1082 			td->td_retval[0] = 0;
1083 		}
1084 	}
1085 	if (fp != NULL)
1086 		fdrop(fp, td);
1087 	free(sa, M_SONAME);
1088 	return (error);
1089 }
1090 
1091 int
1092 linux_accept(struct thread *td, struct linux_accept_args *args)
1093 {
1094 
1095 	return (linux_accept_common(td, args->s, args->addr,
1096 	    args->namelen, 0));
1097 }
1098 
1099 int
1100 linux_accept4(struct thread *td, struct linux_accept4_args *args)
1101 {
1102 
1103 	return (linux_accept_common(td, args->s, args->addr,
1104 	    args->namelen, args->flags));
1105 }
1106 
1107 int
1108 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
1109 {
1110 	struct sockaddr *sa;
1111 	int len, error;
1112 
1113 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1114 	if (error != 0)
1115 		return (error);
1116 
1117 	error = kern_getsockname(td, args->s, &sa, &len);
1118 	if (error != 0)
1119 		return (error);
1120 
1121 	if (len != 0)
1122 		error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
1123 
1124 	free(sa, M_SONAME);
1125 	if (error == 0)
1126 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1127 	return (error);
1128 }
1129 
1130 int
1131 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
1132 {
1133 	struct sockaddr *sa;
1134 	int len, error;
1135 
1136 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1137 	if (error != 0)
1138 		return (error);
1139 	if (len < 0)
1140 		return (EINVAL);
1141 
1142 	error = kern_getpeername(td, args->s, &sa, &len);
1143 	if (error != 0)
1144 		return (error);
1145 
1146 	if (len != 0)
1147 		error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
1148 
1149 	free(sa, M_SONAME);
1150 	if (error == 0)
1151 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1152 	return (error);
1153 }
1154 
1155 int
1156 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
1157 {
1158 	int domain, error, sv[2], type;
1159 
1160 	domain = linux_to_bsd_domain(args->domain);
1161 	if (domain != PF_LOCAL)
1162 		return (EAFNOSUPPORT);
1163 	type = args->type & LINUX_SOCK_TYPE_MASK;
1164 	if (type < 0 || type > LINUX_SOCK_MAX)
1165 		return (EINVAL);
1166 	error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
1167 	    &type);
1168 	if (error != 0)
1169 		return (error);
1170 	if (args->protocol != 0 && args->protocol != PF_UNIX) {
1171 		/*
1172 		 * Use of PF_UNIX as protocol argument is not right,
1173 		 * but Linux does it.
1174 		 * Do not map PF_UNIX as its Linux value is identical
1175 		 * to FreeBSD one.
1176 		 */
1177 		return (EPROTONOSUPPORT);
1178 	}
1179 	error = kern_socketpair(td, domain, type, 0, sv);
1180 	if (error != 0)
1181                 return (error);
1182         error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int));
1183         if (error != 0) {
1184                 (void)kern_close(td, sv[0]);
1185                 (void)kern_close(td, sv[1]);
1186         }
1187 	return (error);
1188 }
1189 
1190 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1191 struct linux_send_args {
1192 	register_t s;
1193 	register_t msg;
1194 	register_t len;
1195 	register_t flags;
1196 };
1197 
1198 static int
1199 linux_send(struct thread *td, struct linux_send_args *args)
1200 {
1201 	struct sendto_args /* {
1202 		int s;
1203 		caddr_t buf;
1204 		int len;
1205 		int flags;
1206 		caddr_t to;
1207 		int tolen;
1208 	} */ bsd_args;
1209 	struct file *fp;
1210 	int error;
1211 
1212 	bsd_args.s = args->s;
1213 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1214 	bsd_args.len = args->len;
1215 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1216 	bsd_args.to = NULL;
1217 	bsd_args.tolen = 0;
1218 	error = sys_sendto(td, &bsd_args);
1219 	if (error == ENOTCONN) {
1220 		/*
1221 		 * Linux doesn't return ENOTCONN for non-blocking sockets.
1222 		 * Instead it returns the EAGAIN.
1223 		 */
1224 		error = getsock(td, args->s, &cap_send_rights, &fp);
1225 		if (error == 0) {
1226 			if (atomic_load_int(&fp->f_flag) & FNONBLOCK)
1227 				error = EAGAIN;
1228 			fdrop(fp, td);
1229 		}
1230 	}
1231 	return (error);
1232 }
1233 
1234 struct linux_recv_args {
1235 	register_t s;
1236 	register_t msg;
1237 	register_t len;
1238 	register_t flags;
1239 };
1240 
1241 static int
1242 linux_recv(struct thread *td, struct linux_recv_args *args)
1243 {
1244 	struct recvfrom_args /* {
1245 		int s;
1246 		caddr_t buf;
1247 		int len;
1248 		int flags;
1249 		struct sockaddr *from;
1250 		socklen_t fromlenaddr;
1251 	} */ bsd_args;
1252 
1253 	bsd_args.s = args->s;
1254 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1255 	bsd_args.len = args->len;
1256 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1257 	bsd_args.from = NULL;
1258 	bsd_args.fromlenaddr = 0;
1259 	return (sys_recvfrom(td, &bsd_args));
1260 }
1261 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1262 
1263 int
1264 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1265 {
1266 	struct msghdr msg;
1267 	struct iovec aiov;
1268 	struct socket *so;
1269 	struct file *fp;
1270 	int error;
1271 
1272 	if (linux_check_hdrincl(td, args->s) == 0)
1273 		/* IP_HDRINCL set, tweak the packet before sending */
1274 		return (linux_sendto_hdrincl(td, args));
1275 
1276 	bzero(&msg, sizeof(msg));
1277 	error = getsock(td, args->s, &cap_send_connect_rights, &fp);
1278 	if (error != 0)
1279 		return (error);
1280 	so = fp->f_data;
1281 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0) {
1282 		msg.msg_name = PTRIN(args->to);
1283 		msg.msg_namelen = args->tolen;
1284 	}
1285 	msg.msg_iov = &aiov;
1286 	msg.msg_iovlen = 1;
1287 	aiov.iov_base = PTRIN(args->msg);
1288 	aiov.iov_len = args->len;
1289 	fdrop(fp, td);
1290 	return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1291 	    UIO_USERSPACE));
1292 }
1293 
1294 int
1295 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1296 {
1297 	struct sockaddr *sa;
1298 	struct msghdr msg;
1299 	struct iovec aiov;
1300 	int error, fromlen;
1301 
1302 	if (PTRIN(args->fromlen) != NULL) {
1303 		error = copyin(PTRIN(args->fromlen), &fromlen,
1304 		    sizeof(fromlen));
1305 		if (error != 0)
1306 			return (error);
1307 		if (fromlen < 0)
1308 			return (EINVAL);
1309 		fromlen = min(fromlen, SOCK_MAXADDRLEN);
1310 		sa = malloc(fromlen, M_SONAME, M_WAITOK);
1311 	} else {
1312 		fromlen = 0;
1313 		sa = NULL;
1314 	}
1315 
1316 	msg.msg_name = sa;
1317 	msg.msg_namelen = fromlen;
1318 	msg.msg_iov = &aiov;
1319 	msg.msg_iovlen = 1;
1320 	aiov.iov_base = PTRIN(args->buf);
1321 	aiov.iov_len = args->len;
1322 	msg.msg_control = 0;
1323 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1324 
1325 	error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL);
1326 	if (error != 0)
1327 		goto out;
1328 
1329 	/*
1330 	 * XXX. Seems that FreeBSD is different from Linux here. Linux
1331 	 * fill source address if underlying protocol provides it, while
1332 	 * FreeBSD fill it if underlying protocol is not connection-oriented.
1333 	 * So, kern_recvit() set msg.msg_namelen to 0 if protocol pr_flags
1334 	 * does not contains PR_ADDR flag.
1335 	 */
1336 	if (PTRIN(args->from) != NULL && msg.msg_namelen != 0)
1337 		error = linux_copyout_sockaddr(sa, PTRIN(args->from),
1338 		    msg.msg_namelen);
1339 
1340 	if (error == 0 && PTRIN(args->fromlen) != NULL)
1341 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1342 		    sizeof(msg.msg_namelen));
1343 out:
1344 	free(sa, M_SONAME);
1345 	return (error);
1346 }
1347 
1348 static int
1349 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1350     l_uint flags)
1351 {
1352 	struct cmsghdr *cmsg;
1353 	struct mbuf *control;
1354 	struct msghdr msg;
1355 	struct l_cmsghdr linux_cmsg;
1356 	struct l_cmsghdr *ptr_cmsg;
1357 	struct l_msghdr linux_msghdr;
1358 	struct iovec *iov;
1359 	socklen_t datalen;
1360 	struct sockaddr *sa;
1361 	struct socket *so;
1362 	sa_family_t sa_family;
1363 	struct file *fp;
1364 	void *data;
1365 	l_size_t len;
1366 	l_size_t clen;
1367 	int error;
1368 
1369 	error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
1370 	if (error != 0)
1371 		return (error);
1372 
1373 	/*
1374 	 * Some Linux applications (ping) define a non-NULL control data
1375 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1376 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1377 	 * order to handle this case.  This should be checked, but allows the
1378 	 * Linux ping to work.
1379 	 */
1380 	if (PTRIN(linux_msghdr.msg_control) != NULL &&
1381 	    linux_msghdr.msg_controllen == 0)
1382 		linux_msghdr.msg_control = PTROUT(NULL);
1383 
1384 	error = linux_to_bsd_msghdr(&msg, &linux_msghdr);
1385 	if (error != 0)
1386 		return (error);
1387 
1388 #ifdef COMPAT_LINUX32
1389 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1390 	    &iov, EMSGSIZE);
1391 #else
1392 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1393 #endif
1394 	if (error != 0)
1395 		return (error);
1396 
1397 	control = NULL;
1398 
1399 	error = kern_getsockname(td, s, &sa, &datalen);
1400 	if (error != 0)
1401 		goto bad;
1402 	sa_family = sa->sa_family;
1403 	free(sa, M_SONAME);
1404 
1405 	if (flags & LINUX_MSG_OOB) {
1406 		error = EOPNOTSUPP;
1407 		if (sa_family == AF_UNIX)
1408 			goto bad;
1409 
1410 		error = getsock(td, s, &cap_send_rights, &fp);
1411 		if (error != 0)
1412 			goto bad;
1413 		so = fp->f_data;
1414 		if (so->so_type != SOCK_STREAM)
1415 			error = EOPNOTSUPP;
1416 		fdrop(fp, td);
1417 		if (error != 0)
1418 			goto bad;
1419 	}
1420 
1421 	if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) {
1422 		error = ENOBUFS;
1423 		control = m_get(M_WAITOK, MT_CONTROL);
1424 		MCLGET(control, M_WAITOK);
1425 		data = mtod(control, void *);
1426 		datalen = 0;
1427 
1428 		ptr_cmsg = PTRIN(linux_msghdr.msg_control);
1429 		clen = linux_msghdr.msg_controllen;
1430 		do {
1431 			error = copyin(ptr_cmsg, &linux_cmsg,
1432 			    sizeof(struct l_cmsghdr));
1433 			if (error != 0)
1434 				goto bad;
1435 
1436 			error = EINVAL;
1437 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1438 			    linux_cmsg.cmsg_len > clen)
1439 				goto bad;
1440 
1441 			if (datalen + CMSG_HDRSZ > MCLBYTES)
1442 				goto bad;
1443 
1444 			/*
1445 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1446 			 * so return EINVAL in any other cmsg_type
1447 			 */
1448 			cmsg = data;
1449 			cmsg->cmsg_type =
1450 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1451 			cmsg->cmsg_level =
1452 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1453 			if (cmsg->cmsg_type == -1
1454 			    || cmsg->cmsg_level != SOL_SOCKET) {
1455 				linux_msg(curthread,
1456 				    "unsupported sendmsg cmsg level %d type %d",
1457 				    linux_cmsg.cmsg_level, linux_cmsg.cmsg_type);
1458 				goto bad;
1459 			}
1460 
1461 			/*
1462 			 * Some applications (e.g. pulseaudio) attempt to
1463 			 * send ancillary data even if the underlying protocol
1464 			 * doesn't support it which is not allowed in the
1465 			 * FreeBSD system call interface.
1466 			 */
1467 			if (sa_family != AF_UNIX)
1468 				goto next;
1469 
1470 			if (cmsg->cmsg_type == SCM_CREDS) {
1471 				len = sizeof(struct cmsgcred);
1472 				if (datalen + CMSG_SPACE(len) > MCLBYTES)
1473 					goto bad;
1474 
1475 				/*
1476 				 * The lower levels will fill in the structure
1477 				 */
1478 				memset(CMSG_DATA(data), 0, len);
1479 			} else {
1480 				len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1481 				if (datalen + CMSG_SPACE(len) < datalen ||
1482 				    datalen + CMSG_SPACE(len) > MCLBYTES)
1483 					goto bad;
1484 
1485 				error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1486 				    CMSG_DATA(data), len);
1487 				if (error != 0)
1488 					goto bad;
1489 			}
1490 
1491 			cmsg->cmsg_len = CMSG_LEN(len);
1492 			data = (char *)data + CMSG_SPACE(len);
1493 			datalen += CMSG_SPACE(len);
1494 
1495 next:
1496 			if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1497 				break;
1498 
1499 			clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1500 			ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1501 			    LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1502 		} while(clen >= sizeof(struct l_cmsghdr));
1503 
1504 		control->m_len = datalen;
1505 		if (datalen == 0) {
1506 			m_freem(control);
1507 			control = NULL;
1508 		}
1509 	}
1510 
1511 	msg.msg_iov = iov;
1512 	msg.msg_flags = 0;
1513 	error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1514 	control = NULL;
1515 
1516 bad:
1517 	m_freem(control);
1518 	free(iov, M_IOV);
1519 	return (error);
1520 }
1521 
1522 int
1523 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1524 {
1525 
1526 	return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1527 	    args->flags));
1528 }
1529 
1530 int
1531 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1532 {
1533 	struct l_mmsghdr *msg;
1534 	l_uint retval;
1535 	int error, datagrams;
1536 
1537 	if (args->vlen > UIO_MAXIOV)
1538 		args->vlen = UIO_MAXIOV;
1539 
1540 	msg = PTRIN(args->msg);
1541 	datagrams = 0;
1542 	while (datagrams < args->vlen) {
1543 		error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1544 		    args->flags);
1545 		if (error != 0)
1546 			break;
1547 
1548 		retval = td->td_retval[0];
1549 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1550 		if (error != 0)
1551 			break;
1552 		++msg;
1553 		++datagrams;
1554 	}
1555 	if (error == 0)
1556 		td->td_retval[0] = datagrams;
1557 	return (error);
1558 }
1559 
1560 static int
1561 recvmsg_scm_rights(struct thread *td, l_uint flags, socklen_t *datalen,
1562     void **data, void **udata)
1563 {
1564 	int i, fd, fds, *fdp;
1565 
1566 	if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1567 		fds = *datalen / sizeof(int);
1568 		fdp = *data;
1569 		for (i = 0; i < fds; i++) {
1570 			fd = *fdp++;
1571 			(void)kern_fcntl(td, fd, F_SETFD, FD_CLOEXEC);
1572 		}
1573 	}
1574 	return (0);
1575 }
1576 
1577 
1578 static int
1579 recvmsg_scm_creds(socklen_t *datalen, void **data, void **udata)
1580 {
1581 	struct cmsgcred *cmcred;
1582 	struct l_ucred lu;
1583 
1584 	cmcred = *data;
1585 	lu.pid = cmcred->cmcred_pid;
1586 	lu.uid = cmcred->cmcred_uid;
1587 	lu.gid = cmcred->cmcred_gid;
1588 	memmove(*data, &lu, sizeof(lu));
1589 	*datalen = sizeof(lu);
1590 	return (0);
1591 }
1592 _Static_assert(sizeof(struct cmsgcred) >= sizeof(struct l_ucred),
1593     "scm_creds sizeof l_ucred");
1594 
1595 static int
1596 recvmsg_scm_creds2(socklen_t *datalen, void **data, void **udata)
1597 {
1598 	struct sockcred2 *scred;
1599 	struct l_ucred lu;
1600 
1601 	scred = *data;
1602 	lu.pid = scred->sc_pid;
1603 	lu.uid = scred->sc_uid;
1604 	lu.gid = scred->sc_gid;
1605 	memmove(*data, &lu, sizeof(lu));
1606 	*datalen = sizeof(lu);
1607 	return (0);
1608 }
1609 _Static_assert(sizeof(struct sockcred2) >= sizeof(struct l_ucred),
1610     "scm_creds2 sizeof l_ucred");
1611 
1612 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1613 static int
1614 recvmsg_scm_timestamp(l_int msg_type, socklen_t *datalen, void **data,
1615     void **udata)
1616 {
1617 	l_sock_timeval ltv64;
1618 	l_timeval ltv;
1619 	struct timeval *tv;
1620 	socklen_t len;
1621 	void *buf;
1622 
1623 	if (*datalen != sizeof(struct timeval))
1624 		return (EMSGSIZE);
1625 
1626 	tv = *data;
1627 #if defined(COMPAT_LINUX32)
1628 	if (msg_type == LINUX_SCM_TIMESTAMPO &&
1629 	    (tv->tv_sec > INT_MAX || tv->tv_sec < INT_MIN))
1630 		return (EOVERFLOW);
1631 #endif
1632 	if (msg_type == LINUX_SCM_TIMESTAMPN)
1633 		len = sizeof(ltv64);
1634 	else
1635 		len = sizeof(ltv);
1636 
1637 	buf = malloc(len, M_LINUX, M_WAITOK);
1638 	if (msg_type == LINUX_SCM_TIMESTAMPN) {
1639 		ltv64.tv_sec = tv->tv_sec;
1640 		ltv64.tv_usec = tv->tv_usec;
1641 		memmove(buf, &ltv64, len);
1642 	} else {
1643 		ltv.tv_sec = tv->tv_sec;
1644 		ltv.tv_usec = tv->tv_usec;
1645 		memmove(buf, &ltv, len);
1646 	}
1647 	*data = *udata = buf;
1648 	*datalen = len;
1649 	return (0);
1650 }
1651 #else
1652 _Static_assert(sizeof(struct timeval) == sizeof(l_timeval),
1653     "scm_timestamp sizeof l_timeval");
1654 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1655 
1656 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1657 static int
1658 recvmsg_scm_timestampns(l_int msg_type, socklen_t *datalen, void **data,
1659     void **udata)
1660 {
1661 	struct l_timespec64 ts64;
1662 	struct l_timespec ts32;
1663 	struct timespec ts;
1664 	socklen_t len;
1665 	void *buf;
1666 
1667 	if (msg_type == LINUX_SCM_TIMESTAMPNSO)
1668 		len = sizeof(ts32);
1669 	else
1670 		len = sizeof(ts64);
1671 
1672 	buf = malloc(len, M_LINUX, M_WAITOK);
1673 	bintime2timespec(*data, &ts);
1674 	if (msg_type == LINUX_SCM_TIMESTAMPNSO) {
1675 		ts32.tv_sec = ts.tv_sec;
1676 		ts32.tv_nsec = ts.tv_nsec;
1677 		memmove(buf, &ts32, len);
1678 	} else {
1679 		ts64.tv_sec = ts.tv_sec;
1680 		ts64.tv_nsec = ts.tv_nsec;
1681 		memmove(buf, &ts64, len);
1682 	}
1683 	*data = *udata = buf;
1684 	*datalen = len;
1685 	return (0);
1686 }
1687 #else
1688 static int
1689 recvmsg_scm_timestampns(l_int msg_type, socklen_t *datalen, void **data,
1690     void **udata)
1691 {
1692 	struct timespec ts;
1693 
1694 	bintime2timespec(*data, &ts);
1695 	memmove(*data, &ts, sizeof(struct timespec));
1696 	*datalen = sizeof(struct timespec);
1697 	return (0);
1698 }
1699 _Static_assert(sizeof(struct bintime) >= sizeof(struct timespec),
1700     "scm_timestampns sizeof timespec");
1701 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1702 
1703 static int
1704 recvmsg_scm_ip_origdstaddr(socklen_t *datalen, void **data, void **udata)
1705 {
1706 	struct l_sockaddr *lsa;
1707 	int error;
1708 
1709 	error = bsd_to_linux_sockaddr(*data, &lsa, *datalen);
1710 	if (error == 0) {
1711 		*data = *udata = lsa;
1712 		*datalen = sizeof(*lsa);
1713 	}
1714 	return (error);
1715 }
1716 
1717 static int
1718 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1719     l_uint flags, struct msghdr *msg)
1720 {
1721 	struct proc *p = td->td_proc;
1722 	struct cmsghdr *cm;
1723 	struct l_cmsghdr *lcm = NULL;
1724 	socklen_t datalen, maxlen, outlen;
1725 	struct l_msghdr l_msghdr;
1726 	struct iovec *iov, *uiov;
1727 	struct mbuf *m, *control = NULL;
1728 	struct mbuf **controlp;
1729 	struct sockaddr *sa;
1730 	caddr_t outbuf;
1731 	void *data, *udata;
1732 	int error;
1733 
1734 	error = copyin(msghdr, &l_msghdr, sizeof(l_msghdr));
1735 	if (error != 0)
1736 		return (error);
1737 
1738 	/*
1739 	 * Pass user-supplied recvmsg() flags in msg_flags field,
1740 	 * following sys_recvmsg() convention.
1741 	*/
1742 	l_msghdr.msg_flags = flags;
1743 
1744 	error = linux_to_bsd_msghdr(msg, &l_msghdr);
1745 	if (error != 0)
1746 		return (error);
1747 
1748 #ifdef COMPAT_LINUX32
1749 	error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1750 	    &iov, EMSGSIZE);
1751 #else
1752 	error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1753 #endif
1754 	if (error != 0)
1755 		return (error);
1756 
1757 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1758 		msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN);
1759 		sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK);
1760 		msg->msg_name = sa;
1761 	} else {
1762 		sa = NULL;
1763 		msg->msg_name = NULL;
1764 	}
1765 
1766 	uiov = msg->msg_iov;
1767 	msg->msg_iov = iov;
1768 	controlp = (msg->msg_control != NULL) ? &control : NULL;
1769 	error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp);
1770 	msg->msg_iov = uiov;
1771 	if (error != 0)
1772 		goto bad;
1773 
1774 	/*
1775 	 * Note that kern_recvit() updates msg->msg_namelen.
1776 	 */
1777 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1778 		msg->msg_name = PTRIN(l_msghdr.msg_name);
1779 		error = linux_copyout_sockaddr(sa, msg->msg_name,
1780 		    msg->msg_namelen);
1781 		if (error != 0)
1782 			goto bad;
1783 	}
1784 
1785 	error = bsd_to_linux_msghdr(msg, &l_msghdr);
1786 	if (error != 0)
1787 		goto bad;
1788 
1789 	maxlen = l_msghdr.msg_controllen;
1790 	l_msghdr.msg_controllen = 0;
1791 	if (control == NULL)
1792 		goto out;
1793 
1794 	lcm = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1795 	msg->msg_control = mtod(control, struct cmsghdr *);
1796 	msg->msg_controllen = control->m_len;
1797 	outbuf = PTRIN(l_msghdr.msg_control);
1798 	outlen = 0;
1799 	for (m = control; m != NULL; m = m->m_next) {
1800 		cm = mtod(m, struct cmsghdr *);
1801 		lcm->cmsg_type = bsd_to_linux_cmsg_type(p, cm->cmsg_type,
1802 		    cm->cmsg_level);
1803 		lcm->cmsg_level = bsd_to_linux_sockopt_level(cm->cmsg_level);
1804 
1805 		data = CMSG_DATA(cm);
1806 		datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1807 		udata = NULL;
1808 		error = 0;
1809 
1810 		/* Process non SOL_SOCKET types. */
1811 		if (cm->cmsg_level == IPPROTO_IP &&
1812 		    lcm->cmsg_type == LINUX_IP_ORIGDSTADDR) {
1813 			error = recvmsg_scm_ip_origdstaddr(&datalen, &data, &udata);
1814 			goto cont;
1815 		}
1816 
1817 		if (lcm->cmsg_type == -1 ||
1818 		    cm->cmsg_level != SOL_SOCKET) {
1819 			LINUX_RATELIMIT_MSG_OPT2(
1820 			    "unsupported recvmsg cmsg level %d type %d",
1821 			    cm->cmsg_level, cm->cmsg_type);
1822 			error = EINVAL;
1823 			goto bad;
1824 		}
1825 
1826 
1827 		switch (cm->cmsg_type) {
1828 		case SCM_RIGHTS:
1829 			error = recvmsg_scm_rights(td, flags,
1830 			    &datalen, &data, &udata);
1831 			break;
1832 		case SCM_CREDS:
1833 			error = recvmsg_scm_creds(&datalen,
1834 			    &data, &udata);
1835 			break;
1836 		case SCM_CREDS2:
1837 			error = recvmsg_scm_creds2(&datalen,
1838 			    &data, &udata);
1839 			break;
1840 		case SCM_TIMESTAMP:
1841 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1842 			error = recvmsg_scm_timestamp(lcm->cmsg_type,
1843 			    &datalen, &data, &udata);
1844 #endif
1845 			break;
1846 		case SCM_BINTIME:
1847 			error = recvmsg_scm_timestampns(lcm->cmsg_type,
1848 			    &datalen, &data, &udata);
1849 			break;
1850 		}
1851 
1852 cont:
1853 		if (error != 0)
1854 			goto bad;
1855 
1856 		if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1857 			if (outlen == 0) {
1858 				error = EMSGSIZE;
1859 				goto err;
1860 			} else {
1861 				l_msghdr.msg_flags |= LINUX_MSG_CTRUNC;
1862 				m_dispose_extcontrolm(control);
1863 				free(udata, M_LINUX);
1864 				goto out;
1865 			}
1866 		}
1867 
1868 		lcm->cmsg_len = LINUX_CMSG_LEN(datalen);
1869 		error = copyout(lcm, outbuf, L_CMSG_HDRSZ);
1870 		if (error == 0) {
1871 			outbuf += L_CMSG_HDRSZ;
1872 			error = copyout(data, outbuf, datalen);
1873 			if (error == 0) {
1874 				outbuf += LINUX_CMSG_ALIGN(datalen);
1875 				outlen += LINUX_CMSG_LEN(datalen);
1876 			}
1877 		}
1878 err:
1879 		free(udata, M_LINUX);
1880 		if (error != 0)
1881 			goto bad;
1882 	}
1883 	l_msghdr.msg_controllen = outlen;
1884 
1885 out:
1886 	error = copyout(&l_msghdr, msghdr, sizeof(l_msghdr));
1887 
1888 bad:
1889 	if (control != NULL) {
1890 		if (error != 0)
1891 			m_dispose_extcontrolm(control);
1892 		m_freem(control);
1893 	}
1894 	free(iov, M_IOV);
1895 	free(lcm, M_LINUX);
1896 	free(sa, M_SONAME);
1897 
1898 	return (error);
1899 }
1900 
1901 int
1902 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1903 {
1904 	struct msghdr bsd_msg;
1905 	struct file *fp;
1906 	int error;
1907 
1908 	error = getsock(td, args->s, &cap_recv_rights, &fp);
1909 	if (error != 0)
1910 		return (error);
1911 	fdrop(fp, td);
1912 	return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1913 	    args->flags, &bsd_msg));
1914 }
1915 
1916 static int
1917 linux_recvmmsg_common(struct thread *td, l_int s, struct l_mmsghdr *msg,
1918     l_uint vlen, l_uint flags, struct timespec *tts)
1919 {
1920 	struct msghdr bsd_msg;
1921 	struct timespec ts;
1922 	struct file *fp;
1923 	l_uint retval;
1924 	int error, datagrams;
1925 
1926 	error = getsock(td, s, &cap_recv_rights, &fp);
1927 	if (error != 0)
1928 		return (error);
1929 	datagrams = 0;
1930 	while (datagrams < vlen) {
1931 		error = linux_recvmsg_common(td, s, &msg->msg_hdr,
1932 		    flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1933 		if (error != 0)
1934 			break;
1935 
1936 		retval = td->td_retval[0];
1937 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1938 		if (error != 0)
1939 			break;
1940 		++msg;
1941 		++datagrams;
1942 
1943 		/*
1944 		 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1945 		 */
1946 		if (flags & LINUX_MSG_WAITFORONE)
1947 			flags |= LINUX_MSG_DONTWAIT;
1948 
1949 		/*
1950 		 * See BUGS section of recvmmsg(2).
1951 		 */
1952 		if (tts) {
1953 			getnanotime(&ts);
1954 			timespecsub(&ts, tts, &ts);
1955 			if (!timespecisset(&ts) || ts.tv_sec > 0)
1956 				break;
1957 		}
1958 		/* Out of band data, return right away. */
1959 		if (bsd_msg.msg_flags & MSG_OOB)
1960 			break;
1961 	}
1962 	if (error == 0)
1963 		td->td_retval[0] = datagrams;
1964 	fdrop(fp, td);
1965 	return (error);
1966 }
1967 
1968 int
1969 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1970 {
1971 	struct timespec ts, tts, *ptts;
1972 	int error;
1973 
1974 	if (args->timeout) {
1975 		error = linux_get_timespec(&ts, args->timeout);
1976 		if (error != 0)
1977 			return (error);
1978 		getnanotime(&tts);
1979 		timespecadd(&tts, &ts, &tts);
1980 		ptts = &tts;
1981 	}
1982 		else ptts = NULL;
1983 
1984 	return (linux_recvmmsg_common(td, args->s, PTRIN(args->msg),
1985 	    args->vlen, args->flags, ptts));
1986 }
1987 
1988 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1989 int
1990 linux_recvmmsg_time64(struct thread *td, struct linux_recvmmsg_time64_args *args)
1991 {
1992 	struct timespec ts, tts, *ptts;
1993 	int error;
1994 
1995 	if (args->timeout) {
1996 		error = linux_get_timespec64(&ts, args->timeout);
1997 		if (error != 0)
1998 			return (error);
1999 		getnanotime(&tts);
2000 		timespecadd(&tts, &ts, &tts);
2001 		ptts = &tts;
2002 	}
2003 		else ptts = NULL;
2004 
2005 	return (linux_recvmmsg_common(td, args->s, PTRIN(args->msg),
2006 	    args->vlen, args->flags, ptts));
2007 }
2008 #endif
2009 
2010 int
2011 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
2012 {
2013 
2014 	return (kern_shutdown(td, args->s, args->how));
2015 }
2016 
2017 int
2018 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
2019 {
2020 	struct proc *p = td->td_proc;
2021 	struct linux_pemuldata *pem;
2022 	l_timeval linux_tv;
2023 	struct sockaddr *sa;
2024 	struct timeval tv;
2025 	socklen_t len;
2026 	int error, level, name, val;
2027 
2028 	level = linux_to_bsd_sockopt_level(args->level);
2029 	switch (level) {
2030 	case SOL_SOCKET:
2031 		name = linux_to_bsd_so_sockopt(args->optname);
2032 		switch (name) {
2033 		case LOCAL_CREDS_PERSISTENT:
2034 			level = SOL_LOCAL;
2035 			break;
2036 		case SO_RCVTIMEO:
2037 			/* FALLTHROUGH */
2038 		case SO_SNDTIMEO:
2039 			error = copyin(PTRIN(args->optval), &linux_tv,
2040 			    sizeof(linux_tv));
2041 			if (error != 0)
2042 				return (error);
2043 			tv.tv_sec = linux_tv.tv_sec;
2044 			tv.tv_usec = linux_tv.tv_usec;
2045 			return (kern_setsockopt(td, args->s, level,
2046 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
2047 			/* NOTREACHED */
2048 		case SO_TIMESTAMP:
2049 			/* overwrite SO_BINTIME */
2050 			val = 0;
2051 			error = kern_setsockopt(td, args->s, level,
2052 			    SO_BINTIME, &val, UIO_SYSSPACE, sizeof(val));
2053 			if (error != 0)
2054 				return (error);
2055 			pem = pem_find(p);
2056 			pem->so_timestamp = args->optname;
2057 			break;
2058 		case SO_BINTIME:
2059 			/* overwrite SO_TIMESTAMP */
2060 			val = 0;
2061 			error = kern_setsockopt(td, args->s, level,
2062 			    SO_TIMESTAMP, &val, UIO_SYSSPACE, sizeof(val));
2063 			if (error != 0)
2064 				return (error);
2065 			pem = pem_find(p);
2066 			pem->so_timestampns = args->optname;
2067 			break;
2068 		default:
2069 			break;
2070 		}
2071 		break;
2072 	case IPPROTO_IP:
2073 		if (args->optname == LINUX_IP_RECVERR &&
2074 		    linux_ignore_ip_recverr) {
2075 			/*
2076 			 * XXX: This is a hack to unbreak DNS resolution
2077 			 *	with glibc 2.30 and above.
2078 			 */
2079 			return (0);
2080 		}
2081 		name = linux_to_bsd_ip_sockopt(args->optname);
2082 		break;
2083 	case IPPROTO_IPV6:
2084 		name = linux_to_bsd_ip6_sockopt(args->optname);
2085 		break;
2086 	case IPPROTO_TCP:
2087 		name = linux_to_bsd_tcp_sockopt(args->optname);
2088 		break;
2089 	case SOL_NETLINK:
2090 		level = SOL_SOCKET;
2091 		name = args->optname;
2092 		break;
2093 	default:
2094 		name = -1;
2095 		break;
2096 	}
2097 	if (name < 0) {
2098 		if (name == -1)
2099 			linux_msg(curthread,
2100 			    "unsupported setsockopt level %d optname %d",
2101 			    args->level, args->optname);
2102 		return (ENOPROTOOPT);
2103 	}
2104 
2105 	if (name == IPV6_NEXTHOP) {
2106 		len = args->optlen;
2107 		error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len);
2108 		if (error != 0)
2109 			return (error);
2110 
2111 		error = kern_setsockopt(td, args->s, level,
2112 		    name, sa, UIO_SYSSPACE, len);
2113 		free(sa, M_SONAME);
2114 	} else {
2115 		error = kern_setsockopt(td, args->s, level,
2116 		    name, PTRIN(args->optval), UIO_USERSPACE, args->optlen);
2117 	}
2118 
2119 	return (error);
2120 }
2121 
2122 static int
2123 linux_sockopt_copyout(struct thread *td, void *val, socklen_t len,
2124     struct linux_getsockopt_args *args)
2125 {
2126 	int error;
2127 
2128 	error = copyout(val, PTRIN(args->optval), len);
2129 	if (error == 0)
2130 		error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2131 	return (error);
2132 }
2133 
2134 static int
2135 linux_getsockopt_so_peergroups(struct thread *td,
2136     struct linux_getsockopt_args *args)
2137 {
2138 	struct xucred xu;
2139 	socklen_t xulen, len;
2140 	int error, i;
2141 
2142 	xulen = sizeof(xu);
2143 	error = kern_getsockopt(td, args->s, 0,
2144 	    LOCAL_PEERCRED, &xu, UIO_SYSSPACE, &xulen);
2145 	if (error != 0)
2146 		return (error);
2147 
2148 	len = xu.cr_ngroups * sizeof(l_gid_t);
2149 	if (args->optlen < len) {
2150 		error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2151 		if (error == 0)
2152 			error = ERANGE;
2153 		return (error);
2154 	}
2155 
2156 	/*
2157 	 * "- 1" to skip the primary group.
2158 	 */
2159 	for (i = 0; i < xu.cr_ngroups - 1; i++) {
2160 		error = copyout(xu.cr_groups + i + 1,
2161 		    (void *)(args->optval + i * sizeof(l_gid_t)),
2162 		    sizeof(l_gid_t));
2163 		if (error != 0)
2164 			return (error);
2165 	}
2166 
2167 	error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2168 	return (error);
2169 }
2170 
2171 static int
2172 linux_getsockopt_so_peersec(struct thread *td,
2173     struct linux_getsockopt_args *args)
2174 {
2175 	socklen_t len;
2176 	int error;
2177 
2178 	len = sizeof(SECURITY_CONTEXT_STRING);
2179 	if (args->optlen < len) {
2180 		error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2181 		if (error == 0)
2182 			error = ERANGE;
2183 		return (error);
2184 	}
2185 
2186 	return (linux_sockopt_copyout(td, SECURITY_CONTEXT_STRING,
2187 	    len, args));
2188 }
2189 
2190 static int
2191 linux_getsockopt_so_linger(struct thread *td,
2192     struct linux_getsockopt_args *args)
2193 {
2194 	struct linger ling;
2195 	socklen_t len;
2196 	int error;
2197 
2198 	len = sizeof(ling);
2199 	error = kern_getsockopt(td, args->s, SOL_SOCKET,
2200 	    SO_LINGER, &ling, UIO_SYSSPACE, &len);
2201 	if (error != 0)
2202 		return (error);
2203 	ling.l_onoff = ((ling.l_onoff & SO_LINGER) != 0);
2204 	return (linux_sockopt_copyout(td, &ling, len, args));
2205 }
2206 
2207 int
2208 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
2209 {
2210 	l_timeval linux_tv;
2211 	struct timeval tv;
2212 	socklen_t tv_len, xulen, len;
2213 	struct sockaddr *sa;
2214 	struct xucred xu;
2215 	struct l_ucred lxu;
2216 	int error, level, name, newval;
2217 
2218 	level = linux_to_bsd_sockopt_level(args->level);
2219 	switch (level) {
2220 	case SOL_SOCKET:
2221 		switch (args->optname) {
2222 		case LINUX_SO_PEERGROUPS:
2223 			return (linux_getsockopt_so_peergroups(td, args));
2224 		case LINUX_SO_PEERSEC:
2225 			return (linux_getsockopt_so_peersec(td, args));
2226 		default:
2227 			break;
2228 		}
2229 
2230 		name = linux_to_bsd_so_sockopt(args->optname);
2231 		switch (name) {
2232 		case LOCAL_CREDS_PERSISTENT:
2233 			level = SOL_LOCAL;
2234 			break;
2235 		case SO_RCVTIMEO:
2236 			/* FALLTHROUGH */
2237 		case SO_SNDTIMEO:
2238 			tv_len = sizeof(tv);
2239 			error = kern_getsockopt(td, args->s, level,
2240 			    name, &tv, UIO_SYSSPACE, &tv_len);
2241 			if (error != 0)
2242 				return (error);
2243 			linux_tv.tv_sec = tv.tv_sec;
2244 			linux_tv.tv_usec = tv.tv_usec;
2245 			return (linux_sockopt_copyout(td, &linux_tv,
2246 			    sizeof(linux_tv), args));
2247 			/* NOTREACHED */
2248 		case LOCAL_PEERCRED:
2249 			if (args->optlen < sizeof(lxu))
2250 				return (EINVAL);
2251 			/*
2252 			 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
2253 			 * but by the Unix socket's level 0.
2254 			 */
2255 			level = 0;
2256 			xulen = sizeof(xu);
2257 			error = kern_getsockopt(td, args->s, level,
2258 			    name, &xu, UIO_SYSSPACE, &xulen);
2259 			if (error != 0)
2260 				return (error);
2261 			lxu.pid = xu.cr_pid;
2262 			lxu.uid = xu.cr_uid;
2263 			lxu.gid = xu.cr_gid;
2264 			return (linux_sockopt_copyout(td, &lxu,
2265 			    sizeof(lxu), args));
2266 			/* NOTREACHED */
2267 		case SO_ERROR:
2268 			len = sizeof(newval);
2269 			error = kern_getsockopt(td, args->s, level,
2270 			    name, &newval, UIO_SYSSPACE, &len);
2271 			if (error != 0)
2272 				return (error);
2273 			newval = -bsd_to_linux_errno(newval);
2274 			return (linux_sockopt_copyout(td, &newval,
2275 			    len, args));
2276 			/* NOTREACHED */
2277 		case SO_DOMAIN:
2278 			len = sizeof(newval);
2279 			error = kern_getsockopt(td, args->s, level,
2280 			    name, &newval, UIO_SYSSPACE, &len);
2281 			if (error != 0)
2282 				return (error);
2283 			newval = bsd_to_linux_domain(newval);
2284 			if (newval == -1)
2285 				return (ENOPROTOOPT);
2286 			return (linux_sockopt_copyout(td, &newval,
2287 			    len, args));
2288 			/* NOTREACHED */
2289 		case SO_LINGER:
2290 			return (linux_getsockopt_so_linger(td, args));
2291 			/* NOTREACHED */
2292 		default:
2293 			break;
2294 		}
2295 		break;
2296 	case IPPROTO_IP:
2297 		name = linux_to_bsd_ip_sockopt(args->optname);
2298 		break;
2299 	case IPPROTO_IPV6:
2300 		name = linux_to_bsd_ip6_sockopt(args->optname);
2301 		break;
2302 	case IPPROTO_TCP:
2303 		name = linux_to_bsd_tcp_sockopt(args->optname);
2304 		break;
2305 	default:
2306 		name = -1;
2307 		break;
2308 	}
2309 	if (name < 0) {
2310 		if (name == -1)
2311 			linux_msg(curthread,
2312 			    "unsupported getsockopt level %d optname %d",
2313 			    args->level, args->optname);
2314 		return (EINVAL);
2315 	}
2316 
2317 	if (name == IPV6_NEXTHOP) {
2318 		error = copyin(PTRIN(args->optlen), &len, sizeof(len));
2319                 if (error != 0)
2320                         return (error);
2321 		sa = malloc(len, M_SONAME, M_WAITOK);
2322 
2323 		error = kern_getsockopt(td, args->s, level,
2324 		    name, sa, UIO_SYSSPACE, &len);
2325 		if (error != 0)
2326 			goto out;
2327 
2328 		error = linux_copyout_sockaddr(sa, PTRIN(args->optval), len);
2329 		if (error == 0)
2330 			error = copyout(&len, PTRIN(args->optlen),
2331 			    sizeof(len));
2332 out:
2333 		free(sa, M_SONAME);
2334 	} else {
2335 		if (args->optval) {
2336 			error = copyin(PTRIN(args->optlen), &len, sizeof(len));
2337 			if (error != 0)
2338 				return (error);
2339 		}
2340 		error = kern_getsockopt(td, args->s, level,
2341 		    name, PTRIN(args->optval), UIO_USERSPACE, &len);
2342 		if (error == 0)
2343 			error = copyout(&len, PTRIN(args->optlen),
2344 			    sizeof(len));
2345 	}
2346 
2347 	return (error);
2348 }
2349 
2350 static int
2351 linux_sendfile_common(struct thread *td, l_int out, l_int in,
2352     l_loff_t *offset, l_size_t count)
2353 {
2354 	off_t bytes_read;
2355 	int error;
2356 	l_loff_t current_offset;
2357 	struct file *fp;
2358 
2359 	AUDIT_ARG_FD(in);
2360 	error = fget_read(td, in, &cap_pread_rights, &fp);
2361 	if (error != 0)
2362 		return (error);
2363 
2364 	if (offset != NULL) {
2365 		current_offset = *offset;
2366 	} else {
2367 		error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2368 		    fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE;
2369 		if (error != 0)
2370 			goto drop;
2371 		current_offset = td->td_uretoff.tdu_off;
2372 	}
2373 
2374 	bytes_read = 0;
2375 
2376 	/* Linux cannot have 0 count. */
2377 	if (count <= 0 || current_offset < 0) {
2378 		error = EINVAL;
2379 		goto drop;
2380 	}
2381 
2382 	error = fo_sendfile(fp, out, NULL, NULL, current_offset, count,
2383 	    &bytes_read, 0, td);
2384 	if (error != 0)
2385 		goto drop;
2386 	current_offset += bytes_read;
2387 
2388 	if (offset != NULL) {
2389 		*offset = current_offset;
2390 	} else {
2391 		error = fo_seek(fp, current_offset, SEEK_SET, td);
2392 		if (error != 0)
2393 			goto drop;
2394 	}
2395 
2396 	td->td_retval[0] = (ssize_t)bytes_read;
2397 drop:
2398 	fdrop(fp, td);
2399 	if (error == ENOTSOCK)
2400 		error = EINVAL;
2401 	return (error);
2402 }
2403 
2404 int
2405 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg)
2406 {
2407 	/*
2408 	 * Differences between FreeBSD and Linux sendfile:
2409 	 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to
2410 	 *   mean send the whole file.)  In linux_sendfile given fds are still
2411 	 *   checked for validity when the count is 0.
2412 	 * - Linux can send to any fd whereas FreeBSD only supports sockets.
2413 	 *   The same restriction follows for linux_sendfile.
2414 	 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr.
2415 	 * - Linux takes an offset pointer and updates it to the read location.
2416 	 *   FreeBSD takes in an offset and a 'bytes read' parameter which is
2417 	 *   only filled if it isn't NULL.  We use this parameter to update the
2418 	 *   offset pointer if it exists.
2419 	 * - Linux sendfile returns bytes read on success while FreeBSD
2420 	 *   returns 0.  We use the 'bytes read' parameter to get this value.
2421 	 */
2422 
2423 	l_loff_t offset64;
2424 	l_long offset;
2425 	int ret;
2426 	int error;
2427 
2428 	if (arg->offset != NULL) {
2429 		error = copyin(arg->offset, &offset, sizeof(offset));
2430 		if (error != 0)
2431 			return (error);
2432 		offset64 = (l_loff_t)offset;
2433 	}
2434 
2435 	ret = linux_sendfile_common(td, arg->out, arg->in,
2436 	    arg->offset != NULL ? &offset64 : NULL, arg->count);
2437 
2438 	if (arg->offset != NULL) {
2439 #if defined(__i386__) || defined(__arm__) || \
2440     (defined(__amd64__) && defined(COMPAT_LINUX32))
2441 		if (offset64 > INT32_MAX)
2442 			return (EOVERFLOW);
2443 #endif
2444 		offset = (l_long)offset64;
2445 		error = copyout(&offset, arg->offset, sizeof(offset));
2446 		if (error != 0)
2447 			return (error);
2448 	}
2449 
2450 	return (ret);
2451 }
2452 
2453 #if defined(__i386__) || defined(__arm__) || \
2454     (defined(__amd64__) && defined(COMPAT_LINUX32))
2455 
2456 int
2457 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg)
2458 {
2459 	l_loff_t offset;
2460 	int ret;
2461 	int error;
2462 
2463 	if (arg->offset != NULL) {
2464 		error = copyin(arg->offset, &offset, sizeof(offset));
2465 		if (error != 0)
2466 			return (error);
2467 	}
2468 
2469 	ret = linux_sendfile_common(td, arg->out, arg->in,
2470 		arg->offset != NULL ? &offset : NULL, arg->count);
2471 
2472 	if (arg->offset != NULL) {
2473 		error = copyout(&offset, arg->offset, sizeof(offset));
2474 		if (error != 0)
2475 			return (error);
2476 	}
2477 
2478 	return (ret);
2479 }
2480 
2481 /* Argument list sizes for linux_socketcall */
2482 static const unsigned char lxs_args_cnt[] = {
2483 	0 /* unused*/,		3 /* socket */,
2484 	3 /* bind */,		3 /* connect */,
2485 	2 /* listen */,		3 /* accept */,
2486 	3 /* getsockname */,	3 /* getpeername */,
2487 	4 /* socketpair */,	4 /* send */,
2488 	4 /* recv */,		6 /* sendto */,
2489 	6 /* recvfrom */,	2 /* shutdown */,
2490 	5 /* setsockopt */,	5 /* getsockopt */,
2491 	3 /* sendmsg */,	3 /* recvmsg */,
2492 	4 /* accept4 */,	5 /* recvmmsg */,
2493 	4 /* sendmmsg */,	4 /* sendfile */
2494 };
2495 #define	LINUX_ARGS_CNT		(nitems(lxs_args_cnt) - 1)
2496 #define	LINUX_ARG_SIZE(x)	(lxs_args_cnt[x] * sizeof(l_ulong))
2497 
2498 int
2499 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
2500 {
2501 	l_ulong a[6];
2502 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2503 	register_t l_args[6];
2504 #endif
2505 	void *arg;
2506 	int error;
2507 
2508 	if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
2509 		return (EINVAL);
2510 	error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
2511 	if (error != 0)
2512 		return (error);
2513 
2514 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2515 	for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
2516 		l_args[i] = a[i];
2517 	arg = l_args;
2518 #else
2519 	arg = a;
2520 #endif
2521 	switch (args->what) {
2522 	case LINUX_SOCKET:
2523 		return (linux_socket(td, arg));
2524 	case LINUX_BIND:
2525 		return (linux_bind(td, arg));
2526 	case LINUX_CONNECT:
2527 		return (linux_connect(td, arg));
2528 	case LINUX_LISTEN:
2529 		return (linux_listen(td, arg));
2530 	case LINUX_ACCEPT:
2531 		return (linux_accept(td, arg));
2532 	case LINUX_GETSOCKNAME:
2533 		return (linux_getsockname(td, arg));
2534 	case LINUX_GETPEERNAME:
2535 		return (linux_getpeername(td, arg));
2536 	case LINUX_SOCKETPAIR:
2537 		return (linux_socketpair(td, arg));
2538 	case LINUX_SEND:
2539 		return (linux_send(td, arg));
2540 	case LINUX_RECV:
2541 		return (linux_recv(td, arg));
2542 	case LINUX_SENDTO:
2543 		return (linux_sendto(td, arg));
2544 	case LINUX_RECVFROM:
2545 		return (linux_recvfrom(td, arg));
2546 	case LINUX_SHUTDOWN:
2547 		return (linux_shutdown(td, arg));
2548 	case LINUX_SETSOCKOPT:
2549 		return (linux_setsockopt(td, arg));
2550 	case LINUX_GETSOCKOPT:
2551 		return (linux_getsockopt(td, arg));
2552 	case LINUX_SENDMSG:
2553 		return (linux_sendmsg(td, arg));
2554 	case LINUX_RECVMSG:
2555 		return (linux_recvmsg(td, arg));
2556 	case LINUX_ACCEPT4:
2557 		return (linux_accept4(td, arg));
2558 	case LINUX_RECVMMSG:
2559 		return (linux_recvmmsg(td, arg));
2560 	case LINUX_SENDMMSG:
2561 		return (linux_sendmmsg(td, arg));
2562 	case LINUX_SENDFILE:
2563 		return (linux_sendfile(td, arg));
2564 	}
2565 
2566 	linux_msg(td, "socket type %d not implemented", args->what);
2567 	return (ENOSYS);
2568 }
2569 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */
2570