xref: /freebsd/sys/compat/linux/linux_socket.c (revision 0c785f06020f3b02e34c97eb27fecd3af8eb2a7b)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
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 "opt_inet6.h"
30 
31 #include <sys/param.h>
32 #include <sys/capsicum.h>
33 #include <sys/filedesc.h>
34 #include <sys/limits.h>
35 #include <sys/malloc.h>
36 #include <sys/mbuf.h>
37 #include <sys/proc.h>
38 #include <sys/protosw.h>
39 #include <sys/socket.h>
40 #include <sys/socketvar.h>
41 #include <sys/syscallsubr.h>
42 #include <sys/sysproto.h>
43 #include <sys/vnode.h>
44 #include <sys/un.h>
45 #include <sys/unistd.h>
46 
47 #include <security/audit/audit.h>
48 
49 #include <net/if.h>
50 #include <net/vnet.h>
51 #include <netinet/in.h>
52 #include <netinet/ip.h>
53 #include <netinet/tcp.h>
54 #ifdef INET6
55 #include <netinet/ip6.h>
56 #include <netinet6/ip6_var.h>
57 #endif
58 
59 #ifdef COMPAT_LINUX32
60 #include <machine/../linux32/linux.h>
61 #include <machine/../linux32/linux32_proto.h>
62 #else
63 #include <machine/../linux/linux.h>
64 #include <machine/../linux/linux_proto.h>
65 #endif
66 #include <compat/linux/linux_common.h>
67 #include <compat/linux/linux_emul.h>
68 #include <compat/linux/linux_file.h>
69 #include <compat/linux/linux_mib.h>
70 #include <compat/linux/linux_socket.h>
71 #include <compat/linux/linux_time.h>
72 #include <compat/linux/linux_util.h>
73 
74 _Static_assert(offsetof(struct l_ifreq, ifr_ifru) ==
75     offsetof(struct ifreq, ifr_ifru),
76     "Linux ifreq members names should be equal to FreeeBSD");
77 _Static_assert(offsetof(struct l_ifreq, ifr_index) ==
78     offsetof(struct ifreq, ifr_index),
79     "Linux ifreq members names should be equal to FreeeBSD");
80 _Static_assert(offsetof(struct l_ifreq, ifr_name) ==
81     offsetof(struct ifreq, ifr_name),
82     "Linux ifreq members names should be equal to FreeeBSD");
83 
84 #define	SECURITY_CONTEXT_STRING	"unconfined"
85 
86 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
87 					l_uint);
88 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
89 					l_uint, struct msghdr *);
90 static int linux_set_socket_flags(int, int *);
91 
92 #define	SOL_NETLINK	270
93 
94 static int
95 linux_to_bsd_sockopt_level(int level)
96 {
97 
98 	if (level == LINUX_SOL_SOCKET)
99 		return (SOL_SOCKET);
100 	/* Remaining values are RFC-defined protocol numbers. */
101 	return (level);
102 }
103 
104 static int
105 bsd_to_linux_sockopt_level(int level)
106 {
107 
108 	if (level == SOL_SOCKET)
109 		return (LINUX_SOL_SOCKET);
110 	return (level);
111 }
112 
113 static int
114 linux_to_bsd_ip_sockopt(int opt)
115 {
116 
117 	switch (opt) {
118 	/* known and translated sockopts */
119 	case LINUX_IP_TOS:
120 		return (IP_TOS);
121 	case LINUX_IP_TTL:
122 		return (IP_TTL);
123 	case LINUX_IP_HDRINCL:
124 		return (IP_HDRINCL);
125 	case LINUX_IP_OPTIONS:
126 		return (IP_OPTIONS);
127 	case LINUX_IP_RECVOPTS:
128 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVOPTS");
129 		return (IP_RECVOPTS);
130 	case LINUX_IP_RETOPTS:
131 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_REETOPTS");
132 		return (IP_RETOPTS);
133 	case LINUX_IP_RECVTTL:
134 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTTL");
135 		return (IP_RECVTTL);
136 	case LINUX_IP_RECVTOS:
137 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTOS");
138 		return (IP_RECVTOS);
139 	case LINUX_IP_FREEBIND:
140 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_FREEBIND");
141 		return (IP_BINDANY);
142 	case LINUX_IP_IPSEC_POLICY:
143 		/* we have this option, but not documented in ip(4) manpage */
144 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_IPSEC_POLICY");
145 		return (IP_IPSEC_POLICY);
146 	case LINUX_IP_MINTTL:
147 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MINTTL");
148 		return (IP_MINTTL);
149 	case LINUX_IP_MULTICAST_IF:
150 		return (IP_MULTICAST_IF);
151 	case LINUX_IP_MULTICAST_TTL:
152 		return (IP_MULTICAST_TTL);
153 	case LINUX_IP_MULTICAST_LOOP:
154 		return (IP_MULTICAST_LOOP);
155 	case LINUX_IP_ADD_MEMBERSHIP:
156 		return (IP_ADD_MEMBERSHIP);
157 	case LINUX_IP_DROP_MEMBERSHIP:
158 		return (IP_DROP_MEMBERSHIP);
159 	case LINUX_IP_UNBLOCK_SOURCE:
160 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_UNBLOCK_SOURCE");
161 		return (IP_UNBLOCK_SOURCE);
162 	case LINUX_IP_BLOCK_SOURCE:
163 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_BLOCK_SOURCE");
164 		return (IP_BLOCK_SOURCE);
165 	case LINUX_IP_ADD_SOURCE_MEMBERSHIP:
166 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_ADD_SOURCE_MEMBERSHIP");
167 		return (IP_ADD_SOURCE_MEMBERSHIP);
168 	case LINUX_IP_DROP_SOURCE_MEMBERSHIP:
169 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_DROP_SOURCE_MEMBERSHIP");
170 		return (IP_DROP_SOURCE_MEMBERSHIP);
171 	case LINUX_MCAST_JOIN_GROUP:
172 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_GROUP");
173 		return (MCAST_JOIN_GROUP);
174 	case LINUX_MCAST_LEAVE_GROUP:
175 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_GROUP");
176 		return (MCAST_LEAVE_GROUP);
177 	case LINUX_MCAST_JOIN_SOURCE_GROUP:
178 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_SOURCE_GROUP");
179 		return (MCAST_JOIN_SOURCE_GROUP);
180 	case LINUX_MCAST_LEAVE_SOURCE_GROUP:
181 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_SOURCE_GROUP");
182 		return (MCAST_LEAVE_SOURCE_GROUP);
183 	case LINUX_IP_RECVORIGDSTADDR:
184 		return (IP_RECVORIGDSTADDR);
185 
186 	/* known but not implemented sockopts */
187 	case LINUX_IP_ROUTER_ALERT:
188 		LINUX_RATELIMIT_MSG_OPT1(
189 		    "unsupported IPv4 socket option IP_ROUTER_ALERT (%d), you can not do user-space routing from linux programs",
190 		    opt);
191 		return (-2);
192 	case LINUX_IP_PKTINFO:
193 		LINUX_RATELIMIT_MSG_OPT1(
194 		    "unsupported IPv4 socket option IP_PKTINFO (%d), you can not get extended packet info for datagram sockets in linux programs",
195 		    opt);
196 		return (-2);
197 	case LINUX_IP_PKTOPTIONS:
198 		LINUX_RATELIMIT_MSG_OPT1(
199 		    "unsupported IPv4 socket option IP_PKTOPTIONS (%d)",
200 		    opt);
201 		return (-2);
202 	case LINUX_IP_MTU_DISCOVER:
203 		LINUX_RATELIMIT_MSG_OPT1(
204 		    "unsupported IPv4 socket option IP_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery",
205 		    opt);
206 		return (-2);
207 	case LINUX_IP_RECVERR:
208 		/* needed by steam */
209 		LINUX_RATELIMIT_MSG_OPT1(
210 		    "unsupported IPv4 socket option IP_RECVERR (%d), you can not get extended reliability info in linux programs",
211 		    opt);
212 		return (-2);
213 	case LINUX_IP_MTU:
214 		LINUX_RATELIMIT_MSG_OPT1(
215 		    "unsupported IPv4 socket option IP_MTU (%d), your linux program can not control the MTU on this socket",
216 		    opt);
217 		return (-2);
218 	case LINUX_IP_XFRM_POLICY:
219 		LINUX_RATELIMIT_MSG_OPT1(
220 		    "unsupported IPv4 socket option IP_XFRM_POLICY (%d)",
221 		    opt);
222 		return (-2);
223 	case LINUX_IP_PASSSEC:
224 		/* needed by steam */
225 		LINUX_RATELIMIT_MSG_OPT1(
226 		    "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",
227 		    opt);
228 		return (-2);
229 	case LINUX_IP_TRANSPARENT:
230 		/* IP_BINDANY or more? */
231 		LINUX_RATELIMIT_MSG_OPT1(
232 		    "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",
233 		    opt);
234 		return (-2);
235 	case LINUX_IP_NODEFRAG:
236 		LINUX_RATELIMIT_MSG_OPT1(
237 		    "unsupported IPv4 socket option IP_NODEFRAG (%d)",
238 		    opt);
239 		return (-2);
240 	case LINUX_IP_CHECKSUM:
241 		LINUX_RATELIMIT_MSG_OPT1(
242 		    "unsupported IPv4 socket option IP_CHECKSUM (%d)",
243 		    opt);
244 		return (-2);
245 	case LINUX_IP_BIND_ADDRESS_NO_PORT:
246 		LINUX_RATELIMIT_MSG_OPT1(
247 		    "unsupported IPv4 socket option IP_BIND_ADDRESS_NO_PORT (%d)",
248 		    opt);
249 		return (-2);
250 	case LINUX_IP_RECVFRAGSIZE:
251 		LINUX_RATELIMIT_MSG_OPT1(
252 		    "unsupported IPv4 socket option IP_RECVFRAGSIZE (%d)",
253 		    opt);
254 		return (-2);
255 	case LINUX_MCAST_MSFILTER:
256 		LINUX_RATELIMIT_MSG_OPT1(
257 		    "unsupported IPv4 socket option IP_MCAST_MSFILTER (%d)",
258 		    opt);
259 		return (-2);
260 	case LINUX_IP_MULTICAST_ALL:
261 		LINUX_RATELIMIT_MSG_OPT1(
262 		    "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",
263 		    opt);
264 		return (-2);
265 	case LINUX_IP_UNICAST_IF:
266 		LINUX_RATELIMIT_MSG_OPT1(
267 		    "unsupported IPv4 socket option IP_UNICAST_IF (%d)",
268 		    opt);
269 		return (-2);
270 
271 	/* unknown sockopts */
272 	default:
273 		return (-1);
274 	}
275 }
276 
277 static int
278 linux_to_bsd_ip6_sockopt(int opt)
279 {
280 
281 	switch (opt) {
282 	/* known and translated sockopts */
283 	case LINUX_IPV6_2292PKTINFO:
284 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTINFO");
285 		return (IPV6_2292PKTINFO);
286 	case LINUX_IPV6_2292HOPOPTS:
287 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPOPTS");
288 		return (IPV6_2292HOPOPTS);
289 	case LINUX_IPV6_2292DSTOPTS:
290 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292DSTOPTS");
291 		return (IPV6_2292DSTOPTS);
292 	case LINUX_IPV6_2292RTHDR:
293 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292RTHDR");
294 		return (IPV6_2292RTHDR);
295 	case LINUX_IPV6_2292PKTOPTIONS:
296 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTOPTIONS");
297 		return (IPV6_2292PKTOPTIONS);
298 	case LINUX_IPV6_CHECKSUM:
299 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_CHECKSUM");
300 		return (IPV6_CHECKSUM);
301 	case LINUX_IPV6_2292HOPLIMIT:
302 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPLIMIT");
303 		return (IPV6_2292HOPLIMIT);
304 	case LINUX_IPV6_NEXTHOP:
305 		return (IPV6_NEXTHOP);
306 	case LINUX_IPV6_UNICAST_HOPS:
307 		return (IPV6_UNICAST_HOPS);
308 	case LINUX_IPV6_MULTICAST_IF:
309 		return (IPV6_MULTICAST_IF);
310 	case LINUX_IPV6_MULTICAST_HOPS:
311 		return (IPV6_MULTICAST_HOPS);
312 	case LINUX_IPV6_MULTICAST_LOOP:
313 		return (IPV6_MULTICAST_LOOP);
314 	case LINUX_IPV6_ADD_MEMBERSHIP:
315 		return (IPV6_JOIN_GROUP);
316 	case LINUX_IPV6_DROP_MEMBERSHIP:
317 		return (IPV6_LEAVE_GROUP);
318 	case LINUX_IPV6_V6ONLY:
319 		return (IPV6_V6ONLY);
320 	case LINUX_IPV6_IPSEC_POLICY:
321 		/* we have this option, but not documented in ip6(4) manpage */
322 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_IPSEC_POLICY");
323 		return (IPV6_IPSEC_POLICY);
324 	case LINUX_MCAST_JOIN_GROUP:
325 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_JOIN_GROUP");
326 		return (IPV6_JOIN_GROUP);
327 	case LINUX_MCAST_LEAVE_GROUP:
328 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_LEAVE_GROUP");
329 		return (IPV6_LEAVE_GROUP);
330 	case LINUX_IPV6_RECVPKTINFO:
331 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPKTINFO");
332 		return (IPV6_RECVPKTINFO);
333 	case LINUX_IPV6_PKTINFO:
334 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PKTINFO");
335 		return (IPV6_PKTINFO);
336 	case LINUX_IPV6_RECVHOPLIMIT:
337 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPLIMIT");
338 		return (IPV6_RECVHOPLIMIT);
339 	case LINUX_IPV6_HOPLIMIT:
340 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPLIMIT");
341 		return (IPV6_HOPLIMIT);
342 	case LINUX_IPV6_RECVHOPOPTS:
343 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPOPTS");
344 		return (IPV6_RECVHOPOPTS);
345 	case LINUX_IPV6_HOPOPTS:
346 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPOPTS");
347 		return (IPV6_HOPOPTS);
348 	case LINUX_IPV6_RTHDRDSTOPTS:
349 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDRDSTOPTS");
350 		return (IPV6_RTHDRDSTOPTS);
351 	case LINUX_IPV6_RECVRTHDR:
352 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVRTHDR");
353 		return (IPV6_RECVRTHDR);
354 	case LINUX_IPV6_RTHDR:
355 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDR");
356 		return (IPV6_RTHDR);
357 	case LINUX_IPV6_RECVDSTOPTS:
358 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVDSTOPTS");
359 		return (IPV6_RECVDSTOPTS);
360 	case LINUX_IPV6_DSTOPTS:
361 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_DSTOPTS");
362 		return (IPV6_DSTOPTS);
363 	case LINUX_IPV6_RECVPATHMTU:
364 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPATHMTU");
365 		return (IPV6_RECVPATHMTU);
366 	case LINUX_IPV6_PATHMTU:
367 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PATHMTU");
368 		return (IPV6_PATHMTU);
369 	case LINUX_IPV6_DONTFRAG:
370 		return (IPV6_DONTFRAG);
371 	case LINUX_IPV6_AUTOFLOWLABEL:
372 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_AUTOFLOWLABEL");
373 		return (IPV6_AUTOFLOWLABEL);
374 	case LINUX_IPV6_ORIGDSTADDR:
375 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_ORIGDSTADDR");
376 		return (IPV6_ORIGDSTADDR);
377 	case LINUX_IPV6_FREEBIND:
378 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_FREEBIND");
379 		return (IPV6_BINDANY);
380 
381 	/* known but not implemented sockopts */
382 	case LINUX_IPV6_ADDRFORM:
383 		LINUX_RATELIMIT_MSG_OPT1(
384 		    "unsupported IPv6 socket option IPV6_ADDRFORM (%d), you linux program can not convert the socket to IPv4",
385 		    opt);
386 		return (-2);
387 	case LINUX_IPV6_AUTHHDR:
388 		LINUX_RATELIMIT_MSG_OPT1(
389 		    "unsupported IPv6 socket option IPV6_AUTHHDR (%d), your linux program can not get the authentication header info of IPv6 packets",
390 		    opt);
391 		return (-2);
392 	case LINUX_IPV6_FLOWINFO:
393 		LINUX_RATELIMIT_MSG_OPT1(
394 		    "unsupported IPv6 socket option IPV6_FLOWINFO (%d), your linux program can not get the flowid of IPv6 packets",
395 		    opt);
396 		return (-2);
397 	case LINUX_IPV6_ROUTER_ALERT:
398 		LINUX_RATELIMIT_MSG_OPT1(
399 		    "unsupported IPv6 socket option IPV6_ROUTER_ALERT (%d), you can not do user-space routing from linux programs",
400 		    opt);
401 		return (-2);
402 	case LINUX_IPV6_MTU_DISCOVER:
403 		LINUX_RATELIMIT_MSG_OPT1(
404 		    "unsupported IPv6 socket option IPV6_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery",
405 		    opt);
406 		return (-2);
407 	case LINUX_IPV6_MTU:
408 		LINUX_RATELIMIT_MSG_OPT1(
409 		    "unsupported IPv6 socket option IPV6_MTU (%d), your linux program can not control the MTU on this socket",
410 		    opt);
411 		return (-2);
412 	case LINUX_IPV6_JOIN_ANYCAST:
413 		LINUX_RATELIMIT_MSG_OPT1(
414 		    "unsupported IPv6 socket option IPV6_JOIN_ANYCAST (%d)",
415 		    opt);
416 		return (-2);
417 	case LINUX_IPV6_LEAVE_ANYCAST:
418 		LINUX_RATELIMIT_MSG_OPT1(
419 		    "unsupported IPv6 socket option IPV6_LEAVE_ANYCAST (%d)",
420 		    opt);
421 		return (-2);
422 	case LINUX_IPV6_MULTICAST_ALL:
423 		LINUX_RATELIMIT_MSG_OPT1(
424 		    "unsupported IPv6 socket option IPV6_MULTICAST_ALL (%d)",
425 		    opt);
426 		return (-2);
427 	case LINUX_IPV6_ROUTER_ALERT_ISOLATE:
428 		LINUX_RATELIMIT_MSG_OPT1(
429 		    "unsupported IPv6 socket option IPV6_ROUTER_ALERT_ISOLATE (%d)",
430 		    opt);
431 		return (-2);
432 	case LINUX_IPV6_FLOWLABEL_MGR:
433 		LINUX_RATELIMIT_MSG_OPT1(
434 		    "unsupported IPv6 socket option IPV6_FLOWLABEL_MGR (%d)",
435 		    opt);
436 		return (-2);
437 	case LINUX_IPV6_FLOWINFO_SEND:
438 		LINUX_RATELIMIT_MSG_OPT1(
439 		    "unsupported IPv6 socket option IPV6_FLOWINFO_SEND (%d)",
440 		    opt);
441 		return (-2);
442 	case LINUX_IPV6_XFRM_POLICY:
443 		LINUX_RATELIMIT_MSG_OPT1(
444 		    "unsupported IPv6 socket option IPV6_XFRM_POLICY (%d)",
445 		    opt);
446 		return (-2);
447 	case LINUX_IPV6_HDRINCL:
448 		LINUX_RATELIMIT_MSG_OPT1(
449 		    "unsupported IPv6 socket option IPV6_HDRINCL (%d)",
450 		    opt);
451 		return (-2);
452 	case LINUX_MCAST_BLOCK_SOURCE:
453 		LINUX_RATELIMIT_MSG_OPT1(
454 		    "unsupported IPv6 socket option MCAST_BLOCK_SOURCE (%d), your linux program may see more multicast stuff than it wants",
455 		    opt);
456 		return (-2);
457 	case LINUX_MCAST_UNBLOCK_SOURCE:
458 		LINUX_RATELIMIT_MSG_OPT1(
459 		    "unsupported IPv6 socket option MCAST_UNBLOCK_SOURCE (%d), your linux program may not see all the multicast stuff it wants",
460 		    opt);
461 		return (-2);
462 	case LINUX_MCAST_JOIN_SOURCE_GROUP:
463 		LINUX_RATELIMIT_MSG_OPT1(
464 		    "unsupported IPv6 socket option MCAST_JOIN_SOURCE_GROUP (%d), your linux program is not able to join a multicast source group",
465 		    opt);
466 		return (-2);
467 	case LINUX_MCAST_LEAVE_SOURCE_GROUP:
468 		LINUX_RATELIMIT_MSG_OPT1(
469 		    "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",
470 		    opt);
471 		return (-2);
472 	case LINUX_MCAST_MSFILTER:
473 		LINUX_RATELIMIT_MSG_OPT1(
474 		    "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",
475 		    opt);
476 		return (-2);
477 	case LINUX_IPV6_ADDR_PREFERENCES:
478 		LINUX_RATELIMIT_MSG_OPT1(
479 		    "unsupported IPv6 socket option IPV6_ADDR_PREFERENCES (%d)",
480 		    opt);
481 		return (-2);
482 	case LINUX_IPV6_MINHOPCOUNT:
483 		LINUX_RATELIMIT_MSG_OPT1(
484 		    "unsupported IPv6 socket option IPV6_MINHOPCOUNT (%d)",
485 		    opt);
486 		return (-2);
487 	case LINUX_IPV6_TRANSPARENT:
488 		/* IP_BINDANY or more? */
489 		LINUX_RATELIMIT_MSG_OPT1(
490 		    "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",
491 		    opt);
492 		return (-2);
493 	case LINUX_IPV6_UNICAST_IF:
494 		LINUX_RATELIMIT_MSG_OPT1(
495 		    "unsupported IPv6 socket option IPV6_UNICAST_IF (%d)",
496 		    opt);
497 		return (-2);
498 	case LINUX_IPV6_RECVFRAGSIZE:
499 		LINUX_RATELIMIT_MSG_OPT1(
500 		    "unsupported IPv6 socket option IPV6_RECVFRAGSIZE (%d)",
501 		    opt);
502 		return (-2);
503 
504 	/* unknown sockopts */
505 	default:
506 		return (-1);
507 	}
508 }
509 
510 static int
511 linux_to_bsd_so_sockopt(int opt)
512 {
513 
514 	switch (opt) {
515 	case LINUX_SO_DEBUG:
516 		return (SO_DEBUG);
517 	case LINUX_SO_REUSEADDR:
518 		return (SO_REUSEADDR);
519 	case LINUX_SO_TYPE:
520 		return (SO_TYPE);
521 	case LINUX_SO_ERROR:
522 		return (SO_ERROR);
523 	case LINUX_SO_DONTROUTE:
524 		return (SO_DONTROUTE);
525 	case LINUX_SO_BROADCAST:
526 		return (SO_BROADCAST);
527 	case LINUX_SO_SNDBUF:
528 	case LINUX_SO_SNDBUFFORCE:
529 		return (SO_SNDBUF);
530 	case LINUX_SO_RCVBUF:
531 	case LINUX_SO_RCVBUFFORCE:
532 		return (SO_RCVBUF);
533 	case LINUX_SO_KEEPALIVE:
534 		return (SO_KEEPALIVE);
535 	case LINUX_SO_OOBINLINE:
536 		return (SO_OOBINLINE);
537 	case LINUX_SO_LINGER:
538 		return (SO_LINGER);
539 	case LINUX_SO_REUSEPORT:
540 		return (SO_REUSEPORT_LB);
541 	case LINUX_SO_PASSCRED:
542 		return (LOCAL_CREDS_PERSISTENT);
543 	case LINUX_SO_PEERCRED:
544 		return (LOCAL_PEERCRED);
545 	case LINUX_SO_RCVLOWAT:
546 		return (SO_RCVLOWAT);
547 	case LINUX_SO_SNDLOWAT:
548 		return (SO_SNDLOWAT);
549 	case LINUX_SO_RCVTIMEO:
550 		return (SO_RCVTIMEO);
551 	case LINUX_SO_SNDTIMEO:
552 		return (SO_SNDTIMEO);
553 	case LINUX_SO_TIMESTAMPO:
554 	case LINUX_SO_TIMESTAMPN:
555 		return (SO_TIMESTAMP);
556 	case LINUX_SO_TIMESTAMPNSO:
557 	case LINUX_SO_TIMESTAMPNSN:
558 		return (SO_BINTIME);
559 	case LINUX_SO_ACCEPTCONN:
560 		return (SO_ACCEPTCONN);
561 	case LINUX_SO_PROTOCOL:
562 		return (SO_PROTOCOL);
563 	case LINUX_SO_DOMAIN:
564 		return (SO_DOMAIN);
565 	}
566 	return (-1);
567 }
568 
569 static int
570 linux_to_bsd_tcp_sockopt(int opt)
571 {
572 
573 	switch (opt) {
574 	case LINUX_TCP_NODELAY:
575 		return (TCP_NODELAY);
576 	case LINUX_TCP_MAXSEG:
577 		return (TCP_MAXSEG);
578 	case LINUX_TCP_CORK:
579 		return (TCP_NOPUSH);
580 	case LINUX_TCP_KEEPIDLE:
581 		return (TCP_KEEPIDLE);
582 	case LINUX_TCP_KEEPINTVL:
583 		return (TCP_KEEPINTVL);
584 	case LINUX_TCP_KEEPCNT:
585 		return (TCP_KEEPCNT);
586 	case LINUX_TCP_INFO:
587 		LINUX_RATELIMIT_MSG_OPT1(
588 		    "unsupported TCP socket option TCP_INFO (%d)", opt);
589 		return (-2);
590 	case LINUX_TCP_MD5SIG:
591 		return (TCP_MD5SIG);
592 	}
593 	return (-1);
594 }
595 
596 static int
597 linux_to_bsd_msg_flags(int flags)
598 {
599 	int ret_flags = 0;
600 
601 	if (flags & LINUX_MSG_OOB)
602 		ret_flags |= MSG_OOB;
603 	if (flags & LINUX_MSG_PEEK)
604 		ret_flags |= MSG_PEEK;
605 	if (flags & LINUX_MSG_DONTROUTE)
606 		ret_flags |= MSG_DONTROUTE;
607 	if (flags & LINUX_MSG_CTRUNC)
608 		ret_flags |= MSG_CTRUNC;
609 	if (flags & LINUX_MSG_TRUNC)
610 		ret_flags |= MSG_TRUNC;
611 	if (flags & LINUX_MSG_DONTWAIT)
612 		ret_flags |= MSG_DONTWAIT;
613 	if (flags & LINUX_MSG_EOR)
614 		ret_flags |= MSG_EOR;
615 	if (flags & LINUX_MSG_WAITALL)
616 		ret_flags |= MSG_WAITALL;
617 	if (flags & LINUX_MSG_NOSIGNAL)
618 		ret_flags |= MSG_NOSIGNAL;
619 	if (flags & LINUX_MSG_PROXY)
620 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_PROXY (%d) not handled",
621 		    LINUX_MSG_PROXY);
622 	if (flags & LINUX_MSG_FIN)
623 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_FIN (%d) not handled",
624 		    LINUX_MSG_FIN);
625 	if (flags & LINUX_MSG_SYN)
626 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_SYN (%d) not handled",
627 		    LINUX_MSG_SYN);
628 	if (flags & LINUX_MSG_CONFIRM)
629 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_CONFIRM (%d) not handled",
630 		    LINUX_MSG_CONFIRM);
631 	if (flags & LINUX_MSG_RST)
632 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_RST (%d) not handled",
633 		    LINUX_MSG_RST);
634 	if (flags & LINUX_MSG_ERRQUEUE)
635 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_ERRQUEUE (%d) not handled",
636 		    LINUX_MSG_ERRQUEUE);
637 	return (ret_flags);
638 }
639 
640 static int
641 linux_to_bsd_cmsg_type(int cmsg_type)
642 {
643 
644 	switch (cmsg_type) {
645 	case LINUX_SCM_RIGHTS:
646 		return (SCM_RIGHTS);
647 	case LINUX_SCM_CREDENTIALS:
648 		return (SCM_CREDS);
649 	}
650 	return (-1);
651 }
652 
653 static int
654 bsd_to_linux_ip_cmsg_type(int cmsg_type)
655 {
656 
657 	switch (cmsg_type) {
658 	case IP_RECVORIGDSTADDR:
659 		return (LINUX_IP_RECVORIGDSTADDR);
660 	}
661 	return (-1);
662 }
663 
664 static int
665 bsd_to_linux_cmsg_type(struct proc *p, int cmsg_type, int cmsg_level)
666 {
667 	struct linux_pemuldata *pem;
668 
669 	if (cmsg_level == IPPROTO_IP)
670 		return (bsd_to_linux_ip_cmsg_type(cmsg_type));
671 	if (cmsg_level != SOL_SOCKET)
672 		return (-1);
673 
674 	pem = pem_find(p);
675 
676 	switch (cmsg_type) {
677 	case SCM_RIGHTS:
678 		return (LINUX_SCM_RIGHTS);
679 	case SCM_CREDS:
680 		return (LINUX_SCM_CREDENTIALS);
681 	case SCM_CREDS2:
682 		return (LINUX_SCM_CREDENTIALS);
683 	case SCM_TIMESTAMP:
684 		return (pem->so_timestamp);
685 	case SCM_BINTIME:
686 		return (pem->so_timestampns);
687 	}
688 	return (-1);
689 }
690 
691 static int
692 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
693 {
694 	if (lhdr->msg_controllen > INT_MAX)
695 		return (ENOBUFS);
696 
697 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
698 	bhdr->msg_namelen	= lhdr->msg_namelen;
699 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
700 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
701 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
702 
703 	/*
704 	 * msg_controllen is skipped since BSD and LINUX control messages
705 	 * are potentially different sizes (e.g. the cred structure used
706 	 * by SCM_CREDS is different between the two operating system).
707 	 *
708 	 * The caller can set it (if necessary) after converting all the
709 	 * control messages.
710 	 */
711 
712 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
713 	return (0);
714 }
715 
716 static int
717 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
718 {
719 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
720 	lhdr->msg_namelen	= bhdr->msg_namelen;
721 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
722 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
723 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
724 
725 	/*
726 	 * msg_controllen is skipped since BSD and LINUX control messages
727 	 * are potentially different sizes (e.g. the cred structure used
728 	 * by SCM_CREDS is different between the two operating system).
729 	 *
730 	 * The caller can set it (if necessary) after converting all the
731 	 * control messages.
732 	 */
733 
734 	/* msg_flags skipped */
735 	return (0);
736 }
737 
738 static int
739 linux_set_socket_flags(int lflags, int *flags)
740 {
741 
742 	if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
743 		return (EINVAL);
744 	if (lflags & LINUX_SOCK_NONBLOCK)
745 		*flags |= SOCK_NONBLOCK;
746 	if (lflags & LINUX_SOCK_CLOEXEC)
747 		*flags |= SOCK_CLOEXEC;
748 	return (0);
749 }
750 
751 static int
752 linux_copyout_sockaddr(const struct sockaddr *sa, void *uaddr, size_t len)
753 {
754 	struct l_sockaddr *lsa;
755 	int error;
756 
757 	error = bsd_to_linux_sockaddr(sa, &lsa, len);
758 	if (error != 0)
759 		return (error);
760 
761 	error = copyout(lsa, uaddr, len);
762 	free(lsa, M_LINUX);
763 
764 	return (error);
765 }
766 
767 static int
768 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
769     struct mbuf *control, enum uio_seg segflg)
770 {
771 	struct sockaddr *to;
772 	int error, len;
773 
774 	if (mp->msg_name != NULL) {
775 		len = mp->msg_namelen;
776 		error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len);
777 		if (error != 0)
778 			return (error);
779 		mp->msg_name = to;
780 	} else
781 		to = NULL;
782 
783 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
784 	    segflg);
785 
786 	if (to)
787 		free(to, M_SONAME);
788 	return (error);
789 }
790 
791 /* Return 0 if IP_HDRINCL is set for the given socket. */
792 static int
793 linux_check_hdrincl(struct thread *td, int s)
794 {
795 	int error, optval;
796 	socklen_t size_val;
797 
798 	size_val = sizeof(optval);
799 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
800 	    &optval, UIO_SYSSPACE, &size_val);
801 	if (error != 0)
802 		return (error);
803 
804 	return (optval == 0);
805 }
806 
807 /*
808  * Updated sendto() when IP_HDRINCL is set:
809  * tweak endian-dependent fields in the IP packet.
810  */
811 static int
812 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
813 {
814 /*
815  * linux_ip_copysize defines how many bytes we should copy
816  * from the beginning of the IP packet before we customize it for BSD.
817  * It should include all the fields we modify (ip_len and ip_off).
818  */
819 #define linux_ip_copysize	8
820 
821 	struct ip *packet;
822 	struct msghdr msg;
823 	struct iovec aiov[1];
824 	int error;
825 
826 	/* Check that the packet isn't too big or too small. */
827 	if (linux_args->len < linux_ip_copysize ||
828 	    linux_args->len > IP_MAXPACKET)
829 		return (EINVAL);
830 
831 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
832 
833 	/* Make kernel copy of the packet to be sent */
834 	if ((error = copyin(PTRIN(linux_args->msg), packet,
835 	    linux_args->len)))
836 		goto goout;
837 
838 	/* Convert fields from Linux to BSD raw IP socket format */
839 	packet->ip_len = linux_args->len;
840 	packet->ip_off = ntohs(packet->ip_off);
841 
842 	/* Prepare the msghdr and iovec structures describing the new packet */
843 	msg.msg_name = PTRIN(linux_args->to);
844 	msg.msg_namelen = linux_args->tolen;
845 	msg.msg_iov = aiov;
846 	msg.msg_iovlen = 1;
847 	msg.msg_control = NULL;
848 	msg.msg_flags = 0;
849 	aiov[0].iov_base = (char *)packet;
850 	aiov[0].iov_len = linux_args->len;
851 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
852 	    NULL, UIO_SYSSPACE);
853 goout:
854 	free(packet, M_LINUX);
855 	return (error);
856 }
857 
858 static const char *linux_netlink_names[] = {
859 	[LINUX_NETLINK_ROUTE] = "ROUTE",
860 	[LINUX_NETLINK_SOCK_DIAG] = "SOCK_DIAG",
861 	[LINUX_NETLINK_NFLOG] = "NFLOG",
862 	[LINUX_NETLINK_SELINUX] = "SELINUX",
863 	[LINUX_NETLINK_AUDIT] = "AUDIT",
864 	[LINUX_NETLINK_FIB_LOOKUP] = "FIB_LOOKUP",
865 	[LINUX_NETLINK_NETFILTER] = "NETFILTER",
866 	[LINUX_NETLINK_KOBJECT_UEVENT] = "KOBJECT_UEVENT",
867 };
868 
869 int
870 linux_socket(struct thread *td, struct linux_socket_args *args)
871 {
872 	int domain, retval_socket, type;
873 
874 	type = args->type & LINUX_SOCK_TYPE_MASK;
875 	if (type < 0 || type > LINUX_SOCK_MAX)
876 		return (EINVAL);
877 	retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
878 		&type);
879 	if (retval_socket != 0)
880 		return (retval_socket);
881 	domain = linux_to_bsd_domain(args->domain);
882 	if (domain == -1) {
883 		/* Mask off SOCK_NONBLOCK / CLOEXEC for error messages. */
884 		type = args->type & LINUX_SOCK_TYPE_MASK;
885 		if (args->domain == LINUX_AF_NETLINK &&
886 		    args->protocol == LINUX_NETLINK_AUDIT) {
887 			; /* Do nothing, quietly. */
888 		} else if (args->domain == LINUX_AF_NETLINK) {
889 			const char *nl_name;
890 
891 			if (args->protocol >= 0 &&
892 			    args->protocol < nitems(linux_netlink_names))
893 				nl_name = linux_netlink_names[args->protocol];
894 			else
895 				nl_name = NULL;
896 			if (nl_name != NULL)
897 				linux_msg(curthread,
898 				    "unsupported socket(AF_NETLINK, %d, "
899 				    "NETLINK_%s)", type, nl_name);
900 			else
901 				linux_msg(curthread,
902 				    "unsupported socket(AF_NETLINK, %d, %d)",
903 				    type, args->protocol);
904 		} else {
905 			linux_msg(curthread, "unsupported socket domain %d, "
906 			    "type %d, protocol %d", args->domain, type,
907 			    args->protocol);
908 		}
909 		return (EAFNOSUPPORT);
910 	}
911 
912 	retval_socket = kern_socket(td, domain, type, args->protocol);
913 	if (retval_socket)
914 		return (retval_socket);
915 
916 	if (type == SOCK_RAW
917 	    && (args->protocol == IPPROTO_RAW || args->protocol == 0)
918 	    && domain == PF_INET) {
919 		/* It's a raw IP socket: set the IP_HDRINCL option. */
920 		int hdrincl;
921 
922 		hdrincl = 1;
923 		/* We ignore any error returned by kern_setsockopt() */
924 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
925 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
926 	}
927 #ifdef INET6
928 	/*
929 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
930 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
931 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
932 	 * sysctl value.
933 	 */
934 	if (domain == PF_INET6) {
935 		int v6only;
936 
937 		v6only = 0;
938 		/* We ignore any error returned by setsockopt() */
939 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
940 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
941 	}
942 #endif
943 
944 	return (retval_socket);
945 }
946 
947 int
948 linux_bind(struct thread *td, struct linux_bind_args *args)
949 {
950 	struct sockaddr *sa;
951 	int error;
952 
953 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
954 	    &args->namelen);
955 	if (error != 0)
956 		return (error);
957 
958 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
959 	free(sa, M_SONAME);
960 
961 	/* XXX */
962 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
963 		return (EINVAL);
964 	return (error);
965 }
966 
967 int
968 linux_connect(struct thread *td, struct linux_connect_args *args)
969 {
970 	struct socket *so;
971 	struct sockaddr *sa;
972 	struct file *fp;
973 	int error;
974 
975 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
976 	    &args->namelen);
977 	if (error != 0)
978 		return (error);
979 
980 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
981 	free(sa, M_SONAME);
982 	if (error != EISCONN)
983 		return (error);
984 
985 	/*
986 	 * Linux doesn't return EISCONN the first time it occurs,
987 	 * when on a non-blocking socket. Instead it returns the
988 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
989 	 */
990 	error = getsock(td, args->s, &cap_connect_rights, &fp);
991 	if (error != 0)
992 		return (error);
993 
994 	error = EISCONN;
995 	so = fp->f_data;
996 	if (atomic_load_int(&fp->f_flag) & FNONBLOCK) {
997 		SOCK_LOCK(so);
998 		if (so->so_emuldata == 0)
999 			error = so->so_error;
1000 		so->so_emuldata = (void *)1;
1001 		SOCK_UNLOCK(so);
1002 	}
1003 	fdrop(fp, td);
1004 
1005 	return (error);
1006 }
1007 
1008 int
1009 linux_listen(struct thread *td, struct linux_listen_args *args)
1010 {
1011 
1012 	return (kern_listen(td, args->s, args->backlog));
1013 }
1014 
1015 static int
1016 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
1017     l_uintptr_t namelen, int flags)
1018 {
1019 	struct sockaddr *sa;
1020 	struct file *fp, *fp1;
1021 	int bflags, len;
1022 	struct socket *so;
1023 	int error, error1;
1024 
1025 	bflags = 0;
1026 	fp = NULL;
1027 	sa = NULL;
1028 
1029 	error = linux_set_socket_flags(flags, &bflags);
1030 	if (error != 0)
1031 		return (error);
1032 
1033 	if (PTRIN(addr) == NULL) {
1034 		len = 0;
1035 		error = kern_accept4(td, s, NULL, NULL, bflags, NULL);
1036 	} else {
1037 		error = copyin(PTRIN(namelen), &len, sizeof(len));
1038 		if (error != 0)
1039 			return (error);
1040 		if (len < 0)
1041 			return (EINVAL);
1042 		error = kern_accept4(td, s, &sa, &len, bflags, &fp);
1043 	}
1044 
1045 	/*
1046 	 * Translate errno values into ones used by Linux.
1047 	 */
1048 	if (error != 0) {
1049 		/*
1050 		 * XXX. This is wrong, different sockaddr structures
1051 		 * have different sizes.
1052 		 */
1053 		switch (error) {
1054 		case EFAULT:
1055 			if (namelen != sizeof(struct sockaddr_in))
1056 				error = EINVAL;
1057 			break;
1058 		case EINVAL:
1059 			error1 = getsock(td, s, &cap_accept_rights, &fp1);
1060 			if (error1 != 0) {
1061 				error = error1;
1062 				break;
1063 			}
1064 			so = fp1->f_data;
1065 			if (so->so_type == SOCK_DGRAM)
1066 				error = EOPNOTSUPP;
1067 			fdrop(fp1, td);
1068 			break;
1069 		}
1070 		return (error);
1071 	}
1072 
1073 	if (len != 0) {
1074 		error = linux_copyout_sockaddr(sa, PTRIN(addr), len);
1075 		if (error == 0)
1076 			error = copyout(&len, PTRIN(namelen),
1077 			    sizeof(len));
1078 		if (error != 0) {
1079 			fdclose(td, fp, td->td_retval[0]);
1080 			td->td_retval[0] = 0;
1081 		}
1082 	}
1083 	if (fp != NULL)
1084 		fdrop(fp, td);
1085 	free(sa, M_SONAME);
1086 	return (error);
1087 }
1088 
1089 int
1090 linux_accept(struct thread *td, struct linux_accept_args *args)
1091 {
1092 
1093 	return (linux_accept_common(td, args->s, args->addr,
1094 	    args->namelen, 0));
1095 }
1096 
1097 int
1098 linux_accept4(struct thread *td, struct linux_accept4_args *args)
1099 {
1100 
1101 	return (linux_accept_common(td, args->s, args->addr,
1102 	    args->namelen, args->flags));
1103 }
1104 
1105 int
1106 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
1107 {
1108 	struct sockaddr *sa;
1109 	int len, error;
1110 
1111 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1112 	if (error != 0)
1113 		return (error);
1114 
1115 	error = kern_getsockname(td, args->s, &sa, &len);
1116 	if (error != 0)
1117 		return (error);
1118 
1119 	if (len != 0)
1120 		error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
1121 
1122 	free(sa, M_SONAME);
1123 	if (error == 0)
1124 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1125 	return (error);
1126 }
1127 
1128 int
1129 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
1130 {
1131 	struct sockaddr *sa;
1132 	int len, error;
1133 
1134 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1135 	if (error != 0)
1136 		return (error);
1137 	if (len < 0)
1138 		return (EINVAL);
1139 
1140 	error = kern_getpeername(td, args->s, &sa, &len);
1141 	if (error != 0)
1142 		return (error);
1143 
1144 	if (len != 0)
1145 		error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
1146 
1147 	free(sa, M_SONAME);
1148 	if (error == 0)
1149 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1150 	return (error);
1151 }
1152 
1153 int
1154 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
1155 {
1156 	int domain, error, sv[2], type;
1157 
1158 	domain = linux_to_bsd_domain(args->domain);
1159 	if (domain != PF_LOCAL)
1160 		return (EAFNOSUPPORT);
1161 	type = args->type & LINUX_SOCK_TYPE_MASK;
1162 	if (type < 0 || type > LINUX_SOCK_MAX)
1163 		return (EINVAL);
1164 	error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
1165 	    &type);
1166 	if (error != 0)
1167 		return (error);
1168 	if (args->protocol != 0 && args->protocol != PF_UNIX) {
1169 		/*
1170 		 * Use of PF_UNIX as protocol argument is not right,
1171 		 * but Linux does it.
1172 		 * Do not map PF_UNIX as its Linux value is identical
1173 		 * to FreeBSD one.
1174 		 */
1175 		return (EPROTONOSUPPORT);
1176 	}
1177 	error = kern_socketpair(td, domain, type, 0, sv);
1178 	if (error != 0)
1179                 return (error);
1180         error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int));
1181         if (error != 0) {
1182                 (void)kern_close(td, sv[0]);
1183                 (void)kern_close(td, sv[1]);
1184         }
1185 	return (error);
1186 }
1187 
1188 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1189 struct linux_send_args {
1190 	register_t s;
1191 	register_t msg;
1192 	register_t len;
1193 	register_t flags;
1194 };
1195 
1196 static int
1197 linux_send(struct thread *td, struct linux_send_args *args)
1198 {
1199 	struct sendto_args /* {
1200 		int s;
1201 		caddr_t buf;
1202 		int len;
1203 		int flags;
1204 		caddr_t to;
1205 		int tolen;
1206 	} */ bsd_args;
1207 	struct file *fp;
1208 	int error;
1209 
1210 	bsd_args.s = args->s;
1211 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1212 	bsd_args.len = args->len;
1213 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1214 	bsd_args.to = NULL;
1215 	bsd_args.tolen = 0;
1216 	error = sys_sendto(td, &bsd_args);
1217 	if (error == ENOTCONN) {
1218 		/*
1219 		 * Linux doesn't return ENOTCONN for non-blocking sockets.
1220 		 * Instead it returns the EAGAIN.
1221 		 */
1222 		error = getsock(td, args->s, &cap_send_rights, &fp);
1223 		if (error == 0) {
1224 			if (atomic_load_int(&fp->f_flag) & FNONBLOCK)
1225 				error = EAGAIN;
1226 			fdrop(fp, td);
1227 		}
1228 	}
1229 	return (error);
1230 }
1231 
1232 struct linux_recv_args {
1233 	register_t s;
1234 	register_t msg;
1235 	register_t len;
1236 	register_t flags;
1237 };
1238 
1239 static int
1240 linux_recv(struct thread *td, struct linux_recv_args *args)
1241 {
1242 	struct recvfrom_args /* {
1243 		int s;
1244 		caddr_t buf;
1245 		int len;
1246 		int flags;
1247 		struct sockaddr *from;
1248 		socklen_t fromlenaddr;
1249 	} */ bsd_args;
1250 
1251 	bsd_args.s = args->s;
1252 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1253 	bsd_args.len = args->len;
1254 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1255 	bsd_args.from = NULL;
1256 	bsd_args.fromlenaddr = 0;
1257 	return (sys_recvfrom(td, &bsd_args));
1258 }
1259 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1260 
1261 int
1262 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1263 {
1264 	struct msghdr msg;
1265 	struct iovec aiov;
1266 	struct socket *so;
1267 	struct file *fp;
1268 	int error;
1269 
1270 	if (linux_check_hdrincl(td, args->s) == 0)
1271 		/* IP_HDRINCL set, tweak the packet before sending */
1272 		return (linux_sendto_hdrincl(td, args));
1273 
1274 	bzero(&msg, sizeof(msg));
1275 	error = getsock(td, args->s, &cap_send_connect_rights, &fp);
1276 	if (error != 0)
1277 		return (error);
1278 	so = fp->f_data;
1279 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0) {
1280 		msg.msg_name = PTRIN(args->to);
1281 		msg.msg_namelen = args->tolen;
1282 	}
1283 	msg.msg_iov = &aiov;
1284 	msg.msg_iovlen = 1;
1285 	aiov.iov_base = PTRIN(args->msg);
1286 	aiov.iov_len = args->len;
1287 	fdrop(fp, td);
1288 	return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1289 	    UIO_USERSPACE));
1290 }
1291 
1292 int
1293 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1294 {
1295 	struct sockaddr *sa;
1296 	struct msghdr msg;
1297 	struct iovec aiov;
1298 	int error, fromlen;
1299 
1300 	if (PTRIN(args->fromlen) != NULL) {
1301 		error = copyin(PTRIN(args->fromlen), &fromlen,
1302 		    sizeof(fromlen));
1303 		if (error != 0)
1304 			return (error);
1305 		if (fromlen < 0)
1306 			return (EINVAL);
1307 		fromlen = min(fromlen, SOCK_MAXADDRLEN);
1308 		sa = malloc(fromlen, M_SONAME, M_WAITOK);
1309 	} else {
1310 		fromlen = 0;
1311 		sa = NULL;
1312 	}
1313 
1314 	msg.msg_name = sa;
1315 	msg.msg_namelen = fromlen;
1316 	msg.msg_iov = &aiov;
1317 	msg.msg_iovlen = 1;
1318 	aiov.iov_base = PTRIN(args->buf);
1319 	aiov.iov_len = args->len;
1320 	msg.msg_control = 0;
1321 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1322 
1323 	error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL);
1324 	if (error != 0)
1325 		goto out;
1326 
1327 	/*
1328 	 * XXX. Seems that FreeBSD is different from Linux here. Linux
1329 	 * fill source address if underlying protocol provides it, while
1330 	 * FreeBSD fill it if underlying protocol is not connection-oriented.
1331 	 * So, kern_recvit() set msg.msg_namelen to 0 if protocol pr_flags
1332 	 * does not contains PR_ADDR flag.
1333 	 */
1334 	if (PTRIN(args->from) != NULL && msg.msg_namelen != 0)
1335 		error = linux_copyout_sockaddr(sa, PTRIN(args->from),
1336 		    msg.msg_namelen);
1337 
1338 	if (error == 0 && PTRIN(args->fromlen) != NULL)
1339 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1340 		    sizeof(msg.msg_namelen));
1341 out:
1342 	free(sa, M_SONAME);
1343 	return (error);
1344 }
1345 
1346 static int
1347 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1348     l_uint flags)
1349 {
1350 	struct cmsghdr *cmsg;
1351 	struct mbuf *control;
1352 	struct msghdr msg;
1353 	struct l_cmsghdr linux_cmsg;
1354 	struct l_cmsghdr *ptr_cmsg;
1355 	struct l_msghdr linux_msghdr;
1356 	struct iovec *iov;
1357 	socklen_t datalen;
1358 	struct sockaddr *sa;
1359 	struct socket *so;
1360 	sa_family_t sa_family;
1361 	struct file *fp;
1362 	void *data;
1363 	l_size_t len;
1364 	l_size_t clen;
1365 	int error;
1366 
1367 	error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
1368 	if (error != 0)
1369 		return (error);
1370 
1371 	/*
1372 	 * Some Linux applications (ping) define a non-NULL control data
1373 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1374 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1375 	 * order to handle this case.  This should be checked, but allows the
1376 	 * Linux ping to work.
1377 	 */
1378 	if (PTRIN(linux_msghdr.msg_control) != NULL &&
1379 	    linux_msghdr.msg_controllen == 0)
1380 		linux_msghdr.msg_control = PTROUT(NULL);
1381 
1382 	error = linux_to_bsd_msghdr(&msg, &linux_msghdr);
1383 	if (error != 0)
1384 		return (error);
1385 
1386 #ifdef COMPAT_LINUX32
1387 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1388 	    &iov, EMSGSIZE);
1389 #else
1390 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1391 #endif
1392 	if (error != 0)
1393 		return (error);
1394 
1395 	control = NULL;
1396 
1397 	error = kern_getsockname(td, s, &sa, &datalen);
1398 	if (error != 0)
1399 		goto bad;
1400 	sa_family = sa->sa_family;
1401 	free(sa, M_SONAME);
1402 
1403 	if (flags & LINUX_MSG_OOB) {
1404 		error = EOPNOTSUPP;
1405 		if (sa_family == AF_UNIX)
1406 			goto bad;
1407 
1408 		error = getsock(td, s, &cap_send_rights, &fp);
1409 		if (error != 0)
1410 			goto bad;
1411 		so = fp->f_data;
1412 		if (so->so_type != SOCK_STREAM)
1413 			error = EOPNOTSUPP;
1414 		fdrop(fp, td);
1415 		if (error != 0)
1416 			goto bad;
1417 	}
1418 
1419 	if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) {
1420 		error = ENOBUFS;
1421 		control = m_get(M_WAITOK, MT_CONTROL);
1422 		MCLGET(control, M_WAITOK);
1423 		data = mtod(control, void *);
1424 		datalen = 0;
1425 
1426 		ptr_cmsg = PTRIN(linux_msghdr.msg_control);
1427 		clen = linux_msghdr.msg_controllen;
1428 		do {
1429 			error = copyin(ptr_cmsg, &linux_cmsg,
1430 			    sizeof(struct l_cmsghdr));
1431 			if (error != 0)
1432 				goto bad;
1433 
1434 			error = EINVAL;
1435 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1436 			    linux_cmsg.cmsg_len > clen)
1437 				goto bad;
1438 
1439 			if (datalen + CMSG_HDRSZ > MCLBYTES)
1440 				goto bad;
1441 
1442 			/*
1443 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1444 			 * so return EINVAL in any other cmsg_type
1445 			 */
1446 			cmsg = data;
1447 			cmsg->cmsg_type =
1448 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1449 			cmsg->cmsg_level =
1450 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1451 			if (cmsg->cmsg_type == -1
1452 			    || cmsg->cmsg_level != SOL_SOCKET) {
1453 				linux_msg(curthread,
1454 				    "unsupported sendmsg cmsg level %d type %d",
1455 				    linux_cmsg.cmsg_level, linux_cmsg.cmsg_type);
1456 				goto bad;
1457 			}
1458 
1459 			/*
1460 			 * Some applications (e.g. pulseaudio) attempt to
1461 			 * send ancillary data even if the underlying protocol
1462 			 * doesn't support it which is not allowed in the
1463 			 * FreeBSD system call interface.
1464 			 */
1465 			if (sa_family != AF_UNIX)
1466 				goto next;
1467 
1468 			if (cmsg->cmsg_type == SCM_CREDS) {
1469 				len = sizeof(struct cmsgcred);
1470 				if (datalen + CMSG_SPACE(len) > MCLBYTES)
1471 					goto bad;
1472 
1473 				/*
1474 				 * The lower levels will fill in the structure
1475 				 */
1476 				memset(CMSG_DATA(data), 0, len);
1477 			} else {
1478 				len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1479 				if (datalen + CMSG_SPACE(len) < datalen ||
1480 				    datalen + CMSG_SPACE(len) > MCLBYTES)
1481 					goto bad;
1482 
1483 				error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1484 				    CMSG_DATA(data), len);
1485 				if (error != 0)
1486 					goto bad;
1487 			}
1488 
1489 			cmsg->cmsg_len = CMSG_LEN(len);
1490 			data = (char *)data + CMSG_SPACE(len);
1491 			datalen += CMSG_SPACE(len);
1492 
1493 next:
1494 			if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1495 				break;
1496 
1497 			clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1498 			ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1499 			    LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1500 		} while(clen >= sizeof(struct l_cmsghdr));
1501 
1502 		control->m_len = datalen;
1503 		if (datalen == 0) {
1504 			m_freem(control);
1505 			control = NULL;
1506 		}
1507 	}
1508 
1509 	msg.msg_iov = iov;
1510 	msg.msg_flags = 0;
1511 	error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1512 	control = NULL;
1513 
1514 bad:
1515 	m_freem(control);
1516 	free(iov, M_IOV);
1517 	return (error);
1518 }
1519 
1520 int
1521 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1522 {
1523 
1524 	return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1525 	    args->flags));
1526 }
1527 
1528 int
1529 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1530 {
1531 	struct l_mmsghdr *msg;
1532 	l_uint retval;
1533 	int error, datagrams;
1534 
1535 	if (args->vlen > UIO_MAXIOV)
1536 		args->vlen = UIO_MAXIOV;
1537 
1538 	msg = PTRIN(args->msg);
1539 	datagrams = 0;
1540 	while (datagrams < args->vlen) {
1541 		error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1542 		    args->flags);
1543 		if (error != 0)
1544 			break;
1545 
1546 		retval = td->td_retval[0];
1547 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1548 		if (error != 0)
1549 			break;
1550 		++msg;
1551 		++datagrams;
1552 	}
1553 	if (error == 0)
1554 		td->td_retval[0] = datagrams;
1555 	return (error);
1556 }
1557 
1558 static int
1559 recvmsg_scm_rights(struct thread *td, l_uint flags, socklen_t *datalen,
1560     void **data, void **udata)
1561 {
1562 	int i, fd, fds, *fdp;
1563 
1564 	if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1565 		fds = *datalen / sizeof(int);
1566 		fdp = *data;
1567 		for (i = 0; i < fds; i++) {
1568 			fd = *fdp++;
1569 			(void)kern_fcntl(td, fd, F_SETFD, FD_CLOEXEC);
1570 		}
1571 	}
1572 	return (0);
1573 }
1574 
1575 
1576 static int
1577 recvmsg_scm_creds(socklen_t *datalen, void **data, void **udata)
1578 {
1579 	struct cmsgcred *cmcred;
1580 	struct l_ucred lu;
1581 
1582 	cmcred = *data;
1583 	lu.pid = cmcred->cmcred_pid;
1584 	lu.uid = cmcred->cmcred_uid;
1585 	lu.gid = cmcred->cmcred_gid;
1586 	memmove(*data, &lu, sizeof(lu));
1587 	*datalen = sizeof(lu);
1588 	return (0);
1589 }
1590 _Static_assert(sizeof(struct cmsgcred) >= sizeof(struct l_ucred),
1591     "scm_creds sizeof l_ucred");
1592 
1593 static int
1594 recvmsg_scm_creds2(socklen_t *datalen, void **data, void **udata)
1595 {
1596 	struct sockcred2 *scred;
1597 	struct l_ucred lu;
1598 
1599 	scred = *data;
1600 	lu.pid = scred->sc_pid;
1601 	lu.uid = scred->sc_uid;
1602 	lu.gid = scred->sc_gid;
1603 	memmove(*data, &lu, sizeof(lu));
1604 	*datalen = sizeof(lu);
1605 	return (0);
1606 }
1607 _Static_assert(sizeof(struct sockcred2) >= sizeof(struct l_ucred),
1608     "scm_creds2 sizeof l_ucred");
1609 
1610 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1611 static int
1612 recvmsg_scm_timestamp(l_int msg_type, socklen_t *datalen, void **data,
1613     void **udata)
1614 {
1615 	l_sock_timeval ltv64;
1616 	l_timeval ltv;
1617 	struct timeval *tv;
1618 	socklen_t len;
1619 	void *buf;
1620 
1621 	if (*datalen != sizeof(struct timeval))
1622 		return (EMSGSIZE);
1623 
1624 	tv = *data;
1625 #if defined(COMPAT_LINUX32)
1626 	if (msg_type == LINUX_SCM_TIMESTAMPO &&
1627 	    (tv->tv_sec > INT_MAX || tv->tv_sec < INT_MIN))
1628 		return (EOVERFLOW);
1629 #endif
1630 	if (msg_type == LINUX_SCM_TIMESTAMPN)
1631 		len = sizeof(ltv64);
1632 	else
1633 		len = sizeof(ltv);
1634 
1635 	buf = malloc(len, M_LINUX, M_WAITOK);
1636 	if (msg_type == LINUX_SCM_TIMESTAMPN) {
1637 		ltv64.tv_sec = tv->tv_sec;
1638 		ltv64.tv_usec = tv->tv_usec;
1639 		memmove(buf, &ltv64, len);
1640 	} else {
1641 		ltv.tv_sec = tv->tv_sec;
1642 		ltv.tv_usec = tv->tv_usec;
1643 		memmove(buf, &ltv, len);
1644 	}
1645 	*data = *udata = buf;
1646 	*datalen = len;
1647 	return (0);
1648 }
1649 #else
1650 _Static_assert(sizeof(struct timeval) == sizeof(l_timeval),
1651     "scm_timestamp sizeof l_timeval");
1652 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1653 
1654 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1655 static int
1656 recvmsg_scm_timestampns(l_int msg_type, socklen_t *datalen, void **data,
1657     void **udata)
1658 {
1659 	struct l_timespec64 ts64;
1660 	struct l_timespec ts32;
1661 	struct timespec ts;
1662 	socklen_t len;
1663 	void *buf;
1664 
1665 	if (msg_type == LINUX_SCM_TIMESTAMPNSO)
1666 		len = sizeof(ts32);
1667 	else
1668 		len = sizeof(ts64);
1669 
1670 	buf = malloc(len, M_LINUX, M_WAITOK);
1671 	bintime2timespec(*data, &ts);
1672 	if (msg_type == LINUX_SCM_TIMESTAMPNSO) {
1673 		ts32.tv_sec = ts.tv_sec;
1674 		ts32.tv_nsec = ts.tv_nsec;
1675 		memmove(buf, &ts32, len);
1676 	} else {
1677 		ts64.tv_sec = ts.tv_sec;
1678 		ts64.tv_nsec = ts.tv_nsec;
1679 		memmove(buf, &ts64, len);
1680 	}
1681 	*data = *udata = buf;
1682 	*datalen = len;
1683 	return (0);
1684 }
1685 #else
1686 static int
1687 recvmsg_scm_timestampns(l_int msg_type, socklen_t *datalen, void **data,
1688     void **udata)
1689 {
1690 	struct timespec ts;
1691 
1692 	bintime2timespec(*data, &ts);
1693 	memmove(*data, &ts, sizeof(struct timespec));
1694 	*datalen = sizeof(struct timespec);
1695 	return (0);
1696 }
1697 _Static_assert(sizeof(struct bintime) >= sizeof(struct timespec),
1698     "scm_timestampns sizeof timespec");
1699 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1700 
1701 static int
1702 recvmsg_scm_sol_socket(struct thread *td, l_int msg_type, l_int lmsg_type,
1703     l_uint flags, socklen_t *datalen, void **data, void **udata)
1704 {
1705 	int error;
1706 
1707 	error = 0;
1708 	switch (msg_type) {
1709 	case SCM_RIGHTS:
1710 		error = recvmsg_scm_rights(td, flags, datalen,
1711 		    data, udata);
1712 		break;
1713 	case SCM_CREDS:
1714 		error = recvmsg_scm_creds(datalen, data, udata);
1715 		break;
1716 	case SCM_CREDS2:
1717 		error = recvmsg_scm_creds2(datalen, data, udata);
1718 		break;
1719 	case SCM_TIMESTAMP:
1720 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1721 		error = recvmsg_scm_timestamp(lmsg_type, datalen,
1722 		    data, udata);
1723 #endif
1724 		break;
1725 	case SCM_BINTIME:
1726 		error = recvmsg_scm_timestampns(lmsg_type, datalen,
1727 		    data, udata);
1728 		break;
1729 	}
1730 
1731 	return (error);
1732 }
1733 
1734 static int
1735 recvmsg_scm_ip_origdstaddr(socklen_t *datalen, void **data, void **udata)
1736 {
1737 	struct l_sockaddr *lsa;
1738 	int error;
1739 
1740 	error = bsd_to_linux_sockaddr(*data, &lsa, *datalen);
1741 	if (error == 0) {
1742 		*data = *udata = lsa;
1743 		*datalen = sizeof(*lsa);
1744 	}
1745 	return (error);
1746 }
1747 
1748 static int
1749 recvmsg_scm_ipproto_ip(l_int msg_type, l_int lmsg_type, socklen_t *datalen,
1750     void **data, void **udata)
1751 {
1752 	int error;
1753 
1754 	error = 0;
1755 	switch (msg_type) {
1756 	case IP_ORIGDSTADDR:
1757 		error = recvmsg_scm_ip_origdstaddr(datalen, data,
1758 		    udata);
1759 		break;
1760 	}
1761 
1762 	return (error);
1763 }
1764 
1765 static int
1766 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1767     l_uint flags, struct msghdr *msg)
1768 {
1769 	struct proc *p = td->td_proc;
1770 	struct cmsghdr *cm;
1771 	struct l_cmsghdr *lcm = NULL;
1772 	socklen_t datalen, maxlen, outlen;
1773 	struct l_msghdr l_msghdr;
1774 	struct iovec *iov, *uiov;
1775 	struct mbuf *m, *control = NULL;
1776 	struct mbuf **controlp;
1777 	struct sockaddr *sa;
1778 	caddr_t outbuf;
1779 	void *data, *udata;
1780 	int error, skiped;
1781 
1782 	error = copyin(msghdr, &l_msghdr, sizeof(l_msghdr));
1783 	if (error != 0)
1784 		return (error);
1785 
1786 	/*
1787 	 * Pass user-supplied recvmsg() flags in msg_flags field,
1788 	 * following sys_recvmsg() convention.
1789 	*/
1790 	l_msghdr.msg_flags = flags;
1791 
1792 	error = linux_to_bsd_msghdr(msg, &l_msghdr);
1793 	if (error != 0)
1794 		return (error);
1795 
1796 #ifdef COMPAT_LINUX32
1797 	error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1798 	    &iov, EMSGSIZE);
1799 #else
1800 	error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1801 #endif
1802 	if (error != 0)
1803 		return (error);
1804 
1805 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1806 		msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN);
1807 		sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK);
1808 		msg->msg_name = sa;
1809 	} else {
1810 		sa = NULL;
1811 		msg->msg_name = NULL;
1812 	}
1813 
1814 	uiov = msg->msg_iov;
1815 	msg->msg_iov = iov;
1816 	controlp = (msg->msg_control != NULL) ? &control : NULL;
1817 	error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp);
1818 	msg->msg_iov = uiov;
1819 	if (error != 0)
1820 		goto bad;
1821 
1822 	/*
1823 	 * Note that kern_recvit() updates msg->msg_namelen.
1824 	 */
1825 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1826 		msg->msg_name = PTRIN(l_msghdr.msg_name);
1827 		error = linux_copyout_sockaddr(sa, msg->msg_name,
1828 		    msg->msg_namelen);
1829 		if (error != 0)
1830 			goto bad;
1831 	}
1832 
1833 	error = bsd_to_linux_msghdr(msg, &l_msghdr);
1834 	if (error != 0)
1835 		goto bad;
1836 
1837 	skiped = outlen = 0;
1838 	maxlen = l_msghdr.msg_controllen;
1839 	if (control == NULL)
1840 		goto out;
1841 
1842 	lcm = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1843 	msg->msg_control = mtod(control, struct cmsghdr *);
1844 	msg->msg_controllen = control->m_len;
1845 	outbuf = PTRIN(l_msghdr.msg_control);
1846 	for (m = control; m != NULL; m = m->m_next) {
1847 		cm = mtod(m, struct cmsghdr *);
1848 		lcm->cmsg_type = bsd_to_linux_cmsg_type(p, cm->cmsg_type,
1849 		    cm->cmsg_level);
1850 		lcm->cmsg_level = bsd_to_linux_sockopt_level(cm->cmsg_level);
1851 
1852 		if (lcm->cmsg_type == -1 ||
1853 		    cm->cmsg_level == -1) {
1854 			LINUX_RATELIMIT_MSG_OPT2(
1855 			    "unsupported recvmsg cmsg level %d type %d",
1856 			    cm->cmsg_level, cm->cmsg_type);
1857 			/* Skip unsupported messages */
1858 			skiped++;
1859 			continue;
1860 		}
1861 		data = CMSG_DATA(cm);
1862 		datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1863 		udata = NULL;
1864 		error = 0;
1865 
1866 		switch (cm->cmsg_level) {
1867 		case IPPROTO_IP:
1868 			error = recvmsg_scm_ipproto_ip(cm->cmsg_type,
1869 			    lcm->cmsg_type, &datalen, &data, &udata);
1870  			break;
1871 		case SOL_SOCKET:
1872 			error = recvmsg_scm_sol_socket(td, cm->cmsg_type,
1873 			    lcm->cmsg_type, flags, &datalen, &data, &udata);
1874  			break;
1875  		}
1876 
1877 		/* The recvmsg_scm_ is responsible to free udata on error. */
1878 		if (error != 0)
1879 			goto bad;
1880 
1881 		if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1882 			if (outlen == 0) {
1883 				error = EMSGSIZE;
1884 				goto err;
1885 			} else {
1886 				l_msghdr.msg_flags |= LINUX_MSG_CTRUNC;
1887 				m_dispose_extcontrolm(control);
1888 				free(udata, M_LINUX);
1889 				goto out;
1890 			}
1891 		}
1892 
1893 		lcm->cmsg_len = LINUX_CMSG_LEN(datalen);
1894 		error = copyout(lcm, outbuf, L_CMSG_HDRSZ);
1895 		if (error == 0) {
1896 			error = copyout(data, LINUX_CMSG_DATA(outbuf), datalen);
1897 			if (error == 0) {
1898 				outbuf += LINUX_CMSG_SPACE(datalen);
1899 				outlen += LINUX_CMSG_SPACE(datalen);
1900 			}
1901 		}
1902 err:
1903 		free(udata, M_LINUX);
1904 		if (error != 0)
1905 			goto bad;
1906 	}
1907 	if (outlen == 0 && skiped > 0) {
1908 		error = EINVAL;
1909 		goto bad;
1910 	}
1911 
1912 out:
1913 	l_msghdr.msg_controllen = outlen;
1914 	error = copyout(&l_msghdr, msghdr, sizeof(l_msghdr));
1915 
1916 bad:
1917 	if (control != NULL) {
1918 		if (error != 0)
1919 			m_dispose_extcontrolm(control);
1920 		m_freem(control);
1921 	}
1922 	free(iov, M_IOV);
1923 	free(lcm, M_LINUX);
1924 	free(sa, M_SONAME);
1925 
1926 	return (error);
1927 }
1928 
1929 int
1930 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1931 {
1932 	struct msghdr bsd_msg;
1933 	struct file *fp;
1934 	int error;
1935 
1936 	error = getsock(td, args->s, &cap_recv_rights, &fp);
1937 	if (error != 0)
1938 		return (error);
1939 	fdrop(fp, td);
1940 	return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1941 	    args->flags, &bsd_msg));
1942 }
1943 
1944 static int
1945 linux_recvmmsg_common(struct thread *td, l_int s, struct l_mmsghdr *msg,
1946     l_uint vlen, l_uint flags, struct timespec *tts)
1947 {
1948 	struct msghdr bsd_msg;
1949 	struct timespec ts;
1950 	struct file *fp;
1951 	l_uint retval;
1952 	int error, datagrams;
1953 
1954 	error = getsock(td, s, &cap_recv_rights, &fp);
1955 	if (error != 0)
1956 		return (error);
1957 	datagrams = 0;
1958 	while (datagrams < vlen) {
1959 		error = linux_recvmsg_common(td, s, &msg->msg_hdr,
1960 		    flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1961 		if (error != 0)
1962 			break;
1963 
1964 		retval = td->td_retval[0];
1965 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1966 		if (error != 0)
1967 			break;
1968 		++msg;
1969 		++datagrams;
1970 
1971 		/*
1972 		 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1973 		 */
1974 		if (flags & LINUX_MSG_WAITFORONE)
1975 			flags |= LINUX_MSG_DONTWAIT;
1976 
1977 		/*
1978 		 * See BUGS section of recvmmsg(2).
1979 		 */
1980 		if (tts) {
1981 			getnanotime(&ts);
1982 			timespecsub(&ts, tts, &ts);
1983 			if (!timespecisset(&ts) || ts.tv_sec > 0)
1984 				break;
1985 		}
1986 		/* Out of band data, return right away. */
1987 		if (bsd_msg.msg_flags & MSG_OOB)
1988 			break;
1989 	}
1990 	if (error == 0)
1991 		td->td_retval[0] = datagrams;
1992 	fdrop(fp, td);
1993 	return (error);
1994 }
1995 
1996 int
1997 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1998 {
1999 	struct timespec ts, tts, *ptts;
2000 	int error;
2001 
2002 	if (args->timeout) {
2003 		error = linux_get_timespec(&ts, args->timeout);
2004 		if (error != 0)
2005 			return (error);
2006 		getnanotime(&tts);
2007 		timespecadd(&tts, &ts, &tts);
2008 		ptts = &tts;
2009 	}
2010 		else ptts = NULL;
2011 
2012 	return (linux_recvmmsg_common(td, args->s, PTRIN(args->msg),
2013 	    args->vlen, args->flags, ptts));
2014 }
2015 
2016 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2017 int
2018 linux_recvmmsg_time64(struct thread *td, struct linux_recvmmsg_time64_args *args)
2019 {
2020 	struct timespec ts, tts, *ptts;
2021 	int error;
2022 
2023 	if (args->timeout) {
2024 		error = linux_get_timespec64(&ts, args->timeout);
2025 		if (error != 0)
2026 			return (error);
2027 		getnanotime(&tts);
2028 		timespecadd(&tts, &ts, &tts);
2029 		ptts = &tts;
2030 	}
2031 		else ptts = NULL;
2032 
2033 	return (linux_recvmmsg_common(td, args->s, PTRIN(args->msg),
2034 	    args->vlen, args->flags, ptts));
2035 }
2036 #endif
2037 
2038 int
2039 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
2040 {
2041 
2042 	return (kern_shutdown(td, args->s, args->how));
2043 }
2044 
2045 int
2046 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
2047 {
2048 	struct proc *p = td->td_proc;
2049 	struct linux_pemuldata *pem;
2050 	l_timeval linux_tv;
2051 	struct sockaddr *sa;
2052 	struct timeval tv;
2053 	socklen_t len;
2054 	int error, level, name, val;
2055 
2056 	level = linux_to_bsd_sockopt_level(args->level);
2057 	switch (level) {
2058 	case SOL_SOCKET:
2059 		name = linux_to_bsd_so_sockopt(args->optname);
2060 		switch (name) {
2061 		case LOCAL_CREDS_PERSISTENT:
2062 			level = SOL_LOCAL;
2063 			break;
2064 		case SO_RCVTIMEO:
2065 			/* FALLTHROUGH */
2066 		case SO_SNDTIMEO:
2067 			error = copyin(PTRIN(args->optval), &linux_tv,
2068 			    sizeof(linux_tv));
2069 			if (error != 0)
2070 				return (error);
2071 			tv.tv_sec = linux_tv.tv_sec;
2072 			tv.tv_usec = linux_tv.tv_usec;
2073 			return (kern_setsockopt(td, args->s, level,
2074 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
2075 			/* NOTREACHED */
2076 		case SO_TIMESTAMP:
2077 			/* overwrite SO_BINTIME */
2078 			val = 0;
2079 			error = kern_setsockopt(td, args->s, level,
2080 			    SO_BINTIME, &val, UIO_SYSSPACE, sizeof(val));
2081 			if (error != 0)
2082 				return (error);
2083 			pem = pem_find(p);
2084 			pem->so_timestamp = args->optname;
2085 			break;
2086 		case SO_BINTIME:
2087 			/* overwrite SO_TIMESTAMP */
2088 			val = 0;
2089 			error = kern_setsockopt(td, args->s, level,
2090 			    SO_TIMESTAMP, &val, UIO_SYSSPACE, sizeof(val));
2091 			if (error != 0)
2092 				return (error);
2093 			pem = pem_find(p);
2094 			pem->so_timestampns = args->optname;
2095 			break;
2096 		default:
2097 			break;
2098 		}
2099 		break;
2100 	case IPPROTO_IP:
2101 		if (args->optname == LINUX_IP_RECVERR &&
2102 		    linux_ignore_ip_recverr) {
2103 			/*
2104 			 * XXX: This is a hack to unbreak DNS resolution
2105 			 *	with glibc 2.30 and above.
2106 			 */
2107 			return (0);
2108 		}
2109 		name = linux_to_bsd_ip_sockopt(args->optname);
2110 		break;
2111 	case IPPROTO_IPV6:
2112 		name = linux_to_bsd_ip6_sockopt(args->optname);
2113 		break;
2114 	case IPPROTO_TCP:
2115 		name = linux_to_bsd_tcp_sockopt(args->optname);
2116 		break;
2117 	case SOL_NETLINK:
2118 		level = SOL_SOCKET;
2119 		name = args->optname;
2120 		break;
2121 	default:
2122 		name = -1;
2123 		break;
2124 	}
2125 	if (name < 0) {
2126 		if (name == -1)
2127 			linux_msg(curthread,
2128 			    "unsupported setsockopt level %d optname %d",
2129 			    args->level, args->optname);
2130 		return (ENOPROTOOPT);
2131 	}
2132 
2133 	if (name == IPV6_NEXTHOP) {
2134 		len = args->optlen;
2135 		error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len);
2136 		if (error != 0)
2137 			return (error);
2138 
2139 		error = kern_setsockopt(td, args->s, level,
2140 		    name, sa, UIO_SYSSPACE, len);
2141 		free(sa, M_SONAME);
2142 	} else {
2143 		error = kern_setsockopt(td, args->s, level,
2144 		    name, PTRIN(args->optval), UIO_USERSPACE, args->optlen);
2145 	}
2146 
2147 	return (error);
2148 }
2149 
2150 static int
2151 linux_sockopt_copyout(struct thread *td, void *val, socklen_t len,
2152     struct linux_getsockopt_args *args)
2153 {
2154 	int error;
2155 
2156 	error = copyout(val, PTRIN(args->optval), len);
2157 	if (error == 0)
2158 		error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2159 	return (error);
2160 }
2161 
2162 static int
2163 linux_getsockopt_so_peergroups(struct thread *td,
2164     struct linux_getsockopt_args *args)
2165 {
2166 	struct xucred xu;
2167 	socklen_t xulen, len;
2168 	int error, i;
2169 
2170 	xulen = sizeof(xu);
2171 	error = kern_getsockopt(td, args->s, 0,
2172 	    LOCAL_PEERCRED, &xu, UIO_SYSSPACE, &xulen);
2173 	if (error != 0)
2174 		return (error);
2175 
2176 	len = xu.cr_ngroups * sizeof(l_gid_t);
2177 	if (args->optlen < len) {
2178 		error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2179 		if (error == 0)
2180 			error = ERANGE;
2181 		return (error);
2182 	}
2183 
2184 	/*
2185 	 * "- 1" to skip the primary group.
2186 	 */
2187 	for (i = 0; i < xu.cr_ngroups - 1; i++) {
2188 		error = copyout(xu.cr_groups + i + 1,
2189 		    (void *)(args->optval + i * sizeof(l_gid_t)),
2190 		    sizeof(l_gid_t));
2191 		if (error != 0)
2192 			return (error);
2193 	}
2194 
2195 	error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2196 	return (error);
2197 }
2198 
2199 static int
2200 linux_getsockopt_so_peersec(struct thread *td,
2201     struct linux_getsockopt_args *args)
2202 {
2203 	socklen_t len;
2204 	int error;
2205 
2206 	len = sizeof(SECURITY_CONTEXT_STRING);
2207 	if (args->optlen < len) {
2208 		error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2209 		if (error == 0)
2210 			error = ERANGE;
2211 		return (error);
2212 	}
2213 
2214 	return (linux_sockopt_copyout(td, SECURITY_CONTEXT_STRING,
2215 	    len, args));
2216 }
2217 
2218 static int
2219 linux_getsockopt_so_linger(struct thread *td,
2220     struct linux_getsockopt_args *args)
2221 {
2222 	struct linger ling;
2223 	socklen_t len;
2224 	int error;
2225 
2226 	len = sizeof(ling);
2227 	error = kern_getsockopt(td, args->s, SOL_SOCKET,
2228 	    SO_LINGER, &ling, UIO_SYSSPACE, &len);
2229 	if (error != 0)
2230 		return (error);
2231 	ling.l_onoff = ((ling.l_onoff & SO_LINGER) != 0);
2232 	return (linux_sockopt_copyout(td, &ling, len, args));
2233 }
2234 
2235 int
2236 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
2237 {
2238 	l_timeval linux_tv;
2239 	struct timeval tv;
2240 	socklen_t tv_len, xulen, len;
2241 	struct sockaddr *sa;
2242 	struct xucred xu;
2243 	struct l_ucred lxu;
2244 	int error, level, name, newval;
2245 
2246 	level = linux_to_bsd_sockopt_level(args->level);
2247 	switch (level) {
2248 	case SOL_SOCKET:
2249 		switch (args->optname) {
2250 		case LINUX_SO_PEERGROUPS:
2251 			return (linux_getsockopt_so_peergroups(td, args));
2252 		case LINUX_SO_PEERSEC:
2253 			return (linux_getsockopt_so_peersec(td, args));
2254 		default:
2255 			break;
2256 		}
2257 
2258 		name = linux_to_bsd_so_sockopt(args->optname);
2259 		switch (name) {
2260 		case LOCAL_CREDS_PERSISTENT:
2261 			level = SOL_LOCAL;
2262 			break;
2263 		case SO_RCVTIMEO:
2264 			/* FALLTHROUGH */
2265 		case SO_SNDTIMEO:
2266 			tv_len = sizeof(tv);
2267 			error = kern_getsockopt(td, args->s, level,
2268 			    name, &tv, UIO_SYSSPACE, &tv_len);
2269 			if (error != 0)
2270 				return (error);
2271 			linux_tv.tv_sec = tv.tv_sec;
2272 			linux_tv.tv_usec = tv.tv_usec;
2273 			return (linux_sockopt_copyout(td, &linux_tv,
2274 			    sizeof(linux_tv), args));
2275 			/* NOTREACHED */
2276 		case LOCAL_PEERCRED:
2277 			if (args->optlen < sizeof(lxu))
2278 				return (EINVAL);
2279 			/*
2280 			 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
2281 			 * but by the Unix socket's level 0.
2282 			 */
2283 			level = 0;
2284 			xulen = sizeof(xu);
2285 			error = kern_getsockopt(td, args->s, level,
2286 			    name, &xu, UIO_SYSSPACE, &xulen);
2287 			if (error != 0)
2288 				return (error);
2289 			lxu.pid = xu.cr_pid;
2290 			lxu.uid = xu.cr_uid;
2291 			lxu.gid = xu.cr_gid;
2292 			return (linux_sockopt_copyout(td, &lxu,
2293 			    sizeof(lxu), args));
2294 			/* NOTREACHED */
2295 		case SO_ERROR:
2296 			len = sizeof(newval);
2297 			error = kern_getsockopt(td, args->s, level,
2298 			    name, &newval, UIO_SYSSPACE, &len);
2299 			if (error != 0)
2300 				return (error);
2301 			newval = -bsd_to_linux_errno(newval);
2302 			return (linux_sockopt_copyout(td, &newval,
2303 			    len, args));
2304 			/* NOTREACHED */
2305 		case SO_DOMAIN:
2306 			len = sizeof(newval);
2307 			error = kern_getsockopt(td, args->s, level,
2308 			    name, &newval, UIO_SYSSPACE, &len);
2309 			if (error != 0)
2310 				return (error);
2311 			newval = bsd_to_linux_domain(newval);
2312 			if (newval == -1)
2313 				return (ENOPROTOOPT);
2314 			return (linux_sockopt_copyout(td, &newval,
2315 			    len, args));
2316 			/* NOTREACHED */
2317 		case SO_LINGER:
2318 			return (linux_getsockopt_so_linger(td, args));
2319 			/* NOTREACHED */
2320 		default:
2321 			break;
2322 		}
2323 		break;
2324 	case IPPROTO_IP:
2325 		name = linux_to_bsd_ip_sockopt(args->optname);
2326 		break;
2327 	case IPPROTO_IPV6:
2328 		name = linux_to_bsd_ip6_sockopt(args->optname);
2329 		break;
2330 	case IPPROTO_TCP:
2331 		name = linux_to_bsd_tcp_sockopt(args->optname);
2332 		break;
2333 	default:
2334 		name = -1;
2335 		break;
2336 	}
2337 	if (name < 0) {
2338 		if (name == -1)
2339 			linux_msg(curthread,
2340 			    "unsupported getsockopt level %d optname %d",
2341 			    args->level, args->optname);
2342 		return (EINVAL);
2343 	}
2344 
2345 	if (name == IPV6_NEXTHOP) {
2346 		error = copyin(PTRIN(args->optlen), &len, sizeof(len));
2347                 if (error != 0)
2348                         return (error);
2349 		sa = malloc(len, M_SONAME, M_WAITOK);
2350 
2351 		error = kern_getsockopt(td, args->s, level,
2352 		    name, sa, UIO_SYSSPACE, &len);
2353 		if (error != 0)
2354 			goto out;
2355 
2356 		error = linux_copyout_sockaddr(sa, PTRIN(args->optval), len);
2357 		if (error == 0)
2358 			error = copyout(&len, PTRIN(args->optlen),
2359 			    sizeof(len));
2360 out:
2361 		free(sa, M_SONAME);
2362 	} else {
2363 		if (args->optval) {
2364 			error = copyin(PTRIN(args->optlen), &len, sizeof(len));
2365 			if (error != 0)
2366 				return (error);
2367 		}
2368 		error = kern_getsockopt(td, args->s, level,
2369 		    name, PTRIN(args->optval), UIO_USERSPACE, &len);
2370 		if (error == 0)
2371 			error = copyout(&len, PTRIN(args->optlen),
2372 			    sizeof(len));
2373 	}
2374 
2375 	return (error);
2376 }
2377 
2378 /*
2379  * Based on sendfile_getsock from kern_sendfile.c
2380  * Determines whether an fd is a stream socket that can be used
2381  * with FreeBSD sendfile.
2382  */
2383 static bool
2384 is_stream_socket(struct file *fp)
2385 {
2386 	struct socket *so;
2387 
2388 	/*
2389 	 * The socket must be a stream socket and connected.
2390 	 */
2391 	if (fp->f_type != DTYPE_SOCKET)
2392 		return (false);
2393 	so = fp->f_data;
2394 	if (so->so_type != SOCK_STREAM)
2395 		return (false);
2396 	/*
2397 	 * SCTP one-to-one style sockets currently don't work with
2398 	 * sendfile().
2399 	 */
2400 	if (so->so_proto->pr_protocol == IPPROTO_SCTP)
2401 		return (false);
2402 	return (!SOLISTENING(so));
2403 }
2404 
2405 static bool
2406 is_regular_file(struct file *fp)
2407 {
2408 
2409 	return (fp->f_type == DTYPE_VNODE && fp->f_vnode != NULL &&
2410 	    fp->f_vnode->v_type == VREG);
2411 }
2412 
2413 static int
2414 sendfile_fallback(struct thread *td, struct file *fp, l_int out,
2415     off_t *offset, l_size_t count, off_t *sbytes)
2416 {
2417 	off_t current_offset, out_offset, to_send;
2418 	l_size_t bytes_sent, n_read;
2419 	struct file *ofp;
2420 	struct iovec aiov;
2421 	struct uio auio;
2422 	bool seekable;
2423 	size_t bufsz;
2424 	void *buf;
2425 	int flags, error;
2426 
2427 	if (offset == NULL) {
2428 		if ((error = fo_seek(fp, 0, SEEK_CUR, td)) != 0)
2429 			return (error);
2430 		current_offset = td->td_uretoff.tdu_off;
2431 	} else {
2432 		if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0)
2433 			return (ESPIPE);
2434 		current_offset = *offset;
2435 	}
2436 	error = fget_write(td, out, &cap_pwrite_rights, &ofp);
2437 	if (error != 0)
2438 		return (error);
2439 	seekable = (ofp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0;
2440 	if (seekable) {
2441 		if ((error = fo_seek(ofp, 0, SEEK_CUR, td)) != 0)
2442 			goto drop;
2443 		out_offset = td->td_uretoff.tdu_off;
2444 	} else
2445 		out_offset = 0;
2446 
2447 	flags = FOF_OFFSET | FOF_NOUPDATE;
2448 	bufsz = min(count, MAXPHYS);
2449 	buf = malloc(bufsz, M_LINUX, M_WAITOK);
2450 	bytes_sent = 0;
2451 	while (bytes_sent < count) {
2452 		to_send = min(count - bytes_sent, bufsz);
2453 		aiov.iov_base = buf;
2454 		aiov.iov_len = bufsz;
2455 		auio.uio_iov = &aiov;
2456 		auio.uio_iovcnt = 1;
2457 		auio.uio_segflg = UIO_SYSSPACE;
2458 		auio.uio_td = td;
2459 		auio.uio_rw = UIO_READ;
2460 		auio.uio_offset = current_offset;
2461 		auio.uio_resid = to_send;
2462 		error = fo_read(fp, &auio, fp->f_cred, flags, td);
2463 		if (error != 0)
2464 			break;
2465 		n_read = to_send - auio.uio_resid;
2466 		if (n_read == 0)
2467 			break;
2468 		aiov.iov_base = buf;
2469 		aiov.iov_len = bufsz;
2470 		auio.uio_iov = &aiov;
2471 		auio.uio_iovcnt = 1;
2472 		auio.uio_segflg = UIO_SYSSPACE;
2473 		auio.uio_td = td;
2474 		auio.uio_rw = UIO_WRITE;
2475 		auio.uio_offset = (seekable) ? out_offset : 0;
2476 		auio.uio_resid = n_read;
2477 		error = fo_write(ofp, &auio, ofp->f_cred, flags, td);
2478 		if (error != 0)
2479 			break;
2480 		bytes_sent += n_read;
2481 		current_offset += n_read;
2482 		out_offset += n_read;
2483 	}
2484 	free(buf, M_LINUX);
2485 
2486 	if (error == 0) {
2487 		*sbytes = bytes_sent;
2488 		if (offset != NULL)
2489 			*offset = current_offset;
2490 		else
2491 			error = fo_seek(fp, current_offset, SEEK_SET, td);
2492 	}
2493 	if (error == 0 && seekable)
2494 		error = fo_seek(ofp, out_offset, SEEK_SET, td);
2495 
2496 drop:
2497 	fdrop(ofp, td);
2498 	return (error);
2499 }
2500 
2501 static int
2502 sendfile_sendfile(struct thread *td, struct file *fp, l_int out,
2503     off_t *offset, l_size_t count, off_t *sbytes)
2504 {
2505 	off_t current_offset;
2506 	int error;
2507 
2508 	if (offset == NULL) {
2509 		if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0)
2510 			return (ESPIPE);
2511 		if ((error = fo_seek(fp, 0, SEEK_CUR, td)) != 0)
2512 			return (error);
2513 		current_offset = td->td_uretoff.tdu_off;
2514 	} else
2515 		current_offset = *offset;
2516 	error = fo_sendfile(fp, out, NULL, NULL, current_offset, count,
2517 	    sbytes, 0, td);
2518 	if (error == 0) {
2519 		current_offset += *sbytes;
2520 		if (offset != NULL)
2521 			*offset = current_offset;
2522 		else
2523 			error = fo_seek(fp, current_offset, SEEK_SET, td);
2524 	}
2525 	return (error);
2526 }
2527 
2528 static int
2529 linux_sendfile_common(struct thread *td, l_int out, l_int in,
2530     off_t *offset, l_size_t count)
2531 {
2532 	struct file *fp, *ofp;
2533 	off_t sbytes;
2534 	int error;
2535 
2536 	/* Linux cannot have 0 count. */
2537 	if (count <= 0 || (offset != NULL && *offset < 0))
2538 		return (EINVAL);
2539 
2540 	AUDIT_ARG_FD(in);
2541 	error = fget_read(td, in, &cap_pread_rights, &fp);
2542 	if (error != 0)
2543 		return (error);
2544 	if ((fp->f_type != DTYPE_SHM && fp->f_type != DTYPE_VNODE) ||
2545 	    (fp->f_type == DTYPE_VNODE &&
2546 	    (fp->f_vnode == NULL || fp->f_vnode->v_type != VREG))) {
2547 		error = EINVAL;
2548 		goto drop;
2549 	}
2550 	error = fget_unlocked(td, out, &cap_no_rights, &ofp);
2551 	if (error != 0)
2552 		goto drop;
2553 
2554 	if (is_regular_file(fp) && is_regular_file(ofp)) {
2555 		error = kern_copy_file_range(td, in, offset, out, NULL, count,
2556 		    0);
2557 	} else {
2558 		sbytes = 0;
2559 		if (is_stream_socket(ofp))
2560 			error = sendfile_sendfile(td, fp, out, offset, count,
2561 			    &sbytes);
2562 		else
2563 			error = sendfile_fallback(td, fp, out, offset, count,
2564 			    &sbytes);
2565 		if (error == 0)
2566 			td->td_retval[0] = sbytes;
2567 	}
2568 	fdrop(ofp, td);
2569 
2570 drop:
2571 	fdrop(fp, td);
2572 	return (error);
2573 }
2574 
2575 int
2576 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg)
2577 {
2578 	/*
2579 	 * Differences between FreeBSD and Linux sendfile:
2580 	 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to
2581 	 *   mean send the whole file).
2582 	 * - Linux can send to any fd whereas FreeBSD only supports sockets.
2583 	 *   We therefore use FreeBSD sendfile where possible for performance,
2584 	 *   but fall back on a manual copy (sendfile_fallback).
2585 	 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr.
2586 	 * - Linux takes an offset pointer and updates it to the read location.
2587 	 *   FreeBSD takes in an offset and a 'bytes read' parameter which is
2588 	 *   only filled if it isn't NULL.  We use this parameter to update the
2589 	 *   offset pointer if it exists.
2590 	 * - Linux sendfile returns bytes read on success while FreeBSD
2591 	 *   returns 0.  We use the 'bytes read' parameter to get this value.
2592 	 */
2593 
2594 	off_t offset64;
2595 	l_off_t offset;
2596 	int error;
2597 
2598 	if (arg->offset != NULL) {
2599 		error = copyin(arg->offset, &offset, sizeof(offset));
2600 		if (error != 0)
2601 			return (error);
2602 		offset64 = offset;
2603 	}
2604 
2605 	error = linux_sendfile_common(td, arg->out, arg->in,
2606 	    arg->offset != NULL ? &offset64 : NULL, arg->count);
2607 
2608 	if (error == 0 && arg->offset != NULL) {
2609 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2610 		if (offset64 > INT32_MAX)
2611 			return (EOVERFLOW);
2612 #endif
2613 		offset = (l_off_t)offset64;
2614 		error = copyout(&offset, arg->offset, sizeof(offset));
2615 	}
2616 
2617 	return (error);
2618 }
2619 
2620 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2621 int
2622 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg)
2623 {
2624 	off_t offset;
2625 	int error;
2626 
2627 	if (arg->offset != NULL) {
2628 		error = copyin(arg->offset, &offset, sizeof(offset));
2629 		if (error != 0)
2630 			return (error);
2631 	}
2632 
2633 	error = linux_sendfile_common(td, arg->out, arg->in,
2634 		arg->offset != NULL ? &offset : NULL, arg->count);
2635 
2636 	if (error == 0 && arg->offset != NULL)
2637 		error = copyout(&offset, arg->offset, sizeof(offset));
2638 
2639 	return (error);
2640 }
2641 
2642 /* Argument list sizes for linux_socketcall */
2643 static const unsigned char lxs_args_cnt[] = {
2644 	0 /* unused*/,		3 /* socket */,
2645 	3 /* bind */,		3 /* connect */,
2646 	2 /* listen */,		3 /* accept */,
2647 	3 /* getsockname */,	3 /* getpeername */,
2648 	4 /* socketpair */,	4 /* send */,
2649 	4 /* recv */,		6 /* sendto */,
2650 	6 /* recvfrom */,	2 /* shutdown */,
2651 	5 /* setsockopt */,	5 /* getsockopt */,
2652 	3 /* sendmsg */,	3 /* recvmsg */,
2653 	4 /* accept4 */,	5 /* recvmmsg */,
2654 	4 /* sendmmsg */,	4 /* sendfile */
2655 };
2656 #define	LINUX_ARGS_CNT		(nitems(lxs_args_cnt) - 1)
2657 #define	LINUX_ARG_SIZE(x)	(lxs_args_cnt[x] * sizeof(l_ulong))
2658 
2659 int
2660 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
2661 {
2662 	l_ulong a[6];
2663 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2664 	register_t l_args[6];
2665 #endif
2666 	void *arg;
2667 	int error;
2668 
2669 	if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
2670 		return (EINVAL);
2671 	error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
2672 	if (error != 0)
2673 		return (error);
2674 
2675 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2676 	for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
2677 		l_args[i] = a[i];
2678 	arg = l_args;
2679 #else
2680 	arg = a;
2681 #endif
2682 	switch (args->what) {
2683 	case LINUX_SOCKET:
2684 		return (linux_socket(td, arg));
2685 	case LINUX_BIND:
2686 		return (linux_bind(td, arg));
2687 	case LINUX_CONNECT:
2688 		return (linux_connect(td, arg));
2689 	case LINUX_LISTEN:
2690 		return (linux_listen(td, arg));
2691 	case LINUX_ACCEPT:
2692 		return (linux_accept(td, arg));
2693 	case LINUX_GETSOCKNAME:
2694 		return (linux_getsockname(td, arg));
2695 	case LINUX_GETPEERNAME:
2696 		return (linux_getpeername(td, arg));
2697 	case LINUX_SOCKETPAIR:
2698 		return (linux_socketpair(td, arg));
2699 	case LINUX_SEND:
2700 		return (linux_send(td, arg));
2701 	case LINUX_RECV:
2702 		return (linux_recv(td, arg));
2703 	case LINUX_SENDTO:
2704 		return (linux_sendto(td, arg));
2705 	case LINUX_RECVFROM:
2706 		return (linux_recvfrom(td, arg));
2707 	case LINUX_SHUTDOWN:
2708 		return (linux_shutdown(td, arg));
2709 	case LINUX_SETSOCKOPT:
2710 		return (linux_setsockopt(td, arg));
2711 	case LINUX_GETSOCKOPT:
2712 		return (linux_getsockopt(td, arg));
2713 	case LINUX_SENDMSG:
2714 		return (linux_sendmsg(td, arg));
2715 	case LINUX_RECVMSG:
2716 		return (linux_recvmsg(td, arg));
2717 	case LINUX_ACCEPT4:
2718 		return (linux_accept4(td, arg));
2719 	case LINUX_RECVMMSG:
2720 		return (linux_recvmmsg(td, arg));
2721 	case LINUX_SENDMMSG:
2722 		return (linux_sendmmsg(td, arg));
2723 	case LINUX_SENDFILE:
2724 		return (linux_sendfile(td, arg));
2725 	}
2726 
2727 	linux_msg(td, "socket type %d not implemented", args->what);
2728 	return (ENOSYS);
2729 }
2730 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2731