xref: /freebsd/sys/compat/linux/linux_socket.c (revision 36d6566e5985030fd2f1100bd9c1387bbe0bd290)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 1995 Søren Schmidt
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 /* XXX we use functions that might not exist. */
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 
36 #include <sys/param.h>
37 #include <sys/proc.h>
38 #include <sys/systm.h>
39 #include <sys/sysproto.h>
40 #include <sys/capsicum.h>
41 #include <sys/fcntl.h>
42 #include <sys/file.h>
43 #include <sys/filedesc.h>
44 #include <sys/limits.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mutex.h>
48 #include <sys/mbuf.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/syscallsubr.h>
52 #include <sys/uio.h>
53 #include <sys/stat.h>
54 #include <sys/syslog.h>
55 #include <sys/un.h>
56 #include <sys/unistd.h>
57 
58 #include <security/audit/audit.h>
59 
60 #include <net/if.h>
61 #include <net/vnet.h>
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/ip.h>
65 #include <netinet/tcp.h>
66 #ifdef INET6
67 #include <netinet/ip6.h>
68 #include <netinet6/ip6_var.h>
69 #endif
70 
71 #ifdef COMPAT_LINUX32
72 #include <machine/../linux32/linux.h>
73 #include <machine/../linux32/linux32_proto.h>
74 #else
75 #include <machine/../linux/linux.h>
76 #include <machine/../linux/linux_proto.h>
77 #endif
78 #include <compat/linux/linux_common.h>
79 #include <compat/linux/linux_file.h>
80 #include <compat/linux/linux_mib.h>
81 #include <compat/linux/linux_socket.h>
82 #include <compat/linux/linux_timer.h>
83 #include <compat/linux/linux_util.h>
84 
85 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
86 					l_uint);
87 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
88 					l_uint, struct msghdr *);
89 static int linux_set_socket_flags(int, int *);
90 
91 static int
92 linux_to_bsd_sockopt_level(int level)
93 {
94 
95 	if (level == LINUX_SOL_SOCKET)
96 		return (SOL_SOCKET);
97 	/* Remaining values are RFC-defined protocol numbers. */
98 	return (level);
99 }
100 
101 static int
102 bsd_to_linux_sockopt_level(int level)
103 {
104 
105 	if (level == SOL_SOCKET)
106 		return (LINUX_SOL_SOCKET);
107 	return (level);
108 }
109 
110 static int
111 linux_to_bsd_ip_sockopt(int opt)
112 {
113 
114 	switch (opt) {
115 	/* known and translated sockopts */
116 	case LINUX_IP_TOS:
117 		return (IP_TOS);
118 	case LINUX_IP_TTL:
119 		return (IP_TTL);
120 	case LINUX_IP_HDRINCL:
121 		return (IP_HDRINCL);
122 	case LINUX_IP_OPTIONS:
123 		return (IP_OPTIONS);
124 	case LINUX_IP_RECVOPTS:
125 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVOPTS");
126 		return (IP_RECVOPTS);
127 	case LINUX_IP_RETOPTS:
128 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_REETOPTS");
129 		return (IP_RETOPTS);
130 	case LINUX_IP_RECVTTL:
131 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTTL");
132 		return (IP_RECVTTL);
133 	case LINUX_IP_RECVTOS:
134 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTOS");
135 		return (IP_RECVTOS);
136 	case LINUX_IP_FREEBIND:
137 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_FREEBIND");
138 		return (IP_BINDANY);
139 	case LINUX_IP_IPSEC_POLICY:
140 		/* we have this option, but not documented in ip(4) manpage */
141 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_IPSEC_POLICY");
142 		return (IP_IPSEC_POLICY);
143 	case LINUX_IP_MINTTL:
144 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MINTTL");
145 		return (IP_MINTTL);
146 	case LINUX_IP_MULTICAST_IF:
147 		return (IP_MULTICAST_IF);
148 	case LINUX_IP_MULTICAST_TTL:
149 		return (IP_MULTICAST_TTL);
150 	case LINUX_IP_MULTICAST_LOOP:
151 		return (IP_MULTICAST_LOOP);
152 	case LINUX_IP_ADD_MEMBERSHIP:
153 		return (IP_ADD_MEMBERSHIP);
154 	case LINUX_IP_DROP_MEMBERSHIP:
155 		return (IP_DROP_MEMBERSHIP);
156 	case LINUX_IP_UNBLOCK_SOURCE:
157 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_UNBLOCK_SOURCE");
158 		return (IP_UNBLOCK_SOURCE);
159 	case LINUX_IP_BLOCK_SOURCE:
160 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_BLOCK_SOURCE");
161 		return (IP_BLOCK_SOURCE);
162 	case LINUX_IP_ADD_SOURCE_MEMBERSHIP:
163 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_ADD_SOURCE_MEMBERSHIP");
164 		return (IP_ADD_SOURCE_MEMBERSHIP);
165 	case LINUX_IP_DROP_SOURCE_MEMBERSHIP:
166 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_DROP_SOURCE_MEMBERSHIP");
167 		return (IP_DROP_SOURCE_MEMBERSHIP);
168 	case LINUX_MCAST_JOIN_GROUP:
169 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_GROUP");
170 		return (MCAST_JOIN_GROUP);
171 	case LINUX_MCAST_LEAVE_GROUP:
172 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_GROUP");
173 		return (MCAST_LEAVE_GROUP);
174 	case LINUX_MCAST_JOIN_SOURCE_GROUP:
175 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_SOURCE_GROUP");
176 		return (MCAST_JOIN_SOURCE_GROUP);
177 	case LINUX_MCAST_LEAVE_SOURCE_GROUP:
178 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_SOURCE_GROUP");
179 		return (MCAST_LEAVE_SOURCE_GROUP);
180 
181 	/* known but not implemented sockopts */
182 	case LINUX_IP_ROUTER_ALERT:
183 		LINUX_RATELIMIT_MSG_OPT1(
184 		    "unsupported IPv4 socket option IP_ROUTER_ALERT (%d), you can not do user-space routing from linux programs",
185 		    opt);
186 		return (-2);
187 	case LINUX_IP_PKTINFO:
188 		LINUX_RATELIMIT_MSG_OPT1(
189 		    "unsupported IPv4 socket option IP_PKTINFO (%d), you can not get extended packet info for datagram sockets in linux programs",
190 		    opt);
191 		return (-2);
192 	case LINUX_IP_PKTOPTIONS:
193 		LINUX_RATELIMIT_MSG_OPT1(
194 		    "unsupported IPv4 socket option IP_PKTOPTIONS (%d)",
195 		    opt);
196 		return (-2);
197 	case LINUX_IP_MTU_DISCOVER:
198 		LINUX_RATELIMIT_MSG_OPT1(
199 		    "unsupported IPv4 socket option IP_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery",
200 		    opt);
201 		return (-2);
202 	case LINUX_IP_RECVERR:
203 		/* needed by steam */
204 		LINUX_RATELIMIT_MSG_OPT1(
205 		    "unsupported IPv4 socket option IP_RECVERR (%d), you can not get extended reliability info in linux programs",
206 		    opt);
207 		return (-2);
208 	case LINUX_IP_MTU:
209 		LINUX_RATELIMIT_MSG_OPT1(
210 		    "unsupported IPv4 socket option IP_MTU (%d), your linux program can not control the MTU on this socket",
211 		    opt);
212 		return (-2);
213 	case LINUX_IP_XFRM_POLICY:
214 		LINUX_RATELIMIT_MSG_OPT1(
215 		    "unsupported IPv4 socket option IP_XFRM_POLICY (%d)",
216 		    opt);
217 		return (-2);
218 	case LINUX_IP_PASSSEC:
219 		/* needed by steam */
220 		LINUX_RATELIMIT_MSG_OPT1(
221 		    "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",
222 		    opt);
223 		return (-2);
224 	case LINUX_IP_TRANSPARENT:
225 		/* IP_BINDANY or more? */
226 		LINUX_RATELIMIT_MSG_OPT1(
227 		    "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",
228 		    opt);
229 		return (-2);
230 	case LINUX_IP_NODEFRAG:
231 		LINUX_RATELIMIT_MSG_OPT1(
232 		    "unsupported IPv4 socket option IP_NODEFRAG (%d)",
233 		    opt);
234 		return (-2);
235 	case LINUX_IP_CHECKSUM:
236 		LINUX_RATELIMIT_MSG_OPT1(
237 		    "unsupported IPv4 socket option IP_CHECKSUM (%d)",
238 		    opt);
239 		return (-2);
240 	case LINUX_IP_BIND_ADDRESS_NO_PORT:
241 		LINUX_RATELIMIT_MSG_OPT1(
242 		    "unsupported IPv4 socket option IP_BIND_ADDRESS_NO_PORT (%d)",
243 		    opt);
244 		return (-2);
245 	case LINUX_IP_RECVFRAGSIZE:
246 		LINUX_RATELIMIT_MSG_OPT1(
247 		    "unsupported IPv4 socket option IP_RECVFRAGSIZE (%d)",
248 		    opt);
249 		return (-2);
250 	case LINUX_MCAST_MSFILTER:
251 		LINUX_RATELIMIT_MSG_OPT1(
252 		    "unsupported IPv4 socket option IP_MCAST_MSFILTER (%d)",
253 		    opt);
254 		return (-2);
255 	case LINUX_IP_MULTICAST_ALL:
256 		LINUX_RATELIMIT_MSG_OPT1(
257 		    "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",
258 		    opt);
259 		return (-2);
260 	case LINUX_IP_UNICAST_IF:
261 		LINUX_RATELIMIT_MSG_OPT1(
262 		    "unsupported IPv4 socket option IP_UNICAST_IF (%d)",
263 		    opt);
264 		return (-2);
265 
266 	/* unknown sockopts */
267 	default:
268 		return (-1);
269 	}
270 }
271 
272 static int
273 linux_to_bsd_ip6_sockopt(int opt)
274 {
275 
276 	switch (opt) {
277 	/* known and translated sockopts */
278 	case LINUX_IPV6_2292PKTINFO:
279 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTINFO");
280 		return (IPV6_2292PKTINFO);
281 	case LINUX_IPV6_2292HOPOPTS:
282 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPOPTS");
283 		return (IPV6_2292HOPOPTS);
284 	case LINUX_IPV6_2292DSTOPTS:
285 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292DSTOPTS");
286 		return (IPV6_2292DSTOPTS);
287 	case LINUX_IPV6_2292RTHDR:
288 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292RTHDR");
289 		return (IPV6_2292RTHDR);
290 	case LINUX_IPV6_2292PKTOPTIONS:
291 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTOPTIONS");
292 		return (IPV6_2292PKTOPTIONS);
293 	case LINUX_IPV6_CHECKSUM:
294 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_CHECKSUM");
295 		return (IPV6_CHECKSUM);
296 	case LINUX_IPV6_2292HOPLIMIT:
297 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPLIMIT");
298 		return (IPV6_2292HOPLIMIT);
299 	case LINUX_IPV6_NEXTHOP:
300 		return (IPV6_NEXTHOP);
301 	case LINUX_IPV6_UNICAST_HOPS:
302 		return (IPV6_UNICAST_HOPS);
303 	case LINUX_IPV6_MULTICAST_IF:
304 		return (IPV6_MULTICAST_IF);
305 	case LINUX_IPV6_MULTICAST_HOPS:
306 		return (IPV6_MULTICAST_HOPS);
307 	case LINUX_IPV6_MULTICAST_LOOP:
308 		return (IPV6_MULTICAST_LOOP);
309 	case LINUX_IPV6_ADD_MEMBERSHIP:
310 		return (IPV6_JOIN_GROUP);
311 	case LINUX_IPV6_DROP_MEMBERSHIP:
312 		return (IPV6_LEAVE_GROUP);
313 	case LINUX_IPV6_V6ONLY:
314 		return (IPV6_V6ONLY);
315 	case LINUX_IPV6_IPSEC_POLICY:
316 		/* we have this option, but not documented in ip6(4) manpage */
317 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_IPSEC_POLICY");
318 		return (IPV6_IPSEC_POLICY);
319 	case LINUX_MCAST_JOIN_GROUP:
320 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_JOIN_GROUP");
321 		return (IPV6_JOIN_GROUP);
322 	case LINUX_MCAST_LEAVE_GROUP:
323 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_LEAVE_GROUP");
324 		return (IPV6_LEAVE_GROUP);
325 	case LINUX_IPV6_RECVPKTINFO:
326 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPKTINFO");
327 		return (IPV6_RECVPKTINFO);
328 	case LINUX_IPV6_PKTINFO:
329 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PKTINFO");
330 		return (IPV6_PKTINFO);
331 	case LINUX_IPV6_RECVHOPLIMIT:
332 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPLIMIT");
333 		return (IPV6_RECVHOPLIMIT);
334 	case LINUX_IPV6_HOPLIMIT:
335 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPLIMIT");
336 		return (IPV6_HOPLIMIT);
337 	case LINUX_IPV6_RECVHOPOPTS:
338 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPOPTS");
339 		return (IPV6_RECVHOPOPTS);
340 	case LINUX_IPV6_HOPOPTS:
341 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPOPTS");
342 		return (IPV6_HOPOPTS);
343 	case LINUX_IPV6_RTHDRDSTOPTS:
344 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDRDSTOPTS");
345 		return (IPV6_RTHDRDSTOPTS);
346 	case LINUX_IPV6_RECVRTHDR:
347 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVRTHDR");
348 		return (IPV6_RECVRTHDR);
349 	case LINUX_IPV6_RTHDR:
350 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDR");
351 		return (IPV6_RTHDR);
352 	case LINUX_IPV6_RECVDSTOPTS:
353 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVDSTOPTS");
354 		return (IPV6_RECVDSTOPTS);
355 	case LINUX_IPV6_DSTOPTS:
356 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_DSTOPTS");
357 		return (IPV6_DSTOPTS);
358 	case LINUX_IPV6_RECVPATHMTU:
359 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPATHMTU");
360 		return (IPV6_RECVPATHMTU);
361 	case LINUX_IPV6_PATHMTU:
362 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PATHMTU");
363 		return (IPV6_PATHMTU);
364 	case LINUX_IPV6_DONTFRAG:
365 		return (IPV6_DONTFRAG);
366 	case LINUX_IPV6_AUTOFLOWLABEL:
367 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_AUTOFLOWLABEL");
368 		return (IPV6_AUTOFLOWLABEL);
369 	case LINUX_IPV6_ORIGDSTADDR:
370 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_ORIGDSTADDR");
371 		return (IPV6_ORIGDSTADDR);
372 	case LINUX_IPV6_FREEBIND:
373 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_FREEBIND");
374 		return (IPV6_BINDANY);
375 
376 	/* known but not implemented sockopts */
377 	case LINUX_IPV6_ADDRFORM:
378 		LINUX_RATELIMIT_MSG_OPT1(
379 		    "unsupported IPv6 socket option IPV6_ADDRFORM (%d), you linux program can not convert the socket to IPv4",
380 		    opt);
381 		return (-2);
382 	case LINUX_IPV6_AUTHHDR:
383 		LINUX_RATELIMIT_MSG_OPT1(
384 		    "unsupported IPv6 socket option IPV6_AUTHHDR (%d), your linux program can not get the authentication header info of IPv6 packets",
385 		    opt);
386 		return (-2);
387 	case LINUX_IPV6_FLOWINFO:
388 		LINUX_RATELIMIT_MSG_OPT1(
389 		    "unsupported IPv6 socket option IPV6_FLOWINFO (%d), your linux program can not get the flowid of IPv6 packets",
390 		    opt);
391 		return (-2);
392 	case LINUX_IPV6_ROUTER_ALERT:
393 		LINUX_RATELIMIT_MSG_OPT1(
394 		    "unsupported IPv6 socket option IPV6_ROUTER_ALERT (%d), you can not do user-space routing from linux programs",
395 		    opt);
396 		return (-2);
397 	case LINUX_IPV6_MTU_DISCOVER:
398 		LINUX_RATELIMIT_MSG_OPT1(
399 		    "unsupported IPv6 socket option IPV6_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery",
400 		    opt);
401 		return (-2);
402 	case LINUX_IPV6_MTU:
403 		LINUX_RATELIMIT_MSG_OPT1(
404 		    "unsupported IPv6 socket option IPV6_MTU (%d), your linux program can not control the MTU on this socket",
405 		    opt);
406 		return (-2);
407 	case LINUX_IPV6_JOIN_ANYCAST:
408 		LINUX_RATELIMIT_MSG_OPT1(
409 		    "unsupported IPv6 socket option IPV6_JOIN_ANYCAST (%d)",
410 		    opt);
411 		return (-2);
412 	case LINUX_IPV6_LEAVE_ANYCAST:
413 		LINUX_RATELIMIT_MSG_OPT1(
414 		    "unsupported IPv6 socket option IPV6_LEAVE_ANYCAST (%d)",
415 		    opt);
416 		return (-2);
417 	case LINUX_IPV6_MULTICAST_ALL:
418 		LINUX_RATELIMIT_MSG_OPT1(
419 		    "unsupported IPv6 socket option IPV6_MULTICAST_ALL (%d)",
420 		    opt);
421 		return (-2);
422 	case LINUX_IPV6_ROUTER_ALERT_ISOLATE:
423 		LINUX_RATELIMIT_MSG_OPT1(
424 		    "unsupported IPv6 socket option IPV6_ROUTER_ALERT_ISOLATE (%d)",
425 		    opt);
426 		return (-2);
427 	case LINUX_IPV6_FLOWLABEL_MGR:
428 		LINUX_RATELIMIT_MSG_OPT1(
429 		    "unsupported IPv6 socket option IPV6_FLOWLABEL_MGR (%d)",
430 		    opt);
431 		return (-2);
432 	case LINUX_IPV6_FLOWINFO_SEND:
433 		LINUX_RATELIMIT_MSG_OPT1(
434 		    "unsupported IPv6 socket option IPV6_FLOWINFO_SEND (%d)",
435 		    opt);
436 		return (-2);
437 	case LINUX_IPV6_XFRM_POLICY:
438 		LINUX_RATELIMIT_MSG_OPT1(
439 		    "unsupported IPv6 socket option IPV6_XFRM_POLICY (%d)",
440 		    opt);
441 		return (-2);
442 	case LINUX_IPV6_HDRINCL:
443 		LINUX_RATELIMIT_MSG_OPT1(
444 		    "unsupported IPv6 socket option IPV6_HDRINCL (%d)",
445 		    opt);
446 		return (-2);
447 	case LINUX_MCAST_BLOCK_SOURCE:
448 		LINUX_RATELIMIT_MSG_OPT1(
449 		    "unsupported IPv6 socket option MCAST_BLOCK_SOURCE (%d), your linux program may see more multicast stuff than it wants",
450 		    opt);
451 		return (-2);
452 	case LINUX_MCAST_UNBLOCK_SOURCE:
453 		LINUX_RATELIMIT_MSG_OPT1(
454 		    "unsupported IPv6 socket option MCAST_UNBLOCK_SOURCE (%d), your linux program may not see all the multicast stuff it wants",
455 		    opt);
456 		return (-2);
457 	case LINUX_MCAST_JOIN_SOURCE_GROUP:
458 		LINUX_RATELIMIT_MSG_OPT1(
459 		    "unsupported IPv6 socket option MCAST_JOIN_SOURCE_GROUP (%d), your linux program is not able to join a multicast source group",
460 		    opt);
461 		return (-2);
462 	case LINUX_MCAST_LEAVE_SOURCE_GROUP:
463 		LINUX_RATELIMIT_MSG_OPT1(
464 		    "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",
465 		    opt);
466 		return (-2);
467 	case LINUX_MCAST_MSFILTER:
468 		LINUX_RATELIMIT_MSG_OPT1(
469 		    "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",
470 		    opt);
471 		return (-2);
472 	case LINUX_IPV6_ADDR_PREFERENCES:
473 		LINUX_RATELIMIT_MSG_OPT1(
474 		    "unsupported IPv6 socket option IPV6_ADDR_PREFERENCES (%d)",
475 		    opt);
476 		return (-2);
477 	case LINUX_IPV6_MINHOPCOUNT:
478 		LINUX_RATELIMIT_MSG_OPT1(
479 		    "unsupported IPv6 socket option IPV6_MINHOPCOUNT (%d)",
480 		    opt);
481 		return (-2);
482 	case LINUX_IPV6_TRANSPARENT:
483 		/* IP_BINDANY or more? */
484 		LINUX_RATELIMIT_MSG_OPT1(
485 		    "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",
486 		    opt);
487 		return (-2);
488 	case LINUX_IPV6_UNICAST_IF:
489 		LINUX_RATELIMIT_MSG_OPT1(
490 		    "unsupported IPv6 socket option IPV6_UNICAST_IF (%d)",
491 		    opt);
492 		return (-2);
493 	case LINUX_IPV6_RECVFRAGSIZE:
494 		LINUX_RATELIMIT_MSG_OPT1(
495 		    "unsupported IPv6 socket option IPV6_RECVFRAGSIZE (%d)",
496 		    opt);
497 		return (-2);
498 
499 	/* unknown sockopts */
500 	default:
501 		return (-1);
502 	}
503 }
504 
505 static int
506 linux_to_bsd_so_sockopt(int opt)
507 {
508 
509 	switch (opt) {
510 	case LINUX_SO_DEBUG:
511 		return (SO_DEBUG);
512 	case LINUX_SO_REUSEADDR:
513 		return (SO_REUSEADDR);
514 	case LINUX_SO_TYPE:
515 		return (SO_TYPE);
516 	case LINUX_SO_ERROR:
517 		return (SO_ERROR);
518 	case LINUX_SO_DONTROUTE:
519 		return (SO_DONTROUTE);
520 	case LINUX_SO_BROADCAST:
521 		return (SO_BROADCAST);
522 	case LINUX_SO_SNDBUF:
523 	case LINUX_SO_SNDBUFFORCE:
524 		return (SO_SNDBUF);
525 	case LINUX_SO_RCVBUF:
526 	case LINUX_SO_RCVBUFFORCE:
527 		return (SO_RCVBUF);
528 	case LINUX_SO_KEEPALIVE:
529 		return (SO_KEEPALIVE);
530 	case LINUX_SO_OOBINLINE:
531 		return (SO_OOBINLINE);
532 	case LINUX_SO_LINGER:
533 		return (SO_LINGER);
534 	case LINUX_SO_REUSEPORT:
535 		return (SO_REUSEPORT_LB);
536 	case LINUX_SO_PASSCRED:
537 		return (LOCAL_CREDS_PERSISTENT);
538 	case LINUX_SO_PEERCRED:
539 		return (LOCAL_PEERCRED);
540 	case LINUX_SO_RCVLOWAT:
541 		return (SO_RCVLOWAT);
542 	case LINUX_SO_SNDLOWAT:
543 		return (SO_SNDLOWAT);
544 	case LINUX_SO_RCVTIMEO:
545 		return (SO_RCVTIMEO);
546 	case LINUX_SO_SNDTIMEO:
547 		return (SO_SNDTIMEO);
548 	case LINUX_SO_TIMESTAMP:
549 		return (SO_TIMESTAMP);
550 	case LINUX_SO_ACCEPTCONN:
551 		return (SO_ACCEPTCONN);
552 	case LINUX_SO_PROTOCOL:
553 		return (SO_PROTOCOL);
554 	}
555 	return (-1);
556 }
557 
558 static int
559 linux_to_bsd_tcp_sockopt(int opt)
560 {
561 
562 	switch (opt) {
563 	case LINUX_TCP_NODELAY:
564 		return (TCP_NODELAY);
565 	case LINUX_TCP_MAXSEG:
566 		return (TCP_MAXSEG);
567 	case LINUX_TCP_CORK:
568 		return (TCP_NOPUSH);
569 	case LINUX_TCP_KEEPIDLE:
570 		return (TCP_KEEPIDLE);
571 	case LINUX_TCP_KEEPINTVL:
572 		return (TCP_KEEPINTVL);
573 	case LINUX_TCP_KEEPCNT:
574 		return (TCP_KEEPCNT);
575 	case LINUX_TCP_MD5SIG:
576 		return (TCP_MD5SIG);
577 	}
578 	return (-1);
579 }
580 
581 static int
582 linux_to_bsd_msg_flags(int flags)
583 {
584 	int ret_flags = 0;
585 
586 	if (flags & LINUX_MSG_OOB)
587 		ret_flags |= MSG_OOB;
588 	if (flags & LINUX_MSG_PEEK)
589 		ret_flags |= MSG_PEEK;
590 	if (flags & LINUX_MSG_DONTROUTE)
591 		ret_flags |= MSG_DONTROUTE;
592 	if (flags & LINUX_MSG_CTRUNC)
593 		ret_flags |= MSG_CTRUNC;
594 	if (flags & LINUX_MSG_TRUNC)
595 		ret_flags |= MSG_TRUNC;
596 	if (flags & LINUX_MSG_DONTWAIT)
597 		ret_flags |= MSG_DONTWAIT;
598 	if (flags & LINUX_MSG_EOR)
599 		ret_flags |= MSG_EOR;
600 	if (flags & LINUX_MSG_WAITALL)
601 		ret_flags |= MSG_WAITALL;
602 	if (flags & LINUX_MSG_NOSIGNAL)
603 		ret_flags |= MSG_NOSIGNAL;
604 	if (flags & LINUX_MSG_PROXY)
605 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_PROXY (%d) not handled",
606 		    LINUX_MSG_PROXY);
607 	if (flags & LINUX_MSG_FIN)
608 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_FIN (%d) not handled",
609 		    LINUX_MSG_FIN);
610 	if (flags & LINUX_MSG_SYN)
611 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_SYN (%d) not handled",
612 		    LINUX_MSG_SYN);
613 	if (flags & LINUX_MSG_CONFIRM)
614 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_CONFIRM (%d) not handled",
615 		    LINUX_MSG_CONFIRM);
616 	if (flags & LINUX_MSG_RST)
617 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_RST (%d) not handled",
618 		    LINUX_MSG_RST);
619 	if (flags & LINUX_MSG_ERRQUEUE)
620 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_ERRQUEUE (%d) not handled",
621 		    LINUX_MSG_ERRQUEUE);
622 	return (ret_flags);
623 }
624 
625 static int
626 linux_to_bsd_cmsg_type(int cmsg_type)
627 {
628 
629 	switch (cmsg_type) {
630 	case LINUX_SCM_RIGHTS:
631 		return (SCM_RIGHTS);
632 	case LINUX_SCM_CREDENTIALS:
633 		return (SCM_CREDS);
634 	}
635 	return (-1);
636 }
637 
638 static int
639 bsd_to_linux_cmsg_type(int cmsg_type)
640 {
641 
642 	switch (cmsg_type) {
643 	case SCM_RIGHTS:
644 		return (LINUX_SCM_RIGHTS);
645 	case SCM_CREDS:
646 		return (LINUX_SCM_CREDENTIALS);
647 	case SCM_TIMESTAMP:
648 		return (LINUX_SCM_TIMESTAMP);
649 	}
650 	return (-1);
651 }
652 
653 static int
654 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
655 {
656 	if (lhdr->msg_controllen > INT_MAX)
657 		return (ENOBUFS);
658 
659 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
660 	bhdr->msg_namelen	= lhdr->msg_namelen;
661 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
662 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
663 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
664 
665 	/*
666 	 * msg_controllen is skipped since BSD and LINUX control messages
667 	 * are potentially different sizes (e.g. the cred structure used
668 	 * by SCM_CREDS is different between the two operating system).
669 	 *
670 	 * The caller can set it (if necessary) after converting all the
671 	 * control messages.
672 	 */
673 
674 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
675 	return (0);
676 }
677 
678 static int
679 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
680 {
681 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
682 	lhdr->msg_namelen	= bhdr->msg_namelen;
683 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
684 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
685 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
686 
687 	/*
688 	 * msg_controllen is skipped since BSD and LINUX control messages
689 	 * are potentially different sizes (e.g. the cred structure used
690 	 * by SCM_CREDS is different between the two operating system).
691 	 *
692 	 * The caller can set it (if necessary) after converting all the
693 	 * control messages.
694 	 */
695 
696 	/* msg_flags skipped */
697 	return (0);
698 }
699 
700 static int
701 linux_set_socket_flags(int lflags, int *flags)
702 {
703 
704 	if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
705 		return (EINVAL);
706 	if (lflags & LINUX_SOCK_NONBLOCK)
707 		*flags |= SOCK_NONBLOCK;
708 	if (lflags & LINUX_SOCK_CLOEXEC)
709 		*flags |= SOCK_CLOEXEC;
710 	return (0);
711 }
712 
713 static int
714 linux_copyout_sockaddr(const struct sockaddr *sa, void *uaddr, size_t len)
715 {
716 	struct l_sockaddr *lsa;
717 	int error;
718 
719 	error = bsd_to_linux_sockaddr(sa, &lsa, len);
720 	if (error != 0)
721 		return (error);
722 
723 	error = copyout(lsa, uaddr, len);
724 	free(lsa, M_SONAME);
725 
726 	return (error);
727 }
728 
729 static int
730 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
731     struct mbuf *control, enum uio_seg segflg)
732 {
733 	struct sockaddr *to;
734 	int error, len;
735 
736 	if (mp->msg_name != NULL) {
737 		len = mp->msg_namelen;
738 		error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len);
739 		if (error != 0)
740 			return (error);
741 		mp->msg_name = to;
742 	} else
743 		to = NULL;
744 
745 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
746 	    segflg);
747 
748 	if (to)
749 		free(to, M_SONAME);
750 	return (error);
751 }
752 
753 /* Return 0 if IP_HDRINCL is set for the given socket. */
754 static int
755 linux_check_hdrincl(struct thread *td, int s)
756 {
757 	int error, optval;
758 	socklen_t size_val;
759 
760 	size_val = sizeof(optval);
761 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
762 	    &optval, UIO_SYSSPACE, &size_val);
763 	if (error != 0)
764 		return (error);
765 
766 	return (optval == 0);
767 }
768 
769 /*
770  * Updated sendto() when IP_HDRINCL is set:
771  * tweak endian-dependent fields in the IP packet.
772  */
773 static int
774 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
775 {
776 /*
777  * linux_ip_copysize defines how many bytes we should copy
778  * from the beginning of the IP packet before we customize it for BSD.
779  * It should include all the fields we modify (ip_len and ip_off).
780  */
781 #define linux_ip_copysize	8
782 
783 	struct ip *packet;
784 	struct msghdr msg;
785 	struct iovec aiov[1];
786 	int error;
787 
788 	/* Check that the packet isn't too big or too small. */
789 	if (linux_args->len < linux_ip_copysize ||
790 	    linux_args->len > IP_MAXPACKET)
791 		return (EINVAL);
792 
793 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
794 
795 	/* Make kernel copy of the packet to be sent */
796 	if ((error = copyin(PTRIN(linux_args->msg), packet,
797 	    linux_args->len)))
798 		goto goout;
799 
800 	/* Convert fields from Linux to BSD raw IP socket format */
801 	packet->ip_len = linux_args->len;
802 	packet->ip_off = ntohs(packet->ip_off);
803 
804 	/* Prepare the msghdr and iovec structures describing the new packet */
805 	msg.msg_name = PTRIN(linux_args->to);
806 	msg.msg_namelen = linux_args->tolen;
807 	msg.msg_iov = aiov;
808 	msg.msg_iovlen = 1;
809 	msg.msg_control = NULL;
810 	msg.msg_flags = 0;
811 	aiov[0].iov_base = (char *)packet;
812 	aiov[0].iov_len = linux_args->len;
813 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
814 	    NULL, UIO_SYSSPACE);
815 goout:
816 	free(packet, M_LINUX);
817 	return (error);
818 }
819 
820 static const char *linux_netlink_names[] = {
821 	[LINUX_NETLINK_ROUTE] = "ROUTE",
822 	[LINUX_NETLINK_SOCK_DIAG] = "SOCK_DIAG",
823 	[LINUX_NETLINK_NFLOG] = "NFLOG",
824 	[LINUX_NETLINK_SELINUX] = "SELINUX",
825 	[LINUX_NETLINK_AUDIT] = "AUDIT",
826 	[LINUX_NETLINK_FIB_LOOKUP] = "FIB_LOOKUP",
827 	[LINUX_NETLINK_NETFILTER] = "NETFILTER",
828 	[LINUX_NETLINK_KOBJECT_UEVENT] = "KOBJECT_UEVENT",
829 };
830 
831 int
832 linux_socket(struct thread *td, struct linux_socket_args *args)
833 {
834 	int domain, retval_socket, type;
835 
836 	type = args->type & LINUX_SOCK_TYPE_MASK;
837 	if (type < 0 || type > LINUX_SOCK_MAX)
838 		return (EINVAL);
839 	retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
840 		&type);
841 	if (retval_socket != 0)
842 		return (retval_socket);
843 	domain = linux_to_bsd_domain(args->domain);
844 	if (domain == -1) {
845 		/* Mask off SOCK_NONBLOCK / CLOEXEC for error messages. */
846 		type = args->type & LINUX_SOCK_TYPE_MASK;
847 		if (args->domain == LINUX_AF_NETLINK) {
848 			const char *nl_name;
849 
850 			if (args->protocol >= 0 &&
851 			    args->protocol < nitems(linux_netlink_names))
852 				nl_name = linux_netlink_names[args->protocol];
853 			else
854 				nl_name = NULL;
855 			if (nl_name != NULL)
856 				linux_msg(curthread,
857 				    "unsupported socket(AF_NETLINK, %d, "
858 				    "NETLINK_%s)", type, nl_name);
859 			else
860 				linux_msg(curthread,
861 				    "unsupported socket(AF_NETLINK, %d, %d)",
862 				    type, args->protocol);
863 		} else {
864 			linux_msg(curthread, "unsupported socket domain %d, "
865 			    "type %d, protocol %d", args->domain, type,
866 			    args->protocol);
867 		}
868 		return (EAFNOSUPPORT);
869 	}
870 
871 	retval_socket = kern_socket(td, domain, type, args->protocol);
872 	if (retval_socket)
873 		return (retval_socket);
874 
875 	if (type == SOCK_RAW
876 	    && (args->protocol == IPPROTO_RAW || args->protocol == 0)
877 	    && domain == PF_INET) {
878 		/* It's a raw IP socket: set the IP_HDRINCL option. */
879 		int hdrincl;
880 
881 		hdrincl = 1;
882 		/* We ignore any error returned by kern_setsockopt() */
883 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
884 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
885 	}
886 #ifdef INET6
887 	/*
888 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
889 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
890 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
891 	 * sysctl value.
892 	 */
893 	if (domain == PF_INET6) {
894 		int v6only;
895 
896 		v6only = 0;
897 		/* We ignore any error returned by setsockopt() */
898 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
899 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
900 	}
901 #endif
902 
903 	return (retval_socket);
904 }
905 
906 int
907 linux_bind(struct thread *td, struct linux_bind_args *args)
908 {
909 	struct sockaddr *sa;
910 	int error;
911 
912 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
913 	    &args->namelen);
914 	if (error != 0)
915 		return (error);
916 
917 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
918 	free(sa, M_SONAME);
919 
920 	/* XXX */
921 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
922 		return (EINVAL);
923 	return (error);
924 }
925 
926 int
927 linux_connect(struct thread *td, struct linux_connect_args *args)
928 {
929 	struct socket *so;
930 	struct sockaddr *sa;
931 	struct file *fp;
932 	u_int fflag;
933 	int error;
934 
935 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
936 	    &args->namelen);
937 	if (error != 0)
938 		return (error);
939 
940 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
941 	free(sa, M_SONAME);
942 	if (error != EISCONN)
943 		return (error);
944 
945 	/*
946 	 * Linux doesn't return EISCONN the first time it occurs,
947 	 * when on a non-blocking socket. Instead it returns the
948 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
949 	 */
950 	error = getsock_cap(td, args->s, &cap_connect_rights,
951 	    &fp, &fflag, NULL);
952 	if (error != 0)
953 		return (error);
954 
955 	error = EISCONN;
956 	so = fp->f_data;
957 	if (fflag & FNONBLOCK) {
958 		SOCK_LOCK(so);
959 		if (so->so_emuldata == 0)
960 			error = so->so_error;
961 		so->so_emuldata = (void *)1;
962 		SOCK_UNLOCK(so);
963 	}
964 	fdrop(fp, td);
965 
966 	return (error);
967 }
968 
969 int
970 linux_listen(struct thread *td, struct linux_listen_args *args)
971 {
972 
973 	return (kern_listen(td, args->s, args->backlog));
974 }
975 
976 static int
977 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
978     l_uintptr_t namelen, int flags)
979 {
980 	struct sockaddr *sa;
981 	struct file *fp, *fp1;
982 	int bflags, len;
983 	struct socket *so;
984 	int error, error1;
985 
986 	bflags = 0;
987 	fp = NULL;
988 	sa = NULL;
989 
990 	error = linux_set_socket_flags(flags, &bflags);
991 	if (error != 0)
992 		return (error);
993 
994 	if (PTRIN(addr) == NULL) {
995 		len = 0;
996 		error = kern_accept4(td, s, NULL, NULL, bflags, NULL);
997 	} else {
998 		error = copyin(PTRIN(namelen), &len, sizeof(len));
999 		if (error != 0)
1000 			return (error);
1001 		if (len < 0)
1002 			return (EINVAL);
1003 		error = kern_accept4(td, s, &sa, &len, bflags, &fp);
1004 	}
1005 
1006 	/*
1007 	 * Translate errno values into ones used by Linux.
1008 	 */
1009 	if (error != 0) {
1010 		/*
1011 		 * XXX. This is wrong, different sockaddr structures
1012 		 * have different sizes.
1013 		 */
1014 		switch (error) {
1015 		case EFAULT:
1016 			if (namelen != sizeof(struct sockaddr_in))
1017 				error = EINVAL;
1018 			break;
1019 		case EINVAL:
1020 			error1 = getsock_cap(td, s, &cap_accept_rights, &fp1, NULL, NULL);
1021 			if (error1 != 0) {
1022 				error = error1;
1023 				break;
1024 			}
1025 			so = fp1->f_data;
1026 			if (so->so_type == SOCK_DGRAM)
1027 				error = EOPNOTSUPP;
1028 			fdrop(fp1, td);
1029 			break;
1030 		}
1031 		return (error);
1032 	}
1033 
1034 	if (len != 0) {
1035 		error = linux_copyout_sockaddr(sa, PTRIN(addr), len);
1036 
1037 		/*
1038 		 * XXX: We should also copyout the len, shouldn't we?
1039 		 */
1040 
1041 		if (error != 0) {
1042 			fdclose(td, fp, td->td_retval[0]);
1043 			td->td_retval[0] = 0;
1044 		}
1045 	}
1046 	if (fp != NULL)
1047 		fdrop(fp, td);
1048 	free(sa, M_SONAME);
1049 	return (error);
1050 }
1051 
1052 int
1053 linux_accept(struct thread *td, struct linux_accept_args *args)
1054 {
1055 
1056 	return (linux_accept_common(td, args->s, args->addr,
1057 	    args->namelen, 0));
1058 }
1059 
1060 int
1061 linux_accept4(struct thread *td, struct linux_accept4_args *args)
1062 {
1063 
1064 	return (linux_accept_common(td, args->s, args->addr,
1065 	    args->namelen, args->flags));
1066 }
1067 
1068 int
1069 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
1070 {
1071 	struct sockaddr *sa;
1072 	int len, error;
1073 
1074 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1075 	if (error != 0)
1076 		return (error);
1077 
1078 	error = kern_getsockname(td, args->s, &sa, &len);
1079 	if (error != 0)
1080 		return (error);
1081 
1082 	if (len != 0)
1083 		error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
1084 
1085 	free(sa, M_SONAME);
1086 	if (error == 0)
1087 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1088 	return (error);
1089 }
1090 
1091 int
1092 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
1093 {
1094 	struct sockaddr *sa;
1095 	int len, error;
1096 
1097 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1098 	if (error != 0)
1099 		return (error);
1100 	if (len < 0)
1101 		return (EINVAL);
1102 
1103 	error = kern_getpeername(td, args->s, &sa, &len);
1104 	if (error != 0)
1105 		return (error);
1106 
1107 	if (len != 0)
1108 		error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
1109 
1110 	free(sa, M_SONAME);
1111 	if (error == 0)
1112 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1113 	return (error);
1114 }
1115 
1116 int
1117 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
1118 {
1119 	int domain, error, sv[2], type;
1120 
1121 	domain = linux_to_bsd_domain(args->domain);
1122 	if (domain != PF_LOCAL)
1123 		return (EAFNOSUPPORT);
1124 	type = args->type & LINUX_SOCK_TYPE_MASK;
1125 	if (type < 0 || type > LINUX_SOCK_MAX)
1126 		return (EINVAL);
1127 	error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
1128 	    &type);
1129 	if (error != 0)
1130 		return (error);
1131 	if (args->protocol != 0 && args->protocol != PF_UNIX) {
1132 		/*
1133 		 * Use of PF_UNIX as protocol argument is not right,
1134 		 * but Linux does it.
1135 		 * Do not map PF_UNIX as its Linux value is identical
1136 		 * to FreeBSD one.
1137 		 */
1138 		return (EPROTONOSUPPORT);
1139 	}
1140 	error = kern_socketpair(td, domain, type, 0, sv);
1141 	if (error != 0)
1142                 return (error);
1143         error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int));
1144         if (error != 0) {
1145                 (void)kern_close(td, sv[0]);
1146                 (void)kern_close(td, sv[1]);
1147         }
1148 	return (error);
1149 }
1150 
1151 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1152 struct linux_send_args {
1153 	register_t s;
1154 	register_t msg;
1155 	register_t len;
1156 	register_t flags;
1157 };
1158 
1159 static int
1160 linux_send(struct thread *td, struct linux_send_args *args)
1161 {
1162 	struct sendto_args /* {
1163 		int s;
1164 		caddr_t buf;
1165 		int len;
1166 		int flags;
1167 		caddr_t to;
1168 		int tolen;
1169 	} */ bsd_args;
1170 	struct file *fp;
1171 	int error, fflag;
1172 
1173 	bsd_args.s = args->s;
1174 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1175 	bsd_args.len = args->len;
1176 	bsd_args.flags = args->flags;
1177 	bsd_args.to = NULL;
1178 	bsd_args.tolen = 0;
1179 	error = sys_sendto(td, &bsd_args);
1180 	if (error == ENOTCONN) {
1181 		/*
1182 		 * Linux doesn't return ENOTCONN for non-blocking sockets.
1183 		 * Instead it returns the EAGAIN.
1184 		 */
1185 		error = getsock_cap(td, args->s, &cap_send_rights, &fp,
1186 		    &fflag, NULL);
1187 		if (error == 0) {
1188 			if (fflag & FNONBLOCK)
1189 				error = EAGAIN;
1190 			fdrop(fp, td);
1191 		}
1192 	}
1193 	return (error);
1194 }
1195 
1196 struct linux_recv_args {
1197 	register_t s;
1198 	register_t msg;
1199 	register_t len;
1200 	register_t flags;
1201 };
1202 
1203 static int
1204 linux_recv(struct thread *td, struct linux_recv_args *args)
1205 {
1206 	struct recvfrom_args /* {
1207 		int s;
1208 		caddr_t buf;
1209 		int len;
1210 		int flags;
1211 		struct sockaddr *from;
1212 		socklen_t fromlenaddr;
1213 	} */ bsd_args;
1214 
1215 	bsd_args.s = args->s;
1216 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1217 	bsd_args.len = args->len;
1218 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1219 	bsd_args.from = NULL;
1220 	bsd_args.fromlenaddr = 0;
1221 	return (sys_recvfrom(td, &bsd_args));
1222 }
1223 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1224 
1225 int
1226 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1227 {
1228 	struct msghdr msg;
1229 	struct iovec aiov;
1230 
1231 	if (linux_check_hdrincl(td, args->s) == 0)
1232 		/* IP_HDRINCL set, tweak the packet before sending */
1233 		return (linux_sendto_hdrincl(td, args));
1234 
1235 	msg.msg_name = PTRIN(args->to);
1236 	msg.msg_namelen = args->tolen;
1237 	msg.msg_iov = &aiov;
1238 	msg.msg_iovlen = 1;
1239 	msg.msg_control = NULL;
1240 	msg.msg_flags = 0;
1241 	aiov.iov_base = PTRIN(args->msg);
1242 	aiov.iov_len = args->len;
1243 	return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1244 	    UIO_USERSPACE));
1245 }
1246 
1247 int
1248 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1249 {
1250 	struct sockaddr *sa;
1251 	struct msghdr msg;
1252 	struct iovec aiov;
1253 	int error, fromlen;
1254 
1255 	if (PTRIN(args->fromlen) != NULL) {
1256 		error = copyin(PTRIN(args->fromlen), &fromlen,
1257 		    sizeof(fromlen));
1258 		if (error != 0)
1259 			return (error);
1260 		if (fromlen < 0)
1261 			return (EINVAL);
1262 		sa = malloc(fromlen, M_SONAME, M_WAITOK);
1263 	} else {
1264 		fromlen = 0;
1265 		sa = NULL;
1266 	}
1267 
1268 	msg.msg_name = sa;
1269 	msg.msg_namelen = fromlen;
1270 	msg.msg_iov = &aiov;
1271 	msg.msg_iovlen = 1;
1272 	aiov.iov_base = PTRIN(args->buf);
1273 	aiov.iov_len = args->len;
1274 	msg.msg_control = 0;
1275 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1276 
1277 	error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL);
1278 	if (error != 0)
1279 		goto out;
1280 
1281 	if (PTRIN(args->from) != NULL)
1282 		error = linux_copyout_sockaddr(sa, PTRIN(args->from), msg.msg_namelen);
1283 
1284 	if (error == 0 && PTRIN(args->fromlen) != NULL)
1285 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1286 		    sizeof(msg.msg_namelen));
1287 out:
1288 	free(sa, M_SONAME);
1289 	return (error);
1290 }
1291 
1292 static int
1293 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1294     l_uint flags)
1295 {
1296 	struct cmsghdr *cmsg;
1297 	struct mbuf *control;
1298 	struct msghdr msg;
1299 	struct l_cmsghdr linux_cmsg;
1300 	struct l_cmsghdr *ptr_cmsg;
1301 	struct l_msghdr linux_msghdr;
1302 	struct iovec *iov;
1303 	socklen_t datalen;
1304 	struct sockaddr *sa;
1305 	struct socket *so;
1306 	sa_family_t sa_family;
1307 	struct file *fp;
1308 	void *data;
1309 	l_size_t len;
1310 	l_size_t clen;
1311 	int error, fflag;
1312 
1313 	error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
1314 	if (error != 0)
1315 		return (error);
1316 
1317 	/*
1318 	 * Some Linux applications (ping) define a non-NULL control data
1319 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1320 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1321 	 * order to handle this case.  This should be checked, but allows the
1322 	 * Linux ping to work.
1323 	 */
1324 	if (PTRIN(linux_msghdr.msg_control) != NULL &&
1325 	    linux_msghdr.msg_controllen == 0)
1326 		linux_msghdr.msg_control = PTROUT(NULL);
1327 
1328 	error = linux_to_bsd_msghdr(&msg, &linux_msghdr);
1329 	if (error != 0)
1330 		return (error);
1331 
1332 #ifdef COMPAT_LINUX32
1333 	error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1334 	    &iov, EMSGSIZE);
1335 #else
1336 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1337 #endif
1338 	if (error != 0)
1339 		return (error);
1340 
1341 	control = NULL;
1342 
1343 	error = kern_getsockname(td, s, &sa, &datalen);
1344 	if (error != 0)
1345 		goto bad;
1346 	sa_family = sa->sa_family;
1347 	free(sa, M_SONAME);
1348 
1349 	if (flags & LINUX_MSG_OOB) {
1350 		error = EOPNOTSUPP;
1351 		if (sa_family == AF_UNIX)
1352 			goto bad;
1353 
1354 		error = getsock_cap(td, s, &cap_send_rights, &fp,
1355 		    &fflag, NULL);
1356 		if (error != 0)
1357 			goto bad;
1358 		so = fp->f_data;
1359 		if (so->so_type != SOCK_STREAM)
1360 			error = EOPNOTSUPP;
1361 		fdrop(fp, td);
1362 		if (error != 0)
1363 			goto bad;
1364 	}
1365 
1366 	if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) {
1367 		error = ENOBUFS;
1368 		control = m_get(M_WAITOK, MT_CONTROL);
1369 		MCLGET(control, M_WAITOK);
1370 		data = mtod(control, void *);
1371 		datalen = 0;
1372 
1373 		ptr_cmsg = PTRIN(linux_msghdr.msg_control);
1374 		clen = linux_msghdr.msg_controllen;
1375 		do {
1376 			error = copyin(ptr_cmsg, &linux_cmsg,
1377 			    sizeof(struct l_cmsghdr));
1378 			if (error != 0)
1379 				goto bad;
1380 
1381 			error = EINVAL;
1382 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1383 			    linux_cmsg.cmsg_len > clen)
1384 				goto bad;
1385 
1386 			if (datalen + CMSG_HDRSZ > MCLBYTES)
1387 				goto bad;
1388 
1389 			/*
1390 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1391 			 * so return EINVAL in any other cmsg_type
1392 			 */
1393 			cmsg = data;
1394 			cmsg->cmsg_type =
1395 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1396 			cmsg->cmsg_level =
1397 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1398 			if (cmsg->cmsg_type == -1
1399 			    || cmsg->cmsg_level != SOL_SOCKET) {
1400 				linux_msg(curthread,
1401 				    "unsupported sendmsg cmsg level %d type %d",
1402 				    linux_cmsg.cmsg_level, linux_cmsg.cmsg_type);
1403 				goto bad;
1404 			}
1405 
1406 			/*
1407 			 * Some applications (e.g. pulseaudio) attempt to
1408 			 * send ancillary data even if the underlying protocol
1409 			 * doesn't support it which is not allowed in the
1410 			 * FreeBSD system call interface.
1411 			 */
1412 			if (sa_family != AF_UNIX)
1413 				goto next;
1414 
1415 			if (cmsg->cmsg_type == SCM_CREDS) {
1416 				len = sizeof(struct cmsgcred);
1417 				if (datalen + CMSG_SPACE(len) > MCLBYTES)
1418 					goto bad;
1419 
1420 				/*
1421 				 * The lower levels will fill in the structure
1422 				 */
1423 				memset(CMSG_DATA(data), 0, len);
1424 			} else {
1425 				len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1426 				if (datalen + CMSG_SPACE(len) < datalen ||
1427 				    datalen + CMSG_SPACE(len) > MCLBYTES)
1428 					goto bad;
1429 
1430 				error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1431 				    CMSG_DATA(data), len);
1432 				if (error != 0)
1433 					goto bad;
1434 			}
1435 
1436 			cmsg->cmsg_len = CMSG_LEN(len);
1437 			data = (char *)data + CMSG_SPACE(len);
1438 			datalen += CMSG_SPACE(len);
1439 
1440 next:
1441 			if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1442 				break;
1443 
1444 			clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1445 			ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1446 			    LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1447 		} while(clen >= sizeof(struct l_cmsghdr));
1448 
1449 		control->m_len = datalen;
1450 		if (datalen == 0) {
1451 			m_freem(control);
1452 			control = NULL;
1453 		}
1454 	}
1455 
1456 	msg.msg_iov = iov;
1457 	msg.msg_flags = 0;
1458 	error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1459 	control = NULL;
1460 
1461 bad:
1462 	m_freem(control);
1463 	free(iov, M_IOV);
1464 	return (error);
1465 }
1466 
1467 int
1468 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1469 {
1470 
1471 	return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1472 	    args->flags));
1473 }
1474 
1475 int
1476 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1477 {
1478 	struct l_mmsghdr *msg;
1479 	l_uint retval;
1480 	int error, datagrams;
1481 
1482 	if (args->vlen > UIO_MAXIOV)
1483 		args->vlen = UIO_MAXIOV;
1484 
1485 	msg = PTRIN(args->msg);
1486 	datagrams = 0;
1487 	while (datagrams < args->vlen) {
1488 		error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1489 		    args->flags);
1490 		if (error != 0)
1491 			break;
1492 
1493 		retval = td->td_retval[0];
1494 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1495 		if (error != 0)
1496 			break;
1497 		++msg;
1498 		++datagrams;
1499 	}
1500 	if (error == 0)
1501 		td->td_retval[0] = datagrams;
1502 	return (error);
1503 }
1504 
1505 static int
1506 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1507     l_uint flags, struct msghdr *msg)
1508 {
1509 	struct cmsghdr *cm;
1510 	struct cmsgcred *cmcred;
1511 	struct l_cmsghdr *linux_cmsg = NULL;
1512 	struct l_ucred linux_ucred;
1513 	socklen_t datalen, maxlen, outlen;
1514 	struct l_msghdr linux_msghdr;
1515 	struct iovec *iov, *uiov;
1516 	struct mbuf *control = NULL;
1517 	struct mbuf **controlp;
1518 	struct timeval *ftmvl;
1519 	struct sockaddr *sa;
1520 	l_timeval ltmvl;
1521 	caddr_t outbuf;
1522 	void *data;
1523 	int error, i, fd, fds, *fdp;
1524 
1525 	error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
1526 	if (error != 0)
1527 		return (error);
1528 
1529 	error = linux_to_bsd_msghdr(msg, &linux_msghdr);
1530 	if (error != 0)
1531 		return (error);
1532 
1533 #ifdef COMPAT_LINUX32
1534 	error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1535 	    &iov, EMSGSIZE);
1536 #else
1537 	error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1538 #endif
1539 	if (error != 0)
1540 		return (error);
1541 
1542 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1543 		msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN);
1544 		sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK);
1545 		msg->msg_name = sa;
1546 	} else {
1547 		sa = NULL;
1548 		msg->msg_name = NULL;
1549 	}
1550 
1551 	uiov = msg->msg_iov;
1552 	msg->msg_iov = iov;
1553 	controlp = (msg->msg_control != NULL) ? &control : NULL;
1554 	error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp);
1555 	msg->msg_iov = uiov;
1556 	if (error != 0)
1557 		goto bad;
1558 
1559 	/*
1560 	 * Note that kern_recvit() updates msg->msg_namelen.
1561 	 */
1562 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1563 		msg->msg_name = PTRIN(linux_msghdr.msg_name);
1564 		error = linux_copyout_sockaddr(sa,
1565 		    PTRIN(msg->msg_name), msg->msg_namelen);
1566 		if (error != 0)
1567 			goto bad;
1568 	}
1569 
1570 	error = bsd_to_linux_msghdr(msg, &linux_msghdr);
1571 	if (error != 0)
1572 		goto bad;
1573 
1574 	maxlen = linux_msghdr.msg_controllen;
1575 	linux_msghdr.msg_controllen = 0;
1576 	if (control) {
1577 		linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1578 
1579 		msg->msg_control = mtod(control, struct cmsghdr *);
1580 		msg->msg_controllen = control->m_len;
1581 
1582 		cm = CMSG_FIRSTHDR(msg);
1583 		outbuf = PTRIN(linux_msghdr.msg_control);
1584 		outlen = 0;
1585 		while (cm != NULL) {
1586 			linux_cmsg->cmsg_type =
1587 			    bsd_to_linux_cmsg_type(cm->cmsg_type);
1588 			linux_cmsg->cmsg_level =
1589 			    bsd_to_linux_sockopt_level(cm->cmsg_level);
1590 			if (linux_cmsg->cmsg_type == -1 ||
1591 			    cm->cmsg_level != SOL_SOCKET) {
1592 				linux_msg(curthread,
1593 				    "unsupported recvmsg cmsg level %d type %d",
1594 				    cm->cmsg_level, cm->cmsg_type);
1595 				error = EINVAL;
1596 				goto bad;
1597 			}
1598 
1599 			data = CMSG_DATA(cm);
1600 			datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1601 
1602 			switch (cm->cmsg_type) {
1603 			case SCM_RIGHTS:
1604 				if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1605 					fds = datalen / sizeof(int);
1606 					fdp = data;
1607 					for (i = 0; i < fds; i++) {
1608 						fd = *fdp++;
1609 						(void)kern_fcntl(td, fd,
1610 						    F_SETFD, FD_CLOEXEC);
1611 					}
1612 				}
1613 				break;
1614 
1615 			case SCM_CREDS:
1616 				/*
1617 				 * Currently LOCAL_CREDS is never in
1618 				 * effect for Linux so no need to worry
1619 				 * about sockcred
1620 				 */
1621 				if (datalen != sizeof(*cmcred)) {
1622 					error = EMSGSIZE;
1623 					goto bad;
1624 				}
1625 				cmcred = (struct cmsgcred *)data;
1626 				bzero(&linux_ucred, sizeof(linux_ucred));
1627 				linux_ucred.pid = cmcred->cmcred_pid;
1628 				linux_ucred.uid = cmcred->cmcred_uid;
1629 				linux_ucred.gid = cmcred->cmcred_gid;
1630 				data = &linux_ucred;
1631 				datalen = sizeof(linux_ucred);
1632 				break;
1633 
1634 			case SCM_TIMESTAMP:
1635 				if (datalen != sizeof(struct timeval)) {
1636 					error = EMSGSIZE;
1637 					goto bad;
1638 				}
1639 				ftmvl = (struct timeval *)data;
1640 				ltmvl.tv_sec = ftmvl->tv_sec;
1641 				ltmvl.tv_usec = ftmvl->tv_usec;
1642 				data = &ltmvl;
1643 				datalen = sizeof(ltmvl);
1644 				break;
1645 			}
1646 
1647 			if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1648 				if (outlen == 0) {
1649 					error = EMSGSIZE;
1650 					goto bad;
1651 				} else {
1652 					linux_msghdr.msg_flags |= LINUX_MSG_CTRUNC;
1653 					m_dispose_extcontrolm(control);
1654 					goto out;
1655 				}
1656 			}
1657 
1658 			linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1659 
1660 			error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1661 			if (error != 0)
1662 				goto bad;
1663 			outbuf += L_CMSG_HDRSZ;
1664 
1665 			error = copyout(data, outbuf, datalen);
1666 			if (error != 0)
1667 				goto bad;
1668 
1669 			outbuf += LINUX_CMSG_ALIGN(datalen);
1670 			outlen += LINUX_CMSG_LEN(datalen);
1671 
1672 			cm = CMSG_NXTHDR(msg, cm);
1673 		}
1674 		linux_msghdr.msg_controllen = outlen;
1675 	}
1676 
1677 out:
1678 	error = copyout(&linux_msghdr, msghdr, sizeof(linux_msghdr));
1679 
1680 bad:
1681 	if (control != NULL) {
1682 		if (error != 0)
1683 			m_dispose_extcontrolm(control);
1684 		m_freem(control);
1685 	}
1686 	free(iov, M_IOV);
1687 	free(linux_cmsg, M_LINUX);
1688 	free(sa, M_SONAME);
1689 
1690 	return (error);
1691 }
1692 
1693 int
1694 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1695 {
1696 	struct msghdr bsd_msg;
1697 
1698 	return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1699 	    args->flags, &bsd_msg));
1700 }
1701 
1702 int
1703 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1704 {
1705 	struct l_mmsghdr *msg;
1706 	struct msghdr bsd_msg;
1707 	struct l_timespec lts;
1708 	struct timespec ts, tts;
1709 	l_uint retval;
1710 	int error, datagrams;
1711 
1712 	if (args->timeout) {
1713 		error = copyin(args->timeout, &lts, sizeof(struct l_timespec));
1714 		if (error != 0)
1715 			return (error);
1716 		error = linux_to_native_timespec(&ts, &lts);
1717 		if (error != 0)
1718 			return (error);
1719 		getnanotime(&tts);
1720 		timespecadd(&tts, &ts, &tts);
1721 	}
1722 
1723 	msg = PTRIN(args->msg);
1724 	datagrams = 0;
1725 	while (datagrams < args->vlen) {
1726 		error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1727 		    args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1728 		if (error != 0)
1729 			break;
1730 
1731 		retval = td->td_retval[0];
1732 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1733 		if (error != 0)
1734 			break;
1735 		++msg;
1736 		++datagrams;
1737 
1738 		/*
1739 		 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1740 		 */
1741 		if (args->flags & LINUX_MSG_WAITFORONE)
1742 			args->flags |= LINUX_MSG_DONTWAIT;
1743 
1744 		/*
1745 		 * See BUGS section of recvmmsg(2).
1746 		 */
1747 		if (args->timeout) {
1748 			getnanotime(&ts);
1749 			timespecsub(&ts, &tts, &ts);
1750 			if (!timespecisset(&ts) || ts.tv_sec > 0)
1751 				break;
1752 		}
1753 		/* Out of band data, return right away. */
1754 		if (bsd_msg.msg_flags & MSG_OOB)
1755 			break;
1756 	}
1757 	if (error == 0)
1758 		td->td_retval[0] = datagrams;
1759 	return (error);
1760 }
1761 
1762 int
1763 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1764 {
1765 
1766 	return (kern_shutdown(td, args->s, args->how));
1767 }
1768 
1769 int
1770 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1771 {
1772 	l_timeval linux_tv;
1773 	struct sockaddr *sa;
1774 	struct timeval tv;
1775 	socklen_t len;
1776 	int error, level, name;
1777 
1778 	level = linux_to_bsd_sockopt_level(args->level);
1779 	switch (level) {
1780 	case SOL_SOCKET:
1781 		name = linux_to_bsd_so_sockopt(args->optname);
1782 		switch (name) {
1783 		case LOCAL_CREDS_PERSISTENT:
1784 			level = SOL_LOCAL;
1785 			break;
1786 		case SO_RCVTIMEO:
1787 			/* FALLTHROUGH */
1788 		case SO_SNDTIMEO:
1789 			error = copyin(PTRIN(args->optval), &linux_tv,
1790 			    sizeof(linux_tv));
1791 			if (error != 0)
1792 				return (error);
1793 			tv.tv_sec = linux_tv.tv_sec;
1794 			tv.tv_usec = linux_tv.tv_usec;
1795 			return (kern_setsockopt(td, args->s, level,
1796 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
1797 			/* NOTREACHED */
1798 		default:
1799 			break;
1800 		}
1801 		break;
1802 	case IPPROTO_IP:
1803 		if (args->optname == LINUX_IP_RECVERR &&
1804 		    linux_ignore_ip_recverr) {
1805 			/*
1806 			 * XXX: This is a hack to unbreak DNS resolution
1807 			 *	with glibc 2.30 and above.
1808 			 */
1809 			return (0);
1810 		}
1811 		name = linux_to_bsd_ip_sockopt(args->optname);
1812 		break;
1813 	case IPPROTO_IPV6:
1814 		name = linux_to_bsd_ip6_sockopt(args->optname);
1815 		break;
1816 	case IPPROTO_TCP:
1817 		name = linux_to_bsd_tcp_sockopt(args->optname);
1818 		break;
1819 	default:
1820 		name = -1;
1821 		break;
1822 	}
1823 	if (name < 0) {
1824 		if (name == -1)
1825 			linux_msg(curthread,
1826 			    "unsupported setsockopt level %d optname %d",
1827 			    args->level, args->optname);
1828 		return (ENOPROTOOPT);
1829 	}
1830 
1831 	if (name == IPV6_NEXTHOP) {
1832 		len = args->optlen;
1833 		error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len);
1834 		if (error != 0)
1835 			return (error);
1836 
1837 		error = kern_setsockopt(td, args->s, level,
1838 		    name, sa, UIO_SYSSPACE, len);
1839 		free(sa, M_SONAME);
1840 	} else {
1841 		error = kern_setsockopt(td, args->s, level,
1842 		    name, PTRIN(args->optval), UIO_USERSPACE, args->optlen);
1843 	}
1844 
1845 	return (error);
1846 }
1847 
1848 int
1849 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1850 {
1851 	l_timeval linux_tv;
1852 	struct timeval tv;
1853 	socklen_t tv_len, xulen, len;
1854 	struct sockaddr *sa;
1855 	struct xucred xu;
1856 	struct l_ucred lxu;
1857 	int error, level, name, newval;
1858 
1859 	level = linux_to_bsd_sockopt_level(args->level);
1860 	switch (level) {
1861 	case SOL_SOCKET:
1862 		name = linux_to_bsd_so_sockopt(args->optname);
1863 		switch (name) {
1864 		case LOCAL_CREDS_PERSISTENT:
1865 			level = SOL_LOCAL;
1866 			break;
1867 		case SO_RCVTIMEO:
1868 			/* FALLTHROUGH */
1869 		case SO_SNDTIMEO:
1870 			tv_len = sizeof(tv);
1871 			error = kern_getsockopt(td, args->s, level,
1872 			    name, &tv, UIO_SYSSPACE, &tv_len);
1873 			if (error != 0)
1874 				return (error);
1875 			linux_tv.tv_sec = tv.tv_sec;
1876 			linux_tv.tv_usec = tv.tv_usec;
1877 			return (copyout(&linux_tv, PTRIN(args->optval),
1878 			    sizeof(linux_tv)));
1879 			/* NOTREACHED */
1880 		case LOCAL_PEERCRED:
1881 			if (args->optlen < sizeof(lxu))
1882 				return (EINVAL);
1883 			/*
1884 			 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
1885 			 * but by the Unix socket's level 0.
1886 			 */
1887 			level = 0;
1888 			xulen = sizeof(xu);
1889 			error = kern_getsockopt(td, args->s, level,
1890 			    name, &xu, UIO_SYSSPACE, &xulen);
1891 			if (error != 0)
1892 				return (error);
1893 			lxu.pid = xu.cr_pid;
1894 			lxu.uid = xu.cr_uid;
1895 			lxu.gid = xu.cr_gid;
1896 			return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1897 			/* NOTREACHED */
1898 		case SO_ERROR:
1899 			len = sizeof(newval);
1900 			error = kern_getsockopt(td, args->s, level,
1901 			    name, &newval, UIO_SYSSPACE, &len);
1902 			if (error != 0)
1903 				return (error);
1904 			newval = -bsd_to_linux_errno(newval);
1905 			return (copyout(&newval, PTRIN(args->optval), len));
1906 			/* NOTREACHED */
1907 		default:
1908 			break;
1909 		}
1910 		break;
1911 	case IPPROTO_IP:
1912 		name = linux_to_bsd_ip_sockopt(args->optname);
1913 		break;
1914 	case IPPROTO_IPV6:
1915 		name = linux_to_bsd_ip6_sockopt(args->optname);
1916 		break;
1917 	case IPPROTO_TCP:
1918 		name = linux_to_bsd_tcp_sockopt(args->optname);
1919 		break;
1920 	default:
1921 		name = -1;
1922 		break;
1923 	}
1924 	if (name < 0) {
1925 		if (name == -1)
1926 			linux_msg(curthread,
1927 			    "unsupported getsockopt level %d optname %d",
1928 			    args->level, args->optname);
1929 		return (EINVAL);
1930 	}
1931 
1932 	if (name == IPV6_NEXTHOP) {
1933 		error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1934                 if (error != 0)
1935                         return (error);
1936 		sa = malloc(len, M_SONAME, M_WAITOK);
1937 
1938 		error = kern_getsockopt(td, args->s, level,
1939 		    name, sa, UIO_SYSSPACE, &len);
1940 		if (error != 0)
1941 			goto out;
1942 
1943 		error = linux_copyout_sockaddr(sa, PTRIN(args->optval), len);
1944 		if (error == 0)
1945 			error = copyout(&len, PTRIN(args->optlen),
1946 			    sizeof(len));
1947 out:
1948 		free(sa, M_SONAME);
1949 	} else {
1950 		if (args->optval) {
1951 			error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1952 			if (error != 0)
1953 				return (error);
1954 		}
1955 		error = kern_getsockopt(td, args->s, level,
1956 		    name, PTRIN(args->optval), UIO_USERSPACE, &len);
1957 		if (error == 0)
1958 			error = copyout(&len, PTRIN(args->optlen),
1959 			    sizeof(len));
1960 	}
1961 
1962 	return (error);
1963 }
1964 
1965 static int
1966 linux_sendfile_common(struct thread *td, l_int out, l_int in,
1967     l_loff_t *offset, l_size_t count)
1968 {
1969 	off_t bytes_read;
1970 	int error;
1971 	l_loff_t current_offset;
1972 	struct file *fp;
1973 
1974 	AUDIT_ARG_FD(in);
1975 	error = fget_read(td, in, &cap_pread_rights, &fp);
1976 	if (error != 0)
1977 		return (error);
1978 
1979 	if (offset != NULL) {
1980 		current_offset = *offset;
1981 	} else {
1982 		error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1983 		    fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE;
1984 		if (error != 0)
1985 			goto drop;
1986 		current_offset = td->td_uretoff.tdu_off;
1987 	}
1988 
1989 	bytes_read = 0;
1990 
1991 	/* Linux cannot have 0 count. */
1992 	if (count <= 0 || current_offset < 0) {
1993 		error = EINVAL;
1994 		goto drop;
1995 	}
1996 
1997 	error = fo_sendfile(fp, out, NULL, NULL, current_offset, count,
1998 	    &bytes_read, 0, td);
1999 	if (error != 0)
2000 		goto drop;
2001 	current_offset += bytes_read;
2002 
2003 	if (offset != NULL) {
2004 		*offset = current_offset;
2005 	} else {
2006 		error = fo_seek(fp, current_offset, SEEK_SET, td);
2007 		if (error != 0)
2008 			goto drop;
2009 	}
2010 
2011 	td->td_retval[0] = (ssize_t)bytes_read;
2012 drop:
2013 	fdrop(fp, td);
2014 	return (error);
2015 }
2016 
2017 int
2018 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg)
2019 {
2020 	/*
2021 	 * Differences between FreeBSD and Linux sendfile:
2022 	 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to
2023 	 *   mean send the whole file.)  In linux_sendfile given fds are still
2024 	 *   checked for validity when the count is 0.
2025 	 * - Linux can send to any fd whereas FreeBSD only supports sockets.
2026 	 *   The same restriction follows for linux_sendfile.
2027 	 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr.
2028 	 * - Linux takes an offset pointer and updates it to the read location.
2029 	 *   FreeBSD takes in an offset and a 'bytes read' parameter which is
2030 	 *   only filled if it isn't NULL.  We use this parameter to update the
2031 	 *   offset pointer if it exists.
2032 	 * - Linux sendfile returns bytes read on success while FreeBSD
2033 	 *   returns 0.  We use the 'bytes read' parameter to get this value.
2034 	 */
2035 
2036 	l_loff_t offset64;
2037 	l_long offset;
2038 	int ret;
2039 	int error;
2040 
2041 	if (arg->offset != NULL) {
2042 		error = copyin(arg->offset, &offset, sizeof(offset));
2043 		if (error != 0)
2044 			return (error);
2045 		offset64 = (l_loff_t)offset;
2046 	}
2047 
2048 	ret = linux_sendfile_common(td, arg->out, arg->in,
2049 	    arg->offset != NULL ? &offset64 : NULL, arg->count);
2050 
2051 	if (arg->offset != NULL) {
2052 #if defined(__i386__) || defined(__arm__) || \
2053     (defined(__amd64__) && defined(COMPAT_LINUX32))
2054 		if (offset64 > INT32_MAX)
2055 			return (EOVERFLOW);
2056 #endif
2057 		offset = (l_long)offset64;
2058 		error = copyout(&offset, arg->offset, sizeof(offset));
2059 		if (error != 0)
2060 			return (error);
2061 	}
2062 
2063 	return (ret);
2064 }
2065 
2066 #if defined(__i386__) || defined(__arm__) || \
2067     (defined(__amd64__) && defined(COMPAT_LINUX32))
2068 
2069 int
2070 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg)
2071 {
2072 	l_loff_t offset;
2073 	int ret;
2074 	int error;
2075 
2076 	if (arg->offset != NULL) {
2077 		error = copyin(arg->offset, &offset, sizeof(offset));
2078 		if (error != 0)
2079 			return (error);
2080 	}
2081 
2082 	ret = linux_sendfile_common(td, arg->out, arg->in,
2083 		arg->offset != NULL ? &offset : NULL, arg->count);
2084 
2085 	if (arg->offset != NULL) {
2086 		error = copyout(&offset, arg->offset, sizeof(offset));
2087 		if (error != 0)
2088 			return (error);
2089 	}
2090 
2091 	return (ret);
2092 }
2093 
2094 /* Argument list sizes for linux_socketcall */
2095 static const unsigned char lxs_args_cnt[] = {
2096 	0 /* unused*/,		3 /* socket */,
2097 	3 /* bind */,		3 /* connect */,
2098 	2 /* listen */,		3 /* accept */,
2099 	3 /* getsockname */,	3 /* getpeername */,
2100 	4 /* socketpair */,	4 /* send */,
2101 	4 /* recv */,		6 /* sendto */,
2102 	6 /* recvfrom */,	2 /* shutdown */,
2103 	5 /* setsockopt */,	5 /* getsockopt */,
2104 	3 /* sendmsg */,	3 /* recvmsg */,
2105 	4 /* accept4 */,	5 /* recvmmsg */,
2106 	4 /* sendmmsg */,	4 /* sendfile */
2107 };
2108 #define	LINUX_ARGS_CNT		(nitems(lxs_args_cnt) - 1)
2109 #define	LINUX_ARG_SIZE(x)	(lxs_args_cnt[x] * sizeof(l_ulong))
2110 
2111 int
2112 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
2113 {
2114 	l_ulong a[6];
2115 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2116 	register_t l_args[6];
2117 #endif
2118 	void *arg;
2119 	int error;
2120 
2121 	if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
2122 		return (EINVAL);
2123 	error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
2124 	if (error != 0)
2125 		return (error);
2126 
2127 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2128 	for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
2129 		l_args[i] = a[i];
2130 	arg = l_args;
2131 #else
2132 	arg = a;
2133 #endif
2134 	switch (args->what) {
2135 	case LINUX_SOCKET:
2136 		return (linux_socket(td, arg));
2137 	case LINUX_BIND:
2138 		return (linux_bind(td, arg));
2139 	case LINUX_CONNECT:
2140 		return (linux_connect(td, arg));
2141 	case LINUX_LISTEN:
2142 		return (linux_listen(td, arg));
2143 	case LINUX_ACCEPT:
2144 		return (linux_accept(td, arg));
2145 	case LINUX_GETSOCKNAME:
2146 		return (linux_getsockname(td, arg));
2147 	case LINUX_GETPEERNAME:
2148 		return (linux_getpeername(td, arg));
2149 	case LINUX_SOCKETPAIR:
2150 		return (linux_socketpair(td, arg));
2151 	case LINUX_SEND:
2152 		return (linux_send(td, arg));
2153 	case LINUX_RECV:
2154 		return (linux_recv(td, arg));
2155 	case LINUX_SENDTO:
2156 		return (linux_sendto(td, arg));
2157 	case LINUX_RECVFROM:
2158 		return (linux_recvfrom(td, arg));
2159 	case LINUX_SHUTDOWN:
2160 		return (linux_shutdown(td, arg));
2161 	case LINUX_SETSOCKOPT:
2162 		return (linux_setsockopt(td, arg));
2163 	case LINUX_GETSOCKOPT:
2164 		return (linux_getsockopt(td, arg));
2165 	case LINUX_SENDMSG:
2166 		return (linux_sendmsg(td, arg));
2167 	case LINUX_RECVMSG:
2168 		return (linux_recvmsg(td, arg));
2169 	case LINUX_ACCEPT4:
2170 		return (linux_accept4(td, arg));
2171 	case LINUX_RECVMMSG:
2172 		return (linux_recvmmsg(td, arg));
2173 	case LINUX_SENDMMSG:
2174 		return (linux_sendmmsg(td, arg));
2175 	case LINUX_SENDFILE:
2176 		return (linux_sendfile(td, arg));
2177 	}
2178 
2179 	linux_msg(td, "socket type %d not implemented", args->what);
2180 	return (ENOSYS);
2181 }
2182 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */
2183