xref: /freebsd/sys/compat/linux/linux_socket.c (revision 6580f5c38dd5b01aeeaed16b370f1a12423437f0)
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 <compat/freebsd32/freebsd32_util.h>
61 #include <machine/../linux32/linux.h>
62 #include <machine/../linux32/linux32_proto.h>
63 #else
64 #include <machine/../linux/linux.h>
65 #include <machine/../linux/linux_proto.h>
66 #endif
67 #include <compat/linux/linux_common.h>
68 #include <compat/linux/linux_emul.h>
69 #include <compat/linux/linux_file.h>
70 #include <compat/linux/linux_mib.h>
71 #include <compat/linux/linux_socket.h>
72 #include <compat/linux/linux_time.h>
73 #include <compat/linux/linux_util.h>
74 
75 _Static_assert(offsetof(struct l_ifreq, ifr_ifru) ==
76     offsetof(struct ifreq, ifr_ifru),
77     "Linux ifreq members names should be equal to FreeeBSD");
78 _Static_assert(offsetof(struct l_ifreq, ifr_index) ==
79     offsetof(struct ifreq, ifr_index),
80     "Linux ifreq members names should be equal to FreeeBSD");
81 _Static_assert(offsetof(struct l_ifreq, ifr_name) ==
82     offsetof(struct ifreq, ifr_name),
83     "Linux ifreq members names should be equal to FreeeBSD");
84 
85 #define	SECURITY_CONTEXT_STRING	"unconfined"
86 
87 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
88 					l_uint);
89 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
90 					l_uint, struct msghdr *);
91 static int linux_set_socket_flags(int, int *);
92 
93 #define	SOL_NETLINK	270
94 
95 static int
96 linux_to_bsd_sockopt_level(int level)
97 {
98 
99 	if (level == LINUX_SOL_SOCKET)
100 		return (SOL_SOCKET);
101 	/* Remaining values are RFC-defined protocol numbers. */
102 	return (level);
103 }
104 
105 static int
106 bsd_to_linux_sockopt_level(int level)
107 {
108 
109 	if (level == SOL_SOCKET)
110 		return (LINUX_SOL_SOCKET);
111 	return (level);
112 }
113 
114 static int
115 linux_to_bsd_ip_sockopt(int opt)
116 {
117 
118 	switch (opt) {
119 	/* known and translated sockopts */
120 	case LINUX_IP_TOS:
121 		return (IP_TOS);
122 	case LINUX_IP_TTL:
123 		return (IP_TTL);
124 	case LINUX_IP_HDRINCL:
125 		return (IP_HDRINCL);
126 	case LINUX_IP_OPTIONS:
127 		return (IP_OPTIONS);
128 	case LINUX_IP_RECVOPTS:
129 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVOPTS");
130 		return (IP_RECVOPTS);
131 	case LINUX_IP_RETOPTS:
132 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_REETOPTS");
133 		return (IP_RETOPTS);
134 	case LINUX_IP_RECVTTL:
135 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTTL");
136 		return (IP_RECVTTL);
137 	case LINUX_IP_RECVTOS:
138 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTOS");
139 		return (IP_RECVTOS);
140 	case LINUX_IP_FREEBIND:
141 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_FREEBIND");
142 		return (IP_BINDANY);
143 	case LINUX_IP_IPSEC_POLICY:
144 		/* we have this option, but not documented in ip(4) manpage */
145 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_IPSEC_POLICY");
146 		return (IP_IPSEC_POLICY);
147 	case LINUX_IP_MINTTL:
148 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MINTTL");
149 		return (IP_MINTTL);
150 	case LINUX_IP_MULTICAST_IF:
151 		return (IP_MULTICAST_IF);
152 	case LINUX_IP_MULTICAST_TTL:
153 		return (IP_MULTICAST_TTL);
154 	case LINUX_IP_MULTICAST_LOOP:
155 		return (IP_MULTICAST_LOOP);
156 	case LINUX_IP_ADD_MEMBERSHIP:
157 		return (IP_ADD_MEMBERSHIP);
158 	case LINUX_IP_DROP_MEMBERSHIP:
159 		return (IP_DROP_MEMBERSHIP);
160 	case LINUX_IP_UNBLOCK_SOURCE:
161 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_UNBLOCK_SOURCE");
162 		return (IP_UNBLOCK_SOURCE);
163 	case LINUX_IP_BLOCK_SOURCE:
164 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_BLOCK_SOURCE");
165 		return (IP_BLOCK_SOURCE);
166 	case LINUX_IP_ADD_SOURCE_MEMBERSHIP:
167 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_ADD_SOURCE_MEMBERSHIP");
168 		return (IP_ADD_SOURCE_MEMBERSHIP);
169 	case LINUX_IP_DROP_SOURCE_MEMBERSHIP:
170 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_DROP_SOURCE_MEMBERSHIP");
171 		return (IP_DROP_SOURCE_MEMBERSHIP);
172 	case LINUX_MCAST_JOIN_GROUP:
173 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_GROUP");
174 		return (MCAST_JOIN_GROUP);
175 	case LINUX_MCAST_LEAVE_GROUP:
176 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_GROUP");
177 		return (MCAST_LEAVE_GROUP);
178 	case LINUX_MCAST_JOIN_SOURCE_GROUP:
179 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_JOIN_SOURCE_GROUP");
180 		return (MCAST_JOIN_SOURCE_GROUP);
181 	case LINUX_MCAST_LEAVE_SOURCE_GROUP:
182 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_MCAST_LEAVE_SOURCE_GROUP");
183 		return (MCAST_LEAVE_SOURCE_GROUP);
184 	case LINUX_IP_RECVORIGDSTADDR:
185 		return (IP_RECVORIGDSTADDR);
186 
187 	/* known but not implemented sockopts */
188 	case LINUX_IP_ROUTER_ALERT:
189 		LINUX_RATELIMIT_MSG_OPT1(
190 		    "unsupported IPv4 socket option IP_ROUTER_ALERT (%d), you can not do user-space routing from linux programs",
191 		    opt);
192 		return (-2);
193 	case LINUX_IP_PKTINFO:
194 		LINUX_RATELIMIT_MSG_OPT1(
195 		    "unsupported IPv4 socket option IP_PKTINFO (%d), you can not get extended packet info for datagram sockets in linux programs",
196 		    opt);
197 		return (-2);
198 	case LINUX_IP_PKTOPTIONS:
199 		LINUX_RATELIMIT_MSG_OPT1(
200 		    "unsupported IPv4 socket option IP_PKTOPTIONS (%d)",
201 		    opt);
202 		return (-2);
203 	case LINUX_IP_MTU_DISCOVER:
204 		LINUX_RATELIMIT_MSG_OPT1(
205 		    "unsupported IPv4 socket option IP_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery",
206 		    opt);
207 		return (-2);
208 	case LINUX_IP_RECVERR:
209 		/* needed by steam */
210 		LINUX_RATELIMIT_MSG_OPT1(
211 		    "unsupported IPv4 socket option IP_RECVERR (%d), you can not get extended reliability info in linux programs",
212 		    opt);
213 		return (-2);
214 	case LINUX_IP_MTU:
215 		LINUX_RATELIMIT_MSG_OPT1(
216 		    "unsupported IPv4 socket option IP_MTU (%d), your linux program can not control the MTU on this socket",
217 		    opt);
218 		return (-2);
219 	case LINUX_IP_XFRM_POLICY:
220 		LINUX_RATELIMIT_MSG_OPT1(
221 		    "unsupported IPv4 socket option IP_XFRM_POLICY (%d)",
222 		    opt);
223 		return (-2);
224 	case LINUX_IP_PASSSEC:
225 		/* needed by steam */
226 		LINUX_RATELIMIT_MSG_OPT1(
227 		    "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",
228 		    opt);
229 		return (-2);
230 	case LINUX_IP_TRANSPARENT:
231 		/* IP_BINDANY or more? */
232 		LINUX_RATELIMIT_MSG_OPT1(
233 		    "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",
234 		    opt);
235 		return (-2);
236 	case LINUX_IP_NODEFRAG:
237 		LINUX_RATELIMIT_MSG_OPT1(
238 		    "unsupported IPv4 socket option IP_NODEFRAG (%d)",
239 		    opt);
240 		return (-2);
241 	case LINUX_IP_CHECKSUM:
242 		LINUX_RATELIMIT_MSG_OPT1(
243 		    "unsupported IPv4 socket option IP_CHECKSUM (%d)",
244 		    opt);
245 		return (-2);
246 	case LINUX_IP_BIND_ADDRESS_NO_PORT:
247 		LINUX_RATELIMIT_MSG_OPT1(
248 		    "unsupported IPv4 socket option IP_BIND_ADDRESS_NO_PORT (%d)",
249 		    opt);
250 		return (-2);
251 	case LINUX_IP_RECVFRAGSIZE:
252 		LINUX_RATELIMIT_MSG_OPT1(
253 		    "unsupported IPv4 socket option IP_RECVFRAGSIZE (%d)",
254 		    opt);
255 		return (-2);
256 	case LINUX_MCAST_MSFILTER:
257 		LINUX_RATELIMIT_MSG_OPT1(
258 		    "unsupported IPv4 socket option IP_MCAST_MSFILTER (%d)",
259 		    opt);
260 		return (-2);
261 	case LINUX_IP_MULTICAST_ALL:
262 		LINUX_RATELIMIT_MSG_OPT1(
263 		    "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",
264 		    opt);
265 		return (-2);
266 	case LINUX_IP_UNICAST_IF:
267 		LINUX_RATELIMIT_MSG_OPT1(
268 		    "unsupported IPv4 socket option IP_UNICAST_IF (%d)",
269 		    opt);
270 		return (-2);
271 
272 	/* unknown sockopts */
273 	default:
274 		return (-1);
275 	}
276 }
277 
278 static int
279 linux_to_bsd_ip6_sockopt(int opt)
280 {
281 
282 	switch (opt) {
283 	/* known and translated sockopts */
284 	case LINUX_IPV6_2292PKTINFO:
285 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTINFO");
286 		return (IPV6_2292PKTINFO);
287 	case LINUX_IPV6_2292HOPOPTS:
288 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPOPTS");
289 		return (IPV6_2292HOPOPTS);
290 	case LINUX_IPV6_2292DSTOPTS:
291 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292DSTOPTS");
292 		return (IPV6_2292DSTOPTS);
293 	case LINUX_IPV6_2292RTHDR:
294 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292RTHDR");
295 		return (IPV6_2292RTHDR);
296 	case LINUX_IPV6_2292PKTOPTIONS:
297 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292PKTOPTIONS");
298 		return (IPV6_2292PKTOPTIONS);
299 	case LINUX_IPV6_CHECKSUM:
300 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_CHECKSUM");
301 		return (IPV6_CHECKSUM);
302 	case LINUX_IPV6_2292HOPLIMIT:
303 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_2292HOPLIMIT");
304 		return (IPV6_2292HOPLIMIT);
305 	case LINUX_IPV6_NEXTHOP:
306 		return (IPV6_NEXTHOP);
307 	case LINUX_IPV6_UNICAST_HOPS:
308 		return (IPV6_UNICAST_HOPS);
309 	case LINUX_IPV6_MULTICAST_IF:
310 		return (IPV6_MULTICAST_IF);
311 	case LINUX_IPV6_MULTICAST_HOPS:
312 		return (IPV6_MULTICAST_HOPS);
313 	case LINUX_IPV6_MULTICAST_LOOP:
314 		return (IPV6_MULTICAST_LOOP);
315 	case LINUX_IPV6_ADD_MEMBERSHIP:
316 		return (IPV6_JOIN_GROUP);
317 	case LINUX_IPV6_DROP_MEMBERSHIP:
318 		return (IPV6_LEAVE_GROUP);
319 	case LINUX_IPV6_V6ONLY:
320 		return (IPV6_V6ONLY);
321 	case LINUX_IPV6_IPSEC_POLICY:
322 		/* we have this option, but not documented in ip6(4) manpage */
323 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_IPSEC_POLICY");
324 		return (IPV6_IPSEC_POLICY);
325 	case LINUX_MCAST_JOIN_GROUP:
326 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_JOIN_GROUP");
327 		return (IPV6_JOIN_GROUP);
328 	case LINUX_MCAST_LEAVE_GROUP:
329 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_LEAVE_GROUP");
330 		return (IPV6_LEAVE_GROUP);
331 	case LINUX_IPV6_RECVPKTINFO:
332 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPKTINFO");
333 		return (IPV6_RECVPKTINFO);
334 	case LINUX_IPV6_PKTINFO:
335 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PKTINFO");
336 		return (IPV6_PKTINFO);
337 	case LINUX_IPV6_RECVHOPLIMIT:
338 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPLIMIT");
339 		return (IPV6_RECVHOPLIMIT);
340 	case LINUX_IPV6_HOPLIMIT:
341 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPLIMIT");
342 		return (IPV6_HOPLIMIT);
343 	case LINUX_IPV6_RECVHOPOPTS:
344 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVHOPOPTS");
345 		return (IPV6_RECVHOPOPTS);
346 	case LINUX_IPV6_HOPOPTS:
347 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_HOPOPTS");
348 		return (IPV6_HOPOPTS);
349 	case LINUX_IPV6_RTHDRDSTOPTS:
350 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDRDSTOPTS");
351 		return (IPV6_RTHDRDSTOPTS);
352 	case LINUX_IPV6_RECVRTHDR:
353 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVRTHDR");
354 		return (IPV6_RECVRTHDR);
355 	case LINUX_IPV6_RTHDR:
356 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RTHDR");
357 		return (IPV6_RTHDR);
358 	case LINUX_IPV6_RECVDSTOPTS:
359 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVDSTOPTS");
360 		return (IPV6_RECVDSTOPTS);
361 	case LINUX_IPV6_DSTOPTS:
362 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_DSTOPTS");
363 		return (IPV6_DSTOPTS);
364 	case LINUX_IPV6_RECVPATHMTU:
365 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_RECVPATHMTU");
366 		return (IPV6_RECVPATHMTU);
367 	case LINUX_IPV6_PATHMTU:
368 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_PATHMTU");
369 		return (IPV6_PATHMTU);
370 	case LINUX_IPV6_DONTFRAG:
371 		return (IPV6_DONTFRAG);
372 	case LINUX_IPV6_AUTOFLOWLABEL:
373 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_AUTOFLOWLABEL");
374 		return (IPV6_AUTOFLOWLABEL);
375 	case LINUX_IPV6_ORIGDSTADDR:
376 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_ORIGDSTADDR");
377 		return (IPV6_ORIGDSTADDR);
378 	case LINUX_IPV6_FREEBIND:
379 		LINUX_RATELIMIT_MSG_NOTTESTED("IPv6 socket option IPV6_FREEBIND");
380 		return (IPV6_BINDANY);
381 
382 	/* known but not implemented sockopts */
383 	case LINUX_IPV6_ADDRFORM:
384 		LINUX_RATELIMIT_MSG_OPT1(
385 		    "unsupported IPv6 socket option IPV6_ADDRFORM (%d), you linux program can not convert the socket to IPv4",
386 		    opt);
387 		return (-2);
388 	case LINUX_IPV6_AUTHHDR:
389 		LINUX_RATELIMIT_MSG_OPT1(
390 		    "unsupported IPv6 socket option IPV6_AUTHHDR (%d), your linux program can not get the authentication header info of IPv6 packets",
391 		    opt);
392 		return (-2);
393 	case LINUX_IPV6_FLOWINFO:
394 		LINUX_RATELIMIT_MSG_OPT1(
395 		    "unsupported IPv6 socket option IPV6_FLOWINFO (%d), your linux program can not get the flowid of IPv6 packets",
396 		    opt);
397 		return (-2);
398 	case LINUX_IPV6_ROUTER_ALERT:
399 		LINUX_RATELIMIT_MSG_OPT1(
400 		    "unsupported IPv6 socket option IPV6_ROUTER_ALERT (%d), you can not do user-space routing from linux programs",
401 		    opt);
402 		return (-2);
403 	case LINUX_IPV6_MTU_DISCOVER:
404 		LINUX_RATELIMIT_MSG_OPT1(
405 		    "unsupported IPv6 socket option IPV6_MTU_DISCOVER (%d), your linux program can not control path-MTU discovery",
406 		    opt);
407 		return (-2);
408 	case LINUX_IPV6_MTU:
409 		LINUX_RATELIMIT_MSG_OPT1(
410 		    "unsupported IPv6 socket option IPV6_MTU (%d), your linux program can not control the MTU on this socket",
411 		    opt);
412 		return (-2);
413 	case LINUX_IPV6_JOIN_ANYCAST:
414 		LINUX_RATELIMIT_MSG_OPT1(
415 		    "unsupported IPv6 socket option IPV6_JOIN_ANYCAST (%d)",
416 		    opt);
417 		return (-2);
418 	case LINUX_IPV6_LEAVE_ANYCAST:
419 		LINUX_RATELIMIT_MSG_OPT1(
420 		    "unsupported IPv6 socket option IPV6_LEAVE_ANYCAST (%d)",
421 		    opt);
422 		return (-2);
423 	case LINUX_IPV6_MULTICAST_ALL:
424 		LINUX_RATELIMIT_MSG_OPT1(
425 		    "unsupported IPv6 socket option IPV6_MULTICAST_ALL (%d)",
426 		    opt);
427 		return (-2);
428 	case LINUX_IPV6_ROUTER_ALERT_ISOLATE:
429 		LINUX_RATELIMIT_MSG_OPT1(
430 		    "unsupported IPv6 socket option IPV6_ROUTER_ALERT_ISOLATE (%d)",
431 		    opt);
432 		return (-2);
433 	case LINUX_IPV6_FLOWLABEL_MGR:
434 		LINUX_RATELIMIT_MSG_OPT1(
435 		    "unsupported IPv6 socket option IPV6_FLOWLABEL_MGR (%d)",
436 		    opt);
437 		return (-2);
438 	case LINUX_IPV6_FLOWINFO_SEND:
439 		LINUX_RATELIMIT_MSG_OPT1(
440 		    "unsupported IPv6 socket option IPV6_FLOWINFO_SEND (%d)",
441 		    opt);
442 		return (-2);
443 	case LINUX_IPV6_XFRM_POLICY:
444 		LINUX_RATELIMIT_MSG_OPT1(
445 		    "unsupported IPv6 socket option IPV6_XFRM_POLICY (%d)",
446 		    opt);
447 		return (-2);
448 	case LINUX_IPV6_HDRINCL:
449 		LINUX_RATELIMIT_MSG_OPT1(
450 		    "unsupported IPv6 socket option IPV6_HDRINCL (%d)",
451 		    opt);
452 		return (-2);
453 	case LINUX_MCAST_BLOCK_SOURCE:
454 		LINUX_RATELIMIT_MSG_OPT1(
455 		    "unsupported IPv6 socket option MCAST_BLOCK_SOURCE (%d), your linux program may see more multicast stuff than it wants",
456 		    opt);
457 		return (-2);
458 	case LINUX_MCAST_UNBLOCK_SOURCE:
459 		LINUX_RATELIMIT_MSG_OPT1(
460 		    "unsupported IPv6 socket option MCAST_UNBLOCK_SOURCE (%d), your linux program may not see all the multicast stuff it wants",
461 		    opt);
462 		return (-2);
463 	case LINUX_MCAST_JOIN_SOURCE_GROUP:
464 		LINUX_RATELIMIT_MSG_OPT1(
465 		    "unsupported IPv6 socket option MCAST_JOIN_SOURCE_GROUP (%d), your linux program is not able to join a multicast source group",
466 		    opt);
467 		return (-2);
468 	case LINUX_MCAST_LEAVE_SOURCE_GROUP:
469 		LINUX_RATELIMIT_MSG_OPT1(
470 		    "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",
471 		    opt);
472 		return (-2);
473 	case LINUX_MCAST_MSFILTER:
474 		LINUX_RATELIMIT_MSG_OPT1(
475 		    "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",
476 		    opt);
477 		return (-2);
478 	case LINUX_IPV6_ADDR_PREFERENCES:
479 		LINUX_RATELIMIT_MSG_OPT1(
480 		    "unsupported IPv6 socket option IPV6_ADDR_PREFERENCES (%d)",
481 		    opt);
482 		return (-2);
483 	case LINUX_IPV6_MINHOPCOUNT:
484 		LINUX_RATELIMIT_MSG_OPT1(
485 		    "unsupported IPv6 socket option IPV6_MINHOPCOUNT (%d)",
486 		    opt);
487 		return (-2);
488 	case LINUX_IPV6_TRANSPARENT:
489 		/* IP_BINDANY or more? */
490 		LINUX_RATELIMIT_MSG_OPT1(
491 		    "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",
492 		    opt);
493 		return (-2);
494 	case LINUX_IPV6_UNICAST_IF:
495 		LINUX_RATELIMIT_MSG_OPT1(
496 		    "unsupported IPv6 socket option IPV6_UNICAST_IF (%d)",
497 		    opt);
498 		return (-2);
499 	case LINUX_IPV6_RECVFRAGSIZE:
500 		LINUX_RATELIMIT_MSG_OPT1(
501 		    "unsupported IPv6 socket option IPV6_RECVFRAGSIZE (%d)",
502 		    opt);
503 		return (-2);
504 	case LINUX_IPV6_RECVERR:
505 		LINUX_RATELIMIT_MSG_OPT1(
506 		    "unsupported IPv6 socket option IPV6_RECVERR (%d), you can not get extended reliability info in linux programs",
507 		    opt);
508 		return (-2);
509 
510 	/* unknown sockopts */
511 	default:
512 		return (-1);
513 	}
514 }
515 
516 static int
517 linux_to_bsd_so_sockopt(int opt)
518 {
519 
520 	switch (opt) {
521 	case LINUX_SO_DEBUG:
522 		return (SO_DEBUG);
523 	case LINUX_SO_REUSEADDR:
524 		return (SO_REUSEADDR);
525 	case LINUX_SO_TYPE:
526 		return (SO_TYPE);
527 	case LINUX_SO_ERROR:
528 		return (SO_ERROR);
529 	case LINUX_SO_DONTROUTE:
530 		return (SO_DONTROUTE);
531 	case LINUX_SO_BROADCAST:
532 		return (SO_BROADCAST);
533 	case LINUX_SO_SNDBUF:
534 	case LINUX_SO_SNDBUFFORCE:
535 		return (SO_SNDBUF);
536 	case LINUX_SO_RCVBUF:
537 	case LINUX_SO_RCVBUFFORCE:
538 		return (SO_RCVBUF);
539 	case LINUX_SO_KEEPALIVE:
540 		return (SO_KEEPALIVE);
541 	case LINUX_SO_OOBINLINE:
542 		return (SO_OOBINLINE);
543 	case LINUX_SO_LINGER:
544 		return (SO_LINGER);
545 	case LINUX_SO_REUSEPORT:
546 		return (SO_REUSEPORT_LB);
547 	case LINUX_SO_PASSCRED:
548 		return (LOCAL_CREDS_PERSISTENT);
549 	case LINUX_SO_PEERCRED:
550 		return (LOCAL_PEERCRED);
551 	case LINUX_SO_RCVLOWAT:
552 		return (SO_RCVLOWAT);
553 	case LINUX_SO_SNDLOWAT:
554 		return (SO_SNDLOWAT);
555 	case LINUX_SO_RCVTIMEO:
556 		return (SO_RCVTIMEO);
557 	case LINUX_SO_SNDTIMEO:
558 		return (SO_SNDTIMEO);
559 	case LINUX_SO_TIMESTAMPO:
560 	case LINUX_SO_TIMESTAMPN:
561 		return (SO_TIMESTAMP);
562 	case LINUX_SO_TIMESTAMPNSO:
563 	case LINUX_SO_TIMESTAMPNSN:
564 		return (SO_BINTIME);
565 	case LINUX_SO_ACCEPTCONN:
566 		return (SO_ACCEPTCONN);
567 	case LINUX_SO_PROTOCOL:
568 		return (SO_PROTOCOL);
569 	case LINUX_SO_DOMAIN:
570 		return (SO_DOMAIN);
571 	}
572 	return (-1);
573 }
574 
575 static int
576 linux_to_bsd_tcp_sockopt(int opt)
577 {
578 
579 	switch (opt) {
580 	case LINUX_TCP_NODELAY:
581 		return (TCP_NODELAY);
582 	case LINUX_TCP_MAXSEG:
583 		return (TCP_MAXSEG);
584 	case LINUX_TCP_CORK:
585 		return (TCP_NOPUSH);
586 	case LINUX_TCP_KEEPIDLE:
587 		return (TCP_KEEPIDLE);
588 	case LINUX_TCP_KEEPINTVL:
589 		return (TCP_KEEPINTVL);
590 	case LINUX_TCP_KEEPCNT:
591 		return (TCP_KEEPCNT);
592 	case LINUX_TCP_INFO:
593 		LINUX_RATELIMIT_MSG_OPT1(
594 		    "unsupported TCP socket option TCP_INFO (%d)", opt);
595 		return (-2);
596 	case LINUX_TCP_MD5SIG:
597 		return (TCP_MD5SIG);
598 	}
599 	return (-1);
600 }
601 
602 static int
603 linux_to_bsd_msg_flags(int flags)
604 {
605 	int ret_flags = 0;
606 
607 	if (flags & LINUX_MSG_OOB)
608 		ret_flags |= MSG_OOB;
609 	if (flags & LINUX_MSG_PEEK)
610 		ret_flags |= MSG_PEEK;
611 	if (flags & LINUX_MSG_DONTROUTE)
612 		ret_flags |= MSG_DONTROUTE;
613 	if (flags & LINUX_MSG_CTRUNC)
614 		ret_flags |= MSG_CTRUNC;
615 	if (flags & LINUX_MSG_TRUNC)
616 		ret_flags |= MSG_TRUNC;
617 	if (flags & LINUX_MSG_DONTWAIT)
618 		ret_flags |= MSG_DONTWAIT;
619 	if (flags & LINUX_MSG_EOR)
620 		ret_flags |= MSG_EOR;
621 	if (flags & LINUX_MSG_WAITALL)
622 		ret_flags |= MSG_WAITALL;
623 	if (flags & LINUX_MSG_NOSIGNAL)
624 		ret_flags |= MSG_NOSIGNAL;
625 	if (flags & LINUX_MSG_PROXY)
626 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_PROXY (%d) not handled",
627 		    LINUX_MSG_PROXY);
628 	if (flags & LINUX_MSG_FIN)
629 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_FIN (%d) not handled",
630 		    LINUX_MSG_FIN);
631 	if (flags & LINUX_MSG_SYN)
632 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_SYN (%d) not handled",
633 		    LINUX_MSG_SYN);
634 	if (flags & LINUX_MSG_CONFIRM)
635 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_CONFIRM (%d) not handled",
636 		    LINUX_MSG_CONFIRM);
637 	if (flags & LINUX_MSG_RST)
638 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_RST (%d) not handled",
639 		    LINUX_MSG_RST);
640 	if (flags & LINUX_MSG_ERRQUEUE)
641 		LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_ERRQUEUE (%d) not handled",
642 		    LINUX_MSG_ERRQUEUE);
643 	return (ret_flags);
644 }
645 
646 static int
647 linux_to_bsd_cmsg_type(int cmsg_type)
648 {
649 
650 	switch (cmsg_type) {
651 	case LINUX_SCM_RIGHTS:
652 		return (SCM_RIGHTS);
653 	case LINUX_SCM_CREDENTIALS:
654 		return (SCM_CREDS);
655 	}
656 	return (-1);
657 }
658 
659 static int
660 bsd_to_linux_ip_cmsg_type(int cmsg_type)
661 {
662 
663 	switch (cmsg_type) {
664 	case IP_RECVORIGDSTADDR:
665 		return (LINUX_IP_RECVORIGDSTADDR);
666 	}
667 	return (-1);
668 }
669 
670 static int
671 bsd_to_linux_cmsg_type(struct proc *p, int cmsg_type, int cmsg_level)
672 {
673 	struct linux_pemuldata *pem;
674 
675 	if (cmsg_level == IPPROTO_IP)
676 		return (bsd_to_linux_ip_cmsg_type(cmsg_type));
677 	if (cmsg_level != SOL_SOCKET)
678 		return (-1);
679 
680 	pem = pem_find(p);
681 
682 	switch (cmsg_type) {
683 	case SCM_RIGHTS:
684 		return (LINUX_SCM_RIGHTS);
685 	case SCM_CREDS:
686 		return (LINUX_SCM_CREDENTIALS);
687 	case SCM_CREDS2:
688 		return (LINUX_SCM_CREDENTIALS);
689 	case SCM_TIMESTAMP:
690 		return (pem->so_timestamp);
691 	case SCM_BINTIME:
692 		return (pem->so_timestampns);
693 	}
694 	return (-1);
695 }
696 
697 static int
698 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
699 {
700 	if (lhdr->msg_controllen > INT_MAX)
701 		return (ENOBUFS);
702 
703 	bhdr->msg_name		= PTRIN(lhdr->msg_name);
704 	bhdr->msg_namelen	= lhdr->msg_namelen;
705 	bhdr->msg_iov		= PTRIN(lhdr->msg_iov);
706 	bhdr->msg_iovlen	= lhdr->msg_iovlen;
707 	bhdr->msg_control	= PTRIN(lhdr->msg_control);
708 
709 	/*
710 	 * msg_controllen is skipped since BSD and LINUX control messages
711 	 * are potentially different sizes (e.g. the cred structure used
712 	 * by SCM_CREDS is different between the two operating system).
713 	 *
714 	 * The caller can set it (if necessary) after converting all the
715 	 * control messages.
716 	 */
717 
718 	bhdr->msg_flags		= linux_to_bsd_msg_flags(lhdr->msg_flags);
719 	return (0);
720 }
721 
722 static int
723 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
724 {
725 	lhdr->msg_name		= PTROUT(bhdr->msg_name);
726 	lhdr->msg_namelen	= bhdr->msg_namelen;
727 	lhdr->msg_iov		= PTROUT(bhdr->msg_iov);
728 	lhdr->msg_iovlen	= bhdr->msg_iovlen;
729 	lhdr->msg_control	= PTROUT(bhdr->msg_control);
730 
731 	/*
732 	 * msg_controllen is skipped since BSD and LINUX control messages
733 	 * are potentially different sizes (e.g. the cred structure used
734 	 * by SCM_CREDS is different between the two operating system).
735 	 *
736 	 * The caller can set it (if necessary) after converting all the
737 	 * control messages.
738 	 */
739 
740 	/* msg_flags skipped */
741 	return (0);
742 }
743 
744 static int
745 linux_set_socket_flags(int lflags, int *flags)
746 {
747 
748 	if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
749 		return (EINVAL);
750 	if (lflags & LINUX_SOCK_NONBLOCK)
751 		*flags |= SOCK_NONBLOCK;
752 	if (lflags & LINUX_SOCK_CLOEXEC)
753 		*flags |= SOCK_CLOEXEC;
754 	return (0);
755 }
756 
757 static int
758 linux_copyout_sockaddr(const struct sockaddr *sa, void *uaddr, size_t len)
759 {
760 	struct l_sockaddr *lsa;
761 	int error;
762 
763 	error = bsd_to_linux_sockaddr(sa, &lsa, len);
764 	if (error != 0)
765 		return (error);
766 
767 	error = copyout(lsa, uaddr, len);
768 	free(lsa, M_LINUX);
769 
770 	return (error);
771 }
772 
773 static int
774 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
775     struct mbuf *control, enum uio_seg segflg)
776 {
777 	struct sockaddr *to;
778 	int error, len;
779 
780 	if (mp->msg_name != NULL) {
781 		len = mp->msg_namelen;
782 		error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len);
783 		if (error != 0)
784 			return (error);
785 		mp->msg_name = to;
786 	} else
787 		to = NULL;
788 
789 	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
790 	    segflg);
791 
792 	if (to)
793 		free(to, M_SONAME);
794 	return (error);
795 }
796 
797 /* Return 0 if IP_HDRINCL is set for the given socket. */
798 static int
799 linux_check_hdrincl(struct thread *td, int s)
800 {
801 	int error, optval;
802 	socklen_t size_val;
803 
804 	size_val = sizeof(optval);
805 	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
806 	    &optval, UIO_SYSSPACE, &size_val);
807 	if (error != 0)
808 		return (error);
809 
810 	return (optval == 0);
811 }
812 
813 /*
814  * Updated sendto() when IP_HDRINCL is set:
815  * tweak endian-dependent fields in the IP packet.
816  */
817 static int
818 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
819 {
820 /*
821  * linux_ip_copysize defines how many bytes we should copy
822  * from the beginning of the IP packet before we customize it for BSD.
823  * It should include all the fields we modify (ip_len and ip_off).
824  */
825 #define linux_ip_copysize	8
826 
827 	struct ip *packet;
828 	struct msghdr msg;
829 	struct iovec aiov[1];
830 	int error;
831 
832 	/* Check that the packet isn't too big or too small. */
833 	if (linux_args->len < linux_ip_copysize ||
834 	    linux_args->len > IP_MAXPACKET)
835 		return (EINVAL);
836 
837 	packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
838 
839 	/* Make kernel copy of the packet to be sent */
840 	if ((error = copyin(PTRIN(linux_args->msg), packet,
841 	    linux_args->len)))
842 		goto goout;
843 
844 	/* Convert fields from Linux to BSD raw IP socket format */
845 	packet->ip_len = linux_args->len;
846 	packet->ip_off = ntohs(packet->ip_off);
847 
848 	/* Prepare the msghdr and iovec structures describing the new packet */
849 	msg.msg_name = PTRIN(linux_args->to);
850 	msg.msg_namelen = linux_args->tolen;
851 	msg.msg_iov = aiov;
852 	msg.msg_iovlen = 1;
853 	msg.msg_control = NULL;
854 	msg.msg_flags = 0;
855 	aiov[0].iov_base = (char *)packet;
856 	aiov[0].iov_len = linux_args->len;
857 	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
858 	    NULL, UIO_SYSSPACE);
859 goout:
860 	free(packet, M_LINUX);
861 	return (error);
862 }
863 
864 static const char *linux_netlink_names[] = {
865 	[LINUX_NETLINK_ROUTE] = "ROUTE",
866 	[LINUX_NETLINK_SOCK_DIAG] = "SOCK_DIAG",
867 	[LINUX_NETLINK_NFLOG] = "NFLOG",
868 	[LINUX_NETLINK_SELINUX] = "SELINUX",
869 	[LINUX_NETLINK_AUDIT] = "AUDIT",
870 	[LINUX_NETLINK_FIB_LOOKUP] = "FIB_LOOKUP",
871 	[LINUX_NETLINK_NETFILTER] = "NETFILTER",
872 	[LINUX_NETLINK_KOBJECT_UEVENT] = "KOBJECT_UEVENT",
873 };
874 
875 int
876 linux_socket(struct thread *td, struct linux_socket_args *args)
877 {
878 	int retval_socket, type;
879 	sa_family_t domain;
880 
881 	type = args->type & LINUX_SOCK_TYPE_MASK;
882 	if (type < 0 || type > LINUX_SOCK_MAX)
883 		return (EINVAL);
884 	retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
885 		&type);
886 	if (retval_socket != 0)
887 		return (retval_socket);
888 	domain = linux_to_bsd_domain(args->domain);
889 	if (domain == AF_UNKNOWN) {
890 		/* Mask off SOCK_NONBLOCK / CLOEXEC for error messages. */
891 		type = args->type & LINUX_SOCK_TYPE_MASK;
892 		if (args->domain == LINUX_AF_NETLINK &&
893 		    args->protocol == LINUX_NETLINK_AUDIT) {
894 			; /* Do nothing, quietly. */
895 		} else if (args->domain == LINUX_AF_NETLINK) {
896 			const char *nl_name;
897 
898 			if (args->protocol >= 0 &&
899 			    args->protocol < nitems(linux_netlink_names))
900 				nl_name = linux_netlink_names[args->protocol];
901 			else
902 				nl_name = NULL;
903 			if (nl_name != NULL)
904 				linux_msg(curthread,
905 				    "unsupported socket(AF_NETLINK, %d, "
906 				    "NETLINK_%s)", type, nl_name);
907 			else
908 				linux_msg(curthread,
909 				    "unsupported socket(AF_NETLINK, %d, %d)",
910 				    type, args->protocol);
911 		} else {
912 			linux_msg(curthread, "unsupported socket domain %d, "
913 			    "type %d, protocol %d", args->domain, type,
914 			    args->protocol);
915 		}
916 		return (EAFNOSUPPORT);
917 	}
918 
919 	retval_socket = kern_socket(td, domain, type, args->protocol);
920 	if (retval_socket)
921 		return (retval_socket);
922 
923 	if (type == SOCK_RAW
924 	    && (args->protocol == IPPROTO_RAW || args->protocol == 0)
925 	    && domain == PF_INET) {
926 		/* It's a raw IP socket: set the IP_HDRINCL option. */
927 		int hdrincl;
928 
929 		hdrincl = 1;
930 		/* We ignore any error returned by kern_setsockopt() */
931 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
932 		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
933 	}
934 #ifdef INET6
935 	/*
936 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
937 	 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
938 	 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
939 	 * sysctl value.
940 	 */
941 	if (domain == PF_INET6) {
942 		int v6only;
943 
944 		v6only = 0;
945 		/* We ignore any error returned by setsockopt() */
946 		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
947 		    &v6only, UIO_SYSSPACE, sizeof(v6only));
948 	}
949 #endif
950 
951 	return (retval_socket);
952 }
953 
954 int
955 linux_bind(struct thread *td, struct linux_bind_args *args)
956 {
957 	struct sockaddr *sa;
958 	int error;
959 
960 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
961 	    &args->namelen);
962 	if (error != 0)
963 		return (error);
964 
965 	error = kern_bindat(td, AT_FDCWD, args->s, sa);
966 	free(sa, M_SONAME);
967 
968 	/* XXX */
969 	if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
970 		return (EINVAL);
971 	return (error);
972 }
973 
974 int
975 linux_connect(struct thread *td, struct linux_connect_args *args)
976 {
977 	struct socket *so;
978 	struct sockaddr *sa;
979 	struct file *fp;
980 	int error;
981 
982 	error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
983 	    &args->namelen);
984 	if (error != 0)
985 		return (error);
986 
987 	error = kern_connectat(td, AT_FDCWD, args->s, sa);
988 	free(sa, M_SONAME);
989 	if (error != EISCONN)
990 		return (error);
991 
992 	/*
993 	 * Linux doesn't return EISCONN the first time it occurs,
994 	 * when on a non-blocking socket. Instead it returns the
995 	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
996 	 */
997 	error = getsock(td, args->s, &cap_connect_rights, &fp);
998 	if (error != 0)
999 		return (error);
1000 
1001 	error = EISCONN;
1002 	so = fp->f_data;
1003 	if (atomic_load_int(&fp->f_flag) & FNONBLOCK) {
1004 		SOCK_LOCK(so);
1005 		if (so->so_emuldata == 0)
1006 			error = so->so_error;
1007 		so->so_emuldata = (void *)1;
1008 		SOCK_UNLOCK(so);
1009 	}
1010 	fdrop(fp, td);
1011 
1012 	return (error);
1013 }
1014 
1015 int
1016 linux_listen(struct thread *td, struct linux_listen_args *args)
1017 {
1018 
1019 	return (kern_listen(td, args->s, args->backlog));
1020 }
1021 
1022 static int
1023 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
1024     l_uintptr_t namelen, int flags)
1025 {
1026 	struct sockaddr_storage ss = { .ss_len = sizeof(ss) };
1027 	struct file *fp, *fp1;
1028 	struct socket *so;
1029 	socklen_t len;
1030 	int bflags, error, error1;
1031 
1032 	bflags = 0;
1033 	fp = NULL;
1034 
1035 	error = linux_set_socket_flags(flags, &bflags);
1036 	if (error != 0)
1037 		return (error);
1038 
1039 	if (PTRIN(addr) != NULL) {
1040 		error = copyin(PTRIN(namelen), &len, sizeof(len));
1041 		if (error != 0)
1042 			return (error);
1043 		if (len < 0)
1044 			return (EINVAL);
1045 	} else
1046 		len = 0;
1047 
1048 	error = kern_accept4(td, s, (struct sockaddr *)&ss, bflags, &fp);
1049 
1050 	/*
1051 	 * Translate errno values into ones used by Linux.
1052 	 */
1053 	if (error != 0) {
1054 		/*
1055 		 * XXX. This is wrong, different sockaddr structures
1056 		 * have different sizes.
1057 		 */
1058 		switch (error) {
1059 		case EFAULT:
1060 			if (namelen != sizeof(struct sockaddr_in))
1061 				error = EINVAL;
1062 			break;
1063 		case EINVAL:
1064 			error1 = getsock(td, s, &cap_accept_rights, &fp1);
1065 			if (error1 != 0) {
1066 				error = error1;
1067 				break;
1068 			}
1069 			so = fp1->f_data;
1070 			if (so->so_type == SOCK_DGRAM)
1071 				error = EOPNOTSUPP;
1072 			fdrop(fp1, td);
1073 			break;
1074 		}
1075 		return (error);
1076 	}
1077 
1078 	if (PTRIN(addr) != NULL) {
1079 		len = min(ss.ss_len, len);
1080 		error = linux_copyout_sockaddr((struct sockaddr *)&ss,
1081 		    PTRIN(addr), len);
1082 		if (error == 0) {
1083 			len = ss.ss_len;
1084 			error = copyout(&len, PTRIN(namelen), sizeof(len));
1085 		}
1086 		if (error != 0) {
1087 			fdclose(td, fp, td->td_retval[0]);
1088 			td->td_retval[0] = 0;
1089 		}
1090 	}
1091 	if (fp != NULL)
1092 		fdrop(fp, td);
1093 	return (error);
1094 }
1095 
1096 int
1097 linux_accept(struct thread *td, struct linux_accept_args *args)
1098 {
1099 
1100 	return (linux_accept_common(td, args->s, args->addr,
1101 	    args->namelen, 0));
1102 }
1103 
1104 int
1105 linux_accept4(struct thread *td, struct linux_accept4_args *args)
1106 {
1107 
1108 	return (linux_accept_common(td, args->s, args->addr,
1109 	    args->namelen, args->flags));
1110 }
1111 
1112 int
1113 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
1114 {
1115 	struct sockaddr_storage ss = { .ss_len = sizeof(ss) };
1116 	socklen_t len;
1117 	int error;
1118 
1119 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1120 	if (error != 0)
1121 		return (error);
1122 
1123 	error = kern_getsockname(td, args->s, (struct sockaddr *)&ss);
1124 	if (error != 0)
1125 		return (error);
1126 
1127 	len = min(ss.ss_len, len);
1128 	error = linux_copyout_sockaddr((struct sockaddr *)&ss,
1129 	    PTRIN(args->addr), len);
1130 	if (error == 0) {
1131 		len = ss.ss_len;
1132 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1133 	}
1134 	return (error);
1135 }
1136 
1137 int
1138 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
1139 {
1140 	struct sockaddr_storage ss = { .ss_len = sizeof(ss) };
1141 	socklen_t len;
1142 	int error;
1143 
1144 	error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1145 	if (error != 0)
1146 		return (error);
1147 
1148 	error = kern_getpeername(td, args->s, (struct sockaddr *)&ss);
1149 	if (error != 0)
1150 		return (error);
1151 
1152 	len = min(ss.ss_len, len);
1153 	error = linux_copyout_sockaddr((struct sockaddr *)&ss,
1154 	    PTRIN(args->addr), len);
1155 	if (error == 0) {
1156 		len = ss.ss_len;
1157 		error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1158 	}
1159 	return (error);
1160 }
1161 
1162 int
1163 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
1164 {
1165 	int domain, error, sv[2], type;
1166 
1167 	domain = linux_to_bsd_domain(args->domain);
1168 	if (domain != PF_LOCAL)
1169 		return (EAFNOSUPPORT);
1170 	type = args->type & LINUX_SOCK_TYPE_MASK;
1171 	if (type < 0 || type > LINUX_SOCK_MAX)
1172 		return (EINVAL);
1173 	error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
1174 	    &type);
1175 	if (error != 0)
1176 		return (error);
1177 	if (args->protocol != 0 && args->protocol != PF_UNIX) {
1178 		/*
1179 		 * Use of PF_UNIX as protocol argument is not right,
1180 		 * but Linux does it.
1181 		 * Do not map PF_UNIX as its Linux value is identical
1182 		 * to FreeBSD one.
1183 		 */
1184 		return (EPROTONOSUPPORT);
1185 	}
1186 	error = kern_socketpair(td, domain, type, 0, sv);
1187 	if (error != 0)
1188                 return (error);
1189         error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int));
1190         if (error != 0) {
1191                 (void)kern_close(td, sv[0]);
1192                 (void)kern_close(td, sv[1]);
1193         }
1194 	return (error);
1195 }
1196 
1197 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1198 struct linux_send_args {
1199 	register_t s;
1200 	register_t msg;
1201 	register_t len;
1202 	register_t flags;
1203 };
1204 
1205 static int
1206 linux_send(struct thread *td, struct linux_send_args *args)
1207 {
1208 	struct sendto_args /* {
1209 		int s;
1210 		caddr_t buf;
1211 		int len;
1212 		int flags;
1213 		caddr_t to;
1214 		int tolen;
1215 	} */ bsd_args;
1216 	struct file *fp;
1217 	int error;
1218 
1219 	bsd_args.s = args->s;
1220 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1221 	bsd_args.len = args->len;
1222 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1223 	bsd_args.to = NULL;
1224 	bsd_args.tolen = 0;
1225 	error = sys_sendto(td, &bsd_args);
1226 	if (error == ENOTCONN) {
1227 		/*
1228 		 * Linux doesn't return ENOTCONN for non-blocking sockets.
1229 		 * Instead it returns the EAGAIN.
1230 		 */
1231 		error = getsock(td, args->s, &cap_send_rights, &fp);
1232 		if (error == 0) {
1233 			if (atomic_load_int(&fp->f_flag) & FNONBLOCK)
1234 				error = EAGAIN;
1235 			fdrop(fp, td);
1236 		}
1237 	}
1238 	return (error);
1239 }
1240 
1241 struct linux_recv_args {
1242 	register_t s;
1243 	register_t msg;
1244 	register_t len;
1245 	register_t flags;
1246 };
1247 
1248 static int
1249 linux_recv(struct thread *td, struct linux_recv_args *args)
1250 {
1251 	struct recvfrom_args /* {
1252 		int s;
1253 		caddr_t buf;
1254 		int len;
1255 		int flags;
1256 		struct sockaddr *from;
1257 		socklen_t fromlenaddr;
1258 	} */ bsd_args;
1259 
1260 	bsd_args.s = args->s;
1261 	bsd_args.buf = (caddr_t)PTRIN(args->msg);
1262 	bsd_args.len = args->len;
1263 	bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1264 	bsd_args.from = NULL;
1265 	bsd_args.fromlenaddr = 0;
1266 	return (sys_recvfrom(td, &bsd_args));
1267 }
1268 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1269 
1270 int
1271 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1272 {
1273 	struct msghdr msg;
1274 	struct iovec aiov;
1275 	struct socket *so;
1276 	struct file *fp;
1277 	int error;
1278 
1279 	if (linux_check_hdrincl(td, args->s) == 0)
1280 		/* IP_HDRINCL set, tweak the packet before sending */
1281 		return (linux_sendto_hdrincl(td, args));
1282 
1283 	bzero(&msg, sizeof(msg));
1284 	error = getsock(td, args->s, &cap_send_connect_rights, &fp);
1285 	if (error != 0)
1286 		return (error);
1287 	so = fp->f_data;
1288 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0) {
1289 		msg.msg_name = PTRIN(args->to);
1290 		msg.msg_namelen = args->tolen;
1291 	}
1292 	msg.msg_iov = &aiov;
1293 	msg.msg_iovlen = 1;
1294 	aiov.iov_base = PTRIN(args->msg);
1295 	aiov.iov_len = args->len;
1296 	fdrop(fp, td);
1297 	return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1298 	    UIO_USERSPACE));
1299 }
1300 
1301 int
1302 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1303 {
1304 	struct sockaddr *sa;
1305 	struct msghdr msg;
1306 	struct iovec aiov;
1307 	int error, fromlen;
1308 
1309 	if (PTRIN(args->fromlen) != NULL) {
1310 		error = copyin(PTRIN(args->fromlen), &fromlen,
1311 		    sizeof(fromlen));
1312 		if (error != 0)
1313 			return (error);
1314 		if (fromlen < 0)
1315 			return (EINVAL);
1316 		fromlen = min(fromlen, SOCK_MAXADDRLEN);
1317 		sa = malloc(fromlen, M_SONAME, M_WAITOK);
1318 	} else {
1319 		fromlen = 0;
1320 		sa = NULL;
1321 	}
1322 
1323 	msg.msg_name = sa;
1324 	msg.msg_namelen = fromlen;
1325 	msg.msg_iov = &aiov;
1326 	msg.msg_iovlen = 1;
1327 	aiov.iov_base = PTRIN(args->buf);
1328 	aiov.iov_len = args->len;
1329 	msg.msg_control = 0;
1330 	msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1331 
1332 	error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL);
1333 	if (error != 0)
1334 		goto out;
1335 
1336 	/*
1337 	 * XXX. Seems that FreeBSD is different from Linux here. Linux
1338 	 * fill source address if underlying protocol provides it, while
1339 	 * FreeBSD fill it if underlying protocol is not connection-oriented.
1340 	 * So, kern_recvit() set msg.msg_namelen to 0 if protocol pr_flags
1341 	 * does not contains PR_ADDR flag.
1342 	 */
1343 	if (PTRIN(args->from) != NULL && msg.msg_namelen != 0)
1344 		error = linux_copyout_sockaddr(sa, PTRIN(args->from),
1345 		    msg.msg_namelen);
1346 
1347 	if (error == 0 && PTRIN(args->fromlen) != NULL)
1348 		error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1349 		    sizeof(msg.msg_namelen));
1350 out:
1351 	free(sa, M_SONAME);
1352 	return (error);
1353 }
1354 
1355 static int
1356 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1357     l_uint flags)
1358 {
1359 	struct sockaddr_storage ss = { .ss_len = sizeof(ss) };
1360 	struct cmsghdr *cmsg;
1361 	struct mbuf *control;
1362 	struct msghdr msg;
1363 	struct l_cmsghdr linux_cmsg;
1364 	struct l_cmsghdr *ptr_cmsg;
1365 	struct l_msghdr linux_msghdr;
1366 	struct iovec *iov;
1367 	socklen_t datalen;
1368 	struct socket *so;
1369 	sa_family_t sa_family;
1370 	struct file *fp;
1371 	void *data;
1372 	l_size_t len;
1373 	l_size_t clen;
1374 	int error;
1375 
1376 	error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
1377 	if (error != 0)
1378 		return (error);
1379 
1380 	/*
1381 	 * Some Linux applications (ping) define a non-NULL control data
1382 	 * pointer, but a msg_controllen of 0, which is not allowed in the
1383 	 * FreeBSD system call interface.  NULL the msg_control pointer in
1384 	 * order to handle this case.  This should be checked, but allows the
1385 	 * Linux ping to work.
1386 	 */
1387 	if (PTRIN(linux_msghdr.msg_control) != NULL &&
1388 	    linux_msghdr.msg_controllen == 0)
1389 		linux_msghdr.msg_control = PTROUT(NULL);
1390 
1391 	error = linux_to_bsd_msghdr(&msg, &linux_msghdr);
1392 	if (error != 0)
1393 		return (error);
1394 
1395 #ifdef COMPAT_LINUX32
1396 	error = freebsd32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1397 	    &iov, EMSGSIZE);
1398 #else
1399 	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1400 #endif
1401 	if (error != 0)
1402 		return (error);
1403 
1404 	control = NULL;
1405 
1406 	error = kern_getsockname(td, s, (struct sockaddr *)&ss);
1407 	if (error != 0)
1408 		goto bad;
1409 	sa_family = ss.ss_family;
1410 
1411 	if (flags & LINUX_MSG_OOB) {
1412 		error = EOPNOTSUPP;
1413 		if (sa_family == AF_UNIX)
1414 			goto bad;
1415 
1416 		error = getsock(td, s, &cap_send_rights, &fp);
1417 		if (error != 0)
1418 			goto bad;
1419 		so = fp->f_data;
1420 		if (so->so_type != SOCK_STREAM)
1421 			error = EOPNOTSUPP;
1422 		fdrop(fp, td);
1423 		if (error != 0)
1424 			goto bad;
1425 	}
1426 
1427 	if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) {
1428 		error = ENOBUFS;
1429 		control = m_get(M_WAITOK, MT_CONTROL);
1430 		MCLGET(control, M_WAITOK);
1431 		data = mtod(control, void *);
1432 		datalen = 0;
1433 
1434 		ptr_cmsg = PTRIN(linux_msghdr.msg_control);
1435 		clen = linux_msghdr.msg_controllen;
1436 		do {
1437 			error = copyin(ptr_cmsg, &linux_cmsg,
1438 			    sizeof(struct l_cmsghdr));
1439 			if (error != 0)
1440 				goto bad;
1441 
1442 			error = EINVAL;
1443 			if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1444 			    linux_cmsg.cmsg_len > clen)
1445 				goto bad;
1446 
1447 			if (datalen + CMSG_HDRSZ > MCLBYTES)
1448 				goto bad;
1449 
1450 			/*
1451 			 * Now we support only SCM_RIGHTS and SCM_CRED,
1452 			 * so return EINVAL in any other cmsg_type
1453 			 */
1454 			cmsg = data;
1455 			cmsg->cmsg_type =
1456 			    linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1457 			cmsg->cmsg_level =
1458 			    linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1459 			if (cmsg->cmsg_type == -1
1460 			    || cmsg->cmsg_level != SOL_SOCKET) {
1461 				linux_msg(curthread,
1462 				    "unsupported sendmsg cmsg level %d type %d",
1463 				    linux_cmsg.cmsg_level, linux_cmsg.cmsg_type);
1464 				goto bad;
1465 			}
1466 
1467 			/*
1468 			 * Some applications (e.g. pulseaudio) attempt to
1469 			 * send ancillary data even if the underlying protocol
1470 			 * doesn't support it which is not allowed in the
1471 			 * FreeBSD system call interface.
1472 			 */
1473 			if (sa_family != AF_UNIX)
1474 				goto next;
1475 
1476 			if (cmsg->cmsg_type == SCM_CREDS) {
1477 				len = sizeof(struct cmsgcred);
1478 				if (datalen + CMSG_SPACE(len) > MCLBYTES)
1479 					goto bad;
1480 
1481 				/*
1482 				 * The lower levels will fill in the structure
1483 				 */
1484 				memset(CMSG_DATA(data), 0, len);
1485 			} else {
1486 				len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1487 				if (datalen + CMSG_SPACE(len) < datalen ||
1488 				    datalen + CMSG_SPACE(len) > MCLBYTES)
1489 					goto bad;
1490 
1491 				error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1492 				    CMSG_DATA(data), len);
1493 				if (error != 0)
1494 					goto bad;
1495 			}
1496 
1497 			cmsg->cmsg_len = CMSG_LEN(len);
1498 			data = (char *)data + CMSG_SPACE(len);
1499 			datalen += CMSG_SPACE(len);
1500 
1501 next:
1502 			if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1503 				break;
1504 
1505 			clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1506 			ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1507 			    LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1508 		} while(clen >= sizeof(struct l_cmsghdr));
1509 
1510 		control->m_len = datalen;
1511 		if (datalen == 0) {
1512 			m_freem(control);
1513 			control = NULL;
1514 		}
1515 	}
1516 
1517 	msg.msg_iov = iov;
1518 	msg.msg_flags = 0;
1519 	error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1520 	control = NULL;
1521 
1522 bad:
1523 	m_freem(control);
1524 	free(iov, M_IOV);
1525 	return (error);
1526 }
1527 
1528 int
1529 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1530 {
1531 
1532 	return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1533 	    args->flags));
1534 }
1535 
1536 int
1537 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1538 {
1539 	struct l_mmsghdr *msg;
1540 	l_uint retval;
1541 	int error, datagrams;
1542 
1543 	if (args->vlen > UIO_MAXIOV)
1544 		args->vlen = UIO_MAXIOV;
1545 
1546 	msg = PTRIN(args->msg);
1547 	datagrams = 0;
1548 	while (datagrams < args->vlen) {
1549 		error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1550 		    args->flags);
1551 		if (error != 0)
1552 			break;
1553 
1554 		retval = td->td_retval[0];
1555 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1556 		if (error != 0)
1557 			break;
1558 		++msg;
1559 		++datagrams;
1560 	}
1561 	if (error == 0)
1562 		td->td_retval[0] = datagrams;
1563 	return (error);
1564 }
1565 
1566 static int
1567 recvmsg_scm_rights(struct thread *td, l_uint flags, socklen_t *datalen,
1568     void **data, void **udata)
1569 {
1570 	int i, fd, fds, *fdp;
1571 
1572 	if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1573 		fds = *datalen / sizeof(int);
1574 		fdp = *data;
1575 		for (i = 0; i < fds; i++) {
1576 			fd = *fdp++;
1577 			(void)kern_fcntl(td, fd, F_SETFD, FD_CLOEXEC);
1578 		}
1579 	}
1580 	return (0);
1581 }
1582 
1583 
1584 static int
1585 recvmsg_scm_creds(socklen_t *datalen, void **data, void **udata)
1586 {
1587 	struct cmsgcred *cmcred;
1588 	struct l_ucred lu;
1589 
1590 	cmcred = *data;
1591 	lu.pid = cmcred->cmcred_pid;
1592 	lu.uid = cmcred->cmcred_uid;
1593 	lu.gid = cmcred->cmcred_gid;
1594 	memmove(*data, &lu, sizeof(lu));
1595 	*datalen = sizeof(lu);
1596 	return (0);
1597 }
1598 _Static_assert(sizeof(struct cmsgcred) >= sizeof(struct l_ucred),
1599     "scm_creds sizeof l_ucred");
1600 
1601 static int
1602 recvmsg_scm_creds2(socklen_t *datalen, void **data, void **udata)
1603 {
1604 	struct sockcred2 *scred;
1605 	struct l_ucred lu;
1606 
1607 	scred = *data;
1608 	lu.pid = scred->sc_pid;
1609 	lu.uid = scred->sc_uid;
1610 	lu.gid = scred->sc_gid;
1611 	memmove(*data, &lu, sizeof(lu));
1612 	*datalen = sizeof(lu);
1613 	return (0);
1614 }
1615 _Static_assert(sizeof(struct sockcred2) >= sizeof(struct l_ucred),
1616     "scm_creds2 sizeof l_ucred");
1617 
1618 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1619 static int
1620 recvmsg_scm_timestamp(l_int msg_type, socklen_t *datalen, void **data,
1621     void **udata)
1622 {
1623 	l_sock_timeval ltv64;
1624 	l_timeval ltv;
1625 	struct timeval *tv;
1626 	socklen_t len;
1627 	void *buf;
1628 
1629 	if (*datalen != sizeof(struct timeval))
1630 		return (EMSGSIZE);
1631 
1632 	tv = *data;
1633 #if defined(COMPAT_LINUX32)
1634 	if (msg_type == LINUX_SCM_TIMESTAMPO &&
1635 	    (tv->tv_sec > INT_MAX || tv->tv_sec < INT_MIN))
1636 		return (EOVERFLOW);
1637 #endif
1638 	if (msg_type == LINUX_SCM_TIMESTAMPN)
1639 		len = sizeof(ltv64);
1640 	else
1641 		len = sizeof(ltv);
1642 
1643 	buf = malloc(len, M_LINUX, M_WAITOK);
1644 	if (msg_type == LINUX_SCM_TIMESTAMPN) {
1645 		ltv64.tv_sec = tv->tv_sec;
1646 		ltv64.tv_usec = tv->tv_usec;
1647 		memmove(buf, &ltv64, len);
1648 	} else {
1649 		ltv.tv_sec = tv->tv_sec;
1650 		ltv.tv_usec = tv->tv_usec;
1651 		memmove(buf, &ltv, len);
1652 	}
1653 	*data = *udata = buf;
1654 	*datalen = len;
1655 	return (0);
1656 }
1657 #else
1658 _Static_assert(sizeof(struct timeval) == sizeof(l_timeval),
1659     "scm_timestamp sizeof l_timeval");
1660 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1661 
1662 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1663 static int
1664 recvmsg_scm_timestampns(l_int msg_type, socklen_t *datalen, void **data,
1665     void **udata)
1666 {
1667 	struct l_timespec64 ts64;
1668 	struct l_timespec ts32;
1669 	struct timespec ts;
1670 	socklen_t len;
1671 	void *buf;
1672 
1673 	if (msg_type == LINUX_SCM_TIMESTAMPNSO)
1674 		len = sizeof(ts32);
1675 	else
1676 		len = sizeof(ts64);
1677 
1678 	buf = malloc(len, M_LINUX, M_WAITOK);
1679 	bintime2timespec(*data, &ts);
1680 	if (msg_type == LINUX_SCM_TIMESTAMPNSO) {
1681 		ts32.tv_sec = ts.tv_sec;
1682 		ts32.tv_nsec = ts.tv_nsec;
1683 		memmove(buf, &ts32, len);
1684 	} else {
1685 		ts64.tv_sec = ts.tv_sec;
1686 		ts64.tv_nsec = ts.tv_nsec;
1687 		memmove(buf, &ts64, len);
1688 	}
1689 	*data = *udata = buf;
1690 	*datalen = len;
1691 	return (0);
1692 }
1693 #else
1694 static int
1695 recvmsg_scm_timestampns(l_int msg_type, socklen_t *datalen, void **data,
1696     void **udata)
1697 {
1698 	struct timespec ts;
1699 
1700 	bintime2timespec(*data, &ts);
1701 	memmove(*data, &ts, sizeof(struct timespec));
1702 	*datalen = sizeof(struct timespec);
1703 	return (0);
1704 }
1705 _Static_assert(sizeof(struct bintime) >= sizeof(struct timespec),
1706     "scm_timestampns sizeof timespec");
1707 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1708 
1709 static int
1710 recvmsg_scm_sol_socket(struct thread *td, l_int msg_type, l_int lmsg_type,
1711     l_uint flags, socklen_t *datalen, void **data, void **udata)
1712 {
1713 	int error;
1714 
1715 	error = 0;
1716 	switch (msg_type) {
1717 	case SCM_RIGHTS:
1718 		error = recvmsg_scm_rights(td, flags, datalen,
1719 		    data, udata);
1720 		break;
1721 	case SCM_CREDS:
1722 		error = recvmsg_scm_creds(datalen, data, udata);
1723 		break;
1724 	case SCM_CREDS2:
1725 		error = recvmsg_scm_creds2(datalen, data, udata);
1726 		break;
1727 	case SCM_TIMESTAMP:
1728 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1729 		error = recvmsg_scm_timestamp(lmsg_type, datalen,
1730 		    data, udata);
1731 #endif
1732 		break;
1733 	case SCM_BINTIME:
1734 		error = recvmsg_scm_timestampns(lmsg_type, datalen,
1735 		    data, udata);
1736 		break;
1737 	}
1738 
1739 	return (error);
1740 }
1741 
1742 static int
1743 recvmsg_scm_ip_origdstaddr(socklen_t *datalen, void **data, void **udata)
1744 {
1745 	struct l_sockaddr *lsa;
1746 	int error;
1747 
1748 	error = bsd_to_linux_sockaddr(*data, &lsa, *datalen);
1749 	if (error == 0) {
1750 		*data = *udata = lsa;
1751 		*datalen = sizeof(*lsa);
1752 	}
1753 	return (error);
1754 }
1755 
1756 static int
1757 recvmsg_scm_ipproto_ip(l_int msg_type, l_int lmsg_type, socklen_t *datalen,
1758     void **data, void **udata)
1759 {
1760 	int error;
1761 
1762 	error = 0;
1763 	switch (msg_type) {
1764 	case IP_ORIGDSTADDR:
1765 		error = recvmsg_scm_ip_origdstaddr(datalen, data,
1766 		    udata);
1767 		break;
1768 	}
1769 
1770 	return (error);
1771 }
1772 
1773 static int
1774 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1775     l_uint flags, struct msghdr *msg)
1776 {
1777 	struct proc *p = td->td_proc;
1778 	struct cmsghdr *cm;
1779 	struct l_cmsghdr *lcm = NULL;
1780 	socklen_t datalen, maxlen, outlen;
1781 	struct l_msghdr l_msghdr;
1782 	struct iovec *iov, *uiov;
1783 	struct mbuf *m, *control = NULL;
1784 	struct mbuf **controlp;
1785 	struct sockaddr *sa;
1786 	caddr_t outbuf;
1787 	void *data, *udata;
1788 	int error, skiped;
1789 
1790 	error = copyin(msghdr, &l_msghdr, sizeof(l_msghdr));
1791 	if (error != 0)
1792 		return (error);
1793 
1794 	/*
1795 	 * Pass user-supplied recvmsg() flags in msg_flags field,
1796 	 * following sys_recvmsg() convention.
1797 	*/
1798 	l_msghdr.msg_flags = flags;
1799 
1800 	error = linux_to_bsd_msghdr(msg, &l_msghdr);
1801 	if (error != 0)
1802 		return (error);
1803 
1804 #ifdef COMPAT_LINUX32
1805 	error = freebsd32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1806 	    &iov, EMSGSIZE);
1807 #else
1808 	error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1809 #endif
1810 	if (error != 0)
1811 		return (error);
1812 
1813 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1814 		msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN);
1815 		sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK);
1816 		msg->msg_name = sa;
1817 	} else {
1818 		sa = NULL;
1819 		msg->msg_name = NULL;
1820 	}
1821 
1822 	uiov = msg->msg_iov;
1823 	msg->msg_iov = iov;
1824 	controlp = (msg->msg_control != NULL) ? &control : NULL;
1825 	error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp);
1826 	msg->msg_iov = uiov;
1827 	if (error != 0)
1828 		goto bad;
1829 
1830 	/*
1831 	 * Note that kern_recvit() updates msg->msg_namelen.
1832 	 */
1833 	if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1834 		msg->msg_name = PTRIN(l_msghdr.msg_name);
1835 		error = linux_copyout_sockaddr(sa, msg->msg_name,
1836 		    msg->msg_namelen);
1837 		if (error != 0)
1838 			goto bad;
1839 	}
1840 
1841 	error = bsd_to_linux_msghdr(msg, &l_msghdr);
1842 	if (error != 0)
1843 		goto bad;
1844 
1845 	skiped = outlen = 0;
1846 	maxlen = l_msghdr.msg_controllen;
1847 	if (control == NULL)
1848 		goto out;
1849 
1850 	lcm = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1851 	msg->msg_control = mtod(control, struct cmsghdr *);
1852 	msg->msg_controllen = control->m_len;
1853 	outbuf = PTRIN(l_msghdr.msg_control);
1854 	for (m = control; m != NULL; m = m->m_next) {
1855 		cm = mtod(m, struct cmsghdr *);
1856 		lcm->cmsg_type = bsd_to_linux_cmsg_type(p, cm->cmsg_type,
1857 		    cm->cmsg_level);
1858 		lcm->cmsg_level = bsd_to_linux_sockopt_level(cm->cmsg_level);
1859 
1860 		if (lcm->cmsg_type == -1 ||
1861 		    cm->cmsg_level == -1) {
1862 			LINUX_RATELIMIT_MSG_OPT2(
1863 			    "unsupported recvmsg cmsg level %d type %d",
1864 			    cm->cmsg_level, cm->cmsg_type);
1865 			/* Skip unsupported messages */
1866 			skiped++;
1867 			continue;
1868 		}
1869 		data = CMSG_DATA(cm);
1870 		datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1871 		udata = NULL;
1872 		error = 0;
1873 
1874 		switch (cm->cmsg_level) {
1875 		case IPPROTO_IP:
1876 			error = recvmsg_scm_ipproto_ip(cm->cmsg_type,
1877 			    lcm->cmsg_type, &datalen, &data, &udata);
1878  			break;
1879 		case SOL_SOCKET:
1880 			error = recvmsg_scm_sol_socket(td, cm->cmsg_type,
1881 			    lcm->cmsg_type, flags, &datalen, &data, &udata);
1882  			break;
1883  		}
1884 
1885 		/* The recvmsg_scm_ is responsible to free udata on error. */
1886 		if (error != 0)
1887 			goto bad;
1888 
1889 		if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1890 			if (outlen == 0) {
1891 				error = EMSGSIZE;
1892 				goto err;
1893 			} else {
1894 				l_msghdr.msg_flags |= LINUX_MSG_CTRUNC;
1895 				m_dispose_extcontrolm(control);
1896 				free(udata, M_LINUX);
1897 				goto out;
1898 			}
1899 		}
1900 
1901 		lcm->cmsg_len = LINUX_CMSG_LEN(datalen);
1902 		error = copyout(lcm, outbuf, L_CMSG_HDRSZ);
1903 		if (error == 0) {
1904 			error = copyout(data, LINUX_CMSG_DATA(outbuf), datalen);
1905 			if (error == 0) {
1906 				outbuf += LINUX_CMSG_SPACE(datalen);
1907 				outlen += LINUX_CMSG_SPACE(datalen);
1908 			}
1909 		}
1910 err:
1911 		free(udata, M_LINUX);
1912 		if (error != 0)
1913 			goto bad;
1914 	}
1915 	if (outlen == 0 && skiped > 0) {
1916 		error = EINVAL;
1917 		goto bad;
1918 	}
1919 
1920 out:
1921 	l_msghdr.msg_controllen = outlen;
1922 	error = copyout(&l_msghdr, msghdr, sizeof(l_msghdr));
1923 
1924 bad:
1925 	if (control != NULL) {
1926 		if (error != 0)
1927 			m_dispose_extcontrolm(control);
1928 		m_freem(control);
1929 	}
1930 	free(iov, M_IOV);
1931 	free(lcm, M_LINUX);
1932 	free(sa, M_SONAME);
1933 
1934 	return (error);
1935 }
1936 
1937 int
1938 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1939 {
1940 	struct msghdr bsd_msg;
1941 	struct file *fp;
1942 	int error;
1943 
1944 	error = getsock(td, args->s, &cap_recv_rights, &fp);
1945 	if (error != 0)
1946 		return (error);
1947 	fdrop(fp, td);
1948 	return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1949 	    args->flags, &bsd_msg));
1950 }
1951 
1952 static int
1953 linux_recvmmsg_common(struct thread *td, l_int s, struct l_mmsghdr *msg,
1954     l_uint vlen, l_uint flags, struct timespec *tts)
1955 {
1956 	struct msghdr bsd_msg;
1957 	struct timespec ts;
1958 	struct file *fp;
1959 	l_uint retval;
1960 	int error, datagrams;
1961 
1962 	error = getsock(td, s, &cap_recv_rights, &fp);
1963 	if (error != 0)
1964 		return (error);
1965 	datagrams = 0;
1966 	while (datagrams < vlen) {
1967 		error = linux_recvmsg_common(td, s, &msg->msg_hdr,
1968 		    flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1969 		if (error != 0)
1970 			break;
1971 
1972 		retval = td->td_retval[0];
1973 		error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1974 		if (error != 0)
1975 			break;
1976 		++msg;
1977 		++datagrams;
1978 
1979 		/*
1980 		 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1981 		 */
1982 		if (flags & LINUX_MSG_WAITFORONE)
1983 			flags |= LINUX_MSG_DONTWAIT;
1984 
1985 		/*
1986 		 * See BUGS section of recvmmsg(2).
1987 		 */
1988 		if (tts) {
1989 			getnanotime(&ts);
1990 			timespecsub(&ts, tts, &ts);
1991 			if (!timespecisset(&ts) || ts.tv_sec > 0)
1992 				break;
1993 		}
1994 		/* Out of band data, return right away. */
1995 		if (bsd_msg.msg_flags & MSG_OOB)
1996 			break;
1997 	}
1998 	if (error == 0)
1999 		td->td_retval[0] = datagrams;
2000 	fdrop(fp, td);
2001 	return (error);
2002 }
2003 
2004 int
2005 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
2006 {
2007 	struct timespec ts, tts, *ptts;
2008 	int error;
2009 
2010 	if (args->timeout) {
2011 		error = linux_get_timespec(&ts, args->timeout);
2012 		if (error != 0)
2013 			return (error);
2014 		getnanotime(&tts);
2015 		timespecadd(&tts, &ts, &tts);
2016 		ptts = &tts;
2017 	}
2018 		else ptts = NULL;
2019 
2020 	return (linux_recvmmsg_common(td, args->s, PTRIN(args->msg),
2021 	    args->vlen, args->flags, ptts));
2022 }
2023 
2024 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2025 int
2026 linux_recvmmsg_time64(struct thread *td, struct linux_recvmmsg_time64_args *args)
2027 {
2028 	struct timespec ts, tts, *ptts;
2029 	int error;
2030 
2031 	if (args->timeout) {
2032 		error = linux_get_timespec64(&ts, args->timeout);
2033 		if (error != 0)
2034 			return (error);
2035 		getnanotime(&tts);
2036 		timespecadd(&tts, &ts, &tts);
2037 		ptts = &tts;
2038 	}
2039 		else ptts = NULL;
2040 
2041 	return (linux_recvmmsg_common(td, args->s, PTRIN(args->msg),
2042 	    args->vlen, args->flags, ptts));
2043 }
2044 #endif
2045 
2046 int
2047 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
2048 {
2049 
2050 	return (kern_shutdown(td, args->s, args->how));
2051 }
2052 
2053 int
2054 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
2055 {
2056 	struct proc *p = td->td_proc;
2057 	struct linux_pemuldata *pem;
2058 	l_timeval linux_tv;
2059 	struct sockaddr *sa;
2060 	struct timeval tv;
2061 	socklen_t len;
2062 	int error, level, name, val;
2063 
2064 	level = linux_to_bsd_sockopt_level(args->level);
2065 	switch (level) {
2066 	case SOL_SOCKET:
2067 		name = linux_to_bsd_so_sockopt(args->optname);
2068 		switch (name) {
2069 		case LOCAL_CREDS_PERSISTENT:
2070 			level = SOL_LOCAL;
2071 			break;
2072 		case SO_RCVTIMEO:
2073 			/* FALLTHROUGH */
2074 		case SO_SNDTIMEO:
2075 			error = copyin(PTRIN(args->optval), &linux_tv,
2076 			    sizeof(linux_tv));
2077 			if (error != 0)
2078 				return (error);
2079 			tv.tv_sec = linux_tv.tv_sec;
2080 			tv.tv_usec = linux_tv.tv_usec;
2081 			return (kern_setsockopt(td, args->s, level,
2082 			    name, &tv, UIO_SYSSPACE, sizeof(tv)));
2083 			/* NOTREACHED */
2084 		case SO_TIMESTAMP:
2085 			/* overwrite SO_BINTIME */
2086 			val = 0;
2087 			error = kern_setsockopt(td, args->s, level,
2088 			    SO_BINTIME, &val, UIO_SYSSPACE, sizeof(val));
2089 			if (error != 0)
2090 				return (error);
2091 			pem = pem_find(p);
2092 			pem->so_timestamp = args->optname;
2093 			break;
2094 		case SO_BINTIME:
2095 			/* overwrite SO_TIMESTAMP */
2096 			val = 0;
2097 			error = kern_setsockopt(td, args->s, level,
2098 			    SO_TIMESTAMP, &val, UIO_SYSSPACE, sizeof(val));
2099 			if (error != 0)
2100 				return (error);
2101 			pem = pem_find(p);
2102 			pem->so_timestampns = args->optname;
2103 			break;
2104 		default:
2105 			break;
2106 		}
2107 		break;
2108 	case IPPROTO_IP:
2109 		if (args->optname == LINUX_IP_RECVERR &&
2110 		    linux_ignore_ip_recverr) {
2111 			/*
2112 			 * XXX: This is a hack to unbreak DNS resolution
2113 			 *	with glibc 2.30 and above.
2114 			 */
2115 			return (0);
2116 		}
2117 		name = linux_to_bsd_ip_sockopt(args->optname);
2118 		break;
2119 	case IPPROTO_IPV6:
2120 		if (args->optname == LINUX_IPV6_RECVERR &&
2121 		    linux_ignore_ip_recverr) {
2122 			/*
2123 			 * XXX: This is a hack to unbreak DNS resolution
2124 			 *	with glibc 2.30 and above.
2125 			 */
2126 			return (0);
2127 		}
2128 		name = linux_to_bsd_ip6_sockopt(args->optname);
2129 		break;
2130 	case IPPROTO_TCP:
2131 		name = linux_to_bsd_tcp_sockopt(args->optname);
2132 		break;
2133 	case SOL_NETLINK:
2134 		name = args->optname;
2135 		break;
2136 	default:
2137 		name = -1;
2138 		break;
2139 	}
2140 	if (name < 0) {
2141 		if (name == -1)
2142 			linux_msg(curthread,
2143 			    "unsupported setsockopt level %d optname %d",
2144 			    args->level, args->optname);
2145 		return (ENOPROTOOPT);
2146 	}
2147 
2148 	if (name == IPV6_NEXTHOP) {
2149 		len = args->optlen;
2150 		error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len);
2151 		if (error != 0)
2152 			return (error);
2153 
2154 		error = kern_setsockopt(td, args->s, level,
2155 		    name, sa, UIO_SYSSPACE, len);
2156 		free(sa, M_SONAME);
2157 	} else {
2158 		error = kern_setsockopt(td, args->s, level,
2159 		    name, PTRIN(args->optval), UIO_USERSPACE, args->optlen);
2160 	}
2161 
2162 	return (error);
2163 }
2164 
2165 static int
2166 linux_sockopt_copyout(struct thread *td, void *val, socklen_t len,
2167     struct linux_getsockopt_args *args)
2168 {
2169 	int error;
2170 
2171 	error = copyout(val, PTRIN(args->optval), len);
2172 	if (error == 0)
2173 		error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2174 	return (error);
2175 }
2176 
2177 static int
2178 linux_getsockopt_so_peergroups(struct thread *td,
2179     struct linux_getsockopt_args *args)
2180 {
2181 	struct xucred xu;
2182 	socklen_t xulen, len;
2183 	int error, i;
2184 
2185 	xulen = sizeof(xu);
2186 	error = kern_getsockopt(td, args->s, 0,
2187 	    LOCAL_PEERCRED, &xu, UIO_SYSSPACE, &xulen);
2188 	if (error != 0)
2189 		return (error);
2190 
2191 	len = xu.cr_ngroups * sizeof(l_gid_t);
2192 	if (args->optlen < len) {
2193 		error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2194 		if (error == 0)
2195 			error = ERANGE;
2196 		return (error);
2197 	}
2198 
2199 	/*
2200 	 * "- 1" to skip the primary group.
2201 	 */
2202 	for (i = 0; i < xu.cr_ngroups - 1; i++) {
2203 		error = copyout(xu.cr_groups + i + 1,
2204 		    (void *)(args->optval + i * sizeof(l_gid_t)),
2205 		    sizeof(l_gid_t));
2206 		if (error != 0)
2207 			return (error);
2208 	}
2209 
2210 	error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2211 	return (error);
2212 }
2213 
2214 static int
2215 linux_getsockopt_so_peersec(struct thread *td,
2216     struct linux_getsockopt_args *args)
2217 {
2218 	socklen_t len;
2219 	int error;
2220 
2221 	len = sizeof(SECURITY_CONTEXT_STRING);
2222 	if (args->optlen < len) {
2223 		error = copyout(&len, PTRIN(args->optlen), sizeof(len));
2224 		if (error == 0)
2225 			error = ERANGE;
2226 		return (error);
2227 	}
2228 
2229 	return (linux_sockopt_copyout(td, SECURITY_CONTEXT_STRING,
2230 	    len, args));
2231 }
2232 
2233 static int
2234 linux_getsockopt_so_linger(struct thread *td,
2235     struct linux_getsockopt_args *args)
2236 {
2237 	struct linger ling;
2238 	socklen_t len;
2239 	int error;
2240 
2241 	len = sizeof(ling);
2242 	error = kern_getsockopt(td, args->s, SOL_SOCKET,
2243 	    SO_LINGER, &ling, UIO_SYSSPACE, &len);
2244 	if (error != 0)
2245 		return (error);
2246 	ling.l_onoff = ((ling.l_onoff & SO_LINGER) != 0);
2247 	return (linux_sockopt_copyout(td, &ling, len, args));
2248 }
2249 
2250 int
2251 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
2252 {
2253 	l_timeval linux_tv;
2254 	struct timeval tv;
2255 	socklen_t tv_len, xulen, len;
2256 	struct sockaddr *sa;
2257 	struct xucred xu;
2258 	struct l_ucred lxu;
2259 	int error, level, name, newval;
2260 
2261 	level = linux_to_bsd_sockopt_level(args->level);
2262 	switch (level) {
2263 	case SOL_SOCKET:
2264 		switch (args->optname) {
2265 		case LINUX_SO_PEERGROUPS:
2266 			return (linux_getsockopt_so_peergroups(td, args));
2267 		case LINUX_SO_PEERSEC:
2268 			return (linux_getsockopt_so_peersec(td, args));
2269 		default:
2270 			break;
2271 		}
2272 
2273 		name = linux_to_bsd_so_sockopt(args->optname);
2274 		switch (name) {
2275 		case LOCAL_CREDS_PERSISTENT:
2276 			level = SOL_LOCAL;
2277 			break;
2278 		case SO_RCVTIMEO:
2279 			/* FALLTHROUGH */
2280 		case SO_SNDTIMEO:
2281 			tv_len = sizeof(tv);
2282 			error = kern_getsockopt(td, args->s, level,
2283 			    name, &tv, UIO_SYSSPACE, &tv_len);
2284 			if (error != 0)
2285 				return (error);
2286 			linux_tv.tv_sec = tv.tv_sec;
2287 			linux_tv.tv_usec = tv.tv_usec;
2288 			return (linux_sockopt_copyout(td, &linux_tv,
2289 			    sizeof(linux_tv), args));
2290 			/* NOTREACHED */
2291 		case LOCAL_PEERCRED:
2292 			if (args->optlen < sizeof(lxu))
2293 				return (EINVAL);
2294 			/*
2295 			 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
2296 			 * but by the Unix socket's level 0.
2297 			 */
2298 			level = 0;
2299 			xulen = sizeof(xu);
2300 			error = kern_getsockopt(td, args->s, level,
2301 			    name, &xu, UIO_SYSSPACE, &xulen);
2302 			if (error != 0)
2303 				return (error);
2304 			lxu.pid = xu.cr_pid;
2305 			lxu.uid = xu.cr_uid;
2306 			lxu.gid = xu.cr_gid;
2307 			return (linux_sockopt_copyout(td, &lxu,
2308 			    sizeof(lxu), args));
2309 			/* NOTREACHED */
2310 		case SO_ERROR:
2311 			len = sizeof(newval);
2312 			error = kern_getsockopt(td, args->s, level,
2313 			    name, &newval, UIO_SYSSPACE, &len);
2314 			if (error != 0)
2315 				return (error);
2316 			newval = -bsd_to_linux_errno(newval);
2317 			return (linux_sockopt_copyout(td, &newval,
2318 			    len, args));
2319 			/* NOTREACHED */
2320 		case SO_DOMAIN:
2321 			len = sizeof(newval);
2322 			error = kern_getsockopt(td, args->s, level,
2323 			    name, &newval, UIO_SYSSPACE, &len);
2324 			if (error != 0)
2325 				return (error);
2326 			newval = bsd_to_linux_domain((sa_family_t)newval);
2327 			if (newval == AF_UNKNOWN)
2328 				return (ENOPROTOOPT);
2329 			return (linux_sockopt_copyout(td, &newval,
2330 			    len, args));
2331 			/* NOTREACHED */
2332 		case SO_LINGER:
2333 			return (linux_getsockopt_so_linger(td, args));
2334 			/* NOTREACHED */
2335 		default:
2336 			break;
2337 		}
2338 		break;
2339 	case IPPROTO_IP:
2340 		name = linux_to_bsd_ip_sockopt(args->optname);
2341 		break;
2342 	case IPPROTO_IPV6:
2343 		name = linux_to_bsd_ip6_sockopt(args->optname);
2344 		break;
2345 	case IPPROTO_TCP:
2346 		name = linux_to_bsd_tcp_sockopt(args->optname);
2347 		break;
2348 	default:
2349 		name = -1;
2350 		break;
2351 	}
2352 	if (name < 0) {
2353 		if (name == -1)
2354 			linux_msg(curthread,
2355 			    "unsupported getsockopt level %d optname %d",
2356 			    args->level, args->optname);
2357 		return (EINVAL);
2358 	}
2359 
2360 	if (name == IPV6_NEXTHOP) {
2361 		error = copyin(PTRIN(args->optlen), &len, sizeof(len));
2362                 if (error != 0)
2363                         return (error);
2364 		sa = malloc(len, M_SONAME, M_WAITOK);
2365 
2366 		error = kern_getsockopt(td, args->s, level,
2367 		    name, sa, UIO_SYSSPACE, &len);
2368 		if (error != 0)
2369 			goto out;
2370 
2371 		error = linux_copyout_sockaddr(sa, PTRIN(args->optval), len);
2372 		if (error == 0)
2373 			error = copyout(&len, PTRIN(args->optlen),
2374 			    sizeof(len));
2375 out:
2376 		free(sa, M_SONAME);
2377 	} else {
2378 		if (args->optval) {
2379 			error = copyin(PTRIN(args->optlen), &len, sizeof(len));
2380 			if (error != 0)
2381 				return (error);
2382 		}
2383 		error = kern_getsockopt(td, args->s, level,
2384 		    name, PTRIN(args->optval), UIO_USERSPACE, &len);
2385 		if (error == 0)
2386 			error = copyout(&len, PTRIN(args->optlen),
2387 			    sizeof(len));
2388 	}
2389 
2390 	return (error);
2391 }
2392 
2393 /*
2394  * Based on sendfile_getsock from kern_sendfile.c
2395  * Determines whether an fd is a stream socket that can be used
2396  * with FreeBSD sendfile.
2397  */
2398 static bool
2399 is_sendfile(struct file *fp, struct file *ofp)
2400 {
2401 	struct socket *so;
2402 
2403 	/*
2404 	 * FreeBSD sendfile() system call sends a regular file or
2405 	 * shared memory object out a stream socket.
2406 	 */
2407 	if ((fp->f_type != DTYPE_SHM && fp->f_type != DTYPE_VNODE) ||
2408 	    (fp->f_type == DTYPE_VNODE &&
2409 	    (fp->f_vnode == NULL || fp->f_vnode->v_type != VREG)))
2410 		return (false);
2411 	/*
2412 	 * The socket must be a stream socket and connected.
2413 	 */
2414 	if (ofp->f_type != DTYPE_SOCKET)
2415 		return (false);
2416 	so = ofp->f_data;
2417 	if (so->so_type != SOCK_STREAM)
2418 		return (false);
2419 	/*
2420 	 * SCTP one-to-one style sockets currently don't work with
2421 	 * sendfile().
2422 	 */
2423 	if (so->so_proto->pr_protocol == IPPROTO_SCTP)
2424 		return (false);
2425 	return (!SOLISTENING(so));
2426 }
2427 
2428 static bool
2429 is_regular_file(struct file *fp)
2430 {
2431 
2432 	return (fp->f_type == DTYPE_VNODE && fp->f_vnode != NULL &&
2433 	    fp->f_vnode->v_type == VREG);
2434 }
2435 
2436 static int
2437 sendfile_fallback(struct thread *td, struct file *fp, l_int out,
2438     off_t *offset, l_size_t count, off_t *sbytes)
2439 {
2440 	off_t current_offset, out_offset, to_send;
2441 	l_size_t bytes_sent, n_read;
2442 	struct file *ofp;
2443 	struct iovec aiov;
2444 	struct uio auio;
2445 	bool seekable;
2446 	size_t bufsz;
2447 	void *buf;
2448 	int flags, error;
2449 
2450 	if (offset == NULL) {
2451 		if ((error = fo_seek(fp, 0, SEEK_CUR, td)) != 0)
2452 			return (error);
2453 		current_offset = td->td_uretoff.tdu_off;
2454 	} else {
2455 		if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0)
2456 			return (ESPIPE);
2457 		current_offset = *offset;
2458 	}
2459 	error = fget_write(td, out, &cap_pwrite_rights, &ofp);
2460 	if (error != 0)
2461 		return (error);
2462 	seekable = (ofp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0;
2463 	if (seekable) {
2464 		if ((error = fo_seek(ofp, 0, SEEK_CUR, td)) != 0)
2465 			goto drop;
2466 		out_offset = td->td_uretoff.tdu_off;
2467 	} else
2468 		out_offset = 0;
2469 
2470 	flags = FOF_OFFSET | FOF_NOUPDATE;
2471 	bufsz = min(count, maxphys);
2472 	buf = malloc(bufsz, M_LINUX, M_WAITOK);
2473 	bytes_sent = 0;
2474 	while (bytes_sent < count) {
2475 		to_send = min(count - bytes_sent, bufsz);
2476 		aiov.iov_base = buf;
2477 		aiov.iov_len = bufsz;
2478 		auio.uio_iov = &aiov;
2479 		auio.uio_iovcnt = 1;
2480 		auio.uio_segflg = UIO_SYSSPACE;
2481 		auio.uio_td = td;
2482 		auio.uio_rw = UIO_READ;
2483 		auio.uio_offset = current_offset;
2484 		auio.uio_resid = to_send;
2485 		error = fo_read(fp, &auio, fp->f_cred, flags, td);
2486 		if (error != 0)
2487 			break;
2488 		n_read = to_send - auio.uio_resid;
2489 		if (n_read == 0)
2490 			break;
2491 		aiov.iov_base = buf;
2492 		aiov.iov_len = bufsz;
2493 		auio.uio_iov = &aiov;
2494 		auio.uio_iovcnt = 1;
2495 		auio.uio_segflg = UIO_SYSSPACE;
2496 		auio.uio_td = td;
2497 		auio.uio_rw = UIO_WRITE;
2498 		auio.uio_offset = (seekable) ? out_offset : 0;
2499 		auio.uio_resid = n_read;
2500 		error = fo_write(ofp, &auio, ofp->f_cred, flags, td);
2501 		if (error != 0)
2502 			break;
2503 		bytes_sent += n_read;
2504 		current_offset += n_read;
2505 		out_offset += n_read;
2506 	}
2507 	free(buf, M_LINUX);
2508 
2509 	if (error == 0) {
2510 		*sbytes = bytes_sent;
2511 		if (offset != NULL)
2512 			*offset = current_offset;
2513 		else
2514 			error = fo_seek(fp, current_offset, SEEK_SET, td);
2515 	}
2516 	if (error == 0 && seekable)
2517 		error = fo_seek(ofp, out_offset, SEEK_SET, td);
2518 
2519 drop:
2520 	fdrop(ofp, td);
2521 	return (error);
2522 }
2523 
2524 static int
2525 sendfile_sendfile(struct thread *td, struct file *fp, l_int out,
2526     off_t *offset, l_size_t count, off_t *sbytes)
2527 {
2528 	off_t current_offset;
2529 	int error;
2530 
2531 	if (offset == NULL) {
2532 		if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0)
2533 			return (ESPIPE);
2534 		if ((error = fo_seek(fp, 0, SEEK_CUR, td)) != 0)
2535 			return (error);
2536 		current_offset = td->td_uretoff.tdu_off;
2537 	} else
2538 		current_offset = *offset;
2539 	error = fo_sendfile(fp, out, NULL, NULL, current_offset, count,
2540 	    sbytes, 0, td);
2541 	if (error == 0) {
2542 		current_offset += *sbytes;
2543 		if (offset != NULL)
2544 			*offset = current_offset;
2545 		else
2546 			error = fo_seek(fp, current_offset, SEEK_SET, td);
2547 	}
2548 	return (error);
2549 }
2550 
2551 static int
2552 linux_sendfile_common(struct thread *td, l_int out, l_int in,
2553     off_t *offset, l_size_t count)
2554 {
2555 	struct file *fp, *ofp;
2556 	off_t sbytes;
2557 	int error;
2558 
2559 	/* Linux cannot have 0 count. */
2560 	if (count <= 0 || (offset != NULL && *offset < 0))
2561 		return (EINVAL);
2562 
2563 	AUDIT_ARG_FD(in);
2564 	error = fget_read(td, in, &cap_pread_rights, &fp);
2565 	if (error != 0)
2566 		return (error);
2567 	if ((fp->f_type != DTYPE_SHM && fp->f_type != DTYPE_VNODE) ||
2568 	    (fp->f_type == DTYPE_VNODE &&
2569 	    (fp->f_vnode == NULL || fp->f_vnode->v_type != VREG))) {
2570 		error = EINVAL;
2571 		goto drop;
2572 	}
2573 	error = fget_unlocked(td, out, &cap_no_rights, &ofp);
2574 	if (error != 0)
2575 		goto drop;
2576 
2577 	if (is_regular_file(fp) && is_regular_file(ofp)) {
2578 		error = kern_copy_file_range(td, in, offset, out, NULL, count,
2579 		    0);
2580 	} else {
2581 		sbytes = 0;
2582 		if (is_sendfile(fp, ofp))
2583 			error = sendfile_sendfile(td, fp, out, offset, count,
2584 			    &sbytes);
2585 		else
2586 			error = sendfile_fallback(td, fp, out, offset, count,
2587 			    &sbytes);
2588 		if (error == ENOBUFS && (ofp->f_flag & FNONBLOCK) != 0)
2589 			error = EAGAIN;
2590 		if (error == 0)
2591 			td->td_retval[0] = sbytes;
2592 	}
2593 	fdrop(ofp, td);
2594 
2595 drop:
2596 	fdrop(fp, td);
2597 	return (error);
2598 }
2599 
2600 int
2601 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg)
2602 {
2603 	/*
2604 	 * Differences between FreeBSD and Linux sendfile:
2605 	 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to
2606 	 *   mean send the whole file).
2607 	 * - Linux can send to any fd whereas FreeBSD only supports sockets.
2608 	 *   We therefore use FreeBSD sendfile where possible for performance,
2609 	 *   but fall back on a manual copy (sendfile_fallback).
2610 	 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr.
2611 	 * - Linux takes an offset pointer and updates it to the read location.
2612 	 *   FreeBSD takes in an offset and a 'bytes read' parameter which is
2613 	 *   only filled if it isn't NULL.  We use this parameter to update the
2614 	 *   offset pointer if it exists.
2615 	 * - Linux sendfile returns bytes read on success while FreeBSD
2616 	 *   returns 0.  We use the 'bytes read' parameter to get this value.
2617 	 */
2618 
2619 	off_t offset64;
2620 	l_off_t offset;
2621 	int error;
2622 
2623 	if (arg->offset != NULL) {
2624 		error = copyin(arg->offset, &offset, sizeof(offset));
2625 		if (error != 0)
2626 			return (error);
2627 		offset64 = offset;
2628 	}
2629 
2630 	error = linux_sendfile_common(td, arg->out, arg->in,
2631 	    arg->offset != NULL ? &offset64 : NULL, arg->count);
2632 
2633 	if (error == 0 && arg->offset != NULL) {
2634 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2635 		if (offset64 > INT32_MAX)
2636 			return (EOVERFLOW);
2637 #endif
2638 		offset = (l_off_t)offset64;
2639 		error = copyout(&offset, arg->offset, sizeof(offset));
2640 	}
2641 
2642 	return (error);
2643 }
2644 
2645 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2646 int
2647 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg)
2648 {
2649 	off_t offset;
2650 	int error;
2651 
2652 	if (arg->offset != NULL) {
2653 		error = copyin(arg->offset, &offset, sizeof(offset));
2654 		if (error != 0)
2655 			return (error);
2656 	}
2657 
2658 	error = linux_sendfile_common(td, arg->out, arg->in,
2659 		arg->offset != NULL ? &offset : NULL, arg->count);
2660 
2661 	if (error == 0 && arg->offset != NULL)
2662 		error = copyout(&offset, arg->offset, sizeof(offset));
2663 
2664 	return (error);
2665 }
2666 
2667 /* Argument list sizes for linux_socketcall */
2668 static const unsigned char lxs_args_cnt[] = {
2669 	0 /* unused*/,		3 /* socket */,
2670 	3 /* bind */,		3 /* connect */,
2671 	2 /* listen */,		3 /* accept */,
2672 	3 /* getsockname */,	3 /* getpeername */,
2673 	4 /* socketpair */,	4 /* send */,
2674 	4 /* recv */,		6 /* sendto */,
2675 	6 /* recvfrom */,	2 /* shutdown */,
2676 	5 /* setsockopt */,	5 /* getsockopt */,
2677 	3 /* sendmsg */,	3 /* recvmsg */,
2678 	4 /* accept4 */,	5 /* recvmmsg */,
2679 	4 /* sendmmsg */,	4 /* sendfile */
2680 };
2681 #define	LINUX_ARGS_CNT		(nitems(lxs_args_cnt) - 1)
2682 #define	LINUX_ARG_SIZE(x)	(lxs_args_cnt[x] * sizeof(l_ulong))
2683 
2684 int
2685 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
2686 {
2687 	l_ulong a[6];
2688 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2689 	register_t l_args[6];
2690 #endif
2691 	void *arg;
2692 	int error;
2693 
2694 	if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
2695 		return (EINVAL);
2696 	error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
2697 	if (error != 0)
2698 		return (error);
2699 
2700 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2701 	for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
2702 		l_args[i] = a[i];
2703 	arg = l_args;
2704 #else
2705 	arg = a;
2706 #endif
2707 	switch (args->what) {
2708 	case LINUX_SOCKET:
2709 		return (linux_socket(td, arg));
2710 	case LINUX_BIND:
2711 		return (linux_bind(td, arg));
2712 	case LINUX_CONNECT:
2713 		return (linux_connect(td, arg));
2714 	case LINUX_LISTEN:
2715 		return (linux_listen(td, arg));
2716 	case LINUX_ACCEPT:
2717 		return (linux_accept(td, arg));
2718 	case LINUX_GETSOCKNAME:
2719 		return (linux_getsockname(td, arg));
2720 	case LINUX_GETPEERNAME:
2721 		return (linux_getpeername(td, arg));
2722 	case LINUX_SOCKETPAIR:
2723 		return (linux_socketpair(td, arg));
2724 	case LINUX_SEND:
2725 		return (linux_send(td, arg));
2726 	case LINUX_RECV:
2727 		return (linux_recv(td, arg));
2728 	case LINUX_SENDTO:
2729 		return (linux_sendto(td, arg));
2730 	case LINUX_RECVFROM:
2731 		return (linux_recvfrom(td, arg));
2732 	case LINUX_SHUTDOWN:
2733 		return (linux_shutdown(td, arg));
2734 	case LINUX_SETSOCKOPT:
2735 		return (linux_setsockopt(td, arg));
2736 	case LINUX_GETSOCKOPT:
2737 		return (linux_getsockopt(td, arg));
2738 	case LINUX_SENDMSG:
2739 		return (linux_sendmsg(td, arg));
2740 	case LINUX_RECVMSG:
2741 		return (linux_recvmsg(td, arg));
2742 	case LINUX_ACCEPT4:
2743 		return (linux_accept4(td, arg));
2744 	case LINUX_RECVMMSG:
2745 		return (linux_recvmmsg(td, arg));
2746 	case LINUX_SENDMMSG:
2747 		return (linux_sendmmsg(td, arg));
2748 	case LINUX_SENDFILE:
2749 		return (linux_sendfile(td, arg));
2750 	}
2751 
2752 	linux_msg(td, "socket type %d not implemented", args->what);
2753 	return (ENOSYS);
2754 }
2755 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2756