2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 1995 Søren Schmidt
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 /* XXX we use functions that might not exist. */
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
36 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/sysproto.h>
40 #include <sys/capsicum.h>
41 #include <sys/fcntl.h>
43 #include <sys/filedesc.h>
44 #include <sys/limits.h>
46 #include <sys/malloc.h>
47 #include <sys/mutex.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/syscallsubr.h>
54 #include <sys/syslog.h>
56 #include <sys/unistd.h>
58 #include <security/audit/audit.h>
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/ip.h>
65 #include <netinet/tcp.h>
67 #include <netinet/ip6.h>
68 #include <netinet6/ip6_var.h>
72 #include <machine/../linux32/linux.h>
73 #include <machine/../linux32/linux32_proto.h>
75 #include <machine/../linux/linux.h>
76 #include <machine/../linux/linux_proto.h>
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>
85 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
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 *);
92 linux_to_bsd_sockopt_level(int level)
95 if (level == LINUX_SOL_SOCKET)
97 /* Remaining values are RFC-defined protocol numbers. */
102 bsd_to_linux_sockopt_level(int level)
105 if (level == SOL_SOCKET)
106 return (LINUX_SOL_SOCKET);
111 linux_to_bsd_ip_sockopt(int opt)
115 /* known and translated sockopts */
120 case LINUX_IP_HDRINCL:
122 case LINUX_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");
130 case LINUX_IP_RECVTTL:
131 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTTL");
133 case LINUX_IP_RECVTOS:
134 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_RECVTOS");
136 case LINUX_IP_FREEBIND:
137 LINUX_RATELIMIT_MSG_NOTTESTED("IPv4 socket option IP_FREEBIND");
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");
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);
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",
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",
192 case LINUX_IP_PKTOPTIONS:
193 LINUX_RATELIMIT_MSG_OPT1(
194 "unsupported IPv4 socket option IP_PKTOPTIONS (%d)",
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",
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",
209 LINUX_RATELIMIT_MSG_OPT1(
210 "unsupported IPv4 socket option IP_MTU (%d), your linux program can not control the MTU on this socket",
213 case LINUX_IP_XFRM_POLICY:
214 LINUX_RATELIMIT_MSG_OPT1(
215 "unsupported IPv4 socket option IP_XFRM_POLICY (%d)",
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",
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",
230 case LINUX_IP_NODEFRAG:
231 LINUX_RATELIMIT_MSG_OPT1(
232 "unsupported IPv4 socket option IP_NODEFRAG (%d)",
235 case LINUX_IP_CHECKSUM:
236 LINUX_RATELIMIT_MSG_OPT1(
237 "unsupported IPv4 socket option IP_CHECKSUM (%d)",
240 case LINUX_IP_BIND_ADDRESS_NO_PORT:
241 LINUX_RATELIMIT_MSG_OPT1(
242 "unsupported IPv4 socket option IP_BIND_ADDRESS_NO_PORT (%d)",
245 case LINUX_IP_RECVFRAGSIZE:
246 LINUX_RATELIMIT_MSG_OPT1(
247 "unsupported IPv4 socket option IP_RECVFRAGSIZE (%d)",
250 case LINUX_MCAST_MSFILTER:
251 LINUX_RATELIMIT_MSG_OPT1(
252 "unsupported IPv4 socket option IP_MCAST_MSFILTER (%d)",
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",
260 case LINUX_IP_UNICAST_IF:
261 LINUX_RATELIMIT_MSG_OPT1(
262 "unsupported IPv4 socket option IP_UNICAST_IF (%d)",
266 /* unknown sockopts */
273 linux_to_bsd_ip6_sockopt(int 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");
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);
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",
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",
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",
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",
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",
403 LINUX_RATELIMIT_MSG_OPT1(
404 "unsupported IPv6 socket option IPV6_MTU (%d), your linux program can not control the MTU on this socket",
407 case LINUX_IPV6_JOIN_ANYCAST:
408 LINUX_RATELIMIT_MSG_OPT1(
409 "unsupported IPv6 socket option IPV6_JOIN_ANYCAST (%d)",
412 case LINUX_IPV6_LEAVE_ANYCAST:
413 LINUX_RATELIMIT_MSG_OPT1(
414 "unsupported IPv6 socket option IPV6_LEAVE_ANYCAST (%d)",
417 case LINUX_IPV6_MULTICAST_ALL:
418 LINUX_RATELIMIT_MSG_OPT1(
419 "unsupported IPv6 socket option IPV6_MULTICAST_ALL (%d)",
422 case LINUX_IPV6_ROUTER_ALERT_ISOLATE:
423 LINUX_RATELIMIT_MSG_OPT1(
424 "unsupported IPv6 socket option IPV6_ROUTER_ALERT_ISOLATE (%d)",
427 case LINUX_IPV6_FLOWLABEL_MGR:
428 LINUX_RATELIMIT_MSG_OPT1(
429 "unsupported IPv6 socket option IPV6_FLOWLABEL_MGR (%d)",
432 case LINUX_IPV6_FLOWINFO_SEND:
433 LINUX_RATELIMIT_MSG_OPT1(
434 "unsupported IPv6 socket option IPV6_FLOWINFO_SEND (%d)",
437 case LINUX_IPV6_XFRM_POLICY:
438 LINUX_RATELIMIT_MSG_OPT1(
439 "unsupported IPv6 socket option IPV6_XFRM_POLICY (%d)",
442 case LINUX_IPV6_HDRINCL:
443 LINUX_RATELIMIT_MSG_OPT1(
444 "unsupported IPv6 socket option IPV6_HDRINCL (%d)",
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",
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",
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",
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",
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",
472 case LINUX_IPV6_ADDR_PREFERENCES:
473 LINUX_RATELIMIT_MSG_OPT1(
474 "unsupported IPv6 socket option IPV6_ADDR_PREFERENCES (%d)",
477 case LINUX_IPV6_MINHOPCOUNT:
478 LINUX_RATELIMIT_MSG_OPT1(
479 "unsupported IPv6 socket option IPV6_MINHOPCOUNT (%d)",
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",
488 case LINUX_IPV6_UNICAST_IF:
489 LINUX_RATELIMIT_MSG_OPT1(
490 "unsupported IPv6 socket option IPV6_UNICAST_IF (%d)",
493 case LINUX_IPV6_RECVFRAGSIZE:
494 LINUX_RATELIMIT_MSG_OPT1(
495 "unsupported IPv6 socket option IPV6_RECVFRAGSIZE (%d)",
499 /* unknown sockopts */
506 linux_to_bsd_so_sockopt(int opt)
512 case LINUX_SO_REUSEADDR:
513 return (SO_REUSEADDR);
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:
525 case LINUX_SO_RCVBUF:
526 case LINUX_SO_RCVBUFFORCE:
528 case LINUX_SO_KEEPALIVE:
529 return (SO_KEEPALIVE);
530 case LINUX_SO_OOBINLINE:
531 return (SO_OOBINLINE);
532 case LINUX_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);
559 linux_to_bsd_tcp_sockopt(int opt)
563 case LINUX_TCP_NODELAY:
564 return (TCP_NODELAY);
565 case LINUX_TCP_MAXSEG:
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:
582 linux_to_bsd_msg_flags(int flags)
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",
607 if (flags & LINUX_MSG_FIN)
608 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_FIN (%d) not handled",
610 if (flags & LINUX_MSG_SYN)
611 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_SYN (%d) not handled",
613 if (flags & LINUX_MSG_CONFIRM)
614 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_CONFIRM (%d) not handled",
616 if (flags & LINUX_MSG_RST)
617 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_RST (%d) not handled",
619 if (flags & LINUX_MSG_ERRQUEUE)
620 LINUX_RATELIMIT_MSG_OPT1("socket message flag MSG_ERRQUEUE (%d) not handled",
626 linux_to_bsd_cmsg_type(int cmsg_type)
630 case LINUX_SCM_RIGHTS:
632 case LINUX_SCM_CREDENTIALS:
639 bsd_to_linux_cmsg_type(int cmsg_type)
644 return (LINUX_SCM_RIGHTS);
646 return (LINUX_SCM_CREDENTIALS);
648 return (LINUX_SCM_CREDENTIALS);
650 return (LINUX_SCM_TIMESTAMP);
656 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
658 if (lhdr->msg_controllen > INT_MAX)
661 bhdr->msg_name = PTRIN(lhdr->msg_name);
662 bhdr->msg_namelen = lhdr->msg_namelen;
663 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
664 bhdr->msg_iovlen = lhdr->msg_iovlen;
665 bhdr->msg_control = PTRIN(lhdr->msg_control);
668 * msg_controllen is skipped since BSD and LINUX control messages
669 * are potentially different sizes (e.g. the cred structure used
670 * by SCM_CREDS is different between the two operating system).
672 * The caller can set it (if necessary) after converting all the
676 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
681 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
683 lhdr->msg_name = PTROUT(bhdr->msg_name);
684 lhdr->msg_namelen = bhdr->msg_namelen;
685 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
686 lhdr->msg_iovlen = bhdr->msg_iovlen;
687 lhdr->msg_control = PTROUT(bhdr->msg_control);
690 * msg_controllen is skipped since BSD and LINUX control messages
691 * are potentially different sizes (e.g. the cred structure used
692 * by SCM_CREDS is different between the two operating system).
694 * The caller can set it (if necessary) after converting all the
698 /* msg_flags skipped */
703 linux_set_socket_flags(int lflags, int *flags)
706 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
708 if (lflags & LINUX_SOCK_NONBLOCK)
709 *flags |= SOCK_NONBLOCK;
710 if (lflags & LINUX_SOCK_CLOEXEC)
711 *flags |= SOCK_CLOEXEC;
716 linux_copyout_sockaddr(const struct sockaddr *sa, void *uaddr, size_t len)
718 struct l_sockaddr *lsa;
721 error = bsd_to_linux_sockaddr(sa, &lsa, len);
725 error = copyout(lsa, uaddr, len);
732 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
733 struct mbuf *control, enum uio_seg segflg)
738 if (mp->msg_name != NULL) {
739 len = mp->msg_namelen;
740 error = linux_to_bsd_sockaddr(mp->msg_name, &to, &len);
747 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
755 /* Return 0 if IP_HDRINCL is set for the given socket. */
757 linux_check_hdrincl(struct thread *td, int s)
762 size_val = sizeof(optval);
763 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
764 &optval, UIO_SYSSPACE, &size_val);
768 return (optval == 0);
772 * Updated sendto() when IP_HDRINCL is set:
773 * tweak endian-dependent fields in the IP packet.
776 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
779 * linux_ip_copysize defines how many bytes we should copy
780 * from the beginning of the IP packet before we customize it for BSD.
781 * It should include all the fields we modify (ip_len and ip_off).
783 #define linux_ip_copysize 8
787 struct iovec aiov[1];
790 /* Check that the packet isn't too big or too small. */
791 if (linux_args->len < linux_ip_copysize ||
792 linux_args->len > IP_MAXPACKET)
795 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
797 /* Make kernel copy of the packet to be sent */
798 if ((error = copyin(PTRIN(linux_args->msg), packet,
802 /* Convert fields from Linux to BSD raw IP socket format */
803 packet->ip_len = linux_args->len;
804 packet->ip_off = ntohs(packet->ip_off);
806 /* Prepare the msghdr and iovec structures describing the new packet */
807 msg.msg_name = PTRIN(linux_args->to);
808 msg.msg_namelen = linux_args->tolen;
811 msg.msg_control = NULL;
813 aiov[0].iov_base = (char *)packet;
814 aiov[0].iov_len = linux_args->len;
815 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
818 free(packet, M_LINUX);
822 static const char *linux_netlink_names[] = {
823 [LINUX_NETLINK_ROUTE] = "ROUTE",
824 [LINUX_NETLINK_SOCK_DIAG] = "SOCK_DIAG",
825 [LINUX_NETLINK_NFLOG] = "NFLOG",
826 [LINUX_NETLINK_SELINUX] = "SELINUX",
827 [LINUX_NETLINK_AUDIT] = "AUDIT",
828 [LINUX_NETLINK_FIB_LOOKUP] = "FIB_LOOKUP",
829 [LINUX_NETLINK_NETFILTER] = "NETFILTER",
830 [LINUX_NETLINK_KOBJECT_UEVENT] = "KOBJECT_UEVENT",
834 linux_socket(struct thread *td, struct linux_socket_args *args)
836 int domain, retval_socket, type;
838 type = args->type & LINUX_SOCK_TYPE_MASK;
839 if (type < 0 || type > LINUX_SOCK_MAX)
841 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
843 if (retval_socket != 0)
844 return (retval_socket);
845 domain = linux_to_bsd_domain(args->domain);
847 /* Mask off SOCK_NONBLOCK / CLOEXEC for error messages. */
848 type = args->type & LINUX_SOCK_TYPE_MASK;
849 if (args->domain == LINUX_AF_NETLINK) {
852 if (args->protocol >= 0 &&
853 args->protocol < nitems(linux_netlink_names))
854 nl_name = linux_netlink_names[args->protocol];
859 "unsupported socket(AF_NETLINK, %d, "
860 "NETLINK_%s)", type, nl_name);
863 "unsupported socket(AF_NETLINK, %d, %d)",
864 type, args->protocol);
866 linux_msg(curthread, "unsupported socket domain %d, "
867 "type %d, protocol %d", args->domain, type,
870 return (EAFNOSUPPORT);
873 retval_socket = kern_socket(td, domain, type, args->protocol);
875 return (retval_socket);
878 && (args->protocol == IPPROTO_RAW || args->protocol == 0)
879 && domain == PF_INET) {
880 /* It's a raw IP socket: set the IP_HDRINCL option. */
884 /* We ignore any error returned by kern_setsockopt() */
885 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
886 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
890 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
891 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
892 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
895 if (domain == PF_INET6) {
899 /* We ignore any error returned by setsockopt() */
900 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
901 &v6only, UIO_SYSSPACE, sizeof(v6only));
905 return (retval_socket);
909 linux_bind(struct thread *td, struct linux_bind_args *args)
914 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
919 error = kern_bindat(td, AT_FDCWD, args->s, sa);
923 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
929 linux_connect(struct thread *td, struct linux_connect_args *args)
937 error = linux_to_bsd_sockaddr(PTRIN(args->name), &sa,
942 error = kern_connectat(td, AT_FDCWD, args->s, sa);
944 if (error != EISCONN)
948 * Linux doesn't return EISCONN the first time it occurs,
949 * when on a non-blocking socket. Instead it returns the
950 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
952 error = getsock_cap(td, args->s, &cap_connect_rights,
959 if (fflag & FNONBLOCK) {
961 if (so->so_emuldata == 0)
962 error = so->so_error;
963 so->so_emuldata = (void *)1;
972 linux_listen(struct thread *td, struct linux_listen_args *args)
975 return (kern_listen(td, args->s, args->backlog));
979 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
980 l_uintptr_t namelen, int flags)
983 struct file *fp, *fp1;
992 error = linux_set_socket_flags(flags, &bflags);
996 if (PTRIN(addr) == NULL) {
998 error = kern_accept4(td, s, NULL, NULL, bflags, NULL);
1000 error = copyin(PTRIN(namelen), &len, sizeof(len));
1005 error = kern_accept4(td, s, &sa, &len, bflags, &fp);
1009 * Translate errno values into ones used by Linux.
1013 * XXX. This is wrong, different sockaddr structures
1014 * have different sizes.
1018 if (namelen != sizeof(struct sockaddr_in))
1022 error1 = getsock_cap(td, s, &cap_accept_rights, &fp1, NULL, NULL);
1028 if (so->so_type == SOCK_DGRAM)
1037 error = linux_copyout_sockaddr(sa, PTRIN(addr), len);
1040 * XXX: We should also copyout the len, shouldn't we?
1044 fdclose(td, fp, td->td_retval[0]);
1045 td->td_retval[0] = 0;
1055 linux_accept(struct thread *td, struct linux_accept_args *args)
1058 return (linux_accept_common(td, args->s, args->addr,
1063 linux_accept4(struct thread *td, struct linux_accept4_args *args)
1066 return (linux_accept_common(td, args->s, args->addr,
1067 args->namelen, args->flags));
1071 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
1073 struct sockaddr *sa;
1076 error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1080 error = kern_getsockname(td, args->s, &sa, &len);
1085 error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
1089 error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1094 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
1096 struct sockaddr *sa;
1099 error = copyin(PTRIN(args->namelen), &len, sizeof(len));
1105 error = kern_getpeername(td, args->s, &sa, &len);
1110 error = linux_copyout_sockaddr(sa, PTRIN(args->addr), len);
1114 error = copyout(&len, PTRIN(args->namelen), sizeof(len));
1119 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
1121 int domain, error, sv[2], type;
1123 domain = linux_to_bsd_domain(args->domain);
1124 if (domain != PF_LOCAL)
1125 return (EAFNOSUPPORT);
1126 type = args->type & LINUX_SOCK_TYPE_MASK;
1127 if (type < 0 || type > LINUX_SOCK_MAX)
1129 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
1133 if (args->protocol != 0 && args->protocol != PF_UNIX) {
1135 * Use of PF_UNIX as protocol argument is not right,
1136 * but Linux does it.
1137 * Do not map PF_UNIX as its Linux value is identical
1140 return (EPROTONOSUPPORT);
1142 error = kern_socketpair(td, domain, type, 0, sv);
1145 error = copyout(sv, PTRIN(args->rsv), 2 * sizeof(int));
1147 (void)kern_close(td, sv[0]);
1148 (void)kern_close(td, sv[1]);
1153 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1154 struct linux_send_args {
1162 linux_send(struct thread *td, struct linux_send_args *args)
1164 struct sendto_args /* {
1175 bsd_args.s = args->s;
1176 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1177 bsd_args.len = args->len;
1178 bsd_args.flags = args->flags;
1181 error = sys_sendto(td, &bsd_args);
1182 if (error == ENOTCONN) {
1184 * Linux doesn't return ENOTCONN for non-blocking sockets.
1185 * Instead it returns the EAGAIN.
1187 error = getsock_cap(td, args->s, &cap_send_rights, &fp,
1190 if (fflag & FNONBLOCK)
1198 struct linux_recv_args {
1206 linux_recv(struct thread *td, struct linux_recv_args *args)
1208 struct recvfrom_args /* {
1213 struct sockaddr *from;
1214 socklen_t fromlenaddr;
1217 bsd_args.s = args->s;
1218 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1219 bsd_args.len = args->len;
1220 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1221 bsd_args.from = NULL;
1222 bsd_args.fromlenaddr = 0;
1223 return (sys_recvfrom(td, &bsd_args));
1225 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1228 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1233 if (linux_check_hdrincl(td, args->s) == 0)
1234 /* IP_HDRINCL set, tweak the packet before sending */
1235 return (linux_sendto_hdrincl(td, args));
1237 msg.msg_name = PTRIN(args->to);
1238 msg.msg_namelen = args->tolen;
1239 msg.msg_iov = &aiov;
1241 msg.msg_control = NULL;
1243 aiov.iov_base = PTRIN(args->msg);
1244 aiov.iov_len = args->len;
1245 return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1250 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1252 struct sockaddr *sa;
1257 if (PTRIN(args->fromlen) != NULL) {
1258 error = copyin(PTRIN(args->fromlen), &fromlen,
1264 sa = malloc(fromlen, M_SONAME, M_WAITOK);
1271 msg.msg_namelen = fromlen;
1272 msg.msg_iov = &aiov;
1274 aiov.iov_base = PTRIN(args->buf);
1275 aiov.iov_len = args->len;
1276 msg.msg_control = 0;
1277 msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1279 error = kern_recvit(td, args->s, &msg, UIO_SYSSPACE, NULL);
1283 if (PTRIN(args->from) != NULL)
1284 error = linux_copyout_sockaddr(sa, PTRIN(args->from), msg.msg_namelen);
1286 if (error == 0 && PTRIN(args->fromlen) != NULL)
1287 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1288 sizeof(msg.msg_namelen));
1295 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1298 struct cmsghdr *cmsg;
1299 struct mbuf *control;
1301 struct l_cmsghdr linux_cmsg;
1302 struct l_cmsghdr *ptr_cmsg;
1303 struct l_msghdr linux_msghdr;
1306 struct sockaddr *sa;
1308 sa_family_t sa_family;
1315 error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
1320 * Some Linux applications (ping) define a non-NULL control data
1321 * pointer, but a msg_controllen of 0, which is not allowed in the
1322 * FreeBSD system call interface. NULL the msg_control pointer in
1323 * order to handle this case. This should be checked, but allows the
1324 * Linux ping to work.
1326 if (PTRIN(linux_msghdr.msg_control) != NULL &&
1327 linux_msghdr.msg_controllen == 0)
1328 linux_msghdr.msg_control = PTROUT(NULL);
1330 error = linux_to_bsd_msghdr(&msg, &linux_msghdr);
1334 #ifdef COMPAT_LINUX32
1335 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1338 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1345 error = kern_getsockname(td, s, &sa, &datalen);
1348 sa_family = sa->sa_family;
1351 if (flags & LINUX_MSG_OOB) {
1353 if (sa_family == AF_UNIX)
1356 error = getsock_cap(td, s, &cap_send_rights, &fp,
1361 if (so->so_type != SOCK_STREAM)
1368 if (linux_msghdr.msg_controllen >= sizeof(struct l_cmsghdr)) {
1370 control = m_get(M_WAITOK, MT_CONTROL);
1371 MCLGET(control, M_WAITOK);
1372 data = mtod(control, void *);
1375 ptr_cmsg = PTRIN(linux_msghdr.msg_control);
1376 clen = linux_msghdr.msg_controllen;
1378 error = copyin(ptr_cmsg, &linux_cmsg,
1379 sizeof(struct l_cmsghdr));
1384 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1385 linux_cmsg.cmsg_len > clen)
1388 if (datalen + CMSG_HDRSZ > MCLBYTES)
1392 * Now we support only SCM_RIGHTS and SCM_CRED,
1393 * so return EINVAL in any other cmsg_type
1397 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1399 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1400 if (cmsg->cmsg_type == -1
1401 || cmsg->cmsg_level != SOL_SOCKET) {
1402 linux_msg(curthread,
1403 "unsupported sendmsg cmsg level %d type %d",
1404 linux_cmsg.cmsg_level, linux_cmsg.cmsg_type);
1409 * Some applications (e.g. pulseaudio) attempt to
1410 * send ancillary data even if the underlying protocol
1411 * doesn't support it which is not allowed in the
1412 * FreeBSD system call interface.
1414 if (sa_family != AF_UNIX)
1417 if (cmsg->cmsg_type == SCM_CREDS) {
1418 len = sizeof(struct cmsgcred);
1419 if (datalen + CMSG_SPACE(len) > MCLBYTES)
1423 * The lower levels will fill in the structure
1425 memset(CMSG_DATA(data), 0, len);
1427 len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1428 if (datalen + CMSG_SPACE(len) < datalen ||
1429 datalen + CMSG_SPACE(len) > MCLBYTES)
1432 error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1433 CMSG_DATA(data), len);
1438 cmsg->cmsg_len = CMSG_LEN(len);
1439 data = (char *)data + CMSG_SPACE(len);
1440 datalen += CMSG_SPACE(len);
1443 if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1446 clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1447 ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1448 LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1449 } while(clen >= sizeof(struct l_cmsghdr));
1451 control->m_len = datalen;
1460 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1470 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1473 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1478 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1480 struct l_mmsghdr *msg;
1482 int error, datagrams;
1484 if (args->vlen > UIO_MAXIOV)
1485 args->vlen = UIO_MAXIOV;
1487 msg = PTRIN(args->msg);
1489 while (datagrams < args->vlen) {
1490 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1495 retval = td->td_retval[0];
1496 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1503 td->td_retval[0] = datagrams;
1508 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1509 l_uint flags, struct msghdr *msg)
1512 struct cmsgcred *cmcred;
1513 struct sockcred2 *scred;
1514 struct l_cmsghdr *linux_cmsg = NULL;
1515 struct l_ucred linux_ucred;
1516 socklen_t datalen, maxlen, outlen;
1517 struct l_msghdr linux_msghdr;
1518 struct iovec *iov, *uiov;
1519 struct mbuf *control = NULL;
1520 struct mbuf **controlp;
1521 struct timeval *ftmvl;
1522 struct sockaddr *sa;
1526 int error, i, fd, fds, *fdp;
1528 error = copyin(msghdr, &linux_msghdr, sizeof(linux_msghdr));
1532 error = linux_to_bsd_msghdr(msg, &linux_msghdr);
1536 #ifdef COMPAT_LINUX32
1537 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1540 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1545 if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1546 msg->msg_namelen = min(msg->msg_namelen, SOCK_MAXADDRLEN);
1547 sa = malloc(msg->msg_namelen, M_SONAME, M_WAITOK);
1551 msg->msg_name = NULL;
1554 uiov = msg->msg_iov;
1556 controlp = (msg->msg_control != NULL) ? &control : NULL;
1557 error = kern_recvit(td, s, msg, UIO_SYSSPACE, controlp);
1558 msg->msg_iov = uiov;
1563 * Note that kern_recvit() updates msg->msg_namelen.
1565 if (msg->msg_name != NULL && msg->msg_namelen > 0) {
1566 msg->msg_name = PTRIN(linux_msghdr.msg_name);
1567 error = linux_copyout_sockaddr(sa,
1568 PTRIN(msg->msg_name), msg->msg_namelen);
1573 error = bsd_to_linux_msghdr(msg, &linux_msghdr);
1577 maxlen = linux_msghdr.msg_controllen;
1578 linux_msghdr.msg_controllen = 0;
1580 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1582 msg->msg_control = mtod(control, struct cmsghdr *);
1583 msg->msg_controllen = control->m_len;
1585 cm = CMSG_FIRSTHDR(msg);
1586 outbuf = PTRIN(linux_msghdr.msg_control);
1588 while (cm != NULL) {
1589 linux_cmsg->cmsg_type =
1590 bsd_to_linux_cmsg_type(cm->cmsg_type);
1591 linux_cmsg->cmsg_level =
1592 bsd_to_linux_sockopt_level(cm->cmsg_level);
1593 if (linux_cmsg->cmsg_type == -1 ||
1594 cm->cmsg_level != SOL_SOCKET) {
1595 linux_msg(curthread,
1596 "unsupported recvmsg cmsg level %d type %d",
1597 cm->cmsg_level, cm->cmsg_type);
1602 data = CMSG_DATA(cm);
1603 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1605 switch (cm->cmsg_type) {
1607 if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1608 fds = datalen / sizeof(int);
1610 for (i = 0; i < fds; i++) {
1612 (void)kern_fcntl(td, fd,
1613 F_SETFD, FD_CLOEXEC);
1620 * Currently LOCAL_CREDS is never in
1621 * effect for Linux so no need to worry
1624 if (datalen != sizeof(*cmcred)) {
1628 cmcred = (struct cmsgcred *)data;
1629 bzero(&linux_ucred, sizeof(linux_ucred));
1630 linux_ucred.pid = cmcred->cmcred_pid;
1631 linux_ucred.uid = cmcred->cmcred_uid;
1632 linux_ucred.gid = cmcred->cmcred_gid;
1633 data = &linux_ucred;
1634 datalen = sizeof(linux_ucred);
1639 bzero(&linux_ucred, sizeof(linux_ucred));
1640 linux_ucred.pid = scred->sc_pid;
1641 linux_ucred.uid = scred->sc_uid;
1642 linux_ucred.gid = scred->sc_gid;
1643 data = &linux_ucred;
1644 datalen = sizeof(linux_ucred);
1648 if (datalen != sizeof(struct timeval)) {
1652 ftmvl = (struct timeval *)data;
1653 ltmvl.tv_sec = ftmvl->tv_sec;
1654 ltmvl.tv_usec = ftmvl->tv_usec;
1656 datalen = sizeof(ltmvl);
1660 if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1665 linux_msghdr.msg_flags |= LINUX_MSG_CTRUNC;
1666 m_dispose_extcontrolm(control);
1671 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1673 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1676 outbuf += L_CMSG_HDRSZ;
1678 error = copyout(data, outbuf, datalen);
1682 outbuf += LINUX_CMSG_ALIGN(datalen);
1683 outlen += LINUX_CMSG_LEN(datalen);
1685 cm = CMSG_NXTHDR(msg, cm);
1687 linux_msghdr.msg_controllen = outlen;
1691 error = copyout(&linux_msghdr, msghdr, sizeof(linux_msghdr));
1694 if (control != NULL) {
1696 m_dispose_extcontrolm(control);
1700 free(linux_cmsg, M_LINUX);
1707 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1709 struct msghdr bsd_msg;
1711 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1712 args->flags, &bsd_msg));
1716 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1718 struct l_mmsghdr *msg;
1719 struct msghdr bsd_msg;
1720 struct l_timespec lts;
1721 struct timespec ts, tts;
1723 int error, datagrams;
1725 if (args->timeout) {
1726 error = copyin(args->timeout, <s, sizeof(struct l_timespec));
1729 error = linux_to_native_timespec(&ts, <s);
1733 timespecadd(&tts, &ts, &tts);
1736 msg = PTRIN(args->msg);
1738 while (datagrams < args->vlen) {
1739 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1740 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1744 retval = td->td_retval[0];
1745 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1752 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1754 if (args->flags & LINUX_MSG_WAITFORONE)
1755 args->flags |= LINUX_MSG_DONTWAIT;
1758 * See BUGS section of recvmmsg(2).
1760 if (args->timeout) {
1762 timespecsub(&ts, &tts, &ts);
1763 if (!timespecisset(&ts) || ts.tv_sec > 0)
1766 /* Out of band data, return right away. */
1767 if (bsd_msg.msg_flags & MSG_OOB)
1771 td->td_retval[0] = datagrams;
1776 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1779 return (kern_shutdown(td, args->s, args->how));
1783 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1786 struct sockaddr *sa;
1789 int error, level, name;
1791 level = linux_to_bsd_sockopt_level(args->level);
1794 name = linux_to_bsd_so_sockopt(args->optname);
1796 case LOCAL_CREDS_PERSISTENT:
1802 error = copyin(PTRIN(args->optval), &linux_tv,
1806 tv.tv_sec = linux_tv.tv_sec;
1807 tv.tv_usec = linux_tv.tv_usec;
1808 return (kern_setsockopt(td, args->s, level,
1809 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1816 if (args->optname == LINUX_IP_RECVERR &&
1817 linux_ignore_ip_recverr) {
1819 * XXX: This is a hack to unbreak DNS resolution
1820 * with glibc 2.30 and above.
1824 name = linux_to_bsd_ip_sockopt(args->optname);
1827 name = linux_to_bsd_ip6_sockopt(args->optname);
1830 name = linux_to_bsd_tcp_sockopt(args->optname);
1838 linux_msg(curthread,
1839 "unsupported setsockopt level %d optname %d",
1840 args->level, args->optname);
1841 return (ENOPROTOOPT);
1844 if (name == IPV6_NEXTHOP) {
1846 error = linux_to_bsd_sockaddr(PTRIN(args->optval), &sa, &len);
1850 error = kern_setsockopt(td, args->s, level,
1851 name, sa, UIO_SYSSPACE, len);
1854 error = kern_setsockopt(td, args->s, level,
1855 name, PTRIN(args->optval), UIO_USERSPACE, args->optlen);
1862 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1866 socklen_t tv_len, xulen, len;
1867 struct sockaddr *sa;
1870 int error, level, name, newval;
1872 level = linux_to_bsd_sockopt_level(args->level);
1875 name = linux_to_bsd_so_sockopt(args->optname);
1877 case LOCAL_CREDS_PERSISTENT:
1883 tv_len = sizeof(tv);
1884 error = kern_getsockopt(td, args->s, level,
1885 name, &tv, UIO_SYSSPACE, &tv_len);
1888 linux_tv.tv_sec = tv.tv_sec;
1889 linux_tv.tv_usec = tv.tv_usec;
1890 return (copyout(&linux_tv, PTRIN(args->optval),
1893 case LOCAL_PEERCRED:
1894 if (args->optlen < sizeof(lxu))
1897 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
1898 * but by the Unix socket's level 0.
1902 error = kern_getsockopt(td, args->s, level,
1903 name, &xu, UIO_SYSSPACE, &xulen);
1906 lxu.pid = xu.cr_pid;
1907 lxu.uid = xu.cr_uid;
1908 lxu.gid = xu.cr_gid;
1909 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1912 len = sizeof(newval);
1913 error = kern_getsockopt(td, args->s, level,
1914 name, &newval, UIO_SYSSPACE, &len);
1917 newval = -bsd_to_linux_errno(newval);
1918 return (copyout(&newval, PTRIN(args->optval), len));
1925 name = linux_to_bsd_ip_sockopt(args->optname);
1928 name = linux_to_bsd_ip6_sockopt(args->optname);
1931 name = linux_to_bsd_tcp_sockopt(args->optname);
1939 linux_msg(curthread,
1940 "unsupported getsockopt level %d optname %d",
1941 args->level, args->optname);
1945 if (name == IPV6_NEXTHOP) {
1946 error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1949 sa = malloc(len, M_SONAME, M_WAITOK);
1951 error = kern_getsockopt(td, args->s, level,
1952 name, sa, UIO_SYSSPACE, &len);
1956 error = linux_copyout_sockaddr(sa, PTRIN(args->optval), len);
1958 error = copyout(&len, PTRIN(args->optlen),
1964 error = copyin(PTRIN(args->optlen), &len, sizeof(len));
1968 error = kern_getsockopt(td, args->s, level,
1969 name, PTRIN(args->optval), UIO_USERSPACE, &len);
1971 error = copyout(&len, PTRIN(args->optlen),
1979 linux_sendfile_common(struct thread *td, l_int out, l_int in,
1980 l_loff_t *offset, l_size_t count)
1984 l_loff_t current_offset;
1988 error = fget_read(td, in, &cap_pread_rights, &fp);
1992 if (offset != NULL) {
1993 current_offset = *offset;
1995 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1996 fo_seek(fp, 0, SEEK_CUR, td) : ESPIPE;
1999 current_offset = td->td_uretoff.tdu_off;
2004 /* Linux cannot have 0 count. */
2005 if (count <= 0 || current_offset < 0) {
2010 error = fo_sendfile(fp, out, NULL, NULL, current_offset, count,
2011 &bytes_read, 0, td);
2014 current_offset += bytes_read;
2016 if (offset != NULL) {
2017 *offset = current_offset;
2019 error = fo_seek(fp, current_offset, SEEK_SET, td);
2024 td->td_retval[0] = (ssize_t)bytes_read;
2031 linux_sendfile(struct thread *td, struct linux_sendfile_args *arg)
2034 * Differences between FreeBSD and Linux sendfile:
2035 * - Linux doesn't send anything when count is 0 (FreeBSD uses 0 to
2036 * mean send the whole file.) In linux_sendfile given fds are still
2037 * checked for validity when the count is 0.
2038 * - Linux can send to any fd whereas FreeBSD only supports sockets.
2039 * The same restriction follows for linux_sendfile.
2040 * - Linux doesn't have an equivalent for FreeBSD's flags and sf_hdtr.
2041 * - Linux takes an offset pointer and updates it to the read location.
2042 * FreeBSD takes in an offset and a 'bytes read' parameter which is
2043 * only filled if it isn't NULL. We use this parameter to update the
2044 * offset pointer if it exists.
2045 * - Linux sendfile returns bytes read on success while FreeBSD
2046 * returns 0. We use the 'bytes read' parameter to get this value.
2054 if (arg->offset != NULL) {
2055 error = copyin(arg->offset, &offset, sizeof(offset));
2058 offset64 = (l_loff_t)offset;
2061 ret = linux_sendfile_common(td, arg->out, arg->in,
2062 arg->offset != NULL ? &offset64 : NULL, arg->count);
2064 if (arg->offset != NULL) {
2065 #if defined(__i386__) || defined(__arm__) || \
2066 (defined(__amd64__) && defined(COMPAT_LINUX32))
2067 if (offset64 > INT32_MAX)
2070 offset = (l_long)offset64;
2071 error = copyout(&offset, arg->offset, sizeof(offset));
2079 #if defined(__i386__) || defined(__arm__) || \
2080 (defined(__amd64__) && defined(COMPAT_LINUX32))
2083 linux_sendfile64(struct thread *td, struct linux_sendfile64_args *arg)
2089 if (arg->offset != NULL) {
2090 error = copyin(arg->offset, &offset, sizeof(offset));
2095 ret = linux_sendfile_common(td, arg->out, arg->in,
2096 arg->offset != NULL ? &offset : NULL, arg->count);
2098 if (arg->offset != NULL) {
2099 error = copyout(&offset, arg->offset, sizeof(offset));
2107 /* Argument list sizes for linux_socketcall */
2108 static const unsigned char lxs_args_cnt[] = {
2109 0 /* unused*/, 3 /* socket */,
2110 3 /* bind */, 3 /* connect */,
2111 2 /* listen */, 3 /* accept */,
2112 3 /* getsockname */, 3 /* getpeername */,
2113 4 /* socketpair */, 4 /* send */,
2114 4 /* recv */, 6 /* sendto */,
2115 6 /* recvfrom */, 2 /* shutdown */,
2116 5 /* setsockopt */, 5 /* getsockopt */,
2117 3 /* sendmsg */, 3 /* recvmsg */,
2118 4 /* accept4 */, 5 /* recvmmsg */,
2119 4 /* sendmmsg */, 4 /* sendfile */
2121 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1)
2122 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong))
2125 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
2128 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2129 register_t l_args[6];
2134 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
2136 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
2140 #if defined(__amd64__) && defined(COMPAT_LINUX32)
2141 for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
2147 switch (args->what) {
2149 return (linux_socket(td, arg));
2151 return (linux_bind(td, arg));
2153 return (linux_connect(td, arg));
2155 return (linux_listen(td, arg));
2157 return (linux_accept(td, arg));
2158 case LINUX_GETSOCKNAME:
2159 return (linux_getsockname(td, arg));
2160 case LINUX_GETPEERNAME:
2161 return (linux_getpeername(td, arg));
2162 case LINUX_SOCKETPAIR:
2163 return (linux_socketpair(td, arg));
2165 return (linux_send(td, arg));
2167 return (linux_recv(td, arg));
2169 return (linux_sendto(td, arg));
2170 case LINUX_RECVFROM:
2171 return (linux_recvfrom(td, arg));
2172 case LINUX_SHUTDOWN:
2173 return (linux_shutdown(td, arg));
2174 case LINUX_SETSOCKOPT:
2175 return (linux_setsockopt(td, arg));
2176 case LINUX_GETSOCKOPT:
2177 return (linux_getsockopt(td, arg));
2179 return (linux_sendmsg(td, arg));
2181 return (linux_recvmsg(td, arg));
2183 return (linux_accept4(td, arg));
2184 case LINUX_RECVMMSG:
2185 return (linux_recvmmsg(td, arg));
2186 case LINUX_SENDMMSG:
2187 return (linux_sendmmsg(td, arg));
2188 case LINUX_SENDFILE:
2189 return (linux_sendfile(td, arg));
2192 linux_msg(td, "socket type %d not implemented", args->what);
2195 #endif /* __i386__ || __arm__ || (__amd64__ && COMPAT_LINUX32) */