1/*-
2 * Copyright (c) 1995 S��ren Schmidt
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: head/sys/compat/linux/linux_socket.c 283413 2015-05-24 15:41:27Z dchagin $");
31
32/* XXX we use functions that might not exist. */
33#include "opt_compat.h"
34#include "opt_inet6.h"
35
36#include <sys/param.h>
37#include <sys/proc.h>
38#include <sys/systm.h>
39#include <sys/sysproto.h>
40#include <sys/capsicum.h>
41#include <sys/fcntl.h>
42#include <sys/file.h>
43#include <sys/limits.h>
44#include <sys/lock.h>
45#include <sys/malloc.h>
46#include <sys/mutex.h>
47#include <sys/mbuf.h>
48#include <sys/socket.h>
49#include <sys/socketvar.h>
50#include <sys/syscallsubr.h>
51#include <sys/uio.h>
52#include <sys/syslog.h>
53#include <sys/un.h>
54
55#include <net/if.h>
56#include <net/vnet.h>
57#include <netinet/in.h>
58#include <netinet/in_systm.h>
59#include <netinet/ip.h>
60#include <netinet/tcp.h>
61#ifdef INET6
62#include <netinet/ip6.h>
63#include <netinet6/ip6_var.h>
64#endif
65
66#ifdef COMPAT_LINUX32
67#include <machine/../linux32/linux.h>
68#include <machine/../linux32/linux32_proto.h>
69#else
70#include <machine/../linux/linux.h>
71#include <machine/../linux/linux_proto.h>
72#endif
73#include <compat/linux/linux_socket.h>
74#include <compat/linux/linux_util.h>
75
76static int linux_to_bsd_domain(int);
77
78/*
79 * Reads a linux sockaddr and does any necessary translation.
80 * Linux sockaddrs don't have a length field, only a family.
81 * Copy the osockaddr structure pointed to by osa to kernel, adjust
82 * family and convert to sockaddr.
83 */
84static int
85linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
86{
87 struct sockaddr *sa;
88 struct osockaddr *kosa;
89#ifdef INET6
90 struct sockaddr_in6 *sin6;
91 int oldv6size;
92#endif
93 char *name;
94 int bdom, error, hdrlen, namelen;
95
96 if (salen < 2 || salen > UCHAR_MAX || !osa)
97 return (EINVAL);
98
99#ifdef INET6
100 oldv6size = 0;
101 /*
102 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
103 * if it's a v4-mapped address, so reserve the proper space
104 * for it.
105 */
106 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
107 salen += sizeof(uint32_t);
108 oldv6size = 1;
109 }
110#endif
111
112 kosa = malloc(salen, M_SONAME, M_WAITOK);
113
114 if ((error = copyin(osa, kosa, salen)))
115 goto out;
116
117 bdom = linux_to_bsd_domain(kosa->sa_family);
118 if (bdom == -1) {
119 error = EAFNOSUPPORT;
120 goto out;
121 }
122
123#ifdef INET6
124 /*
125 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
126 * which lacks the scope id compared with RFC2553 one. If we detect
127 * the situation, reject the address and write a message to system log.
128 *
129 * Still accept addresses for which the scope id is not used.
130 */
131 if (oldv6size) {
132 if (bdom == AF_INET6) {
133 sin6 = (struct sockaddr_in6 *)kosa;
134 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
135 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
136 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
137 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
138 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
139 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
140 sin6->sin6_scope_id = 0;
141 } else {
142 log(LOG_DEBUG,
143 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
144 error = EINVAL;
145 goto out;
146 }
147 } else
148 salen -= sizeof(uint32_t);
149 }
150#endif
151 if (bdom == AF_INET) {
152 if (salen < sizeof(struct sockaddr_in)) {
153 error = EINVAL;
154 goto out;
155 }
156 salen = sizeof(struct sockaddr_in);
157 }
158
159 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
160 hdrlen = offsetof(struct sockaddr_un, sun_path);
161 name = ((struct sockaddr_un *)kosa)->sun_path;
162 if (*name == '\0') {
163 /*
164 * Linux abstract namespace starts with a NULL byte.
165 * XXX We do not support abstract namespace yet.
166 */
167 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
168 } else
169 namelen = strnlen(name, salen - hdrlen);
170 salen = hdrlen + namelen;
171 if (salen > sizeof(struct sockaddr_un)) {
172 error = ENAMETOOLONG;
173 goto out;
174 }
175 }
176
177 sa = (struct sockaddr *)kosa;
178 sa->sa_family = bdom;
179 sa->sa_len = salen;
180
181 *sap = sa;
182 return (0);
183
184out:
185 free(kosa, M_SONAME);
186 return (error);
187}
188
189static int
190linux_to_bsd_domain(int domain)
191{
192
193 switch (domain) {
194 case LINUX_AF_UNSPEC:
195 return (AF_UNSPEC);
196 case LINUX_AF_UNIX:
197 return (AF_LOCAL);
198 case LINUX_AF_INET:
199 return (AF_INET);
200 case LINUX_AF_INET6:
201 return (AF_INET6);
202 case LINUX_AF_AX25:
203 return (AF_CCITT);
204 case LINUX_AF_IPX:
205 return (AF_IPX);
206 case LINUX_AF_APPLETALK:
207 return (AF_APPLETALK);
208 }
209 return (-1);
210}
211
212static int
213bsd_to_linux_domain(int domain)
214{
215
216 switch (domain) {
217 case AF_UNSPEC:
218 return (LINUX_AF_UNSPEC);
219 case AF_LOCAL:
220 return (LINUX_AF_UNIX);
221 case AF_INET:
222 return (LINUX_AF_INET);
223 case AF_INET6:
224 return (LINUX_AF_INET6);
225 case AF_CCITT:
226 return (LINUX_AF_AX25);
227 case AF_IPX:
228 return (LINUX_AF_IPX);
229 case AF_APPLETALK:
230 return (LINUX_AF_APPLETALK);
231 }
232 return (-1);
233}
234
235static int
236linux_to_bsd_sockopt_level(int level)
237{
238
239 switch (level) {
240 case LINUX_SOL_SOCKET:
241 return (SOL_SOCKET);
242 }
243 return (level);
244}
245
246static int
247bsd_to_linux_sockopt_level(int level)
248{
249
250 switch (level) {
251 case SOL_SOCKET:
252 return (LINUX_SOL_SOCKET);
253 }
254 return (level);
255}
256
257static int
258linux_to_bsd_ip_sockopt(int opt)
259{
260
261 switch (opt) {
262 case LINUX_IP_TOS:
263 return (IP_TOS);
264 case LINUX_IP_TTL:
265 return (IP_TTL);
266 case LINUX_IP_OPTIONS:
267 return (IP_OPTIONS);
268 case LINUX_IP_MULTICAST_IF:
269 return (IP_MULTICAST_IF);
270 case LINUX_IP_MULTICAST_TTL:
271 return (IP_MULTICAST_TTL);
272 case LINUX_IP_MULTICAST_LOOP:
273 return (IP_MULTICAST_LOOP);
274 case LINUX_IP_ADD_MEMBERSHIP:
275 return (IP_ADD_MEMBERSHIP);
276 case LINUX_IP_DROP_MEMBERSHIP:
277 return (IP_DROP_MEMBERSHIP);
278 case LINUX_IP_HDRINCL:
279 return (IP_HDRINCL);
280 }
281 return (-1);
282}
283
284static int
285linux_to_bsd_so_sockopt(int opt)
286{
287
288 switch (opt) {
289 case LINUX_SO_DEBUG:
290 return (SO_DEBUG);
291 case LINUX_SO_REUSEADDR:
292 return (SO_REUSEADDR);
293 case LINUX_SO_TYPE:
294 return (SO_TYPE);
295 case LINUX_SO_ERROR:
296 return (SO_ERROR);
297 case LINUX_SO_DONTROUTE:
298 return (SO_DONTROUTE);
299 case LINUX_SO_BROADCAST:
300 return (SO_BROADCAST);
301 case LINUX_SO_SNDBUF:
302 return (SO_SNDBUF);
303 case LINUX_SO_RCVBUF:
304 return (SO_RCVBUF);
305 case LINUX_SO_KEEPALIVE:
306 return (SO_KEEPALIVE);
307 case LINUX_SO_OOBINLINE:
308 return (SO_OOBINLINE);
309 case LINUX_SO_LINGER:
310 return (SO_LINGER);
311 case LINUX_SO_PEERCRED:
312 return (LOCAL_PEERCRED);
313 case LINUX_SO_RCVLOWAT:
314 return (SO_RCVLOWAT);
315 case LINUX_SO_SNDLOWAT:
316 return (SO_SNDLOWAT);
317 case LINUX_SO_RCVTIMEO:
318 return (SO_RCVTIMEO);
319 case LINUX_SO_SNDTIMEO:
320 return (SO_SNDTIMEO);
321 case LINUX_SO_TIMESTAMP:
322 return (SO_TIMESTAMP);
323 case LINUX_SO_ACCEPTCONN:
324 return (SO_ACCEPTCONN);
325 }
326 return (-1);
327}
328
329static int
330linux_to_bsd_tcp_sockopt(int opt)
331{
332
333 switch (opt) {
334 case LINUX_TCP_NODELAY:
335 return (TCP_NODELAY);
336 case LINUX_TCP_MAXSEG:
337 return (TCP_MAXSEG);
338 case LINUX_TCP_KEEPIDLE:
339 return (TCP_KEEPIDLE);
340 case LINUX_TCP_KEEPINTVL:
341 return (TCP_KEEPINTVL);
342 case LINUX_TCP_KEEPCNT:
343 return (TCP_KEEPCNT);
344 case LINUX_TCP_MD5SIG:
345 return (TCP_MD5SIG);
346 }
347 return (-1);
348}
349
350static int
351linux_to_bsd_msg_flags(int flags)
352{
353 int ret_flags = 0;
354
355 if (flags & LINUX_MSG_OOB)
356 ret_flags |= MSG_OOB;
357 if (flags & LINUX_MSG_PEEK)
358 ret_flags |= MSG_PEEK;
359 if (flags & LINUX_MSG_DONTROUTE)
360 ret_flags |= MSG_DONTROUTE;
361 if (flags & LINUX_MSG_CTRUNC)
362 ret_flags |= MSG_CTRUNC;
363 if (flags & LINUX_MSG_TRUNC)
364 ret_flags |= MSG_TRUNC;
365 if (flags & LINUX_MSG_DONTWAIT)
366 ret_flags |= MSG_DONTWAIT;
367 if (flags & LINUX_MSG_EOR)
368 ret_flags |= MSG_EOR;
369 if (flags & LINUX_MSG_WAITALL)
370 ret_flags |= MSG_WAITALL;
371 if (flags & LINUX_MSG_NOSIGNAL)
372 ret_flags |= MSG_NOSIGNAL;
373#if 0 /* not handled */
374 if (flags & LINUX_MSG_PROXY)
375 ;
376 if (flags & LINUX_MSG_FIN)
377 ;
378 if (flags & LINUX_MSG_SYN)
379 ;
380 if (flags & LINUX_MSG_CONFIRM)
381 ;
382 if (flags & LINUX_MSG_RST)
383 ;
384 if (flags & LINUX_MSG_ERRQUEUE)
385 ;
386#endif
387 return ret_flags;
388}
389
390/*
391* If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
392* native syscall will fault. Thus, we don't really need to check the
393* return values for these functions.
394*/
395
396static int
397bsd_to_linux_sockaddr(struct sockaddr *arg)
398{
399 struct sockaddr sa;
400 size_t sa_len = sizeof(struct sockaddr);
401 int error;
402
403 if ((error = copyin(arg, &sa, sa_len)))
404 return (error);
405
406 *(u_short *)&sa = sa.sa_family;
407
408 error = copyout(&sa, arg, sa_len);
409
410 return (error);
411}
412
413static int
414linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
415{
416 struct sockaddr sa;
417 size_t sa_len = sizeof(struct sockaddr);
418 int error;
419
420 if ((error = copyin(arg, &sa, sa_len)))
421 return (error);
422
423 sa.sa_family = *(sa_family_t *)&sa;
424 sa.sa_len = len;
425
426 error = copyout(&sa, arg, sa_len);
427
428 return (error);
429}
430
431
432static int
433linux_sa_put(struct osockaddr *osa)
434{
435 struct osockaddr sa;
436 int error, bdom;
437
438 /*
439 * Only read/write the osockaddr family part, the rest is
440 * not changed.
441 */
442 error = copyin(osa, &sa, sizeof(sa.sa_family));
443 if (error)
444 return (error);
445
446 bdom = bsd_to_linux_domain(sa.sa_family);
447 if (bdom == -1)
448 return (EINVAL);
449
450 sa.sa_family = bdom;
451 error = copyout(&sa, osa, sizeof(sa.sa_family));
452 if (error)
453 return (error);
454
455 return (0);
456}
457
458static int
459linux_to_bsd_cmsg_type(int cmsg_type)
460{
461
462 switch (cmsg_type) {
463 case LINUX_SCM_RIGHTS:
464 return (SCM_RIGHTS);
465 case LINUX_SCM_CREDENTIALS:
466 return (SCM_CREDS);
467 }
468 return (-1);
469}
470
471static int
472bsd_to_linux_cmsg_type(int cmsg_type)
473{
474
475 switch (cmsg_type) {
476 case SCM_RIGHTS:
477 return (LINUX_SCM_RIGHTS);
478 case SCM_CREDS:
479 return (LINUX_SCM_CREDENTIALS);
480 }
481 return (-1);
482}
483
484static int
485linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
486{
487 if (lhdr->msg_controllen > INT_MAX)
488 return (ENOBUFS);
489
490 bhdr->msg_name = PTRIN(lhdr->msg_name);
491 bhdr->msg_namelen = lhdr->msg_namelen;
492 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
493 bhdr->msg_iovlen = lhdr->msg_iovlen;
494 bhdr->msg_control = PTRIN(lhdr->msg_control);
495
496 /*
497 * msg_controllen is skipped since BSD and LINUX control messages
498 * are potentially different sizes (e.g. the cred structure used
499 * by SCM_CREDS is different between the two operating system).
500 *
501 * The caller can set it (if necessary) after converting all the
502 * control messages.
503 */
504
505 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
506 return (0);
507}
508
509static int
510bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
511{
512 lhdr->msg_name = PTROUT(bhdr->msg_name);
513 lhdr->msg_namelen = bhdr->msg_namelen;
514 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
515 lhdr->msg_iovlen = bhdr->msg_iovlen;
516 lhdr->msg_control = PTROUT(bhdr->msg_control);
517
518 /*
519 * msg_controllen is skipped since BSD and LINUX control messages
520 * are potentially different sizes (e.g. the cred structure used
521 * by SCM_CREDS is different between the two operating system).
522 *
523 * The caller can set it (if necessary) after converting all the
524 * control messages.
525 */
526
527 /* msg_flags skipped */
528 return (0);
529}
530
531static int
532linux_set_socket_flags(struct thread *td, int s, int flags)
533{
534 int error;
535
536 if (flags & LINUX_SOCK_NONBLOCK) {
537 error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
538 if (error)
539 return (error);
540 }
541 if (flags & LINUX_SOCK_CLOEXEC) {
542 error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
543 if (error)
544 return (error);
545 }
546 return (0);
547}
548
549static int
550linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
551 struct mbuf *control, enum uio_seg segflg)
552{
553 struct sockaddr *to;
554 int error;
555
556 if (mp->msg_name != NULL) {
557 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
558 if (error)
559 return (error);
560 mp->msg_name = to;
561 } else
562 to = NULL;
563
564 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
565 segflg);
566
567 if (to)
568 free(to, M_SONAME);
569 return (error);
570}
571
572/* Return 0 if IP_HDRINCL is set for the given socket. */
573static int
574linux_check_hdrincl(struct thread *td, int s)
575{
576 int error, optval;
577 socklen_t size_val;
578
579 size_val = sizeof(optval);
580 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
581 &optval, UIO_SYSSPACE, &size_val);
582 if (error)
583 return (error);
584
585 return (optval == 0);
586}
587
588/*
589 * Updated sendto() when IP_HDRINCL is set:
590 * tweak endian-dependent fields in the IP packet.
591 */
592static int
593linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
594{
595/*
596 * linux_ip_copysize defines how many bytes we should copy
597 * from the beginning of the IP packet before we customize it for BSD.
598 * It should include all the fields we modify (ip_len and ip_off).
599 */
600#define linux_ip_copysize 8
601
602 struct ip *packet;
603 struct msghdr msg;
604 struct iovec aiov[1];
605 int error;
606
607 /* Check that the packet isn't too big or too small. */
608 if (linux_args->len < linux_ip_copysize ||
609 linux_args->len > IP_MAXPACKET)
610 return (EINVAL);
611
612 packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
613
614 /* Make kernel copy of the packet to be sent */
615 if ((error = copyin(PTRIN(linux_args->msg), packet,
616 linux_args->len)))
617 goto goout;
618
619 /* Convert fields from Linux to BSD raw IP socket format */
620 packet->ip_len = linux_args->len;
621 packet->ip_off = ntohs(packet->ip_off);
622
623 /* Prepare the msghdr and iovec structures describing the new packet */
624 msg.msg_name = PTRIN(linux_args->to);
625 msg.msg_namelen = linux_args->tolen;
626 msg.msg_iov = aiov;
627 msg.msg_iovlen = 1;
628 msg.msg_control = NULL;
629 msg.msg_flags = 0;
630 aiov[0].iov_base = (char *)packet;
631 aiov[0].iov_len = linux_args->len;
632 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
633 NULL, UIO_SYSSPACE);
634goout:
635 free(packet, M_TEMP);
636 return (error);
637}
638
639int
640linux_socket(struct thread *td, struct linux_socket_args *args)
641{
642 struct socket_args /* {
643 int domain;
644 int type;
645 int protocol;
646 } */ bsd_args;
647 int retval_socket, socket_flags;
648
649 bsd_args.protocol = args->protocol;
650 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
651 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
652 return (EINVAL);
653 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
654 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
655 return (EINVAL);
656 bsd_args.domain = linux_to_bsd_domain(args->domain);
657 if (bsd_args.domain == -1)
658 return (EAFNOSUPPORT);
659
660 retval_socket = sys_socket(td, &bsd_args);
661 if (retval_socket)
662 return (retval_socket);
663
664 retval_socket = linux_set_socket_flags(td, td->td_retval[0],
665 socket_flags);
666 if (retval_socket) {
667 (void)kern_close(td, td->td_retval[0]);
668 goto out;
669 }
670
671 if (bsd_args.type == SOCK_RAW
672 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
673 && bsd_args.domain == PF_INET) {
674 /* It's a raw IP socket: set the IP_HDRINCL option. */
675 int hdrincl;
676
677 hdrincl = 1;
678 /* We ignore any error returned by kern_setsockopt() */
679 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
680 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
681 }
682#ifdef INET6
683 /*
684 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
685 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
686 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
687 * sysctl value.
688 */
689 if (bsd_args.domain == PF_INET6) {
690 int v6only;
691
692 v6only = 0;
693 /* We ignore any error returned by setsockopt() */
694 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
695 &v6only, UIO_SYSSPACE, sizeof(v6only));
696 }
697#endif
698
699out:
700 return (retval_socket);
701}
702
703int
704linux_bind(struct thread *td, struct linux_bind_args *args)
705{
706 struct sockaddr *sa;
707 int error;
708
709 error = linux_getsockaddr(&sa, PTRIN(args->name),
710 args->namelen);
711 if (error)
712 return (error);
713
714 error = kern_bindat(td, AT_FDCWD, args->s, sa);
715 free(sa, M_SONAME);
716 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
717 return (EINVAL);
718 return (error);
719}
720
721int
722linux_connect(struct thread *td, struct linux_connect_args *args)
723{
724 cap_rights_t rights;
725 struct socket *so;
726 struct sockaddr *sa;
727 u_int fflag;
728 int error;
729
730 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
731 args->namelen);
732 if (error)
733 return (error);
734
735 error = kern_connectat(td, AT_FDCWD, args->s, sa);
736 free(sa, M_SONAME);
737 if (error != EISCONN)
738 return (error);
739
740 /*
741 * Linux doesn't return EISCONN the first time it occurs,
742 * when on a non-blocking socket. Instead it returns the
743 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
744 *
745 * XXXRW: Instead of using fgetsock(), check that it is a
746 * socket and use the file descriptor reference instead of
747 * creating a new one.
748 */
749 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
750 &so, &fflag);
751 if (error == 0) {
752 error = EISCONN;
753 if (fflag & FNONBLOCK) {
754 SOCK_LOCK(so);
755 if (so->so_emuldata == 0)
756 error = so->so_error;
757 so->so_emuldata = (void *)1;
758 SOCK_UNLOCK(so);
759 }
760 fputsock(so);
761 }
762 return (error);
763}
764
765int
766linux_listen(struct thread *td, struct linux_listen_args *args)
767{
768 struct listen_args /* {
769 int s;
770 int backlog;
771 } */ bsd_args;
772
773 bsd_args.s = args->s;
774 bsd_args.backlog = args->backlog;
775 return (sys_listen(td, &bsd_args));
776}
777
778static int
779linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
780 l_uintptr_t namelen, int flags)
781{
782 struct accept_args /* {
783 int s;
784 struct sockaddr * __restrict name;
785 socklen_t * __restrict anamelen;
786 } */ bsd_args;
787 int error;
788
789 if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
790 return (EINVAL);
791
792 bsd_args.s = s;
793 /* XXX: */
794 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
795 bsd_args.anamelen = PTRIN(namelen);/* XXX */
796 error = sys_accept(td, &bsd_args);
797 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
798 if (error) {
799 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
800 return (EINVAL);
801 return (error);
802 }
803
804 /*
805 * linux appears not to copy flags from the parent socket to the
806 * accepted one, so we must clear the flags in the new descriptor
807 * and apply the requested flags.
808 */
809 error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
810 if (error)
811 goto out;
812 error = linux_set_socket_flags(td, td->td_retval[0], flags);
813 if (error)
814 goto out;
815 if (addr)
816 error = linux_sa_put(PTRIN(addr));
817
818out:
819 if (error) {
820 (void)kern_close(td, td->td_retval[0]);
821 td->td_retval[0] = 0;
822 }
823 return (error);
824}
825
826int
827linux_accept(struct thread *td, struct linux_accept_args *args)
828{
829
830 return (linux_accept_common(td, args->s, args->addr,
831 args->namelen, 0));
832}
833
834int
835linux_accept4(struct thread *td, struct linux_accept4_args *args)
836{
837
838 return (linux_accept_common(td, args->s, args->addr,
839 args->namelen, args->flags));
840}
841
842int
843linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
844{
845 struct getsockname_args /* {
846 int fdes;
847 struct sockaddr * __restrict asa;
848 socklen_t * __restrict alen;
849 } */ bsd_args;
850 int error;
851
852 bsd_args.fdes = args->s;
853 /* XXX: */
854 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
855 bsd_args.alen = PTRIN(args->namelen); /* XXX */
856 error = sys_getsockname(td, &bsd_args);
857 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
858 if (error)
859 return (error);
860 error = linux_sa_put(PTRIN(args->addr));
861 if (error)
862 return (error);
863 return (0);
864}
865
866int
867linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
868{
869 struct getpeername_args /* {
870 int fdes;
871 caddr_t asa;
872 int *alen;
873 } */ bsd_args;
874 int error;
875
876 bsd_args.fdes = args->s;
877 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
878 bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
879 error = sys_getpeername(td, &bsd_args);
880 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
881 if (error)
882 return (error);
883 error = linux_sa_put(PTRIN(args->addr));
884 if (error)
885 return (error);
886 return (0);
887}
888
889int
890linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
891{
892 struct socketpair_args /* {
893 int domain;
894 int type;
895 int protocol;
896 int *rsv;
897 } */ bsd_args;
898 int error, socket_flags;
899 int sv[2];
900
901 bsd_args.domain = linux_to_bsd_domain(args->domain);
902 if (bsd_args.domain != PF_LOCAL)
903 return (EAFNOSUPPORT);
904
905 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
906 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
907 return (EINVAL);
908 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
909 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
910 return (EINVAL);
911
912 if (args->protocol != 0 && args->protocol != PF_UNIX)
913
914 /*
915 * Use of PF_UNIX as protocol argument is not right,
916 * but Linux does it.
917 * Do not map PF_UNIX as its Linux value is identical
918 * to FreeBSD one.
919 */
920 return (EPROTONOSUPPORT);
921 else
922 bsd_args.protocol = 0;
923 bsd_args.rsv = (int *)PTRIN(args->rsv);
924 error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
925 bsd_args.protocol, sv);
926 if (error)
927 return (error);
928 error = linux_set_socket_flags(td, sv[0], socket_flags);
929 if (error)
930 goto out;
931 error = linux_set_socket_flags(td, sv[1], socket_flags);
932 if (error)
933 goto out;
934
935 error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
936
937out:
938 if (error) {
939 (void)kern_close(td, sv[0]);
940 (void)kern_close(td, sv[1]);
941 }
942 return (error);
943}
944
945#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
946struct linux_send_args {
947 int s;
948 l_uintptr_t msg;
949 int len;
950 int flags;
951};
952
953static int
954linux_send(struct thread *td, struct linux_send_args *args)
955{
956 struct sendto_args /* {
957 int s;
958 caddr_t buf;
959 int len;
960 int flags;
961 caddr_t to;
962 int tolen;
963 } */ bsd_args;
964
965 bsd_args.s = args->s;
966 bsd_args.buf = (caddr_t)PTRIN(args->msg);
967 bsd_args.len = args->len;
968 bsd_args.flags = args->flags;
969 bsd_args.to = NULL;
970 bsd_args.tolen = 0;
971 return sys_sendto(td, &bsd_args);
972}
973
974struct linux_recv_args {
975 int s;
976 l_uintptr_t msg;
977 int len;
978 int flags;
979};
980
981static int
982linux_recv(struct thread *td, struct linux_recv_args *args)
983{
984 struct recvfrom_args /* {
985 int s;
986 caddr_t buf;
987 int len;
988 int flags;
989 struct sockaddr *from;
990 socklen_t fromlenaddr;
991 } */ bsd_args;
992
993 bsd_args.s = args->s;
994 bsd_args.buf = (caddr_t)PTRIN(args->msg);
995 bsd_args.len = args->len;
996 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
997 bsd_args.from = NULL;
998 bsd_args.fromlenaddr = 0;
999 return (sys_recvfrom(td, &bsd_args));
1000}
1001#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1002
1003int
1004linux_sendto(struct thread *td, struct linux_sendto_args *args)
1005{
1006 struct msghdr msg;
1007 struct iovec aiov;
1008 int error;
1009
1010 if (linux_check_hdrincl(td, args->s) == 0)
1011 /* IP_HDRINCL set, tweak the packet before sending */
1012 return (linux_sendto_hdrincl(td, args));
1013
1014 msg.msg_name = PTRIN(args->to);
1015 msg.msg_namelen = args->tolen;
1016 msg.msg_iov = &aiov;
1017 msg.msg_iovlen = 1;
1018 msg.msg_control = NULL;
1019 msg.msg_flags = 0;
1020 aiov.iov_base = PTRIN(args->msg);
1021 aiov.iov_len = args->len;
1022 error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1023 UIO_USERSPACE);
1024 return (error);
1025}
1026
1027int
1028linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1029{
1030 struct recvfrom_args /* {
1031 int s;
1032 caddr_t buf;
1033 size_t len;
1034 int flags;
1035 struct sockaddr * __restrict from;
1036 socklen_t * __restrict fromlenaddr;
1037 } */ bsd_args;
1038 size_t len;
1039 int error;
1040
1041 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1042 return (error);
1043
1044 bsd_args.s = args->s;
1045 bsd_args.buf = PTRIN(args->buf);
1046 bsd_args.len = args->len;
1047 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1048 /* XXX: */
1049 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1050 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1051
1052 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1053 error = sys_recvfrom(td, &bsd_args);
1054 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1055
1056 if (error)
1057 return (error);
1058 if (args->from) {
1059 error = linux_sa_put((struct osockaddr *)
1060 PTRIN(args->from));
1061 if (error)
1062 return (error);
1063 }
1064 return (0);
1065}
1066
1067int
1068linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1069{
1070 struct cmsghdr *cmsg;
1071 struct cmsgcred cmcred;
1072 struct mbuf *control;
1073 struct msghdr msg;
1074 struct l_cmsghdr linux_cmsg;
1075 struct l_cmsghdr *ptr_cmsg;
1076 struct l_msghdr linux_msg;
1077 struct iovec *iov;
1078 socklen_t datalen;
1079 struct sockaddr *sa;
1080 sa_family_t sa_family;
1081 void *data;
1082 int error;
1083
1084 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1085 if (error)
1086 return (error);
1087
1088 /*
1089 * Some Linux applications (ping) define a non-NULL control data
1090 * pointer, but a msg_controllen of 0, which is not allowed in the
1091 * FreeBSD system call interface. NULL the msg_control pointer in
1092 * order to handle this case. This should be checked, but allows the
1093 * Linux ping to work.
1094 */
1095 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1096 linux_msg.msg_control = PTROUT(NULL);
1097
1098 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1099 if (error)
1100 return (error);
1101
1102#ifdef COMPAT_LINUX32
1103 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1104 &iov, EMSGSIZE);
1105#else
1106 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1107#endif
1108 if (error)
1109 return (error);
1110
1111 control = NULL;
1112 cmsg = NULL;
1113
1114 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1115 error = kern_getsockname(td, args->s, &sa, &datalen);
1116 if (error)
1117 goto bad;
1118 sa_family = sa->sa_family;
1119 free(sa, M_SONAME);
1120
1121 error = ENOBUFS;
1122 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1123 control = m_get(M_WAITOK, MT_CONTROL);
1124 if (control == NULL)
1125 goto bad;
1126
1127 do {
1128 error = copyin(ptr_cmsg, &linux_cmsg,
1129 sizeof(struct l_cmsghdr));
1130 if (error)
1131 goto bad;
1132
1133 error = EINVAL;
1134 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1135 goto bad;
1136
1137 /*
1138 * Now we support only SCM_RIGHTS and SCM_CRED,
1139 * so return EINVAL in any other cmsg_type
1140 */
1141 cmsg->cmsg_type =
1142 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1143 cmsg->cmsg_level =
1144 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1145 if (cmsg->cmsg_type == -1
1146 || cmsg->cmsg_level != SOL_SOCKET)
1147 goto bad;
1148
1149 /*
1150 * Some applications (e.g. pulseaudio) attempt to
1151 * send ancillary data even if the underlying protocol
1152 * doesn't support it which is not allowed in the
1153 * FreeBSD system call interface.
1154 */
1155 if (sa_family != AF_UNIX)
1156 continue;
1157
1158 data = LINUX_CMSG_DATA(ptr_cmsg);
1159 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1160
1161 switch (cmsg->cmsg_type)
1162 {
1163 case SCM_RIGHTS:
1164 break;
1165
1166 case SCM_CREDS:
1167 data = &cmcred;
1168 datalen = sizeof(cmcred);
1169
1170 /*
1171 * The lower levels will fill in the structure
1172 */
1173 bzero(data, datalen);
1174 break;
1175 }
1176
1177 cmsg->cmsg_len = CMSG_LEN(datalen);
1178
1179 error = ENOBUFS;
1180 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1181 goto bad;
1182 if (!m_append(control, datalen, (c_caddr_t)data))
1183 goto bad;
1184 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1185
1186 if (m_length(control, NULL) == 0) {
1187 m_freem(control);
1188 control = NULL;
1189 }
1190 }
1191
1192 msg.msg_iov = iov;
1193 msg.msg_flags = 0;
1194 error = linux_sendit(td, args->s, &msg, args->flags, control,
1195 UIO_USERSPACE);
1196
1197bad:
1198 free(iov, M_IOV);
1199 if (cmsg)
1200 free(cmsg, M_TEMP);
1201 return (error);
1202}
1203
1204int
1205linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1206{
1207 struct cmsghdr *cm;
1208 struct cmsgcred *cmcred;
1209 struct msghdr msg;
1210 struct l_cmsghdr *linux_cmsg = NULL;
1211 struct l_ucred linux_ucred;
1212 socklen_t datalen, outlen;
1213 struct l_msghdr linux_msg;
1214 struct iovec *iov, *uiov;
1215 struct mbuf *control = NULL;
1216 struct mbuf **controlp;
1217 caddr_t outbuf;
1218 void *data;
1219 int error, i, fd, fds, *fdp;
1220
1221 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1222 if (error)
1223 return (error);
1224
1225 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1226 if (error)
1227 return (error);
1228
1229#ifdef COMPAT_LINUX32
1230 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1231 &iov, EMSGSIZE);
1232#else
1233 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1234#endif
1235 if (error)
1236 return (error);
1237
1238 if (msg.msg_name) {
1239 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1240 msg.msg_namelen);
1241 if (error)
1242 goto bad;
1243 }
1244
1245 uiov = msg.msg_iov;
1246 msg.msg_iov = iov;
1247 controlp = (msg.msg_control != NULL) ? &control : NULL;
1248 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1249 msg.msg_iov = uiov;
1250 if (error)
1251 goto bad;
1252
1253 error = bsd_to_linux_msghdr(&msg, &linux_msg);
1254 if (error)
1255 goto bad;
1256
1257 if (linux_msg.msg_name) {
1258 error = bsd_to_linux_sockaddr((struct sockaddr *)
1259 PTRIN(linux_msg.msg_name));
1260 if (error)
1261 goto bad;
1262 }
1263 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1264 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1265 if (error)
1266 goto bad;
1267 }
1268
1269 outbuf = PTRIN(linux_msg.msg_control);
1270 outlen = 0;
1271
1272 if (control) {
1273 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1274
1275 msg.msg_control = mtod(control, struct cmsghdr *);
1276 msg.msg_controllen = control->m_len;
1277
1278 cm = CMSG_FIRSTHDR(&msg);
1279
1280 while (cm != NULL) {
1281 linux_cmsg->cmsg_type =
1282 bsd_to_linux_cmsg_type(cm->cmsg_type);
1283 linux_cmsg->cmsg_level =
1284 bsd_to_linux_sockopt_level(cm->cmsg_level);
1285 if (linux_cmsg->cmsg_type == -1
1286 || cm->cmsg_level != SOL_SOCKET)
1287 {
1288 error = EINVAL;
1289 goto bad;
1290 }
1291
1292 data = CMSG_DATA(cm);
1293 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1294
1295 switch (cm->cmsg_type)
1296 {
1297 case SCM_RIGHTS:
1298 if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1299 fds = datalen / sizeof(int);
1300 fdp = data;
1301 for (i = 0; i < fds; i++) {
1302 fd = *fdp++;
1303 (void)kern_fcntl(td, fd,
1304 F_SETFD, FD_CLOEXEC);
1305 }
1306 }
1307 break;
1308
1309 case SCM_CREDS:
1310 /*
1311 * Currently LOCAL_CREDS is never in
1312 * effect for Linux so no need to worry
1313 * about sockcred
1314 */
1315 if (datalen != sizeof(*cmcred)) {
1316 error = EMSGSIZE;
1317 goto bad;
1318 }
1319 cmcred = (struct cmsgcred *)data;
1320 bzero(&linux_ucred, sizeof(linux_ucred));
1321 linux_ucred.pid = cmcred->cmcred_pid;
1322 linux_ucred.uid = cmcred->cmcred_uid;
1323 linux_ucred.gid = cmcred->cmcred_gid;
1324 data = &linux_ucred;
1325 datalen = sizeof(linux_ucred);
1326 break;
1327 }
1328
1329 if (outlen + LINUX_CMSG_LEN(datalen) >
1330 linux_msg.msg_controllen) {
1331 if (outlen == 0) {
1332 error = EMSGSIZE;
1333 goto bad;
1334 } else {
1335 linux_msg.msg_flags |=
1336 LINUX_MSG_CTRUNC;
1337 goto out;
1338 }
1339 }
1340
1341 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1342
1343 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1344 if (error)
1345 goto bad;
1346 outbuf += L_CMSG_HDRSZ;
1347
1348 error = copyout(data, outbuf, datalen);
1349 if (error)
1350 goto bad;
1351
1352 outbuf += LINUX_CMSG_ALIGN(datalen);
1353 outlen += LINUX_CMSG_LEN(datalen);
1354
1355 cm = CMSG_NXTHDR(&msg, cm);
1356 }
1357 }
1358
1359out:
1360 linux_msg.msg_controllen = outlen;
1361 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1362
1363bad:
1364 free(iov, M_IOV);
1365 m_freem(control);
1366 free(linux_cmsg, M_TEMP);
1367
1368 return (error);
1369}
1370
1371int
1372linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1373{
1374 struct shutdown_args /* {
1375 int s;
1376 int how;
1377 } */ bsd_args;
1378
1379 bsd_args.s = args->s;
1380 bsd_args.how = args->how;
1381 return (sys_shutdown(td, &bsd_args));
1382}
1383
1384int
1385linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1386{
1387 struct setsockopt_args /* {
1388 int s;
1389 int level;
1390 int name;
1391 caddr_t val;
1392 int valsize;
1393 } */ bsd_args;
1394 l_timeval linux_tv;
1395 struct timeval tv;
1396 int error, name;
1397
1398 bsd_args.s = args->s;
1399 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1400 switch (bsd_args.level) {
1401 case SOL_SOCKET:
1402 name = linux_to_bsd_so_sockopt(args->optname);
1403 switch (name) {
1404 case SO_RCVTIMEO:
1405 /* FALLTHROUGH */
1406 case SO_SNDTIMEO:
1407 error = copyin(PTRIN(args->optval), &linux_tv,
1408 sizeof(linux_tv));
1409 if (error)
1410 return (error);
1411 tv.tv_sec = linux_tv.tv_sec;
1412 tv.tv_usec = linux_tv.tv_usec;
1413 return (kern_setsockopt(td, args->s, bsd_args.level,
1414 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1415 /* NOTREACHED */
1416 break;
1417 default:
1418 break;
1419 }
1420 break;
1421 case IPPROTO_IP:
1422 name = linux_to_bsd_ip_sockopt(args->optname);
1423 break;
1424 case IPPROTO_TCP:
1425 name = linux_to_bsd_tcp_sockopt(args->optname);
1426 break;
1427 default:
1428 name = -1;
1429 break;
1430 }
1431 if (name == -1)
1432 return (ENOPROTOOPT);
1433
1434 bsd_args.name = name;
1435 bsd_args.val = PTRIN(args->optval);
1436 bsd_args.valsize = args->optlen;
1437
1438 if (name == IPV6_NEXTHOP) {
1439 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1440 bsd_args.valsize);
1441 error = sys_setsockopt(td, &bsd_args);
1442 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1443 } else
1444 error = sys_setsockopt(td, &bsd_args);
1445
1446 return (error);
1447}
1448
1449int
1450linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1451{
1452 struct getsockopt_args /* {
1453 int s;
1454 int level;
1455 int name;
1456 caddr_t val;
1457 int *avalsize;
1458 } */ bsd_args;
1459 l_timeval linux_tv;
1460 struct timeval tv;
1461 socklen_t tv_len, xulen;
1462 struct xucred xu;
1463 struct l_ucred lxu;
1464 int error, name;
1465
1466 bsd_args.s = args->s;
1467 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1468 switch (bsd_args.level) {
1469 case SOL_SOCKET:
1470 name = linux_to_bsd_so_sockopt(args->optname);
1471 switch (name) {
1472 case SO_RCVTIMEO:
1473 /* FALLTHROUGH */
1474 case SO_SNDTIMEO:
1475 tv_len = sizeof(tv);
1476 error = kern_getsockopt(td, args->s, bsd_args.level,
1477 name, &tv, UIO_SYSSPACE, &tv_len);
1478 if (error)
1479 return (error);
1480 linux_tv.tv_sec = tv.tv_sec;
1481 linux_tv.tv_usec = tv.tv_usec;
1482 return (copyout(&linux_tv, PTRIN(args->optval),
1483 sizeof(linux_tv)));
1484 /* NOTREACHED */
1485 break;
1486 case LOCAL_PEERCRED:
1487 if (args->optlen != sizeof(lxu))
1488 return (EINVAL);
1489 xulen = sizeof(xu);
1490 error = kern_getsockopt(td, args->s, bsd_args.level,
1491 name, &xu, UIO_SYSSPACE, &xulen);
1492 if (error)
1493 return (error);
1494 /*
1495 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1496 */
1497 lxu.pid = 0;
1498 lxu.uid = xu.cr_uid;
1499 lxu.gid = xu.cr_gid;
1500 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1501 /* NOTREACHED */
1502 break;
1503 default:
1504 break;
1505 }
1506 break;
1507 case IPPROTO_IP:
1508 name = linux_to_bsd_ip_sockopt(args->optname);
1509 break;
1510 case IPPROTO_TCP:
1511 name = linux_to_bsd_tcp_sockopt(args->optname);
1512 break;
1513 default:
1514 name = -1;
1515 break;
1516 }
1517 if (name == -1)
1518 return (EINVAL);
1519
1520 bsd_args.name = name;
1521 bsd_args.val = PTRIN(args->optval);
1522 bsd_args.avalsize = PTRIN(args->optlen);
1523
1524 if (name == IPV6_NEXTHOP) {
1525 error = sys_getsockopt(td, &bsd_args);
1526 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1527 } else
1528 error = sys_getsockopt(td, &bsd_args);
1529
1530 return (error);
1531}
1532
1533/* Argument list sizes for linux_socketcall */
1534
1535#define LINUX_AL(x) ((x) * sizeof(l_ulong))
1536
1537static const unsigned char lxs_args[] = {
1538 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1539 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1540 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1541 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1542 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1543 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1544 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1545 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1546 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1547 LINUX_AL(4) /* accept4 */
1548};
1549
1550#define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1551
1552#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1553int
1554linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1555{
1556 l_ulong a[6];
1557 void *arg;
1558 int error;
1559
1560 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1561 return (EINVAL);
1562 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1563 if (error)
1564 return (error);
1565
1566 arg = a;
1567 switch (args->what) {
1568 case LINUX_SOCKET:
1569 return (linux_socket(td, arg));
1570 case LINUX_BIND:
1571 return (linux_bind(td, arg));
1572 case LINUX_CONNECT:
1573 return (linux_connect(td, arg));
1574 case LINUX_LISTEN:
1575 return (linux_listen(td, arg));
1576 case LINUX_ACCEPT:
1577 return (linux_accept(td, arg));
1578 case LINUX_GETSOCKNAME:
1579 return (linux_getsockname(td, arg));
1580 case LINUX_GETPEERNAME:
1581 return (linux_getpeername(td, arg));
1582 case LINUX_SOCKETPAIR:
1583 return (linux_socketpair(td, arg));
1584 case LINUX_SEND:
1585 return (linux_send(td, arg));
1586 case LINUX_RECV:
1587 return (linux_recv(td, arg));
1588 case LINUX_SENDTO:
1589 return (linux_sendto(td, arg));
1590 case LINUX_RECVFROM:
1591 return (linux_recvfrom(td, arg));
1592 case LINUX_SHUTDOWN:
1593 return (linux_shutdown(td, arg));
1594 case LINUX_SETSOCKOPT:
1595 return (linux_setsockopt(td, arg));
1596 case LINUX_GETSOCKOPT:
1597 return (linux_getsockopt(td, arg));
1598 case LINUX_SENDMSG:
1599 return (linux_sendmsg(td, arg));
1600 case LINUX_RECVMSG:
1601 return (linux_recvmsg(td, arg));
1602 case LINUX_ACCEPT4:
1603 return (linux_accept4(td, arg));
1604 }
1605
1606 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1607 return (ENOSYS);
1608}
1609#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2 * Copyright (c) 1995 S��ren Schmidt
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: head/sys/compat/linux/linux_socket.c 283413 2015-05-24 15:41:27Z dchagin $");
31
32/* XXX we use functions that might not exist. */
33#include "opt_compat.h"
34#include "opt_inet6.h"
35
36#include <sys/param.h>
37#include <sys/proc.h>
38#include <sys/systm.h>
39#include <sys/sysproto.h>
40#include <sys/capsicum.h>
41#include <sys/fcntl.h>
42#include <sys/file.h>
43#include <sys/limits.h>
44#include <sys/lock.h>
45#include <sys/malloc.h>
46#include <sys/mutex.h>
47#include <sys/mbuf.h>
48#include <sys/socket.h>
49#include <sys/socketvar.h>
50#include <sys/syscallsubr.h>
51#include <sys/uio.h>
52#include <sys/syslog.h>
53#include <sys/un.h>
54
55#include <net/if.h>
56#include <net/vnet.h>
57#include <netinet/in.h>
58#include <netinet/in_systm.h>
59#include <netinet/ip.h>
60#include <netinet/tcp.h>
61#ifdef INET6
62#include <netinet/ip6.h>
63#include <netinet6/ip6_var.h>
64#endif
65
66#ifdef COMPAT_LINUX32
67#include <machine/../linux32/linux.h>
68#include <machine/../linux32/linux32_proto.h>
69#else
70#include <machine/../linux/linux.h>
71#include <machine/../linux/linux_proto.h>
72#endif
73#include <compat/linux/linux_socket.h>
74#include <compat/linux/linux_util.h>
75
76static int linux_to_bsd_domain(int);
77
78/*
79 * Reads a linux sockaddr and does any necessary translation.
80 * Linux sockaddrs don't have a length field, only a family.
81 * Copy the osockaddr structure pointed to by osa to kernel, adjust
82 * family and convert to sockaddr.
83 */
84static int
85linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
86{
87 struct sockaddr *sa;
88 struct osockaddr *kosa;
89#ifdef INET6
90 struct sockaddr_in6 *sin6;
91 int oldv6size;
92#endif
93 char *name;
94 int bdom, error, hdrlen, namelen;
95
96 if (salen < 2 || salen > UCHAR_MAX || !osa)
97 return (EINVAL);
98
99#ifdef INET6
100 oldv6size = 0;
101 /*
102 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
103 * if it's a v4-mapped address, so reserve the proper space
104 * for it.
105 */
106 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
107 salen += sizeof(uint32_t);
108 oldv6size = 1;
109 }
110#endif
111
112 kosa = malloc(salen, M_SONAME, M_WAITOK);
113
114 if ((error = copyin(osa, kosa, salen)))
115 goto out;
116
117 bdom = linux_to_bsd_domain(kosa->sa_family);
118 if (bdom == -1) {
119 error = EAFNOSUPPORT;
120 goto out;
121 }
122
123#ifdef INET6
124 /*
125 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
126 * which lacks the scope id compared with RFC2553 one. If we detect
127 * the situation, reject the address and write a message to system log.
128 *
129 * Still accept addresses for which the scope id is not used.
130 */
131 if (oldv6size) {
132 if (bdom == AF_INET6) {
133 sin6 = (struct sockaddr_in6 *)kosa;
134 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
135 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
136 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
137 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
138 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
139 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
140 sin6->sin6_scope_id = 0;
141 } else {
142 log(LOG_DEBUG,
143 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
144 error = EINVAL;
145 goto out;
146 }
147 } else
148 salen -= sizeof(uint32_t);
149 }
150#endif
151 if (bdom == AF_INET) {
152 if (salen < sizeof(struct sockaddr_in)) {
153 error = EINVAL;
154 goto out;
155 }
156 salen = sizeof(struct sockaddr_in);
157 }
158
159 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
160 hdrlen = offsetof(struct sockaddr_un, sun_path);
161 name = ((struct sockaddr_un *)kosa)->sun_path;
162 if (*name == '\0') {
163 /*
164 * Linux abstract namespace starts with a NULL byte.
165 * XXX We do not support abstract namespace yet.
166 */
167 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
168 } else
169 namelen = strnlen(name, salen - hdrlen);
170 salen = hdrlen + namelen;
171 if (salen > sizeof(struct sockaddr_un)) {
172 error = ENAMETOOLONG;
173 goto out;
174 }
175 }
176
177 sa = (struct sockaddr *)kosa;
178 sa->sa_family = bdom;
179 sa->sa_len = salen;
180
181 *sap = sa;
182 return (0);
183
184out:
185 free(kosa, M_SONAME);
186 return (error);
187}
188
189static int
190linux_to_bsd_domain(int domain)
191{
192
193 switch (domain) {
194 case LINUX_AF_UNSPEC:
195 return (AF_UNSPEC);
196 case LINUX_AF_UNIX:
197 return (AF_LOCAL);
198 case LINUX_AF_INET:
199 return (AF_INET);
200 case LINUX_AF_INET6:
201 return (AF_INET6);
202 case LINUX_AF_AX25:
203 return (AF_CCITT);
204 case LINUX_AF_IPX:
205 return (AF_IPX);
206 case LINUX_AF_APPLETALK:
207 return (AF_APPLETALK);
208 }
209 return (-1);
210}
211
212static int
213bsd_to_linux_domain(int domain)
214{
215
216 switch (domain) {
217 case AF_UNSPEC:
218 return (LINUX_AF_UNSPEC);
219 case AF_LOCAL:
220 return (LINUX_AF_UNIX);
221 case AF_INET:
222 return (LINUX_AF_INET);
223 case AF_INET6:
224 return (LINUX_AF_INET6);
225 case AF_CCITT:
226 return (LINUX_AF_AX25);
227 case AF_IPX:
228 return (LINUX_AF_IPX);
229 case AF_APPLETALK:
230 return (LINUX_AF_APPLETALK);
231 }
232 return (-1);
233}
234
235static int
236linux_to_bsd_sockopt_level(int level)
237{
238
239 switch (level) {
240 case LINUX_SOL_SOCKET:
241 return (SOL_SOCKET);
242 }
243 return (level);
244}
245
246static int
247bsd_to_linux_sockopt_level(int level)
248{
249
250 switch (level) {
251 case SOL_SOCKET:
252 return (LINUX_SOL_SOCKET);
253 }
254 return (level);
255}
256
257static int
258linux_to_bsd_ip_sockopt(int opt)
259{
260
261 switch (opt) {
262 case LINUX_IP_TOS:
263 return (IP_TOS);
264 case LINUX_IP_TTL:
265 return (IP_TTL);
266 case LINUX_IP_OPTIONS:
267 return (IP_OPTIONS);
268 case LINUX_IP_MULTICAST_IF:
269 return (IP_MULTICAST_IF);
270 case LINUX_IP_MULTICAST_TTL:
271 return (IP_MULTICAST_TTL);
272 case LINUX_IP_MULTICAST_LOOP:
273 return (IP_MULTICAST_LOOP);
274 case LINUX_IP_ADD_MEMBERSHIP:
275 return (IP_ADD_MEMBERSHIP);
276 case LINUX_IP_DROP_MEMBERSHIP:
277 return (IP_DROP_MEMBERSHIP);
278 case LINUX_IP_HDRINCL:
279 return (IP_HDRINCL);
280 }
281 return (-1);
282}
283
284static int
285linux_to_bsd_so_sockopt(int opt)
286{
287
288 switch (opt) {
289 case LINUX_SO_DEBUG:
290 return (SO_DEBUG);
291 case LINUX_SO_REUSEADDR:
292 return (SO_REUSEADDR);
293 case LINUX_SO_TYPE:
294 return (SO_TYPE);
295 case LINUX_SO_ERROR:
296 return (SO_ERROR);
297 case LINUX_SO_DONTROUTE:
298 return (SO_DONTROUTE);
299 case LINUX_SO_BROADCAST:
300 return (SO_BROADCAST);
301 case LINUX_SO_SNDBUF:
302 return (SO_SNDBUF);
303 case LINUX_SO_RCVBUF:
304 return (SO_RCVBUF);
305 case LINUX_SO_KEEPALIVE:
306 return (SO_KEEPALIVE);
307 case LINUX_SO_OOBINLINE:
308 return (SO_OOBINLINE);
309 case LINUX_SO_LINGER:
310 return (SO_LINGER);
311 case LINUX_SO_PEERCRED:
312 return (LOCAL_PEERCRED);
313 case LINUX_SO_RCVLOWAT:
314 return (SO_RCVLOWAT);
315 case LINUX_SO_SNDLOWAT:
316 return (SO_SNDLOWAT);
317 case LINUX_SO_RCVTIMEO:
318 return (SO_RCVTIMEO);
319 case LINUX_SO_SNDTIMEO:
320 return (SO_SNDTIMEO);
321 case LINUX_SO_TIMESTAMP:
322 return (SO_TIMESTAMP);
323 case LINUX_SO_ACCEPTCONN:
324 return (SO_ACCEPTCONN);
325 }
326 return (-1);
327}
328
329static int
330linux_to_bsd_tcp_sockopt(int opt)
331{
332
333 switch (opt) {
334 case LINUX_TCP_NODELAY:
335 return (TCP_NODELAY);
336 case LINUX_TCP_MAXSEG:
337 return (TCP_MAXSEG);
338 case LINUX_TCP_KEEPIDLE:
339 return (TCP_KEEPIDLE);
340 case LINUX_TCP_KEEPINTVL:
341 return (TCP_KEEPINTVL);
342 case LINUX_TCP_KEEPCNT:
343 return (TCP_KEEPCNT);
344 case LINUX_TCP_MD5SIG:
345 return (TCP_MD5SIG);
346 }
347 return (-1);
348}
349
350static int
351linux_to_bsd_msg_flags(int flags)
352{
353 int ret_flags = 0;
354
355 if (flags & LINUX_MSG_OOB)
356 ret_flags |= MSG_OOB;
357 if (flags & LINUX_MSG_PEEK)
358 ret_flags |= MSG_PEEK;
359 if (flags & LINUX_MSG_DONTROUTE)
360 ret_flags |= MSG_DONTROUTE;
361 if (flags & LINUX_MSG_CTRUNC)
362 ret_flags |= MSG_CTRUNC;
363 if (flags & LINUX_MSG_TRUNC)
364 ret_flags |= MSG_TRUNC;
365 if (flags & LINUX_MSG_DONTWAIT)
366 ret_flags |= MSG_DONTWAIT;
367 if (flags & LINUX_MSG_EOR)
368 ret_flags |= MSG_EOR;
369 if (flags & LINUX_MSG_WAITALL)
370 ret_flags |= MSG_WAITALL;
371 if (flags & LINUX_MSG_NOSIGNAL)
372 ret_flags |= MSG_NOSIGNAL;
373#if 0 /* not handled */
374 if (flags & LINUX_MSG_PROXY)
375 ;
376 if (flags & LINUX_MSG_FIN)
377 ;
378 if (flags & LINUX_MSG_SYN)
379 ;
380 if (flags & LINUX_MSG_CONFIRM)
381 ;
382 if (flags & LINUX_MSG_RST)
383 ;
384 if (flags & LINUX_MSG_ERRQUEUE)
385 ;
386#endif
387 return ret_flags;
388}
389
390/*
391* If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
392* native syscall will fault. Thus, we don't really need to check the
393* return values for these functions.
394*/
395
396static int
397bsd_to_linux_sockaddr(struct sockaddr *arg)
398{
399 struct sockaddr sa;
400 size_t sa_len = sizeof(struct sockaddr);
401 int error;
402
403 if ((error = copyin(arg, &sa, sa_len)))
404 return (error);
405
406 *(u_short *)&sa = sa.sa_family;
407
408 error = copyout(&sa, arg, sa_len);
409
410 return (error);
411}
412
413static int
414linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
415{
416 struct sockaddr sa;
417 size_t sa_len = sizeof(struct sockaddr);
418 int error;
419
420 if ((error = copyin(arg, &sa, sa_len)))
421 return (error);
422
423 sa.sa_family = *(sa_family_t *)&sa;
424 sa.sa_len = len;
425
426 error = copyout(&sa, arg, sa_len);
427
428 return (error);
429}
430
431
432static int
433linux_sa_put(struct osockaddr *osa)
434{
435 struct osockaddr sa;
436 int error, bdom;
437
438 /*
439 * Only read/write the osockaddr family part, the rest is
440 * not changed.
441 */
442 error = copyin(osa, &sa, sizeof(sa.sa_family));
443 if (error)
444 return (error);
445
446 bdom = bsd_to_linux_domain(sa.sa_family);
447 if (bdom == -1)
448 return (EINVAL);
449
450 sa.sa_family = bdom;
451 error = copyout(&sa, osa, sizeof(sa.sa_family));
452 if (error)
453 return (error);
454
455 return (0);
456}
457
458static int
459linux_to_bsd_cmsg_type(int cmsg_type)
460{
461
462 switch (cmsg_type) {
463 case LINUX_SCM_RIGHTS:
464 return (SCM_RIGHTS);
465 case LINUX_SCM_CREDENTIALS:
466 return (SCM_CREDS);
467 }
468 return (-1);
469}
470
471static int
472bsd_to_linux_cmsg_type(int cmsg_type)
473{
474
475 switch (cmsg_type) {
476 case SCM_RIGHTS:
477 return (LINUX_SCM_RIGHTS);
478 case SCM_CREDS:
479 return (LINUX_SCM_CREDENTIALS);
480 }
481 return (-1);
482}
483
484static int
485linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
486{
487 if (lhdr->msg_controllen > INT_MAX)
488 return (ENOBUFS);
489
490 bhdr->msg_name = PTRIN(lhdr->msg_name);
491 bhdr->msg_namelen = lhdr->msg_namelen;
492 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
493 bhdr->msg_iovlen = lhdr->msg_iovlen;
494 bhdr->msg_control = PTRIN(lhdr->msg_control);
495
496 /*
497 * msg_controllen is skipped since BSD and LINUX control messages
498 * are potentially different sizes (e.g. the cred structure used
499 * by SCM_CREDS is different between the two operating system).
500 *
501 * The caller can set it (if necessary) after converting all the
502 * control messages.
503 */
504
505 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
506 return (0);
507}
508
509static int
510bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
511{
512 lhdr->msg_name = PTROUT(bhdr->msg_name);
513 lhdr->msg_namelen = bhdr->msg_namelen;
514 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
515 lhdr->msg_iovlen = bhdr->msg_iovlen;
516 lhdr->msg_control = PTROUT(bhdr->msg_control);
517
518 /*
519 * msg_controllen is skipped since BSD and LINUX control messages
520 * are potentially different sizes (e.g. the cred structure used
521 * by SCM_CREDS is different between the two operating system).
522 *
523 * The caller can set it (if necessary) after converting all the
524 * control messages.
525 */
526
527 /* msg_flags skipped */
528 return (0);
529}
530
531static int
532linux_set_socket_flags(struct thread *td, int s, int flags)
533{
534 int error;
535
536 if (flags & LINUX_SOCK_NONBLOCK) {
537 error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
538 if (error)
539 return (error);
540 }
541 if (flags & LINUX_SOCK_CLOEXEC) {
542 error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
543 if (error)
544 return (error);
545 }
546 return (0);
547}
548
549static int
550linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
551 struct mbuf *control, enum uio_seg segflg)
552{
553 struct sockaddr *to;
554 int error;
555
556 if (mp->msg_name != NULL) {
557 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
558 if (error)
559 return (error);
560 mp->msg_name = to;
561 } else
562 to = NULL;
563
564 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
565 segflg);
566
567 if (to)
568 free(to, M_SONAME);
569 return (error);
570}
571
572/* Return 0 if IP_HDRINCL is set for the given socket. */
573static int
574linux_check_hdrincl(struct thread *td, int s)
575{
576 int error, optval;
577 socklen_t size_val;
578
579 size_val = sizeof(optval);
580 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
581 &optval, UIO_SYSSPACE, &size_val);
582 if (error)
583 return (error);
584
585 return (optval == 0);
586}
587
588/*
589 * Updated sendto() when IP_HDRINCL is set:
590 * tweak endian-dependent fields in the IP packet.
591 */
592static int
593linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
594{
595/*
596 * linux_ip_copysize defines how many bytes we should copy
597 * from the beginning of the IP packet before we customize it for BSD.
598 * It should include all the fields we modify (ip_len and ip_off).
599 */
600#define linux_ip_copysize 8
601
602 struct ip *packet;
603 struct msghdr msg;
604 struct iovec aiov[1];
605 int error;
606
607 /* Check that the packet isn't too big or too small. */
608 if (linux_args->len < linux_ip_copysize ||
609 linux_args->len > IP_MAXPACKET)
610 return (EINVAL);
611
612 packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
613
614 /* Make kernel copy of the packet to be sent */
615 if ((error = copyin(PTRIN(linux_args->msg), packet,
616 linux_args->len)))
617 goto goout;
618
619 /* Convert fields from Linux to BSD raw IP socket format */
620 packet->ip_len = linux_args->len;
621 packet->ip_off = ntohs(packet->ip_off);
622
623 /* Prepare the msghdr and iovec structures describing the new packet */
624 msg.msg_name = PTRIN(linux_args->to);
625 msg.msg_namelen = linux_args->tolen;
626 msg.msg_iov = aiov;
627 msg.msg_iovlen = 1;
628 msg.msg_control = NULL;
629 msg.msg_flags = 0;
630 aiov[0].iov_base = (char *)packet;
631 aiov[0].iov_len = linux_args->len;
632 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
633 NULL, UIO_SYSSPACE);
634goout:
635 free(packet, M_TEMP);
636 return (error);
637}
638
639int
640linux_socket(struct thread *td, struct linux_socket_args *args)
641{
642 struct socket_args /* {
643 int domain;
644 int type;
645 int protocol;
646 } */ bsd_args;
647 int retval_socket, socket_flags;
648
649 bsd_args.protocol = args->protocol;
650 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
651 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
652 return (EINVAL);
653 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
654 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
655 return (EINVAL);
656 bsd_args.domain = linux_to_bsd_domain(args->domain);
657 if (bsd_args.domain == -1)
658 return (EAFNOSUPPORT);
659
660 retval_socket = sys_socket(td, &bsd_args);
661 if (retval_socket)
662 return (retval_socket);
663
664 retval_socket = linux_set_socket_flags(td, td->td_retval[0],
665 socket_flags);
666 if (retval_socket) {
667 (void)kern_close(td, td->td_retval[0]);
668 goto out;
669 }
670
671 if (bsd_args.type == SOCK_RAW
672 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
673 && bsd_args.domain == PF_INET) {
674 /* It's a raw IP socket: set the IP_HDRINCL option. */
675 int hdrincl;
676
677 hdrincl = 1;
678 /* We ignore any error returned by kern_setsockopt() */
679 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
680 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
681 }
682#ifdef INET6
683 /*
684 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
685 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
686 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
687 * sysctl value.
688 */
689 if (bsd_args.domain == PF_INET6) {
690 int v6only;
691
692 v6only = 0;
693 /* We ignore any error returned by setsockopt() */
694 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
695 &v6only, UIO_SYSSPACE, sizeof(v6only));
696 }
697#endif
698
699out:
700 return (retval_socket);
701}
702
703int
704linux_bind(struct thread *td, struct linux_bind_args *args)
705{
706 struct sockaddr *sa;
707 int error;
708
709 error = linux_getsockaddr(&sa, PTRIN(args->name),
710 args->namelen);
711 if (error)
712 return (error);
713
714 error = kern_bindat(td, AT_FDCWD, args->s, sa);
715 free(sa, M_SONAME);
716 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
717 return (EINVAL);
718 return (error);
719}
720
721int
722linux_connect(struct thread *td, struct linux_connect_args *args)
723{
724 cap_rights_t rights;
725 struct socket *so;
726 struct sockaddr *sa;
727 u_int fflag;
728 int error;
729
730 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
731 args->namelen);
732 if (error)
733 return (error);
734
735 error = kern_connectat(td, AT_FDCWD, args->s, sa);
736 free(sa, M_SONAME);
737 if (error != EISCONN)
738 return (error);
739
740 /*
741 * Linux doesn't return EISCONN the first time it occurs,
742 * when on a non-blocking socket. Instead it returns the
743 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
744 *
745 * XXXRW: Instead of using fgetsock(), check that it is a
746 * socket and use the file descriptor reference instead of
747 * creating a new one.
748 */
749 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
750 &so, &fflag);
751 if (error == 0) {
752 error = EISCONN;
753 if (fflag & FNONBLOCK) {
754 SOCK_LOCK(so);
755 if (so->so_emuldata == 0)
756 error = so->so_error;
757 so->so_emuldata = (void *)1;
758 SOCK_UNLOCK(so);
759 }
760 fputsock(so);
761 }
762 return (error);
763}
764
765int
766linux_listen(struct thread *td, struct linux_listen_args *args)
767{
768 struct listen_args /* {
769 int s;
770 int backlog;
771 } */ bsd_args;
772
773 bsd_args.s = args->s;
774 bsd_args.backlog = args->backlog;
775 return (sys_listen(td, &bsd_args));
776}
777
778static int
779linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
780 l_uintptr_t namelen, int flags)
781{
782 struct accept_args /* {
783 int s;
784 struct sockaddr * __restrict name;
785 socklen_t * __restrict anamelen;
786 } */ bsd_args;
787 int error;
788
789 if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
790 return (EINVAL);
791
792 bsd_args.s = s;
793 /* XXX: */
794 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
795 bsd_args.anamelen = PTRIN(namelen);/* XXX */
796 error = sys_accept(td, &bsd_args);
797 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
798 if (error) {
799 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
800 return (EINVAL);
801 return (error);
802 }
803
804 /*
805 * linux appears not to copy flags from the parent socket to the
806 * accepted one, so we must clear the flags in the new descriptor
807 * and apply the requested flags.
808 */
809 error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
810 if (error)
811 goto out;
812 error = linux_set_socket_flags(td, td->td_retval[0], flags);
813 if (error)
814 goto out;
815 if (addr)
816 error = linux_sa_put(PTRIN(addr));
817
818out:
819 if (error) {
820 (void)kern_close(td, td->td_retval[0]);
821 td->td_retval[0] = 0;
822 }
823 return (error);
824}
825
826int
827linux_accept(struct thread *td, struct linux_accept_args *args)
828{
829
830 return (linux_accept_common(td, args->s, args->addr,
831 args->namelen, 0));
832}
833
834int
835linux_accept4(struct thread *td, struct linux_accept4_args *args)
836{
837
838 return (linux_accept_common(td, args->s, args->addr,
839 args->namelen, args->flags));
840}
841
842int
843linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
844{
845 struct getsockname_args /* {
846 int fdes;
847 struct sockaddr * __restrict asa;
848 socklen_t * __restrict alen;
849 } */ bsd_args;
850 int error;
851
852 bsd_args.fdes = args->s;
853 /* XXX: */
854 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
855 bsd_args.alen = PTRIN(args->namelen); /* XXX */
856 error = sys_getsockname(td, &bsd_args);
857 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
858 if (error)
859 return (error);
860 error = linux_sa_put(PTRIN(args->addr));
861 if (error)
862 return (error);
863 return (0);
864}
865
866int
867linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
868{
869 struct getpeername_args /* {
870 int fdes;
871 caddr_t asa;
872 int *alen;
873 } */ bsd_args;
874 int error;
875
876 bsd_args.fdes = args->s;
877 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
878 bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
879 error = sys_getpeername(td, &bsd_args);
880 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
881 if (error)
882 return (error);
883 error = linux_sa_put(PTRIN(args->addr));
884 if (error)
885 return (error);
886 return (0);
887}
888
889int
890linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
891{
892 struct socketpair_args /* {
893 int domain;
894 int type;
895 int protocol;
896 int *rsv;
897 } */ bsd_args;
898 int error, socket_flags;
899 int sv[2];
900
901 bsd_args.domain = linux_to_bsd_domain(args->domain);
902 if (bsd_args.domain != PF_LOCAL)
903 return (EAFNOSUPPORT);
904
905 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
906 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
907 return (EINVAL);
908 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
909 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
910 return (EINVAL);
911
912 if (args->protocol != 0 && args->protocol != PF_UNIX)
913
914 /*
915 * Use of PF_UNIX as protocol argument is not right,
916 * but Linux does it.
917 * Do not map PF_UNIX as its Linux value is identical
918 * to FreeBSD one.
919 */
920 return (EPROTONOSUPPORT);
921 else
922 bsd_args.protocol = 0;
923 bsd_args.rsv = (int *)PTRIN(args->rsv);
924 error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
925 bsd_args.protocol, sv);
926 if (error)
927 return (error);
928 error = linux_set_socket_flags(td, sv[0], socket_flags);
929 if (error)
930 goto out;
931 error = linux_set_socket_flags(td, sv[1], socket_flags);
932 if (error)
933 goto out;
934
935 error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
936
937out:
938 if (error) {
939 (void)kern_close(td, sv[0]);
940 (void)kern_close(td, sv[1]);
941 }
942 return (error);
943}
944
945#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
946struct linux_send_args {
947 int s;
948 l_uintptr_t msg;
949 int len;
950 int flags;
951};
952
953static int
954linux_send(struct thread *td, struct linux_send_args *args)
955{
956 struct sendto_args /* {
957 int s;
958 caddr_t buf;
959 int len;
960 int flags;
961 caddr_t to;
962 int tolen;
963 } */ bsd_args;
964
965 bsd_args.s = args->s;
966 bsd_args.buf = (caddr_t)PTRIN(args->msg);
967 bsd_args.len = args->len;
968 bsd_args.flags = args->flags;
969 bsd_args.to = NULL;
970 bsd_args.tolen = 0;
971 return sys_sendto(td, &bsd_args);
972}
973
974struct linux_recv_args {
975 int s;
976 l_uintptr_t msg;
977 int len;
978 int flags;
979};
980
981static int
982linux_recv(struct thread *td, struct linux_recv_args *args)
983{
984 struct recvfrom_args /* {
985 int s;
986 caddr_t buf;
987 int len;
988 int flags;
989 struct sockaddr *from;
990 socklen_t fromlenaddr;
991 } */ bsd_args;
992
993 bsd_args.s = args->s;
994 bsd_args.buf = (caddr_t)PTRIN(args->msg);
995 bsd_args.len = args->len;
996 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
997 bsd_args.from = NULL;
998 bsd_args.fromlenaddr = 0;
999 return (sys_recvfrom(td, &bsd_args));
1000}
1001#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1002
1003int
1004linux_sendto(struct thread *td, struct linux_sendto_args *args)
1005{
1006 struct msghdr msg;
1007 struct iovec aiov;
1008 int error;
1009
1010 if (linux_check_hdrincl(td, args->s) == 0)
1011 /* IP_HDRINCL set, tweak the packet before sending */
1012 return (linux_sendto_hdrincl(td, args));
1013
1014 msg.msg_name = PTRIN(args->to);
1015 msg.msg_namelen = args->tolen;
1016 msg.msg_iov = &aiov;
1017 msg.msg_iovlen = 1;
1018 msg.msg_control = NULL;
1019 msg.msg_flags = 0;
1020 aiov.iov_base = PTRIN(args->msg);
1021 aiov.iov_len = args->len;
1022 error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1023 UIO_USERSPACE);
1024 return (error);
1025}
1026
1027int
1028linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1029{
1030 struct recvfrom_args /* {
1031 int s;
1032 caddr_t buf;
1033 size_t len;
1034 int flags;
1035 struct sockaddr * __restrict from;
1036 socklen_t * __restrict fromlenaddr;
1037 } */ bsd_args;
1038 size_t len;
1039 int error;
1040
1041 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1042 return (error);
1043
1044 bsd_args.s = args->s;
1045 bsd_args.buf = PTRIN(args->buf);
1046 bsd_args.len = args->len;
1047 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1048 /* XXX: */
1049 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1050 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1051
1052 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1053 error = sys_recvfrom(td, &bsd_args);
1054 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1055
1056 if (error)
1057 return (error);
1058 if (args->from) {
1059 error = linux_sa_put((struct osockaddr *)
1060 PTRIN(args->from));
1061 if (error)
1062 return (error);
1063 }
1064 return (0);
1065}
1066
1067int
1068linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1069{
1070 struct cmsghdr *cmsg;
1071 struct cmsgcred cmcred;
1072 struct mbuf *control;
1073 struct msghdr msg;
1074 struct l_cmsghdr linux_cmsg;
1075 struct l_cmsghdr *ptr_cmsg;
1076 struct l_msghdr linux_msg;
1077 struct iovec *iov;
1078 socklen_t datalen;
1079 struct sockaddr *sa;
1080 sa_family_t sa_family;
1081 void *data;
1082 int error;
1083
1084 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1085 if (error)
1086 return (error);
1087
1088 /*
1089 * Some Linux applications (ping) define a non-NULL control data
1090 * pointer, but a msg_controllen of 0, which is not allowed in the
1091 * FreeBSD system call interface. NULL the msg_control pointer in
1092 * order to handle this case. This should be checked, but allows the
1093 * Linux ping to work.
1094 */
1095 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1096 linux_msg.msg_control = PTROUT(NULL);
1097
1098 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1099 if (error)
1100 return (error);
1101
1102#ifdef COMPAT_LINUX32
1103 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1104 &iov, EMSGSIZE);
1105#else
1106 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1107#endif
1108 if (error)
1109 return (error);
1110
1111 control = NULL;
1112 cmsg = NULL;
1113
1114 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1115 error = kern_getsockname(td, args->s, &sa, &datalen);
1116 if (error)
1117 goto bad;
1118 sa_family = sa->sa_family;
1119 free(sa, M_SONAME);
1120
1121 error = ENOBUFS;
1122 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1123 control = m_get(M_WAITOK, MT_CONTROL);
1124 if (control == NULL)
1125 goto bad;
1126
1127 do {
1128 error = copyin(ptr_cmsg, &linux_cmsg,
1129 sizeof(struct l_cmsghdr));
1130 if (error)
1131 goto bad;
1132
1133 error = EINVAL;
1134 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1135 goto bad;
1136
1137 /*
1138 * Now we support only SCM_RIGHTS and SCM_CRED,
1139 * so return EINVAL in any other cmsg_type
1140 */
1141 cmsg->cmsg_type =
1142 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1143 cmsg->cmsg_level =
1144 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1145 if (cmsg->cmsg_type == -1
1146 || cmsg->cmsg_level != SOL_SOCKET)
1147 goto bad;
1148
1149 /*
1150 * Some applications (e.g. pulseaudio) attempt to
1151 * send ancillary data even if the underlying protocol
1152 * doesn't support it which is not allowed in the
1153 * FreeBSD system call interface.
1154 */
1155 if (sa_family != AF_UNIX)
1156 continue;
1157
1158 data = LINUX_CMSG_DATA(ptr_cmsg);
1159 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1160
1161 switch (cmsg->cmsg_type)
1162 {
1163 case SCM_RIGHTS:
1164 break;
1165
1166 case SCM_CREDS:
1167 data = &cmcred;
1168 datalen = sizeof(cmcred);
1169
1170 /*
1171 * The lower levels will fill in the structure
1172 */
1173 bzero(data, datalen);
1174 break;
1175 }
1176
1177 cmsg->cmsg_len = CMSG_LEN(datalen);
1178
1179 error = ENOBUFS;
1180 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1181 goto bad;
1182 if (!m_append(control, datalen, (c_caddr_t)data))
1183 goto bad;
1184 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1185
1186 if (m_length(control, NULL) == 0) {
1187 m_freem(control);
1188 control = NULL;
1189 }
1190 }
1191
1192 msg.msg_iov = iov;
1193 msg.msg_flags = 0;
1194 error = linux_sendit(td, args->s, &msg, args->flags, control,
1195 UIO_USERSPACE);
1196
1197bad:
1198 free(iov, M_IOV);
1199 if (cmsg)
1200 free(cmsg, M_TEMP);
1201 return (error);
1202}
1203
1204int
1205linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1206{
1207 struct cmsghdr *cm;
1208 struct cmsgcred *cmcred;
1209 struct msghdr msg;
1210 struct l_cmsghdr *linux_cmsg = NULL;
1211 struct l_ucred linux_ucred;
1212 socklen_t datalen, outlen;
1213 struct l_msghdr linux_msg;
1214 struct iovec *iov, *uiov;
1215 struct mbuf *control = NULL;
1216 struct mbuf **controlp;
1217 caddr_t outbuf;
1218 void *data;
1219 int error, i, fd, fds, *fdp;
1220
1221 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1222 if (error)
1223 return (error);
1224
1225 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1226 if (error)
1227 return (error);
1228
1229#ifdef COMPAT_LINUX32
1230 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1231 &iov, EMSGSIZE);
1232#else
1233 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1234#endif
1235 if (error)
1236 return (error);
1237
1238 if (msg.msg_name) {
1239 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1240 msg.msg_namelen);
1241 if (error)
1242 goto bad;
1243 }
1244
1245 uiov = msg.msg_iov;
1246 msg.msg_iov = iov;
1247 controlp = (msg.msg_control != NULL) ? &control : NULL;
1248 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1249 msg.msg_iov = uiov;
1250 if (error)
1251 goto bad;
1252
1253 error = bsd_to_linux_msghdr(&msg, &linux_msg);
1254 if (error)
1255 goto bad;
1256
1257 if (linux_msg.msg_name) {
1258 error = bsd_to_linux_sockaddr((struct sockaddr *)
1259 PTRIN(linux_msg.msg_name));
1260 if (error)
1261 goto bad;
1262 }
1263 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1264 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1265 if (error)
1266 goto bad;
1267 }
1268
1269 outbuf = PTRIN(linux_msg.msg_control);
1270 outlen = 0;
1271
1272 if (control) {
1273 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1274
1275 msg.msg_control = mtod(control, struct cmsghdr *);
1276 msg.msg_controllen = control->m_len;
1277
1278 cm = CMSG_FIRSTHDR(&msg);
1279
1280 while (cm != NULL) {
1281 linux_cmsg->cmsg_type =
1282 bsd_to_linux_cmsg_type(cm->cmsg_type);
1283 linux_cmsg->cmsg_level =
1284 bsd_to_linux_sockopt_level(cm->cmsg_level);
1285 if (linux_cmsg->cmsg_type == -1
1286 || cm->cmsg_level != SOL_SOCKET)
1287 {
1288 error = EINVAL;
1289 goto bad;
1290 }
1291
1292 data = CMSG_DATA(cm);
1293 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1294
1295 switch (cm->cmsg_type)
1296 {
1297 case SCM_RIGHTS:
1298 if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1299 fds = datalen / sizeof(int);
1300 fdp = data;
1301 for (i = 0; i < fds; i++) {
1302 fd = *fdp++;
1303 (void)kern_fcntl(td, fd,
1304 F_SETFD, FD_CLOEXEC);
1305 }
1306 }
1307 break;
1308
1309 case SCM_CREDS:
1310 /*
1311 * Currently LOCAL_CREDS is never in
1312 * effect for Linux so no need to worry
1313 * about sockcred
1314 */
1315 if (datalen != sizeof(*cmcred)) {
1316 error = EMSGSIZE;
1317 goto bad;
1318 }
1319 cmcred = (struct cmsgcred *)data;
1320 bzero(&linux_ucred, sizeof(linux_ucred));
1321 linux_ucred.pid = cmcred->cmcred_pid;
1322 linux_ucred.uid = cmcred->cmcred_uid;
1323 linux_ucred.gid = cmcred->cmcred_gid;
1324 data = &linux_ucred;
1325 datalen = sizeof(linux_ucred);
1326 break;
1327 }
1328
1329 if (outlen + LINUX_CMSG_LEN(datalen) >
1330 linux_msg.msg_controllen) {
1331 if (outlen == 0) {
1332 error = EMSGSIZE;
1333 goto bad;
1334 } else {
1335 linux_msg.msg_flags |=
1336 LINUX_MSG_CTRUNC;
1337 goto out;
1338 }
1339 }
1340
1341 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1342
1343 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1344 if (error)
1345 goto bad;
1346 outbuf += L_CMSG_HDRSZ;
1347
1348 error = copyout(data, outbuf, datalen);
1349 if (error)
1350 goto bad;
1351
1352 outbuf += LINUX_CMSG_ALIGN(datalen);
1353 outlen += LINUX_CMSG_LEN(datalen);
1354
1355 cm = CMSG_NXTHDR(&msg, cm);
1356 }
1357 }
1358
1359out:
1360 linux_msg.msg_controllen = outlen;
1361 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1362
1363bad:
1364 free(iov, M_IOV);
1365 m_freem(control);
1366 free(linux_cmsg, M_TEMP);
1367
1368 return (error);
1369}
1370
1371int
1372linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1373{
1374 struct shutdown_args /* {
1375 int s;
1376 int how;
1377 } */ bsd_args;
1378
1379 bsd_args.s = args->s;
1380 bsd_args.how = args->how;
1381 return (sys_shutdown(td, &bsd_args));
1382}
1383
1384int
1385linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1386{
1387 struct setsockopt_args /* {
1388 int s;
1389 int level;
1390 int name;
1391 caddr_t val;
1392 int valsize;
1393 } */ bsd_args;
1394 l_timeval linux_tv;
1395 struct timeval tv;
1396 int error, name;
1397
1398 bsd_args.s = args->s;
1399 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1400 switch (bsd_args.level) {
1401 case SOL_SOCKET:
1402 name = linux_to_bsd_so_sockopt(args->optname);
1403 switch (name) {
1404 case SO_RCVTIMEO:
1405 /* FALLTHROUGH */
1406 case SO_SNDTIMEO:
1407 error = copyin(PTRIN(args->optval), &linux_tv,
1408 sizeof(linux_tv));
1409 if (error)
1410 return (error);
1411 tv.tv_sec = linux_tv.tv_sec;
1412 tv.tv_usec = linux_tv.tv_usec;
1413 return (kern_setsockopt(td, args->s, bsd_args.level,
1414 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1415 /* NOTREACHED */
1416 break;
1417 default:
1418 break;
1419 }
1420 break;
1421 case IPPROTO_IP:
1422 name = linux_to_bsd_ip_sockopt(args->optname);
1423 break;
1424 case IPPROTO_TCP:
1425 name = linux_to_bsd_tcp_sockopt(args->optname);
1426 break;
1427 default:
1428 name = -1;
1429 break;
1430 }
1431 if (name == -1)
1432 return (ENOPROTOOPT);
1433
1434 bsd_args.name = name;
1435 bsd_args.val = PTRIN(args->optval);
1436 bsd_args.valsize = args->optlen;
1437
1438 if (name == IPV6_NEXTHOP) {
1439 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1440 bsd_args.valsize);
1441 error = sys_setsockopt(td, &bsd_args);
1442 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1443 } else
1444 error = sys_setsockopt(td, &bsd_args);
1445
1446 return (error);
1447}
1448
1449int
1450linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1451{
1452 struct getsockopt_args /* {
1453 int s;
1454 int level;
1455 int name;
1456 caddr_t val;
1457 int *avalsize;
1458 } */ bsd_args;
1459 l_timeval linux_tv;
1460 struct timeval tv;
1461 socklen_t tv_len, xulen;
1462 struct xucred xu;
1463 struct l_ucred lxu;
1464 int error, name;
1465
1466 bsd_args.s = args->s;
1467 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1468 switch (bsd_args.level) {
1469 case SOL_SOCKET:
1470 name = linux_to_bsd_so_sockopt(args->optname);
1471 switch (name) {
1472 case SO_RCVTIMEO:
1473 /* FALLTHROUGH */
1474 case SO_SNDTIMEO:
1475 tv_len = sizeof(tv);
1476 error = kern_getsockopt(td, args->s, bsd_args.level,
1477 name, &tv, UIO_SYSSPACE, &tv_len);
1478 if (error)
1479 return (error);
1480 linux_tv.tv_sec = tv.tv_sec;
1481 linux_tv.tv_usec = tv.tv_usec;
1482 return (copyout(&linux_tv, PTRIN(args->optval),
1483 sizeof(linux_tv)));
1484 /* NOTREACHED */
1485 break;
1486 case LOCAL_PEERCRED:
1487 if (args->optlen != sizeof(lxu))
1488 return (EINVAL);
1489 xulen = sizeof(xu);
1490 error = kern_getsockopt(td, args->s, bsd_args.level,
1491 name, &xu, UIO_SYSSPACE, &xulen);
1492 if (error)
1493 return (error);
1494 /*
1495 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1496 */
1497 lxu.pid = 0;
1498 lxu.uid = xu.cr_uid;
1499 lxu.gid = xu.cr_gid;
1500 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1501 /* NOTREACHED */
1502 break;
1503 default:
1504 break;
1505 }
1506 break;
1507 case IPPROTO_IP:
1508 name = linux_to_bsd_ip_sockopt(args->optname);
1509 break;
1510 case IPPROTO_TCP:
1511 name = linux_to_bsd_tcp_sockopt(args->optname);
1512 break;
1513 default:
1514 name = -1;
1515 break;
1516 }
1517 if (name == -1)
1518 return (EINVAL);
1519
1520 bsd_args.name = name;
1521 bsd_args.val = PTRIN(args->optval);
1522 bsd_args.avalsize = PTRIN(args->optlen);
1523
1524 if (name == IPV6_NEXTHOP) {
1525 error = sys_getsockopt(td, &bsd_args);
1526 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1527 } else
1528 error = sys_getsockopt(td, &bsd_args);
1529
1530 return (error);
1531}
1532
1533/* Argument list sizes for linux_socketcall */
1534
1535#define LINUX_AL(x) ((x) * sizeof(l_ulong))
1536
1537static const unsigned char lxs_args[] = {
1538 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1539 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1540 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1541 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1542 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1543 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1544 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1545 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1546 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1547 LINUX_AL(4) /* accept4 */
1548};
1549
1550#define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1551
1552#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1553int
1554linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1555{
1556 l_ulong a[6];
1557 void *arg;
1558 int error;
1559
1560 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1561 return (EINVAL);
1562 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1563 if (error)
1564 return (error);
1565
1566 arg = a;
1567 switch (args->what) {
1568 case LINUX_SOCKET:
1569 return (linux_socket(td, arg));
1570 case LINUX_BIND:
1571 return (linux_bind(td, arg));
1572 case LINUX_CONNECT:
1573 return (linux_connect(td, arg));
1574 case LINUX_LISTEN:
1575 return (linux_listen(td, arg));
1576 case LINUX_ACCEPT:
1577 return (linux_accept(td, arg));
1578 case LINUX_GETSOCKNAME:
1579 return (linux_getsockname(td, arg));
1580 case LINUX_GETPEERNAME:
1581 return (linux_getpeername(td, arg));
1582 case LINUX_SOCKETPAIR:
1583 return (linux_socketpair(td, arg));
1584 case LINUX_SEND:
1585 return (linux_send(td, arg));
1586 case LINUX_RECV:
1587 return (linux_recv(td, arg));
1588 case LINUX_SENDTO:
1589 return (linux_sendto(td, arg));
1590 case LINUX_RECVFROM:
1591 return (linux_recvfrom(td, arg));
1592 case LINUX_SHUTDOWN:
1593 return (linux_shutdown(td, arg));
1594 case LINUX_SETSOCKOPT:
1595 return (linux_setsockopt(td, arg));
1596 case LINUX_GETSOCKOPT:
1597 return (linux_getsockopt(td, arg));
1598 case LINUX_SENDMSG:
1599 return (linux_sendmsg(td, arg));
1600 case LINUX_RECVMSG:
1601 return (linux_recvmsg(td, arg));
1602 case LINUX_ACCEPT4:
1603 return (linux_accept4(td, arg));
1604 }
1605
1606 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1607 return (ENOSYS);
1608}
1609#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */