2 * Copyright (c) 1995 Søren Schmidt
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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
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.
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/limits.h>
45 #include <sys/malloc.h>
46 #include <sys/mutex.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/syscallsubr.h>
52 #include <sys/syslog.h>
57 #include <netinet/in.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/ip.h>
60 #include <netinet/tcp.h>
62 #include <netinet/ip6.h>
63 #include <netinet6/ip6_var.h>
67 #include <machine/../linux32/linux.h>
68 #include <machine/../linux32/linux32_proto.h>
70 #include <machine/../linux/linux.h>
71 #include <machine/../linux/linux_proto.h>
73 #include <compat/linux/linux_file.h>
74 #include <compat/linux/linux_socket.h>
75 #include <compat/linux/linux_timer.h>
76 #include <compat/linux/linux_util.h>
78 static int linux_to_bsd_domain(int);
79 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
81 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
82 l_uint, struct msghdr *);
83 static int linux_set_socket_flags(int, int *);
86 * Reads a linux sockaddr and does any necessary translation.
87 * Linux sockaddrs don't have a length field, only a family.
88 * Copy the osockaddr structure pointed to by osa to kernel, adjust
89 * family and convert to sockaddr.
92 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
95 struct osockaddr *kosa;
97 struct sockaddr_in6 *sin6;
101 int bdom, error, hdrlen, namelen;
103 if (salen < 2 || salen > UCHAR_MAX || !osa)
109 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
110 * if it's a v4-mapped address, so reserve the proper space
113 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
114 salen += sizeof(uint32_t);
119 kosa = malloc(salen, M_SONAME, M_WAITOK);
121 if ((error = copyin(osa, kosa, salen)))
124 bdom = linux_to_bsd_domain(kosa->sa_family);
126 error = EAFNOSUPPORT;
132 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
133 * which lacks the scope id compared with RFC2553 one. If we detect
134 * the situation, reject the address and write a message to system log.
136 * Still accept addresses for which the scope id is not used.
139 if (bdom == AF_INET6) {
140 sin6 = (struct sockaddr_in6 *)kosa;
141 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
142 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
143 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
144 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
145 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
146 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
147 sin6->sin6_scope_id = 0;
150 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
155 salen -= sizeof(uint32_t);
158 if (bdom == AF_INET) {
159 if (salen < sizeof(struct sockaddr_in)) {
163 salen = sizeof(struct sockaddr_in);
166 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
167 hdrlen = offsetof(struct sockaddr_un, sun_path);
168 name = ((struct sockaddr_un *)kosa)->sun_path;
171 * Linux abstract namespace starts with a NULL byte.
172 * XXX We do not support abstract namespace yet.
174 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
176 namelen = strnlen(name, salen - hdrlen);
177 salen = hdrlen + namelen;
178 if (salen > sizeof(struct sockaddr_un)) {
179 error = ENAMETOOLONG;
184 sa = (struct sockaddr *)kosa;
185 sa->sa_family = bdom;
192 free(kosa, M_SONAME);
197 linux_to_bsd_domain(int domain)
201 case LINUX_AF_UNSPEC:
213 case LINUX_AF_APPLETALK:
214 return (AF_APPLETALK);
220 bsd_to_linux_domain(int domain)
225 return (LINUX_AF_UNSPEC);
227 return (LINUX_AF_UNIX);
229 return (LINUX_AF_INET);
231 return (LINUX_AF_INET6);
233 return (LINUX_AF_AX25);
235 return (LINUX_AF_IPX);
237 return (LINUX_AF_APPLETALK);
243 linux_to_bsd_sockopt_level(int level)
247 case LINUX_SOL_SOCKET:
254 bsd_to_linux_sockopt_level(int level)
259 return (LINUX_SOL_SOCKET);
265 linux_to_bsd_ip_sockopt(int opt)
273 case LINUX_IP_OPTIONS:
275 case LINUX_IP_MULTICAST_IF:
276 return (IP_MULTICAST_IF);
277 case LINUX_IP_MULTICAST_TTL:
278 return (IP_MULTICAST_TTL);
279 case LINUX_IP_MULTICAST_LOOP:
280 return (IP_MULTICAST_LOOP);
281 case LINUX_IP_ADD_MEMBERSHIP:
282 return (IP_ADD_MEMBERSHIP);
283 case LINUX_IP_DROP_MEMBERSHIP:
284 return (IP_DROP_MEMBERSHIP);
285 case LINUX_IP_HDRINCL:
292 linux_to_bsd_ip6_sockopt(int opt)
296 case LINUX_IPV6_NEXTHOP:
297 return (IPV6_NEXTHOP);
298 case LINUX_IPV6_UNICAST_HOPS:
299 return (IPV6_UNICAST_HOPS);
300 case LINUX_IPV6_MULTICAST_IF:
301 return (IPV6_MULTICAST_IF);
302 case LINUX_IPV6_MULTICAST_HOPS:
303 return (IPV6_MULTICAST_HOPS);
304 case LINUX_IPV6_MULTICAST_LOOP:
305 return (IPV6_MULTICAST_LOOP);
306 case LINUX_IPV6_ADD_MEMBERSHIP:
307 return (IPV6_JOIN_GROUP);
308 case LINUX_IPV6_DROP_MEMBERSHIP:
309 return (IPV6_LEAVE_GROUP);
310 case LINUX_IPV6_V6ONLY:
311 return (IPV6_V6ONLY);
312 case LINUX_IPV6_DONTFRAG:
313 return (IPV6_DONTFRAG);
315 case LINUX_IPV6_CHECKSUM:
316 return (IPV6_CHECKSUM);
317 case LINUX_IPV6_RECVPKTINFO:
318 return (IPV6_RECVPKTINFO);
319 case LINUX_IPV6_PKTINFO:
320 return (IPV6_PKTINFO);
321 case LINUX_IPV6_RECVHOPLIMIT:
322 return (IPV6_RECVHOPLIMIT);
323 case LINUX_IPV6_HOPLIMIT:
324 return (IPV6_HOPLIMIT);
325 case LINUX_IPV6_RECVHOPOPTS:
326 return (IPV6_RECVHOPOPTS);
327 case LINUX_IPV6_HOPOPTS:
328 return (IPV6_HOPOPTS);
329 case LINUX_IPV6_RTHDRDSTOPTS:
330 return (IPV6_RTHDRDSTOPTS);
331 case LINUX_IPV6_RECVRTHDR:
332 return (IPV6_RECVRTHDR);
333 case LINUX_IPV6_RTHDR:
335 case LINUX_IPV6_RECVDSTOPTS:
336 return (IPV6_RECVDSTOPTS);
337 case LINUX_IPV6_DSTOPTS:
338 return (IPV6_DSTOPTS);
339 case LINUX_IPV6_RECVPATHMTU:
340 return (IPV6_RECVPATHMTU);
341 case LINUX_IPV6_PATHMTU:
342 return (IPV6_PATHMTU);
349 linux_to_bsd_so_sockopt(int opt)
355 case LINUX_SO_REUSEADDR:
356 return (SO_REUSEADDR);
361 case LINUX_SO_DONTROUTE:
362 return (SO_DONTROUTE);
363 case LINUX_SO_BROADCAST:
364 return (SO_BROADCAST);
365 case LINUX_SO_SNDBUF:
367 case LINUX_SO_RCVBUF:
369 case LINUX_SO_KEEPALIVE:
370 return (SO_KEEPALIVE);
371 case LINUX_SO_OOBINLINE:
372 return (SO_OOBINLINE);
373 case LINUX_SO_LINGER:
375 case LINUX_SO_PEERCRED:
376 return (LOCAL_PEERCRED);
377 case LINUX_SO_RCVLOWAT:
378 return (SO_RCVLOWAT);
379 case LINUX_SO_SNDLOWAT:
380 return (SO_SNDLOWAT);
381 case LINUX_SO_RCVTIMEO:
382 return (SO_RCVTIMEO);
383 case LINUX_SO_SNDTIMEO:
384 return (SO_SNDTIMEO);
385 case LINUX_SO_TIMESTAMP:
386 return (SO_TIMESTAMP);
387 case LINUX_SO_ACCEPTCONN:
388 return (SO_ACCEPTCONN);
394 linux_to_bsd_tcp_sockopt(int opt)
398 case LINUX_TCP_NODELAY:
399 return (TCP_NODELAY);
400 case LINUX_TCP_MAXSEG:
402 case LINUX_TCP_KEEPIDLE:
403 return (TCP_KEEPIDLE);
404 case LINUX_TCP_KEEPINTVL:
405 return (TCP_KEEPINTVL);
406 case LINUX_TCP_KEEPCNT:
407 return (TCP_KEEPCNT);
408 case LINUX_TCP_MD5SIG:
415 linux_to_bsd_msg_flags(int flags)
419 if (flags & LINUX_MSG_OOB)
420 ret_flags |= MSG_OOB;
421 if (flags & LINUX_MSG_PEEK)
422 ret_flags |= MSG_PEEK;
423 if (flags & LINUX_MSG_DONTROUTE)
424 ret_flags |= MSG_DONTROUTE;
425 if (flags & LINUX_MSG_CTRUNC)
426 ret_flags |= MSG_CTRUNC;
427 if (flags & LINUX_MSG_TRUNC)
428 ret_flags |= MSG_TRUNC;
429 if (flags & LINUX_MSG_DONTWAIT)
430 ret_flags |= MSG_DONTWAIT;
431 if (flags & LINUX_MSG_EOR)
432 ret_flags |= MSG_EOR;
433 if (flags & LINUX_MSG_WAITALL)
434 ret_flags |= MSG_WAITALL;
435 if (flags & LINUX_MSG_NOSIGNAL)
436 ret_flags |= MSG_NOSIGNAL;
437 #if 0 /* not handled */
438 if (flags & LINUX_MSG_PROXY)
440 if (flags & LINUX_MSG_FIN)
442 if (flags & LINUX_MSG_SYN)
444 if (flags & LINUX_MSG_CONFIRM)
446 if (flags & LINUX_MSG_RST)
448 if (flags & LINUX_MSG_ERRQUEUE)
455 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
456 * native syscall will fault. Thus, we don't really need to check the
457 * return values for these functions.
461 bsd_to_linux_sockaddr(struct sockaddr *arg)
464 size_t sa_len = sizeof(struct sockaddr);
467 if ((error = copyin(arg, &sa, sa_len)))
470 bdom = bsd_to_linux_domain(sa.sa_family);
472 return (EAFNOSUPPORT);
474 *(u_short *)&sa = bdom;
475 return (copyout(&sa, arg, sa_len));
479 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
482 size_t sa_len = sizeof(struct sockaddr);
485 if ((error = copyin(arg, &sa, sa_len)))
488 bdom = linux_to_bsd_domain(*(sa_family_t *)&sa);
490 return (EAFNOSUPPORT);
494 return (copyout(&sa, arg, sa_len));
498 linux_sa_put(struct osockaddr *osa)
504 * Only read/write the osockaddr family part, the rest is
507 error = copyin(osa, &sa, sizeof(sa.sa_family));
511 bdom = bsd_to_linux_domain(sa.sa_family);
516 return (copyout(&sa, osa, sizeof(sa.sa_family)));
520 linux_to_bsd_cmsg_type(int cmsg_type)
524 case LINUX_SCM_RIGHTS:
526 case LINUX_SCM_CREDENTIALS:
533 bsd_to_linux_cmsg_type(int cmsg_type)
538 return (LINUX_SCM_RIGHTS);
540 return (LINUX_SCM_CREDENTIALS);
542 return (LINUX_SCM_TIMESTAMP);
548 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
550 if (lhdr->msg_controllen > INT_MAX)
553 bhdr->msg_name = PTRIN(lhdr->msg_name);
554 bhdr->msg_namelen = lhdr->msg_namelen;
555 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
556 bhdr->msg_iovlen = lhdr->msg_iovlen;
557 bhdr->msg_control = PTRIN(lhdr->msg_control);
560 * msg_controllen is skipped since BSD and LINUX control messages
561 * are potentially different sizes (e.g. the cred structure used
562 * by SCM_CREDS is different between the two operating system).
564 * The caller can set it (if necessary) after converting all the
568 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
573 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
575 lhdr->msg_name = PTROUT(bhdr->msg_name);
576 lhdr->msg_namelen = bhdr->msg_namelen;
577 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
578 lhdr->msg_iovlen = bhdr->msg_iovlen;
579 lhdr->msg_control = PTROUT(bhdr->msg_control);
582 * msg_controllen is skipped since BSD and LINUX control messages
583 * are potentially different sizes (e.g. the cred structure used
584 * by SCM_CREDS is different between the two operating system).
586 * The caller can set it (if necessary) after converting all the
590 /* msg_flags skipped */
595 linux_set_socket_flags(int lflags, int *flags)
598 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
600 if (lflags & LINUX_SOCK_NONBLOCK)
601 *flags |= SOCK_NONBLOCK;
602 if (lflags & LINUX_SOCK_CLOEXEC)
603 *flags |= SOCK_CLOEXEC;
608 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
609 struct mbuf *control, enum uio_seg segflg)
614 if (mp->msg_name != NULL) {
615 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
622 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
630 /* Return 0 if IP_HDRINCL is set for the given socket. */
632 linux_check_hdrincl(struct thread *td, int s)
637 size_val = sizeof(optval);
638 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
639 &optval, UIO_SYSSPACE, &size_val);
643 return (optval == 0);
647 * Updated sendto() when IP_HDRINCL is set:
648 * tweak endian-dependent fields in the IP packet.
651 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
654 * linux_ip_copysize defines how many bytes we should copy
655 * from the beginning of the IP packet before we customize it for BSD.
656 * It should include all the fields we modify (ip_len and ip_off).
658 #define linux_ip_copysize 8
662 struct iovec aiov[1];
665 /* Check that the packet isn't too big or too small. */
666 if (linux_args->len < linux_ip_copysize ||
667 linux_args->len > IP_MAXPACKET)
670 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
672 /* Make kernel copy of the packet to be sent */
673 if ((error = copyin(PTRIN(linux_args->msg), packet,
677 /* Convert fields from Linux to BSD raw IP socket format */
678 packet->ip_len = linux_args->len;
679 packet->ip_off = ntohs(packet->ip_off);
681 /* Prepare the msghdr and iovec structures describing the new packet */
682 msg.msg_name = PTRIN(linux_args->to);
683 msg.msg_namelen = linux_args->tolen;
686 msg.msg_control = NULL;
688 aiov[0].iov_base = (char *)packet;
689 aiov[0].iov_len = linux_args->len;
690 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
693 free(packet, M_LINUX);
698 linux_socket(struct thread *td, struct linux_socket_args *args)
700 struct socket_args /* {
707 bsd_args.protocol = args->protocol;
708 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
709 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
711 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
713 if (retval_socket != 0)
714 return (retval_socket);
715 bsd_args.domain = linux_to_bsd_domain(args->domain);
716 if (bsd_args.domain == -1)
717 return (EAFNOSUPPORT);
719 retval_socket = sys_socket(td, &bsd_args);
721 return (retval_socket);
723 if (bsd_args.type == SOCK_RAW
724 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
725 && bsd_args.domain == PF_INET) {
726 /* It's a raw IP socket: set the IP_HDRINCL option. */
730 /* We ignore any error returned by kern_setsockopt() */
731 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
732 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
736 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
737 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
738 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
741 if (bsd_args.domain == PF_INET6) {
745 /* We ignore any error returned by setsockopt() */
746 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
747 &v6only, UIO_SYSSPACE, sizeof(v6only));
751 return (retval_socket);
755 linux_bind(struct thread *td, struct linux_bind_args *args)
760 error = linux_getsockaddr(&sa, PTRIN(args->name),
765 error = kern_bindat(td, AT_FDCWD, args->s, sa);
767 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
773 linux_connect(struct thread *td, struct linux_connect_args *args)
782 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
787 error = kern_connectat(td, AT_FDCWD, args->s, sa);
789 if (error != EISCONN)
793 * Linux doesn't return EISCONN the first time it occurs,
794 * when on a non-blocking socket. Instead it returns the
795 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
797 error = getsock_cap(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
804 if (fflag & FNONBLOCK) {
806 if (so->so_emuldata == 0)
807 error = so->so_error;
808 so->so_emuldata = (void *)1;
817 linux_listen(struct thread *td, struct linux_listen_args *args)
819 struct listen_args /* {
824 bsd_args.s = args->s;
825 bsd_args.backlog = args->backlog;
826 return (sys_listen(td, &bsd_args));
830 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
831 l_uintptr_t namelen, int flags)
833 struct accept4_args /* {
835 struct sockaddr * __restrict name;
836 socklen_t * __restrict anamelen;
846 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
847 bsd_args.anamelen = PTRIN(namelen);/* XXX */
849 error = linux_set_socket_flags(flags, &bsd_args.flags);
852 error = sys_accept4(td, &bsd_args);
853 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
855 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
857 if (error == EINVAL) {
858 error1 = getsock_cap(td, s, &rights, &fp, NULL, NULL);
862 if (so->so_type == SOCK_DGRAM) {
871 error = linux_sa_put(PTRIN(addr));
873 (void)kern_close(td, td->td_retval[0]);
874 td->td_retval[0] = 0;
880 linux_accept(struct thread *td, struct linux_accept_args *args)
883 return (linux_accept_common(td, args->s, args->addr,
888 linux_accept4(struct thread *td, struct linux_accept4_args *args)
891 return (linux_accept_common(td, args->s, args->addr,
892 args->namelen, args->flags));
896 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
898 struct getsockname_args /* {
900 struct sockaddr * __restrict asa;
901 socklen_t * __restrict alen;
905 bsd_args.fdes = args->s;
907 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
908 bsd_args.alen = PTRIN(args->namelen); /* XXX */
909 error = sys_getsockname(td, &bsd_args);
910 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
913 return (linux_sa_put(PTRIN(args->addr)));
917 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
919 struct getpeername_args /* {
926 bsd_args.fdes = args->s;
927 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
928 bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
929 error = sys_getpeername(td, &bsd_args);
930 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
933 return (linux_sa_put(PTRIN(args->addr)));
937 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
939 struct socketpair_args /* {
947 bsd_args.domain = linux_to_bsd_domain(args->domain);
948 if (bsd_args.domain != PF_LOCAL)
949 return (EAFNOSUPPORT);
950 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
951 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
953 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
957 if (args->protocol != 0 && args->protocol != PF_UNIX)
960 * Use of PF_UNIX as protocol argument is not right,
962 * Do not map PF_UNIX as its Linux value is identical
965 return (EPROTONOSUPPORT);
967 bsd_args.protocol = 0;
968 bsd_args.rsv = (int *)PTRIN(args->rsv);
969 return (sys_socketpair(td, &bsd_args));
972 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
973 struct linux_send_args {
981 linux_send(struct thread *td, struct linux_send_args *args)
983 struct sendto_args /* {
992 bsd_args.s = args->s;
993 bsd_args.buf = (caddr_t)PTRIN(args->msg);
994 bsd_args.len = args->len;
995 bsd_args.flags = args->flags;
998 return (sys_sendto(td, &bsd_args));
1001 struct linux_recv_args {
1009 linux_recv(struct thread *td, struct linux_recv_args *args)
1011 struct recvfrom_args /* {
1016 struct sockaddr *from;
1017 socklen_t fromlenaddr;
1020 bsd_args.s = args->s;
1021 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1022 bsd_args.len = args->len;
1023 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1024 bsd_args.from = NULL;
1025 bsd_args.fromlenaddr = 0;
1026 return (sys_recvfrom(td, &bsd_args));
1028 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1031 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1036 if (linux_check_hdrincl(td, args->s) == 0)
1037 /* IP_HDRINCL set, tweak the packet before sending */
1038 return (linux_sendto_hdrincl(td, args));
1040 msg.msg_name = PTRIN(args->to);
1041 msg.msg_namelen = args->tolen;
1042 msg.msg_iov = &aiov;
1044 msg.msg_control = NULL;
1046 aiov.iov_base = PTRIN(args->msg);
1047 aiov.iov_len = args->len;
1048 return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1053 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1059 if (PTRIN(args->fromlen) != NULL) {
1060 error = copyin(PTRIN(args->fromlen), &fromlen,
1066 msg.msg_namelen = fromlen;
1068 msg.msg_namelen = 0;
1070 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from);
1071 msg.msg_iov = &aiov;
1073 aiov.iov_base = PTRIN(args->buf);
1074 aiov.iov_len = args->len;
1075 msg.msg_control = 0;
1076 msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1078 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL);
1082 if (PTRIN(args->from) != NULL) {
1083 error = bsd_to_linux_sockaddr((struct sockaddr *)
1088 error = linux_sa_put((struct osockaddr *)
1092 if (PTRIN(args->fromlen) != NULL)
1093 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1094 sizeof(msg.msg_namelen));
1100 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1103 struct cmsghdr *cmsg;
1104 struct cmsgcred cmcred;
1105 struct mbuf *control;
1107 struct l_cmsghdr linux_cmsg;
1108 struct l_cmsghdr *ptr_cmsg;
1109 struct l_msghdr linux_msg;
1112 struct sockaddr *sa;
1113 sa_family_t sa_family;
1117 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1122 * Some Linux applications (ping) define a non-NULL control data
1123 * pointer, but a msg_controllen of 0, which is not allowed in the
1124 * FreeBSD system call interface. NULL the msg_control pointer in
1125 * order to handle this case. This should be checked, but allows the
1126 * Linux ping to work.
1128 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1129 linux_msg.msg_control = PTROUT(NULL);
1131 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1135 #ifdef COMPAT_LINUX32
1136 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1139 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1147 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1148 error = kern_getsockname(td, s, &sa, &datalen);
1151 sa_family = sa->sa_family;
1155 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO);
1156 control = m_get(M_WAITOK, MT_CONTROL);
1159 error = copyin(ptr_cmsg, &linux_cmsg,
1160 sizeof(struct l_cmsghdr));
1165 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1169 * Now we support only SCM_RIGHTS and SCM_CRED,
1170 * so return EINVAL in any other cmsg_type
1173 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1175 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1176 if (cmsg->cmsg_type == -1
1177 || cmsg->cmsg_level != SOL_SOCKET)
1181 * Some applications (e.g. pulseaudio) attempt to
1182 * send ancillary data even if the underlying protocol
1183 * doesn't support it which is not allowed in the
1184 * FreeBSD system call interface.
1186 if (sa_family != AF_UNIX)
1189 data = LINUX_CMSG_DATA(ptr_cmsg);
1190 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1192 switch (cmsg->cmsg_type)
1199 datalen = sizeof(cmcred);
1202 * The lower levels will fill in the structure
1204 bzero(data, datalen);
1208 cmsg->cmsg_len = CMSG_LEN(datalen);
1211 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1213 if (!m_append(control, datalen, (c_caddr_t)data))
1215 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1217 if (m_length(control, NULL) == 0) {
1225 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1232 free(cmsg, M_LINUX);
1237 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1240 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1245 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1247 struct l_mmsghdr *msg;
1249 int error, datagrams;
1251 if (args->vlen > UIO_MAXIOV)
1252 args->vlen = UIO_MAXIOV;
1254 msg = PTRIN(args->msg);
1256 while (datagrams < args->vlen) {
1257 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1262 retval = td->td_retval[0];
1263 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1270 td->td_retval[0] = datagrams;
1275 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1276 l_uint flags, struct msghdr *msg)
1279 struct cmsgcred *cmcred;
1280 struct l_cmsghdr *linux_cmsg = NULL;
1281 struct l_ucred linux_ucred;
1282 socklen_t datalen, outlen;
1283 struct l_msghdr linux_msg;
1284 struct iovec *iov, *uiov;
1285 struct mbuf *control = NULL;
1286 struct mbuf **controlp;
1287 struct timeval *ftmvl;
1291 int error, i, fd, fds, *fdp;
1293 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1297 error = linux_to_bsd_msghdr(msg, &linux_msg);
1301 #ifdef COMPAT_LINUX32
1302 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1305 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1310 if (msg->msg_name) {
1311 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name,
1317 uiov = msg->msg_iov;
1319 controlp = (msg->msg_control != NULL) ? &control : NULL;
1320 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp);
1321 msg->msg_iov = uiov;
1325 error = bsd_to_linux_msghdr(msg, &linux_msg);
1329 if (linux_msg.msg_name) {
1330 error = bsd_to_linux_sockaddr((struct sockaddr *)
1331 PTRIN(linux_msg.msg_name));
1335 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1336 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1341 outbuf = PTRIN(linux_msg.msg_control);
1345 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1347 msg->msg_control = mtod(control, struct cmsghdr *);
1348 msg->msg_controllen = control->m_len;
1350 cm = CMSG_FIRSTHDR(msg);
1352 while (cm != NULL) {
1353 linux_cmsg->cmsg_type =
1354 bsd_to_linux_cmsg_type(cm->cmsg_type);
1355 linux_cmsg->cmsg_level =
1356 bsd_to_linux_sockopt_level(cm->cmsg_level);
1357 if (linux_cmsg->cmsg_type == -1
1358 || cm->cmsg_level != SOL_SOCKET)
1364 data = CMSG_DATA(cm);
1365 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1367 switch (cm->cmsg_type)
1370 if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1371 fds = datalen / sizeof(int);
1373 for (i = 0; i < fds; i++) {
1375 (void)kern_fcntl(td, fd,
1376 F_SETFD, FD_CLOEXEC);
1383 * Currently LOCAL_CREDS is never in
1384 * effect for Linux so no need to worry
1387 if (datalen != sizeof(*cmcred)) {
1391 cmcred = (struct cmsgcred *)data;
1392 bzero(&linux_ucred, sizeof(linux_ucred));
1393 linux_ucred.pid = cmcred->cmcred_pid;
1394 linux_ucred.uid = cmcred->cmcred_uid;
1395 linux_ucred.gid = cmcred->cmcred_gid;
1396 data = &linux_ucred;
1397 datalen = sizeof(linux_ucred);
1401 if (datalen != sizeof(struct timeval)) {
1405 ftmvl = (struct timeval *)data;
1406 ltmvl.tv_sec = ftmvl->tv_sec;
1407 ltmvl.tv_usec = ftmvl->tv_usec;
1409 datalen = sizeof(ltmvl);
1413 if (outlen + LINUX_CMSG_LEN(datalen) >
1414 linux_msg.msg_controllen) {
1419 linux_msg.msg_flags |=
1425 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1427 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1430 outbuf += L_CMSG_HDRSZ;
1432 error = copyout(data, outbuf, datalen);
1436 outbuf += LINUX_CMSG_ALIGN(datalen);
1437 outlen += LINUX_CMSG_LEN(datalen);
1439 cm = CMSG_NXTHDR(msg, cm);
1444 linux_msg.msg_controllen = outlen;
1445 error = copyout(&linux_msg, msghdr, sizeof(linux_msg));
1450 free(linux_cmsg, M_LINUX);
1456 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1458 struct msghdr bsd_msg;
1460 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1461 args->flags, &bsd_msg));
1465 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1467 struct l_mmsghdr *msg;
1468 struct msghdr bsd_msg;
1469 struct l_timespec lts;
1470 struct timespec ts, tts;
1472 int error, datagrams;
1474 if (args->timeout) {
1475 error = copyin(args->timeout, <s, sizeof(struct l_timespec));
1478 error = linux_to_native_timespec(&ts, <s);
1482 timespecadd(&tts, &ts);
1485 msg = PTRIN(args->msg);
1487 while (datagrams < args->vlen) {
1488 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1489 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1493 retval = td->td_retval[0];
1494 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1501 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1503 if (args->flags & LINUX_MSG_WAITFORONE)
1504 args->flags |= LINUX_MSG_DONTWAIT;
1507 * See BUGS section of recvmmsg(2).
1509 if (args->timeout) {
1511 timespecsub(&ts, &tts);
1512 if (!timespecisset(&ts) || ts.tv_sec > 0)
1515 /* Out of band data, return right away. */
1516 if (bsd_msg.msg_flags & MSG_OOB)
1520 td->td_retval[0] = datagrams;
1525 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1527 struct shutdown_args /* {
1532 bsd_args.s = args->s;
1533 bsd_args.how = args->how;
1534 return (sys_shutdown(td, &bsd_args));
1538 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1540 struct setsockopt_args /* {
1551 bsd_args.s = args->s;
1552 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1553 switch (bsd_args.level) {
1555 name = linux_to_bsd_so_sockopt(args->optname);
1560 error = copyin(PTRIN(args->optval), &linux_tv,
1564 tv.tv_sec = linux_tv.tv_sec;
1565 tv.tv_usec = linux_tv.tv_usec;
1566 return (kern_setsockopt(td, args->s, bsd_args.level,
1567 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1575 name = linux_to_bsd_ip_sockopt(args->optname);
1578 name = linux_to_bsd_ip6_sockopt(args->optname);
1581 name = linux_to_bsd_tcp_sockopt(args->optname);
1588 return (ENOPROTOOPT);
1590 bsd_args.name = name;
1591 bsd_args.val = PTRIN(args->optval);
1592 bsd_args.valsize = args->optlen;
1594 if (name == IPV6_NEXTHOP) {
1595 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1597 error = sys_setsockopt(td, &bsd_args);
1598 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1600 error = sys_setsockopt(td, &bsd_args);
1606 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1608 struct getsockopt_args /* {
1617 socklen_t tv_len, xulen, len;
1620 int error, name, newval;
1622 bsd_args.s = args->s;
1623 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1624 switch (bsd_args.level) {
1626 name = linux_to_bsd_so_sockopt(args->optname);
1631 tv_len = sizeof(tv);
1632 error = kern_getsockopt(td, args->s, bsd_args.level,
1633 name, &tv, UIO_SYSSPACE, &tv_len);
1636 linux_tv.tv_sec = tv.tv_sec;
1637 linux_tv.tv_usec = tv.tv_usec;
1638 return (copyout(&linux_tv, PTRIN(args->optval),
1642 case LOCAL_PEERCRED:
1643 if (args->optlen != sizeof(lxu))
1646 error = kern_getsockopt(td, args->s, bsd_args.level,
1647 name, &xu, UIO_SYSSPACE, &xulen);
1651 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1654 lxu.uid = xu.cr_uid;
1655 lxu.gid = xu.cr_gid;
1656 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1660 len = sizeof(newval);
1661 error = kern_getsockopt(td, args->s, bsd_args.level,
1662 name, &newval, UIO_SYSSPACE, &len);
1665 newval = -SV_ABI_ERRNO(td->td_proc, newval);
1666 return (copyout(&newval, PTRIN(args->optval), len));
1673 name = linux_to_bsd_ip_sockopt(args->optname);
1676 name = linux_to_bsd_ip6_sockopt(args->optname);
1679 name = linux_to_bsd_tcp_sockopt(args->optname);
1688 bsd_args.name = name;
1689 bsd_args.val = PTRIN(args->optval);
1690 bsd_args.avalsize = PTRIN(args->optlen);
1692 if (name == IPV6_NEXTHOP) {
1693 error = sys_getsockopt(td, &bsd_args);
1694 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1696 error = sys_getsockopt(td, &bsd_args);
1701 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1703 /* Argument list sizes for linux_socketcall */
1705 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1707 static const unsigned char lxs_args[] = {
1708 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1709 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1710 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1711 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1712 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1713 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1714 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1715 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1716 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1717 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */,
1718 LINUX_AL(4) /* sendmmsg */
1721 #define LINUX_AL_SIZE (nitems(lxs_args) - 1)
1724 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1730 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1732 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1737 switch (args->what) {
1739 return (linux_socket(td, arg));
1741 return (linux_bind(td, arg));
1743 return (linux_connect(td, arg));
1745 return (linux_listen(td, arg));
1747 return (linux_accept(td, arg));
1748 case LINUX_GETSOCKNAME:
1749 return (linux_getsockname(td, arg));
1750 case LINUX_GETPEERNAME:
1751 return (linux_getpeername(td, arg));
1752 case LINUX_SOCKETPAIR:
1753 return (linux_socketpair(td, arg));
1755 return (linux_send(td, arg));
1757 return (linux_recv(td, arg));
1759 return (linux_sendto(td, arg));
1760 case LINUX_RECVFROM:
1761 return (linux_recvfrom(td, arg));
1762 case LINUX_SHUTDOWN:
1763 return (linux_shutdown(td, arg));
1764 case LINUX_SETSOCKOPT:
1765 return (linux_setsockopt(td, arg));
1766 case LINUX_GETSOCKOPT:
1767 return (linux_getsockopt(td, arg));
1769 return (linux_sendmsg(td, arg));
1771 return (linux_recvmsg(td, arg));
1773 return (linux_accept4(td, arg));
1774 case LINUX_RECVMMSG:
1775 return (linux_recvmmsg(td, arg));
1776 case LINUX_SENDMMSG:
1777 return (linux_sendmmsg(td, arg));
1780 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1783 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */