2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 1995 Søren Schmidt
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 /* XXX we use functions that might not exist. */
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
36 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/sysproto.h>
40 #include <sys/capsicum.h>
41 #include <sys/fcntl.h>
43 #include <sys/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 int domain, retval_socket, type;
702 type = args->type & LINUX_SOCK_TYPE_MASK;
703 if (type < 0 || type > LINUX_SOCK_MAX)
705 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
707 if (retval_socket != 0)
708 return (retval_socket);
709 domain = linux_to_bsd_domain(args->domain);
711 return (EAFNOSUPPORT);
713 retval_socket = kern_socket(td, domain, type, args->protocol);
715 return (retval_socket);
718 && (args->protocol == IPPROTO_RAW || args->protocol == 0)
719 && domain == PF_INET) {
720 /* It's a raw IP socket: set the IP_HDRINCL option. */
724 /* We ignore any error returned by kern_setsockopt() */
725 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
726 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
730 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
731 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
732 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
735 if (domain == PF_INET6) {
739 /* We ignore any error returned by setsockopt() */
740 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
741 &v6only, UIO_SYSSPACE, sizeof(v6only));
745 return (retval_socket);
749 linux_bind(struct thread *td, struct linux_bind_args *args)
754 error = linux_getsockaddr(&sa, PTRIN(args->name),
759 error = kern_bindat(td, AT_FDCWD, args->s, sa);
761 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
767 linux_connect(struct thread *td, struct linux_connect_args *args)
775 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
780 error = kern_connectat(td, AT_FDCWD, args->s, sa);
782 if (error != EISCONN)
786 * Linux doesn't return EISCONN the first time it occurs,
787 * when on a non-blocking socket. Instead it returns the
788 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
790 error = getsock_cap(td, args->s, &cap_connect_rights,
797 if (fflag & FNONBLOCK) {
799 if (so->so_emuldata == 0)
800 error = so->so_error;
801 so->so_emuldata = (void *)1;
810 linux_listen(struct thread *td, struct linux_listen_args *args)
813 return (kern_listen(td, args->s, args->backlog));
817 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
818 l_uintptr_t namelen, int flags)
820 struct accept4_args /* {
822 struct sockaddr * __restrict name;
823 socklen_t * __restrict anamelen;
831 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
832 bsd_args.anamelen = PTRIN(namelen);
834 error = linux_set_socket_flags(flags, &bsd_args.flags);
837 error = sys_accept4(td, &bsd_args);
838 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
840 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
842 if (error == EINVAL) {
843 error1 = getsock_cap(td, s, &cap_accept_rights, &fp, NULL, NULL);
847 if (so->so_type == SOCK_DGRAM) {
856 error = linux_sa_put(PTRIN(addr));
858 (void)kern_close(td, td->td_retval[0]);
859 td->td_retval[0] = 0;
865 linux_accept(struct thread *td, struct linux_accept_args *args)
868 return (linux_accept_common(td, args->s, args->addr,
873 linux_accept4(struct thread *td, struct linux_accept4_args *args)
876 return (linux_accept_common(td, args->s, args->addr,
877 args->namelen, args->flags));
881 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
883 struct getsockname_args /* {
885 struct sockaddr * __restrict asa;
886 socklen_t * __restrict alen;
890 bsd_args.fdes = args->s;
891 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
892 bsd_args.alen = PTRIN(args->namelen);
893 error = sys_getsockname(td, &bsd_args);
894 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
897 return (linux_sa_put(PTRIN(args->addr)));
901 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
903 struct getpeername_args /* {
910 bsd_args.fdes = args->s;
911 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
912 bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
913 error = sys_getpeername(td, &bsd_args);
914 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
917 return (linux_sa_put(PTRIN(args->addr)));
921 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
923 struct socketpair_args /* {
931 bsd_args.domain = linux_to_bsd_domain(args->domain);
932 if (bsd_args.domain != PF_LOCAL)
933 return (EAFNOSUPPORT);
934 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
935 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
937 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
941 if (args->protocol != 0 && args->protocol != PF_UNIX)
944 * Use of PF_UNIX as protocol argument is not right,
946 * Do not map PF_UNIX as its Linux value is identical
949 return (EPROTONOSUPPORT);
951 bsd_args.protocol = 0;
952 bsd_args.rsv = (int *)PTRIN(args->rsv);
953 return (sys_socketpair(td, &bsd_args));
956 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
957 struct linux_send_args {
965 linux_send(struct thread *td, struct linux_send_args *args)
967 struct sendto_args /* {
976 bsd_args.s = args->s;
977 bsd_args.buf = (caddr_t)PTRIN(args->msg);
978 bsd_args.len = args->len;
979 bsd_args.flags = args->flags;
982 return (sys_sendto(td, &bsd_args));
985 struct linux_recv_args {
993 linux_recv(struct thread *td, struct linux_recv_args *args)
995 struct recvfrom_args /* {
1000 struct sockaddr *from;
1001 socklen_t fromlenaddr;
1004 bsd_args.s = args->s;
1005 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1006 bsd_args.len = args->len;
1007 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1008 bsd_args.from = NULL;
1009 bsd_args.fromlenaddr = 0;
1010 return (sys_recvfrom(td, &bsd_args));
1012 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1015 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1020 if (linux_check_hdrincl(td, args->s) == 0)
1021 /* IP_HDRINCL set, tweak the packet before sending */
1022 return (linux_sendto_hdrincl(td, args));
1024 msg.msg_name = PTRIN(args->to);
1025 msg.msg_namelen = args->tolen;
1026 msg.msg_iov = &aiov;
1028 msg.msg_control = NULL;
1030 aiov.iov_base = PTRIN(args->msg);
1031 aiov.iov_len = args->len;
1032 return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1037 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1043 if (PTRIN(args->fromlen) != NULL) {
1044 error = copyin(PTRIN(args->fromlen), &fromlen,
1050 msg.msg_namelen = fromlen;
1052 msg.msg_namelen = 0;
1054 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from);
1055 msg.msg_iov = &aiov;
1057 aiov.iov_base = PTRIN(args->buf);
1058 aiov.iov_len = args->len;
1059 msg.msg_control = 0;
1060 msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1062 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL);
1066 if (PTRIN(args->from) != NULL) {
1067 error = bsd_to_linux_sockaddr((struct sockaddr *)
1072 error = linux_sa_put((struct osockaddr *)
1076 if (PTRIN(args->fromlen) != NULL)
1077 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1078 sizeof(msg.msg_namelen));
1084 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1087 struct cmsghdr *cmsg;
1088 struct mbuf *control;
1090 struct l_cmsghdr linux_cmsg;
1091 struct l_cmsghdr *ptr_cmsg;
1092 struct l_msghdr linux_msg;
1095 struct sockaddr *sa;
1096 sa_family_t sa_family;
1102 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1107 * Some Linux applications (ping) define a non-NULL control data
1108 * pointer, but a msg_controllen of 0, which is not allowed in the
1109 * FreeBSD system call interface. NULL the msg_control pointer in
1110 * order to handle this case. This should be checked, but allows the
1111 * Linux ping to work.
1113 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1114 linux_msg.msg_control = PTROUT(NULL);
1116 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1120 #ifdef COMPAT_LINUX32
1121 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1124 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1131 if (linux_msg.msg_controllen >= sizeof(struct l_cmsghdr)) {
1132 error = kern_getsockname(td, s, &sa, &datalen);
1135 sa_family = sa->sa_family;
1139 control = m_get(M_WAITOK, MT_CONTROL);
1140 MCLGET(control, M_WAITOK);
1141 data = mtod(control, void *);
1144 ptr_cmsg = PTRIN(linux_msg.msg_control);
1145 clen = linux_msg.msg_controllen;
1147 error = copyin(ptr_cmsg, &linux_cmsg,
1148 sizeof(struct l_cmsghdr));
1153 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr) ||
1154 linux_cmsg.cmsg_len > clen)
1157 if (datalen + CMSG_HDRSZ > MCLBYTES)
1161 * Now we support only SCM_RIGHTS and SCM_CRED,
1162 * so return EINVAL in any other cmsg_type
1166 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1168 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1169 if (cmsg->cmsg_type == -1
1170 || cmsg->cmsg_level != SOL_SOCKET)
1174 * Some applications (e.g. pulseaudio) attempt to
1175 * send ancillary data even if the underlying protocol
1176 * doesn't support it which is not allowed in the
1177 * FreeBSD system call interface.
1179 if (sa_family != AF_UNIX)
1182 if (cmsg->cmsg_type == SCM_CREDS) {
1183 len = sizeof(struct cmsgcred);
1184 if (datalen + CMSG_SPACE(len) > MCLBYTES)
1188 * The lower levels will fill in the structure
1190 memset(CMSG_DATA(data), 0, len);
1192 len = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1193 if (datalen + CMSG_SPACE(len) < datalen ||
1194 datalen + CMSG_SPACE(len) > MCLBYTES)
1197 error = copyin(LINUX_CMSG_DATA(ptr_cmsg),
1198 CMSG_DATA(data), len);
1203 cmsg->cmsg_len = CMSG_LEN(len);
1204 data = (char *)data + CMSG_SPACE(len);
1205 datalen += CMSG_SPACE(len);
1207 if (clen <= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len))
1210 clen -= LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len);
1211 ptr_cmsg = (struct l_cmsghdr *)((char *)ptr_cmsg +
1212 LINUX_CMSG_ALIGN(linux_cmsg.cmsg_len));
1213 } while(clen >= sizeof(struct l_cmsghdr));
1215 control->m_len = datalen;
1224 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1234 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1237 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1242 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1244 struct l_mmsghdr *msg;
1246 int error, datagrams;
1248 if (args->vlen > UIO_MAXIOV)
1249 args->vlen = UIO_MAXIOV;
1251 msg = PTRIN(args->msg);
1253 while (datagrams < args->vlen) {
1254 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1259 retval = td->td_retval[0];
1260 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1267 td->td_retval[0] = datagrams;
1272 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1273 l_uint flags, struct msghdr *msg)
1276 struct cmsgcred *cmcred;
1277 struct l_cmsghdr *linux_cmsg = NULL;
1278 struct l_ucred linux_ucred;
1279 socklen_t datalen, maxlen, outlen;
1280 struct l_msghdr linux_msg;
1281 struct iovec *iov, *uiov;
1282 struct mbuf *control = NULL;
1283 struct mbuf **controlp;
1284 struct timeval *ftmvl;
1288 int error, i, fd, fds, *fdp;
1290 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1294 error = linux_to_bsd_msghdr(msg, &linux_msg);
1298 #ifdef COMPAT_LINUX32
1299 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1302 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1307 if (msg->msg_name) {
1308 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name,
1314 uiov = msg->msg_iov;
1316 controlp = (msg->msg_control != NULL) ? &control : NULL;
1317 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp);
1318 msg->msg_iov = uiov;
1322 error = bsd_to_linux_msghdr(msg, &linux_msg);
1326 if (linux_msg.msg_name) {
1327 error = bsd_to_linux_sockaddr((struct sockaddr *)
1328 PTRIN(linux_msg.msg_name));
1332 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1333 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1338 maxlen = linux_msg.msg_controllen;
1339 linux_msg.msg_controllen = 0;
1341 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1343 msg->msg_control = mtod(control, struct cmsghdr *);
1344 msg->msg_controllen = control->m_len;
1346 cm = CMSG_FIRSTHDR(msg);
1347 outbuf = PTRIN(linux_msg.msg_control);
1349 while (cm != NULL) {
1350 linux_cmsg->cmsg_type =
1351 bsd_to_linux_cmsg_type(cm->cmsg_type);
1352 linux_cmsg->cmsg_level =
1353 bsd_to_linux_sockopt_level(cm->cmsg_level);
1354 if (linux_cmsg->cmsg_type == -1 ||
1355 cm->cmsg_level != SOL_SOCKET) {
1360 data = CMSG_DATA(cm);
1361 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1363 switch (cm->cmsg_type) {
1365 if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1366 fds = datalen / sizeof(int);
1368 for (i = 0; i < fds; i++) {
1370 (void)kern_fcntl(td, fd,
1371 F_SETFD, FD_CLOEXEC);
1378 * Currently LOCAL_CREDS is never in
1379 * effect for Linux so no need to worry
1382 if (datalen != sizeof(*cmcred)) {
1386 cmcred = (struct cmsgcred *)data;
1387 bzero(&linux_ucred, sizeof(linux_ucred));
1388 linux_ucred.pid = cmcred->cmcred_pid;
1389 linux_ucred.uid = cmcred->cmcred_uid;
1390 linux_ucred.gid = cmcred->cmcred_gid;
1391 data = &linux_ucred;
1392 datalen = sizeof(linux_ucred);
1396 if (datalen != sizeof(struct timeval)) {
1400 ftmvl = (struct timeval *)data;
1401 ltmvl.tv_sec = ftmvl->tv_sec;
1402 ltmvl.tv_usec = ftmvl->tv_usec;
1404 datalen = sizeof(ltmvl);
1408 if (outlen + LINUX_CMSG_LEN(datalen) > maxlen) {
1413 linux_msg.msg_flags |= LINUX_MSG_CTRUNC;
1414 m_dispose_extcontrolm(control);
1419 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1421 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1424 outbuf += L_CMSG_HDRSZ;
1426 error = copyout(data, outbuf, datalen);
1430 outbuf += LINUX_CMSG_ALIGN(datalen);
1431 outlen += LINUX_CMSG_LEN(datalen);
1433 cm = CMSG_NXTHDR(msg, cm);
1435 linux_msg.msg_controllen = outlen;
1439 error = copyout(&linux_msg, msghdr, sizeof(linux_msg));
1442 if (control != NULL) {
1444 m_dispose_extcontrolm(control);
1448 free(linux_cmsg, M_LINUX);
1454 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1456 struct msghdr bsd_msg;
1458 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1459 args->flags, &bsd_msg));
1463 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1465 struct l_mmsghdr *msg;
1466 struct msghdr bsd_msg;
1467 struct l_timespec lts;
1468 struct timespec ts, tts;
1470 int error, datagrams;
1472 if (args->timeout) {
1473 error = copyin(args->timeout, <s, sizeof(struct l_timespec));
1476 error = linux_to_native_timespec(&ts, <s);
1480 timespecadd(&tts, &ts, &tts);
1483 msg = PTRIN(args->msg);
1485 while (datagrams < args->vlen) {
1486 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1487 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1491 retval = td->td_retval[0];
1492 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1499 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1501 if (args->flags & LINUX_MSG_WAITFORONE)
1502 args->flags |= LINUX_MSG_DONTWAIT;
1505 * See BUGS section of recvmmsg(2).
1507 if (args->timeout) {
1509 timespecsub(&ts, &tts, &ts);
1510 if (!timespecisset(&ts) || ts.tv_sec > 0)
1513 /* Out of band data, return right away. */
1514 if (bsd_msg.msg_flags & MSG_OOB)
1518 td->td_retval[0] = datagrams;
1523 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1526 return (kern_shutdown(td, args->s, args->how));
1530 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1532 struct setsockopt_args /* {
1543 bsd_args.s = args->s;
1544 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1545 switch (bsd_args.level) {
1547 name = linux_to_bsd_so_sockopt(args->optname);
1552 error = copyin(PTRIN(args->optval), &linux_tv,
1556 tv.tv_sec = linux_tv.tv_sec;
1557 tv.tv_usec = linux_tv.tv_usec;
1558 return (kern_setsockopt(td, args->s, bsd_args.level,
1559 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1566 name = linux_to_bsd_ip_sockopt(args->optname);
1569 name = linux_to_bsd_ip6_sockopt(args->optname);
1572 name = linux_to_bsd_tcp_sockopt(args->optname);
1579 return (ENOPROTOOPT);
1581 bsd_args.name = name;
1582 bsd_args.val = PTRIN(args->optval);
1583 bsd_args.valsize = args->optlen;
1585 if (name == IPV6_NEXTHOP) {
1586 linux_to_bsd_sockaddr(__DECONST(struct sockaddr *,
1587 bsd_args.val), bsd_args.valsize);
1588 error = sys_setsockopt(td, &bsd_args);
1589 bsd_to_linux_sockaddr(__DECONST(struct sockaddr *,
1592 error = sys_setsockopt(td, &bsd_args);
1598 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1600 struct getsockopt_args /* {
1609 socklen_t tv_len, xulen, len;
1612 int error, name, newval;
1614 bsd_args.s = args->s;
1615 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1616 switch (bsd_args.level) {
1618 name = linux_to_bsd_so_sockopt(args->optname);
1623 tv_len = sizeof(tv);
1624 error = kern_getsockopt(td, args->s, bsd_args.level,
1625 name, &tv, UIO_SYSSPACE, &tv_len);
1628 linux_tv.tv_sec = tv.tv_sec;
1629 linux_tv.tv_usec = tv.tv_usec;
1630 return (copyout(&linux_tv, PTRIN(args->optval),
1633 case LOCAL_PEERCRED:
1634 if (args->optlen < sizeof(lxu))
1637 * LOCAL_PEERCRED is not served at the SOL_SOCKET level,
1638 * but by the Unix socket's level 0.
1642 error = kern_getsockopt(td, args->s, bsd_args.level,
1643 name, &xu, UIO_SYSSPACE, &xulen);
1647 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1650 lxu.uid = xu.cr_uid;
1651 lxu.gid = xu.cr_gid;
1652 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1655 len = sizeof(newval);
1656 error = kern_getsockopt(td, args->s, bsd_args.level,
1657 name, &newval, UIO_SYSSPACE, &len);
1660 newval = -SV_ABI_ERRNO(td->td_proc, newval);
1661 return (copyout(&newval, PTRIN(args->optval), len));
1668 name = linux_to_bsd_ip_sockopt(args->optname);
1671 name = linux_to_bsd_ip6_sockopt(args->optname);
1674 name = linux_to_bsd_tcp_sockopt(args->optname);
1683 bsd_args.name = name;
1684 bsd_args.val = PTRIN(args->optval);
1685 bsd_args.avalsize = PTRIN(args->optlen);
1687 if (name == IPV6_NEXTHOP) {
1688 error = sys_getsockopt(td, &bsd_args);
1689 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1691 error = sys_getsockopt(td, &bsd_args);
1696 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1698 /* Argument list sizes for linux_socketcall */
1699 static const unsigned char lxs_args_cnt[] = {
1700 0 /* unused*/, 3 /* socket */,
1701 3 /* bind */, 3 /* connect */,
1702 2 /* listen */, 3 /* accept */,
1703 3 /* getsockname */, 3 /* getpeername */,
1704 4 /* socketpair */, 4 /* send */,
1705 4 /* recv */, 6 /* sendto */,
1706 6 /* recvfrom */, 2 /* shutdown */,
1707 5 /* setsockopt */, 5 /* getsockopt */,
1708 3 /* sendmsg */, 3 /* recvmsg */,
1709 4 /* accept4 */, 5 /* recvmmsg */,
1712 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1)
1713 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong))
1716 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1719 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1720 register_t l_args[6];
1725 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
1727 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
1731 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1732 for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
1738 switch (args->what) {
1740 return (linux_socket(td, arg));
1742 return (linux_bind(td, arg));
1744 return (linux_connect(td, arg));
1746 return (linux_listen(td, arg));
1748 return (linux_accept(td, arg));
1749 case LINUX_GETSOCKNAME:
1750 return (linux_getsockname(td, arg));
1751 case LINUX_GETPEERNAME:
1752 return (linux_getpeername(td, arg));
1753 case LINUX_SOCKETPAIR:
1754 return (linux_socketpair(td, arg));
1756 return (linux_send(td, arg));
1758 return (linux_recv(td, arg));
1760 return (linux_sendto(td, arg));
1761 case LINUX_RECVFROM:
1762 return (linux_recvfrom(td, arg));
1763 case LINUX_SHUTDOWN:
1764 return (linux_shutdown(td, arg));
1765 case LINUX_SETSOCKOPT:
1766 return (linux_setsockopt(td, arg));
1767 case LINUX_GETSOCKOPT:
1768 return (linux_getsockopt(td, arg));
1770 return (linux_sendmsg(td, arg));
1772 return (linux_recvmsg(td, arg));
1774 return (linux_accept4(td, arg));
1775 case LINUX_RECVMMSG:
1776 return (linux_recvmmsg(td, arg));
1777 case LINUX_SENDMMSG:
1778 return (linux_sendmmsg(td, arg));
1781 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1784 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */