2 * SPDX-License-Identifier: BSD-3-Clause
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 * in this position and unchanged.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. The name of the author may not be used to endorse or promote products
17 * derived from this software without specific prior written permission
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
20 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
21 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
22 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
23 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
24 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
28 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
34 /* XXX we use functions that might not exist. */
35 #include "opt_compat.h"
36 #include "opt_inet6.h"
38 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/sysproto.h>
42 #include <sys/capsicum.h>
43 #include <sys/fcntl.h>
45 #include <sys/limits.h>
47 #include <sys/malloc.h>
48 #include <sys/mutex.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/syscallsubr.h>
54 #include <sys/syslog.h>
59 #include <netinet/in.h>
60 #include <netinet/in_systm.h>
61 #include <netinet/ip.h>
62 #include <netinet/tcp.h>
64 #include <netinet/ip6.h>
65 #include <netinet6/ip6_var.h>
69 #include <machine/../linux32/linux.h>
70 #include <machine/../linux32/linux32_proto.h>
72 #include <machine/../linux/linux.h>
73 #include <machine/../linux/linux_proto.h>
75 #include <compat/linux/linux_file.h>
76 #include <compat/linux/linux_socket.h>
77 #include <compat/linux/linux_timer.h>
78 #include <compat/linux/linux_util.h>
80 static int linux_to_bsd_domain(int);
81 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
83 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
84 l_uint, struct msghdr *);
85 static int linux_set_socket_flags(int, int *);
88 * Reads a linux sockaddr and does any necessary translation.
89 * Linux sockaddrs don't have a length field, only a family.
90 * Copy the osockaddr structure pointed to by osa to kernel, adjust
91 * family and convert to sockaddr.
94 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
97 struct osockaddr *kosa;
99 struct sockaddr_in6 *sin6;
103 int bdom, error, hdrlen, namelen;
105 if (salen < 2 || salen > UCHAR_MAX || !osa)
111 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
112 * if it's a v4-mapped address, so reserve the proper space
115 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
116 salen += sizeof(uint32_t);
121 kosa = malloc(salen, M_SONAME, M_WAITOK);
123 if ((error = copyin(osa, kosa, salen)))
126 bdom = linux_to_bsd_domain(kosa->sa_family);
128 error = EAFNOSUPPORT;
134 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
135 * which lacks the scope id compared with RFC2553 one. If we detect
136 * the situation, reject the address and write a message to system log.
138 * Still accept addresses for which the scope id is not used.
141 if (bdom == AF_INET6) {
142 sin6 = (struct sockaddr_in6 *)kosa;
143 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
144 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
145 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
146 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
147 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
148 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
149 sin6->sin6_scope_id = 0;
152 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
157 salen -= sizeof(uint32_t);
160 if (bdom == AF_INET) {
161 if (salen < sizeof(struct sockaddr_in)) {
165 salen = sizeof(struct sockaddr_in);
168 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
169 hdrlen = offsetof(struct sockaddr_un, sun_path);
170 name = ((struct sockaddr_un *)kosa)->sun_path;
173 * Linux abstract namespace starts with a NULL byte.
174 * XXX We do not support abstract namespace yet.
176 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
178 namelen = strnlen(name, salen - hdrlen);
179 salen = hdrlen + namelen;
180 if (salen > sizeof(struct sockaddr_un)) {
181 error = ENAMETOOLONG;
186 sa = (struct sockaddr *)kosa;
187 sa->sa_family = bdom;
194 free(kosa, M_SONAME);
199 linux_to_bsd_domain(int domain)
203 case LINUX_AF_UNSPEC:
215 case LINUX_AF_APPLETALK:
216 return (AF_APPLETALK);
222 bsd_to_linux_domain(int domain)
227 return (LINUX_AF_UNSPEC);
229 return (LINUX_AF_UNIX);
231 return (LINUX_AF_INET);
233 return (LINUX_AF_INET6);
235 return (LINUX_AF_AX25);
237 return (LINUX_AF_IPX);
239 return (LINUX_AF_APPLETALK);
245 linux_to_bsd_sockopt_level(int level)
249 case LINUX_SOL_SOCKET:
256 bsd_to_linux_sockopt_level(int level)
261 return (LINUX_SOL_SOCKET);
267 linux_to_bsd_ip_sockopt(int opt)
275 case LINUX_IP_OPTIONS:
277 case LINUX_IP_MULTICAST_IF:
278 return (IP_MULTICAST_IF);
279 case LINUX_IP_MULTICAST_TTL:
280 return (IP_MULTICAST_TTL);
281 case LINUX_IP_MULTICAST_LOOP:
282 return (IP_MULTICAST_LOOP);
283 case LINUX_IP_ADD_MEMBERSHIP:
284 return (IP_ADD_MEMBERSHIP);
285 case LINUX_IP_DROP_MEMBERSHIP:
286 return (IP_DROP_MEMBERSHIP);
287 case LINUX_IP_HDRINCL:
294 linux_to_bsd_ip6_sockopt(int opt)
298 case LINUX_IPV6_NEXTHOP:
299 return (IPV6_NEXTHOP);
300 case LINUX_IPV6_UNICAST_HOPS:
301 return (IPV6_UNICAST_HOPS);
302 case LINUX_IPV6_MULTICAST_IF:
303 return (IPV6_MULTICAST_IF);
304 case LINUX_IPV6_MULTICAST_HOPS:
305 return (IPV6_MULTICAST_HOPS);
306 case LINUX_IPV6_MULTICAST_LOOP:
307 return (IPV6_MULTICAST_LOOP);
308 case LINUX_IPV6_ADD_MEMBERSHIP:
309 return (IPV6_JOIN_GROUP);
310 case LINUX_IPV6_DROP_MEMBERSHIP:
311 return (IPV6_LEAVE_GROUP);
312 case LINUX_IPV6_V6ONLY:
313 return (IPV6_V6ONLY);
314 case LINUX_IPV6_DONTFRAG:
315 return (IPV6_DONTFRAG);
317 case LINUX_IPV6_CHECKSUM:
318 return (IPV6_CHECKSUM);
319 case LINUX_IPV6_RECVPKTINFO:
320 return (IPV6_RECVPKTINFO);
321 case LINUX_IPV6_PKTINFO:
322 return (IPV6_PKTINFO);
323 case LINUX_IPV6_RECVHOPLIMIT:
324 return (IPV6_RECVHOPLIMIT);
325 case LINUX_IPV6_HOPLIMIT:
326 return (IPV6_HOPLIMIT);
327 case LINUX_IPV6_RECVHOPOPTS:
328 return (IPV6_RECVHOPOPTS);
329 case LINUX_IPV6_HOPOPTS:
330 return (IPV6_HOPOPTS);
331 case LINUX_IPV6_RTHDRDSTOPTS:
332 return (IPV6_RTHDRDSTOPTS);
333 case LINUX_IPV6_RECVRTHDR:
334 return (IPV6_RECVRTHDR);
335 case LINUX_IPV6_RTHDR:
337 case LINUX_IPV6_RECVDSTOPTS:
338 return (IPV6_RECVDSTOPTS);
339 case LINUX_IPV6_DSTOPTS:
340 return (IPV6_DSTOPTS);
341 case LINUX_IPV6_RECVPATHMTU:
342 return (IPV6_RECVPATHMTU);
343 case LINUX_IPV6_PATHMTU:
344 return (IPV6_PATHMTU);
351 linux_to_bsd_so_sockopt(int opt)
357 case LINUX_SO_REUSEADDR:
358 return (SO_REUSEADDR);
363 case LINUX_SO_DONTROUTE:
364 return (SO_DONTROUTE);
365 case LINUX_SO_BROADCAST:
366 return (SO_BROADCAST);
367 case LINUX_SO_SNDBUF:
369 case LINUX_SO_RCVBUF:
371 case LINUX_SO_KEEPALIVE:
372 return (SO_KEEPALIVE);
373 case LINUX_SO_OOBINLINE:
374 return (SO_OOBINLINE);
375 case LINUX_SO_LINGER:
377 case LINUX_SO_PEERCRED:
378 return (LOCAL_PEERCRED);
379 case LINUX_SO_RCVLOWAT:
380 return (SO_RCVLOWAT);
381 case LINUX_SO_SNDLOWAT:
382 return (SO_SNDLOWAT);
383 case LINUX_SO_RCVTIMEO:
384 return (SO_RCVTIMEO);
385 case LINUX_SO_SNDTIMEO:
386 return (SO_SNDTIMEO);
387 case LINUX_SO_TIMESTAMP:
388 return (SO_TIMESTAMP);
389 case LINUX_SO_ACCEPTCONN:
390 return (SO_ACCEPTCONN);
396 linux_to_bsd_tcp_sockopt(int opt)
400 case LINUX_TCP_NODELAY:
401 return (TCP_NODELAY);
402 case LINUX_TCP_MAXSEG:
404 case LINUX_TCP_KEEPIDLE:
405 return (TCP_KEEPIDLE);
406 case LINUX_TCP_KEEPINTVL:
407 return (TCP_KEEPINTVL);
408 case LINUX_TCP_KEEPCNT:
409 return (TCP_KEEPCNT);
410 case LINUX_TCP_MD5SIG:
417 linux_to_bsd_msg_flags(int flags)
421 if (flags & LINUX_MSG_OOB)
422 ret_flags |= MSG_OOB;
423 if (flags & LINUX_MSG_PEEK)
424 ret_flags |= MSG_PEEK;
425 if (flags & LINUX_MSG_DONTROUTE)
426 ret_flags |= MSG_DONTROUTE;
427 if (flags & LINUX_MSG_CTRUNC)
428 ret_flags |= MSG_CTRUNC;
429 if (flags & LINUX_MSG_TRUNC)
430 ret_flags |= MSG_TRUNC;
431 if (flags & LINUX_MSG_DONTWAIT)
432 ret_flags |= MSG_DONTWAIT;
433 if (flags & LINUX_MSG_EOR)
434 ret_flags |= MSG_EOR;
435 if (flags & LINUX_MSG_WAITALL)
436 ret_flags |= MSG_WAITALL;
437 if (flags & LINUX_MSG_NOSIGNAL)
438 ret_flags |= MSG_NOSIGNAL;
439 #if 0 /* not handled */
440 if (flags & LINUX_MSG_PROXY)
442 if (flags & LINUX_MSG_FIN)
444 if (flags & LINUX_MSG_SYN)
446 if (flags & LINUX_MSG_CONFIRM)
448 if (flags & LINUX_MSG_RST)
450 if (flags & LINUX_MSG_ERRQUEUE)
457 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
458 * native syscall will fault. Thus, we don't really need to check the
459 * return values for these functions.
463 bsd_to_linux_sockaddr(struct sockaddr *arg)
466 size_t sa_len = sizeof(struct sockaddr);
469 if ((error = copyin(arg, &sa, sa_len)))
472 bdom = bsd_to_linux_domain(sa.sa_family);
474 return (EAFNOSUPPORT);
476 *(u_short *)&sa = bdom;
477 return (copyout(&sa, arg, sa_len));
481 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
484 size_t sa_len = sizeof(struct sockaddr);
487 if ((error = copyin(arg, &sa, sa_len)))
490 bdom = linux_to_bsd_domain(*(sa_family_t *)&sa);
492 return (EAFNOSUPPORT);
496 return (copyout(&sa, arg, sa_len));
500 linux_sa_put(struct osockaddr *osa)
506 * Only read/write the osockaddr family part, the rest is
509 error = copyin(osa, &sa, sizeof(sa.sa_family));
513 bdom = bsd_to_linux_domain(sa.sa_family);
518 return (copyout(&sa, osa, sizeof(sa.sa_family)));
522 linux_to_bsd_cmsg_type(int cmsg_type)
526 case LINUX_SCM_RIGHTS:
528 case LINUX_SCM_CREDENTIALS:
535 bsd_to_linux_cmsg_type(int cmsg_type)
540 return (LINUX_SCM_RIGHTS);
542 return (LINUX_SCM_CREDENTIALS);
544 return (LINUX_SCM_TIMESTAMP);
550 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
552 if (lhdr->msg_controllen > INT_MAX)
555 bhdr->msg_name = PTRIN(lhdr->msg_name);
556 bhdr->msg_namelen = lhdr->msg_namelen;
557 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
558 bhdr->msg_iovlen = lhdr->msg_iovlen;
559 bhdr->msg_control = PTRIN(lhdr->msg_control);
562 * msg_controllen is skipped since BSD and LINUX control messages
563 * are potentially different sizes (e.g. the cred structure used
564 * by SCM_CREDS is different between the two operating system).
566 * The caller can set it (if necessary) after converting all the
570 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
575 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
577 lhdr->msg_name = PTROUT(bhdr->msg_name);
578 lhdr->msg_namelen = bhdr->msg_namelen;
579 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
580 lhdr->msg_iovlen = bhdr->msg_iovlen;
581 lhdr->msg_control = PTROUT(bhdr->msg_control);
584 * msg_controllen is skipped since BSD and LINUX control messages
585 * are potentially different sizes (e.g. the cred structure used
586 * by SCM_CREDS is different between the two operating system).
588 * The caller can set it (if necessary) after converting all the
592 /* msg_flags skipped */
597 linux_set_socket_flags(int lflags, int *flags)
600 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
602 if (lflags & LINUX_SOCK_NONBLOCK)
603 *flags |= SOCK_NONBLOCK;
604 if (lflags & LINUX_SOCK_CLOEXEC)
605 *flags |= SOCK_CLOEXEC;
610 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
611 struct mbuf *control, enum uio_seg segflg)
616 if (mp->msg_name != NULL) {
617 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
624 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
632 /* Return 0 if IP_HDRINCL is set for the given socket. */
634 linux_check_hdrincl(struct thread *td, int s)
639 size_val = sizeof(optval);
640 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
641 &optval, UIO_SYSSPACE, &size_val);
645 return (optval == 0);
649 * Updated sendto() when IP_HDRINCL is set:
650 * tweak endian-dependent fields in the IP packet.
653 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
656 * linux_ip_copysize defines how many bytes we should copy
657 * from the beginning of the IP packet before we customize it for BSD.
658 * It should include all the fields we modify (ip_len and ip_off).
660 #define linux_ip_copysize 8
664 struct iovec aiov[1];
667 /* Check that the packet isn't too big or too small. */
668 if (linux_args->len < linux_ip_copysize ||
669 linux_args->len > IP_MAXPACKET)
672 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
674 /* Make kernel copy of the packet to be sent */
675 if ((error = copyin(PTRIN(linux_args->msg), packet,
679 /* Convert fields from Linux to BSD raw IP socket format */
680 packet->ip_len = linux_args->len;
681 packet->ip_off = ntohs(packet->ip_off);
683 /* Prepare the msghdr and iovec structures describing the new packet */
684 msg.msg_name = PTRIN(linux_args->to);
685 msg.msg_namelen = linux_args->tolen;
688 msg.msg_control = NULL;
690 aiov[0].iov_base = (char *)packet;
691 aiov[0].iov_len = linux_args->len;
692 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
695 free(packet, M_LINUX);
700 linux_socket(struct thread *td, struct linux_socket_args *args)
702 int domain, retval_socket, type;
704 type = args->type & LINUX_SOCK_TYPE_MASK;
705 if (type < 0 || type > LINUX_SOCK_MAX)
707 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
709 if (retval_socket != 0)
710 return (retval_socket);
711 domain = linux_to_bsd_domain(args->domain);
713 return (EAFNOSUPPORT);
715 retval_socket = kern_socket(td, domain, type, args->protocol);
717 return (retval_socket);
720 && (args->protocol == IPPROTO_RAW || args->protocol == 0)
721 && domain == PF_INET) {
722 /* It's a raw IP socket: set the IP_HDRINCL option. */
726 /* We ignore any error returned by kern_setsockopt() */
727 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
728 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
732 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
733 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
734 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
737 if (domain == PF_INET6) {
741 /* We ignore any error returned by setsockopt() */
742 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
743 &v6only, UIO_SYSSPACE, sizeof(v6only));
747 return (retval_socket);
751 linux_bind(struct thread *td, struct linux_bind_args *args)
756 error = linux_getsockaddr(&sa, PTRIN(args->name),
761 error = kern_bindat(td, AT_FDCWD, args->s, sa);
763 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
769 linux_connect(struct thread *td, struct linux_connect_args *args)
778 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
783 error = kern_connectat(td, AT_FDCWD, args->s, sa);
785 if (error != EISCONN)
789 * Linux doesn't return EISCONN the first time it occurs,
790 * when on a non-blocking socket. Instead it returns the
791 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
793 error = getsock_cap(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
800 if (fflag & FNONBLOCK) {
802 if (so->so_emuldata == 0)
803 error = so->so_error;
804 so->so_emuldata = (void *)1;
813 linux_listen(struct thread *td, struct linux_listen_args *args)
816 return (kern_listen(td, args->s, args->backlog));
820 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
821 l_uintptr_t namelen, int flags)
823 struct accept4_args /* {
825 struct sockaddr * __restrict name;
826 socklen_t * __restrict anamelen;
835 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
836 bsd_args.anamelen = PTRIN(namelen);
838 error = linux_set_socket_flags(flags, &bsd_args.flags);
841 error = sys_accept4(td, &bsd_args);
842 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
844 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
846 if (error == EINVAL) {
847 error1 = getsock_cap(td, s,
848 cap_rights_init(&rights, CAP_ACCEPT), &fp, NULL, NULL);
852 if (so->so_type == SOCK_DGRAM) {
861 error = linux_sa_put(PTRIN(addr));
863 (void)kern_close(td, td->td_retval[0]);
864 td->td_retval[0] = 0;
870 linux_accept(struct thread *td, struct linux_accept_args *args)
873 return (linux_accept_common(td, args->s, args->addr,
878 linux_accept4(struct thread *td, struct linux_accept4_args *args)
881 return (linux_accept_common(td, args->s, args->addr,
882 args->namelen, args->flags));
886 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
888 struct getsockname_args /* {
890 struct sockaddr * __restrict asa;
891 socklen_t * __restrict alen;
895 bsd_args.fdes = args->s;
896 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
897 bsd_args.alen = PTRIN(args->namelen);
898 error = sys_getsockname(td, &bsd_args);
899 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
902 return (linux_sa_put(PTRIN(args->addr)));
906 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
908 struct getpeername_args /* {
915 bsd_args.fdes = args->s;
916 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
917 bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
918 error = sys_getpeername(td, &bsd_args);
919 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
922 return (linux_sa_put(PTRIN(args->addr)));
926 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
928 struct socketpair_args /* {
936 bsd_args.domain = linux_to_bsd_domain(args->domain);
937 if (bsd_args.domain != PF_LOCAL)
938 return (EAFNOSUPPORT);
939 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
940 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
942 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
946 if (args->protocol != 0 && args->protocol != PF_UNIX)
949 * Use of PF_UNIX as protocol argument is not right,
951 * Do not map PF_UNIX as its Linux value is identical
954 return (EPROTONOSUPPORT);
956 bsd_args.protocol = 0;
957 bsd_args.rsv = (int *)PTRIN(args->rsv);
958 return (sys_socketpair(td, &bsd_args));
961 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
962 struct linux_send_args {
970 linux_send(struct thread *td, struct linux_send_args *args)
972 struct sendto_args /* {
981 bsd_args.s = args->s;
982 bsd_args.buf = (caddr_t)PTRIN(args->msg);
983 bsd_args.len = args->len;
984 bsd_args.flags = args->flags;
987 return (sys_sendto(td, &bsd_args));
990 struct linux_recv_args {
998 linux_recv(struct thread *td, struct linux_recv_args *args)
1000 struct recvfrom_args /* {
1005 struct sockaddr *from;
1006 socklen_t fromlenaddr;
1009 bsd_args.s = args->s;
1010 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1011 bsd_args.len = args->len;
1012 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1013 bsd_args.from = NULL;
1014 bsd_args.fromlenaddr = 0;
1015 return (sys_recvfrom(td, &bsd_args));
1017 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1020 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1025 if (linux_check_hdrincl(td, args->s) == 0)
1026 /* IP_HDRINCL set, tweak the packet before sending */
1027 return (linux_sendto_hdrincl(td, args));
1029 msg.msg_name = PTRIN(args->to);
1030 msg.msg_namelen = args->tolen;
1031 msg.msg_iov = &aiov;
1033 msg.msg_control = NULL;
1035 aiov.iov_base = PTRIN(args->msg);
1036 aiov.iov_len = args->len;
1037 return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1042 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1048 if (PTRIN(args->fromlen) != NULL) {
1049 error = copyin(PTRIN(args->fromlen), &fromlen,
1055 msg.msg_namelen = fromlen;
1057 msg.msg_namelen = 0;
1059 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from);
1060 msg.msg_iov = &aiov;
1062 aiov.iov_base = PTRIN(args->buf);
1063 aiov.iov_len = args->len;
1064 msg.msg_control = 0;
1065 msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1067 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL);
1071 if (PTRIN(args->from) != NULL) {
1072 error = bsd_to_linux_sockaddr((struct sockaddr *)
1077 error = linux_sa_put((struct osockaddr *)
1081 if (PTRIN(args->fromlen) != NULL)
1082 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1083 sizeof(msg.msg_namelen));
1089 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1092 struct cmsghdr *cmsg;
1093 struct cmsgcred cmcred;
1094 struct mbuf *control;
1096 struct l_cmsghdr linux_cmsg;
1097 struct l_cmsghdr *ptr_cmsg;
1098 struct l_msghdr linux_msg;
1101 struct sockaddr *sa;
1102 sa_family_t sa_family;
1106 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1111 * Some Linux applications (ping) define a non-NULL control data
1112 * pointer, but a msg_controllen of 0, which is not allowed in the
1113 * FreeBSD system call interface. NULL the msg_control pointer in
1114 * order to handle this case. This should be checked, but allows the
1115 * Linux ping to work.
1117 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1118 linux_msg.msg_control = PTROUT(NULL);
1120 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1124 #ifdef COMPAT_LINUX32
1125 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1128 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1136 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1137 error = kern_getsockname(td, s, &sa, &datalen);
1140 sa_family = sa->sa_family;
1144 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO);
1145 control = m_get(M_WAITOK, MT_CONTROL);
1148 error = copyin(ptr_cmsg, &linux_cmsg,
1149 sizeof(struct l_cmsghdr));
1154 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1158 * Now we support only SCM_RIGHTS and SCM_CRED,
1159 * so return EINVAL in any other cmsg_type
1162 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1164 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1165 if (cmsg->cmsg_type == -1
1166 || cmsg->cmsg_level != SOL_SOCKET)
1170 * Some applications (e.g. pulseaudio) attempt to
1171 * send ancillary data even if the underlying protocol
1172 * doesn't support it which is not allowed in the
1173 * FreeBSD system call interface.
1175 if (sa_family != AF_UNIX)
1178 data = LINUX_CMSG_DATA(ptr_cmsg);
1179 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1181 switch (cmsg->cmsg_type)
1188 datalen = sizeof(cmcred);
1191 * The lower levels will fill in the structure
1193 bzero(data, datalen);
1197 cmsg->cmsg_len = CMSG_LEN(datalen);
1200 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1202 if (!m_append(control, datalen, (c_caddr_t)data))
1204 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1206 if (m_length(control, NULL) == 0) {
1214 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1221 free(cmsg, M_LINUX);
1226 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1229 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1234 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1236 struct l_mmsghdr *msg;
1238 int error, datagrams;
1240 if (args->vlen > UIO_MAXIOV)
1241 args->vlen = UIO_MAXIOV;
1243 msg = PTRIN(args->msg);
1245 while (datagrams < args->vlen) {
1246 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1251 retval = td->td_retval[0];
1252 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1259 td->td_retval[0] = datagrams;
1264 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1265 l_uint flags, struct msghdr *msg)
1268 struct cmsgcred *cmcred;
1269 struct l_cmsghdr *linux_cmsg = NULL;
1270 struct l_ucred linux_ucred;
1271 socklen_t datalen, outlen;
1272 struct l_msghdr linux_msg;
1273 struct iovec *iov, *uiov;
1274 struct mbuf *control = NULL;
1275 struct mbuf **controlp;
1276 struct timeval *ftmvl;
1280 int error, i, fd, fds, *fdp;
1282 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1286 error = linux_to_bsd_msghdr(msg, &linux_msg);
1290 #ifdef COMPAT_LINUX32
1291 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1294 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1299 if (msg->msg_name) {
1300 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name,
1306 uiov = msg->msg_iov;
1308 controlp = (msg->msg_control != NULL) ? &control : NULL;
1309 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp);
1310 msg->msg_iov = uiov;
1314 error = bsd_to_linux_msghdr(msg, &linux_msg);
1318 if (linux_msg.msg_name) {
1319 error = bsd_to_linux_sockaddr((struct sockaddr *)
1320 PTRIN(linux_msg.msg_name));
1324 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1325 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1330 outbuf = PTRIN(linux_msg.msg_control);
1334 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1336 msg->msg_control = mtod(control, struct cmsghdr *);
1337 msg->msg_controllen = control->m_len;
1339 cm = CMSG_FIRSTHDR(msg);
1341 while (cm != NULL) {
1342 linux_cmsg->cmsg_type =
1343 bsd_to_linux_cmsg_type(cm->cmsg_type);
1344 linux_cmsg->cmsg_level =
1345 bsd_to_linux_sockopt_level(cm->cmsg_level);
1346 if (linux_cmsg->cmsg_type == -1
1347 || cm->cmsg_level != SOL_SOCKET)
1353 data = CMSG_DATA(cm);
1354 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1356 switch (cm->cmsg_type)
1359 if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1360 fds = datalen / sizeof(int);
1362 for (i = 0; i < fds; i++) {
1364 (void)kern_fcntl(td, fd,
1365 F_SETFD, FD_CLOEXEC);
1372 * Currently LOCAL_CREDS is never in
1373 * effect for Linux so no need to worry
1376 if (datalen != sizeof(*cmcred)) {
1380 cmcred = (struct cmsgcred *)data;
1381 bzero(&linux_ucred, sizeof(linux_ucred));
1382 linux_ucred.pid = cmcred->cmcred_pid;
1383 linux_ucred.uid = cmcred->cmcred_uid;
1384 linux_ucred.gid = cmcred->cmcred_gid;
1385 data = &linux_ucred;
1386 datalen = sizeof(linux_ucred);
1390 if (datalen != sizeof(struct timeval)) {
1394 ftmvl = (struct timeval *)data;
1395 ltmvl.tv_sec = ftmvl->tv_sec;
1396 ltmvl.tv_usec = ftmvl->tv_usec;
1398 datalen = sizeof(ltmvl);
1402 if (outlen + LINUX_CMSG_LEN(datalen) >
1403 linux_msg.msg_controllen) {
1408 linux_msg.msg_flags |=
1414 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1416 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1419 outbuf += L_CMSG_HDRSZ;
1421 error = copyout(data, outbuf, datalen);
1425 outbuf += LINUX_CMSG_ALIGN(datalen);
1426 outlen += LINUX_CMSG_LEN(datalen);
1428 cm = CMSG_NXTHDR(msg, cm);
1433 linux_msg.msg_controllen = outlen;
1434 error = copyout(&linux_msg, msghdr, sizeof(linux_msg));
1439 free(linux_cmsg, M_LINUX);
1445 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1447 struct msghdr bsd_msg;
1449 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1450 args->flags, &bsd_msg));
1454 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1456 struct l_mmsghdr *msg;
1457 struct msghdr bsd_msg;
1458 struct l_timespec lts;
1459 struct timespec ts, tts;
1461 int error, datagrams;
1463 if (args->timeout) {
1464 error = copyin(args->timeout, <s, sizeof(struct l_timespec));
1467 error = linux_to_native_timespec(&ts, <s);
1471 timespecadd(&tts, &ts);
1474 msg = PTRIN(args->msg);
1476 while (datagrams < args->vlen) {
1477 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1478 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1482 retval = td->td_retval[0];
1483 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1490 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1492 if (args->flags & LINUX_MSG_WAITFORONE)
1493 args->flags |= LINUX_MSG_DONTWAIT;
1496 * See BUGS section of recvmmsg(2).
1498 if (args->timeout) {
1500 timespecsub(&ts, &tts);
1501 if (!timespecisset(&ts) || ts.tv_sec > 0)
1504 /* Out of band data, return right away. */
1505 if (bsd_msg.msg_flags & MSG_OOB)
1509 td->td_retval[0] = datagrams;
1514 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1517 return (kern_shutdown(td, args->s, args->how));
1521 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1523 struct setsockopt_args /* {
1534 bsd_args.s = args->s;
1535 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1536 switch (bsd_args.level) {
1538 name = linux_to_bsd_so_sockopt(args->optname);
1543 error = copyin(PTRIN(args->optval), &linux_tv,
1547 tv.tv_sec = linux_tv.tv_sec;
1548 tv.tv_usec = linux_tv.tv_usec;
1549 return (kern_setsockopt(td, args->s, bsd_args.level,
1550 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1557 name = linux_to_bsd_ip_sockopt(args->optname);
1560 name = linux_to_bsd_ip6_sockopt(args->optname);
1563 name = linux_to_bsd_tcp_sockopt(args->optname);
1570 return (ENOPROTOOPT);
1572 bsd_args.name = name;
1573 bsd_args.val = PTRIN(args->optval);
1574 bsd_args.valsize = args->optlen;
1576 if (name == IPV6_NEXTHOP) {
1577 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1579 error = sys_setsockopt(td, &bsd_args);
1580 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1582 error = sys_setsockopt(td, &bsd_args);
1588 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1590 struct getsockopt_args /* {
1599 socklen_t tv_len, xulen, len;
1602 int error, name, newval;
1604 bsd_args.s = args->s;
1605 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1606 switch (bsd_args.level) {
1608 name = linux_to_bsd_so_sockopt(args->optname);
1613 tv_len = sizeof(tv);
1614 error = kern_getsockopt(td, args->s, bsd_args.level,
1615 name, &tv, UIO_SYSSPACE, &tv_len);
1618 linux_tv.tv_sec = tv.tv_sec;
1619 linux_tv.tv_usec = tv.tv_usec;
1620 return (copyout(&linux_tv, PTRIN(args->optval),
1623 case LOCAL_PEERCRED:
1624 if (args->optlen < sizeof(lxu))
1627 error = kern_getsockopt(td, args->s, bsd_args.level,
1628 name, &xu, UIO_SYSSPACE, &xulen);
1632 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1635 lxu.uid = xu.cr_uid;
1636 lxu.gid = xu.cr_gid;
1637 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1640 len = sizeof(newval);
1641 error = kern_getsockopt(td, args->s, bsd_args.level,
1642 name, &newval, UIO_SYSSPACE, &len);
1645 newval = -SV_ABI_ERRNO(td->td_proc, newval);
1646 return (copyout(&newval, PTRIN(args->optval), len));
1653 name = linux_to_bsd_ip_sockopt(args->optname);
1656 name = linux_to_bsd_ip6_sockopt(args->optname);
1659 name = linux_to_bsd_tcp_sockopt(args->optname);
1668 bsd_args.name = name;
1669 bsd_args.val = PTRIN(args->optval);
1670 bsd_args.avalsize = PTRIN(args->optlen);
1672 if (name == IPV6_NEXTHOP) {
1673 error = sys_getsockopt(td, &bsd_args);
1674 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1676 error = sys_getsockopt(td, &bsd_args);
1681 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1683 /* Argument list sizes for linux_socketcall */
1684 static const unsigned char lxs_args_cnt[] = {
1685 0 /* unused*/, 3 /* socket */,
1686 3 /* bind */, 3 /* connect */,
1687 2 /* listen */, 3 /* accept */,
1688 3 /* getsockname */, 3 /* getpeername */,
1689 4 /* socketpair */, 4 /* send */,
1690 4 /* recv */, 6 /* sendto */,
1691 6 /* recvfrom */, 2 /* shutdown */,
1692 5 /* setsockopt */, 5 /* getsockopt */,
1693 3 /* sendmsg */, 3 /* recvmsg */,
1694 4 /* accept4 */, 5 /* recvmmsg */,
1697 #define LINUX_ARGS_CNT (nitems(lxs_args_cnt) - 1)
1698 #define LINUX_ARG_SIZE(x) (lxs_args_cnt[x] * sizeof(l_ulong))
1701 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1704 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1705 register_t l_args[6];
1710 if (args->what < LINUX_SOCKET || args->what > LINUX_ARGS_CNT)
1712 error = copyin(PTRIN(args->args), a, LINUX_ARG_SIZE(args->what));
1716 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1717 for (int i = 0; i < lxs_args_cnt[args->what]; ++i)
1723 switch (args->what) {
1725 return (linux_socket(td, arg));
1727 return (linux_bind(td, arg));
1729 return (linux_connect(td, arg));
1731 return (linux_listen(td, arg));
1733 return (linux_accept(td, arg));
1734 case LINUX_GETSOCKNAME:
1735 return (linux_getsockname(td, arg));
1736 case LINUX_GETPEERNAME:
1737 return (linux_getpeername(td, arg));
1738 case LINUX_SOCKETPAIR:
1739 return (linux_socketpair(td, arg));
1741 return (linux_send(td, arg));
1743 return (linux_recv(td, arg));
1745 return (linux_sendto(td, arg));
1746 case LINUX_RECVFROM:
1747 return (linux_recvfrom(td, arg));
1748 case LINUX_SHUTDOWN:
1749 return (linux_shutdown(td, arg));
1750 case LINUX_SETSOCKOPT:
1751 return (linux_setsockopt(td, arg));
1752 case LINUX_GETSOCKOPT:
1753 return (linux_getsockopt(td, arg));
1755 return (linux_sendmsg(td, arg));
1757 return (linux_recvmsg(td, arg));
1759 return (linux_accept4(td, arg));
1760 case LINUX_RECVMMSG:
1761 return (linux_recvmmsg(td, arg));
1762 case LINUX_SENDMMSG:
1763 return (linux_sendmmsg(td, arg));
1766 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1769 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */