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/capability.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>
56 #include <netinet/in.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/ip.h>
60 #include <netinet/ip6.h>
61 #include <netinet6/ip6_var.h>
62 #include <netinet6/in6_var.h>
66 #include <machine/../linux32/linux.h>
67 #include <machine/../linux32/linux32_proto.h>
69 #include <machine/../linux/linux.h>
70 #include <machine/../linux/linux_proto.h>
72 #include <compat/linux/linux_socket.h>
73 #include <compat/linux/linux_util.h>
75 static int linux_to_bsd_domain(int);
78 * Reads a linux sockaddr and does any necessary translation.
79 * Linux sockaddrs don't have a length field, only a family.
80 * Copy the osockaddr structure pointed to by osa to kernel, adjust
81 * family and convert to sockaddr.
84 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
87 struct osockaddr *kosa;
89 struct sockaddr_in6 *sin6;
93 int bdom, error, hdrlen, namelen;
95 if (salen < 2 || salen > UCHAR_MAX || !osa)
101 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
102 * if it's a v4-mapped address, so reserve the proper space
105 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
106 salen += sizeof(uint32_t);
111 kosa = malloc(salen, M_SONAME, M_WAITOK);
113 if ((error = copyin(osa, kosa, salen)))
116 bdom = linux_to_bsd_domain(kosa->sa_family);
118 error = EAFNOSUPPORT;
124 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
125 * which lacks the scope id compared with RFC2553 one. If we detect
126 * the situation, reject the address and write a message to system log.
128 * Still accept addresses for which the scope id is not used.
131 if (bdom == AF_INET6) {
132 sin6 = (struct sockaddr_in6 *)kosa;
133 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
134 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
135 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
136 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
137 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
138 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
139 sin6->sin6_scope_id = 0;
142 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
147 salen -= sizeof(uint32_t);
150 if (bdom == AF_INET) {
151 if (salen < sizeof(struct sockaddr_in)) {
155 salen = sizeof(struct sockaddr_in);
158 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
159 hdrlen = offsetof(struct sockaddr_un, sun_path);
160 name = ((struct sockaddr_un *)kosa)->sun_path;
163 * Linux abstract namespace starts with a NULL byte.
164 * XXX We do not support abstract namespace yet.
166 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
168 namelen = strnlen(name, salen - hdrlen);
169 salen = hdrlen + namelen;
170 if (salen > sizeof(struct sockaddr_un)) {
171 error = ENAMETOOLONG;
176 sa = (struct sockaddr *)kosa;
177 sa->sa_family = bdom;
184 free(kosa, M_SONAME);
189 linux_to_bsd_domain(int domain)
193 case LINUX_AF_UNSPEC:
205 case LINUX_AF_APPLETALK:
206 return (AF_APPLETALK);
212 bsd_to_linux_domain(int domain)
217 return (LINUX_AF_UNSPEC);
219 return (LINUX_AF_UNIX);
221 return (LINUX_AF_INET);
223 return (LINUX_AF_INET6);
225 return (LINUX_AF_AX25);
227 return (LINUX_AF_IPX);
229 return (LINUX_AF_APPLETALK);
235 linux_to_bsd_sockopt_level(int level)
239 case LINUX_SOL_SOCKET:
246 bsd_to_linux_sockopt_level(int level)
251 return (LINUX_SOL_SOCKET);
257 linux_to_bsd_ip_sockopt(int opt)
265 case LINUX_IP_OPTIONS:
267 case LINUX_IP_MULTICAST_IF:
268 return (IP_MULTICAST_IF);
269 case LINUX_IP_MULTICAST_TTL:
270 return (IP_MULTICAST_TTL);
271 case LINUX_IP_MULTICAST_LOOP:
272 return (IP_MULTICAST_LOOP);
273 case LINUX_IP_ADD_MEMBERSHIP:
274 return (IP_ADD_MEMBERSHIP);
275 case LINUX_IP_DROP_MEMBERSHIP:
276 return (IP_DROP_MEMBERSHIP);
277 case LINUX_IP_HDRINCL:
284 linux_to_bsd_so_sockopt(int opt)
290 case LINUX_SO_REUSEADDR:
291 return (SO_REUSEADDR);
296 case LINUX_SO_DONTROUTE:
297 return (SO_DONTROUTE);
298 case LINUX_SO_BROADCAST:
299 return (SO_BROADCAST);
300 case LINUX_SO_SNDBUF:
302 case LINUX_SO_RCVBUF:
304 case LINUX_SO_KEEPALIVE:
305 return (SO_KEEPALIVE);
306 case LINUX_SO_OOBINLINE:
307 return (SO_OOBINLINE);
308 case LINUX_SO_LINGER:
310 case LINUX_SO_PEERCRED:
311 return (LOCAL_PEERCRED);
312 case LINUX_SO_RCVLOWAT:
313 return (SO_RCVLOWAT);
314 case LINUX_SO_SNDLOWAT:
315 return (SO_SNDLOWAT);
316 case LINUX_SO_RCVTIMEO:
317 return (SO_RCVTIMEO);
318 case LINUX_SO_SNDTIMEO:
319 return (SO_SNDTIMEO);
320 case LINUX_SO_TIMESTAMP:
321 return (SO_TIMESTAMP);
322 case LINUX_SO_ACCEPTCONN:
323 return (SO_ACCEPTCONN);
329 linux_to_bsd_msg_flags(int flags)
333 if (flags & LINUX_MSG_OOB)
334 ret_flags |= MSG_OOB;
335 if (flags & LINUX_MSG_PEEK)
336 ret_flags |= MSG_PEEK;
337 if (flags & LINUX_MSG_DONTROUTE)
338 ret_flags |= MSG_DONTROUTE;
339 if (flags & LINUX_MSG_CTRUNC)
340 ret_flags |= MSG_CTRUNC;
341 if (flags & LINUX_MSG_TRUNC)
342 ret_flags |= MSG_TRUNC;
343 if (flags & LINUX_MSG_DONTWAIT)
344 ret_flags |= MSG_DONTWAIT;
345 if (flags & LINUX_MSG_EOR)
346 ret_flags |= MSG_EOR;
347 if (flags & LINUX_MSG_WAITALL)
348 ret_flags |= MSG_WAITALL;
349 if (flags & LINUX_MSG_NOSIGNAL)
350 ret_flags |= MSG_NOSIGNAL;
351 #if 0 /* not handled */
352 if (flags & LINUX_MSG_PROXY)
354 if (flags & LINUX_MSG_FIN)
356 if (flags & LINUX_MSG_SYN)
358 if (flags & LINUX_MSG_CONFIRM)
360 if (flags & LINUX_MSG_RST)
362 if (flags & LINUX_MSG_ERRQUEUE)
369 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
370 * native syscall will fault. Thus, we don't really need to check the
371 * return values for these functions.
375 bsd_to_linux_sockaddr(struct sockaddr *arg)
378 size_t sa_len = sizeof(struct sockaddr);
381 if ((error = copyin(arg, &sa, sa_len)))
384 *(u_short *)&sa = sa.sa_family;
386 error = copyout(&sa, arg, sa_len);
392 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
395 size_t sa_len = sizeof(struct sockaddr);
398 if ((error = copyin(arg, &sa, sa_len)))
401 sa.sa_family = *(sa_family_t *)&sa;
404 error = copyout(&sa, arg, sa_len);
411 linux_sa_put(struct osockaddr *osa)
417 * Only read/write the osockaddr family part, the rest is
420 error = copyin(osa, &sa, sizeof(sa.sa_family));
424 bdom = bsd_to_linux_domain(sa.sa_family);
429 error = copyout(&sa, osa, sizeof(sa.sa_family));
437 linux_to_bsd_cmsg_type(int cmsg_type)
441 case LINUX_SCM_RIGHTS:
443 case LINUX_SCM_CREDENTIALS:
450 bsd_to_linux_cmsg_type(int cmsg_type)
455 return (LINUX_SCM_RIGHTS);
457 return (LINUX_SCM_CREDENTIALS);
463 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
465 if (lhdr->msg_controllen > INT_MAX)
468 bhdr->msg_name = PTRIN(lhdr->msg_name);
469 bhdr->msg_namelen = lhdr->msg_namelen;
470 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
471 bhdr->msg_iovlen = lhdr->msg_iovlen;
472 bhdr->msg_control = PTRIN(lhdr->msg_control);
475 * msg_controllen is skipped since BSD and LINUX control messages
476 * are potentially different sizes (e.g. the cred structure used
477 * by SCM_CREDS is different between the two operating system).
479 * The caller can set it (if necessary) after converting all the
483 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
488 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
490 lhdr->msg_name = PTROUT(bhdr->msg_name);
491 lhdr->msg_namelen = bhdr->msg_namelen;
492 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
493 lhdr->msg_iovlen = bhdr->msg_iovlen;
494 lhdr->msg_control = PTROUT(bhdr->msg_control);
497 * msg_controllen is skipped since BSD and LINUX control messages
498 * are potentially different sizes (e.g. the cred structure used
499 * by SCM_CREDS is different between the two operating system).
501 * The caller can set it (if necessary) after converting all the
505 /* msg_flags skipped */
510 linux_set_socket_flags(struct thread *td, int s, int flags)
514 if (flags & LINUX_SOCK_NONBLOCK) {
515 error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
519 if (flags & LINUX_SOCK_CLOEXEC) {
520 error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
528 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
529 struct mbuf *control, enum uio_seg segflg)
534 if (mp->msg_name != NULL) {
535 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
542 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
550 /* Return 0 if IP_HDRINCL is set for the given socket. */
552 linux_check_hdrincl(struct thread *td, int s)
554 int error, optval, size_val;
556 size_val = sizeof(optval);
557 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
558 &optval, UIO_SYSSPACE, &size_val);
562 return (optval == 0);
565 struct linux_sendto_args {
575 * Updated sendto() when IP_HDRINCL is set:
576 * tweak endian-dependent fields in the IP packet.
579 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
582 * linux_ip_copysize defines how many bytes we should copy
583 * from the beginning of the IP packet before we customize it for BSD.
584 * It should include all the fields we modify (ip_len and ip_off).
586 #define linux_ip_copysize 8
590 struct iovec aiov[1];
593 /* Check that the packet isn't too big or too small. */
594 if (linux_args->len < linux_ip_copysize ||
595 linux_args->len > IP_MAXPACKET)
598 packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
600 /* Make kernel copy of the packet to be sent */
601 if ((error = copyin(PTRIN(linux_args->msg), packet,
605 /* Convert fields from Linux to BSD raw IP socket format */
606 packet->ip_len = linux_args->len;
607 packet->ip_off = ntohs(packet->ip_off);
609 /* Prepare the msghdr and iovec structures describing the new packet */
610 msg.msg_name = PTRIN(linux_args->to);
611 msg.msg_namelen = linux_args->tolen;
614 msg.msg_control = NULL;
616 aiov[0].iov_base = (char *)packet;
617 aiov[0].iov_len = linux_args->len;
618 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
621 free(packet, M_TEMP);
625 struct linux_socket_args {
632 linux_socket(struct thread *td, struct linux_socket_args *args)
634 struct socket_args /* {
639 int retval_socket, socket_flags;
641 bsd_args.protocol = args->protocol;
642 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
643 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
645 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
646 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
648 bsd_args.domain = linux_to_bsd_domain(args->domain);
649 if (bsd_args.domain == -1)
650 return (EAFNOSUPPORT);
652 retval_socket = sys_socket(td, &bsd_args);
654 return (retval_socket);
656 retval_socket = linux_set_socket_flags(td, td->td_retval[0],
659 (void)kern_close(td, td->td_retval[0]);
663 if (bsd_args.type == SOCK_RAW
664 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
665 && bsd_args.domain == PF_INET) {
666 /* It's a raw IP socket: set the IP_HDRINCL option. */
670 /* We ignore any error returned by kern_setsockopt() */
671 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
672 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
676 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
677 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
678 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
681 if (bsd_args.domain == PF_INET6) {
685 /* We ignore any error returned by setsockopt() */
686 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
687 &v6only, UIO_SYSSPACE, sizeof(v6only));
692 return (retval_socket);
695 struct linux_bind_args {
702 linux_bind(struct thread *td, struct linux_bind_args *args)
707 error = linux_getsockaddr(&sa, PTRIN(args->name),
712 error = kern_bind(td, args->s, sa);
714 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
719 struct linux_connect_args {
724 int linux_connect(struct thread *, struct linux_connect_args *);
727 linux_connect(struct thread *td, struct linux_connect_args *args)
734 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
739 error = kern_connect(td, args->s, sa);
741 if (error != EISCONN)
745 * Linux doesn't return EISCONN the first time it occurs,
746 * when on a non-blocking socket. Instead it returns the
747 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
749 * XXXRW: Instead of using fgetsock(), check that it is a
750 * socket and use the file descriptor reference instead of
751 * creating a new one.
753 error = fgetsock(td, args->s, CAP_CONNECT, &so, &fflag);
756 if (fflag & FNONBLOCK) {
758 if (so->so_emuldata == 0)
759 error = so->so_error;
760 so->so_emuldata = (void *)1;
768 struct linux_listen_args {
774 linux_listen(struct thread *td, struct linux_listen_args *args)
776 struct listen_args /* {
781 bsd_args.s = args->s;
782 bsd_args.backlog = args->backlog;
783 return (sys_listen(td, &bsd_args));
787 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
788 l_uintptr_t namelen, int flags)
790 struct accept_args /* {
792 struct sockaddr * __restrict name;
793 socklen_t * __restrict anamelen;
797 if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
802 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
803 bsd_args.anamelen = PTRIN(namelen);/* XXX */
804 error = sys_accept(td, &bsd_args);
805 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
807 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
813 * linux appears not to copy flags from the parent socket to the
814 * accepted one, so we must clear the flags in the new descriptor
815 * and apply the requested flags.
817 error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
820 error = linux_set_socket_flags(td, td->td_retval[0], flags);
824 error = linux_sa_put(PTRIN(addr));
828 (void)kern_close(td, td->td_retval[0]);
829 td->td_retval[0] = 0;
834 struct linux_accept_args {
841 linux_accept(struct thread *td, struct linux_accept_args *args)
844 return (linux_accept_common(td, args->s, args->addr,
848 struct linux_accept4_args {
856 linux_accept4(struct thread *td, struct linux_accept4_args *args)
859 return (linux_accept_common(td, args->s, args->addr,
860 args->namelen, args->flags));
863 struct linux_getsockname_args {
870 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
872 struct getsockname_args /* {
874 struct sockaddr * __restrict asa;
875 socklen_t * __restrict alen;
879 bsd_args.fdes = args->s;
881 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
882 bsd_args.alen = PTRIN(args->namelen); /* XXX */
883 error = sys_getsockname(td, &bsd_args);
884 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
887 error = linux_sa_put(PTRIN(args->addr));
893 struct linux_getpeername_args {
900 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
902 struct getpeername_args /* {
909 bsd_args.fdes = args->s;
910 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
911 bsd_args.alen = (int *)PTRIN(args->namelen);
912 error = sys_getpeername(td, &bsd_args);
913 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
916 error = linux_sa_put(PTRIN(args->addr));
922 struct linux_socketpair_args {
930 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
932 struct socketpair_args /* {
938 int error, socket_flags;
941 bsd_args.domain = linux_to_bsd_domain(args->domain);
942 if (bsd_args.domain != PF_LOCAL)
943 return (EAFNOSUPPORT);
945 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
946 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
948 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
949 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
952 if (args->protocol != 0 && args->protocol != PF_UNIX)
955 * Use of PF_UNIX as protocol argument is not right,
957 * Do not map PF_UNIX as its Linux value is identical
960 return (EPROTONOSUPPORT);
962 bsd_args.protocol = 0;
963 bsd_args.rsv = (int *)PTRIN(args->rsv);
964 error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
965 bsd_args.protocol, sv);
968 error = linux_set_socket_flags(td, sv[0], socket_flags);
971 error = linux_set_socket_flags(td, sv[1], socket_flags);
975 error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
979 (void)kern_close(td, sv[0]);
980 (void)kern_close(td, sv[1]);
985 struct linux_send_args {
993 linux_send(struct thread *td, struct linux_send_args *args)
995 struct sendto_args /* {
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 = args->flags;
1010 return sys_sendto(td, &bsd_args);
1013 struct linux_recv_args {
1021 linux_recv(struct thread *td, struct linux_recv_args *args)
1023 struct recvfrom_args /* {
1028 struct sockaddr *from;
1029 socklen_t fromlenaddr;
1032 bsd_args.s = args->s;
1033 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1034 bsd_args.len = args->len;
1035 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1036 bsd_args.from = NULL;
1037 bsd_args.fromlenaddr = 0;
1038 return (sys_recvfrom(td, &bsd_args));
1042 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1048 if (linux_check_hdrincl(td, args->s) == 0)
1049 /* IP_HDRINCL set, tweak the packet before sending */
1050 return (linux_sendto_hdrincl(td, args));
1052 msg.msg_name = PTRIN(args->to);
1053 msg.msg_namelen = args->tolen;
1054 msg.msg_iov = &aiov;
1056 msg.msg_control = NULL;
1058 aiov.iov_base = PTRIN(args->msg);
1059 aiov.iov_len = args->len;
1060 error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1065 struct linux_recvfrom_args {
1071 l_uintptr_t fromlen;
1075 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1077 struct recvfrom_args /* {
1082 struct sockaddr * __restrict from;
1083 socklen_t * __restrict fromlenaddr;
1088 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1091 bsd_args.s = args->s;
1092 bsd_args.buf = PTRIN(args->buf);
1093 bsd_args.len = args->len;
1094 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1096 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1097 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1099 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1100 error = sys_recvfrom(td, &bsd_args);
1101 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1106 error = linux_sa_put((struct osockaddr *)
1114 struct linux_sendmsg_args {
1121 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1123 struct cmsghdr *cmsg;
1124 struct cmsgcred cmcred;
1125 struct mbuf *control;
1127 struct l_cmsghdr linux_cmsg;
1128 struct l_cmsghdr *ptr_cmsg;
1129 struct l_msghdr linux_msg;
1132 struct sockaddr *sa;
1133 sa_family_t sa_family;
1137 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1142 * Some Linux applications (ping) define a non-NULL control data
1143 * pointer, but a msg_controllen of 0, which is not allowed in the
1144 * FreeBSD system call interface. NULL the msg_control pointer in
1145 * order to handle this case. This should be checked, but allows the
1146 * Linux ping to work.
1148 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1149 linux_msg.msg_control = PTROUT(NULL);
1151 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1155 #ifdef COMPAT_LINUX32
1156 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1159 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1167 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1168 error = kern_getsockname(td, args->s, &sa, &datalen);
1171 sa_family = sa->sa_family;
1175 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1176 control = m_get(M_WAIT, MT_CONTROL);
1177 if (control == NULL)
1181 error = copyin(ptr_cmsg, &linux_cmsg,
1182 sizeof(struct l_cmsghdr));
1187 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1191 * Now we support only SCM_RIGHTS and SCM_CRED,
1192 * so return EINVAL in any other cmsg_type
1195 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1197 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1198 if (cmsg->cmsg_type == -1
1199 || cmsg->cmsg_level != SOL_SOCKET)
1203 * Some applications (e.g. pulseaudio) attempt to
1204 * send ancillary data even if the underlying protocol
1205 * doesn't support it which is not allowed in the
1206 * FreeBSD system call interface.
1208 if (sa_family != AF_UNIX)
1211 data = LINUX_CMSG_DATA(ptr_cmsg);
1212 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1214 switch (cmsg->cmsg_type)
1221 datalen = sizeof(cmcred);
1224 * The lower levels will fill in the structure
1226 bzero(data, datalen);
1230 cmsg->cmsg_len = CMSG_LEN(datalen);
1233 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1235 if (!m_append(control, datalen, (c_caddr_t)data))
1237 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1239 if (m_length(control, NULL) == 0) {
1247 error = linux_sendit(td, args->s, &msg, args->flags, control,
1257 struct linux_recvmsg_args {
1264 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1267 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;
1278 int error, i, fd, fds, *fdp;
1280 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1284 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1288 #ifdef COMPAT_LINUX32
1289 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1292 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1298 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1306 controlp = (msg.msg_control != NULL) ? &control : NULL;
1307 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1312 error = bsd_to_linux_msghdr(&msg, &linux_msg);
1316 if (linux_msg.msg_name) {
1317 error = bsd_to_linux_sockaddr((struct sockaddr *)
1318 PTRIN(linux_msg.msg_name));
1322 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1323 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1328 outbuf = PTRIN(linux_msg.msg_control);
1332 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1334 msg.msg_control = mtod(control, struct cmsghdr *);
1335 msg.msg_controllen = control->m_len;
1337 cm = CMSG_FIRSTHDR(&msg);
1339 while (cm != NULL) {
1340 linux_cmsg->cmsg_type =
1341 bsd_to_linux_cmsg_type(cm->cmsg_type);
1342 linux_cmsg->cmsg_level =
1343 bsd_to_linux_sockopt_level(cm->cmsg_level);
1344 if (linux_cmsg->cmsg_type == -1
1345 || cm->cmsg_level != SOL_SOCKET)
1351 data = CMSG_DATA(cm);
1352 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1354 switch (cm->cmsg_type)
1357 if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1358 fds = datalen / sizeof(int);
1360 for (i = 0; i < fds; i++) {
1362 (void)kern_fcntl(td, fd,
1363 F_SETFD, FD_CLOEXEC);
1370 * Currently LOCAL_CREDS is never in
1371 * effect for Linux so no need to worry
1374 if (datalen != sizeof(*cmcred)) {
1378 cmcred = (struct cmsgcred *)data;
1379 bzero(&linux_ucred, sizeof(linux_ucred));
1380 linux_ucred.pid = cmcred->cmcred_pid;
1381 linux_ucred.uid = cmcred->cmcred_uid;
1382 linux_ucred.gid = cmcred->cmcred_gid;
1383 data = &linux_ucred;
1384 datalen = sizeof(linux_ucred);
1388 if (outlen + LINUX_CMSG_LEN(datalen) >
1389 linux_msg.msg_controllen) {
1394 linux_msg.msg_flags |=
1400 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1402 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1405 outbuf += L_CMSG_HDRSZ;
1407 error = copyout(data, outbuf, datalen);
1411 outbuf += LINUX_CMSG_ALIGN(datalen);
1412 outlen += LINUX_CMSG_LEN(datalen);
1414 cm = CMSG_NXTHDR(&msg, cm);
1419 linux_msg.msg_controllen = outlen;
1420 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1424 if (control != NULL)
1426 if (linux_cmsg != NULL)
1427 free(linux_cmsg, M_TEMP);
1432 struct linux_shutdown_args {
1438 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1440 struct shutdown_args /* {
1445 bsd_args.s = args->s;
1446 bsd_args.how = args->how;
1447 return (sys_shutdown(td, &bsd_args));
1450 struct linux_setsockopt_args {
1459 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1461 struct setsockopt_args /* {
1472 bsd_args.s = args->s;
1473 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1474 switch (bsd_args.level) {
1476 name = linux_to_bsd_so_sockopt(args->optname);
1481 error = copyin(PTRIN(args->optval), &linux_tv,
1485 tv.tv_sec = linux_tv.tv_sec;
1486 tv.tv_usec = linux_tv.tv_usec;
1487 return (kern_setsockopt(td, args->s, bsd_args.level,
1488 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1496 name = linux_to_bsd_ip_sockopt(args->optname);
1499 /* Linux TCP option values match BSD's */
1500 name = args->optname;
1507 return (ENOPROTOOPT);
1509 bsd_args.name = name;
1510 bsd_args.val = PTRIN(args->optval);
1511 bsd_args.valsize = args->optlen;
1513 if (name == IPV6_NEXTHOP) {
1514 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1516 error = sys_setsockopt(td, &bsd_args);
1517 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1519 error = sys_setsockopt(td, &bsd_args);
1524 struct linux_getsockopt_args {
1533 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1535 struct getsockopt_args /* {
1544 socklen_t tv_len, xulen;
1549 bsd_args.s = args->s;
1550 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1551 switch (bsd_args.level) {
1553 name = linux_to_bsd_so_sockopt(args->optname);
1558 tv_len = sizeof(tv);
1559 error = kern_getsockopt(td, args->s, bsd_args.level,
1560 name, &tv, UIO_SYSSPACE, &tv_len);
1563 linux_tv.tv_sec = tv.tv_sec;
1564 linux_tv.tv_usec = tv.tv_usec;
1565 return (copyout(&linux_tv, PTRIN(args->optval),
1569 case LOCAL_PEERCRED:
1570 if (args->optlen != sizeof(lxu))
1573 error = kern_getsockopt(td, args->s, bsd_args.level,
1574 name, &xu, UIO_SYSSPACE, &xulen);
1578 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1581 lxu.uid = xu.cr_uid;
1582 lxu.gid = xu.cr_gid;
1583 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1591 name = linux_to_bsd_ip_sockopt(args->optname);
1594 /* Linux TCP option values match BSD's */
1595 name = args->optname;
1604 bsd_args.name = name;
1605 bsd_args.val = PTRIN(args->optval);
1606 bsd_args.avalsize = PTRIN(args->optlen);
1608 if (name == IPV6_NEXTHOP) {
1609 error = sys_getsockopt(td, &bsd_args);
1610 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1612 error = sys_getsockopt(td, &bsd_args);
1617 /* Argument list sizes for linux_socketcall */
1619 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1621 static const unsigned char lxs_args[] = {
1622 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1623 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1624 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1625 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1626 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1627 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1628 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1629 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1630 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1631 LINUX_AL(4) /* accept4 */
1634 #define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1637 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1643 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1645 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1650 switch (args->what) {
1652 return (linux_socket(td, arg));
1654 return (linux_bind(td, arg));
1656 return (linux_connect(td, arg));
1658 return (linux_listen(td, arg));
1660 return (linux_accept(td, arg));
1661 case LINUX_GETSOCKNAME:
1662 return (linux_getsockname(td, arg));
1663 case LINUX_GETPEERNAME:
1664 return (linux_getpeername(td, arg));
1665 case LINUX_SOCKETPAIR:
1666 return (linux_socketpair(td, arg));
1668 return (linux_send(td, arg));
1670 return (linux_recv(td, arg));
1672 return (linux_sendto(td, arg));
1673 case LINUX_RECVFROM:
1674 return (linux_recvfrom(td, arg));
1675 case LINUX_SHUTDOWN:
1676 return (linux_shutdown(td, arg));
1677 case LINUX_SETSOCKOPT:
1678 return (linux_setsockopt(td, arg));
1679 case LINUX_GETSOCKOPT:
1680 return (linux_getsockopt(td, arg));
1682 return (linux_sendmsg(td, arg));
1684 return (linux_recvmsg(td, arg));
1686 return (linux_accept4(td, arg));
1689 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);