2 * Copyright (c) 1995 Søren Schmidt
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 /* XXX we use functions that might not exist. */
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
36 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/sysproto.h>
40 #include <sys/capsicum.h>
41 #include <sys/fcntl.h>
43 #include <sys/limits.h>
45 #include <sys/malloc.h>
46 #include <sys/mutex.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/syscallsubr.h>
52 #include <sys/syslog.h>
57 #include <netinet/in.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/ip.h>
60 #include <netinet/tcp.h>
62 #include <netinet/ip6.h>
63 #include <netinet6/ip6_var.h>
67 #include <machine/../linux32/linux.h>
68 #include <machine/../linux32/linux32_proto.h>
70 #include <machine/../linux/linux.h>
71 #include <machine/../linux/linux_proto.h>
73 #include <compat/linux/linux_socket.h>
74 #include <compat/linux/linux_util.h>
76 static int linux_to_bsd_domain(int);
79 * Reads a linux sockaddr and does any necessary translation.
80 * Linux sockaddrs don't have a length field, only a family.
81 * Copy the osockaddr structure pointed to by osa to kernel, adjust
82 * family and convert to sockaddr.
85 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
88 struct osockaddr *kosa;
90 struct sockaddr_in6 *sin6;
94 int bdom, error, hdrlen, namelen;
96 if (salen < 2 || salen > UCHAR_MAX || !osa)
102 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
103 * if it's a v4-mapped address, so reserve the proper space
106 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
107 salen += sizeof(uint32_t);
112 kosa = malloc(salen, M_SONAME, M_WAITOK);
114 if ((error = copyin(osa, kosa, salen)))
117 bdom = linux_to_bsd_domain(kosa->sa_family);
119 error = EAFNOSUPPORT;
125 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
126 * which lacks the scope id compared with RFC2553 one. If we detect
127 * the situation, reject the address and write a message to system log.
129 * Still accept addresses for which the scope id is not used.
132 if (bdom == AF_INET6) {
133 sin6 = (struct sockaddr_in6 *)kosa;
134 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
135 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
136 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
137 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
138 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
139 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
140 sin6->sin6_scope_id = 0;
143 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
148 salen -= sizeof(uint32_t);
151 if (bdom == AF_INET) {
152 if (salen < sizeof(struct sockaddr_in)) {
156 salen = sizeof(struct sockaddr_in);
159 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
160 hdrlen = offsetof(struct sockaddr_un, sun_path);
161 name = ((struct sockaddr_un *)kosa)->sun_path;
164 * Linux abstract namespace starts with a NULL byte.
165 * XXX We do not support abstract namespace yet.
167 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
169 namelen = strnlen(name, salen - hdrlen);
170 salen = hdrlen + namelen;
171 if (salen > sizeof(struct sockaddr_un)) {
172 error = ENAMETOOLONG;
177 sa = (struct sockaddr *)kosa;
178 sa->sa_family = bdom;
185 free(kosa, M_SONAME);
190 linux_to_bsd_domain(int domain)
194 case LINUX_AF_UNSPEC:
206 case LINUX_AF_APPLETALK:
207 return (AF_APPLETALK);
213 bsd_to_linux_domain(int domain)
218 return (LINUX_AF_UNSPEC);
220 return (LINUX_AF_UNIX);
222 return (LINUX_AF_INET);
224 return (LINUX_AF_INET6);
226 return (LINUX_AF_AX25);
228 return (LINUX_AF_IPX);
230 return (LINUX_AF_APPLETALK);
236 linux_to_bsd_sockopt_level(int level)
240 case LINUX_SOL_SOCKET:
247 bsd_to_linux_sockopt_level(int level)
252 return (LINUX_SOL_SOCKET);
258 linux_to_bsd_ip_sockopt(int opt)
266 case LINUX_IP_OPTIONS:
268 case LINUX_IP_MULTICAST_IF:
269 return (IP_MULTICAST_IF);
270 case LINUX_IP_MULTICAST_TTL:
271 return (IP_MULTICAST_TTL);
272 case LINUX_IP_MULTICAST_LOOP:
273 return (IP_MULTICAST_LOOP);
274 case LINUX_IP_ADD_MEMBERSHIP:
275 return (IP_ADD_MEMBERSHIP);
276 case LINUX_IP_DROP_MEMBERSHIP:
277 return (IP_DROP_MEMBERSHIP);
278 case LINUX_IP_HDRINCL:
285 linux_to_bsd_so_sockopt(int opt)
291 case LINUX_SO_REUSEADDR:
292 return (SO_REUSEADDR);
297 case LINUX_SO_DONTROUTE:
298 return (SO_DONTROUTE);
299 case LINUX_SO_BROADCAST:
300 return (SO_BROADCAST);
301 case LINUX_SO_SNDBUF:
303 case LINUX_SO_RCVBUF:
305 case LINUX_SO_KEEPALIVE:
306 return (SO_KEEPALIVE);
307 case LINUX_SO_OOBINLINE:
308 return (SO_OOBINLINE);
309 case LINUX_SO_LINGER:
311 case LINUX_SO_PEERCRED:
312 return (LOCAL_PEERCRED);
313 case LINUX_SO_RCVLOWAT:
314 return (SO_RCVLOWAT);
315 case LINUX_SO_SNDLOWAT:
316 return (SO_SNDLOWAT);
317 case LINUX_SO_RCVTIMEO:
318 return (SO_RCVTIMEO);
319 case LINUX_SO_SNDTIMEO:
320 return (SO_SNDTIMEO);
321 case LINUX_SO_TIMESTAMP:
322 return (SO_TIMESTAMP);
323 case LINUX_SO_ACCEPTCONN:
324 return (SO_ACCEPTCONN);
330 linux_to_bsd_tcp_sockopt(int opt)
334 case LINUX_TCP_NODELAY:
335 return (TCP_NODELAY);
336 case LINUX_TCP_MAXSEG:
338 case LINUX_TCP_KEEPIDLE:
339 return (TCP_KEEPIDLE);
340 case LINUX_TCP_KEEPINTVL:
341 return (TCP_KEEPINTVL);
342 case LINUX_TCP_KEEPCNT:
343 return (TCP_KEEPCNT);
344 case LINUX_TCP_MD5SIG:
351 linux_to_bsd_msg_flags(int flags)
355 if (flags & LINUX_MSG_OOB)
356 ret_flags |= MSG_OOB;
357 if (flags & LINUX_MSG_PEEK)
358 ret_flags |= MSG_PEEK;
359 if (flags & LINUX_MSG_DONTROUTE)
360 ret_flags |= MSG_DONTROUTE;
361 if (flags & LINUX_MSG_CTRUNC)
362 ret_flags |= MSG_CTRUNC;
363 if (flags & LINUX_MSG_TRUNC)
364 ret_flags |= MSG_TRUNC;
365 if (flags & LINUX_MSG_DONTWAIT)
366 ret_flags |= MSG_DONTWAIT;
367 if (flags & LINUX_MSG_EOR)
368 ret_flags |= MSG_EOR;
369 if (flags & LINUX_MSG_WAITALL)
370 ret_flags |= MSG_WAITALL;
371 if (flags & LINUX_MSG_NOSIGNAL)
372 ret_flags |= MSG_NOSIGNAL;
373 #if 0 /* not handled */
374 if (flags & LINUX_MSG_PROXY)
376 if (flags & LINUX_MSG_FIN)
378 if (flags & LINUX_MSG_SYN)
380 if (flags & LINUX_MSG_CONFIRM)
382 if (flags & LINUX_MSG_RST)
384 if (flags & LINUX_MSG_ERRQUEUE)
391 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
392 * native syscall will fault. Thus, we don't really need to check the
393 * return values for these functions.
397 bsd_to_linux_sockaddr(struct sockaddr *arg)
400 size_t sa_len = sizeof(struct sockaddr);
403 if ((error = copyin(arg, &sa, sa_len)))
406 *(u_short *)&sa = sa.sa_family;
408 error = copyout(&sa, arg, sa_len);
414 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
417 size_t sa_len = sizeof(struct sockaddr);
420 if ((error = copyin(arg, &sa, sa_len)))
423 sa.sa_family = *(sa_family_t *)&sa;
426 error = copyout(&sa, arg, sa_len);
433 linux_sa_put(struct osockaddr *osa)
439 * Only read/write the osockaddr family part, the rest is
442 error = copyin(osa, &sa, sizeof(sa.sa_family));
446 bdom = bsd_to_linux_domain(sa.sa_family);
451 error = copyout(&sa, osa, sizeof(sa.sa_family));
459 linux_to_bsd_cmsg_type(int cmsg_type)
463 case LINUX_SCM_RIGHTS:
465 case LINUX_SCM_CREDENTIALS:
472 bsd_to_linux_cmsg_type(int cmsg_type)
477 return (LINUX_SCM_RIGHTS);
479 return (LINUX_SCM_CREDENTIALS);
485 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
487 if (lhdr->msg_controllen > INT_MAX)
490 bhdr->msg_name = PTRIN(lhdr->msg_name);
491 bhdr->msg_namelen = lhdr->msg_namelen;
492 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
493 bhdr->msg_iovlen = lhdr->msg_iovlen;
494 bhdr->msg_control = PTRIN(lhdr->msg_control);
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 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
510 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
512 lhdr->msg_name = PTROUT(bhdr->msg_name);
513 lhdr->msg_namelen = bhdr->msg_namelen;
514 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
515 lhdr->msg_iovlen = bhdr->msg_iovlen;
516 lhdr->msg_control = PTROUT(bhdr->msg_control);
519 * msg_controllen is skipped since BSD and LINUX control messages
520 * are potentially different sizes (e.g. the cred structure used
521 * by SCM_CREDS is different between the two operating system).
523 * The caller can set it (if necessary) after converting all the
527 /* msg_flags skipped */
532 linux_set_socket_flags(struct thread *td, int s, int flags)
536 if (flags & LINUX_SOCK_NONBLOCK) {
537 error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
541 if (flags & LINUX_SOCK_CLOEXEC) {
542 error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
550 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
551 struct mbuf *control, enum uio_seg segflg)
556 if (mp->msg_name != NULL) {
557 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
564 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
572 /* Return 0 if IP_HDRINCL is set for the given socket. */
574 linux_check_hdrincl(struct thread *td, int s)
579 size_val = sizeof(optval);
580 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
581 &optval, UIO_SYSSPACE, &size_val);
585 return (optval == 0);
589 * Updated sendto() when IP_HDRINCL is set:
590 * tweak endian-dependent fields in the IP packet.
593 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
596 * linux_ip_copysize defines how many bytes we should copy
597 * from the beginning of the IP packet before we customize it for BSD.
598 * It should include all the fields we modify (ip_len and ip_off).
600 #define linux_ip_copysize 8
604 struct iovec aiov[1];
607 /* Check that the packet isn't too big or too small. */
608 if (linux_args->len < linux_ip_copysize ||
609 linux_args->len > IP_MAXPACKET)
612 packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
614 /* Make kernel copy of the packet to be sent */
615 if ((error = copyin(PTRIN(linux_args->msg), packet,
619 /* Convert fields from Linux to BSD raw IP socket format */
620 packet->ip_len = linux_args->len;
621 packet->ip_off = ntohs(packet->ip_off);
623 /* Prepare the msghdr and iovec structures describing the new packet */
624 msg.msg_name = PTRIN(linux_args->to);
625 msg.msg_namelen = linux_args->tolen;
628 msg.msg_control = NULL;
630 aiov[0].iov_base = (char *)packet;
631 aiov[0].iov_len = linux_args->len;
632 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
635 free(packet, M_TEMP);
640 linux_socket(struct thread *td, struct linux_socket_args *args)
642 struct socket_args /* {
647 int retval_socket, socket_flags;
649 bsd_args.protocol = args->protocol;
650 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
651 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
653 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
654 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
656 bsd_args.domain = linux_to_bsd_domain(args->domain);
657 if (bsd_args.domain == -1)
658 return (EAFNOSUPPORT);
660 retval_socket = sys_socket(td, &bsd_args);
662 return (retval_socket);
664 retval_socket = linux_set_socket_flags(td, td->td_retval[0],
667 (void)kern_close(td, td->td_retval[0]);
671 if (bsd_args.type == SOCK_RAW
672 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
673 && bsd_args.domain == PF_INET) {
674 /* It's a raw IP socket: set the IP_HDRINCL option. */
678 /* We ignore any error returned by kern_setsockopt() */
679 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
680 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
684 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
685 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
686 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
689 if (bsd_args.domain == PF_INET6) {
693 /* We ignore any error returned by setsockopt() */
694 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
695 &v6only, UIO_SYSSPACE, sizeof(v6only));
700 return (retval_socket);
704 linux_bind(struct thread *td, struct linux_bind_args *args)
709 error = linux_getsockaddr(&sa, PTRIN(args->name),
714 error = kern_bindat(td, AT_FDCWD, args->s, sa);
716 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
722 linux_connect(struct thread *td, struct linux_connect_args *args)
730 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
735 error = kern_connectat(td, AT_FDCWD, args->s, sa);
737 if (error != EISCONN)
741 * Linux doesn't return EISCONN the first time it occurs,
742 * when on a non-blocking socket. Instead it returns the
743 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
745 * XXXRW: Instead of using fgetsock(), check that it is a
746 * socket and use the file descriptor reference instead of
747 * creating a new one.
749 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
753 if (fflag & FNONBLOCK) {
755 if (so->so_emuldata == 0)
756 error = so->so_error;
757 so->so_emuldata = (void *)1;
766 linux_listen(struct thread *td, struct linux_listen_args *args)
768 struct listen_args /* {
773 bsd_args.s = args->s;
774 bsd_args.backlog = args->backlog;
775 return (sys_listen(td, &bsd_args));
779 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
780 l_uintptr_t namelen, int flags)
782 struct accept_args /* {
784 struct sockaddr * __restrict name;
785 socklen_t * __restrict anamelen;
789 if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
794 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
795 bsd_args.anamelen = PTRIN(namelen);/* XXX */
796 error = sys_accept(td, &bsd_args);
797 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
799 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
805 * linux appears not to copy flags from the parent socket to the
806 * accepted one, so we must clear the flags in the new descriptor
807 * and apply the requested flags.
809 error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
812 error = linux_set_socket_flags(td, td->td_retval[0], flags);
816 error = linux_sa_put(PTRIN(addr));
820 (void)kern_close(td, td->td_retval[0]);
821 td->td_retval[0] = 0;
827 linux_accept(struct thread *td, struct linux_accept_args *args)
830 return (linux_accept_common(td, args->s, args->addr,
835 linux_accept4(struct thread *td, struct linux_accept4_args *args)
838 return (linux_accept_common(td, args->s, args->addr,
839 args->namelen, args->flags));
843 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
845 struct getsockname_args /* {
847 struct sockaddr * __restrict asa;
848 socklen_t * __restrict alen;
852 bsd_args.fdes = args->s;
854 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
855 bsd_args.alen = PTRIN(args->namelen); /* XXX */
856 error = sys_getsockname(td, &bsd_args);
857 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
860 error = linux_sa_put(PTRIN(args->addr));
867 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
869 struct getpeername_args /* {
876 bsd_args.fdes = args->s;
877 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
878 bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
879 error = sys_getpeername(td, &bsd_args);
880 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
883 error = linux_sa_put(PTRIN(args->addr));
890 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
892 struct socketpair_args /* {
898 int error, socket_flags;
901 bsd_args.domain = linux_to_bsd_domain(args->domain);
902 if (bsd_args.domain != PF_LOCAL)
903 return (EAFNOSUPPORT);
905 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
906 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
908 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
909 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
912 if (args->protocol != 0 && args->protocol != PF_UNIX)
915 * Use of PF_UNIX as protocol argument is not right,
917 * Do not map PF_UNIX as its Linux value is identical
920 return (EPROTONOSUPPORT);
922 bsd_args.protocol = 0;
923 bsd_args.rsv = (int *)PTRIN(args->rsv);
924 error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
925 bsd_args.protocol, sv);
928 error = linux_set_socket_flags(td, sv[0], socket_flags);
931 error = linux_set_socket_flags(td, sv[1], socket_flags);
935 error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
939 (void)kern_close(td, sv[0]);
940 (void)kern_close(td, sv[1]);
945 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
946 struct linux_send_args {
954 linux_send(struct thread *td, struct linux_send_args *args)
956 struct sendto_args /* {
965 bsd_args.s = args->s;
966 bsd_args.buf = (caddr_t)PTRIN(args->msg);
967 bsd_args.len = args->len;
968 bsd_args.flags = args->flags;
971 return sys_sendto(td, &bsd_args);
974 struct linux_recv_args {
982 linux_recv(struct thread *td, struct linux_recv_args *args)
984 struct recvfrom_args /* {
989 struct sockaddr *from;
990 socklen_t fromlenaddr;
993 bsd_args.s = args->s;
994 bsd_args.buf = (caddr_t)PTRIN(args->msg);
995 bsd_args.len = args->len;
996 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
997 bsd_args.from = NULL;
998 bsd_args.fromlenaddr = 0;
999 return (sys_recvfrom(td, &bsd_args));
1001 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1004 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1010 if (linux_check_hdrincl(td, args->s) == 0)
1011 /* IP_HDRINCL set, tweak the packet before sending */
1012 return (linux_sendto_hdrincl(td, args));
1014 msg.msg_name = PTRIN(args->to);
1015 msg.msg_namelen = args->tolen;
1016 msg.msg_iov = &aiov;
1018 msg.msg_control = NULL;
1020 aiov.iov_base = PTRIN(args->msg);
1021 aiov.iov_len = args->len;
1022 error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1028 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1030 struct recvfrom_args /* {
1035 struct sockaddr * __restrict from;
1036 socklen_t * __restrict fromlenaddr;
1041 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1044 bsd_args.s = args->s;
1045 bsd_args.buf = PTRIN(args->buf);
1046 bsd_args.len = args->len;
1047 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1049 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1050 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1052 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1053 error = sys_recvfrom(td, &bsd_args);
1054 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1059 error = linux_sa_put((struct osockaddr *)
1068 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1070 struct cmsghdr *cmsg;
1071 struct cmsgcred cmcred;
1072 struct mbuf *control;
1074 struct l_cmsghdr linux_cmsg;
1075 struct l_cmsghdr *ptr_cmsg;
1076 struct l_msghdr linux_msg;
1079 struct sockaddr *sa;
1080 sa_family_t sa_family;
1084 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1089 * Some Linux applications (ping) define a non-NULL control data
1090 * pointer, but a msg_controllen of 0, which is not allowed in the
1091 * FreeBSD system call interface. NULL the msg_control pointer in
1092 * order to handle this case. This should be checked, but allows the
1093 * Linux ping to work.
1095 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1096 linux_msg.msg_control = PTROUT(NULL);
1098 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1102 #ifdef COMPAT_LINUX32
1103 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1106 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1114 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1115 error = kern_getsockname(td, args->s, &sa, &datalen);
1118 sa_family = sa->sa_family;
1122 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1123 control = m_get(M_WAITOK, MT_CONTROL);
1124 if (control == NULL)
1128 error = copyin(ptr_cmsg, &linux_cmsg,
1129 sizeof(struct l_cmsghdr));
1134 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1138 * Now we support only SCM_RIGHTS and SCM_CRED,
1139 * so return EINVAL in any other cmsg_type
1142 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1144 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1145 if (cmsg->cmsg_type == -1
1146 || cmsg->cmsg_level != SOL_SOCKET)
1150 * Some applications (e.g. pulseaudio) attempt to
1151 * send ancillary data even if the underlying protocol
1152 * doesn't support it which is not allowed in the
1153 * FreeBSD system call interface.
1155 if (sa_family != AF_UNIX)
1158 data = LINUX_CMSG_DATA(ptr_cmsg);
1159 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1161 switch (cmsg->cmsg_type)
1168 datalen = sizeof(cmcred);
1171 * The lower levels will fill in the structure
1173 bzero(data, datalen);
1177 cmsg->cmsg_len = CMSG_LEN(datalen);
1180 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1182 if (!m_append(control, datalen, (c_caddr_t)data))
1184 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1186 if (m_length(control, NULL) == 0) {
1194 error = linux_sendit(td, args->s, &msg, args->flags, control,
1205 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1208 struct cmsgcred *cmcred;
1210 struct l_cmsghdr *linux_cmsg = NULL;
1211 struct l_ucred linux_ucred;
1212 socklen_t datalen, outlen;
1213 struct l_msghdr linux_msg;
1214 struct iovec *iov, *uiov;
1215 struct mbuf *control = NULL;
1216 struct mbuf **controlp;
1219 int error, i, fd, fds, *fdp;
1221 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1225 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1229 #ifdef COMPAT_LINUX32
1230 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1233 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1239 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1247 controlp = (msg.msg_control != NULL) ? &control : NULL;
1248 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1253 error = bsd_to_linux_msghdr(&msg, &linux_msg);
1257 if (linux_msg.msg_name) {
1258 error = bsd_to_linux_sockaddr((struct sockaddr *)
1259 PTRIN(linux_msg.msg_name));
1263 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1264 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1269 outbuf = PTRIN(linux_msg.msg_control);
1273 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1275 msg.msg_control = mtod(control, struct cmsghdr *);
1276 msg.msg_controllen = control->m_len;
1278 cm = CMSG_FIRSTHDR(&msg);
1280 while (cm != NULL) {
1281 linux_cmsg->cmsg_type =
1282 bsd_to_linux_cmsg_type(cm->cmsg_type);
1283 linux_cmsg->cmsg_level =
1284 bsd_to_linux_sockopt_level(cm->cmsg_level);
1285 if (linux_cmsg->cmsg_type == -1
1286 || cm->cmsg_level != SOL_SOCKET)
1292 data = CMSG_DATA(cm);
1293 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1295 switch (cm->cmsg_type)
1298 if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1299 fds = datalen / sizeof(int);
1301 for (i = 0; i < fds; i++) {
1303 (void)kern_fcntl(td, fd,
1304 F_SETFD, FD_CLOEXEC);
1311 * Currently LOCAL_CREDS is never in
1312 * effect for Linux so no need to worry
1315 if (datalen != sizeof(*cmcred)) {
1319 cmcred = (struct cmsgcred *)data;
1320 bzero(&linux_ucred, sizeof(linux_ucred));
1321 linux_ucred.pid = cmcred->cmcred_pid;
1322 linux_ucred.uid = cmcred->cmcred_uid;
1323 linux_ucred.gid = cmcred->cmcred_gid;
1324 data = &linux_ucred;
1325 datalen = sizeof(linux_ucred);
1329 if (outlen + LINUX_CMSG_LEN(datalen) >
1330 linux_msg.msg_controllen) {
1335 linux_msg.msg_flags |=
1341 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1343 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1346 outbuf += L_CMSG_HDRSZ;
1348 error = copyout(data, outbuf, datalen);
1352 outbuf += LINUX_CMSG_ALIGN(datalen);
1353 outlen += LINUX_CMSG_LEN(datalen);
1355 cm = CMSG_NXTHDR(&msg, cm);
1360 linux_msg.msg_controllen = outlen;
1361 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1366 free(linux_cmsg, M_TEMP);
1372 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1374 struct shutdown_args /* {
1379 bsd_args.s = args->s;
1380 bsd_args.how = args->how;
1381 return (sys_shutdown(td, &bsd_args));
1385 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1387 struct setsockopt_args /* {
1398 bsd_args.s = args->s;
1399 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1400 switch (bsd_args.level) {
1402 name = linux_to_bsd_so_sockopt(args->optname);
1407 error = copyin(PTRIN(args->optval), &linux_tv,
1411 tv.tv_sec = linux_tv.tv_sec;
1412 tv.tv_usec = linux_tv.tv_usec;
1413 return (kern_setsockopt(td, args->s, bsd_args.level,
1414 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1422 name = linux_to_bsd_ip_sockopt(args->optname);
1425 name = linux_to_bsd_tcp_sockopt(args->optname);
1432 return (ENOPROTOOPT);
1434 bsd_args.name = name;
1435 bsd_args.val = PTRIN(args->optval);
1436 bsd_args.valsize = args->optlen;
1438 if (name == IPV6_NEXTHOP) {
1439 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1441 error = sys_setsockopt(td, &bsd_args);
1442 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1444 error = sys_setsockopt(td, &bsd_args);
1450 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1452 struct getsockopt_args /* {
1461 socklen_t tv_len, xulen;
1466 bsd_args.s = args->s;
1467 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1468 switch (bsd_args.level) {
1470 name = linux_to_bsd_so_sockopt(args->optname);
1475 tv_len = sizeof(tv);
1476 error = kern_getsockopt(td, args->s, bsd_args.level,
1477 name, &tv, UIO_SYSSPACE, &tv_len);
1480 linux_tv.tv_sec = tv.tv_sec;
1481 linux_tv.tv_usec = tv.tv_usec;
1482 return (copyout(&linux_tv, PTRIN(args->optval),
1486 case LOCAL_PEERCRED:
1487 if (args->optlen != sizeof(lxu))
1490 error = kern_getsockopt(td, args->s, bsd_args.level,
1491 name, &xu, UIO_SYSSPACE, &xulen);
1495 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1498 lxu.uid = xu.cr_uid;
1499 lxu.gid = xu.cr_gid;
1500 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1508 name = linux_to_bsd_ip_sockopt(args->optname);
1511 name = linux_to_bsd_tcp_sockopt(args->optname);
1520 bsd_args.name = name;
1521 bsd_args.val = PTRIN(args->optval);
1522 bsd_args.avalsize = PTRIN(args->optlen);
1524 if (name == IPV6_NEXTHOP) {
1525 error = sys_getsockopt(td, &bsd_args);
1526 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1528 error = sys_getsockopt(td, &bsd_args);
1533 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1535 /* Argument list sizes for linux_socketcall */
1537 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1539 static const unsigned char lxs_args[] = {
1540 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1541 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1542 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1543 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1544 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1545 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1546 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1547 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1548 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1549 LINUX_AL(4) /* accept4 */
1552 #define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1555 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1561 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1563 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1568 switch (args->what) {
1570 return (linux_socket(td, arg));
1572 return (linux_bind(td, arg));
1574 return (linux_connect(td, arg));
1576 return (linux_listen(td, arg));
1578 return (linux_accept(td, arg));
1579 case LINUX_GETSOCKNAME:
1580 return (linux_getsockname(td, arg));
1581 case LINUX_GETPEERNAME:
1582 return (linux_getpeername(td, arg));
1583 case LINUX_SOCKETPAIR:
1584 return (linux_socketpair(td, arg));
1586 return (linux_send(td, arg));
1588 return (linux_recv(td, arg));
1590 return (linux_sendto(td, arg));
1591 case LINUX_RECVFROM:
1592 return (linux_recvfrom(td, arg));
1593 case LINUX_SHUTDOWN:
1594 return (linux_shutdown(td, arg));
1595 case LINUX_SETSOCKOPT:
1596 return (linux_setsockopt(td, arg));
1597 case LINUX_GETSOCKOPT:
1598 return (linux_getsockopt(td, arg));
1600 return (linux_sendmsg(td, arg));
1602 return (linux_recvmsg(td, arg));
1604 return (linux_accept4(td, arg));
1607 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1610 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */