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/fcntl.h>
42 #include <sys/limits.h>
44 #include <sys/malloc.h>
45 #include <sys/mutex.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/syscallsubr.h>
51 #include <sys/syslog.h>
55 #include <netinet/in.h>
56 #include <netinet/in_systm.h>
57 #include <netinet/ip.h>
59 #include <netinet/ip6.h>
60 #include <netinet6/ip6_var.h>
61 #include <netinet6/in6_var.h>
65 #include <machine/../linux32/linux.h>
66 #include <machine/../linux32/linux32_proto.h>
68 #include <machine/../linux/linux.h>
69 #include <machine/../linux/linux_proto.h>
71 #include <compat/linux/linux_socket.h>
72 #include <compat/linux/linux_util.h>
74 static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
75 struct malloc_type *);
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.
83 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
87 return (do_sa_get(sap, osa, &osalen, M_SONAME));
91 * Copy the osockaddr structure pointed to by osa to kernel, adjust
92 * family and convert to sockaddr.
95 do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
96 struct malloc_type *mtype)
100 struct osockaddr *kosa;
104 struct sockaddr_in6 *sin6;
107 if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
114 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
115 * if it's a v4-mapped address, so reserve the proper space
118 if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
119 alloclen = sizeof (struct sockaddr_in6);
124 kosa = malloc(alloclen, mtype, M_WAITOK);
126 if ((error = copyin(osa, kosa, *osalen)))
129 bdom = linux_to_bsd_domain(kosa->sa_family);
137 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
138 * which lacks the scope id compared with RFC2553 one. If we detect
139 * the situation, reject the address and write a message to system log.
141 * Still accept addresses for which the scope id is not used.
143 if (oldv6size && bdom == AF_INET6) {
144 sin6 = (struct sockaddr_in6 *)kosa;
145 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
146 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
147 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
148 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
149 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
150 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
151 sin6->sin6_scope_id = 0;
154 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
161 alloclen = sizeof(struct sockaddr_in);
163 sa = (struct sockaddr *) kosa;
164 sa->sa_family = bdom;
165 sa->sa_len = alloclen;
177 linux_to_bsd_domain(int domain)
181 case LINUX_AF_UNSPEC:
193 case LINUX_AF_APPLETALK:
194 return (AF_APPLETALK);
200 bsd_to_linux_domain(int domain)
205 return (LINUX_AF_UNSPEC);
207 return (LINUX_AF_UNIX);
209 return (LINUX_AF_INET);
211 return (LINUX_AF_INET6);
213 return (LINUX_AF_AX25);
215 return (LINUX_AF_IPX);
217 return (LINUX_AF_APPLETALK);
223 linux_to_bsd_sockopt_level(int level)
227 case LINUX_SOL_SOCKET:
234 bsd_to_linux_sockopt_level(int level)
239 return (LINUX_SOL_SOCKET);
245 linux_to_bsd_ip_sockopt(int opt)
253 case LINUX_IP_OPTIONS:
255 case LINUX_IP_MULTICAST_IF:
256 return (IP_MULTICAST_IF);
257 case LINUX_IP_MULTICAST_TTL:
258 return (IP_MULTICAST_TTL);
259 case LINUX_IP_MULTICAST_LOOP:
260 return (IP_MULTICAST_LOOP);
261 case LINUX_IP_ADD_MEMBERSHIP:
262 return (IP_ADD_MEMBERSHIP);
263 case LINUX_IP_DROP_MEMBERSHIP:
264 return (IP_DROP_MEMBERSHIP);
265 case LINUX_IP_HDRINCL:
272 linux_to_bsd_so_sockopt(int opt)
278 case LINUX_SO_REUSEADDR:
279 return (SO_REUSEADDR);
284 case LINUX_SO_DONTROUTE:
285 return (SO_DONTROUTE);
286 case LINUX_SO_BROADCAST:
287 return (SO_BROADCAST);
288 case LINUX_SO_SNDBUF:
290 case LINUX_SO_RCVBUF:
292 case LINUX_SO_KEEPALIVE:
293 return (SO_KEEPALIVE);
294 case LINUX_SO_OOBINLINE:
295 return (SO_OOBINLINE);
296 case LINUX_SO_LINGER:
298 case LINUX_SO_PEERCRED:
299 return (LOCAL_PEERCRED);
300 case LINUX_SO_RCVLOWAT:
301 return (SO_RCVLOWAT);
302 case LINUX_SO_SNDLOWAT:
303 return (SO_SNDLOWAT);
304 case LINUX_SO_RCVTIMEO:
305 return (SO_RCVTIMEO);
306 case LINUX_SO_SNDTIMEO:
307 return (SO_SNDTIMEO);
308 case LINUX_SO_TIMESTAMP:
309 return (SO_TIMESTAMP);
310 case LINUX_SO_ACCEPTCONN:
311 return (SO_ACCEPTCONN);
317 linux_to_bsd_msg_flags(int flags)
321 if (flags & LINUX_MSG_OOB)
322 ret_flags |= MSG_OOB;
323 if (flags & LINUX_MSG_PEEK)
324 ret_flags |= MSG_PEEK;
325 if (flags & LINUX_MSG_DONTROUTE)
326 ret_flags |= MSG_DONTROUTE;
327 if (flags & LINUX_MSG_CTRUNC)
328 ret_flags |= MSG_CTRUNC;
329 if (flags & LINUX_MSG_TRUNC)
330 ret_flags |= MSG_TRUNC;
331 if (flags & LINUX_MSG_DONTWAIT)
332 ret_flags |= MSG_DONTWAIT;
333 if (flags & LINUX_MSG_EOR)
334 ret_flags |= MSG_EOR;
335 if (flags & LINUX_MSG_WAITALL)
336 ret_flags |= MSG_WAITALL;
337 if (flags & LINUX_MSG_NOSIGNAL)
338 ret_flags |= MSG_NOSIGNAL;
339 #if 0 /* not handled */
340 if (flags & LINUX_MSG_PROXY)
342 if (flags & LINUX_MSG_FIN)
344 if (flags & LINUX_MSG_SYN)
346 if (flags & LINUX_MSG_CONFIRM)
348 if (flags & LINUX_MSG_RST)
350 if (flags & LINUX_MSG_ERRQUEUE)
357 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
358 * native syscall will fault. Thus, we don't really need to check the
359 * return values for these functions.
363 bsd_to_linux_sockaddr(struct sockaddr *arg)
366 size_t sa_len = sizeof(struct sockaddr);
369 if ((error = copyin(arg, &sa, sa_len)))
372 *(u_short *)&sa = sa.sa_family;
374 error = copyout(&sa, arg, sa_len);
380 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
383 size_t sa_len = sizeof(struct sockaddr);
386 if ((error = copyin(arg, &sa, sa_len)))
389 sa.sa_family = *(sa_family_t *)&sa;
392 error = copyout(&sa, arg, sa_len);
399 linux_sa_put(struct osockaddr *osa)
405 * Only read/write the osockaddr family part, the rest is
408 error = copyin(osa, &sa, sizeof(sa.sa_family));
412 bdom = bsd_to_linux_domain(sa.sa_family);
417 error = copyout(&sa, osa, sizeof(sa.sa_family));
425 linux_to_bsd_cmsg_type(int cmsg_type)
429 case LINUX_SCM_RIGHTS:
436 bsd_to_linux_cmsg_type(int cmsg_type)
441 return (LINUX_SCM_RIGHTS);
447 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
449 if (lhdr->msg_controllen > INT_MAX)
452 bhdr->msg_name = PTRIN(lhdr->msg_name);
453 bhdr->msg_namelen = lhdr->msg_namelen;
454 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
455 bhdr->msg_iovlen = lhdr->msg_iovlen;
456 bhdr->msg_control = PTRIN(lhdr->msg_control);
457 bhdr->msg_controllen = lhdr->msg_controllen;
458 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
463 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
465 lhdr->msg_name = PTROUT(bhdr->msg_name);
466 lhdr->msg_namelen = bhdr->msg_namelen;
467 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
468 lhdr->msg_iovlen = bhdr->msg_iovlen;
469 lhdr->msg_control = PTROUT(bhdr->msg_control);
470 lhdr->msg_controllen = bhdr->msg_controllen;
471 /* msg_flags skipped */
476 linux_set_socket_flags(struct thread *td, int s, int flags)
480 if (flags & LINUX_SOCK_NONBLOCK) {
481 error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
485 if (flags & LINUX_SOCK_CLOEXEC) {
486 error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
494 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
495 struct mbuf *control, enum uio_seg segflg)
500 if (mp->msg_name != NULL) {
501 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
508 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
516 /* Return 0 if IP_HDRINCL is set for the given socket. */
518 linux_check_hdrincl(struct thread *td, int s)
520 int error, optval, size_val;
522 size_val = sizeof(optval);
523 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
524 &optval, UIO_SYSSPACE, &size_val);
528 return (optval == 0);
531 struct linux_sendto_args {
541 * Updated sendto() when IP_HDRINCL is set:
542 * tweak endian-dependent fields in the IP packet.
545 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
548 * linux_ip_copysize defines how many bytes we should copy
549 * from the beginning of the IP packet before we customize it for BSD.
550 * It should include all the fields we modify (ip_len and ip_off).
552 #define linux_ip_copysize 8
556 struct iovec aiov[1];
559 /* Check that the packet isn't too big or too small. */
560 if (linux_args->len < linux_ip_copysize ||
561 linux_args->len > IP_MAXPACKET)
564 packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
566 /* Make kernel copy of the packet to be sent */
567 if ((error = copyin(PTRIN(linux_args->msg), packet,
571 /* Convert fields from Linux to BSD raw IP socket format */
572 packet->ip_len = linux_args->len;
573 packet->ip_off = ntohs(packet->ip_off);
575 /* Prepare the msghdr and iovec structures describing the new packet */
576 msg.msg_name = PTRIN(linux_args->to);
577 msg.msg_namelen = linux_args->tolen;
580 msg.msg_control = NULL;
582 aiov[0].iov_base = (char *)packet;
583 aiov[0].iov_len = linux_args->len;
584 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
587 free(packet, M_TEMP);
591 struct linux_socket_args {
598 linux_socket(struct thread *td, struct linux_socket_args *args)
600 struct socket_args /* {
605 int retval_socket, socket_flags;
607 bsd_args.protocol = args->protocol;
608 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
609 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
611 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
612 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
614 bsd_args.domain = linux_to_bsd_domain(args->domain);
615 if (bsd_args.domain == -1)
616 return (EAFNOSUPPORT);
618 retval_socket = socket(td, &bsd_args);
620 return (retval_socket);
622 retval_socket = linux_set_socket_flags(td, td->td_retval[0],
625 (void)kern_close(td, td->td_retval[0]);
629 if (bsd_args.type == SOCK_RAW
630 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
631 && bsd_args.domain == PF_INET) {
632 /* It's a raw IP socket: set the IP_HDRINCL option. */
636 /* We ignore any error returned by kern_setsockopt() */
637 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
638 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
642 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
643 * default and some apps depend on this. So, set V6ONLY to 0
644 * for Linux apps if the sysctl value is set to 1.
646 if (bsd_args.domain == PF_INET6
649 * XXX: Avoid undefined symbol error with an IPv4 only
658 /* We ignore any error returned by setsockopt() */
659 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
660 &v6only, UIO_SYSSPACE, sizeof(v6only));
665 return (retval_socket);
668 struct linux_bind_args {
675 linux_bind(struct thread *td, struct linux_bind_args *args)
680 error = linux_getsockaddr(&sa, PTRIN(args->name),
685 error = kern_bind(td, args->s, sa);
687 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
692 struct linux_connect_args {
697 int linux_connect(struct thread *, struct linux_connect_args *);
700 linux_connect(struct thread *td, struct linux_connect_args *args)
707 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
712 error = kern_connect(td, args->s, sa);
714 if (error != EISCONN)
718 * Linux doesn't return EISCONN the first time it occurs,
719 * when on a non-blocking socket. Instead it returns the
720 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
722 * XXXRW: Instead of using fgetsock(), check that it is a
723 * socket and use the file descriptor reference instead of
724 * creating a new one.
726 error = fgetsock(td, args->s, &so, &fflag);
729 if (fflag & FNONBLOCK) {
731 if (so->so_emuldata == 0)
732 error = so->so_error;
733 so->so_emuldata = (void *)1;
741 struct linux_listen_args {
747 linux_listen(struct thread *td, struct linux_listen_args *args)
749 struct listen_args /* {
754 bsd_args.s = args->s;
755 bsd_args.backlog = args->backlog;
756 return (listen(td, &bsd_args));
760 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
761 l_uintptr_t namelen, int flags)
763 struct accept_args /* {
765 struct sockaddr * __restrict name;
766 socklen_t * __restrict anamelen;
770 if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
775 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
776 bsd_args.anamelen = PTRIN(namelen);/* XXX */
777 error = accept(td, &bsd_args);
778 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
780 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
786 * linux appears not to copy flags from the parent socket to the
787 * accepted one, so we must clear the flags in the new descriptor
788 * and apply the requested flags.
790 error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
793 error = linux_set_socket_flags(td, td->td_retval[0], flags);
797 error = linux_sa_put(PTRIN(addr));
801 (void)kern_close(td, td->td_retval[0]);
802 td->td_retval[0] = 0;
807 struct linux_accept_args {
814 linux_accept(struct thread *td, struct linux_accept_args *args)
817 return (linux_accept_common(td, args->s, args->addr,
821 struct linux_accept4_args {
829 linux_accept4(struct thread *td, struct linux_accept4_args *args)
832 return (linux_accept_common(td, args->s, args->addr,
833 args->namelen, args->flags));
836 struct linux_getsockname_args {
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 = getsockname(td, &bsd_args);
857 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
860 error = linux_sa_put(PTRIN(args->addr));
866 struct linux_getpeername_args {
873 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
875 struct getpeername_args /* {
882 bsd_args.fdes = args->s;
883 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
884 bsd_args.alen = (int *)PTRIN(args->namelen);
885 error = getpeername(td, &bsd_args);
886 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
889 error = linux_sa_put(PTRIN(args->addr));
895 struct linux_socketpair_args {
903 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
905 struct socketpair_args /* {
911 int error, socket_flags;
914 bsd_args.domain = linux_to_bsd_domain(args->domain);
915 if (bsd_args.domain != PF_LOCAL)
916 return (EAFNOSUPPORT);
918 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
919 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
921 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
922 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
925 if (args->protocol != 0 && args->protocol != PF_UNIX)
928 * Use of PF_UNIX as protocol argument is not right,
930 * Do not map PF_UNIX as its Linux value is identical
933 return (EPROTONOSUPPORT);
935 bsd_args.protocol = 0;
936 bsd_args.rsv = (int *)PTRIN(args->rsv);
937 error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
938 bsd_args.protocol, sv);
941 error = linux_set_socket_flags(td, sv[0], socket_flags);
944 error = linux_set_socket_flags(td, sv[1], socket_flags);
948 error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
952 (void)kern_close(td, sv[0]);
953 (void)kern_close(td, sv[1]);
958 struct linux_send_args {
966 linux_send(struct thread *td, struct linux_send_args *args)
968 struct sendto_args /* {
977 bsd_args.s = args->s;
978 bsd_args.buf = (caddr_t)PTRIN(args->msg);
979 bsd_args.len = args->len;
980 bsd_args.flags = args->flags;
983 return sendto(td, &bsd_args);
986 struct linux_recv_args {
994 linux_recv(struct thread *td, struct linux_recv_args *args)
996 struct recvfrom_args /* {
1001 struct sockaddr *from;
1002 socklen_t fromlenaddr;
1005 bsd_args.s = args->s;
1006 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1007 bsd_args.len = args->len;
1008 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1009 bsd_args.from = NULL;
1010 bsd_args.fromlenaddr = 0;
1011 return (recvfrom(td, &bsd_args));
1015 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1021 if (linux_check_hdrincl(td, args->s) == 0)
1022 /* IP_HDRINCL set, tweak the packet before sending */
1023 return (linux_sendto_hdrincl(td, args));
1025 msg.msg_name = PTRIN(args->to);
1026 msg.msg_namelen = args->tolen;
1027 msg.msg_iov = &aiov;
1029 msg.msg_control = NULL;
1031 aiov.iov_base = PTRIN(args->msg);
1032 aiov.iov_len = args->len;
1033 error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1038 struct linux_recvfrom_args {
1044 l_uintptr_t fromlen;
1048 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1050 struct recvfrom_args /* {
1055 struct sockaddr * __restrict from;
1056 socklen_t * __restrict fromlenaddr;
1061 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1064 bsd_args.s = args->s;
1065 bsd_args.buf = PTRIN(args->buf);
1066 bsd_args.len = args->len;
1067 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1069 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1070 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1072 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1073 error = recvfrom(td, &bsd_args);
1074 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1079 error = linux_sa_put((struct osockaddr *)
1087 struct linux_sendmsg_args {
1094 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1096 struct cmsghdr *cmsg;
1097 struct mbuf *control;
1099 struct l_cmsghdr linux_cmsg;
1100 struct l_cmsghdr *ptr_cmsg;
1101 struct l_msghdr linux_msg;
1107 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1110 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1115 * Some Linux applications (ping) define a non-NULL control data
1116 * pointer, but a msg_controllen of 0, which is not allowed in the
1117 * FreeBSD system call interface. NULL the msg_control pointer in
1118 * order to handle this case. This should be checked, but allows the
1119 * Linux ping to work.
1121 if (msg.msg_control != NULL && msg.msg_controllen == 0)
1122 msg.msg_control = NULL;
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);
1133 if (msg.msg_control != NULL) {
1135 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1136 control = m_get(M_WAIT, MT_CONTROL);
1137 if (control == NULL)
1139 ptr_cmsg = LINUX_CMSG_FIRSTHDR(&msg);
1142 error = copyin(ptr_cmsg, &linux_cmsg,
1143 sizeof(struct l_cmsghdr));
1148 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1152 * Now we support only SCM_RIGHTS, so return EINVAL
1153 * in any other cmsg_type
1155 if ((cmsg->cmsg_type =
1156 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type)) == -1)
1159 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1161 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1162 cmsg->cmsg_len = CMSG_LEN(datalen);
1163 data = LINUX_CMSG_DATA(ptr_cmsg);
1166 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg))
1168 if (!m_append(control, datalen, (c_caddr_t) data))
1170 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&msg, ptr_cmsg)));
1178 error = linux_sendit(td, args->s, &msg, args->flags, control,
1188 struct linux_recvmsg_args {
1195 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1199 struct l_cmsghdr *linux_cmsg = NULL;
1200 socklen_t datalen, outlen, clen;
1201 struct l_msghdr linux_msg;
1202 struct iovec *iov, *uiov;
1203 struct mbuf *control = NULL;
1204 struct mbuf **controlp;
1207 int error, i, fd, fds, *fdp;
1209 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1213 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1217 #ifdef COMPAT_LINUX32
1218 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1221 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1227 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1235 controlp = (msg.msg_control != NULL) ? &control : NULL;
1236 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1241 error = bsd_to_linux_msghdr(&msg, &linux_msg);
1245 if (linux_msg.msg_name) {
1246 error = bsd_to_linux_sockaddr((struct sockaddr *)
1247 PTRIN(linux_msg.msg_name));
1251 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1252 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1259 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1260 outbuf = PTRIN(linux_msg.msg_control);
1261 cm = mtod(control, struct cmsghdr *);
1263 clen = control->m_len;
1265 while (cm != NULL) {
1267 if ((linux_cmsg->cmsg_type =
1268 bsd_to_linux_cmsg_type(cm->cmsg_type)) == -1)
1273 data = CMSG_DATA(cm);
1274 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1276 switch (linux_cmsg->cmsg_type)
1278 case LINUX_SCM_RIGHTS:
1279 if (outlen + LINUX_CMSG_LEN(datalen) >
1280 linux_msg.msg_controllen) {
1285 linux_msg.msg_flags |=
1290 if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1291 fds = datalen / sizeof(int);
1293 for (i = 0; i < fds; i++) {
1295 (void)kern_fcntl(td, fd,
1296 F_SETFD, FD_CLOEXEC);
1302 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1303 linux_cmsg->cmsg_level =
1304 bsd_to_linux_sockopt_level(cm->cmsg_level);
1306 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1309 outbuf += L_CMSG_HDRSZ;
1311 error = copyout(data, outbuf, datalen);
1315 outbuf += LINUX_CMSG_ALIGN(datalen);
1316 outlen += LINUX_CMSG_LEN(datalen);
1317 linux_msg.msg_controllen = outlen;
1319 if (CMSG_SPACE(datalen) < clen) {
1320 clen -= CMSG_SPACE(datalen);
1321 cm = (struct cmsghdr *)
1322 ((caddr_t)cm + CMSG_SPACE(datalen));
1329 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1333 if (control != NULL)
1335 if (linux_cmsg != NULL)
1336 free(linux_cmsg, M_TEMP);
1341 struct linux_shutdown_args {
1347 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1349 struct shutdown_args /* {
1354 bsd_args.s = args->s;
1355 bsd_args.how = args->how;
1356 return (shutdown(td, &bsd_args));
1359 struct linux_setsockopt_args {
1368 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1370 struct setsockopt_args /* {
1381 bsd_args.s = args->s;
1382 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1383 switch (bsd_args.level) {
1385 name = linux_to_bsd_so_sockopt(args->optname);
1390 error = copyin(PTRIN(args->optval), &linux_tv,
1394 tv.tv_sec = linux_tv.tv_sec;
1395 tv.tv_usec = linux_tv.tv_usec;
1396 return (kern_setsockopt(td, args->s, bsd_args.level,
1397 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1405 name = linux_to_bsd_ip_sockopt(args->optname);
1408 /* Linux TCP option values match BSD's */
1409 name = args->optname;
1416 return (ENOPROTOOPT);
1418 bsd_args.name = name;
1419 bsd_args.val = PTRIN(args->optval);
1420 bsd_args.valsize = args->optlen;
1422 if (name == IPV6_NEXTHOP) {
1423 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1425 error = setsockopt(td, &bsd_args);
1426 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1428 error = setsockopt(td, &bsd_args);
1433 struct linux_getsockopt_args {
1442 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1444 struct getsockopt_args /* {
1453 socklen_t tv_len, xulen;
1458 bsd_args.s = args->s;
1459 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1460 switch (bsd_args.level) {
1462 name = linux_to_bsd_so_sockopt(args->optname);
1467 tv_len = sizeof(tv);
1468 error = kern_getsockopt(td, args->s, bsd_args.level,
1469 name, &tv, UIO_SYSSPACE, &tv_len);
1472 linux_tv.tv_sec = tv.tv_sec;
1473 linux_tv.tv_usec = tv.tv_usec;
1474 return (copyout(&linux_tv, PTRIN(args->optval),
1478 case LOCAL_PEERCRED:
1479 if (args->optlen != sizeof(lxu))
1482 error = kern_getsockopt(td, args->s, bsd_args.level,
1483 name, &xu, UIO_SYSSPACE, &xulen);
1487 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1490 lxu.uid = xu.cr_uid;
1491 lxu.gid = xu.cr_gid;
1492 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1500 name = linux_to_bsd_ip_sockopt(args->optname);
1503 /* Linux TCP option values match BSD's */
1504 name = args->optname;
1513 bsd_args.name = name;
1514 bsd_args.val = PTRIN(args->optval);
1515 bsd_args.avalsize = PTRIN(args->optlen);
1517 if (name == IPV6_NEXTHOP) {
1518 error = getsockopt(td, &bsd_args);
1519 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1521 error = getsockopt(td, &bsd_args);
1526 /* Argument list sizes for linux_socketcall */
1528 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1530 static const unsigned char lxs_args[] = {
1531 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1532 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1533 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1534 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1535 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1536 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1537 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1538 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1539 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1540 LINUX_AL(4) /* accept4 */
1543 #define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1546 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1552 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1554 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1559 switch (args->what) {
1561 return (linux_socket(td, arg));
1563 return (linux_bind(td, arg));
1565 return (linux_connect(td, arg));
1567 return (linux_listen(td, arg));
1569 return (linux_accept(td, arg));
1570 case LINUX_GETSOCKNAME:
1571 return (linux_getsockname(td, arg));
1572 case LINUX_GETPEERNAME:
1573 return (linux_getpeername(td, arg));
1574 case LINUX_SOCKETPAIR:
1575 return (linux_socketpair(td, arg));
1577 return (linux_send(td, arg));
1579 return (linux_recv(td, arg));
1581 return (linux_sendto(td, arg));
1582 case LINUX_RECVFROM:
1583 return (linux_recvfrom(td, arg));
1584 case LINUX_SHUTDOWN:
1585 return (linux_shutdown(td, arg));
1586 case LINUX_SETSOCKOPT:
1587 return (linux_setsockopt(td, arg));
1588 case LINUX_GETSOCKOPT:
1589 return (linux_getsockopt(td, arg));
1591 return (linux_sendmsg(td, arg));
1593 return (linux_recvmsg(td, arg));
1595 return (linux_accept4(td, arg));
1598 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);