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 do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
76 struct malloc_type *);
77 static int linux_to_bsd_domain(int);
80 * Reads a linux sockaddr and does any necessary translation.
81 * Linux sockaddrs don't have a length field, only a family.
84 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
88 return (do_sa_get(sap, osa, &osalen, M_SONAME));
92 * Copy the osockaddr structure pointed to by osa to kernel, adjust
93 * family and convert to sockaddr.
96 do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
97 struct malloc_type *mtype)
101 struct osockaddr *kosa;
105 struct sockaddr_in6 *sin6;
108 if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
115 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
116 * if it's a v4-mapped address, so reserve the proper space
119 if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
120 alloclen = sizeof (struct sockaddr_in6);
125 kosa = malloc(alloclen, mtype, M_WAITOK);
127 if ((error = copyin(osa, kosa, *osalen)))
130 bdom = linux_to_bsd_domain(kosa->sa_family);
132 error = EAFNOSUPPORT;
138 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
139 * which lacks the scope id compared with RFC2553 one. If we detect
140 * the situation, reject the address and write a message to system log.
142 * Still accept addresses for which the scope id is not used.
144 if (oldv6size && bdom == AF_INET6) {
145 sin6 = (struct sockaddr_in6 *)kosa;
146 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
147 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
148 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
149 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
150 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
151 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
152 sin6->sin6_scope_id = 0;
155 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
161 if (bdom == AF_INET) {
162 alloclen = sizeof(struct sockaddr_in);
163 if (*osalen < alloclen) {
169 sa = (struct sockaddr *) kosa;
170 sa->sa_family = bdom;
171 sa->sa_len = alloclen;
183 linux_to_bsd_domain(int domain)
187 case LINUX_AF_UNSPEC:
199 case LINUX_AF_APPLETALK:
200 return (AF_APPLETALK);
206 bsd_to_linux_domain(int domain)
211 return (LINUX_AF_UNSPEC);
213 return (LINUX_AF_UNIX);
215 return (LINUX_AF_INET);
217 return (LINUX_AF_INET6);
219 return (LINUX_AF_AX25);
221 return (LINUX_AF_IPX);
223 return (LINUX_AF_APPLETALK);
229 linux_to_bsd_sockopt_level(int level)
233 case LINUX_SOL_SOCKET:
240 bsd_to_linux_sockopt_level(int level)
245 return (LINUX_SOL_SOCKET);
251 linux_to_bsd_ip_sockopt(int opt)
259 case LINUX_IP_OPTIONS:
261 case LINUX_IP_MULTICAST_IF:
262 return (IP_MULTICAST_IF);
263 case LINUX_IP_MULTICAST_TTL:
264 return (IP_MULTICAST_TTL);
265 case LINUX_IP_MULTICAST_LOOP:
266 return (IP_MULTICAST_LOOP);
267 case LINUX_IP_ADD_MEMBERSHIP:
268 return (IP_ADD_MEMBERSHIP);
269 case LINUX_IP_DROP_MEMBERSHIP:
270 return (IP_DROP_MEMBERSHIP);
271 case LINUX_IP_HDRINCL:
278 linux_to_bsd_so_sockopt(int opt)
284 case LINUX_SO_REUSEADDR:
285 return (SO_REUSEADDR);
290 case LINUX_SO_DONTROUTE:
291 return (SO_DONTROUTE);
292 case LINUX_SO_BROADCAST:
293 return (SO_BROADCAST);
294 case LINUX_SO_SNDBUF:
296 case LINUX_SO_RCVBUF:
298 case LINUX_SO_KEEPALIVE:
299 return (SO_KEEPALIVE);
300 case LINUX_SO_OOBINLINE:
301 return (SO_OOBINLINE);
302 case LINUX_SO_LINGER:
304 case LINUX_SO_PEERCRED:
305 return (LOCAL_PEERCRED);
306 case LINUX_SO_RCVLOWAT:
307 return (SO_RCVLOWAT);
308 case LINUX_SO_SNDLOWAT:
309 return (SO_SNDLOWAT);
310 case LINUX_SO_RCVTIMEO:
311 return (SO_RCVTIMEO);
312 case LINUX_SO_SNDTIMEO:
313 return (SO_SNDTIMEO);
314 case LINUX_SO_TIMESTAMP:
315 return (SO_TIMESTAMP);
316 case LINUX_SO_ACCEPTCONN:
317 return (SO_ACCEPTCONN);
323 linux_to_bsd_msg_flags(int flags)
327 if (flags & LINUX_MSG_OOB)
328 ret_flags |= MSG_OOB;
329 if (flags & LINUX_MSG_PEEK)
330 ret_flags |= MSG_PEEK;
331 if (flags & LINUX_MSG_DONTROUTE)
332 ret_flags |= MSG_DONTROUTE;
333 if (flags & LINUX_MSG_CTRUNC)
334 ret_flags |= MSG_CTRUNC;
335 if (flags & LINUX_MSG_TRUNC)
336 ret_flags |= MSG_TRUNC;
337 if (flags & LINUX_MSG_DONTWAIT)
338 ret_flags |= MSG_DONTWAIT;
339 if (flags & LINUX_MSG_EOR)
340 ret_flags |= MSG_EOR;
341 if (flags & LINUX_MSG_WAITALL)
342 ret_flags |= MSG_WAITALL;
343 if (flags & LINUX_MSG_NOSIGNAL)
344 ret_flags |= MSG_NOSIGNAL;
345 #if 0 /* not handled */
346 if (flags & LINUX_MSG_PROXY)
348 if (flags & LINUX_MSG_FIN)
350 if (flags & LINUX_MSG_SYN)
352 if (flags & LINUX_MSG_CONFIRM)
354 if (flags & LINUX_MSG_RST)
356 if (flags & LINUX_MSG_ERRQUEUE)
363 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
364 * native syscall will fault. Thus, we don't really need to check the
365 * return values for these functions.
369 bsd_to_linux_sockaddr(struct sockaddr *arg)
372 size_t sa_len = sizeof(struct sockaddr);
375 if ((error = copyin(arg, &sa, sa_len)))
378 *(u_short *)&sa = sa.sa_family;
380 error = copyout(&sa, arg, sa_len);
386 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
389 size_t sa_len = sizeof(struct sockaddr);
392 if ((error = copyin(arg, &sa, sa_len)))
395 sa.sa_family = *(sa_family_t *)&sa;
398 error = copyout(&sa, arg, sa_len);
405 linux_sa_put(struct osockaddr *osa)
411 * Only read/write the osockaddr family part, the rest is
414 error = copyin(osa, &sa, sizeof(sa.sa_family));
418 bdom = bsd_to_linux_domain(sa.sa_family);
423 error = copyout(&sa, osa, sizeof(sa.sa_family));
431 linux_to_bsd_cmsg_type(int cmsg_type)
435 case LINUX_SCM_RIGHTS:
437 case LINUX_SCM_CREDENTIALS:
444 bsd_to_linux_cmsg_type(int cmsg_type)
449 return (LINUX_SCM_RIGHTS);
451 return (LINUX_SCM_CREDENTIALS);
457 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
459 if (lhdr->msg_controllen > INT_MAX)
462 bhdr->msg_name = PTRIN(lhdr->msg_name);
463 bhdr->msg_namelen = lhdr->msg_namelen;
464 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
465 bhdr->msg_iovlen = lhdr->msg_iovlen;
466 bhdr->msg_control = PTRIN(lhdr->msg_control);
469 * msg_controllen is skipped since BSD and LINUX control messages
470 * are potentially different sizes (e.g. the cred structure used
471 * by SCM_CREDS is different between the two operating system).
473 * The caller can set it (if necessary) after converting all the
477 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
482 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
484 lhdr->msg_name = PTROUT(bhdr->msg_name);
485 lhdr->msg_namelen = bhdr->msg_namelen;
486 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
487 lhdr->msg_iovlen = bhdr->msg_iovlen;
488 lhdr->msg_control = PTROUT(bhdr->msg_control);
491 * msg_controllen is skipped since BSD and LINUX control messages
492 * are potentially different sizes (e.g. the cred structure used
493 * by SCM_CREDS is different between the two operating system).
495 * The caller can set it (if necessary) after converting all the
499 /* msg_flags skipped */
504 linux_set_socket_flags(struct thread *td, int s, int flags)
508 if (flags & LINUX_SOCK_NONBLOCK) {
509 error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
513 if (flags & LINUX_SOCK_CLOEXEC) {
514 error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
522 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
523 struct mbuf *control, enum uio_seg segflg)
528 if (mp->msg_name != NULL) {
529 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
536 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
544 /* Return 0 if IP_HDRINCL is set for the given socket. */
546 linux_check_hdrincl(struct thread *td, int s)
548 int error, optval, size_val;
550 size_val = sizeof(optval);
551 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
552 &optval, UIO_SYSSPACE, &size_val);
556 return (optval == 0);
559 struct linux_sendto_args {
569 * Updated sendto() when IP_HDRINCL is set:
570 * tweak endian-dependent fields in the IP packet.
573 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
576 * linux_ip_copysize defines how many bytes we should copy
577 * from the beginning of the IP packet before we customize it for BSD.
578 * It should include all the fields we modify (ip_len and ip_off).
580 #define linux_ip_copysize 8
584 struct iovec aiov[1];
587 /* Check that the packet isn't too big or too small. */
588 if (linux_args->len < linux_ip_copysize ||
589 linux_args->len > IP_MAXPACKET)
592 packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
594 /* Make kernel copy of the packet to be sent */
595 if ((error = copyin(PTRIN(linux_args->msg), packet,
599 /* Convert fields from Linux to BSD raw IP socket format */
600 packet->ip_len = linux_args->len;
601 packet->ip_off = ntohs(packet->ip_off);
603 /* Prepare the msghdr and iovec structures describing the new packet */
604 msg.msg_name = PTRIN(linux_args->to);
605 msg.msg_namelen = linux_args->tolen;
608 msg.msg_control = NULL;
610 aiov[0].iov_base = (char *)packet;
611 aiov[0].iov_len = linux_args->len;
612 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
615 free(packet, M_TEMP);
619 struct linux_socket_args {
626 linux_socket(struct thread *td, struct linux_socket_args *args)
628 struct socket_args /* {
633 int retval_socket, socket_flags;
635 bsd_args.protocol = args->protocol;
636 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
637 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
639 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
640 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
642 bsd_args.domain = linux_to_bsd_domain(args->domain);
643 if (bsd_args.domain == -1)
644 return (EAFNOSUPPORT);
646 retval_socket = sys_socket(td, &bsd_args);
648 return (retval_socket);
650 retval_socket = linux_set_socket_flags(td, td->td_retval[0],
653 (void)kern_close(td, td->td_retval[0]);
657 if (bsd_args.type == SOCK_RAW
658 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
659 && bsd_args.domain == PF_INET) {
660 /* It's a raw IP socket: set the IP_HDRINCL option. */
664 /* We ignore any error returned by kern_setsockopt() */
665 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
666 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
670 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
671 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
672 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
675 if (bsd_args.domain == PF_INET6) {
679 /* We ignore any error returned by setsockopt() */
680 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
681 &v6only, UIO_SYSSPACE, sizeof(v6only));
686 return (retval_socket);
689 struct linux_bind_args {
696 linux_bind(struct thread *td, struct linux_bind_args *args)
701 error = linux_getsockaddr(&sa, PTRIN(args->name),
706 error = kern_bind(td, args->s, sa);
708 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
713 struct linux_connect_args {
718 int linux_connect(struct thread *, struct linux_connect_args *);
721 linux_connect(struct thread *td, struct linux_connect_args *args)
728 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
733 error = kern_connect(td, args->s, sa);
735 if (error != EISCONN)
739 * Linux doesn't return EISCONN the first time it occurs,
740 * when on a non-blocking socket. Instead it returns the
741 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
743 * XXXRW: Instead of using fgetsock(), check that it is a
744 * socket and use the file descriptor reference instead of
745 * creating a new one.
747 error = fgetsock(td, args->s, CAP_CONNECT, &so, &fflag);
750 if (fflag & FNONBLOCK) {
752 if (so->so_emuldata == 0)
753 error = so->so_error;
754 so->so_emuldata = (void *)1;
762 struct linux_listen_args {
768 linux_listen(struct thread *td, struct linux_listen_args *args)
770 struct listen_args /* {
775 bsd_args.s = args->s;
776 bsd_args.backlog = args->backlog;
777 return (sys_listen(td, &bsd_args));
781 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
782 l_uintptr_t namelen, int flags)
784 struct accept_args /* {
786 struct sockaddr * __restrict name;
787 socklen_t * __restrict anamelen;
791 if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
796 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
797 bsd_args.anamelen = PTRIN(namelen);/* XXX */
798 error = sys_accept(td, &bsd_args);
799 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
801 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
807 * linux appears not to copy flags from the parent socket to the
808 * accepted one, so we must clear the flags in the new descriptor
809 * and apply the requested flags.
811 error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
814 error = linux_set_socket_flags(td, td->td_retval[0], flags);
818 error = linux_sa_put(PTRIN(addr));
822 (void)kern_close(td, td->td_retval[0]);
823 td->td_retval[0] = 0;
828 struct linux_accept_args {
835 linux_accept(struct thread *td, struct linux_accept_args *args)
838 return (linux_accept_common(td, args->s, args->addr,
842 struct linux_accept4_args {
850 linux_accept4(struct thread *td, struct linux_accept4_args *args)
853 return (linux_accept_common(td, args->s, args->addr,
854 args->namelen, args->flags));
857 struct linux_getsockname_args {
864 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
866 struct getsockname_args /* {
868 struct sockaddr * __restrict asa;
869 socklen_t * __restrict alen;
873 bsd_args.fdes = args->s;
875 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
876 bsd_args.alen = PTRIN(args->namelen); /* XXX */
877 error = sys_getsockname(td, &bsd_args);
878 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
881 error = linux_sa_put(PTRIN(args->addr));
887 struct linux_getpeername_args {
894 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
896 struct getpeername_args /* {
903 bsd_args.fdes = args->s;
904 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
905 bsd_args.alen = (int *)PTRIN(args->namelen);
906 error = sys_getpeername(td, &bsd_args);
907 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
910 error = linux_sa_put(PTRIN(args->addr));
916 struct linux_socketpair_args {
924 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
926 struct socketpair_args /* {
932 int error, socket_flags;
935 bsd_args.domain = linux_to_bsd_domain(args->domain);
936 if (bsd_args.domain != PF_LOCAL)
937 return (EAFNOSUPPORT);
939 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
940 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
942 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
943 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
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 error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
959 bsd_args.protocol, sv);
962 error = linux_set_socket_flags(td, sv[0], socket_flags);
965 error = linux_set_socket_flags(td, sv[1], socket_flags);
969 error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
973 (void)kern_close(td, sv[0]);
974 (void)kern_close(td, sv[1]);
979 struct linux_send_args {
987 linux_send(struct thread *td, struct linux_send_args *args)
989 struct sendto_args /* {
998 bsd_args.s = args->s;
999 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1000 bsd_args.len = args->len;
1001 bsd_args.flags = args->flags;
1004 return sys_sendto(td, &bsd_args);
1007 struct linux_recv_args {
1015 linux_recv(struct thread *td, struct linux_recv_args *args)
1017 struct recvfrom_args /* {
1022 struct sockaddr *from;
1023 socklen_t fromlenaddr;
1026 bsd_args.s = args->s;
1027 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1028 bsd_args.len = args->len;
1029 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1030 bsd_args.from = NULL;
1031 bsd_args.fromlenaddr = 0;
1032 return (sys_recvfrom(td, &bsd_args));
1036 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1042 if (linux_check_hdrincl(td, args->s) == 0)
1043 /* IP_HDRINCL set, tweak the packet before sending */
1044 return (linux_sendto_hdrincl(td, args));
1046 msg.msg_name = PTRIN(args->to);
1047 msg.msg_namelen = args->tolen;
1048 msg.msg_iov = &aiov;
1050 msg.msg_control = NULL;
1052 aiov.iov_base = PTRIN(args->msg);
1053 aiov.iov_len = args->len;
1054 error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1059 struct linux_recvfrom_args {
1065 l_uintptr_t fromlen;
1069 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1071 struct recvfrom_args /* {
1076 struct sockaddr * __restrict from;
1077 socklen_t * __restrict fromlenaddr;
1082 if ((error = copyin(PTRIN(args->fromlen), &len, sizeof(size_t))))
1085 bsd_args.s = args->s;
1086 bsd_args.buf = PTRIN(args->buf);
1087 bsd_args.len = args->len;
1088 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1090 bsd_args.from = (struct sockaddr * __restrict)PTRIN(args->from);
1091 bsd_args.fromlenaddr = PTRIN(args->fromlen);/* XXX */
1093 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1094 error = sys_recvfrom(td, &bsd_args);
1095 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1100 error = linux_sa_put((struct osockaddr *)
1108 struct linux_sendmsg_args {
1115 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1117 struct cmsghdr *cmsg;
1118 struct cmsgcred cmcred;
1119 struct mbuf *control;
1121 struct l_cmsghdr linux_cmsg;
1122 struct l_cmsghdr *ptr_cmsg;
1123 struct l_msghdr linux_msg;
1126 struct sockaddr *sa;
1127 sa_family_t sa_family;
1131 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1136 * Some Linux applications (ping) define a non-NULL control data
1137 * pointer, but a msg_controllen of 0, which is not allowed in the
1138 * FreeBSD system call interface. NULL the msg_control pointer in
1139 * order to handle this case. This should be checked, but allows the
1140 * Linux ping to work.
1142 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1143 linux_msg.msg_control = PTROUT(NULL);
1145 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1149 #ifdef COMPAT_LINUX32
1150 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1153 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1161 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1162 error = kern_getsockname(td, args->s, &sa, &datalen);
1165 sa_family = sa->sa_family;
1169 cmsg = malloc(CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1170 control = m_get(M_WAIT, MT_CONTROL);
1171 if (control == NULL)
1175 error = copyin(ptr_cmsg, &linux_cmsg,
1176 sizeof(struct l_cmsghdr));
1181 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1185 * Now we support only SCM_RIGHTS and SCM_CRED,
1186 * so return EINVAL in any other cmsg_type
1189 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1191 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1192 if (cmsg->cmsg_type == -1
1193 || cmsg->cmsg_level != SOL_SOCKET)
1197 * Some applications (e.g. pulseaudio) attempt to
1198 * send ancillary data even if the underlying protocol
1199 * doesn't support it which is not allowed in the
1200 * FreeBSD system call interface.
1202 if (sa_family != AF_UNIX)
1205 data = LINUX_CMSG_DATA(ptr_cmsg);
1206 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1208 switch (cmsg->cmsg_type)
1215 datalen = sizeof(cmcred);
1218 * The lower levels will fill in the structure
1220 bzero(data, datalen);
1224 cmsg->cmsg_len = CMSG_LEN(datalen);
1227 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t) cmsg))
1229 if (!m_append(control, datalen, (c_caddr_t) data))
1231 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1233 if (m_length(control, NULL) == 0) {
1241 error = linux_sendit(td, args->s, &msg, args->flags, control,
1251 struct linux_recvmsg_args {
1258 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1261 struct cmsgcred *cmcred;
1263 struct l_cmsghdr *linux_cmsg = NULL;
1264 struct l_ucred linux_ucred;
1265 socklen_t datalen, outlen;
1266 struct l_msghdr linux_msg;
1267 struct iovec *iov, *uiov;
1268 struct mbuf *control = NULL;
1269 struct mbuf **controlp;
1272 int error, i, fd, fds, *fdp;
1274 error = copyin(PTRIN(args->msg), &linux_msg, sizeof(linux_msg));
1278 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1282 #ifdef COMPAT_LINUX32
1283 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1286 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1292 error = linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1300 controlp = (msg.msg_control != NULL) ? &control : NULL;
1301 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, controlp);
1306 error = bsd_to_linux_msghdr(&msg, &linux_msg);
1310 if (linux_msg.msg_name) {
1311 error = bsd_to_linux_sockaddr((struct sockaddr *)
1312 PTRIN(linux_msg.msg_name));
1316 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1317 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1322 outbuf = PTRIN(linux_msg.msg_control);
1326 linux_cmsg = malloc(L_CMSG_HDRSZ, M_TEMP, M_WAITOK | M_ZERO);
1328 msg.msg_control = mtod(control, struct cmsghdr *);
1329 msg.msg_controllen = control->m_len;
1331 cm = CMSG_FIRSTHDR(&msg);
1333 while (cm != NULL) {
1334 linux_cmsg->cmsg_type =
1335 bsd_to_linux_cmsg_type(cm->cmsg_type);
1336 linux_cmsg->cmsg_level =
1337 bsd_to_linux_sockopt_level(cm->cmsg_level);
1338 if (linux_cmsg->cmsg_type == -1
1339 || cm->cmsg_level != SOL_SOCKET)
1345 data = CMSG_DATA(cm);
1346 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1348 switch (cm->cmsg_type)
1351 if (args->flags & LINUX_MSG_CMSG_CLOEXEC) {
1352 fds = datalen / sizeof(int);
1354 for (i = 0; i < fds; i++) {
1356 (void)kern_fcntl(td, fd,
1357 F_SETFD, FD_CLOEXEC);
1364 * Currently LOCAL_CREDS is never in
1365 * effect for Linux so no need to worry
1368 if (datalen != sizeof (*cmcred)) {
1372 cmcred = (struct cmsgcred *)data;
1373 bzero(&linux_ucred, sizeof(linux_ucred));
1374 linux_ucred.pid = cmcred->cmcred_pid;
1375 linux_ucred.uid = cmcred->cmcred_uid;
1376 linux_ucred.gid = cmcred->cmcred_gid;
1377 data = &linux_ucred;
1378 datalen = sizeof(linux_ucred);
1382 if (outlen + LINUX_CMSG_LEN(datalen) >
1383 linux_msg.msg_controllen) {
1388 linux_msg.msg_flags |=
1394 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1396 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1399 outbuf += L_CMSG_HDRSZ;
1401 error = copyout(data, outbuf, datalen);
1405 outbuf += LINUX_CMSG_ALIGN(datalen);
1406 outlen += LINUX_CMSG_LEN(datalen);
1408 cm = CMSG_NXTHDR(&msg, cm);
1413 linux_msg.msg_controllen = outlen;
1414 error = copyout(&linux_msg, PTRIN(args->msg), sizeof(linux_msg));
1418 if (control != NULL)
1420 if (linux_cmsg != NULL)
1421 free(linux_cmsg, M_TEMP);
1426 struct linux_shutdown_args {
1432 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1434 struct shutdown_args /* {
1439 bsd_args.s = args->s;
1440 bsd_args.how = args->how;
1441 return (sys_shutdown(td, &bsd_args));
1444 struct linux_setsockopt_args {
1453 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1455 struct setsockopt_args /* {
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 error = copyin(PTRIN(args->optval), &linux_tv,
1479 tv.tv_sec = linux_tv.tv_sec;
1480 tv.tv_usec = linux_tv.tv_usec;
1481 return (kern_setsockopt(td, args->s, bsd_args.level,
1482 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1490 name = linux_to_bsd_ip_sockopt(args->optname);
1493 /* Linux TCP option values match BSD's */
1494 name = args->optname;
1501 return (ENOPROTOOPT);
1503 bsd_args.name = name;
1504 bsd_args.val = PTRIN(args->optval);
1505 bsd_args.valsize = args->optlen;
1507 if (name == IPV6_NEXTHOP) {
1508 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1510 error = sys_setsockopt(td, &bsd_args);
1511 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1513 error = sys_setsockopt(td, &bsd_args);
1518 struct linux_getsockopt_args {
1527 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1529 struct getsockopt_args /* {
1538 socklen_t tv_len, xulen;
1543 bsd_args.s = args->s;
1544 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1545 switch (bsd_args.level) {
1547 name = linux_to_bsd_so_sockopt(args->optname);
1552 tv_len = sizeof(tv);
1553 error = kern_getsockopt(td, args->s, bsd_args.level,
1554 name, &tv, UIO_SYSSPACE, &tv_len);
1557 linux_tv.tv_sec = tv.tv_sec;
1558 linux_tv.tv_usec = tv.tv_usec;
1559 return (copyout(&linux_tv, PTRIN(args->optval),
1563 case LOCAL_PEERCRED:
1564 if (args->optlen != sizeof(lxu))
1567 error = kern_getsockopt(td, args->s, bsd_args.level,
1568 name, &xu, UIO_SYSSPACE, &xulen);
1572 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1575 lxu.uid = xu.cr_uid;
1576 lxu.gid = xu.cr_gid;
1577 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1585 name = linux_to_bsd_ip_sockopt(args->optname);
1588 /* Linux TCP option values match BSD's */
1589 name = args->optname;
1598 bsd_args.name = name;
1599 bsd_args.val = PTRIN(args->optval);
1600 bsd_args.avalsize = PTRIN(args->optlen);
1602 if (name == IPV6_NEXTHOP) {
1603 error = sys_getsockopt(td, &bsd_args);
1604 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1606 error = sys_getsockopt(td, &bsd_args);
1611 /* Argument list sizes for linux_socketcall */
1613 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1615 static const unsigned char lxs_args[] = {
1616 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1617 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1618 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1619 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1620 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1621 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1622 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1623 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1624 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1625 LINUX_AL(4) /* accept4 */
1628 #define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1631 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1637 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1639 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1644 switch (args->what) {
1646 return (linux_socket(td, arg));
1648 return (linux_bind(td, arg));
1650 return (linux_connect(td, arg));
1652 return (linux_listen(td, arg));
1654 return (linux_accept(td, arg));
1655 case LINUX_GETSOCKNAME:
1656 return (linux_getsockname(td, arg));
1657 case LINUX_GETPEERNAME:
1658 return (linux_getpeername(td, arg));
1659 case LINUX_SOCKETPAIR:
1660 return (linux_socketpair(td, arg));
1662 return (linux_send(td, arg));
1664 return (linux_recv(td, arg));
1666 return (linux_sendto(td, arg));
1667 case LINUX_RECVFROM:
1668 return (linux_recvfrom(td, arg));
1669 case LINUX_SHUTDOWN:
1670 return (linux_shutdown(td, arg));
1671 case LINUX_SETSOCKOPT:
1672 return (linux_setsockopt(td, arg));
1673 case LINUX_GETSOCKOPT:
1674 return (linux_getsockopt(td, arg));
1676 return (linux_sendmsg(td, arg));
1678 return (linux_recvmsg(td, arg));
1680 return (linux_accept4(td, arg));
1683 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);