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>
56 #include <netinet/in.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/ip.h>
59 #include <netinet/tcp.h>
61 #include <netinet/ip6.h>
62 #include <netinet6/ip6_var.h>
63 #include <netinet6/in6_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_file.h>
74 #include <compat/linux/linux_socket.h>
75 #include <compat/linux/linux_timer.h>
76 #include <compat/linux/linux_util.h>
78 static int linux_to_bsd_domain(int);
79 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
81 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
82 l_uint, struct msghdr *);
85 * Reads a linux sockaddr and does any necessary translation.
86 * Linux sockaddrs don't have a length field, only a family.
87 * Copy the osockaddr structure pointed to by osa to kernel, adjust
88 * family and convert to sockaddr.
91 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
94 struct osockaddr *kosa;
96 struct sockaddr_in6 *sin6;
100 int bdom, error, hdrlen, namelen;
102 if (salen < 2 || salen > UCHAR_MAX || !osa)
108 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
109 * if it's a v4-mapped address, so reserve the proper space
112 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
113 salen += sizeof(uint32_t);
118 kosa = malloc(salen, M_SONAME, M_WAITOK);
120 if ((error = copyin(osa, kosa, salen)))
123 bdom = linux_to_bsd_domain(kosa->sa_family);
125 error = EAFNOSUPPORT;
131 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
132 * which lacks the scope id compared with RFC2553 one. If we detect
133 * the situation, reject the address and write a message to system log.
135 * Still accept addresses for which the scope id is not used.
138 if (bdom == AF_INET6) {
139 sin6 = (struct sockaddr_in6 *)kosa;
140 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
141 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
142 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
143 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
144 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
145 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
146 sin6->sin6_scope_id = 0;
149 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
154 salen -= sizeof(uint32_t);
157 if (bdom == AF_INET) {
158 if (salen < sizeof(struct sockaddr_in)) {
162 salen = sizeof(struct sockaddr_in);
165 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
166 hdrlen = offsetof(struct sockaddr_un, sun_path);
167 name = ((struct sockaddr_un *)kosa)->sun_path;
170 * Linux abstract namespace starts with a NULL byte.
171 * XXX We do not support abstract namespace yet.
173 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
175 namelen = strnlen(name, salen - hdrlen);
176 salen = hdrlen + namelen;
177 if (salen > sizeof(struct sockaddr_un)) {
178 error = ENAMETOOLONG;
183 sa = (struct sockaddr *)kosa;
184 sa->sa_family = bdom;
191 free(kosa, M_SONAME);
196 linux_to_bsd_domain(int domain)
200 case LINUX_AF_UNSPEC:
212 case LINUX_AF_APPLETALK:
213 return (AF_APPLETALK);
219 bsd_to_linux_domain(int domain)
224 return (LINUX_AF_UNSPEC);
226 return (LINUX_AF_UNIX);
228 return (LINUX_AF_INET);
230 return (LINUX_AF_INET6);
232 return (LINUX_AF_AX25);
234 return (LINUX_AF_IPX);
236 return (LINUX_AF_APPLETALK);
242 linux_to_bsd_sockopt_level(int level)
246 case LINUX_SOL_SOCKET:
253 bsd_to_linux_sockopt_level(int level)
258 return (LINUX_SOL_SOCKET);
264 linux_to_bsd_ip_sockopt(int opt)
272 case LINUX_IP_OPTIONS:
274 case LINUX_IP_MULTICAST_IF:
275 return (IP_MULTICAST_IF);
276 case LINUX_IP_MULTICAST_TTL:
277 return (IP_MULTICAST_TTL);
278 case LINUX_IP_MULTICAST_LOOP:
279 return (IP_MULTICAST_LOOP);
280 case LINUX_IP_ADD_MEMBERSHIP:
281 return (IP_ADD_MEMBERSHIP);
282 case LINUX_IP_DROP_MEMBERSHIP:
283 return (IP_DROP_MEMBERSHIP);
284 case LINUX_IP_HDRINCL:
291 linux_to_bsd_so_sockopt(int opt)
297 case LINUX_SO_REUSEADDR:
298 return (SO_REUSEADDR);
303 case LINUX_SO_DONTROUTE:
304 return (SO_DONTROUTE);
305 case LINUX_SO_BROADCAST:
306 return (SO_BROADCAST);
307 case LINUX_SO_SNDBUF:
309 case LINUX_SO_RCVBUF:
311 case LINUX_SO_KEEPALIVE:
312 return (SO_KEEPALIVE);
313 case LINUX_SO_OOBINLINE:
314 return (SO_OOBINLINE);
315 case LINUX_SO_LINGER:
317 case LINUX_SO_PEERCRED:
318 return (LOCAL_PEERCRED);
319 case LINUX_SO_RCVLOWAT:
320 return (SO_RCVLOWAT);
321 case LINUX_SO_SNDLOWAT:
322 return (SO_SNDLOWAT);
323 case LINUX_SO_RCVTIMEO:
324 return (SO_RCVTIMEO);
325 case LINUX_SO_SNDTIMEO:
326 return (SO_SNDTIMEO);
327 case LINUX_SO_TIMESTAMP:
328 return (SO_TIMESTAMP);
329 case LINUX_SO_ACCEPTCONN:
330 return (SO_ACCEPTCONN);
336 linux_to_bsd_tcp_sockopt(int opt)
340 case LINUX_TCP_NODELAY:
341 return (TCP_NODELAY);
342 case LINUX_TCP_MAXSEG:
344 case LINUX_TCP_KEEPIDLE:
345 return (TCP_KEEPIDLE);
346 case LINUX_TCP_KEEPINTVL:
347 return (TCP_KEEPINTVL);
348 case LINUX_TCP_KEEPCNT:
349 return (TCP_KEEPCNT);
350 case LINUX_TCP_MD5SIG:
357 linux_to_bsd_msg_flags(int flags)
361 if (flags & LINUX_MSG_OOB)
362 ret_flags |= MSG_OOB;
363 if (flags & LINUX_MSG_PEEK)
364 ret_flags |= MSG_PEEK;
365 if (flags & LINUX_MSG_DONTROUTE)
366 ret_flags |= MSG_DONTROUTE;
367 if (flags & LINUX_MSG_CTRUNC)
368 ret_flags |= MSG_CTRUNC;
369 if (flags & LINUX_MSG_TRUNC)
370 ret_flags |= MSG_TRUNC;
371 if (flags & LINUX_MSG_DONTWAIT)
372 ret_flags |= MSG_DONTWAIT;
373 if (flags & LINUX_MSG_EOR)
374 ret_flags |= MSG_EOR;
375 if (flags & LINUX_MSG_WAITALL)
376 ret_flags |= MSG_WAITALL;
377 if (flags & LINUX_MSG_NOSIGNAL)
378 ret_flags |= MSG_NOSIGNAL;
379 #if 0 /* not handled */
380 if (flags & LINUX_MSG_PROXY)
382 if (flags & LINUX_MSG_FIN)
384 if (flags & LINUX_MSG_SYN)
386 if (flags & LINUX_MSG_CONFIRM)
388 if (flags & LINUX_MSG_RST)
390 if (flags & LINUX_MSG_ERRQUEUE)
397 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
398 * native syscall will fault. Thus, we don't really need to check the
399 * return values for these functions.
403 bsd_to_linux_sockaddr(struct sockaddr *arg)
406 size_t sa_len = sizeof(struct sockaddr);
409 if ((error = copyin(arg, &sa, sa_len)))
412 *(u_short *)&sa = sa.sa_family;
414 error = copyout(&sa, arg, sa_len);
420 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
423 size_t sa_len = sizeof(struct sockaddr);
426 if ((error = copyin(arg, &sa, sa_len)))
429 sa.sa_family = *(sa_family_t *)&sa;
432 error = copyout(&sa, arg, sa_len);
438 linux_sa_put(struct osockaddr *osa)
444 * Only read/write the osockaddr family part, the rest is
447 error = copyin(osa, &sa, sizeof(sa.sa_family));
451 bdom = bsd_to_linux_domain(sa.sa_family);
456 error = copyout(&sa, osa, sizeof(sa.sa_family));
464 linux_to_bsd_cmsg_type(int cmsg_type)
468 case LINUX_SCM_RIGHTS:
470 case LINUX_SCM_CREDENTIALS:
477 bsd_to_linux_cmsg_type(int cmsg_type)
482 return (LINUX_SCM_RIGHTS);
484 return (LINUX_SCM_CREDENTIALS);
490 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
492 if (lhdr->msg_controllen > INT_MAX)
495 bhdr->msg_name = PTRIN(lhdr->msg_name);
496 bhdr->msg_namelen = lhdr->msg_namelen;
497 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
498 bhdr->msg_iovlen = lhdr->msg_iovlen;
499 bhdr->msg_control = PTRIN(lhdr->msg_control);
502 * msg_controllen is skipped since BSD and LINUX control messages
503 * are potentially different sizes (e.g. the cred structure used
504 * by SCM_CREDS is different between the two operating system).
506 * The caller can set it (if necessary) after converting all the
510 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
515 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
517 lhdr->msg_name = PTROUT(bhdr->msg_name);
518 lhdr->msg_namelen = bhdr->msg_namelen;
519 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
520 lhdr->msg_iovlen = bhdr->msg_iovlen;
521 lhdr->msg_control = PTROUT(bhdr->msg_control);
524 * msg_controllen is skipped since BSD and LINUX control messages
525 * are potentially different sizes (e.g. the cred structure used
526 * by SCM_CREDS is different between the two operating system).
528 * The caller can set it (if necessary) after converting all the
532 /* msg_flags skipped */
537 linux_set_socket_flags(struct thread *td, int s, int flags)
541 if (flags & LINUX_SOCK_NONBLOCK) {
542 error = kern_fcntl(td, s, F_SETFL, O_NONBLOCK);
546 if (flags & LINUX_SOCK_CLOEXEC) {
547 error = kern_fcntl(td, s, F_SETFD, FD_CLOEXEC);
555 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
556 struct mbuf *control, enum uio_seg segflg)
561 if (mp->msg_name != NULL) {
562 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
569 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
577 /* Return 0 if IP_HDRINCL is set for the given socket. */
579 linux_check_hdrincl(struct thread *td, int s)
584 size_val = sizeof(optval);
585 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
586 &optval, UIO_SYSSPACE, &size_val);
590 return (optval == 0);
594 * Updated sendto() when IP_HDRINCL is set:
595 * tweak endian-dependent fields in the IP packet.
598 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
601 * linux_ip_copysize defines how many bytes we should copy
602 * from the beginning of the IP packet before we customize it for BSD.
603 * It should include all the fields we modify (ip_len and ip_off).
605 #define linux_ip_copysize 8
609 struct iovec aiov[1];
612 /* Check that the packet isn't too big or too small. */
613 if (linux_args->len < linux_ip_copysize ||
614 linux_args->len > IP_MAXPACKET)
617 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
619 /* Make kernel copy of the packet to be sent */
620 if ((error = copyin(PTRIN(linux_args->msg), packet,
624 /* Convert fields from Linux to BSD raw IP socket format */
625 packet->ip_len = linux_args->len;
626 packet->ip_off = ntohs(packet->ip_off);
628 /* Prepare the msghdr and iovec structures describing the new packet */
629 msg.msg_name = PTRIN(linux_args->to);
630 msg.msg_namelen = linux_args->tolen;
633 msg.msg_control = NULL;
635 aiov[0].iov_base = (char *)packet;
636 aiov[0].iov_len = linux_args->len;
637 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
640 free(packet, M_LINUX);
645 linux_socket(struct thread *td, struct linux_socket_args *args)
647 struct socket_args /* {
652 int retval_socket, socket_flags;
654 bsd_args.protocol = args->protocol;
655 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
656 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
658 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
659 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
661 bsd_args.domain = linux_to_bsd_domain(args->domain);
662 if (bsd_args.domain == -1)
663 return (EAFNOSUPPORT);
665 retval_socket = sys_socket(td, &bsd_args);
667 return (retval_socket);
669 retval_socket = linux_set_socket_flags(td, td->td_retval[0],
672 (void)kern_close(td, td->td_retval[0]);
676 if (bsd_args.type == SOCK_RAW
677 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
678 && bsd_args.domain == PF_INET) {
679 /* It's a raw IP socket: set the IP_HDRINCL option. */
683 /* We ignore any error returned by kern_setsockopt() */
684 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
685 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
689 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
690 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
691 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
694 if (bsd_args.domain == PF_INET6) {
698 /* We ignore any error returned by setsockopt() */
699 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
700 &v6only, UIO_SYSSPACE, sizeof(v6only));
705 return (retval_socket);
709 linux_bind(struct thread *td, struct linux_bind_args *args)
714 error = linux_getsockaddr(&sa, PTRIN(args->name),
719 error = kern_bind(td, args->s, sa);
721 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
727 linux_connect(struct thread *td, struct linux_connect_args *args)
735 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
740 error = kern_connect(td, args->s, sa);
742 if (error != EISCONN)
746 * Linux doesn't return EISCONN the first time it occurs,
747 * when on a non-blocking socket. Instead it returns the
748 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
750 * XXXRW: Instead of using fgetsock(), check that it is a
751 * socket and use the file descriptor reference instead of
752 * creating a new one.
754 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
758 if (fflag & FNONBLOCK) {
760 if (so->so_emuldata == 0)
761 error = so->so_error;
762 so->so_emuldata = (void *)1;
771 linux_listen(struct thread *td, struct linux_listen_args *args)
773 struct listen_args /* {
778 bsd_args.s = args->s;
779 bsd_args.backlog = args->backlog;
780 return (sys_listen(td, &bsd_args));
784 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
785 l_uintptr_t namelen, int flags)
787 struct accept_args /* {
789 struct sockaddr * __restrict name;
790 socklen_t * __restrict anamelen;
794 if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
799 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
800 bsd_args.anamelen = PTRIN(namelen);/* XXX */
801 error = sys_accept(td, &bsd_args);
802 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
804 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
810 * linux appears not to copy flags from the parent socket to the
811 * accepted one, so we must clear the flags in the new descriptor
812 * and apply the requested flags.
814 error = kern_fcntl(td, td->td_retval[0], F_SETFL, 0);
817 error = linux_set_socket_flags(td, td->td_retval[0], flags);
821 error = linux_sa_put(PTRIN(addr));
825 (void)kern_close(td, td->td_retval[0]);
826 td->td_retval[0] = 0;
832 linux_accept(struct thread *td, struct linux_accept_args *args)
835 return (linux_accept_common(td, args->s, args->addr,
840 linux_accept4(struct thread *td, struct linux_accept4_args *args)
843 return (linux_accept_common(td, args->s, args->addr,
844 args->namelen, args->flags));
848 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
850 struct getsockname_args /* {
852 struct sockaddr * __restrict asa;
853 socklen_t * __restrict alen;
857 bsd_args.fdes = args->s;
859 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
860 bsd_args.alen = PTRIN(args->namelen); /* XXX */
861 error = sys_getsockname(td, &bsd_args);
862 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
865 error = linux_sa_put(PTRIN(args->addr));
872 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
874 struct getpeername_args /* {
881 bsd_args.fdes = args->s;
882 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
883 bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
884 error = sys_getpeername(td, &bsd_args);
885 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
888 error = linux_sa_put(PTRIN(args->addr));
895 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
897 struct socketpair_args /* {
903 int error, socket_flags;
906 bsd_args.domain = linux_to_bsd_domain(args->domain);
907 if (bsd_args.domain != PF_LOCAL)
908 return (EAFNOSUPPORT);
910 socket_flags = args->type & ~LINUX_SOCK_TYPE_MASK;
911 if (socket_flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
913 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
914 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
917 if (args->protocol != 0 && args->protocol != PF_UNIX)
920 * Use of PF_UNIX as protocol argument is not right,
922 * Do not map PF_UNIX as its Linux value is identical
925 return (EPROTONOSUPPORT);
927 bsd_args.protocol = 0;
928 bsd_args.rsv = (int *)PTRIN(args->rsv);
929 error = kern_socketpair(td, bsd_args.domain, bsd_args.type,
930 bsd_args.protocol, sv);
933 error = linux_set_socket_flags(td, sv[0], socket_flags);
936 error = linux_set_socket_flags(td, sv[1], socket_flags);
940 error = copyout(sv, bsd_args.rsv, 2 * sizeof(int));
944 (void)kern_close(td, sv[0]);
945 (void)kern_close(td, sv[1]);
950 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
951 struct linux_send_args {
959 linux_send(struct thread *td, struct linux_send_args *args)
961 struct sendto_args /* {
970 bsd_args.s = args->s;
971 bsd_args.buf = (caddr_t)PTRIN(args->msg);
972 bsd_args.len = args->len;
973 bsd_args.flags = args->flags;
976 return sys_sendto(td, &bsd_args);
979 struct linux_recv_args {
987 linux_recv(struct thread *td, struct linux_recv_args *args)
989 struct recvfrom_args /* {
994 struct sockaddr *from;
995 socklen_t fromlenaddr;
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 = linux_to_bsd_msg_flags(args->flags);
1002 bsd_args.from = NULL;
1003 bsd_args.fromlenaddr = 0;
1004 return (sys_recvfrom(td, &bsd_args));
1006 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1009 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1015 if (linux_check_hdrincl(td, args->s) == 0)
1016 /* IP_HDRINCL set, tweak the packet before sending */
1017 return (linux_sendto_hdrincl(td, args));
1019 msg.msg_name = PTRIN(args->to);
1020 msg.msg_namelen = args->tolen;
1021 msg.msg_iov = &aiov;
1023 msg.msg_control = NULL;
1025 aiov.iov_base = PTRIN(args->msg);
1026 aiov.iov_len = args->len;
1027 error = linux_sendit(td, args->s, &msg, args->flags, NULL,
1033 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1039 if (PTRIN(args->fromlen) != NULL) {
1040 error = copyin(PTRIN(args->fromlen), &msg.msg_namelen,
1041 sizeof(msg.msg_namelen));
1045 error = linux_to_bsd_sockaddr((struct sockaddr *)PTRIN(args->from),
1050 msg.msg_namelen = 0;
1052 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from);
1053 msg.msg_iov = &aiov;
1055 aiov.iov_base = PTRIN(args->buf);
1056 aiov.iov_len = args->len;
1057 msg.msg_control = 0;
1058 msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1060 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL);
1064 if (PTRIN(args->from) != NULL) {
1065 error = bsd_to_linux_sockaddr((struct sockaddr *)
1070 error = linux_sa_put((struct osockaddr *)
1074 if (PTRIN(args->fromlen) != NULL)
1075 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1076 sizeof(msg.msg_namelen));
1082 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1085 struct cmsghdr *cmsg;
1086 struct cmsgcred cmcred;
1087 struct mbuf *control;
1089 struct l_cmsghdr linux_cmsg;
1090 struct l_cmsghdr *ptr_cmsg;
1091 struct l_msghdr linux_msg;
1094 struct sockaddr *sa;
1095 sa_family_t sa_family;
1099 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1104 * Some Linux applications (ping) define a non-NULL control data
1105 * pointer, but a msg_controllen of 0, which is not allowed in the
1106 * FreeBSD system call interface. NULL the msg_control pointer in
1107 * order to handle this case. This should be checked, but allows the
1108 * Linux ping to work.
1110 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1111 linux_msg.msg_control = PTROUT(NULL);
1113 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1117 #ifdef COMPAT_LINUX32
1118 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1121 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1129 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1130 error = kern_getsockname(td, s, &sa, &datalen);
1133 sa_family = sa->sa_family;
1137 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO);
1138 control = m_get(M_WAITOK, MT_CONTROL);
1139 if (control == NULL)
1143 error = copyin(ptr_cmsg, &linux_cmsg,
1144 sizeof(struct l_cmsghdr));
1149 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1153 * Now we support only SCM_RIGHTS and SCM_CRED,
1154 * so return EINVAL in any other cmsg_type
1157 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1159 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1160 if (cmsg->cmsg_type == -1
1161 || cmsg->cmsg_level != SOL_SOCKET)
1165 * Some applications (e.g. pulseaudio) attempt to
1166 * send ancillary data even if the underlying protocol
1167 * doesn't support it which is not allowed in the
1168 * FreeBSD system call interface.
1170 if (sa_family != AF_UNIX)
1173 data = LINUX_CMSG_DATA(ptr_cmsg);
1174 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1176 switch (cmsg->cmsg_type)
1183 datalen = sizeof(cmcred);
1186 * The lower levels will fill in the structure
1188 bzero(data, datalen);
1192 cmsg->cmsg_len = CMSG_LEN(datalen);
1195 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1197 if (!m_append(control, datalen, (c_caddr_t)data))
1199 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1201 if (m_length(control, NULL) == 0) {
1209 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1214 free(cmsg, M_LINUX);
1219 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1222 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1227 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1229 struct l_mmsghdr *msg;
1231 int error, datagrams;
1233 if (args->vlen > UIO_MAXIOV)
1234 args->vlen = UIO_MAXIOV;
1236 msg = PTRIN(args->msg);
1238 while (datagrams < args->vlen) {
1239 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1244 retval = td->td_retval[0];
1245 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1252 td->td_retval[0] = datagrams;
1257 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1258 l_uint flags, struct msghdr *msg)
1261 struct cmsgcred *cmcred;
1262 struct l_cmsghdr *linux_cmsg = NULL;
1263 struct l_ucred linux_ucred;
1264 socklen_t datalen, outlen;
1265 struct l_msghdr linux_msg;
1266 struct iovec *iov, *uiov;
1267 struct mbuf *control = NULL;
1268 struct mbuf **controlp;
1271 int error, i, fd, fds, *fdp;
1273 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1277 error = linux_to_bsd_msghdr(msg, &linux_msg);
1281 #ifdef COMPAT_LINUX32
1282 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1285 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1290 if (msg->msg_name) {
1291 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name,
1297 uiov = msg->msg_iov;
1299 controlp = (msg->msg_control != NULL) ? &control : NULL;
1300 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp);
1301 msg->msg_iov = uiov;
1305 error = bsd_to_linux_msghdr(msg, &linux_msg);
1309 if (linux_msg.msg_name) {
1310 error = bsd_to_linux_sockaddr((struct sockaddr *)
1311 PTRIN(linux_msg.msg_name));
1315 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1316 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1321 outbuf = PTRIN(linux_msg.msg_control);
1325 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1327 msg->msg_control = mtod(control, struct cmsghdr *);
1328 msg->msg_controllen = control->m_len;
1330 cm = CMSG_FIRSTHDR(msg);
1332 while (cm != NULL) {
1333 linux_cmsg->cmsg_type =
1334 bsd_to_linux_cmsg_type(cm->cmsg_type);
1335 linux_cmsg->cmsg_level =
1336 bsd_to_linux_sockopt_level(cm->cmsg_level);
1337 if (linux_cmsg->cmsg_type == -1
1338 || cm->cmsg_level != SOL_SOCKET)
1344 data = CMSG_DATA(cm);
1345 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1347 switch (cm->cmsg_type)
1350 if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1351 fds = datalen / sizeof(int);
1353 for (i = 0; i < fds; i++) {
1355 (void)kern_fcntl(td, fd,
1356 F_SETFD, FD_CLOEXEC);
1363 * Currently LOCAL_CREDS is never in
1364 * effect for Linux so no need to worry
1367 if (datalen != sizeof(*cmcred)) {
1371 cmcred = (struct cmsgcred *)data;
1372 bzero(&linux_ucred, sizeof(linux_ucred));
1373 linux_ucred.pid = cmcred->cmcred_pid;
1374 linux_ucred.uid = cmcred->cmcred_uid;
1375 linux_ucred.gid = cmcred->cmcred_gid;
1376 data = &linux_ucred;
1377 datalen = sizeof(linux_ucred);
1381 if (outlen + LINUX_CMSG_LEN(datalen) >
1382 linux_msg.msg_controllen) {
1387 linux_msg.msg_flags |=
1393 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1395 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1398 outbuf += L_CMSG_HDRSZ;
1400 error = copyout(data, outbuf, datalen);
1404 outbuf += LINUX_CMSG_ALIGN(datalen);
1405 outlen += LINUX_CMSG_LEN(datalen);
1407 cm = CMSG_NXTHDR(msg, cm);
1412 linux_msg.msg_controllen = outlen;
1413 error = copyout(&linux_msg, msghdr, sizeof(linux_msg));
1418 free(linux_cmsg, M_LINUX);
1424 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1426 struct msghdr bsd_msg;
1428 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1429 args->flags, &bsd_msg));
1433 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1435 struct l_mmsghdr *msg;
1436 struct msghdr bsd_msg;
1437 struct l_timespec lts;
1438 struct timespec ts, tts;
1440 int error, datagrams;
1442 if (args->timeout) {
1443 error = copyin(args->timeout, <s, sizeof(struct l_timespec));
1446 error = linux_to_native_timespec(&ts, <s);
1450 timespecadd(&tts, &ts);
1453 msg = PTRIN(args->msg);
1455 while (datagrams < args->vlen) {
1456 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1457 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1461 retval = td->td_retval[0];
1462 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1469 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1471 if (args->flags & LINUX_MSG_WAITFORONE)
1472 args->flags |= LINUX_MSG_DONTWAIT;
1475 * See BUGS section of recvmmsg(2).
1477 if (args->timeout) {
1479 timespecsub(&ts, &tts);
1480 if (!timespecisset(&ts) || ts.tv_sec > 0)
1483 /* Out of band data, return right away. */
1484 if (bsd_msg.msg_flags & MSG_OOB)
1488 td->td_retval[0] = datagrams;
1493 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1495 struct shutdown_args /* {
1500 bsd_args.s = args->s;
1501 bsd_args.how = args->how;
1502 return (sys_shutdown(td, &bsd_args));
1506 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1508 struct setsockopt_args /* {
1519 bsd_args.s = args->s;
1520 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1521 switch (bsd_args.level) {
1523 name = linux_to_bsd_so_sockopt(args->optname);
1528 error = copyin(PTRIN(args->optval), &linux_tv,
1532 tv.tv_sec = linux_tv.tv_sec;
1533 tv.tv_usec = linux_tv.tv_usec;
1534 return (kern_setsockopt(td, args->s, bsd_args.level,
1535 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1543 name = linux_to_bsd_ip_sockopt(args->optname);
1546 name = linux_to_bsd_tcp_sockopt(args->optname);
1553 return (ENOPROTOOPT);
1555 bsd_args.name = name;
1556 bsd_args.val = PTRIN(args->optval);
1557 bsd_args.valsize = args->optlen;
1559 if (name == IPV6_NEXTHOP) {
1560 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1562 error = sys_setsockopt(td, &bsd_args);
1563 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1565 error = sys_setsockopt(td, &bsd_args);
1571 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1573 struct getsockopt_args /* {
1582 socklen_t tv_len, xulen;
1587 bsd_args.s = args->s;
1588 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1589 switch (bsd_args.level) {
1591 name = linux_to_bsd_so_sockopt(args->optname);
1596 tv_len = sizeof(tv);
1597 error = kern_getsockopt(td, args->s, bsd_args.level,
1598 name, &tv, UIO_SYSSPACE, &tv_len);
1601 linux_tv.tv_sec = tv.tv_sec;
1602 linux_tv.tv_usec = tv.tv_usec;
1603 return (copyout(&linux_tv, PTRIN(args->optval),
1607 case LOCAL_PEERCRED:
1608 if (args->optlen != sizeof(lxu))
1611 error = kern_getsockopt(td, args->s, bsd_args.level,
1612 name, &xu, UIO_SYSSPACE, &xulen);
1616 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1619 lxu.uid = xu.cr_uid;
1620 lxu.gid = xu.cr_gid;
1621 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1629 name = linux_to_bsd_ip_sockopt(args->optname);
1632 name = linux_to_bsd_tcp_sockopt(args->optname);
1641 bsd_args.name = name;
1642 bsd_args.val = PTRIN(args->optval);
1643 bsd_args.avalsize = PTRIN(args->optlen);
1645 if (name == IPV6_NEXTHOP) {
1646 error = sys_getsockopt(td, &bsd_args);
1647 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1649 error = sys_getsockopt(td, &bsd_args);
1654 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1656 /* Argument list sizes for linux_socketcall */
1658 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1660 static const unsigned char lxs_args[] = {
1661 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1662 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1663 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1664 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1665 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1666 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1667 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1668 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1669 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1670 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */,
1671 LINUX_AL(4) /* sendmmsg */
1674 #define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1677 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1683 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1685 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1690 switch (args->what) {
1692 return (linux_socket(td, arg));
1694 return (linux_bind(td, arg));
1696 return (linux_connect(td, arg));
1698 return (linux_listen(td, arg));
1700 return (linux_accept(td, arg));
1701 case LINUX_GETSOCKNAME:
1702 return (linux_getsockname(td, arg));
1703 case LINUX_GETPEERNAME:
1704 return (linux_getpeername(td, arg));
1705 case LINUX_SOCKETPAIR:
1706 return (linux_socketpair(td, arg));
1708 return (linux_send(td, arg));
1710 return (linux_recv(td, arg));
1712 return (linux_sendto(td, arg));
1713 case LINUX_RECVFROM:
1714 return (linux_recvfrom(td, arg));
1715 case LINUX_SHUTDOWN:
1716 return (linux_shutdown(td, arg));
1717 case LINUX_SETSOCKOPT:
1718 return (linux_setsockopt(td, arg));
1719 case LINUX_GETSOCKOPT:
1720 return (linux_getsockopt(td, arg));
1722 return (linux_sendmsg(td, arg));
1724 return (linux_recvmsg(td, arg));
1726 return (linux_accept4(td, arg));
1727 case LINUX_RECVMMSG:
1728 return (linux_recvmmsg(td, arg));
1729 case LINUX_SENDMMSG:
1730 return (linux_sendmmsg(td, arg));
1733 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1736 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */