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 *);
83 static int linux_set_socket_flags(int, int *);
86 * Reads a linux sockaddr and does any necessary translation.
87 * Linux sockaddrs don't have a length field, only a family.
88 * Copy the osockaddr structure pointed to by osa to kernel, adjust
89 * family and convert to sockaddr.
92 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
95 struct osockaddr *kosa;
97 struct sockaddr_in6 *sin6;
101 int bdom, error, hdrlen, namelen;
103 if (salen < 2 || salen > UCHAR_MAX || !osa)
109 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
110 * if it's a v4-mapped address, so reserve the proper space
113 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
114 salen += sizeof(uint32_t);
119 kosa = malloc(salen, M_SONAME, M_WAITOK);
121 if ((error = copyin(osa, kosa, salen)))
124 bdom = linux_to_bsd_domain(kosa->sa_family);
126 error = EAFNOSUPPORT;
132 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
133 * which lacks the scope id compared with RFC2553 one. If we detect
134 * the situation, reject the address and write a message to system log.
136 * Still accept addresses for which the scope id is not used.
139 if (bdom == AF_INET6) {
140 sin6 = (struct sockaddr_in6 *)kosa;
141 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
142 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
143 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
144 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
145 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
146 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
147 sin6->sin6_scope_id = 0;
150 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
155 salen -= sizeof(uint32_t);
158 if (bdom == AF_INET) {
159 if (salen < sizeof(struct sockaddr_in)) {
163 salen = sizeof(struct sockaddr_in);
166 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
167 hdrlen = offsetof(struct sockaddr_un, sun_path);
168 name = ((struct sockaddr_un *)kosa)->sun_path;
171 * Linux abstract namespace starts with a NULL byte.
172 * XXX We do not support abstract namespace yet.
174 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
176 namelen = strnlen(name, salen - hdrlen);
177 salen = hdrlen + namelen;
178 if (salen > sizeof(struct sockaddr_un)) {
179 error = ENAMETOOLONG;
184 sa = (struct sockaddr *)kosa;
185 sa->sa_family = bdom;
192 free(kosa, M_SONAME);
197 linux_to_bsd_domain(int domain)
201 case LINUX_AF_UNSPEC:
213 case LINUX_AF_APPLETALK:
214 return (AF_APPLETALK);
220 bsd_to_linux_domain(int domain)
225 return (LINUX_AF_UNSPEC);
227 return (LINUX_AF_UNIX);
229 return (LINUX_AF_INET);
231 return (LINUX_AF_INET6);
233 return (LINUX_AF_AX25);
235 return (LINUX_AF_IPX);
237 return (LINUX_AF_APPLETALK);
243 linux_to_bsd_sockopt_level(int level)
247 case LINUX_SOL_SOCKET:
254 bsd_to_linux_sockopt_level(int level)
259 return (LINUX_SOL_SOCKET);
265 linux_to_bsd_ip_sockopt(int opt)
273 case LINUX_IP_OPTIONS:
275 case LINUX_IP_MULTICAST_IF:
276 return (IP_MULTICAST_IF);
277 case LINUX_IP_MULTICAST_TTL:
278 return (IP_MULTICAST_TTL);
279 case LINUX_IP_MULTICAST_LOOP:
280 return (IP_MULTICAST_LOOP);
281 case LINUX_IP_ADD_MEMBERSHIP:
282 return (IP_ADD_MEMBERSHIP);
283 case LINUX_IP_DROP_MEMBERSHIP:
284 return (IP_DROP_MEMBERSHIP);
285 case LINUX_IP_HDRINCL:
292 linux_to_bsd_so_sockopt(int opt)
298 case LINUX_SO_REUSEADDR:
299 return (SO_REUSEADDR);
304 case LINUX_SO_DONTROUTE:
305 return (SO_DONTROUTE);
306 case LINUX_SO_BROADCAST:
307 return (SO_BROADCAST);
308 case LINUX_SO_SNDBUF:
310 case LINUX_SO_RCVBUF:
312 case LINUX_SO_KEEPALIVE:
313 return (SO_KEEPALIVE);
314 case LINUX_SO_OOBINLINE:
315 return (SO_OOBINLINE);
316 case LINUX_SO_LINGER:
318 case LINUX_SO_PEERCRED:
319 return (LOCAL_PEERCRED);
320 case LINUX_SO_RCVLOWAT:
321 return (SO_RCVLOWAT);
322 case LINUX_SO_SNDLOWAT:
323 return (SO_SNDLOWAT);
324 case LINUX_SO_RCVTIMEO:
325 return (SO_RCVTIMEO);
326 case LINUX_SO_SNDTIMEO:
327 return (SO_SNDTIMEO);
328 case LINUX_SO_TIMESTAMP:
329 return (SO_TIMESTAMP);
330 case LINUX_SO_ACCEPTCONN:
331 return (SO_ACCEPTCONN);
337 linux_to_bsd_tcp_sockopt(int opt)
341 case LINUX_TCP_NODELAY:
342 return (TCP_NODELAY);
343 case LINUX_TCP_MAXSEG:
345 case LINUX_TCP_KEEPIDLE:
346 return (TCP_KEEPIDLE);
347 case LINUX_TCP_KEEPINTVL:
348 return (TCP_KEEPINTVL);
349 case LINUX_TCP_KEEPCNT:
350 return (TCP_KEEPCNT);
351 case LINUX_TCP_MD5SIG:
358 linux_to_bsd_msg_flags(int flags)
362 if (flags & LINUX_MSG_OOB)
363 ret_flags |= MSG_OOB;
364 if (flags & LINUX_MSG_PEEK)
365 ret_flags |= MSG_PEEK;
366 if (flags & LINUX_MSG_DONTROUTE)
367 ret_flags |= MSG_DONTROUTE;
368 if (flags & LINUX_MSG_CTRUNC)
369 ret_flags |= MSG_CTRUNC;
370 if (flags & LINUX_MSG_TRUNC)
371 ret_flags |= MSG_TRUNC;
372 if (flags & LINUX_MSG_DONTWAIT)
373 ret_flags |= MSG_DONTWAIT;
374 if (flags & LINUX_MSG_EOR)
375 ret_flags |= MSG_EOR;
376 if (flags & LINUX_MSG_WAITALL)
377 ret_flags |= MSG_WAITALL;
378 if (flags & LINUX_MSG_NOSIGNAL)
379 ret_flags |= MSG_NOSIGNAL;
380 #if 0 /* not handled */
381 if (flags & LINUX_MSG_PROXY)
383 if (flags & LINUX_MSG_FIN)
385 if (flags & LINUX_MSG_SYN)
387 if (flags & LINUX_MSG_CONFIRM)
389 if (flags & LINUX_MSG_RST)
391 if (flags & LINUX_MSG_ERRQUEUE)
398 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
399 * native syscall will fault. Thus, we don't really need to check the
400 * return values for these functions.
404 bsd_to_linux_sockaddr(struct sockaddr *arg)
407 size_t sa_len = sizeof(struct sockaddr);
410 if ((error = copyin(arg, &sa, sa_len)))
413 *(u_short *)&sa = sa.sa_family;
415 error = copyout(&sa, arg, sa_len);
421 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
424 size_t sa_len = sizeof(struct sockaddr);
427 if ((error = copyin(arg, &sa, sa_len)))
430 sa.sa_family = *(sa_family_t *)&sa;
433 error = copyout(&sa, arg, sa_len);
439 linux_sa_put(struct osockaddr *osa)
445 * Only read/write the osockaddr family part, the rest is
448 error = copyin(osa, &sa, sizeof(sa.sa_family));
452 bdom = bsd_to_linux_domain(sa.sa_family);
457 error = copyout(&sa, osa, sizeof(sa.sa_family));
465 linux_to_bsd_cmsg_type(int cmsg_type)
469 case LINUX_SCM_RIGHTS:
471 case LINUX_SCM_CREDENTIALS:
478 bsd_to_linux_cmsg_type(int cmsg_type)
483 return (LINUX_SCM_RIGHTS);
485 return (LINUX_SCM_CREDENTIALS);
491 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
493 if (lhdr->msg_controllen > INT_MAX)
496 bhdr->msg_name = PTRIN(lhdr->msg_name);
497 bhdr->msg_namelen = lhdr->msg_namelen;
498 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
499 bhdr->msg_iovlen = lhdr->msg_iovlen;
500 bhdr->msg_control = PTRIN(lhdr->msg_control);
503 * msg_controllen is skipped since BSD and LINUX control messages
504 * are potentially different sizes (e.g. the cred structure used
505 * by SCM_CREDS is different between the two operating system).
507 * The caller can set it (if necessary) after converting all the
511 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
516 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
518 lhdr->msg_name = PTROUT(bhdr->msg_name);
519 lhdr->msg_namelen = bhdr->msg_namelen;
520 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
521 lhdr->msg_iovlen = bhdr->msg_iovlen;
522 lhdr->msg_control = PTROUT(bhdr->msg_control);
525 * msg_controllen is skipped since BSD and LINUX control messages
526 * are potentially different sizes (e.g. the cred structure used
527 * by SCM_CREDS is different between the two operating system).
529 * The caller can set it (if necessary) after converting all the
533 /* msg_flags skipped */
538 linux_set_socket_flags(int lflags, int *flags)
541 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
543 if (lflags & LINUX_SOCK_NONBLOCK)
544 *flags |= SOCK_NONBLOCK;
545 if (lflags & LINUX_SOCK_CLOEXEC)
546 *flags |= SOCK_CLOEXEC;
551 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
552 struct mbuf *control, enum uio_seg segflg)
557 if (mp->msg_name != NULL) {
558 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
565 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
573 /* Return 0 if IP_HDRINCL is set for the given socket. */
575 linux_check_hdrincl(struct thread *td, int s)
580 size_val = sizeof(optval);
581 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
582 &optval, UIO_SYSSPACE, &size_val);
586 return (optval == 0);
590 * Updated sendto() when IP_HDRINCL is set:
591 * tweak endian-dependent fields in the IP packet.
594 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
597 * linux_ip_copysize defines how many bytes we should copy
598 * from the beginning of the IP packet before we customize it for BSD.
599 * It should include all the fields we modify (ip_len and ip_off).
601 #define linux_ip_copysize 8
605 struct iovec aiov[1];
608 /* Check that the packet isn't too big or too small. */
609 if (linux_args->len < linux_ip_copysize ||
610 linux_args->len > IP_MAXPACKET)
613 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
615 /* Make kernel copy of the packet to be sent */
616 if ((error = copyin(PTRIN(linux_args->msg), packet,
620 /* Convert fields from Linux to BSD raw IP socket format */
621 packet->ip_len = linux_args->len;
622 packet->ip_off = ntohs(packet->ip_off);
624 /* Prepare the msghdr and iovec structures describing the new packet */
625 msg.msg_name = PTRIN(linux_args->to);
626 msg.msg_namelen = linux_args->tolen;
629 msg.msg_control = NULL;
631 aiov[0].iov_base = (char *)packet;
632 aiov[0].iov_len = linux_args->len;
633 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
636 free(packet, M_LINUX);
641 linux_socket(struct thread *td, struct linux_socket_args *args)
643 struct socket_args /* {
650 bsd_args.protocol = args->protocol;
651 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
652 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
654 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
656 if (retval_socket != 0)
657 return (retval_socket);
658 bsd_args.domain = linux_to_bsd_domain(args->domain);
659 if (bsd_args.domain == -1)
660 return (EAFNOSUPPORT);
662 retval_socket = sys_socket(td, &bsd_args);
664 return (retval_socket);
666 if (bsd_args.type == SOCK_RAW
667 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
668 && bsd_args.domain == PF_INET) {
669 /* It's a raw IP socket: set the IP_HDRINCL option. */
673 /* We ignore any error returned by kern_setsockopt() */
674 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
675 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
679 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
680 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
681 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
684 if (bsd_args.domain == PF_INET6) {
688 /* We ignore any error returned by setsockopt() */
689 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
690 &v6only, UIO_SYSSPACE, sizeof(v6only));
694 return (retval_socket);
698 linux_bind(struct thread *td, struct linux_bind_args *args)
703 error = linux_getsockaddr(&sa, PTRIN(args->name),
708 error = kern_bind(td, args->s, sa);
710 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
716 linux_connect(struct thread *td, struct linux_connect_args *args)
724 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
729 error = kern_connect(td, args->s, sa);
731 if (error != EISCONN)
735 * Linux doesn't return EISCONN the first time it occurs,
736 * when on a non-blocking socket. Instead it returns the
737 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
739 * XXXRW: Instead of using fgetsock(), check that it is a
740 * socket and use the file descriptor reference instead of
741 * creating a new one.
743 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
747 if (fflag & FNONBLOCK) {
749 if (so->so_emuldata == 0)
750 error = so->so_error;
751 so->so_emuldata = (void *)1;
760 linux_listen(struct thread *td, struct linux_listen_args *args)
762 struct listen_args /* {
767 bsd_args.s = args->s;
768 bsd_args.backlog = args->backlog;
769 return (sys_listen(td, &bsd_args));
773 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
774 l_uintptr_t namelen, int flags)
776 struct accept4_args /* {
778 struct sockaddr * __restrict name;
779 socklen_t * __restrict anamelen;
786 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
787 bsd_args.anamelen = PTRIN(namelen);/* XXX */
788 error = linux_set_socket_flags(flags, &bsd_args.flags);
791 error = sys_accept4(td, &bsd_args);
792 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
794 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
799 error = linux_sa_put(PTRIN(addr));
801 (void)kern_close(td, td->td_retval[0]);
802 td->td_retval[0] = 0;
808 linux_accept(struct thread *td, struct linux_accept_args *args)
811 return (linux_accept_common(td, args->s, args->addr,
816 linux_accept4(struct thread *td, struct linux_accept4_args *args)
819 return (linux_accept_common(td, args->s, args->addr,
820 args->namelen, args->flags));
824 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
826 struct getsockname_args /* {
828 struct sockaddr * __restrict asa;
829 socklen_t * __restrict alen;
833 bsd_args.fdes = args->s;
835 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
836 bsd_args.alen = PTRIN(args->namelen); /* XXX */
837 error = sys_getsockname(td, &bsd_args);
838 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
841 error = linux_sa_put(PTRIN(args->addr));
848 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
850 struct getpeername_args /* {
857 bsd_args.fdes = args->s;
858 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
859 bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
860 error = sys_getpeername(td, &bsd_args);
861 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
864 error = linux_sa_put(PTRIN(args->addr));
871 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
873 struct socketpair_args /* {
881 bsd_args.domain = linux_to_bsd_domain(args->domain);
882 if (bsd_args.domain != PF_LOCAL)
883 return (EAFNOSUPPORT);
884 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
885 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
887 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
891 if (args->protocol != 0 && args->protocol != PF_UNIX)
894 * Use of PF_UNIX as protocol argument is not right,
896 * Do not map PF_UNIX as its Linux value is identical
899 return (EPROTONOSUPPORT);
901 bsd_args.protocol = 0;
902 bsd_args.rsv = (int *)PTRIN(args->rsv);
903 return (sys_socketpair(td, &bsd_args));
906 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
907 struct linux_send_args {
915 linux_send(struct thread *td, struct linux_send_args *args)
917 struct sendto_args /* {
926 bsd_args.s = args->s;
927 bsd_args.buf = (caddr_t)PTRIN(args->msg);
928 bsd_args.len = args->len;
929 bsd_args.flags = args->flags;
932 return sys_sendto(td, &bsd_args);
935 struct linux_recv_args {
943 linux_recv(struct thread *td, struct linux_recv_args *args)
945 struct recvfrom_args /* {
950 struct sockaddr *from;
951 socklen_t fromlenaddr;
954 bsd_args.s = args->s;
955 bsd_args.buf = (caddr_t)PTRIN(args->msg);
956 bsd_args.len = args->len;
957 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
958 bsd_args.from = NULL;
959 bsd_args.fromlenaddr = 0;
960 return (sys_recvfrom(td, &bsd_args));
962 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
965 linux_sendto(struct thread *td, struct linux_sendto_args *args)
971 if (linux_check_hdrincl(td, args->s) == 0)
972 /* IP_HDRINCL set, tweak the packet before sending */
973 return (linux_sendto_hdrincl(td, args));
975 msg.msg_name = PTRIN(args->to);
976 msg.msg_namelen = args->tolen;
979 msg.msg_control = NULL;
981 aiov.iov_base = PTRIN(args->msg);
982 aiov.iov_len = args->len;
983 error = linux_sendit(td, args->s, &msg, args->flags, NULL,
989 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
995 if (PTRIN(args->fromlen) != NULL) {
996 error = copyin(PTRIN(args->fromlen), &msg.msg_namelen,
997 sizeof(msg.msg_namelen));
1001 error = linux_to_bsd_sockaddr((struct sockaddr *)PTRIN(args->from),
1006 msg.msg_namelen = 0;
1008 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from);
1009 msg.msg_iov = &aiov;
1011 aiov.iov_base = PTRIN(args->buf);
1012 aiov.iov_len = args->len;
1013 msg.msg_control = 0;
1014 msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1016 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL);
1020 if (PTRIN(args->from) != NULL) {
1021 error = bsd_to_linux_sockaddr((struct sockaddr *)
1026 error = linux_sa_put((struct osockaddr *)
1030 if (PTRIN(args->fromlen) != NULL)
1031 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1032 sizeof(msg.msg_namelen));
1038 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1041 struct cmsghdr *cmsg;
1042 struct cmsgcred cmcred;
1043 struct mbuf *control;
1045 struct l_cmsghdr linux_cmsg;
1046 struct l_cmsghdr *ptr_cmsg;
1047 struct l_msghdr linux_msg;
1050 struct sockaddr *sa;
1051 sa_family_t sa_family;
1055 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1060 * Some Linux applications (ping) define a non-NULL control data
1061 * pointer, but a msg_controllen of 0, which is not allowed in the
1062 * FreeBSD system call interface. NULL the msg_control pointer in
1063 * order to handle this case. This should be checked, but allows the
1064 * Linux ping to work.
1066 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1067 linux_msg.msg_control = PTROUT(NULL);
1069 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1073 #ifdef COMPAT_LINUX32
1074 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1077 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1085 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1086 error = kern_getsockname(td, s, &sa, &datalen);
1089 sa_family = sa->sa_family;
1093 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO);
1094 control = m_get(M_WAITOK, MT_CONTROL);
1095 if (control == NULL)
1099 error = copyin(ptr_cmsg, &linux_cmsg,
1100 sizeof(struct l_cmsghdr));
1105 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1109 * Now we support only SCM_RIGHTS and SCM_CRED,
1110 * so return EINVAL in any other cmsg_type
1113 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1115 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1116 if (cmsg->cmsg_type == -1
1117 || cmsg->cmsg_level != SOL_SOCKET)
1121 * Some applications (e.g. pulseaudio) attempt to
1122 * send ancillary data even if the underlying protocol
1123 * doesn't support it which is not allowed in the
1124 * FreeBSD system call interface.
1126 if (sa_family != AF_UNIX)
1129 data = LINUX_CMSG_DATA(ptr_cmsg);
1130 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1132 switch (cmsg->cmsg_type)
1139 datalen = sizeof(cmcred);
1142 * The lower levels will fill in the structure
1144 bzero(data, datalen);
1148 cmsg->cmsg_len = CMSG_LEN(datalen);
1151 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1153 if (!m_append(control, datalen, (c_caddr_t)data))
1155 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1157 if (m_length(control, NULL) == 0) {
1165 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1170 free(cmsg, M_LINUX);
1175 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1178 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1183 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1185 struct l_mmsghdr *msg;
1187 int error, datagrams;
1189 if (args->vlen > UIO_MAXIOV)
1190 args->vlen = UIO_MAXIOV;
1192 msg = PTRIN(args->msg);
1194 while (datagrams < args->vlen) {
1195 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1200 retval = td->td_retval[0];
1201 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1208 td->td_retval[0] = datagrams;
1213 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1214 l_uint flags, struct msghdr *msg)
1217 struct cmsgcred *cmcred;
1218 struct l_cmsghdr *linux_cmsg = NULL;
1219 struct l_ucred linux_ucred;
1220 socklen_t datalen, outlen;
1221 struct l_msghdr linux_msg;
1222 struct iovec *iov, *uiov;
1223 struct mbuf *control = NULL;
1224 struct mbuf **controlp;
1227 int error, i, fd, fds, *fdp;
1229 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1233 error = linux_to_bsd_msghdr(msg, &linux_msg);
1237 #ifdef COMPAT_LINUX32
1238 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1241 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1246 if (msg->msg_name) {
1247 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name,
1253 uiov = msg->msg_iov;
1255 controlp = (msg->msg_control != NULL) ? &control : NULL;
1256 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp);
1257 msg->msg_iov = uiov;
1261 error = bsd_to_linux_msghdr(msg, &linux_msg);
1265 if (linux_msg.msg_name) {
1266 error = bsd_to_linux_sockaddr((struct sockaddr *)
1267 PTRIN(linux_msg.msg_name));
1271 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1272 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1277 outbuf = PTRIN(linux_msg.msg_control);
1281 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1283 msg->msg_control = mtod(control, struct cmsghdr *);
1284 msg->msg_controllen = control->m_len;
1286 cm = CMSG_FIRSTHDR(msg);
1288 while (cm != NULL) {
1289 linux_cmsg->cmsg_type =
1290 bsd_to_linux_cmsg_type(cm->cmsg_type);
1291 linux_cmsg->cmsg_level =
1292 bsd_to_linux_sockopt_level(cm->cmsg_level);
1293 if (linux_cmsg->cmsg_type == -1
1294 || cm->cmsg_level != SOL_SOCKET)
1300 data = CMSG_DATA(cm);
1301 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1303 switch (cm->cmsg_type)
1306 if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1307 fds = datalen / sizeof(int);
1309 for (i = 0; i < fds; i++) {
1311 (void)kern_fcntl(td, fd,
1312 F_SETFD, FD_CLOEXEC);
1319 * Currently LOCAL_CREDS is never in
1320 * effect for Linux so no need to worry
1323 if (datalen != sizeof(*cmcred)) {
1327 cmcred = (struct cmsgcred *)data;
1328 bzero(&linux_ucred, sizeof(linux_ucred));
1329 linux_ucred.pid = cmcred->cmcred_pid;
1330 linux_ucred.uid = cmcred->cmcred_uid;
1331 linux_ucred.gid = cmcred->cmcred_gid;
1332 data = &linux_ucred;
1333 datalen = sizeof(linux_ucred);
1337 if (outlen + LINUX_CMSG_LEN(datalen) >
1338 linux_msg.msg_controllen) {
1343 linux_msg.msg_flags |=
1349 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1351 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1354 outbuf += L_CMSG_HDRSZ;
1356 error = copyout(data, outbuf, datalen);
1360 outbuf += LINUX_CMSG_ALIGN(datalen);
1361 outlen += LINUX_CMSG_LEN(datalen);
1363 cm = CMSG_NXTHDR(msg, cm);
1368 linux_msg.msg_controllen = outlen;
1369 error = copyout(&linux_msg, msghdr, sizeof(linux_msg));
1374 free(linux_cmsg, M_LINUX);
1380 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1382 struct msghdr bsd_msg;
1384 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1385 args->flags, &bsd_msg));
1389 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1391 struct l_mmsghdr *msg;
1392 struct msghdr bsd_msg;
1393 struct l_timespec lts;
1394 struct timespec ts, tts;
1396 int error, datagrams;
1398 if (args->timeout) {
1399 error = copyin(args->timeout, <s, sizeof(struct l_timespec));
1402 error = linux_to_native_timespec(&ts, <s);
1406 timespecadd(&tts, &ts);
1409 msg = PTRIN(args->msg);
1411 while (datagrams < args->vlen) {
1412 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1413 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1417 retval = td->td_retval[0];
1418 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1425 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1427 if (args->flags & LINUX_MSG_WAITFORONE)
1428 args->flags |= LINUX_MSG_DONTWAIT;
1431 * See BUGS section of recvmmsg(2).
1433 if (args->timeout) {
1435 timespecsub(&ts, &tts);
1436 if (!timespecisset(&ts) || ts.tv_sec > 0)
1439 /* Out of band data, return right away. */
1440 if (bsd_msg.msg_flags & MSG_OOB)
1444 td->td_retval[0] = datagrams;
1449 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1451 struct shutdown_args /* {
1456 bsd_args.s = args->s;
1457 bsd_args.how = args->how;
1458 return (sys_shutdown(td, &bsd_args));
1462 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1464 struct setsockopt_args /* {
1475 bsd_args.s = args->s;
1476 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1477 switch (bsd_args.level) {
1479 name = linux_to_bsd_so_sockopt(args->optname);
1484 error = copyin(PTRIN(args->optval), &linux_tv,
1488 tv.tv_sec = linux_tv.tv_sec;
1489 tv.tv_usec = linux_tv.tv_usec;
1490 return (kern_setsockopt(td, args->s, bsd_args.level,
1491 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1499 name = linux_to_bsd_ip_sockopt(args->optname);
1502 name = linux_to_bsd_tcp_sockopt(args->optname);
1509 return (ENOPROTOOPT);
1511 bsd_args.name = name;
1512 bsd_args.val = PTRIN(args->optval);
1513 bsd_args.valsize = args->optlen;
1515 if (name == IPV6_NEXTHOP) {
1516 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1518 error = sys_setsockopt(td, &bsd_args);
1519 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1521 error = sys_setsockopt(td, &bsd_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 name = linux_to_bsd_tcp_sockopt(args->optname);
1597 bsd_args.name = name;
1598 bsd_args.val = PTRIN(args->optval);
1599 bsd_args.avalsize = PTRIN(args->optlen);
1601 if (name == IPV6_NEXTHOP) {
1602 error = sys_getsockopt(td, &bsd_args);
1603 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1605 error = sys_getsockopt(td, &bsd_args);
1610 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1612 /* Argument list sizes for linux_socketcall */
1614 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1616 static const unsigned char lxs_args[] = {
1617 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1618 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1619 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1620 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1621 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1622 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1623 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1624 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1625 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1626 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */,
1627 LINUX_AL(4) /* sendmmsg */
1630 #define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1633 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1639 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1641 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1646 switch (args->what) {
1648 return (linux_socket(td, arg));
1650 return (linux_bind(td, arg));
1652 return (linux_connect(td, arg));
1654 return (linux_listen(td, arg));
1656 return (linux_accept(td, arg));
1657 case LINUX_GETSOCKNAME:
1658 return (linux_getsockname(td, arg));
1659 case LINUX_GETPEERNAME:
1660 return (linux_getpeername(td, arg));
1661 case LINUX_SOCKETPAIR:
1662 return (linux_socketpair(td, arg));
1664 return (linux_send(td, arg));
1666 return (linux_recv(td, arg));
1668 return (linux_sendto(td, arg));
1669 case LINUX_RECVFROM:
1670 return (linux_recvfrom(td, arg));
1671 case LINUX_SHUTDOWN:
1672 return (linux_shutdown(td, arg));
1673 case LINUX_SETSOCKOPT:
1674 return (linux_setsockopt(td, arg));
1675 case LINUX_GETSOCKOPT:
1676 return (linux_getsockopt(td, arg));
1678 return (linux_sendmsg(td, arg));
1680 return (linux_recvmsg(td, arg));
1682 return (linux_accept4(td, arg));
1683 case LINUX_RECVMMSG:
1684 return (linux_recvmmsg(td, arg));
1685 case LINUX_SENDMMSG:
1686 return (linux_sendmmsg(td, arg));
1689 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1692 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */