linux_socket.c revision 157369
1/*-
2 * Copyright (c) 1995 S�ren Schmidt
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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
16 *
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.
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: head/sys/compat/linux/linux_socket.c 157369 2006-04-01 15:25:01Z rwatson $");
31
32/* XXX we use functions that might not exist. */
33#include "opt_compat.h"
34#include "opt_inet6.h"
35
36#include <sys/param.h>
37#include <sys/proc.h>
38#include <sys/systm.h>
39#include <sys/sysproto.h>
40#include <sys/fcntl.h>
41#include <sys/file.h>
42#include <sys/limits.h>
43#include <sys/lock.h>
44#include <sys/malloc.h>
45#include <sys/mutex.h>
46#include <sys/mbuf.h>
47#include <sys/socket.h>
48#include <sys/socketvar.h>
49#include <sys/syscallsubr.h>
50#include <sys/uio.h>
51#include <sys/syslog.h>
52
53#include <netinet/in.h>
54#include <netinet/in_systm.h>
55#include <netinet/ip.h>
56#ifdef INET6
57#include <netinet/ip6.h>
58#include <netinet6/ip6_var.h>
59#endif
60
61#ifdef COMPAT_LINUX32
62#include <machine/../linux32/linux.h>
63#include <machine/../linux32/linux32_proto.h>
64#else
65#include <machine/../linux/linux.h>
66#include <machine/../linux/linux_proto.h>
67#endif
68#include <compat/linux/linux_socket.h>
69#include <compat/linux/linux_util.h>
70
71static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
72    struct malloc_type *);
73static int linux_to_bsd_domain(int);
74
75#ifdef __alpha__
76#define	STATIC
77#else
78#define	STATIC static
79#endif
80
81/*
82 * Reads a linux sockaddr and does any necessary translation.
83 * Linux sockaddrs don't have a length field, only a family.
84 */
85static int
86linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
87{
88	int osalen = len;
89
90	return (do_sa_get(sap, osa, &osalen, M_SONAME));
91}
92
93/*
94 * Copy the osockaddr structure pointed to by osa to kernel, adjust
95 * family and convert to sockaddr.
96 */
97static int
98do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
99    struct malloc_type *mtype)
100{
101	int error=0, bdom;
102	struct sockaddr *sa;
103	struct osockaddr *kosa;
104	int alloclen;
105#ifdef INET6
106	int oldv6size;
107	struct sockaddr_in6 *sin6;
108#endif
109
110	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
111		return (EINVAL);
112
113	alloclen = *osalen;
114#ifdef INET6
115	oldv6size = 0;
116	/*
117	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
118	 * if it's a v4-mapped address, so reserve the proper space
119	 * for it.
120	 */
121	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
122		alloclen = sizeof (struct sockaddr_in6);
123		oldv6size = 1;
124	}
125#endif
126
127	MALLOC(kosa, struct osockaddr *, alloclen, mtype, M_WAITOK);
128
129	if ((error = copyin(osa, kosa, *osalen)))
130		goto out;
131
132	bdom = linux_to_bsd_domain(kosa->sa_family);
133	if (bdom == -1) {
134		error = EINVAL;
135		goto out;
136	}
137
138#ifdef INET6
139	/*
140	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
141	 * which lacks the scope id compared with RFC2553 one. If we detect
142	 * the situation, reject the address and write a message to system log.
143	 *
144	 * Still accept addresses for which the scope id is not used.
145	 */
146	if (oldv6size && bdom == AF_INET6) {
147		sin6 = (struct sockaddr_in6 *)kosa;
148		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
149		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
150		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
151		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
152		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
153		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
154			sin6->sin6_scope_id = 0;
155		} else {
156			log(LOG_DEBUG,
157			    "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
158			error = EINVAL;
159			goto out;
160		}
161	} else
162#endif
163	if (bdom == AF_INET)
164		alloclen = sizeof(struct sockaddr_in);
165
166	sa = (struct sockaddr *) kosa;
167	sa->sa_family = bdom;
168	sa->sa_len = alloclen;
169
170	*sap = sa;
171	*osalen = alloclen;
172	return (0);
173
174out:
175	FREE(kosa, mtype);
176	return (error);
177}
178
179static int
180linux_to_bsd_domain(int domain)
181{
182
183	switch (domain) {
184	case LINUX_AF_UNSPEC:
185		return (AF_UNSPEC);
186	case LINUX_AF_UNIX:
187		return (AF_LOCAL);
188	case LINUX_AF_INET:
189		return (AF_INET);
190	case LINUX_AF_INET6:
191		return (AF_INET6);
192	case LINUX_AF_AX25:
193		return (AF_CCITT);
194	case LINUX_AF_IPX:
195		return (AF_IPX);
196	case LINUX_AF_APPLETALK:
197		return (AF_APPLETALK);
198	}
199	return (-1);
200}
201
202static int
203bsd_to_linux_domain(int domain)
204{
205
206	switch (domain) {
207	case AF_UNSPEC:
208		return (LINUX_AF_UNSPEC);
209	case AF_LOCAL:
210		return (LINUX_AF_UNIX);
211	case AF_INET:
212		return (LINUX_AF_INET);
213	case AF_INET6:
214		return (LINUX_AF_INET6);
215	case AF_CCITT:
216		return (LINUX_AF_AX25);
217	case AF_IPX:
218		return (LINUX_AF_IPX);
219	case AF_APPLETALK:
220		return (LINUX_AF_APPLETALK);
221	}
222	return (-1);
223}
224
225static int
226linux_to_bsd_sockopt_level(int level)
227{
228
229	switch (level) {
230	case LINUX_SOL_SOCKET:
231		return (SOL_SOCKET);
232	}
233	return (level);
234}
235
236static int
237bsd_to_linux_sockopt_level(int level)
238{
239
240	switch (level) {
241	case SOL_SOCKET:
242		return (LINUX_SOL_SOCKET);
243	}
244	return (level);
245}
246
247#ifndef __alpha__
248static int
249linux_to_bsd_ip_sockopt(int opt)
250{
251
252	switch (opt) {
253	case LINUX_IP_TOS:
254		return (IP_TOS);
255	case LINUX_IP_TTL:
256		return (IP_TTL);
257	case LINUX_IP_OPTIONS:
258		return (IP_OPTIONS);
259	case LINUX_IP_MULTICAST_IF:
260		return (IP_MULTICAST_IF);
261	case LINUX_IP_MULTICAST_TTL:
262		return (IP_MULTICAST_TTL);
263	case LINUX_IP_MULTICAST_LOOP:
264		return (IP_MULTICAST_LOOP);
265	case LINUX_IP_ADD_MEMBERSHIP:
266		return (IP_ADD_MEMBERSHIP);
267	case LINUX_IP_DROP_MEMBERSHIP:
268		return (IP_DROP_MEMBERSHIP);
269	case LINUX_IP_HDRINCL:
270		return (IP_HDRINCL);
271	}
272	return (-1);
273}
274
275static int
276linux_to_bsd_so_sockopt(int opt)
277{
278
279	switch (opt) {
280	case LINUX_SO_DEBUG:
281		return (SO_DEBUG);
282	case LINUX_SO_REUSEADDR:
283		return (SO_REUSEADDR);
284	case LINUX_SO_TYPE:
285		return (SO_TYPE);
286	case LINUX_SO_ERROR:
287		return (SO_ERROR);
288	case LINUX_SO_DONTROUTE:
289		return (SO_DONTROUTE);
290	case LINUX_SO_BROADCAST:
291		return (SO_BROADCAST);
292	case LINUX_SO_SNDBUF:
293		return (SO_SNDBUF);
294	case LINUX_SO_RCVBUF:
295		return (SO_RCVBUF);
296	case LINUX_SO_KEEPALIVE:
297		return (SO_KEEPALIVE);
298	case LINUX_SO_OOBINLINE:
299		return (SO_OOBINLINE);
300	case LINUX_SO_LINGER:
301		return (SO_LINGER);
302	}
303	return (-1);
304}
305#endif /* !__alpha__*/
306
307static int
308linux_to_bsd_msg_flags(int flags)
309{
310	int ret_flags = 0;
311
312	if (flags & LINUX_MSG_OOB)
313		ret_flags |= MSG_OOB;
314	if (flags & LINUX_MSG_PEEK)
315		ret_flags |= MSG_PEEK;
316	if (flags & LINUX_MSG_DONTROUTE)
317		ret_flags |= MSG_DONTROUTE;
318	if (flags & LINUX_MSG_CTRUNC)
319		ret_flags |= MSG_CTRUNC;
320	if (flags & LINUX_MSG_TRUNC)
321		ret_flags |= MSG_TRUNC;
322	if (flags & LINUX_MSG_DONTWAIT)
323		ret_flags |= MSG_DONTWAIT;
324	if (flags & LINUX_MSG_EOR)
325		ret_flags |= MSG_EOR;
326	if (flags & LINUX_MSG_WAITALL)
327		ret_flags |= MSG_WAITALL;
328	if (flags & LINUX_MSG_NOSIGNAL)
329		ret_flags |= MSG_NOSIGNAL;
330#if 0 /* not handled */
331	if (flags & LINUX_MSG_PROXY)
332		;
333	if (flags & LINUX_MSG_FIN)
334		;
335	if (flags & LINUX_MSG_SYN)
336		;
337	if (flags & LINUX_MSG_CONFIRM)
338		;
339	if (flags & LINUX_MSG_RST)
340		;
341	if (flags & LINUX_MSG_ERRQUEUE)
342		;
343#endif
344	return ret_flags;
345}
346
347/*
348* If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
349* native syscall will fault.  Thus, we don't really need to check the
350* return values for these functions.
351*/
352
353static int
354bsd_to_linux_sockaddr(struct sockaddr *arg)
355{
356	struct sockaddr sa;
357	size_t sa_len = sizeof(struct sockaddr);
358	int error;
359
360	if ((error = copyin(arg, &sa, sa_len)))
361		return (error);
362
363	*(u_short *)&sa = sa.sa_family;
364
365	error = copyout(&sa, arg, sa_len);
366
367	return (error);
368}
369
370static int
371linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
372{
373	struct sockaddr sa;
374	size_t sa_len = sizeof(struct sockaddr);
375	int error;
376
377	if ((error = copyin(arg, &sa, sa_len)))
378		return (error);
379
380	sa.sa_family = *(sa_family_t *)&sa;
381	sa.sa_len = len;
382
383	error = copyout(&sa, arg, sa_len);
384
385	return (error);
386}
387
388
389static int
390linux_sa_put(struct osockaddr *osa)
391{
392	struct osockaddr sa;
393	int error, bdom;
394
395	/*
396	 * Only read/write the osockaddr family part, the rest is
397	 * not changed.
398	 */
399	error = copyin(osa, &sa, sizeof(sa.sa_family));
400	if (error)
401		return (error);
402
403	bdom = bsd_to_linux_domain(sa.sa_family);
404	if (bdom == -1)
405		return (EINVAL);
406
407	sa.sa_family = bdom;
408	error = copyout(&sa, osa, sizeof(sa.sa_family));
409	if (error)
410		return (error);
411
412	return (0);
413}
414
415static int
416linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
417    enum uio_seg segflg)
418{
419	struct mbuf *control;
420	struct sockaddr *to;
421	int error;
422
423	if (mp->msg_name != NULL) {
424		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
425		if (error)
426			return (error);
427		mp->msg_name = to;
428	} else
429		to = NULL;
430
431	if (mp->msg_control != NULL) {
432		struct cmsghdr *cmsg;
433
434		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
435			error = EINVAL;
436			goto bad;
437		}
438		error = sockargs(&control, mp->msg_control,
439		    mp->msg_controllen, MT_CONTROL);
440		if (error)
441			goto bad;
442
443		cmsg = mtod(control, struct cmsghdr *);
444		cmsg->cmsg_level = linux_to_bsd_sockopt_level(cmsg->cmsg_level);
445	} else
446		control = NULL;
447
448	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
449	    segflg);
450
451bad:
452	if (to)
453		FREE(to, M_SONAME);
454	return (error);
455}
456
457#ifndef __alpha__
458/* Return 0 if IP_HDRINCL is set for the given socket. */
459static int
460linux_check_hdrincl(struct thread *td, int s)
461{
462	int error, optval, size_val;
463
464	size_val = sizeof(optval);
465	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
466	    &optval, UIO_SYSSPACE, &size_val);
467	if (error)
468		return (error);
469
470	return (optval == 0);
471}
472
473struct linux_sendto_args {
474	int s;
475	l_uintptr_t msg;
476	int len;
477	int flags;
478	l_uintptr_t to;
479	int tolen;
480};
481
482/*
483 * Updated sendto() when IP_HDRINCL is set:
484 * tweak endian-dependent fields in the IP packet.
485 */
486static int
487linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
488{
489/*
490 * linux_ip_copysize defines how many bytes we should copy
491 * from the beginning of the IP packet before we customize it for BSD.
492 * It should include all the fields we modify (ip_len and ip_off).
493 */
494#define linux_ip_copysize	8
495
496	struct ip *packet;
497	struct msghdr msg;
498	struct iovec aiov[1];
499	int error;
500
501	/* Check that the packet isn't too big or too small. */
502	if (linux_args->len < linux_ip_copysize ||
503	    linux_args->len > IP_MAXPACKET)
504		return (EINVAL);
505
506	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
507
508	/* Make kernel copy of the packet to be sent */
509	if ((error = copyin(PTRIN(linux_args->msg), packet,
510	    linux_args->len)))
511		goto goout;
512
513	/* Convert fields from Linux to BSD raw IP socket format */
514	packet->ip_len = linux_args->len;
515	packet->ip_off = ntohs(packet->ip_off);
516
517	/* Prepare the msghdr and iovec structures describing the new packet */
518	msg.msg_name = PTRIN(linux_args->to);
519	msg.msg_namelen = linux_args->tolen;
520	msg.msg_iov = aiov;
521	msg.msg_iovlen = 1;
522	msg.msg_control = NULL;
523	msg.msg_flags = 0;
524	aiov[0].iov_base = (char *)packet;
525	aiov[0].iov_len = linux_args->len;
526	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
527	    UIO_SYSSPACE);
528goout:
529	free(packet, M_TEMP);
530	return (error);
531}
532
533struct linux_socket_args {
534	int domain;
535	int type;
536	int protocol;
537};
538
539static int
540linux_socket(struct thread *td, struct linux_socket_args *args)
541{
542	struct linux_socket_args linux_args;
543	struct socket_args /* {
544		int domain;
545		int type;
546		int protocol;
547	} */ bsd_args;
548	int error;
549	int retval_socket;
550
551	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
552		return (error);
553
554	bsd_args.protocol = linux_args.protocol;
555	bsd_args.type = linux_args.type;
556	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
557	if (bsd_args.domain == -1)
558		return (EINVAL);
559
560	retval_socket = socket(td, &bsd_args);
561	if (bsd_args.type == SOCK_RAW
562	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
563	    && bsd_args.domain == AF_INET
564	    && retval_socket >= 0) {
565		/* It's a raw IP socket: set the IP_HDRINCL option. */
566		int hdrincl;
567
568		hdrincl = 1;
569		/* We ignore any error returned by kern_setsockopt() */
570		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
571		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
572	}
573#ifdef INET6
574	/*
575	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
576	 * default and some apps depend on this. So, set V6ONLY to 0
577	 * for Linux apps if the sysctl value is set to 1.
578	 */
579	if (bsd_args.domain == PF_INET6 && retval_socket >= 0
580#ifndef KLD_MODULE
581	    /*
582	     * XXX: Avoid undefined symbol error with an IPv4 only
583	     * kernel.
584	     */
585	    && ip6_v6only
586#endif
587	    ) {
588		int v6only;
589
590		v6only = 0;
591		/* We ignore any error returned by setsockopt() */
592		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
593		    &v6only, UIO_SYSSPACE, sizeof(v6only));
594	}
595#endif
596
597	return (retval_socket);
598}
599
600struct linux_bind_args {
601	int s;
602	l_uintptr_t name;
603	int namelen;
604};
605
606static int
607linux_bind(struct thread *td, struct linux_bind_args *args)
608{
609	struct linux_bind_args linux_args;
610	struct sockaddr *sa;
611	int error;
612
613	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
614		return (error);
615
616	error = linux_getsockaddr(&sa, PTRIN(linux_args.name),
617	    linux_args.namelen);
618	if (error)
619		return (error);
620
621	return (kern_bind(td, linux_args.s, sa));
622}
623#endif /* !__alpha__*/
624
625struct l_connect_args {
626	int s;
627	l_uintptr_t name;
628	int namelen;
629};
630#ifndef __alpha__
631#define	linux_connect_args l_connect_args
632int linux_connect(struct thread *, struct linux_connect_args *);
633#endif /* !__alpha__*/
634
635int
636linux_connect(struct thread *td, struct linux_connect_args *args)
637{
638	struct l_connect_args linux_args;
639	struct socket *so;
640	struct sockaddr *sa;
641	u_int fflag;
642	int error;
643
644#ifdef __alpha__
645	bcopy(args, &linux_args, sizeof(linux_args));
646#else
647	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
648		return (error);
649#endif /* __alpha__ */
650
651	error = linux_getsockaddr(&sa,
652	    (struct osockaddr *)PTRIN(linux_args.name),
653	    linux_args.namelen);
654	if (error)
655		return (error);
656
657	error = kern_connect(td, linux_args.s, sa);
658	if (error != EISCONN)
659		return (error);
660
661	/*
662	 * Linux doesn't return EISCONN the first time it occurs,
663	 * when on a non-blocking socket. Instead it returns the
664	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
665	 *
666	 * XXXRW: Instead of using fgetsock(), check that it is a
667	 * socket and use the file descriptor reference instead of
668	 * creating a new one.
669	 */
670	NET_LOCK_GIANT();
671	error = fgetsock(td, linux_args.s, &so, &fflag);
672	if (error == 0) {
673		error = EISCONN;
674		if (fflag & FNONBLOCK) {
675			SOCK_LOCK(so);
676			if (so->so_emuldata == 0)
677				error = so->so_error;
678			so->so_emuldata = (void *)1;
679			SOCK_UNLOCK(so);
680		}
681		fputsock(so);
682	}
683	NET_UNLOCK_GIANT();
684	return (error);
685}
686
687#ifndef __alpha__
688struct linux_listen_args {
689	int s;
690	int backlog;
691};
692
693static int
694linux_listen(struct thread *td, struct linux_listen_args *args)
695{
696	struct linux_listen_args linux_args;
697	struct listen_args /* {
698		int s;
699		int backlog;
700	} */ bsd_args;
701	int error;
702
703	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
704		return (error);
705
706	bsd_args.s = linux_args.s;
707	bsd_args.backlog = linux_args.backlog;
708	return (listen(td, &bsd_args));
709}
710#endif /* !__alpha__*/
711
712struct l_accept_args {
713	int s;
714	l_uintptr_t addr;
715	l_uintptr_t namelen;
716};
717
718#ifndef __alpha__
719#define	linux_accept_args l_accept_args
720#endif
721
722STATIC int
723linux_accept(struct thread *td, struct linux_accept_args *args)
724{
725	struct l_accept_args linux_args;
726	struct accept_args /* {
727		int	s;
728		struct sockaddr * __restrict name;
729		socklen_t * __restrict anamelen;
730	} */ bsd_args;
731	struct close_args /* {
732		int     fd;
733	} */ c_args;
734	int error, fd;
735
736	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
737		return (error);
738
739	bsd_args.s = linux_args.s;
740	/* XXX: */
741	bsd_args.name = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
742	bsd_args.anamelen = PTRIN(linux_args.namelen);/* XXX */
743	error = accept(td, &bsd_args);
744	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
745	if (error)
746		return (error);
747	if (linux_args.addr) {
748		error = linux_sa_put(PTRIN(linux_args.addr));
749		if (error) {
750			c_args.fd = td->td_retval[0];
751			(void)close(td, &c_args);
752			return (error);
753		}
754	}
755
756	/*
757	 * linux appears not to copy flags from the parent socket to the
758	 * accepted one, so we must clear the flags in the new descriptor.
759	 * Ignore any errors, because we already have an open fd.
760	 */
761	fd = td->td_retval[0];
762	(void)kern_fcntl(td, fd, F_SETFL, 0);
763	td->td_retval[0] = fd;
764	return (0);
765}
766
767struct l_getsockname_args {
768	int s;
769	l_uintptr_t addr;
770	l_uintptr_t namelen;
771};
772
773#ifndef __alpha__
774#define linux_getsockname_args l_getsockname_args
775#endif
776
777STATIC int
778linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
779{
780	struct l_getsockname_args linux_args;
781	struct getsockname_args /* {
782		int	fdes;
783		struct sockaddr * __restrict asa;
784		socklen_t * __restrict alen;
785	} */ bsd_args;
786	int error;
787
788	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
789		return (error);
790
791	bsd_args.fdes = linux_args.s;
792	/* XXX: */
793	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
794	bsd_args.alen = PTRIN(linux_args.namelen);	/* XXX */
795	error = getsockname(td, &bsd_args);
796	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
797	if (error)
798		return (error);
799	error = linux_sa_put(PTRIN(linux_args.addr));
800	if (error)
801		return (error);
802	return (0);
803}
804
805struct l_getpeername_args {
806	int s;
807	l_uintptr_t addr;
808	l_uintptr_t namelen;
809};
810
811#ifndef __alpha__
812#define linux_getpeername_args l_getpeername_args
813#endif
814
815STATIC int
816linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
817{
818	struct l_getpeername_args linux_args;
819	struct getpeername_args /* {
820		int fdes;
821		caddr_t asa;
822		int *alen;
823	} */ bsd_args;
824	int error;
825
826	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
827		return (error);
828
829	bsd_args.fdes = linux_args.s;
830	bsd_args.asa = (struct sockaddr *)PTRIN(linux_args.addr);
831	bsd_args.alen = (int *)PTRIN(linux_args.namelen);
832	error = getpeername(td, &bsd_args);
833	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
834	if (error)
835		return (error);
836	error = linux_sa_put(PTRIN(linux_args.addr));
837	if (error)
838		return (error);
839	return (0);
840}
841
842struct l_socketpair_args {
843	int domain;
844	int type;
845	int protocol;
846	l_uintptr_t rsv;
847};
848
849#ifndef __alpha__
850#define linux_socketpair_args l_socketpair_args
851#endif
852
853STATIC int
854linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
855{
856	struct l_socketpair_args linux_args;
857	struct socketpair_args /* {
858		int domain;
859		int type;
860		int protocol;
861		int *rsv;
862	} */ bsd_args;
863	int error;
864
865	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
866		return (error);
867
868	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
869	if (bsd_args.domain == -1)
870		return (EINVAL);
871
872	bsd_args.type = linux_args.type;
873	bsd_args.protocol = linux_args.protocol;
874	bsd_args.rsv = (int *)PTRIN(linux_args.rsv);
875	return (socketpair(td, &bsd_args));
876}
877
878struct l_send_args {
879	int s;
880	l_uintptr_t msg;
881	int len;
882	int flags;
883};
884
885#ifndef __alpha__
886#define linux_send_args l_send_args
887#endif
888
889STATIC int
890linux_send(struct thread *td, struct linux_send_args *args)
891{
892	struct l_send_args linux_args;
893	struct sendto_args /* {
894		int s;
895		caddr_t buf;
896		int len;
897		int flags;
898		caddr_t to;
899		int tolen;
900	} */ bsd_args;
901	int error;
902
903	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
904		return (error);
905
906	bsd_args.s = linux_args.s;
907	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
908	bsd_args.len = linux_args.len;
909	bsd_args.flags = linux_args.flags;
910	bsd_args.to = NULL;
911	bsd_args.tolen = 0;
912	return sendto(td, &bsd_args);
913}
914
915struct l_recv_args {
916	int s;
917	l_uintptr_t msg;
918	int len;
919	int flags;
920};
921
922#ifndef __alpha__
923#define linux_recv_args l_recv_args
924#endif
925
926STATIC int
927linux_recv(struct thread *td, struct linux_recv_args *args)
928{
929	struct l_recv_args linux_args;
930	struct recvfrom_args /* {
931		int s;
932		caddr_t buf;
933		int len;
934		int flags;
935		struct sockaddr *from;
936		socklen_t fromlenaddr;
937	} */ bsd_args;
938	int error;
939
940	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
941		return (error);
942
943	bsd_args.s = linux_args.s;
944	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
945	bsd_args.len = linux_args.len;
946	bsd_args.flags = linux_args.flags;
947	bsd_args.from = NULL;
948	bsd_args.fromlenaddr = 0;
949	return (recvfrom(td, &bsd_args));
950}
951
952#ifndef __alpha__
953static int
954linux_sendto(struct thread *td, struct linux_sendto_args *args)
955{
956	struct linux_sendto_args linux_args;
957	struct msghdr msg;
958	struct iovec aiov;
959	int error;
960
961	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
962		return (error);
963
964	if (linux_check_hdrincl(td, linux_args.s) == 0)
965		/* IP_HDRINCL set, tweak the packet before sending */
966		return (linux_sendto_hdrincl(td, &linux_args));
967
968	msg.msg_name = PTRIN(linux_args.to);
969	msg.msg_namelen = linux_args.tolen;
970	msg.msg_iov = &aiov;
971	msg.msg_iovlen = 1;
972	msg.msg_control = NULL;
973	msg.msg_flags = 0;
974	aiov.iov_base = PTRIN(linux_args.msg);
975	aiov.iov_len = linux_args.len;
976	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
977	    UIO_USERSPACE);
978	return (error);
979}
980#endif /* !__alpha__*/
981
982struct l_recvfrom_args {
983	int s;
984	l_uintptr_t buf;
985	int len;
986	int flags;
987	l_uintptr_t from;
988	l_uintptr_t fromlen;
989};
990
991#ifndef __alpha__
992#define linux_recvfrom_args l_recvfrom_args
993#endif
994
995STATIC int
996linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
997{
998	struct l_recvfrom_args linux_args;
999	struct recvfrom_args /* {
1000		int	s;
1001		caddr_t	buf;
1002		size_t	len;
1003		int	flags;
1004		struct sockaddr * __restrict from;
1005		socklen_t * __restrict fromlenaddr;
1006	} */ bsd_args;
1007	size_t len;
1008	int error;
1009
1010	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1011		return (error);
1012
1013	if ((error = copyin(PTRIN(linux_args.fromlen), &len, sizeof(size_t))))
1014		return (error);
1015
1016	bsd_args.s = linux_args.s;
1017	bsd_args.buf = PTRIN(linux_args.buf);
1018	bsd_args.len = linux_args.len;
1019	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
1020	/* XXX: */
1021	bsd_args.from = (struct sockaddr * __restrict)PTRIN(linux_args.from);
1022	bsd_args.fromlenaddr = PTRIN(linux_args.fromlen);/* XXX */
1023
1024	linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
1025	error = recvfrom(td, &bsd_args);
1026	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
1027
1028	if (error)
1029		return (error);
1030	if (linux_args.from) {
1031		error = linux_sa_put((struct osockaddr *)
1032		    PTRIN(linux_args.from));
1033		if (error)
1034			return (error);
1035	}
1036	return (0);
1037}
1038
1039struct l_sendmsg_args {
1040	int s;
1041	l_uintptr_t msg;
1042	int flags;
1043};
1044
1045#ifndef __alpha__
1046#define linux_sendmsg_args l_sendmsg_args
1047#endif
1048
1049STATIC int
1050linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1051{
1052	struct l_sendmsg_args linux_args;
1053	struct msghdr msg;
1054	struct iovec *iov;
1055	int error;
1056
1057	/* XXXTJR sendmsg is broken on amd64 */
1058
1059	error = copyin(args, &linux_args, sizeof(linux_args));
1060	if (error)
1061		return (error);
1062	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
1063	if (error)
1064		return (error);
1065	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1066	if (error)
1067		return (error);
1068	msg.msg_iov = iov;
1069	msg.msg_flags = 0;
1070	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
1071	    UIO_USERSPACE);
1072	free(iov, M_IOV);
1073	return (error);
1074}
1075
1076struct l_recvmsg_args {
1077	int s;
1078	l_uintptr_t msg;
1079	int flags;
1080};
1081
1082#ifndef __alpha__
1083#define linux_recvmsg_args l_recvmsg_args
1084#endif
1085
1086STATIC int
1087linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1088{
1089	struct l_recvmsg_args linux_args;
1090	struct recvmsg_args /* {
1091		int	s;
1092		struct	msghdr *msg;
1093		int	flags;
1094	} */ bsd_args;
1095	struct msghdr msg;
1096	struct cmsghdr *cmsg;
1097	int error;
1098
1099	/* XXXTJR recvmsg is broken on amd64 */
1100
1101	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1102		return (error);
1103
1104	bsd_args.s = linux_args.s;
1105	bsd_args.msg = PTRIN(linux_args.msg);
1106	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
1107	if (msg.msg_name) {
1108	   	linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1109		      msg.msg_namelen);
1110		error = recvmsg(td, &bsd_args);
1111		bsd_to_linux_sockaddr((struct sockaddr *)msg.msg_name);
1112	} else
1113	   	error = recvmsg(td, &bsd_args);
1114	if (error)
1115		return (error);
1116
1117	if (bsd_args.msg->msg_control != NULL &&
1118	    bsd_args.msg->msg_controllen > 0) {
1119		cmsg = (struct cmsghdr*)bsd_args.msg->msg_control;
1120		cmsg->cmsg_level = bsd_to_linux_sockopt_level(cmsg->cmsg_level);
1121	}
1122
1123	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
1124	if (error)
1125		return (error);
1126	if (msg.msg_name && msg.msg_namelen > 2)
1127		error = linux_sa_put(msg.msg_name);
1128	return (error);
1129}
1130
1131#ifndef __alpha__
1132struct linux_shutdown_args {
1133	int s;
1134	int how;
1135};
1136
1137static int
1138linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1139{
1140	struct linux_shutdown_args linux_args;
1141	struct shutdown_args /* {
1142		int s;
1143		int how;
1144	} */ bsd_args;
1145	int error;
1146
1147	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1148		return (error);
1149
1150	bsd_args.s = linux_args.s;
1151	bsd_args.how = linux_args.how;
1152	return (shutdown(td, &bsd_args));
1153}
1154
1155struct linux_setsockopt_args {
1156	int s;
1157	int level;
1158	int optname;
1159	l_uintptr_t optval;
1160	int optlen;
1161};
1162
1163static int
1164linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1165{
1166	struct linux_setsockopt_args linux_args;
1167	struct setsockopt_args /* {
1168		int s;
1169		int level;
1170		int name;
1171		caddr_t val;
1172		int valsize;
1173	} */ bsd_args;
1174	int error, name;
1175
1176	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1177		return (error);
1178
1179	bsd_args.s = linux_args.s;
1180	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1181	switch (bsd_args.level) {
1182	case SOL_SOCKET:
1183		name = linux_to_bsd_so_sockopt(linux_args.optname);
1184		break;
1185	case IPPROTO_IP:
1186		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1187		break;
1188	case IPPROTO_TCP:
1189		/* Linux TCP option values match BSD's */
1190		name = linux_args.optname;
1191		break;
1192	default:
1193		name = -1;
1194		break;
1195	}
1196	if (name == -1)
1197		return (EINVAL);
1198
1199	bsd_args.name = name;
1200	bsd_args.val = PTRIN(linux_args.optval);
1201	bsd_args.valsize = linux_args.optlen;
1202
1203	if (name == IPV6_NEXTHOP) {
1204		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1205			bsd_args.valsize);
1206		error = setsockopt(td, &bsd_args);
1207		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1208	} else
1209		error = setsockopt(td, &bsd_args);
1210
1211	return (error);
1212}
1213
1214struct linux_getsockopt_args {
1215	int s;
1216	int level;
1217	int optname;
1218	l_uintptr_t optval;
1219	l_uintptr_t optlen;
1220};
1221
1222static int
1223linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1224{
1225	struct linux_getsockopt_args linux_args;
1226	struct getsockopt_args /* {
1227		int s;
1228		int level;
1229		int name;
1230		caddr_t val;
1231		int *avalsize;
1232	} */ bsd_args;
1233	int error, name;
1234
1235	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1236		return (error);
1237
1238	bsd_args.s = linux_args.s;
1239	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1240	switch (bsd_args.level) {
1241	case SOL_SOCKET:
1242		name = linux_to_bsd_so_sockopt(linux_args.optname);
1243		break;
1244	case IPPROTO_IP:
1245		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1246		break;
1247	case IPPROTO_TCP:
1248		/* Linux TCP option values match BSD's */
1249		name = linux_args.optname;
1250		break;
1251	default:
1252		name = -1;
1253		break;
1254	}
1255	if (name == -1)
1256		return (EINVAL);
1257
1258	bsd_args.name = name;
1259	bsd_args.val = PTRIN(linux_args.optval);
1260	bsd_args.avalsize = PTRIN(linux_args.optlen);
1261
1262	if (name == IPV6_NEXTHOP) {
1263		error = getsockopt(td, &bsd_args);
1264		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1265	} else
1266		error = getsockopt(td, &bsd_args);
1267
1268	return (error);
1269}
1270
1271int
1272linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1273{
1274	void *arg = (void *)(intptr_t)args->args;
1275
1276	switch (args->what) {
1277	case LINUX_SOCKET:
1278		return (linux_socket(td, arg));
1279	case LINUX_BIND:
1280		return (linux_bind(td, arg));
1281	case LINUX_CONNECT:
1282		return (linux_connect(td, arg));
1283	case LINUX_LISTEN:
1284		return (linux_listen(td, arg));
1285	case LINUX_ACCEPT:
1286		return (linux_accept(td, arg));
1287	case LINUX_GETSOCKNAME:
1288		return (linux_getsockname(td, arg));
1289	case LINUX_GETPEERNAME:
1290		return (linux_getpeername(td, arg));
1291	case LINUX_SOCKETPAIR:
1292		return (linux_socketpair(td, arg));
1293	case LINUX_SEND:
1294		return (linux_send(td, arg));
1295	case LINUX_RECV:
1296		return (linux_recv(td, arg));
1297	case LINUX_SENDTO:
1298		return (linux_sendto(td, arg));
1299	case LINUX_RECVFROM:
1300		return (linux_recvfrom(td, arg));
1301	case LINUX_SHUTDOWN:
1302		return (linux_shutdown(td, arg));
1303	case LINUX_SETSOCKOPT:
1304		return (linux_setsockopt(td, arg));
1305	case LINUX_GETSOCKOPT:
1306		return (linux_getsockopt(td, arg));
1307	case LINUX_SENDMSG:
1308		return (linux_sendmsg(td, arg));
1309	case LINUX_RECVMSG:
1310		return (linux_recvmsg(td, arg));
1311	}
1312
1313	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1314	return (ENOSYS);
1315}
1316#endif /*!__alpha__*/
1317