linux_socket.c revision 162585
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 162585 2006-09-23 19:06:54Z netchild $");
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/*
76 * Reads a linux sockaddr and does any necessary translation.
77 * Linux sockaddrs don't have a length field, only a family.
78 */
79static int
80linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
81{
82	int osalen = len;
83
84	return (do_sa_get(sap, osa, &osalen, M_SONAME));
85}
86
87/*
88 * Copy the osockaddr structure pointed to by osa to kernel, adjust
89 * family and convert to sockaddr.
90 */
91static int
92do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
93    struct malloc_type *mtype)
94{
95	int error=0, bdom;
96	struct sockaddr *sa;
97	struct osockaddr *kosa;
98	int alloclen;
99#ifdef INET6
100	int oldv6size;
101	struct sockaddr_in6 *sin6;
102#endif
103
104	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
105		return (EINVAL);
106
107	alloclen = *osalen;
108#ifdef INET6
109	oldv6size = 0;
110	/*
111	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
112	 * if it's a v4-mapped address, so reserve the proper space
113	 * for it.
114	 */
115	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
116		alloclen = sizeof (struct sockaddr_in6);
117		oldv6size = 1;
118	}
119#endif
120
121	MALLOC(kosa, struct osockaddr *, alloclen, mtype, M_WAITOK);
122
123	if ((error = copyin(osa, kosa, *osalen)))
124		goto out;
125
126	bdom = linux_to_bsd_domain(kosa->sa_family);
127	if (bdom == -1) {
128		error = EINVAL;
129		goto out;
130	}
131
132#ifdef INET6
133	/*
134	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
135	 * which lacks the scope id compared with RFC2553 one. If we detect
136	 * the situation, reject the address and write a message to system log.
137	 *
138	 * Still accept addresses for which the scope id is not used.
139	 */
140	if (oldv6size && bdom == AF_INET6) {
141		sin6 = (struct sockaddr_in6 *)kosa;
142		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
143		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
144		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
145		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
146		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
147		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
148			sin6->sin6_scope_id = 0;
149		} else {
150			log(LOG_DEBUG,
151			    "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
152			error = EINVAL;
153			goto out;
154		}
155	} else
156#endif
157	if (bdom == AF_INET)
158		alloclen = sizeof(struct sockaddr_in);
159
160	sa = (struct sockaddr *) kosa;
161	sa->sa_family = bdom;
162	sa->sa_len = alloclen;
163
164	*sap = sa;
165	*osalen = alloclen;
166	return (0);
167
168out:
169	FREE(kosa, mtype);
170	return (error);
171}
172
173static int
174linux_to_bsd_domain(int domain)
175{
176
177	switch (domain) {
178	case LINUX_AF_UNSPEC:
179		return (AF_UNSPEC);
180	case LINUX_AF_UNIX:
181		return (AF_LOCAL);
182	case LINUX_AF_INET:
183		return (AF_INET);
184	case LINUX_AF_INET6:
185		return (AF_INET6);
186	case LINUX_AF_AX25:
187		return (AF_CCITT);
188	case LINUX_AF_IPX:
189		return (AF_IPX);
190	case LINUX_AF_APPLETALK:
191		return (AF_APPLETALK);
192	}
193	return (-1);
194}
195
196static int
197bsd_to_linux_domain(int domain)
198{
199
200	switch (domain) {
201	case AF_UNSPEC:
202		return (LINUX_AF_UNSPEC);
203	case AF_LOCAL:
204		return (LINUX_AF_UNIX);
205	case AF_INET:
206		return (LINUX_AF_INET);
207	case AF_INET6:
208		return (LINUX_AF_INET6);
209	case AF_CCITT:
210		return (LINUX_AF_AX25);
211	case AF_IPX:
212		return (LINUX_AF_IPX);
213	case AF_APPLETALK:
214		return (LINUX_AF_APPLETALK);
215	}
216	return (-1);
217}
218
219static int
220linux_to_bsd_sockopt_level(int level)
221{
222
223	switch (level) {
224	case LINUX_SOL_SOCKET:
225		return (SOL_SOCKET);
226	}
227	return (level);
228}
229
230static int
231bsd_to_linux_sockopt_level(int level)
232{
233
234	switch (level) {
235	case SOL_SOCKET:
236		return (LINUX_SOL_SOCKET);
237	}
238	return (level);
239}
240
241static int
242linux_to_bsd_ip_sockopt(int opt)
243{
244
245	switch (opt) {
246	case LINUX_IP_TOS:
247		return (IP_TOS);
248	case LINUX_IP_TTL:
249		return (IP_TTL);
250	case LINUX_IP_OPTIONS:
251		return (IP_OPTIONS);
252	case LINUX_IP_MULTICAST_IF:
253		return (IP_MULTICAST_IF);
254	case LINUX_IP_MULTICAST_TTL:
255		return (IP_MULTICAST_TTL);
256	case LINUX_IP_MULTICAST_LOOP:
257		return (IP_MULTICAST_LOOP);
258	case LINUX_IP_ADD_MEMBERSHIP:
259		return (IP_ADD_MEMBERSHIP);
260	case LINUX_IP_DROP_MEMBERSHIP:
261		return (IP_DROP_MEMBERSHIP);
262	case LINUX_IP_HDRINCL:
263		return (IP_HDRINCL);
264	}
265	return (-1);
266}
267
268static int
269linux_to_bsd_so_sockopt(int opt)
270{
271
272	switch (opt) {
273	case LINUX_SO_DEBUG:
274		return (SO_DEBUG);
275	case LINUX_SO_REUSEADDR:
276		return (SO_REUSEADDR);
277	case LINUX_SO_TYPE:
278		return (SO_TYPE);
279	case LINUX_SO_ERROR:
280		return (SO_ERROR);
281	case LINUX_SO_DONTROUTE:
282		return (SO_DONTROUTE);
283	case LINUX_SO_BROADCAST:
284		return (SO_BROADCAST);
285	case LINUX_SO_SNDBUF:
286		return (SO_SNDBUF);
287	case LINUX_SO_RCVBUF:
288		return (SO_RCVBUF);
289	case LINUX_SO_KEEPALIVE:
290		return (SO_KEEPALIVE);
291	case LINUX_SO_OOBINLINE:
292		return (SO_OOBINLINE);
293	case LINUX_SO_LINGER:
294		return (SO_LINGER);
295	}
296	return (-1);
297}
298
299static int
300linux_to_bsd_msg_flags(int flags)
301{
302	int ret_flags = 0;
303
304	if (flags & LINUX_MSG_OOB)
305		ret_flags |= MSG_OOB;
306	if (flags & LINUX_MSG_PEEK)
307		ret_flags |= MSG_PEEK;
308	if (flags & LINUX_MSG_DONTROUTE)
309		ret_flags |= MSG_DONTROUTE;
310	if (flags & LINUX_MSG_CTRUNC)
311		ret_flags |= MSG_CTRUNC;
312	if (flags & LINUX_MSG_TRUNC)
313		ret_flags |= MSG_TRUNC;
314	if (flags & LINUX_MSG_DONTWAIT)
315		ret_flags |= MSG_DONTWAIT;
316	if (flags & LINUX_MSG_EOR)
317		ret_flags |= MSG_EOR;
318	if (flags & LINUX_MSG_WAITALL)
319		ret_flags |= MSG_WAITALL;
320	if (flags & LINUX_MSG_NOSIGNAL)
321		ret_flags |= MSG_NOSIGNAL;
322#if 0 /* not handled */
323	if (flags & LINUX_MSG_PROXY)
324		;
325	if (flags & LINUX_MSG_FIN)
326		;
327	if (flags & LINUX_MSG_SYN)
328		;
329	if (flags & LINUX_MSG_CONFIRM)
330		;
331	if (flags & LINUX_MSG_RST)
332		;
333	if (flags & LINUX_MSG_ERRQUEUE)
334		;
335#endif
336	return ret_flags;
337}
338
339/*
340* If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
341* native syscall will fault.  Thus, we don't really need to check the
342* return values for these functions.
343*/
344
345static int
346bsd_to_linux_sockaddr(struct sockaddr *arg)
347{
348	struct sockaddr sa;
349	size_t sa_len = sizeof(struct sockaddr);
350	int error;
351
352	if ((error = copyin(arg, &sa, sa_len)))
353		return (error);
354
355	*(u_short *)&sa = sa.sa_family;
356
357	error = copyout(&sa, arg, sa_len);
358
359	return (error);
360}
361
362static int
363linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
364{
365	struct sockaddr sa;
366	size_t sa_len = sizeof(struct sockaddr);
367	int error;
368
369	if ((error = copyin(arg, &sa, sa_len)))
370		return (error);
371
372	sa.sa_family = *(sa_family_t *)&sa;
373	sa.sa_len = len;
374
375	error = copyout(&sa, arg, sa_len);
376
377	return (error);
378}
379
380
381static int
382linux_sa_put(struct osockaddr *osa)
383{
384	struct osockaddr sa;
385	int error, bdom;
386
387	/*
388	 * Only read/write the osockaddr family part, the rest is
389	 * not changed.
390	 */
391	error = copyin(osa, &sa, sizeof(sa.sa_family));
392	if (error)
393		return (error);
394
395	bdom = bsd_to_linux_domain(sa.sa_family);
396	if (bdom == -1)
397		return (EINVAL);
398
399	sa.sa_family = bdom;
400	error = copyout(&sa, osa, sizeof(sa.sa_family));
401	if (error)
402		return (error);
403
404	return (0);
405}
406
407static int
408linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
409    enum uio_seg segflg)
410{
411	struct mbuf *control;
412	struct sockaddr *to;
413	int error;
414
415	if (mp->msg_name != NULL) {
416		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
417		if (error)
418			return (error);
419		mp->msg_name = to;
420	} else
421		to = NULL;
422
423	if (mp->msg_control != NULL) {
424		struct cmsghdr *cmsg;
425
426		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
427			error = EINVAL;
428			goto bad;
429		}
430		error = sockargs(&control, mp->msg_control,
431		    mp->msg_controllen, MT_CONTROL);
432		if (error)
433			goto bad;
434
435		cmsg = mtod(control, struct cmsghdr *);
436		cmsg->cmsg_level = linux_to_bsd_sockopt_level(cmsg->cmsg_level);
437	} else
438		control = NULL;
439
440	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
441	    segflg);
442
443bad:
444	if (to)
445		FREE(to, M_SONAME);
446	return (error);
447}
448
449/* Return 0 if IP_HDRINCL is set for the given socket. */
450static int
451linux_check_hdrincl(struct thread *td, int s)
452{
453	int error, optval, size_val;
454
455	size_val = sizeof(optval);
456	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
457	    &optval, UIO_SYSSPACE, &size_val);
458	if (error)
459		return (error);
460
461	return (optval == 0);
462}
463
464struct linux_sendto_args {
465	int s;
466	l_uintptr_t msg;
467	int len;
468	int flags;
469	l_uintptr_t to;
470	int tolen;
471};
472
473/*
474 * Updated sendto() when IP_HDRINCL is set:
475 * tweak endian-dependent fields in the IP packet.
476 */
477static int
478linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
479{
480/*
481 * linux_ip_copysize defines how many bytes we should copy
482 * from the beginning of the IP packet before we customize it for BSD.
483 * It should include all the fields we modify (ip_len and ip_off).
484 */
485#define linux_ip_copysize	8
486
487	struct ip *packet;
488	struct msghdr msg;
489	struct iovec aiov[1];
490	int error;
491
492	/* Check that the packet isn't too big or too small. */
493	if (linux_args->len < linux_ip_copysize ||
494	    linux_args->len > IP_MAXPACKET)
495		return (EINVAL);
496
497	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
498
499	/* Make kernel copy of the packet to be sent */
500	if ((error = copyin(PTRIN(linux_args->msg), packet,
501	    linux_args->len)))
502		goto goout;
503
504	/* Convert fields from Linux to BSD raw IP socket format */
505	packet->ip_len = linux_args->len;
506	packet->ip_off = ntohs(packet->ip_off);
507
508	/* Prepare the msghdr and iovec structures describing the new packet */
509	msg.msg_name = PTRIN(linux_args->to);
510	msg.msg_namelen = linux_args->tolen;
511	msg.msg_iov = aiov;
512	msg.msg_iovlen = 1;
513	msg.msg_control = NULL;
514	msg.msg_flags = 0;
515	aiov[0].iov_base = (char *)packet;
516	aiov[0].iov_len = linux_args->len;
517	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
518	    UIO_SYSSPACE);
519goout:
520	free(packet, M_TEMP);
521	return (error);
522}
523
524struct linux_socket_args {
525	int domain;
526	int type;
527	int protocol;
528};
529
530static int
531linux_socket(struct thread *td, struct linux_socket_args *args)
532{
533	struct linux_socket_args linux_args;
534	struct socket_args /* {
535		int domain;
536		int type;
537		int protocol;
538	} */ bsd_args;
539	int error;
540	int retval_socket;
541
542	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
543		return (error);
544
545	bsd_args.protocol = linux_args.protocol;
546	bsd_args.type = linux_args.type;
547	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
548	if (bsd_args.domain == -1)
549		return (EINVAL);
550
551	retval_socket = socket(td, &bsd_args);
552	if (bsd_args.type == SOCK_RAW
553	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
554	    && bsd_args.domain == AF_INET
555	    && retval_socket >= 0) {
556		/* It's a raw IP socket: set the IP_HDRINCL option. */
557		int hdrincl;
558
559		hdrincl = 1;
560		/* We ignore any error returned by kern_setsockopt() */
561		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
562		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
563	}
564#ifdef INET6
565	/*
566	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
567	 * default and some apps depend on this. So, set V6ONLY to 0
568	 * for Linux apps if the sysctl value is set to 1.
569	 */
570	if (bsd_args.domain == PF_INET6 && retval_socket >= 0
571#ifndef KLD_MODULE
572	    /*
573	     * XXX: Avoid undefined symbol error with an IPv4 only
574	     * kernel.
575	     */
576	    && ip6_v6only
577#endif
578	    ) {
579		int v6only;
580
581		v6only = 0;
582		/* We ignore any error returned by setsockopt() */
583		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
584		    &v6only, UIO_SYSSPACE, sizeof(v6only));
585	}
586#endif
587
588	return (retval_socket);
589}
590
591struct linux_bind_args {
592	int s;
593	l_uintptr_t name;
594	int namelen;
595};
596
597static int
598linux_bind(struct thread *td, struct linux_bind_args *args)
599{
600	struct linux_bind_args linux_args;
601	struct sockaddr *sa;
602	int error;
603
604	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
605		return (error);
606
607	error = linux_getsockaddr(&sa, PTRIN(linux_args.name),
608	    linux_args.namelen);
609	if (error)
610		return (error);
611
612	error = kern_bind(td, linux_args.s, sa);
613	free(sa, M_SONAME);
614	if (error == EADDRNOTAVAIL && linux_args.namelen != sizeof(struct sockaddr_in))
615	   	return (EINVAL);
616	return (error);
617}
618
619struct linux_connect_args {
620	int s;
621	l_uintptr_t name;
622	int namelen;
623};
624int linux_connect(struct thread *, struct linux_connect_args *);
625
626int
627linux_connect(struct thread *td, struct linux_connect_args *args)
628{
629	struct linux_connect_args linux_args;
630	struct socket *so;
631	struct sockaddr *sa;
632	u_int fflag;
633	int error;
634
635	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
636		return (error);
637
638	error = linux_getsockaddr(&sa,
639	    (struct osockaddr *)PTRIN(linux_args.name),
640	    linux_args.namelen);
641	if (error)
642		return (error);
643
644	error = kern_connect(td, linux_args.s, sa);
645	free(sa, M_SONAME);
646	if (error != EISCONN)
647		return (error);
648
649	/*
650	 * Linux doesn't return EISCONN the first time it occurs,
651	 * when on a non-blocking socket. Instead it returns the
652	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
653	 *
654	 * XXXRW: Instead of using fgetsock(), check that it is a
655	 * socket and use the file descriptor reference instead of
656	 * creating a new one.
657	 */
658	NET_LOCK_GIANT();
659	error = fgetsock(td, linux_args.s, &so, &fflag);
660	if (error == 0) {
661		error = EISCONN;
662		if (fflag & FNONBLOCK) {
663			SOCK_LOCK(so);
664			if (so->so_emuldata == 0)
665				error = so->so_error;
666			so->so_emuldata = (void *)1;
667			SOCK_UNLOCK(so);
668		}
669		fputsock(so);
670	}
671	NET_UNLOCK_GIANT();
672	return (error);
673}
674
675struct linux_listen_args {
676	int s;
677	int backlog;
678};
679
680static int
681linux_listen(struct thread *td, struct linux_listen_args *args)
682{
683	struct linux_listen_args linux_args;
684	struct listen_args /* {
685		int s;
686		int backlog;
687	} */ bsd_args;
688	int error;
689
690	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
691		return (error);
692
693	bsd_args.s = linux_args.s;
694	bsd_args.backlog = linux_args.backlog;
695	return (listen(td, &bsd_args));
696}
697
698struct linux_accept_args {
699	int s;
700	l_uintptr_t addr;
701	l_uintptr_t namelen;
702};
703
704static int
705linux_accept(struct thread *td, struct linux_accept_args *args)
706{
707	struct linux_accept_args linux_args;
708	struct accept_args /* {
709		int	s;
710		struct sockaddr * __restrict name;
711		socklen_t * __restrict anamelen;
712	} */ bsd_args;
713	int error, fd;
714
715	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
716		return (error);
717
718	bsd_args.s = linux_args.s;
719	/* XXX: */
720	bsd_args.name = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
721	bsd_args.anamelen = PTRIN(linux_args.namelen);/* XXX */
722	error = accept(td, &bsd_args);
723	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
724	if (error) {
725	   	if (error == EFAULT && linux_args.namelen != sizeof(struct sockaddr_in))
726		   	return (EINVAL);
727		return (error);
728	}
729	if (linux_args.addr) {
730		error = linux_sa_put(PTRIN(linux_args.addr));
731		if (error) {
732			(void)kern_close(td, td->td_retval[0]);
733			return (error);
734		}
735	}
736
737	/*
738	 * linux appears not to copy flags from the parent socket to the
739	 * accepted one, so we must clear the flags in the new descriptor.
740	 * Ignore any errors, because we already have an open fd.
741	 */
742	fd = td->td_retval[0];
743	(void)kern_fcntl(td, fd, F_SETFL, 0);
744	td->td_retval[0] = fd;
745	return (0);
746}
747
748struct linux_getsockname_args {
749	int s;
750	l_uintptr_t addr;
751	l_uintptr_t namelen;
752};
753
754static int
755linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
756{
757	struct linux_getsockname_args linux_args;
758	struct getsockname_args /* {
759		int	fdes;
760		struct sockaddr * __restrict asa;
761		socklen_t * __restrict alen;
762	} */ bsd_args;
763	int error;
764
765	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
766		return (error);
767
768	bsd_args.fdes = linux_args.s;
769	/* XXX: */
770	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
771	bsd_args.alen = PTRIN(linux_args.namelen);	/* XXX */
772	error = getsockname(td, &bsd_args);
773	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
774	if (error)
775		return (error);
776	error = linux_sa_put(PTRIN(linux_args.addr));
777	if (error)
778		return (error);
779	return (0);
780}
781
782struct linux_getpeername_args {
783	int s;
784	l_uintptr_t addr;
785	l_uintptr_t namelen;
786};
787
788static int
789linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
790{
791	struct linux_getpeername_args linux_args;
792	struct getpeername_args /* {
793		int fdes;
794		caddr_t asa;
795		int *alen;
796	} */ bsd_args;
797	int error;
798
799	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
800		return (error);
801
802	bsd_args.fdes = linux_args.s;
803	bsd_args.asa = (struct sockaddr *)PTRIN(linux_args.addr);
804	bsd_args.alen = (int *)PTRIN(linux_args.namelen);
805	error = getpeername(td, &bsd_args);
806	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
807	if (error)
808		return (error);
809	error = linux_sa_put(PTRIN(linux_args.addr));
810	if (error)
811		return (error);
812	return (0);
813}
814
815struct linux_socketpair_args {
816	int domain;
817	int type;
818	int protocol;
819	l_uintptr_t rsv;
820};
821
822static int
823linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
824{
825	struct linux_socketpair_args linux_args;
826	struct socketpair_args /* {
827		int domain;
828		int type;
829		int protocol;
830		int *rsv;
831	} */ bsd_args;
832	int error;
833
834	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
835		return (error);
836
837	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
838	if (bsd_args.domain == -1)
839		return (EINVAL);
840
841	bsd_args.type = linux_args.type;
842	bsd_args.protocol = linux_args.protocol;
843	bsd_args.rsv = (int *)PTRIN(linux_args.rsv);
844	return (socketpair(td, &bsd_args));
845}
846
847struct linux_send_args {
848	int s;
849	l_uintptr_t msg;
850	int len;
851	int flags;
852};
853
854static int
855linux_send(struct thread *td, struct linux_send_args *args)
856{
857	struct linux_send_args linux_args;
858	struct sendto_args /* {
859		int s;
860		caddr_t buf;
861		int len;
862		int flags;
863		caddr_t to;
864		int tolen;
865	} */ bsd_args;
866	int error;
867
868	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
869		return (error);
870
871	bsd_args.s = linux_args.s;
872	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
873	bsd_args.len = linux_args.len;
874	bsd_args.flags = linux_args.flags;
875	bsd_args.to = NULL;
876	bsd_args.tolen = 0;
877	return sendto(td, &bsd_args);
878}
879
880struct linux_recv_args {
881	int s;
882	l_uintptr_t msg;
883	int len;
884	int flags;
885};
886
887static int
888linux_recv(struct thread *td, struct linux_recv_args *args)
889{
890	struct linux_recv_args linux_args;
891	struct recvfrom_args /* {
892		int s;
893		caddr_t buf;
894		int len;
895		int flags;
896		struct sockaddr *from;
897		socklen_t fromlenaddr;
898	} */ bsd_args;
899	int error;
900
901	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
902		return (error);
903
904	bsd_args.s = linux_args.s;
905	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
906	bsd_args.len = linux_args.len;
907	bsd_args.flags = linux_args.flags;
908	bsd_args.from = NULL;
909	bsd_args.fromlenaddr = 0;
910	return (recvfrom(td, &bsd_args));
911}
912
913static int
914linux_sendto(struct thread *td, struct linux_sendto_args *args)
915{
916	struct linux_sendto_args linux_args;
917	struct msghdr msg;
918	struct iovec aiov;
919	int error;
920
921	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
922		return (error);
923
924	if (linux_check_hdrincl(td, linux_args.s) == 0)
925		/* IP_HDRINCL set, tweak the packet before sending */
926		return (linux_sendto_hdrincl(td, &linux_args));
927
928	msg.msg_name = PTRIN(linux_args.to);
929	msg.msg_namelen = linux_args.tolen;
930	msg.msg_iov = &aiov;
931	msg.msg_iovlen = 1;
932	msg.msg_control = NULL;
933	msg.msg_flags = 0;
934	aiov.iov_base = PTRIN(linux_args.msg);
935	aiov.iov_len = linux_args.len;
936	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
937	    UIO_USERSPACE);
938	return (error);
939}
940
941struct linux_recvfrom_args {
942	int s;
943	l_uintptr_t buf;
944	int len;
945	int flags;
946	l_uintptr_t from;
947	l_uintptr_t fromlen;
948};
949
950static int
951linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
952{
953	struct linux_recvfrom_args linux_args;
954	struct recvfrom_args /* {
955		int	s;
956		caddr_t	buf;
957		size_t	len;
958		int	flags;
959		struct sockaddr * __restrict from;
960		socklen_t * __restrict fromlenaddr;
961	} */ bsd_args;
962	size_t len;
963	int error;
964
965	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
966		return (error);
967
968	if ((error = copyin(PTRIN(linux_args.fromlen), &len, sizeof(size_t))))
969		return (error);
970
971	bsd_args.s = linux_args.s;
972	bsd_args.buf = PTRIN(linux_args.buf);
973	bsd_args.len = linux_args.len;
974	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
975	/* XXX: */
976	bsd_args.from = (struct sockaddr * __restrict)PTRIN(linux_args.from);
977	bsd_args.fromlenaddr = PTRIN(linux_args.fromlen);/* XXX */
978
979	linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.from, len);
980	error = recvfrom(td, &bsd_args);
981	bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.from);
982
983	if (error)
984		return (error);
985	if (linux_args.from) {
986		error = linux_sa_put((struct osockaddr *)
987		    PTRIN(linux_args.from));
988		if (error)
989			return (error);
990	}
991	return (0);
992}
993
994struct linux_sendmsg_args {
995	int s;
996	l_uintptr_t msg;
997	int flags;
998};
999
1000static int
1001linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1002{
1003	struct linux_sendmsg_args linux_args;
1004	struct msghdr msg;
1005	struct iovec *iov;
1006	int error;
1007
1008	/* XXXTJR sendmsg is broken on amd64 */
1009
1010	error = copyin(args, &linux_args, sizeof(linux_args));
1011	if (error)
1012		return (error);
1013	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
1014	if (error)
1015		return (error);
1016	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1017	if (error)
1018		return (error);
1019	msg.msg_iov = iov;
1020	msg.msg_flags = 0;
1021	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
1022	    UIO_USERSPACE);
1023	free(iov, M_IOV);
1024	return (error);
1025}
1026
1027struct linux_recvmsg_args {
1028	int s;
1029	l_uintptr_t msg;
1030	int flags;
1031};
1032
1033static int
1034linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1035{
1036	struct linux_recvmsg_args linux_args;
1037	struct recvmsg_args /* {
1038		int	s;
1039		struct	msghdr *msg;
1040		int	flags;
1041	} */ bsd_args;
1042	struct msghdr msg;
1043	struct cmsghdr *cmsg;
1044	int error;
1045
1046	/* XXXTJR recvmsg is broken on amd64 */
1047
1048	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1049		return (error);
1050
1051	bsd_args.s = linux_args.s;
1052	bsd_args.msg = PTRIN(linux_args.msg);
1053	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
1054	if (msg.msg_name) {
1055	   	linux_to_bsd_sockaddr((struct sockaddr *)msg.msg_name,
1056		      msg.msg_namelen);
1057		error = recvmsg(td, &bsd_args);
1058		bsd_to_linux_sockaddr((struct sockaddr *)msg.msg_name);
1059	} else
1060	   	error = recvmsg(td, &bsd_args);
1061	if (error)
1062		return (error);
1063
1064	if (bsd_args.msg->msg_control != NULL &&
1065	    bsd_args.msg->msg_controllen > 0) {
1066		cmsg = (struct cmsghdr*)bsd_args.msg->msg_control;
1067		cmsg->cmsg_level = bsd_to_linux_sockopt_level(cmsg->cmsg_level);
1068	}
1069
1070	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
1071	if (error)
1072		return (error);
1073	if (msg.msg_name && msg.msg_namelen > 2)
1074		error = linux_sa_put(msg.msg_name);
1075	return (error);
1076}
1077
1078struct linux_shutdown_args {
1079	int s;
1080	int how;
1081};
1082
1083static int
1084linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1085{
1086	struct linux_shutdown_args linux_args;
1087	struct shutdown_args /* {
1088		int s;
1089		int how;
1090	} */ bsd_args;
1091	int error;
1092
1093	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1094		return (error);
1095
1096	bsd_args.s = linux_args.s;
1097	bsd_args.how = linux_args.how;
1098	return (shutdown(td, &bsd_args));
1099}
1100
1101struct linux_setsockopt_args {
1102	int s;
1103	int level;
1104	int optname;
1105	l_uintptr_t optval;
1106	int optlen;
1107};
1108
1109static int
1110linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1111{
1112	struct linux_setsockopt_args linux_args;
1113	struct setsockopt_args /* {
1114		int s;
1115		int level;
1116		int name;
1117		caddr_t val;
1118		int valsize;
1119	} */ bsd_args;
1120	int error, name;
1121
1122	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1123		return (error);
1124
1125	bsd_args.s = linux_args.s;
1126	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1127	switch (bsd_args.level) {
1128	case SOL_SOCKET:
1129		name = linux_to_bsd_so_sockopt(linux_args.optname);
1130		break;
1131	case IPPROTO_IP:
1132		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1133		break;
1134	case IPPROTO_TCP:
1135		/* Linux TCP option values match BSD's */
1136		name = linux_args.optname;
1137		break;
1138	default:
1139		name = -1;
1140		break;
1141	}
1142	if (name == -1)
1143		return (ENOPROTOOPT);
1144
1145	bsd_args.name = name;
1146	bsd_args.val = PTRIN(linux_args.optval);
1147	bsd_args.valsize = linux_args.optlen;
1148
1149	if (name == IPV6_NEXTHOP) {
1150		linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1151			bsd_args.valsize);
1152		error = setsockopt(td, &bsd_args);
1153		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1154	} else
1155		error = setsockopt(td, &bsd_args);
1156
1157	return (error);
1158}
1159
1160struct linux_getsockopt_args {
1161	int s;
1162	int level;
1163	int optname;
1164	l_uintptr_t optval;
1165	l_uintptr_t optlen;
1166};
1167
1168static int
1169linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1170{
1171	struct linux_getsockopt_args linux_args;
1172	struct getsockopt_args /* {
1173		int s;
1174		int level;
1175		int name;
1176		caddr_t val;
1177		int *avalsize;
1178	} */ bsd_args;
1179	int error, name;
1180
1181	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1182		return (error);
1183
1184	bsd_args.s = linux_args.s;
1185	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1186	switch (bsd_args.level) {
1187	case SOL_SOCKET:
1188		name = linux_to_bsd_so_sockopt(linux_args.optname);
1189		break;
1190	case IPPROTO_IP:
1191		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1192		break;
1193	case IPPROTO_TCP:
1194		/* Linux TCP option values match BSD's */
1195		name = linux_args.optname;
1196		break;
1197	default:
1198		name = -1;
1199		break;
1200	}
1201	if (name == -1)
1202		return (EINVAL);
1203
1204	bsd_args.name = name;
1205	bsd_args.val = PTRIN(linux_args.optval);
1206	bsd_args.avalsize = PTRIN(linux_args.optlen);
1207
1208	if (name == IPV6_NEXTHOP) {
1209		error = getsockopt(td, &bsd_args);
1210		bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1211	} else
1212		error = getsockopt(td, &bsd_args);
1213
1214	return (error);
1215}
1216
1217int
1218linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1219{
1220	void *arg = (void *)(intptr_t)args->args;
1221
1222	switch (args->what) {
1223	case LINUX_SOCKET:
1224		return (linux_socket(td, arg));
1225	case LINUX_BIND:
1226		return (linux_bind(td, arg));
1227	case LINUX_CONNECT:
1228		return (linux_connect(td, arg));
1229	case LINUX_LISTEN:
1230		return (linux_listen(td, arg));
1231	case LINUX_ACCEPT:
1232		return (linux_accept(td, arg));
1233	case LINUX_GETSOCKNAME:
1234		return (linux_getsockname(td, arg));
1235	case LINUX_GETPEERNAME:
1236		return (linux_getpeername(td, arg));
1237	case LINUX_SOCKETPAIR:
1238		return (linux_socketpair(td, arg));
1239	case LINUX_SEND:
1240		return (linux_send(td, arg));
1241	case LINUX_RECV:
1242		return (linux_recv(td, arg));
1243	case LINUX_SENDTO:
1244		return (linux_sendto(td, arg));
1245	case LINUX_RECVFROM:
1246		return (linux_recvfrom(td, arg));
1247	case LINUX_SHUTDOWN:
1248		return (linux_shutdown(td, arg));
1249	case LINUX_SETSOCKOPT:
1250		return (linux_setsockopt(td, arg));
1251	case LINUX_GETSOCKOPT:
1252		return (linux_getsockopt(td, arg));
1253	case LINUX_SENDMSG:
1254		return (linux_sendmsg(td, arg));
1255	case LINUX_RECVMSG:
1256		return (linux_recvmsg(td, arg));
1257	}
1258
1259	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1260	return (ENOSYS);
1261}
1262