linux_socket.c revision 144012
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 144012 2005-03-23 08:28:00Z das $");
31
32/* XXX we use functions that might not exist. */
33#include "opt_compat.h"
34#include "opt_inet6.h"
35
36#ifndef COMPAT_43
37#error "Unable to compile Linux-emulator due to missing COMPAT_43 option!"
38#endif
39
40#include <sys/param.h>
41#include <sys/proc.h>
42#include <sys/systm.h>
43#include <sys/sysproto.h>
44#include <sys/fcntl.h>
45#include <sys/file.h>
46#include <sys/limits.h>
47#include <sys/malloc.h>
48#include <sys/mbuf.h>
49#include <sys/socket.h>
50#include <sys/socketvar.h>
51#include <sys/syscallsubr.h>
52#include <sys/uio.h>
53#include <sys/syslog.h>
54
55#include <netinet/in.h>
56#include <netinet/in_systm.h>
57#include <netinet/ip.h>
58#ifdef INET6
59#include <netinet/ip6.h>
60#include <netinet6/ip6_var.h>
61#endif
62
63#include "opt_compat.h"
64
65#ifdef COMPAT_LINUX32
66#include <machine/../linux32/linux.h>
67#include <machine/../linux32/linux32_proto.h>
68#else
69#include <machine/../linux/linux.h>
70#include <machine/../linux/linux_proto.h>
71#endif
72#include <compat/linux/linux_socket.h>
73#include <compat/linux/linux_util.h>
74
75static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
76    struct malloc_type *);
77static int linux_to_bsd_domain(int);
78
79/*
80 * Reads a linux sockaddr and does any necessary translation.
81 * Linux sockaddrs don't have a length field, only a family.
82 */
83static int
84linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
85{
86	int osalen = len;
87
88	return (do_sa_get(sap, osa, &osalen, M_SONAME));
89}
90
91/*
92 * Copy the osockaddr structure pointed to by osa to kernel, adjust
93 * family and convert to sockaddr.
94 */
95static int
96do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
97    struct malloc_type *mtype)
98{
99	int error=0, bdom;
100	struct sockaddr *sa;
101	struct osockaddr *kosa;
102	int alloclen;
103#ifdef INET6
104	int oldv6size;
105	struct sockaddr_in6 *sin6;
106#endif
107
108	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
109		return (EINVAL);
110
111	alloclen = *osalen;
112#ifdef INET6
113	oldv6size = 0;
114	/*
115	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
116	 * if it's a v4-mapped address, so reserve the proper space
117	 * for it.
118	 */
119	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
120		alloclen = sizeof (struct sockaddr_in6);
121		oldv6size = 1;
122	}
123#endif
124
125	MALLOC(kosa, struct osockaddr *, alloclen, mtype, M_WAITOK);
126
127	if ((error = copyin(osa, kosa, *osalen)))
128		goto out;
129
130	bdom = linux_to_bsd_domain(kosa->sa_family);
131	if (bdom == -1) {
132		error = EINVAL;
133		goto out;
134	}
135
136#ifdef INET6
137	/*
138	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
139	 * which lacks the scope id compared with RFC2553 one. If we detect
140	 * the situation, reject the address and write a message to system log.
141	 *
142	 * Still accept addresses for which the scope id is not used.
143	 */
144	if (oldv6size && bdom == AF_INET6) {
145		sin6 = (struct sockaddr_in6 *)kosa;
146		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
147		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
148		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
149		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
150		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
151		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
152			sin6->sin6_scope_id = 0;
153		} else {
154			log(LOG_DEBUG,
155			    "obsolete pre-RFC2553 sockaddr_in6 rejected");
156			error = EINVAL;
157			goto out;
158		}
159	} else
160#endif
161	if (bdom == AF_INET)
162		alloclen = sizeof(struct sockaddr_in);
163
164	sa = (struct sockaddr *) kosa;
165	sa->sa_family = bdom;
166	sa->sa_len = alloclen;
167
168	*sap = sa;
169	*osalen = alloclen;
170	return (0);
171
172out:
173	FREE(kosa, mtype);
174	return (error);
175}
176
177static int
178linux_to_bsd_domain(int domain)
179{
180
181	switch (domain) {
182	case LINUX_AF_UNSPEC:
183		return (AF_UNSPEC);
184	case LINUX_AF_UNIX:
185		return (AF_LOCAL);
186	case LINUX_AF_INET:
187		return (AF_INET);
188	case LINUX_AF_INET6:
189		return (AF_INET6);
190	case LINUX_AF_AX25:
191		return (AF_CCITT);
192	case LINUX_AF_IPX:
193		return (AF_IPX);
194	case LINUX_AF_APPLETALK:
195		return (AF_APPLETALK);
196	}
197	return (-1);
198}
199
200#ifndef __alpha__
201static int
202bsd_to_linux_domain(int domain)
203{
204
205	switch (domain) {
206	case AF_UNSPEC:
207		return (LINUX_AF_UNSPEC);
208	case AF_LOCAL:
209		return (LINUX_AF_UNIX);
210	case AF_INET:
211		return (LINUX_AF_INET);
212	case AF_INET6:
213		return (LINUX_AF_INET6);
214	case AF_CCITT:
215		return (LINUX_AF_AX25);
216	case AF_IPX:
217		return (LINUX_AF_IPX);
218	case AF_APPLETALK:
219		return (LINUX_AF_APPLETALK);
220	}
221	return (-1);
222}
223
224static int
225linux_to_bsd_sockopt_level(int level)
226{
227
228	switch (level) {
229	case LINUX_SOL_SOCKET:
230		return (SOL_SOCKET);
231	}
232	return (level);
233}
234
235static int
236bsd_to_linux_sockopt_level(int level)
237{
238
239	switch (level) {
240	case SOL_SOCKET:
241		return (LINUX_SOL_SOCKET);
242	}
243	return (level);
244}
245
246static int
247linux_to_bsd_ip_sockopt(int opt)
248{
249
250	switch (opt) {
251	case LINUX_IP_TOS:
252		return (IP_TOS);
253	case LINUX_IP_TTL:
254		return (IP_TTL);
255	case LINUX_IP_OPTIONS:
256		return (IP_OPTIONS);
257	case LINUX_IP_MULTICAST_IF:
258		return (IP_MULTICAST_IF);
259	case LINUX_IP_MULTICAST_TTL:
260		return (IP_MULTICAST_TTL);
261	case LINUX_IP_MULTICAST_LOOP:
262		return (IP_MULTICAST_LOOP);
263	case LINUX_IP_ADD_MEMBERSHIP:
264		return (IP_ADD_MEMBERSHIP);
265	case LINUX_IP_DROP_MEMBERSHIP:
266		return (IP_DROP_MEMBERSHIP);
267	case LINUX_IP_HDRINCL:
268		return (IP_HDRINCL);
269	}
270	return (-1);
271}
272
273static int
274linux_to_bsd_so_sockopt(int opt)
275{
276
277	switch (opt) {
278	case LINUX_SO_DEBUG:
279		return (SO_DEBUG);
280	case LINUX_SO_REUSEADDR:
281		return (SO_REUSEADDR);
282	case LINUX_SO_TYPE:
283		return (SO_TYPE);
284	case LINUX_SO_ERROR:
285		return (SO_ERROR);
286	case LINUX_SO_DONTROUTE:
287		return (SO_DONTROUTE);
288	case LINUX_SO_BROADCAST:
289		return (SO_BROADCAST);
290	case LINUX_SO_SNDBUF:
291		return (SO_SNDBUF);
292	case LINUX_SO_RCVBUF:
293		return (SO_RCVBUF);
294	case LINUX_SO_KEEPALIVE:
295		return (SO_KEEPALIVE);
296	case LINUX_SO_OOBINLINE:
297		return (SO_OOBINLINE);
298	case LINUX_SO_LINGER:
299		return (SO_LINGER);
300	}
301	return (-1);
302}
303
304static int
305linux_to_bsd_msg_flags(int flags)
306{
307	int ret_flags = 0;
308
309	if (flags & LINUX_MSG_OOB)
310		ret_flags |= MSG_OOB;
311	if (flags & LINUX_MSG_PEEK)
312		ret_flags |= MSG_PEEK;
313	if (flags & LINUX_MSG_DONTROUTE)
314		ret_flags |= MSG_DONTROUTE;
315	if (flags & LINUX_MSG_CTRUNC)
316		ret_flags |= MSG_CTRUNC;
317	if (flags & LINUX_MSG_TRUNC)
318		ret_flags |= MSG_TRUNC;
319	if (flags & LINUX_MSG_DONTWAIT)
320		ret_flags |= MSG_DONTWAIT;
321	if (flags & LINUX_MSG_EOR)
322		ret_flags |= MSG_EOR;
323	if (flags & LINUX_MSG_WAITALL)
324		ret_flags |= MSG_WAITALL;
325	if (flags & LINUX_MSG_NOSIGNAL)
326		ret_flags |= MSG_NOSIGNAL;
327#if 0 /* not handled */
328	if (flags & LINUX_MSG_PROXY)
329		;
330	if (flags & LINUX_MSG_FIN)
331		;
332	if (flags & LINUX_MSG_SYN)
333		;
334	if (flags & LINUX_MSG_CONFIRM)
335		;
336	if (flags & LINUX_MSG_RST)
337		;
338	if (flags & LINUX_MSG_ERRQUEUE)
339		;
340#endif
341	return ret_flags;
342}
343
344static int
345linux_sa_put(struct osockaddr *osa)
346{
347	struct osockaddr sa;
348	int error, bdom;
349
350	/*
351	 * Only read/write the osockaddr family part, the rest is
352	 * not changed.
353	 */
354	error = copyin(osa, &sa, sizeof(sa.sa_family));
355	if (error)
356		return (error);
357
358	bdom = bsd_to_linux_domain(sa.sa_family);
359	if (bdom == -1)
360		return (EINVAL);
361
362	sa.sa_family = bdom;
363	error = copyout(&sa, osa, sizeof(sa.sa_family));
364	if (error)
365		return (error);
366
367	return (0);
368}
369
370static int
371linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
372    enum uio_seg segflg)
373{
374	struct mbuf *control;
375	struct sockaddr *to;
376	int error;
377
378	if (mp->msg_name != NULL) {
379		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
380		if (error)
381			return (error);
382		mp->msg_name = to;
383	} else
384		to = NULL;
385
386	if (mp->msg_control != NULL) {
387		struct cmsghdr *cmsg;
388
389		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
390			error = EINVAL;
391			goto bad;
392		}
393		error = sockargs(&control, mp->msg_control,
394		    mp->msg_controllen, MT_CONTROL);
395		if (error)
396			goto bad;
397
398		cmsg = mtod(control, struct cmsghdr *);
399		cmsg->cmsg_level = linux_to_bsd_sockopt_level(cmsg->cmsg_level);
400	} else
401		control = NULL;
402
403	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
404	    segflg);
405
406bad:
407	if (to)
408		FREE(to, M_SONAME);
409	return (error);
410}
411
412/* Return 0 if IP_HDRINCL is set for the given socket. */
413static int
414linux_check_hdrincl(struct thread *td, int s)
415{
416	int error, optval, size_val;
417
418	size_val = sizeof(optval);
419	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
420	    &optval, UIO_SYSSPACE, &size_val);
421	if (error)
422		return (error);
423
424	return (optval == 0);
425}
426
427struct linux_sendto_args {
428	int s;
429	l_uintptr_t msg;
430	int len;
431	int flags;
432	l_uintptr_t to;
433	int tolen;
434};
435
436/*
437 * Updated sendto() when IP_HDRINCL is set:
438 * tweak endian-dependent fields in the IP packet.
439 */
440static int
441linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
442{
443/*
444 * linux_ip_copysize defines how many bytes we should copy
445 * from the beginning of the IP packet before we customize it for BSD.
446 * It should include all the fields we modify (ip_len and ip_off).
447 */
448#define linux_ip_copysize	8
449
450	struct ip *packet;
451	struct msghdr msg;
452	struct iovec aiov[1];
453	int error;
454
455	/* Check that the packet isn't too big or too small. */
456	if (linux_args->len < linux_ip_copysize ||
457	    linux_args->len > IP_MAXPACKET)
458		return (EINVAL);
459
460	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
461
462	/* Make kernel copy of the packet to be sent */
463	if ((error = copyin(PTRIN(linux_args->msg), packet,
464	    linux_args->len)))
465		goto goout;
466
467	/* Convert fields from Linux to BSD raw IP socket format */
468	packet->ip_len = linux_args->len;
469	packet->ip_off = ntohs(packet->ip_off);
470
471	/* Prepare the msghdr and iovec structures describing the new packet */
472	msg.msg_name = PTRIN(linux_args->to);
473	msg.msg_namelen = linux_args->tolen;
474	msg.msg_iov = aiov;
475	msg.msg_iovlen = 1;
476	msg.msg_control = NULL;
477	msg.msg_flags = 0;
478	aiov[0].iov_base = (char *)packet;
479	aiov[0].iov_len = linux_args->len;
480	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
481	    UIO_SYSSPACE);
482goout:
483	free(packet, M_TEMP);
484	return (error);
485}
486
487struct linux_socket_args {
488	int domain;
489	int type;
490	int protocol;
491};
492
493static int
494linux_socket(struct thread *td, struct linux_socket_args *args)
495{
496	struct linux_socket_args linux_args;
497	struct socket_args /* {
498		int domain;
499		int type;
500		int protocol;
501	} */ bsd_args;
502	int error;
503	int retval_socket;
504
505	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
506		return (error);
507
508	bsd_args.protocol = linux_args.protocol;
509	bsd_args.type = linux_args.type;
510	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
511	if (bsd_args.domain == -1)
512		return (EINVAL);
513
514	retval_socket = socket(td, &bsd_args);
515	if (bsd_args.type == SOCK_RAW
516	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
517	    && bsd_args.domain == AF_INET
518	    && retval_socket >= 0) {
519		/* It's a raw IP socket: set the IP_HDRINCL option. */
520		int hdrincl;
521
522		hdrincl = 1;
523		/* We ignore any error returned by kern_setsockopt() */
524		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
525		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
526	}
527#ifdef INET6
528	/*
529	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
530	 * default and some apps depend on this. So, set V6ONLY to 0
531	 * for Linux apps if the sysctl value is set to 1.
532	 */
533	if (bsd_args.domain == PF_INET6 && retval_socket >= 0
534#ifndef KLD_MODULE
535	    /*
536	     * XXX: Avoid undefined symbol error with an IPv4 only
537	     * kernel.
538	     */
539	    && ip6_v6only
540#endif
541	    ) {
542		int v6only;
543
544		v6only = 0;
545		/* We ignore any error returned by setsockopt() */
546		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
547		    &v6only, UIO_SYSSPACE, sizeof(v6only));
548	}
549#endif
550
551	return (retval_socket);
552}
553
554struct linux_bind_args {
555	int s;
556	l_uintptr_t name;
557	int namelen;
558};
559
560static int
561linux_bind(struct thread *td, struct linux_bind_args *args)
562{
563	struct linux_bind_args linux_args;
564	struct sockaddr *sa;
565	int error;
566
567	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
568		return (error);
569
570	error = linux_getsockaddr(&sa, PTRIN(linux_args.name),
571	    linux_args.namelen);
572	if (error)
573		return (error);
574
575	return (kern_bind(td, linux_args.s, sa));
576}
577
578struct linux_connect_args {
579	int s;
580	l_uintptr_t name;
581	int namelen;
582};
583int linux_connect(struct thread *, struct linux_connect_args *);
584#endif /* !__alpha__*/
585
586int
587linux_connect(struct thread *td, struct linux_connect_args *args)
588{
589	struct linux_connect_args linux_args;
590	struct socket *so;
591	struct sockaddr *sa;
592	u_int fflag;
593	int error;
594
595#ifdef __alpha__
596	bcopy(args, &linux_args, sizeof(linux_args));
597#else
598	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
599		return (error);
600#endif /* __alpha__ */
601
602	error = linux_getsockaddr(&sa,
603	    (struct osockaddr *)PTRIN(linux_args.name),
604	    linux_args.namelen);
605	if (error)
606		return (error);
607
608	error = kern_connect(td, linux_args.s, sa);
609	if (error != EISCONN)
610		return (error);
611
612	/*
613	 * Linux doesn't return EISCONN the first time it occurs,
614	 * when on a non-blocking socket. Instead it returns the
615	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
616	 */
617	if ((error = fgetsock(td, linux_args.s, &so, &fflag)) != 0)
618		return(error);
619	error = EISCONN;
620	if (fflag & FNONBLOCK) {
621		if (so->so_emuldata == 0)
622			error = so->so_error;
623		so->so_emuldata = (void *)1;
624	}
625	fputsock(so);
626	return (error);
627}
628
629#ifndef __alpha__
630
631struct linux_listen_args {
632	int s;
633	int backlog;
634};
635
636static int
637linux_listen(struct thread *td, struct linux_listen_args *args)
638{
639	struct linux_listen_args linux_args;
640	struct listen_args /* {
641		int s;
642		int backlog;
643	} */ bsd_args;
644	int error;
645
646	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
647		return (error);
648
649	bsd_args.s = linux_args.s;
650	bsd_args.backlog = linux_args.backlog;
651	return (listen(td, &bsd_args));
652}
653
654struct linux_accept_args {
655	int s;
656	l_uintptr_t addr;
657	l_uintptr_t namelen;
658};
659
660static int
661linux_accept(struct thread *td, struct linux_accept_args *args)
662{
663	struct linux_accept_args linux_args;
664	struct accept_args /* {
665		int	s;
666		struct sockaddr * __restrict name;
667		socklen_t * __restrict anamelen;
668	} */ bsd_args;
669	struct close_args /* {
670		int     fd;
671	} */ c_args;
672	int error, fd;
673
674	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
675		return (error);
676
677	bsd_args.s = linux_args.s;
678	/* XXX: */
679	bsd_args.name = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
680	bsd_args.anamelen = PTRIN(linux_args.namelen);/* XXX */
681	error = oaccept(td, &bsd_args);
682	if (error)
683		return (error);
684	if (linux_args.addr) {
685		error = linux_sa_put(PTRIN(linux_args.addr));
686		if (error) {
687			c_args.fd = td->td_retval[0];
688			(void)close(td, &c_args);
689			return (error);
690		}
691	}
692
693	/*
694	 * linux appears not to copy flags from the parent socket to the
695	 * accepted one, so we must clear the flags in the new descriptor.
696	 * Ignore any errors, because we already have an open fd.
697	 */
698	fd = td->td_retval[0];
699	(void)kern_fcntl(td, fd, F_SETFL, 0);
700	td->td_retval[0] = fd;
701	return (0);
702}
703
704struct linux_getsockname_args {
705	int s;
706	l_uintptr_t addr;
707	l_uintptr_t namelen;
708};
709
710static int
711linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
712{
713	struct linux_getsockname_args linux_args;
714	struct getsockname_args /* {
715		int	fdes;
716		struct sockaddr * __restrict asa;
717		socklen_t * __restrict alen;
718	} */ bsd_args;
719	int error;
720
721	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
722		return (error);
723
724	bsd_args.fdes = linux_args.s;
725	/* XXX: */
726	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
727	bsd_args.alen = PTRIN(linux_args.namelen);	/* XXX */
728	error = ogetsockname(td, &bsd_args);
729	if (error)
730		return (error);
731	error = linux_sa_put(PTRIN(linux_args.addr));
732	if (error)
733		return (error);
734	return (0);
735}
736
737struct linux_getpeername_args {
738	int s;
739	l_uintptr_t addr;
740	l_uintptr_t namelen;
741};
742
743static int
744linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
745{
746	struct linux_getpeername_args linux_args;
747	struct ogetpeername_args /* {
748		int fdes;
749		caddr_t asa;
750		int *alen;
751	} */ bsd_args;
752	int error;
753
754	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
755		return (error);
756
757	bsd_args.fdes = linux_args.s;
758	bsd_args.asa = (caddr_t)PTRIN(linux_args.addr);
759	bsd_args.alen = (int *)PTRIN(linux_args.namelen);
760	error = ogetpeername(td, &bsd_args);
761	if (error)
762		return (error);
763	error = linux_sa_put(PTRIN(linux_args.addr));
764	if (error)
765		return (error);
766	return (0);
767}
768
769struct linux_socketpair_args {
770	int domain;
771	int type;
772	int protocol;
773	l_uintptr_t rsv;
774};
775
776static int
777linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
778{
779	struct linux_socketpair_args linux_args;
780	struct socketpair_args /* {
781		int domain;
782		int type;
783		int protocol;
784		int *rsv;
785	} */ bsd_args;
786	int error;
787
788	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
789		return (error);
790
791	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
792	if (bsd_args.domain == -1)
793		return (EINVAL);
794
795	bsd_args.type = linux_args.type;
796	bsd_args.protocol = linux_args.protocol;
797	bsd_args.rsv = (int *)PTRIN(linux_args.rsv);
798	return (socketpair(td, &bsd_args));
799}
800
801struct linux_send_args {
802	int s;
803	l_uintptr_t msg;
804	int len;
805	int flags;
806};
807
808static int
809linux_send(struct thread *td, struct linux_send_args *args)
810{
811	struct linux_send_args linux_args;
812	struct sendto_args /* {
813		int s;
814		caddr_t buf;
815		int len;
816		int flags;
817		caddr_t to;
818		int tolen;
819	} */ bsd_args;
820	int error;
821
822	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
823		return (error);
824
825	bsd_args.s = linux_args.s;
826	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
827	bsd_args.len = linux_args.len;
828	bsd_args.flags = linux_args.flags;
829	bsd_args.to = NULL;
830	bsd_args.tolen = 0;
831	return sendto(td, &bsd_args);
832}
833
834struct linux_recv_args {
835	int s;
836	l_uintptr_t msg;
837	int len;
838	int flags;
839};
840
841static int
842linux_recv(struct thread *td, struct linux_recv_args *args)
843{
844	struct linux_recv_args linux_args;
845	struct recvfrom_args /* {
846		int s;
847		caddr_t buf;
848		int len;
849		int flags;
850		struct sockaddr *from;
851		socklen_t fromlenaddr;
852	} */ bsd_args;
853	int error;
854
855	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
856		return (error);
857
858	bsd_args.s = linux_args.s;
859	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
860	bsd_args.len = linux_args.len;
861	bsd_args.flags = linux_args.flags;
862	bsd_args.from = NULL;
863	bsd_args.fromlenaddr = 0;
864	return (recvfrom(td, &bsd_args));
865}
866
867static int
868linux_sendto(struct thread *td, struct linux_sendto_args *args)
869{
870	struct linux_sendto_args linux_args;
871	struct msghdr msg;
872	struct iovec aiov;
873	int error;
874
875	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
876		return (error);
877
878	if (linux_check_hdrincl(td, linux_args.s) == 0)
879		/* IP_HDRINCL set, tweak the packet before sending */
880		return (linux_sendto_hdrincl(td, &linux_args));
881
882	msg.msg_name = PTRIN(linux_args.to);
883	msg.msg_namelen = linux_args.tolen;
884	msg.msg_iov = &aiov;
885	msg.msg_iovlen = 1;
886	msg.msg_control = NULL;
887	msg.msg_flags = 0;
888	aiov.iov_base = PTRIN(linux_args.msg);
889	aiov.iov_len = linux_args.len;
890	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
891	    UIO_USERSPACE);
892	return (error);
893}
894
895struct linux_recvfrom_args {
896	int s;
897	l_uintptr_t buf;
898	int len;
899	int flags;
900	l_uintptr_t from;
901	l_uintptr_t fromlen;
902};
903
904static int
905linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
906{
907	struct linux_recvfrom_args linux_args;
908	struct recvfrom_args /* {
909		int	s;
910		caddr_t	buf;
911		size_t	len;
912		int	flags;
913		struct sockaddr * __restrict from;
914		socklen_t * __restrict fromlenaddr;
915	} */ bsd_args;
916	int error;
917
918	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
919		return (error);
920
921	bsd_args.s = linux_args.s;
922	bsd_args.buf = PTRIN(linux_args.buf);
923	bsd_args.len = linux_args.len;
924	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
925	/* XXX: */
926	bsd_args.from = (struct sockaddr * __restrict)PTRIN(linux_args.from);
927	bsd_args.fromlenaddr = PTRIN(linux_args.fromlen);/* XXX */
928	error = orecvfrom(td, &bsd_args);
929	if (error)
930		return (error);
931	if (linux_args.from) {
932		error = linux_sa_put((struct osockaddr *)
933		    PTRIN(linux_args.from));
934		if (error)
935			return (error);
936	}
937	return (0);
938}
939
940struct linux_sendmsg_args {
941	int s;
942	l_uintptr_t msg;
943	int flags;
944};
945
946static int
947linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
948{
949	struct linux_sendmsg_args linux_args;
950	struct msghdr msg;
951	struct iovec *iov;
952	int error;
953
954	/* XXXTJR sendmsg is broken on amd64 */
955
956	error = copyin(args, &linux_args, sizeof(linux_args));
957	if (error)
958		return (error);
959	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
960	if (error)
961		return (error);
962	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
963	if (error)
964		return (error);
965	msg.msg_iov = iov;
966	msg.msg_flags = 0;
967	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
968	    UIO_USERSPACE);
969	free(iov, M_IOV);
970	return (error);
971}
972
973struct linux_recvmsg_args {
974	int s;
975	l_uintptr_t msg;
976	int flags;
977};
978
979static int
980linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
981{
982	struct linux_recvmsg_args linux_args;
983	struct recvmsg_args /* {
984		int	s;
985		struct	msghdr *msg;
986		int	flags;
987	} */ bsd_args;
988	struct msghdr msg;
989	struct cmsghdr *cmsg;
990	int error;
991
992	/* XXXTJR recvmsg is broken on amd64 */
993
994	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
995		return (error);
996
997	bsd_args.s = linux_args.s;
998	bsd_args.msg = PTRIN(linux_args.msg);
999	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
1000	error = recvmsg(td, &bsd_args);
1001	if (error)
1002		return (error);
1003
1004	if (bsd_args.msg->msg_control != NULL &&
1005	    bsd_args.msg->msg_controllen > 0) {
1006		cmsg = (struct cmsghdr*)bsd_args.msg->msg_control;
1007		cmsg->cmsg_level = bsd_to_linux_sockopt_level(cmsg->cmsg_level);
1008	}
1009
1010	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
1011	if (error)
1012		return (error);
1013	if (msg.msg_name && msg.msg_namelen > 2)
1014		error = linux_sa_put(msg.msg_name);
1015	return (error);
1016}
1017
1018struct linux_shutdown_args {
1019	int s;
1020	int how;
1021};
1022
1023static int
1024linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1025{
1026	struct linux_shutdown_args linux_args;
1027	struct shutdown_args /* {
1028		int s;
1029		int how;
1030	} */ bsd_args;
1031	int error;
1032
1033	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1034		return (error);
1035
1036	bsd_args.s = linux_args.s;
1037	bsd_args.how = linux_args.how;
1038	return (shutdown(td, &bsd_args));
1039}
1040
1041struct linux_setsockopt_args {
1042	int s;
1043	int level;
1044	int optname;
1045	l_uintptr_t optval;
1046	int optlen;
1047};
1048
1049static int
1050linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1051{
1052	struct linux_setsockopt_args linux_args;
1053	struct setsockopt_args /* {
1054		int s;
1055		int level;
1056		int name;
1057		caddr_t val;
1058		int valsize;
1059	} */ bsd_args;
1060	int error, name;
1061
1062	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1063		return (error);
1064
1065	bsd_args.s = linux_args.s;
1066	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1067	switch (bsd_args.level) {
1068	case SOL_SOCKET:
1069		name = linux_to_bsd_so_sockopt(linux_args.optname);
1070		break;
1071	case IPPROTO_IP:
1072		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1073		break;
1074	case IPPROTO_TCP:
1075		/* Linux TCP option values match BSD's */
1076		name = linux_args.optname;
1077		break;
1078	default:
1079		name = -1;
1080		break;
1081	}
1082	if (name == -1)
1083		return (EINVAL);
1084
1085	bsd_args.name = name;
1086	bsd_args.val = PTRIN(linux_args.optval);
1087	bsd_args.valsize = linux_args.optlen;
1088	return (setsockopt(td, &bsd_args));
1089}
1090
1091struct linux_getsockopt_args {
1092	int s;
1093	int level;
1094	int optname;
1095	l_uintptr_t optval;
1096	l_uintptr_t optlen;
1097};
1098
1099static int
1100linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1101{
1102	struct linux_getsockopt_args linux_args;
1103	struct getsockopt_args /* {
1104		int s;
1105		int level;
1106		int name;
1107		caddr_t val;
1108		int *avalsize;
1109	} */ bsd_args;
1110	int error, name;
1111
1112	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1113		return (error);
1114
1115	bsd_args.s = linux_args.s;
1116	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1117	switch (bsd_args.level) {
1118	case SOL_SOCKET:
1119		name = linux_to_bsd_so_sockopt(linux_args.optname);
1120		break;
1121	case IPPROTO_IP:
1122		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1123		break;
1124	case IPPROTO_TCP:
1125		/* Linux TCP option values match BSD's */
1126		name = linux_args.optname;
1127		break;
1128	default:
1129		name = -1;
1130		break;
1131	}
1132	if (name == -1)
1133		return (EINVAL);
1134
1135	bsd_args.name = name;
1136	bsd_args.val = PTRIN(linux_args.optval);
1137	bsd_args.avalsize = PTRIN(linux_args.optlen);
1138	return (getsockopt(td, &bsd_args));
1139}
1140
1141int
1142linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1143{
1144	void *arg = (void *)(intptr_t)args->args;
1145
1146	switch (args->what) {
1147	case LINUX_SOCKET:
1148		return (linux_socket(td, arg));
1149	case LINUX_BIND:
1150		return (linux_bind(td, arg));
1151	case LINUX_CONNECT:
1152		return (linux_connect(td, arg));
1153	case LINUX_LISTEN:
1154		return (linux_listen(td, arg));
1155	case LINUX_ACCEPT:
1156		return (linux_accept(td, arg));
1157	case LINUX_GETSOCKNAME:
1158		return (linux_getsockname(td, arg));
1159	case LINUX_GETPEERNAME:
1160		return (linux_getpeername(td, arg));
1161	case LINUX_SOCKETPAIR:
1162		return (linux_socketpair(td, arg));
1163	case LINUX_SEND:
1164		return (linux_send(td, arg));
1165	case LINUX_RECV:
1166		return (linux_recv(td, arg));
1167	case LINUX_SENDTO:
1168		return (linux_sendto(td, arg));
1169	case LINUX_RECVFROM:
1170		return (linux_recvfrom(td, arg));
1171	case LINUX_SHUTDOWN:
1172		return (linux_shutdown(td, arg));
1173	case LINUX_SETSOCKOPT:
1174		return (linux_setsockopt(td, arg));
1175	case LINUX_GETSOCKOPT:
1176		return (linux_getsockopt(td, arg));
1177	case LINUX_SENDMSG:
1178		return (linux_sendmsg(td, arg));
1179	case LINUX_RECVMSG:
1180		return (linux_recvmsg(td, arg));
1181	}
1182
1183	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1184	return (ENOSYS);
1185}
1186#endif	/*!__alpha__*/
1187