linux_socket.c revision 132347 
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 132347 2004-07-18 09:26:34Z dwmalone $");
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 <machine/../linux/linux.h>
64#include <machine/../linux/linux_proto.h>
65#include <compat/linux/linux_socket.h>
66#include <compat/linux/linux_util.h>
67
68static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
69    struct malloc_type *);
70static int linux_to_bsd_domain(int);
71
72/*
73 * Reads a linux sockaddr and does any necessary translation.
74 * Linux sockaddrs don't have a length field, only a family.
75 */
76static int
77linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
78{
79	int osalen = len;
80
81	return (do_sa_get(sap, osa, &osalen, M_SONAME));
82}
83
84/*
85 * Copy the osockaddr structure pointed to by osa to kernel, adjust
86 * family and convert to sockaddr.
87 */
88static int
89do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
90    struct malloc_type *mtype)
91{
92	int error=0, bdom;
93	struct sockaddr *sa;
94	struct osockaddr *kosa;
95	int alloclen;
96#ifdef INET6
97	int oldv6size;
98	struct sockaddr_in6 *sin6;
99#endif
100
101	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
102		return (EINVAL);
103
104	alloclen = *osalen;
105#ifdef INET6
106	oldv6size = 0;
107	/*
108	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
109	 * if it's a v4-mapped address, so reserve the proper space
110	 * for it.
111	 */
112	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
113		alloclen = sizeof (struct sockaddr_in6);
114		oldv6size = 1;
115	}
116#endif
117
118	MALLOC(kosa, struct osockaddr *, alloclen, mtype, M_WAITOK);
119
120	if ((error = copyin(osa, kosa, *osalen)))
121		goto out;
122
123	bdom = linux_to_bsd_domain(kosa->sa_family);
124	if (bdom == -1) {
125		error = EINVAL;
126		goto out;
127	}
128
129#ifdef INET6
130	/*
131	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
132	 * which lacks the scope id compared with RFC2553 one. If we detect
133	 * the situation, reject the address and write a message to system log.
134	 *
135	 * Still accept addresses for which the scope id is not used.
136	 */
137	if (oldv6size && bdom == AF_INET6) {
138		sin6 = (struct sockaddr_in6 *)kosa;
139		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
140		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
141		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
142		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
143		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
144		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
145			sin6->sin6_scope_id = 0;
146		} else {
147			log(LOG_DEBUG,
148			    "obsolete pre-RFC2553 sockaddr_in6 rejected");
149			error = EINVAL;
150			goto out;
151		}
152	} else
153#endif
154	if (bdom == AF_INET)
155		alloclen = sizeof(struct sockaddr_in);
156
157	sa = (struct sockaddr *) kosa;
158	sa->sa_family = bdom;
159	sa->sa_len = alloclen;
160
161	*sap = sa;
162	*osalen = alloclen;
163	return (0);
164
165out:
166	FREE(kosa, mtype);
167	return (error);
168}
169
170static int
171linux_to_bsd_domain(int domain)
172{
173
174	switch (domain) {
175	case LINUX_AF_UNSPEC:
176		return (AF_UNSPEC);
177	case LINUX_AF_UNIX:
178		return (AF_LOCAL);
179	case LINUX_AF_INET:
180		return (AF_INET);
181	case LINUX_AF_INET6:
182		return (AF_INET6);
183	case LINUX_AF_AX25:
184		return (AF_CCITT);
185	case LINUX_AF_IPX:
186		return (AF_IPX);
187	case LINUX_AF_APPLETALK:
188		return (AF_APPLETALK);
189	}
190	return (-1);
191}
192
193#ifndef __alpha__
194static int
195bsd_to_linux_domain(int domain)
196{
197
198	switch (domain) {
199	case AF_UNSPEC:
200		return (LINUX_AF_UNSPEC);
201	case AF_LOCAL:
202		return (LINUX_AF_UNIX);
203	case AF_INET:
204		return (LINUX_AF_INET);
205	case AF_INET6:
206		return (LINUX_AF_INET6);
207	case AF_CCITT:
208		return (LINUX_AF_AX25);
209	case AF_IPX:
210		return (LINUX_AF_IPX);
211	case AF_APPLETALK:
212		return (LINUX_AF_APPLETALK);
213	}
214	return (-1);
215}
216
217static int
218linux_to_bsd_sockopt_level(int level)
219{
220
221	switch (level) {
222	case LINUX_SOL_SOCKET:
223		return (SOL_SOCKET);
224	}
225	return (level);
226}
227
228static int
229bsd_to_linux_sockopt_level(int level)
230{
231
232	switch (level) {
233	case SOL_SOCKET:
234		return (LINUX_SOL_SOCKET);
235	}
236	return (level);
237}
238
239static int
240linux_to_bsd_ip_sockopt(int opt)
241{
242
243	switch (opt) {
244	case LINUX_IP_TOS:
245		return (IP_TOS);
246	case LINUX_IP_TTL:
247		return (IP_TTL);
248	case LINUX_IP_OPTIONS:
249		return (IP_OPTIONS);
250	case LINUX_IP_MULTICAST_IF:
251		return (IP_MULTICAST_IF);
252	case LINUX_IP_MULTICAST_TTL:
253		return (IP_MULTICAST_TTL);
254	case LINUX_IP_MULTICAST_LOOP:
255		return (IP_MULTICAST_LOOP);
256	case LINUX_IP_ADD_MEMBERSHIP:
257		return (IP_ADD_MEMBERSHIP);
258	case LINUX_IP_DROP_MEMBERSHIP:
259		return (IP_DROP_MEMBERSHIP);
260	case LINUX_IP_HDRINCL:
261		return (IP_HDRINCL);
262	}
263	return (-1);
264}
265
266static int
267linux_to_bsd_so_sockopt(int opt)
268{
269
270	switch (opt) {
271	case LINUX_SO_DEBUG:
272		return (SO_DEBUG);
273	case LINUX_SO_REUSEADDR:
274		return (SO_REUSEADDR);
275	case LINUX_SO_TYPE:
276		return (SO_TYPE);
277	case LINUX_SO_ERROR:
278		return (SO_ERROR);
279	case LINUX_SO_DONTROUTE:
280		return (SO_DONTROUTE);
281	case LINUX_SO_BROADCAST:
282		return (SO_BROADCAST);
283	case LINUX_SO_SNDBUF:
284		return (SO_SNDBUF);
285	case LINUX_SO_RCVBUF:
286		return (SO_RCVBUF);
287	case LINUX_SO_KEEPALIVE:
288		return (SO_KEEPALIVE);
289	case LINUX_SO_OOBINLINE:
290		return (SO_OOBINLINE);
291	case LINUX_SO_LINGER:
292		return (SO_LINGER);
293	}
294	return (-1);
295}
296
297static int
298linux_to_bsd_msg_flags(int flags)
299{
300	int ret_flags = 0;
301
302	if (flags & LINUX_MSG_OOB)
303		ret_flags |= MSG_OOB;
304	if (flags & LINUX_MSG_PEEK)
305		ret_flags |= MSG_PEEK;
306	if (flags & LINUX_MSG_DONTROUTE)
307		ret_flags |= MSG_DONTROUTE;
308	if (flags & LINUX_MSG_CTRUNC)
309		ret_flags |= MSG_CTRUNC;
310	if (flags & LINUX_MSG_TRUNC)
311		ret_flags |= MSG_TRUNC;
312	if (flags & LINUX_MSG_DONTWAIT)
313		ret_flags |= MSG_DONTWAIT;
314	if (flags & LINUX_MSG_EOR)
315		ret_flags |= MSG_EOR;
316	if (flags & LINUX_MSG_WAITALL)
317		ret_flags |= MSG_WAITALL;
318#if 0 /* not handled */
319	if (flags & LINUX_MSG_PROXY)
320		;
321	if (flags & LINUX_MSG_FIN)
322		;
323	if (flags & LINUX_MSG_SYN)
324		;
325	if (flags & LINUX_MSG_CONFIRM)
326		;
327	if (flags & LINUX_MSG_RST)
328		;
329	if (flags & LINUX_MSG_ERRQUEUE)
330		;
331	if (flags & LINUX_MSG_NOSIGNAL)
332		;
333#endif
334	return ret_flags;
335}
336
337static int
338linux_sa_put(struct osockaddr *osa)
339{
340	struct osockaddr sa;
341	int error, bdom;
342
343	/*
344	 * Only read/write the osockaddr family part, the rest is
345	 * not changed.
346	 */
347	error = copyin(osa, &sa, sizeof(sa.sa_family));
348	if (error)
349		return (error);
350
351	bdom = bsd_to_linux_domain(sa.sa_family);
352	if (bdom == -1)
353		return (EINVAL);
354
355	sa.sa_family = bdom;
356	error = copyout(&sa, osa, sizeof(sa.sa_family));
357	if (error)
358		return (error);
359
360	return (0);
361}
362
363static int
364linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags)
365{
366	struct mbuf *control;
367	struct sockaddr *to;
368	int error;
369
370	if (mp->msg_name != NULL) {
371		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
372		if (error)
373			return (error);
374		mp->msg_name = to;
375	} else
376		to = NULL;
377
378	if (mp->msg_control != NULL) {
379		struct cmsghdr *cmsg;
380
381		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
382			error = EINVAL;
383			goto bad;
384		}
385		error = sockargs(&control, mp->msg_control,
386		    mp->msg_controllen, MT_CONTROL);
387		if (error)
388			goto bad;
389
390		cmsg = mtod(control, struct cmsghdr *);
391		cmsg->cmsg_level = linux_to_bsd_sockopt_level(cmsg->cmsg_level);
392	} else
393		control = NULL;
394
395	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control);
396
397bad:
398	if (to)
399		FREE(to, M_SONAME);
400	return (error);
401}
402
403/* Return 0 if IP_HDRINCL is set for the given socket. */
404static int
405linux_check_hdrincl(struct thread *td, int s)
406{
407	int error, optval, size_val;
408
409	size_val = sizeof(optval);
410	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
411	    &optval, UIO_SYSSPACE, &size_val);
412	if (error)
413		return (error);
414
415	return (optval == 0);
416}
417
418struct linux_sendto_args {
419	int s;
420	void *msg;
421	int len;
422	int flags;
423	caddr_t to;
424	int tolen;
425};
426
427/*
428 * Updated sendto() when IP_HDRINCL is set:
429 * tweak endian-dependent fields in the IP packet.
430 */
431static int
432linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
433{
434/*
435 * linux_ip_copysize defines how many bytes we should copy
436 * from the beginning of the IP packet before we customize it for BSD.
437 * It should include all the fields we modify (ip_len and ip_off)
438 * and be as small as possible to minimize copying overhead.
439 */
440#define linux_ip_copysize	8
441
442	struct ip *packet;
443	caddr_t sg = stackgap_init();
444	struct msghdr msg;
445	struct iovec aiov[2];
446	int error;
447
448	/* Check the packet isn't too small before we mess with it */
449	if (linux_args->len < linux_ip_copysize)
450		return (EINVAL);
451
452	/*
453	 * Tweaking the user buffer in place would be bad manners.
454	 * We create a corrected IP header with just the needed length,
455	 * then use an iovec to glue it to the rest of the user packet
456	 * when calling sendit().
457	 */
458	packet = (struct ip *)stackgap_alloc(&sg, linux_ip_copysize);
459
460	/* Make a copy of the beginning of the packet to be sent */
461	if ((error = copyin(linux_args->msg, packet, linux_ip_copysize)))
462		return (error);
463
464	/* Convert fields from Linux to BSD raw IP socket format */
465	packet->ip_len = linux_args->len;
466	packet->ip_off = ntohs(packet->ip_off);
467
468	/* Prepare the msghdr and iovec structures describing the new packet */
469	msg.msg_name = linux_args->to;
470	msg.msg_namelen = linux_args->tolen;
471	msg.msg_iov = aiov;
472	msg.msg_iovlen = 2;
473	msg.msg_control = NULL;
474	msg.msg_flags = 0;
475	aiov[0].iov_base = (char *)packet;
476	aiov[0].iov_len = linux_ip_copysize;
477	aiov[1].iov_base = (char *)(linux_args->msg) + linux_ip_copysize;
478	aiov[1].iov_len = linux_args->len - linux_ip_copysize;
479	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags);
480	return (error);
481}
482
483struct linux_socket_args {
484	int domain;
485	int type;
486	int protocol;
487};
488
489static int
490linux_socket(struct thread *td, struct linux_socket_args *args)
491{
492	struct linux_socket_args linux_args;
493	struct socket_args /* {
494		int domain;
495		int type;
496		int protocol;
497	} */ bsd_args;
498	int error;
499	int retval_socket;
500
501	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
502		return (error);
503
504	bsd_args.protocol = linux_args.protocol;
505	bsd_args.type = linux_args.type;
506	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
507	if (bsd_args.domain == -1)
508		return (EINVAL);
509
510	retval_socket = socket(td, &bsd_args);
511	if (bsd_args.type == SOCK_RAW
512	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
513	    && bsd_args.domain == AF_INET
514	    && retval_socket >= 0) {
515		/* It's a raw IP socket: set the IP_HDRINCL option. */
516		int hdrincl;
517
518		hdrincl = 1;
519		/* We ignore any error returned by kern_setsockopt() */
520		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
521		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
522	}
523#ifdef INET6
524	/*
525	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
526	 * default and some apps depend on this. So, set V6ONLY to 0
527	 * for Linux apps if the sysctl value is set to 1.
528	 */
529	if (bsd_args.domain == PF_INET6 && retval_socket >= 0
530#ifndef KLD_MODULE
531	    /*
532	     * XXX: Avoid undefined symbol error with an IPv4 only
533	     * kernel.
534	     */
535	    && ip6_v6only
536#endif
537	    ) {
538		int v6only;
539
540		v6only = 0;
541		/* We ignore any error returned by setsockopt() */
542		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
543		    &v6only, UIO_SYSSPACE, sizeof(v6only));
544	}
545#endif
546
547	return (retval_socket);
548}
549
550struct linux_bind_args {
551	int s;
552	struct osockaddr *name;
553	int namelen;
554};
555
556static int
557linux_bind(struct thread *td, struct linux_bind_args *args)
558{
559	struct linux_bind_args linux_args;
560	struct sockaddr *sa;
561	int error;
562
563	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
564		return (error);
565
566	error = linux_getsockaddr(&sa, linux_args.name, linux_args.namelen);
567	if (error)
568		return (error);
569
570	return (kern_bind(td, linux_args.s, sa));
571}
572
573struct linux_connect_args {
574	int s;
575	struct osockaddr * name;
576	int namelen;
577};
578int linux_connect(struct thread *, struct linux_connect_args *);
579#endif /* !__alpha__*/
580
581int
582linux_connect(struct thread *td, struct linux_connect_args *args)
583{
584	struct linux_connect_args linux_args;
585	struct socket *so;
586	struct sockaddr *sa;
587	u_int fflag;
588	int error;
589
590#ifdef __alpha__
591	bcopy(args, &linux_args, sizeof(linux_args));
592#else
593	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
594		return (error);
595#endif /* __alpha__ */
596
597	error = linux_getsockaddr(&sa, (struct osockaddr *)linux_args.name,
598	    linux_args.namelen);
599	if (error)
600		return (error);
601
602	error = kern_connect(td, linux_args.s, sa);
603	if (error != EISCONN)
604		return (error);
605
606	/*
607	 * Linux doesn't return EISCONN the first time it occurs,
608	 * when on a non-blocking socket. Instead it returns the
609	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
610	 */
611	if ((error = fgetsock(td, linux_args.s, &so, &fflag)) != 0)
612		return(error);
613	error = EISCONN;
614	if (fflag & FNONBLOCK) {
615		if (so->so_emuldata == 0)
616			error = so->so_error;
617		so->so_emuldata = (void *)1;
618	}
619	fputsock(so);
620	return (error);
621}
622
623#ifndef __alpha__
624
625struct linux_listen_args {
626	int s;
627	int backlog;
628};
629
630static int
631linux_listen(struct thread *td, struct linux_listen_args *args)
632{
633	struct linux_listen_args linux_args;
634	struct listen_args /* {
635		int s;
636		int backlog;
637	} */ bsd_args;
638	int error;
639
640	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
641		return (error);
642
643	bsd_args.s = linux_args.s;
644	bsd_args.backlog = linux_args.backlog;
645	return (listen(td, &bsd_args));
646}
647
648struct linux_accept_args {
649	int s;
650	struct osockaddr *addr;
651	int *namelen;
652};
653
654static int
655linux_accept(struct thread *td, struct linux_accept_args *args)
656{
657	struct linux_accept_args linux_args;
658	struct accept_args /* {
659		int	s;
660		struct sockaddr * __restrict name;
661		socklen_t * __restrict anamelen;
662	} */ bsd_args;
663	struct close_args /* {
664		int     fd;
665	} */ c_args;
666	struct fcntl_args /* {
667		int	fd;
668		int	cmd;
669		long	arg;
670	} */ f_args;
671	int error;
672
673	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
674		return (error);
675
676	bsd_args.s = linux_args.s;
677	/* XXX: */
678	bsd_args.name = (struct sockaddr * __restrict)linux_args.addr;
679	bsd_args.anamelen = linux_args.namelen;		/* XXX */
680	error = oaccept(td, &bsd_args);
681	if (error)
682		return (error);
683	if (linux_args.addr) {
684		error = linux_sa_put(linux_args.addr);
685		if (error) {
686			c_args.fd = td->td_retval[0];
687			(void)close(td, &c_args);
688			return (error);
689		}
690	}
691
692	/*
693	 * linux appears not to copy flags from the parent socket to the
694	 * accepted one, so we must clear the flags in the new descriptor.
695	 * Ignore any errors, because we already have an open fd.
696	 */
697	f_args.fd = td->td_retval[0];
698	f_args.cmd = F_SETFL;
699	f_args.arg = 0;
700	(void)fcntl(td, &f_args);
701	td->td_retval[0] = f_args.fd;
702	return (0);
703}
704
705struct linux_getsockname_args {
706	int s;
707	struct osockaddr *addr;
708	int *namelen;
709};
710
711static int
712linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
713{
714	struct linux_getsockname_args linux_args;
715	struct getsockname_args /* {
716		int	fdes;
717		struct sockaddr * __restrict asa;
718		socklen_t * __restrict alen;
719	} */ bsd_args;
720	int error;
721
722	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
723		return (error);
724
725	bsd_args.fdes = linux_args.s;
726	/* XXX: */
727	bsd_args.asa = (struct sockaddr * __restrict)linux_args.addr;
728	bsd_args.alen = linux_args.namelen;	/* XXX */
729	error = ogetsockname(td, &bsd_args);
730	if (error)
731		return (error);
732	error = linux_sa_put(linux_args.addr);
733	if (error)
734		return (error);
735	return (0);
736}
737
738struct linux_getpeername_args {
739	int s;
740	struct osockaddr *addr;
741	int *namelen;
742};
743
744static int
745linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
746{
747	struct linux_getpeername_args linux_args;
748	struct ogetpeername_args /* {
749		int fdes;
750		caddr_t asa;
751		int *alen;
752	} */ bsd_args;
753	int error;
754
755	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
756		return (error);
757
758	bsd_args.fdes = linux_args.s;
759	bsd_args.asa = (caddr_t) linux_args.addr;
760	bsd_args.alen = linux_args.namelen;
761	error = ogetpeername(td, &bsd_args);
762	if (error)
763		return (error);
764	error = linux_sa_put(linux_args.addr);
765	if (error)
766		return (error);
767	return (0);
768}
769
770struct linux_socketpair_args {
771	int domain;
772	int type;
773	int protocol;
774	int *rsv;
775};
776
777static int
778linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
779{
780	struct linux_socketpair_args linux_args;
781	struct socketpair_args /* {
782		int domain;
783		int type;
784		int protocol;
785		int *rsv;
786	} */ bsd_args;
787	int error;
788
789	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
790		return (error);
791
792	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
793	if (bsd_args.domain == -1)
794		return (EINVAL);
795
796	bsd_args.type = linux_args.type;
797	bsd_args.protocol = linux_args.protocol;
798	bsd_args.rsv = linux_args.rsv;
799	return (socketpair(td, &bsd_args));
800}
801
802struct linux_send_args {
803	int s;
804	void *msg;
805	int len;
806	int flags;
807};
808
809static int
810linux_send(struct thread *td, struct linux_send_args *args)
811{
812	struct linux_send_args linux_args;
813	struct sendto_args /* {
814		int s;
815		caddr_t buf;
816		int len;
817		int flags;
818		caddr_t to;
819		int tolen;
820	} */ bsd_args;
821	int error;
822
823	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
824		return (error);
825
826	bsd_args.s = linux_args.s;
827	bsd_args.buf = linux_args.msg;
828	bsd_args.len = linux_args.len;
829	bsd_args.flags = linux_args.flags;
830	bsd_args.to = NULL;
831	bsd_args.tolen = 0;
832	return (sendto(td, &bsd_args));
833}
834
835struct linux_recv_args {
836	int s;
837	void *msg;
838	int len;
839	int flags;
840};
841
842static int
843linux_recv(struct thread *td, struct linux_recv_args *args)
844{
845	struct linux_recv_args linux_args;
846	struct recvfrom_args /* {
847		int s;
848		caddr_t buf;
849		int len;
850		int flags;
851		struct sockaddr *from;
852		socklen_t fromlenaddr;
853	} */ bsd_args;
854	int error;
855
856	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
857		return (error);
858
859	bsd_args.s = linux_args.s;
860	bsd_args.buf = linux_args.msg;
861	bsd_args.len = linux_args.len;
862	bsd_args.flags = linux_args.flags;
863	bsd_args.from = NULL;
864	bsd_args.fromlenaddr = 0;
865	return (recvfrom(td, &bsd_args));
866}
867
868static int
869linux_sendto(struct thread *td, struct linux_sendto_args *args)
870{
871	struct linux_sendto_args linux_args;
872	struct msghdr msg;
873	struct iovec aiov;
874	int error;
875
876	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
877		return (error);
878
879	if (linux_check_hdrincl(td, linux_args.s) == 0)
880		/* IP_HDRINCL set, tweak the packet before sending */
881		return (linux_sendto_hdrincl(td, &linux_args));
882
883	msg.msg_name = linux_args.to;
884	msg.msg_namelen = linux_args.tolen;
885	msg.msg_iov = &aiov;
886	msg.msg_iovlen = 1;
887	msg.msg_control = NULL;
888	msg.msg_flags = 0;
889	aiov.iov_base = linux_args.msg;
890	aiov.iov_len = linux_args.len;
891	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags);
892	return (error);
893}
894
895struct linux_recvfrom_args {
896	int s;
897	void *buf;
898	int len;
899	int flags;
900	caddr_t from;
901	int *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 = 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)linux_args.from;
927	bsd_args.fromlenaddr = 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 *) linux_args.from);
933		if (error)
934			return (error);
935	}
936	return (0);
937}
938
939struct linux_sendmsg_args {
940	int s;
941	const struct msghdr *msg;
942	int flags;
943};
944
945static int
946linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
947{
948	struct linux_sendmsg_args linux_args;
949	struct msghdr msg;
950	struct iovec *iov;
951	int error;
952
953	error = copyin(args, &linux_args, sizeof(linux_args));
954	if (error)
955		return (error);
956	error = copyin(linux_args.msg, &msg, sizeof(msg));
957	if (error)
958		return (error);
959	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
960	if (error)
961		return (error);
962	msg.msg_iov = iov;
963	msg.msg_flags = 0;
964	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags);
965	free(iov, M_IOV);
966	return (error);
967}
968
969struct linux_recvmsg_args {
970	int s;
971	struct msghdr *msg;
972	int flags;
973};
974
975static int
976linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
977{
978	struct linux_recvmsg_args linux_args;
979	struct recvmsg_args /* {
980		int	s;
981		struct	msghdr *msg;
982		int	flags;
983	} */ bsd_args;
984	struct msghdr msg;
985	struct cmsghdr *cmsg;
986	int error;
987
988	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
989		return (error);
990
991	bsd_args.s = linux_args.s;
992	bsd_args.msg = linux_args.msg;
993	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
994	error = recvmsg(td, &bsd_args);
995	if (error)
996		return (error);
997
998	if (bsd_args.msg->msg_control != NULL) {
999		cmsg = (struct cmsghdr*)bsd_args.msg->msg_control;
1000		cmsg->cmsg_level = bsd_to_linux_sockopt_level(cmsg->cmsg_level);
1001	}
1002
1003	error = copyin(linux_args.msg, &msg, sizeof(msg));
1004	if (error)
1005		return (error);
1006	if (msg.msg_name && msg.msg_namelen > 2)
1007		error = linux_sa_put(msg.msg_name);
1008	return (error);
1009}
1010
1011struct linux_shutdown_args {
1012	int s;
1013	int how;
1014};
1015
1016static int
1017linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1018{
1019	struct linux_shutdown_args linux_args;
1020	struct shutdown_args /* {
1021		int s;
1022		int how;
1023	} */ bsd_args;
1024	int error;
1025
1026	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1027		return (error);
1028
1029	bsd_args.s = linux_args.s;
1030	bsd_args.how = linux_args.how;
1031	return (shutdown(td, &bsd_args));
1032}
1033
1034struct linux_setsockopt_args {
1035	int s;
1036	int level;
1037	int optname;
1038	void *optval;
1039	int optlen;
1040};
1041
1042static int
1043linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1044{
1045	struct linux_setsockopt_args linux_args;
1046	struct setsockopt_args /* {
1047		int s;
1048		int level;
1049		int name;
1050		caddr_t val;
1051		int valsize;
1052	} */ bsd_args;
1053	int error, name;
1054
1055	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1056		return (error);
1057
1058	bsd_args.s = linux_args.s;
1059	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1060	switch (bsd_args.level) {
1061	case SOL_SOCKET:
1062		name = linux_to_bsd_so_sockopt(linux_args.optname);
1063		break;
1064	case IPPROTO_IP:
1065		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1066		break;
1067	case IPPROTO_TCP:
1068		/* Linux TCP option values match BSD's */
1069		name = linux_args.optname;
1070		break;
1071	default:
1072		name = -1;
1073		break;
1074	}
1075	if (name == -1)
1076		return (EINVAL);
1077
1078	bsd_args.name = name;
1079	bsd_args.val = linux_args.optval;
1080	bsd_args.valsize = linux_args.optlen;
1081	return (setsockopt(td, &bsd_args));
1082}
1083
1084struct linux_getsockopt_args {
1085	int s;
1086	int level;
1087	int optname;
1088	void *optval;
1089	int *optlen;
1090};
1091
1092static int
1093linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1094{
1095	struct linux_getsockopt_args linux_args;
1096	struct getsockopt_args /* {
1097		int s;
1098		int level;
1099		int name;
1100		caddr_t val;
1101		int *avalsize;
1102	} */ bsd_args;
1103	int error, name;
1104
1105	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1106		return (error);
1107
1108	bsd_args.s = linux_args.s;
1109	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1110	switch (bsd_args.level) {
1111	case SOL_SOCKET:
1112		name = linux_to_bsd_so_sockopt(linux_args.optname);
1113		break;
1114	case IPPROTO_IP:
1115		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1116		break;
1117	case IPPROTO_TCP:
1118		/* Linux TCP option values match BSD's */
1119		name = linux_args.optname;
1120		break;
1121	default:
1122		name = -1;
1123		break;
1124	}
1125	if (name == -1)
1126		return (EINVAL);
1127
1128	bsd_args.name = name;
1129	bsd_args.val = linux_args.optval;
1130	bsd_args.avalsize = linux_args.optlen;
1131	return (getsockopt(td, &bsd_args));
1132}
1133
1134int
1135linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1136{
1137	void *arg = (void *)args->args;
1138
1139	switch (args->what) {
1140	case LINUX_SOCKET:
1141		return (linux_socket(td, arg));
1142	case LINUX_BIND:
1143		return (linux_bind(td, arg));
1144	case LINUX_CONNECT:
1145		return (linux_connect(td, arg));
1146	case LINUX_LISTEN:
1147		return (linux_listen(td, arg));
1148	case LINUX_ACCEPT:
1149		return (linux_accept(td, arg));
1150	case LINUX_GETSOCKNAME:
1151		return (linux_getsockname(td, arg));
1152	case LINUX_GETPEERNAME:
1153		return (linux_getpeername(td, arg));
1154	case LINUX_SOCKETPAIR:
1155		return (linux_socketpair(td, arg));
1156	case LINUX_SEND:
1157		return (linux_send(td, arg));
1158	case LINUX_RECV:
1159		return (linux_recv(td, arg));
1160	case LINUX_SENDTO:
1161		return (linux_sendto(td, arg));
1162	case LINUX_RECVFROM:
1163		return (linux_recvfrom(td, arg));
1164	case LINUX_SHUTDOWN:
1165		return (linux_shutdown(td, arg));
1166	case LINUX_SETSOCKOPT:
1167		return (linux_setsockopt(td, arg));
1168	case LINUX_GETSOCKOPT:
1169		return (linux_getsockopt(td, arg));
1170	case LINUX_SENDMSG:
1171		return (linux_sendmsg(td, arg));
1172	case LINUX_RECVMSG:
1173		return (linux_recvmsg(td, arg));
1174	}
1175
1176	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1177	return (ENOSYS);
1178}
1179#endif	/*!__alpha__*/
1180