linux_socket.c revision 150335
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 150335 2005-09-19 16:51:43Z rwatson $");
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/lock.h>
48#include <sys/malloc.h>
49#include <sys/mutex.h>
50#include <sys/mbuf.h>
51#include <sys/socket.h>
52#include <sys/socketvar.h>
53#include <sys/syscallsubr.h>
54#include <sys/uio.h>
55#include <sys/syslog.h>
56
57#include <netinet/in.h>
58#include <netinet/in_systm.h>
59#include <netinet/ip.h>
60#ifdef INET6
61#include <netinet/ip6.h>
62#include <netinet6/ip6_var.h>
63#endif
64
65#include "opt_compat.h"
66
67#ifdef COMPAT_LINUX32
68#include <machine/../linux32/linux.h>
69#include <machine/../linux32/linux32_proto.h>
70#else
71#include <machine/../linux/linux.h>
72#include <machine/../linux/linux_proto.h>
73#endif
74#include <compat/linux/linux_socket.h>
75#include <compat/linux/linux_util.h>
76
77static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
78    struct malloc_type *);
79static int linux_to_bsd_domain(int);
80
81/*
82 * Reads a linux sockaddr and does any necessary translation.
83 * Linux sockaddrs don't have a length field, only a family.
84 */
85static int
86linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
87{
88	int osalen = len;
89
90	return (do_sa_get(sap, osa, &osalen, M_SONAME));
91}
92
93/*
94 * Copy the osockaddr structure pointed to by osa to kernel, adjust
95 * family and convert to sockaddr.
96 */
97static int
98do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
99    struct malloc_type *mtype)
100{
101	int error=0, bdom;
102	struct sockaddr *sa;
103	struct osockaddr *kosa;
104	int alloclen;
105#ifdef INET6
106	int oldv6size;
107	struct sockaddr_in6 *sin6;
108#endif
109
110	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
111		return (EINVAL);
112
113	alloclen = *osalen;
114#ifdef INET6
115	oldv6size = 0;
116	/*
117	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
118	 * if it's a v4-mapped address, so reserve the proper space
119	 * for it.
120	 */
121	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
122		alloclen = sizeof (struct sockaddr_in6);
123		oldv6size = 1;
124	}
125#endif
126
127	MALLOC(kosa, struct osockaddr *, alloclen, mtype, M_WAITOK);
128
129	if ((error = copyin(osa, kosa, *osalen)))
130		goto out;
131
132	bdom = linux_to_bsd_domain(kosa->sa_family);
133	if (bdom == -1) {
134		error = EINVAL;
135		goto out;
136	}
137
138#ifdef INET6
139	/*
140	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
141	 * which lacks the scope id compared with RFC2553 one. If we detect
142	 * the situation, reject the address and write a message to system log.
143	 *
144	 * Still accept addresses for which the scope id is not used.
145	 */
146	if (oldv6size && bdom == AF_INET6) {
147		sin6 = (struct sockaddr_in6 *)kosa;
148		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
149		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
150		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
151		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
152		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
153		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
154			sin6->sin6_scope_id = 0;
155		} else {
156			log(LOG_DEBUG,
157			    "obsolete pre-RFC2553 sockaddr_in6 rejected");
158			error = EINVAL;
159			goto out;
160		}
161	} else
162#endif
163	if (bdom == AF_INET)
164		alloclen = sizeof(struct sockaddr_in);
165
166	sa = (struct sockaddr *) kosa;
167	sa->sa_family = bdom;
168	sa->sa_len = alloclen;
169
170	*sap = sa;
171	*osalen = alloclen;
172	return (0);
173
174out:
175	FREE(kosa, mtype);
176	return (error);
177}
178
179static int
180linux_to_bsd_domain(int domain)
181{
182
183	switch (domain) {
184	case LINUX_AF_UNSPEC:
185		return (AF_UNSPEC);
186	case LINUX_AF_UNIX:
187		return (AF_LOCAL);
188	case LINUX_AF_INET:
189		return (AF_INET);
190	case LINUX_AF_INET6:
191		return (AF_INET6);
192	case LINUX_AF_AX25:
193		return (AF_CCITT);
194	case LINUX_AF_IPX:
195		return (AF_IPX);
196	case LINUX_AF_APPLETALK:
197		return (AF_APPLETALK);
198	}
199	return (-1);
200}
201
202#ifndef __alpha__
203static int
204bsd_to_linux_domain(int domain)
205{
206
207	switch (domain) {
208	case AF_UNSPEC:
209		return (LINUX_AF_UNSPEC);
210	case AF_LOCAL:
211		return (LINUX_AF_UNIX);
212	case AF_INET:
213		return (LINUX_AF_INET);
214	case AF_INET6:
215		return (LINUX_AF_INET6);
216	case AF_CCITT:
217		return (LINUX_AF_AX25);
218	case AF_IPX:
219		return (LINUX_AF_IPX);
220	case AF_APPLETALK:
221		return (LINUX_AF_APPLETALK);
222	}
223	return (-1);
224}
225
226static int
227linux_to_bsd_sockopt_level(int level)
228{
229
230	switch (level) {
231	case LINUX_SOL_SOCKET:
232		return (SOL_SOCKET);
233	}
234	return (level);
235}
236
237static int
238bsd_to_linux_sockopt_level(int level)
239{
240
241	switch (level) {
242	case SOL_SOCKET:
243		return (LINUX_SOL_SOCKET);
244	}
245	return (level);
246}
247
248static int
249linux_to_bsd_ip_sockopt(int opt)
250{
251
252	switch (opt) {
253	case LINUX_IP_TOS:
254		return (IP_TOS);
255	case LINUX_IP_TTL:
256		return (IP_TTL);
257	case LINUX_IP_OPTIONS:
258		return (IP_OPTIONS);
259	case LINUX_IP_MULTICAST_IF:
260		return (IP_MULTICAST_IF);
261	case LINUX_IP_MULTICAST_TTL:
262		return (IP_MULTICAST_TTL);
263	case LINUX_IP_MULTICAST_LOOP:
264		return (IP_MULTICAST_LOOP);
265	case LINUX_IP_ADD_MEMBERSHIP:
266		return (IP_ADD_MEMBERSHIP);
267	case LINUX_IP_DROP_MEMBERSHIP:
268		return (IP_DROP_MEMBERSHIP);
269	case LINUX_IP_HDRINCL:
270		return (IP_HDRINCL);
271	}
272	return (-1);
273}
274
275static int
276linux_to_bsd_so_sockopt(int opt)
277{
278
279	switch (opt) {
280	case LINUX_SO_DEBUG:
281		return (SO_DEBUG);
282	case LINUX_SO_REUSEADDR:
283		return (SO_REUSEADDR);
284	case LINUX_SO_TYPE:
285		return (SO_TYPE);
286	case LINUX_SO_ERROR:
287		return (SO_ERROR);
288	case LINUX_SO_DONTROUTE:
289		return (SO_DONTROUTE);
290	case LINUX_SO_BROADCAST:
291		return (SO_BROADCAST);
292	case LINUX_SO_SNDBUF:
293		return (SO_SNDBUF);
294	case LINUX_SO_RCVBUF:
295		return (SO_RCVBUF);
296	case LINUX_SO_KEEPALIVE:
297		return (SO_KEEPALIVE);
298	case LINUX_SO_OOBINLINE:
299		return (SO_OOBINLINE);
300	case LINUX_SO_LINGER:
301		return (SO_LINGER);
302	}
303	return (-1);
304}
305
306static int
307linux_to_bsd_msg_flags(int flags)
308{
309	int ret_flags = 0;
310
311	if (flags & LINUX_MSG_OOB)
312		ret_flags |= MSG_OOB;
313	if (flags & LINUX_MSG_PEEK)
314		ret_flags |= MSG_PEEK;
315	if (flags & LINUX_MSG_DONTROUTE)
316		ret_flags |= MSG_DONTROUTE;
317	if (flags & LINUX_MSG_CTRUNC)
318		ret_flags |= MSG_CTRUNC;
319	if (flags & LINUX_MSG_TRUNC)
320		ret_flags |= MSG_TRUNC;
321	if (flags & LINUX_MSG_DONTWAIT)
322		ret_flags |= MSG_DONTWAIT;
323	if (flags & LINUX_MSG_EOR)
324		ret_flags |= MSG_EOR;
325	if (flags & LINUX_MSG_WAITALL)
326		ret_flags |= MSG_WAITALL;
327	if (flags & LINUX_MSG_NOSIGNAL)
328		ret_flags |= MSG_NOSIGNAL;
329#if 0 /* not handled */
330	if (flags & LINUX_MSG_PROXY)
331		;
332	if (flags & LINUX_MSG_FIN)
333		;
334	if (flags & LINUX_MSG_SYN)
335		;
336	if (flags & LINUX_MSG_CONFIRM)
337		;
338	if (flags & LINUX_MSG_RST)
339		;
340	if (flags & LINUX_MSG_ERRQUEUE)
341		;
342#endif
343	return ret_flags;
344}
345
346static int
347linux_sa_put(struct osockaddr *osa)
348{
349	struct osockaddr sa;
350	int error, bdom;
351
352	/*
353	 * Only read/write the osockaddr family part, the rest is
354	 * not changed.
355	 */
356	error = copyin(osa, &sa, sizeof(sa.sa_family));
357	if (error)
358		return (error);
359
360	bdom = bsd_to_linux_domain(sa.sa_family);
361	if (bdom == -1)
362		return (EINVAL);
363
364	sa.sa_family = bdom;
365	error = copyout(&sa, osa, sizeof(sa.sa_family));
366	if (error)
367		return (error);
368
369	return (0);
370}
371
372static int
373linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
374    enum uio_seg segflg)
375{
376	struct mbuf *control;
377	struct sockaddr *to;
378	int error;
379
380	if (mp->msg_name != NULL) {
381		error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
382		if (error)
383			return (error);
384		mp->msg_name = to;
385	} else
386		to = NULL;
387
388	if (mp->msg_control != NULL) {
389		struct cmsghdr *cmsg;
390
391		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
392			error = EINVAL;
393			goto bad;
394		}
395		error = sockargs(&control, mp->msg_control,
396		    mp->msg_controllen, MT_CONTROL);
397		if (error)
398			goto bad;
399
400		cmsg = mtod(control, struct cmsghdr *);
401		cmsg->cmsg_level = linux_to_bsd_sockopt_level(cmsg->cmsg_level);
402	} else
403		control = NULL;
404
405	error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
406	    segflg);
407
408bad:
409	if (to)
410		FREE(to, M_SONAME);
411	return (error);
412}
413
414/* Return 0 if IP_HDRINCL is set for the given socket. */
415static int
416linux_check_hdrincl(struct thread *td, int s)
417{
418	int error, optval, size_val;
419
420	size_val = sizeof(optval);
421	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
422	    &optval, UIO_SYSSPACE, &size_val);
423	if (error)
424		return (error);
425
426	return (optval == 0);
427}
428
429struct linux_sendto_args {
430	int s;
431	l_uintptr_t msg;
432	int len;
433	int flags;
434	l_uintptr_t to;
435	int tolen;
436};
437
438/*
439 * Updated sendto() when IP_HDRINCL is set:
440 * tweak endian-dependent fields in the IP packet.
441 */
442static int
443linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
444{
445/*
446 * linux_ip_copysize defines how many bytes we should copy
447 * from the beginning of the IP packet before we customize it for BSD.
448 * It should include all the fields we modify (ip_len and ip_off).
449 */
450#define linux_ip_copysize	8
451
452	struct ip *packet;
453	struct msghdr msg;
454	struct iovec aiov[1];
455	int error;
456
457	/* Check that the packet isn't too big or too small. */
458	if (linux_args->len < linux_ip_copysize ||
459	    linux_args->len > IP_MAXPACKET)
460		return (EINVAL);
461
462	packet = (struct ip *)malloc(linux_args->len, M_TEMP, M_WAITOK);
463
464	/* Make kernel copy of the packet to be sent */
465	if ((error = copyin(PTRIN(linux_args->msg), packet,
466	    linux_args->len)))
467		goto goout;
468
469	/* Convert fields from Linux to BSD raw IP socket format */
470	packet->ip_len = linux_args->len;
471	packet->ip_off = ntohs(packet->ip_off);
472
473	/* Prepare the msghdr and iovec structures describing the new packet */
474	msg.msg_name = PTRIN(linux_args->to);
475	msg.msg_namelen = linux_args->tolen;
476	msg.msg_iov = aiov;
477	msg.msg_iovlen = 1;
478	msg.msg_control = NULL;
479	msg.msg_flags = 0;
480	aiov[0].iov_base = (char *)packet;
481	aiov[0].iov_len = linux_args->len;
482	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
483	    UIO_SYSSPACE);
484goout:
485	free(packet, M_TEMP);
486	return (error);
487}
488
489struct linux_socket_args {
490	int domain;
491	int type;
492	int protocol;
493};
494
495static int
496linux_socket(struct thread *td, struct linux_socket_args *args)
497{
498	struct linux_socket_args linux_args;
499	struct socket_args /* {
500		int domain;
501		int type;
502		int protocol;
503	} */ bsd_args;
504	int error;
505	int retval_socket;
506
507	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
508		return (error);
509
510	bsd_args.protocol = linux_args.protocol;
511	bsd_args.type = linux_args.type;
512	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
513	if (bsd_args.domain == -1)
514		return (EINVAL);
515
516	retval_socket = socket(td, &bsd_args);
517	if (bsd_args.type == SOCK_RAW
518	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
519	    && bsd_args.domain == AF_INET
520	    && retval_socket >= 0) {
521		/* It's a raw IP socket: set the IP_HDRINCL option. */
522		int hdrincl;
523
524		hdrincl = 1;
525		/* We ignore any error returned by kern_setsockopt() */
526		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
527		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
528	}
529#ifdef INET6
530	/*
531	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
532	 * default and some apps depend on this. So, set V6ONLY to 0
533	 * for Linux apps if the sysctl value is set to 1.
534	 */
535	if (bsd_args.domain == PF_INET6 && retval_socket >= 0
536#ifndef KLD_MODULE
537	    /*
538	     * XXX: Avoid undefined symbol error with an IPv4 only
539	     * kernel.
540	     */
541	    && ip6_v6only
542#endif
543	    ) {
544		int v6only;
545
546		v6only = 0;
547		/* We ignore any error returned by setsockopt() */
548		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
549		    &v6only, UIO_SYSSPACE, sizeof(v6only));
550	}
551#endif
552
553	return (retval_socket);
554}
555
556struct linux_bind_args {
557	int s;
558	l_uintptr_t name;
559	int namelen;
560};
561
562static int
563linux_bind(struct thread *td, struct linux_bind_args *args)
564{
565	struct linux_bind_args linux_args;
566	struct sockaddr *sa;
567	int error;
568
569	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
570		return (error);
571
572	error = linux_getsockaddr(&sa, PTRIN(linux_args.name),
573	    linux_args.namelen);
574	if (error)
575		return (error);
576
577	return (kern_bind(td, linux_args.s, sa));
578}
579
580struct linux_connect_args {
581	int s;
582	l_uintptr_t name;
583	int namelen;
584};
585int linux_connect(struct thread *, struct linux_connect_args *);
586#endif /* !__alpha__*/
587
588int
589linux_connect(struct thread *td, struct linux_connect_args *args)
590{
591	struct linux_connect_args linux_args;
592	struct socket *so;
593	struct sockaddr *sa;
594	u_int fflag;
595	int error;
596
597#ifdef __alpha__
598	bcopy(args, &linux_args, sizeof(linux_args));
599#else
600	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
601		return (error);
602#endif /* __alpha__ */
603
604	error = linux_getsockaddr(&sa,
605	    (struct osockaddr *)PTRIN(linux_args.name),
606	    linux_args.namelen);
607	if (error)
608		return (error);
609
610	error = kern_connect(td, linux_args.s, sa);
611	if (error != EISCONN)
612		return (error);
613
614	/*
615	 * Linux doesn't return EISCONN the first time it occurs,
616	 * when on a non-blocking socket. Instead it returns the
617	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
618	 */
619	NET_LOCK_GIANT();
620	error = fgetsock(td, linux_args.s, &so, &fflag);
621	if (error == 0) {
622		error = EISCONN;
623		if (fflag & FNONBLOCK) {
624			SOCK_LOCK(so);
625			if (so->so_emuldata == 0)
626				error = so->so_error;
627			so->so_emuldata = (void *)1;
628			SOCK_UNLOCK(so);
629		}
630		fputsock(so);
631	}
632	NET_UNLOCK_GIANT();
633	return (error);
634}
635
636#ifndef __alpha__
637
638struct linux_listen_args {
639	int s;
640	int backlog;
641};
642
643static int
644linux_listen(struct thread *td, struct linux_listen_args *args)
645{
646	struct linux_listen_args linux_args;
647	struct listen_args /* {
648		int s;
649		int backlog;
650	} */ bsd_args;
651	int error;
652
653	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
654		return (error);
655
656	bsd_args.s = linux_args.s;
657	bsd_args.backlog = linux_args.backlog;
658	return (listen(td, &bsd_args));
659}
660
661struct linux_accept_args {
662	int s;
663	l_uintptr_t addr;
664	l_uintptr_t namelen;
665};
666
667static int
668linux_accept(struct thread *td, struct linux_accept_args *args)
669{
670	struct linux_accept_args linux_args;
671	struct accept_args /* {
672		int	s;
673		struct sockaddr * __restrict name;
674		socklen_t * __restrict anamelen;
675	} */ bsd_args;
676	struct close_args /* {
677		int     fd;
678	} */ c_args;
679	int error, fd;
680
681	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
682		return (error);
683
684	bsd_args.s = linux_args.s;
685	/* XXX: */
686	bsd_args.name = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
687	bsd_args.anamelen = PTRIN(linux_args.namelen);/* XXX */
688	error = oaccept(td, &bsd_args);
689	if (error)
690		return (error);
691	if (linux_args.addr) {
692		error = linux_sa_put(PTRIN(linux_args.addr));
693		if (error) {
694			c_args.fd = td->td_retval[0];
695			(void)close(td, &c_args);
696			return (error);
697		}
698	}
699
700	/*
701	 * linux appears not to copy flags from the parent socket to the
702	 * accepted one, so we must clear the flags in the new descriptor.
703	 * Ignore any errors, because we already have an open fd.
704	 */
705	fd = td->td_retval[0];
706	(void)kern_fcntl(td, fd, F_SETFL, 0);
707	td->td_retval[0] = fd;
708	return (0);
709}
710
711struct linux_getsockname_args {
712	int s;
713	l_uintptr_t addr;
714	l_uintptr_t namelen;
715};
716
717static int
718linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
719{
720	struct linux_getsockname_args linux_args;
721	struct getsockname_args /* {
722		int	fdes;
723		struct sockaddr * __restrict asa;
724		socklen_t * __restrict alen;
725	} */ bsd_args;
726	int error;
727
728	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
729		return (error);
730
731	bsd_args.fdes = linux_args.s;
732	/* XXX: */
733	bsd_args.asa = (struct sockaddr * __restrict)PTRIN(linux_args.addr);
734	bsd_args.alen = PTRIN(linux_args.namelen);	/* XXX */
735	error = ogetsockname(td, &bsd_args);
736	if (error)
737		return (error);
738	error = linux_sa_put(PTRIN(linux_args.addr));
739	if (error)
740		return (error);
741	return (0);
742}
743
744struct linux_getpeername_args {
745	int s;
746	l_uintptr_t addr;
747	l_uintptr_t namelen;
748};
749
750static int
751linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
752{
753	struct linux_getpeername_args linux_args;
754	struct ogetpeername_args /* {
755		int fdes;
756		caddr_t asa;
757		int *alen;
758	} */ bsd_args;
759	int error;
760
761	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
762		return (error);
763
764	bsd_args.fdes = linux_args.s;
765	bsd_args.asa = (caddr_t)PTRIN(linux_args.addr);
766	bsd_args.alen = (int *)PTRIN(linux_args.namelen);
767	error = ogetpeername(td, &bsd_args);
768	if (error)
769		return (error);
770	error = linux_sa_put(PTRIN(linux_args.addr));
771	if (error)
772		return (error);
773	return (0);
774}
775
776struct linux_socketpair_args {
777	int domain;
778	int type;
779	int protocol;
780	l_uintptr_t rsv;
781};
782
783static int
784linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
785{
786	struct linux_socketpair_args linux_args;
787	struct socketpair_args /* {
788		int domain;
789		int type;
790		int protocol;
791		int *rsv;
792	} */ bsd_args;
793	int error;
794
795	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
796		return (error);
797
798	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
799	if (bsd_args.domain == -1)
800		return (EINVAL);
801
802	bsd_args.type = linux_args.type;
803	bsd_args.protocol = linux_args.protocol;
804	bsd_args.rsv = (int *)PTRIN(linux_args.rsv);
805	return (socketpair(td, &bsd_args));
806}
807
808struct linux_send_args {
809	int s;
810	l_uintptr_t msg;
811	int len;
812	int flags;
813};
814
815static int
816linux_send(struct thread *td, struct linux_send_args *args)
817{
818	struct linux_send_args linux_args;
819	struct sendto_args /* {
820		int s;
821		caddr_t buf;
822		int len;
823		int flags;
824		caddr_t to;
825		int tolen;
826	} */ bsd_args;
827	int error;
828
829	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
830		return (error);
831
832	bsd_args.s = linux_args.s;
833	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
834	bsd_args.len = linux_args.len;
835	bsd_args.flags = linux_args.flags;
836	bsd_args.to = NULL;
837	bsd_args.tolen = 0;
838	return sendto(td, &bsd_args);
839}
840
841struct linux_recv_args {
842	int s;
843	l_uintptr_t msg;
844	int len;
845	int flags;
846};
847
848static int
849linux_recv(struct thread *td, struct linux_recv_args *args)
850{
851	struct linux_recv_args linux_args;
852	struct recvfrom_args /* {
853		int s;
854		caddr_t buf;
855		int len;
856		int flags;
857		struct sockaddr *from;
858		socklen_t fromlenaddr;
859	} */ bsd_args;
860	int error;
861
862	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
863		return (error);
864
865	bsd_args.s = linux_args.s;
866	bsd_args.buf = (caddr_t)PTRIN(linux_args.msg);
867	bsd_args.len = linux_args.len;
868	bsd_args.flags = linux_args.flags;
869	bsd_args.from = NULL;
870	bsd_args.fromlenaddr = 0;
871	return (recvfrom(td, &bsd_args));
872}
873
874static int
875linux_sendto(struct thread *td, struct linux_sendto_args *args)
876{
877	struct linux_sendto_args linux_args;
878	struct msghdr msg;
879	struct iovec aiov;
880	int error;
881
882	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
883		return (error);
884
885	if (linux_check_hdrincl(td, linux_args.s) == 0)
886		/* IP_HDRINCL set, tweak the packet before sending */
887		return (linux_sendto_hdrincl(td, &linux_args));
888
889	msg.msg_name = PTRIN(linux_args.to);
890	msg.msg_namelen = linux_args.tolen;
891	msg.msg_iov = &aiov;
892	msg.msg_iovlen = 1;
893	msg.msg_control = NULL;
894	msg.msg_flags = 0;
895	aiov.iov_base = PTRIN(linux_args.msg);
896	aiov.iov_len = linux_args.len;
897	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
898	    UIO_USERSPACE);
899	return (error);
900}
901
902struct linux_recvfrom_args {
903	int s;
904	l_uintptr_t buf;
905	int len;
906	int flags;
907	l_uintptr_t from;
908	l_uintptr_t fromlen;
909};
910
911static int
912linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
913{
914	struct linux_recvfrom_args linux_args;
915	struct recvfrom_args /* {
916		int	s;
917		caddr_t	buf;
918		size_t	len;
919		int	flags;
920		struct sockaddr * __restrict from;
921		socklen_t * __restrict fromlenaddr;
922	} */ bsd_args;
923	int error;
924
925	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
926		return (error);
927
928	bsd_args.s = linux_args.s;
929	bsd_args.buf = PTRIN(linux_args.buf);
930	bsd_args.len = linux_args.len;
931	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
932	/* XXX: */
933	bsd_args.from = (struct sockaddr * __restrict)PTRIN(linux_args.from);
934	bsd_args.fromlenaddr = PTRIN(linux_args.fromlen);/* XXX */
935	error = orecvfrom(td, &bsd_args);
936	if (error)
937		return (error);
938	if (linux_args.from) {
939		error = linux_sa_put((struct osockaddr *)
940		    PTRIN(linux_args.from));
941		if (error)
942			return (error);
943	}
944	return (0);
945}
946
947struct linux_sendmsg_args {
948	int s;
949	l_uintptr_t msg;
950	int flags;
951};
952
953static int
954linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
955{
956	struct linux_sendmsg_args linux_args;
957	struct msghdr msg;
958	struct iovec *iov;
959	int error;
960
961	/* XXXTJR sendmsg is broken on amd64 */
962
963	error = copyin(args, &linux_args, sizeof(linux_args));
964	if (error)
965		return (error);
966	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
967	if (error)
968		return (error);
969	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
970	if (error)
971		return (error);
972	msg.msg_iov = iov;
973	msg.msg_flags = 0;
974	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags,
975	    UIO_USERSPACE);
976	free(iov, M_IOV);
977	return (error);
978}
979
980struct linux_recvmsg_args {
981	int s;
982	l_uintptr_t msg;
983	int flags;
984};
985
986static int
987linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
988{
989	struct linux_recvmsg_args linux_args;
990	struct recvmsg_args /* {
991		int	s;
992		struct	msghdr *msg;
993		int	flags;
994	} */ bsd_args;
995	struct msghdr msg;
996	struct cmsghdr *cmsg;
997	int error;
998
999	/* XXXTJR recvmsg is broken on amd64 */
1000
1001	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1002		return (error);
1003
1004	bsd_args.s = linux_args.s;
1005	bsd_args.msg = PTRIN(linux_args.msg);
1006	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
1007	error = recvmsg(td, &bsd_args);
1008	if (error)
1009		return (error);
1010
1011	if (bsd_args.msg->msg_control != NULL &&
1012	    bsd_args.msg->msg_controllen > 0) {
1013		cmsg = (struct cmsghdr*)bsd_args.msg->msg_control;
1014		cmsg->cmsg_level = bsd_to_linux_sockopt_level(cmsg->cmsg_level);
1015	}
1016
1017	error = copyin(PTRIN(linux_args.msg), &msg, sizeof(msg));
1018	if (error)
1019		return (error);
1020	if (msg.msg_name && msg.msg_namelen > 2)
1021		error = linux_sa_put(msg.msg_name);
1022	return (error);
1023}
1024
1025struct linux_shutdown_args {
1026	int s;
1027	int how;
1028};
1029
1030static int
1031linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1032{
1033	struct linux_shutdown_args linux_args;
1034	struct shutdown_args /* {
1035		int s;
1036		int how;
1037	} */ bsd_args;
1038	int error;
1039
1040	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1041		return (error);
1042
1043	bsd_args.s = linux_args.s;
1044	bsd_args.how = linux_args.how;
1045	return (shutdown(td, &bsd_args));
1046}
1047
1048struct linux_setsockopt_args {
1049	int s;
1050	int level;
1051	int optname;
1052	l_uintptr_t optval;
1053	int optlen;
1054};
1055
1056static int
1057linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1058{
1059	struct linux_setsockopt_args linux_args;
1060	struct setsockopt_args /* {
1061		int s;
1062		int level;
1063		int name;
1064		caddr_t val;
1065		int valsize;
1066	} */ bsd_args;
1067	int error, name;
1068
1069	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1070		return (error);
1071
1072	bsd_args.s = linux_args.s;
1073	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1074	switch (bsd_args.level) {
1075	case SOL_SOCKET:
1076		name = linux_to_bsd_so_sockopt(linux_args.optname);
1077		break;
1078	case IPPROTO_IP:
1079		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1080		break;
1081	case IPPROTO_TCP:
1082		/* Linux TCP option values match BSD's */
1083		name = linux_args.optname;
1084		break;
1085	default:
1086		name = -1;
1087		break;
1088	}
1089	if (name == -1)
1090		return (EINVAL);
1091
1092	bsd_args.name = name;
1093	bsd_args.val = PTRIN(linux_args.optval);
1094	bsd_args.valsize = linux_args.optlen;
1095	return (setsockopt(td, &bsd_args));
1096}
1097
1098struct linux_getsockopt_args {
1099	int s;
1100	int level;
1101	int optname;
1102	l_uintptr_t optval;
1103	l_uintptr_t optlen;
1104};
1105
1106static int
1107linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1108{
1109	struct linux_getsockopt_args linux_args;
1110	struct getsockopt_args /* {
1111		int s;
1112		int level;
1113		int name;
1114		caddr_t val;
1115		int *avalsize;
1116	} */ bsd_args;
1117	int error, name;
1118
1119	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1120		return (error);
1121
1122	bsd_args.s = linux_args.s;
1123	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1124	switch (bsd_args.level) {
1125	case SOL_SOCKET:
1126		name = linux_to_bsd_so_sockopt(linux_args.optname);
1127		break;
1128	case IPPROTO_IP:
1129		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1130		break;
1131	case IPPROTO_TCP:
1132		/* Linux TCP option values match BSD's */
1133		name = linux_args.optname;
1134		break;
1135	default:
1136		name = -1;
1137		break;
1138	}
1139	if (name == -1)
1140		return (EINVAL);
1141
1142	bsd_args.name = name;
1143	bsd_args.val = PTRIN(linux_args.optval);
1144	bsd_args.avalsize = PTRIN(linux_args.optlen);
1145	return (getsockopt(td, &bsd_args));
1146}
1147
1148int
1149linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1150{
1151	void *arg = (void *)(intptr_t)args->args;
1152
1153	switch (args->what) {
1154	case LINUX_SOCKET:
1155		return (linux_socket(td, arg));
1156	case LINUX_BIND:
1157		return (linux_bind(td, arg));
1158	case LINUX_CONNECT:
1159		return (linux_connect(td, arg));
1160	case LINUX_LISTEN:
1161		return (linux_listen(td, arg));
1162	case LINUX_ACCEPT:
1163		return (linux_accept(td, arg));
1164	case LINUX_GETSOCKNAME:
1165		return (linux_getsockname(td, arg));
1166	case LINUX_GETPEERNAME:
1167		return (linux_getpeername(td, arg));
1168	case LINUX_SOCKETPAIR:
1169		return (linux_socketpair(td, arg));
1170	case LINUX_SEND:
1171		return (linux_send(td, arg));
1172	case LINUX_RECV:
1173		return (linux_recv(td, arg));
1174	case LINUX_SENDTO:
1175		return (linux_sendto(td, arg));
1176	case LINUX_RECVFROM:
1177		return (linux_recvfrom(td, arg));
1178	case LINUX_SHUTDOWN:
1179		return (linux_shutdown(td, arg));
1180	case LINUX_SETSOCKOPT:
1181		return (linux_setsockopt(td, arg));
1182	case LINUX_GETSOCKOPT:
1183		return (linux_getsockopt(td, arg));
1184	case LINUX_SENDMSG:
1185		return (linux_sendmsg(td, arg));
1186	case LINUX_RECVMSG:
1187		return (linux_recvmsg(td, arg));
1188	}
1189
1190	mtx_lock(&Giant);
1191	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1192	mtx_unlock(&Giant);
1193	return (ENOSYS);
1194}
1195#endif	/*!__alpha__*/
1196