linux_socket.c revision 131897
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 131897 2004-07-10 15:42:16Z phk $");
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, caddr_t *sg, int s)
406{
407	struct getsockopt_args /* {
408		int	s;
409		int	level;
410		int	name;
411		void * __restrict val;
412		socklen_t * __restrict avalsize;
413	} */ bsd_args;
414	void * __restrict val;
415	socklen_t * __restrict valsize;
416	int error, optval, size_val;
417
418	val = stackgap_alloc(sg, sizeof(size_val));
419	valsize = stackgap_alloc(sg, sizeof(socklen_t));
420
421	size_val = sizeof(val);
422	if ((error = copyout(&size_val, valsize, sizeof(size_val))))
423		return (error);
424
425	bsd_args.s = s;
426	bsd_args.level = IPPROTO_IP;
427	bsd_args.name = IP_HDRINCL;
428	bsd_args.val = val;
429	bsd_args.avalsize = valsize;
430	if ((error = getsockopt(td, &bsd_args)))
431		return (error);
432
433	if ((error = copyin(val, &optval, sizeof(optval))))
434		return (error);
435
436	return (optval == 0);
437}
438
439struct linux_sendto_args {
440	int s;
441	void *msg;
442	int len;
443	int flags;
444	caddr_t to;
445	int tolen;
446};
447
448/*
449 * Updated sendto() when IP_HDRINCL is set:
450 * tweak endian-dependent fields in the IP packet.
451 */
452static int
453linux_sendto_hdrincl(struct thread *td, caddr_t *sg, struct linux_sendto_args *linux_args)
454{
455/*
456 * linux_ip_copysize defines how many bytes we should copy
457 * from the beginning of the IP packet before we customize it for BSD.
458 * It should include all the fields we modify (ip_len and ip_off)
459 * and be as small as possible to minimize copying overhead.
460 */
461#define linux_ip_copysize	8
462
463	struct ip *packet;
464	struct msghdr msg;
465	struct iovec aiov[2];
466	int error;
467
468	/* Check the packet isn't too small before we mess with it */
469	if (linux_args->len < linux_ip_copysize)
470		return (EINVAL);
471
472	/*
473	 * Tweaking the user buffer in place would be bad manners.
474	 * We create a corrected IP header with just the needed length,
475	 * then use an iovec to glue it to the rest of the user packet
476	 * when calling sendit().
477	 */
478	packet = (struct ip *)stackgap_alloc(sg, linux_ip_copysize);
479
480	/* Make a copy of the beginning of the packet to be sent */
481	if ((error = copyin(linux_args->msg, packet, linux_ip_copysize)))
482		return (error);
483
484	/* Convert fields from Linux to BSD raw IP socket format */
485	packet->ip_len = linux_args->len;
486	packet->ip_off = ntohs(packet->ip_off);
487
488	/* Prepare the msghdr and iovec structures describing the new packet */
489	msg.msg_name = linux_args->to;
490	msg.msg_namelen = linux_args->tolen;
491	msg.msg_iov = aiov;
492	msg.msg_iovlen = 2;
493	msg.msg_control = NULL;
494	msg.msg_flags = 0;
495	aiov[0].iov_base = (char *)packet;
496	aiov[0].iov_len = linux_ip_copysize;
497	aiov[1].iov_base = (char *)(linux_args->msg) + linux_ip_copysize;
498	aiov[1].iov_len = linux_args->len - linux_ip_copysize;
499	error = linux_sendit(td, linux_args->s, &msg, linux_args->flags);
500	return (error);
501}
502
503struct linux_socket_args {
504	int domain;
505	int type;
506	int protocol;
507};
508
509static int
510linux_socket(struct thread *td, struct linux_socket_args *args)
511{
512	struct linux_socket_args linux_args;
513	struct socket_args /* {
514		int domain;
515		int type;
516		int protocol;
517	} */ bsd_args;
518	struct setsockopt_args /* {
519		int s;
520		int level;
521		int name;
522		caddr_t val;
523		int valsize;
524	} */ bsd_setsockopt_args;
525	int error;
526	int retval_socket;
527
528	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
529		return (error);
530
531	bsd_args.protocol = linux_args.protocol;
532	bsd_args.type = linux_args.type;
533	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
534	if (bsd_args.domain == -1)
535		return (EINVAL);
536
537	retval_socket = socket(td, &bsd_args);
538	if (bsd_args.type == SOCK_RAW
539	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
540	    && bsd_args.domain == AF_INET
541	    && retval_socket >= 0) {
542		/* It's a raw IP socket: set the IP_HDRINCL option. */
543		caddr_t sg;
544		int *hdrincl;
545
546		sg = stackgap_init();
547		hdrincl = (int *)stackgap_alloc(&sg, sizeof(*hdrincl));
548		*hdrincl = 1;
549		bsd_setsockopt_args.s = td->td_retval[0];
550		bsd_setsockopt_args.level = IPPROTO_IP;
551		bsd_setsockopt_args.name = IP_HDRINCL;
552		bsd_setsockopt_args.val = (caddr_t)hdrincl;
553		bsd_setsockopt_args.valsize = sizeof(*hdrincl);
554		/* We ignore any error returned by setsockopt() */
555		setsockopt(td, &bsd_setsockopt_args);
556		/* Copy back the return value from socket() */
557		td->td_retval[0] = bsd_setsockopt_args.s;
558	}
559#ifdef INET6
560	/*
561	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
562	 * default and some apps depend on this. So, set V6ONLY to 0
563	 * for Linux apps if the sysctl value is set to 1.
564	 */
565	if (bsd_args.domain == PF_INET6 && retval_socket >= 0
566#ifndef KLD_MODULE
567	    /*
568	     * XXX: Avoid undefined symbol error with an IPv4 only
569	     * kernel.
570	     */
571	    && ip6_v6only
572#endif
573	    ) {
574		caddr_t sg;
575		int *v6only;
576
577		sg = stackgap_init();
578		v6only = (int *)stackgap_alloc(&sg, sizeof(*v6only));
579		*v6only = 0;
580		bsd_setsockopt_args.s = td->td_retval[0];
581		bsd_setsockopt_args.level = IPPROTO_IPV6;
582		bsd_setsockopt_args.name = IPV6_V6ONLY;
583		bsd_setsockopt_args.val = (caddr_t)v6only;
584		bsd_setsockopt_args.valsize = sizeof(*v6only);
585		/* We ignore any error returned by setsockopt() */
586		setsockopt(td, &bsd_setsockopt_args);
587		/* Copy back the return value from socket() */
588		td->td_retval[0] = bsd_setsockopt_args.s;
589	}
590#endif
591
592	return (retval_socket);
593}
594
595struct linux_bind_args {
596	int s;
597	struct osockaddr *name;
598	int namelen;
599};
600
601static int
602linux_bind(struct thread *td, struct linux_bind_args *args)
603{
604	struct linux_bind_args linux_args;
605	struct sockaddr *sa;
606	int error;
607
608	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
609		return (error);
610
611	error = linux_getsockaddr(&sa, linux_args.name, linux_args.namelen);
612	if (error)
613		return (error);
614
615	return (kern_bind(td, linux_args.s, sa));
616}
617
618struct linux_connect_args {
619	int s;
620	struct osockaddr * name;
621	int namelen;
622};
623int linux_connect(struct thread *, struct linux_connect_args *);
624#endif /* !__alpha__*/
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#ifdef __alpha__
636	bcopy(args, &linux_args, sizeof(linux_args));
637#else
638	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
639		return (error);
640#endif /* __alpha__ */
641
642	error = linux_getsockaddr(&sa, (struct osockaddr *)linux_args.name,
643	    linux_args.namelen);
644	if (error)
645		return (error);
646
647	error = kern_connect(td, linux_args.s, sa);
648	if (error != EISCONN)
649		return (error);
650
651	/*
652	 * Linux doesn't return EISCONN the first time it occurs,
653	 * when on a non-blocking socket. Instead it returns the
654	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
655	 */
656	if ((error = fgetsock(td, linux_args.s, &so, &fflag)) != 0)
657		return(error);
658	error = EISCONN;
659	if (fflag & FNONBLOCK) {
660		if (so->so_emuldata == 0)
661			error = so->so_error;
662		so->so_emuldata = (void *)1;
663	}
664	fputsock(so);
665	return (error);
666}
667
668#ifndef __alpha__
669
670struct linux_listen_args {
671	int s;
672	int backlog;
673};
674
675static int
676linux_listen(struct thread *td, struct linux_listen_args *args)
677{
678	struct linux_listen_args linux_args;
679	struct listen_args /* {
680		int s;
681		int backlog;
682	} */ bsd_args;
683	int error;
684
685	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
686		return (error);
687
688	bsd_args.s = linux_args.s;
689	bsd_args.backlog = linux_args.backlog;
690	return (listen(td, &bsd_args));
691}
692
693struct linux_accept_args {
694	int s;
695	struct osockaddr *addr;
696	int *namelen;
697};
698
699static int
700linux_accept(struct thread *td, struct linux_accept_args *args)
701{
702	struct linux_accept_args linux_args;
703	struct accept_args /* {
704		int	s;
705		struct sockaddr * __restrict name;
706		socklen_t * __restrict anamelen;
707	} */ bsd_args;
708	struct close_args /* {
709		int     fd;
710	} */ c_args;
711	struct fcntl_args /* {
712		int	fd;
713		int	cmd;
714		long	arg;
715	} */ f_args;
716	int error;
717
718	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
719		return (error);
720
721	bsd_args.s = linux_args.s;
722	/* XXX: */
723	bsd_args.name = (struct sockaddr * __restrict)linux_args.addr;
724	bsd_args.anamelen = linux_args.namelen;		/* XXX */
725	error = oaccept(td, &bsd_args);
726	if (error)
727		return (error);
728	if (linux_args.addr) {
729		error = linux_sa_put(linux_args.addr);
730		if (error) {
731			c_args.fd = td->td_retval[0];
732			(void)close(td, &c_args);
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	f_args.fd = td->td_retval[0];
743	f_args.cmd = F_SETFL;
744	f_args.arg = 0;
745	(void)fcntl(td, &f_args);
746	td->td_retval[0] = f_args.fd;
747	return (0);
748}
749
750struct linux_getsockname_args {
751	int s;
752	struct osockaddr *addr;
753	int *namelen;
754};
755
756static int
757linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
758{
759	struct linux_getsockname_args linux_args;
760	struct getsockname_args /* {
761		int	fdes;
762		struct sockaddr * __restrict asa;
763		socklen_t * __restrict alen;
764	} */ bsd_args;
765	int error;
766
767	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
768		return (error);
769
770	bsd_args.fdes = linux_args.s;
771	/* XXX: */
772	bsd_args.asa = (struct sockaddr * __restrict)linux_args.addr;
773	bsd_args.alen = linux_args.namelen;	/* XXX */
774	error = ogetsockname(td, &bsd_args);
775	if (error)
776		return (error);
777	error = linux_sa_put(linux_args.addr);
778	if (error)
779		return (error);
780	return (0);
781}
782
783struct linux_getpeername_args {
784	int s;
785	struct osockaddr *addr;
786	int *namelen;
787};
788
789static int
790linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
791{
792	struct linux_getpeername_args linux_args;
793	struct ogetpeername_args /* {
794		int fdes;
795		caddr_t asa;
796		int *alen;
797	} */ bsd_args;
798	int error;
799
800	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
801		return (error);
802
803	bsd_args.fdes = linux_args.s;
804	bsd_args.asa = (caddr_t) linux_args.addr;
805	bsd_args.alen = linux_args.namelen;
806	error = ogetpeername(td, &bsd_args);
807	if (error)
808		return (error);
809	error = linux_sa_put(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	int *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 = linux_args.rsv;
844	return (socketpair(td, &bsd_args));
845}
846
847struct linux_send_args {
848	int s;
849	void *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 = 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	void *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 = 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	caddr_t sg = stackgap_init();
920	int error;
921
922	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
923		return (error);
924
925	if (linux_check_hdrincl(td, &sg, linux_args.s) == 0)
926		/* IP_HDRINCL set, tweak the packet before sending */
927		return (linux_sendto_hdrincl(td, &sg, &linux_args));
928
929	msg.msg_name = linux_args.to;
930	msg.msg_namelen = linux_args.tolen;
931	msg.msg_iov = &aiov;
932	msg.msg_iovlen = 1;
933	msg.msg_control = NULL;
934	msg.msg_flags = 0;
935	aiov.iov_base = linux_args.msg;
936	aiov.iov_len = linux_args.len;
937	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags);
938	return (error);
939}
940
941struct linux_recvfrom_args {
942	int s;
943	void *buf;
944	int len;
945	int flags;
946	caddr_t from;
947	int *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	int error;
963
964	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
965		return (error);
966
967	bsd_args.s = linux_args.s;
968	bsd_args.buf = linux_args.buf;
969	bsd_args.len = linux_args.len;
970	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
971	/* XXX: */
972	bsd_args.from = (struct sockaddr * __restrict)linux_args.from;
973	bsd_args.fromlenaddr = linux_args.fromlen;	/* XXX */
974	error = orecvfrom(td, &bsd_args);
975	if (error)
976		return (error);
977	if (linux_args.from) {
978		error = linux_sa_put((struct osockaddr *) linux_args.from);
979		if (error)
980			return (error);
981	}
982	return (0);
983}
984
985struct linux_sendmsg_args {
986	int s;
987	const struct msghdr *msg;
988	int flags;
989};
990
991static int
992linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
993{
994	struct linux_sendmsg_args linux_args;
995	struct msghdr msg;
996	struct iovec *iov;
997	int error;
998
999	error = copyin(args, &linux_args, sizeof(linux_args));
1000	if (error)
1001		return (error);
1002	error = copyin(linux_args.msg, &msg, sizeof(msg));
1003	if (error)
1004		return (error);
1005	error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1006	if (error)
1007		return (error);
1008	msg.msg_iov = iov;
1009	msg.msg_flags = 0;
1010	error = linux_sendit(td, linux_args.s, &msg, linux_args.flags);
1011	free(iov, M_IOV);
1012	return (error);
1013}
1014
1015struct linux_recvmsg_args {
1016	int s;
1017	struct msghdr *msg;
1018	int flags;
1019};
1020
1021static int
1022linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1023{
1024	struct linux_recvmsg_args linux_args;
1025	struct recvmsg_args /* {
1026		int	s;
1027		struct	msghdr *msg;
1028		int	flags;
1029	} */ bsd_args;
1030	struct msghdr msg;
1031	struct cmsghdr *cmsg;
1032	int error;
1033
1034	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1035		return (error);
1036
1037	bsd_args.s = linux_args.s;
1038	bsd_args.msg = linux_args.msg;
1039	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
1040	error = recvmsg(td, &bsd_args);
1041	if (error)
1042		return (error);
1043
1044	if (bsd_args.msg->msg_control != NULL) {
1045		cmsg = (struct cmsghdr*)bsd_args.msg->msg_control;
1046		cmsg->cmsg_level = bsd_to_linux_sockopt_level(cmsg->cmsg_level);
1047	}
1048
1049	error = copyin(linux_args.msg, &msg, sizeof(msg));
1050	if (error)
1051		return (error);
1052	if (msg.msg_name && msg.msg_namelen > 2)
1053		error = linux_sa_put(msg.msg_name);
1054	return (error);
1055}
1056
1057struct linux_shutdown_args {
1058	int s;
1059	int how;
1060};
1061
1062static int
1063linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1064{
1065	struct linux_shutdown_args linux_args;
1066	struct shutdown_args /* {
1067		int s;
1068		int how;
1069	} */ bsd_args;
1070	int error;
1071
1072	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1073		return (error);
1074
1075	bsd_args.s = linux_args.s;
1076	bsd_args.how = linux_args.how;
1077	return (shutdown(td, &bsd_args));
1078}
1079
1080struct linux_setsockopt_args {
1081	int s;
1082	int level;
1083	int optname;
1084	void *optval;
1085	int optlen;
1086};
1087
1088static int
1089linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1090{
1091	struct linux_setsockopt_args linux_args;
1092	struct setsockopt_args /* {
1093		int s;
1094		int level;
1095		int name;
1096		caddr_t val;
1097		int valsize;
1098	} */ bsd_args;
1099	int error, name;
1100
1101	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1102		return (error);
1103
1104	bsd_args.s = linux_args.s;
1105	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1106	switch (bsd_args.level) {
1107	case SOL_SOCKET:
1108		name = linux_to_bsd_so_sockopt(linux_args.optname);
1109		break;
1110	case IPPROTO_IP:
1111		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1112		break;
1113	case IPPROTO_TCP:
1114		/* Linux TCP option values match BSD's */
1115		name = linux_args.optname;
1116		break;
1117	default:
1118		name = -1;
1119		break;
1120	}
1121	if (name == -1)
1122		return (EINVAL);
1123
1124	bsd_args.name = name;
1125	bsd_args.val = linux_args.optval;
1126	bsd_args.valsize = linux_args.optlen;
1127	return (setsockopt(td, &bsd_args));
1128}
1129
1130struct linux_getsockopt_args {
1131	int s;
1132	int level;
1133	int optname;
1134	void *optval;
1135	int *optlen;
1136};
1137
1138static int
1139linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1140{
1141	struct linux_getsockopt_args linux_args;
1142	struct getsockopt_args /* {
1143		int s;
1144		int level;
1145		int name;
1146		caddr_t val;
1147		int *avalsize;
1148	} */ bsd_args;
1149	int error, name;
1150
1151	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1152		return (error);
1153
1154	bsd_args.s = linux_args.s;
1155	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1156	switch (bsd_args.level) {
1157	case SOL_SOCKET:
1158		name = linux_to_bsd_so_sockopt(linux_args.optname);
1159		break;
1160	case IPPROTO_IP:
1161		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1162		break;
1163	case IPPROTO_TCP:
1164		/* Linux TCP option values match BSD's */
1165		name = linux_args.optname;
1166		break;
1167	default:
1168		name = -1;
1169		break;
1170	}
1171	if (name == -1)
1172		return (EINVAL);
1173
1174	bsd_args.name = name;
1175	bsd_args.val = linux_args.optval;
1176	bsd_args.avalsize = linux_args.optlen;
1177	return (getsockopt(td, &bsd_args));
1178}
1179
1180int
1181linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1182{
1183	void *arg = (void *)args->args;
1184
1185	switch (args->what) {
1186	case LINUX_SOCKET:
1187		return (linux_socket(td, arg));
1188	case LINUX_BIND:
1189		return (linux_bind(td, arg));
1190	case LINUX_CONNECT:
1191		return (linux_connect(td, arg));
1192	case LINUX_LISTEN:
1193		return (linux_listen(td, arg));
1194	case LINUX_ACCEPT:
1195		return (linux_accept(td, arg));
1196	case LINUX_GETSOCKNAME:
1197		return (linux_getsockname(td, arg));
1198	case LINUX_GETPEERNAME:
1199		return (linux_getpeername(td, arg));
1200	case LINUX_SOCKETPAIR:
1201		return (linux_socketpair(td, arg));
1202	case LINUX_SEND:
1203		return (linux_send(td, arg));
1204	case LINUX_RECV:
1205		return (linux_recv(td, arg));
1206	case LINUX_SENDTO:
1207		return (linux_sendto(td, arg));
1208	case LINUX_RECVFROM:
1209		return (linux_recvfrom(td, arg));
1210	case LINUX_SHUTDOWN:
1211		return (linux_shutdown(td, arg));
1212	case LINUX_SETSOCKOPT:
1213		return (linux_setsockopt(td, arg));
1214	case LINUX_GETSOCKOPT:
1215		return (linux_getsockopt(td, arg));
1216	case LINUX_SENDMSG:
1217		return (linux_sendmsg(td, arg));
1218	case LINUX_RECVMSG:
1219		return (linux_recvmsg(td, arg));
1220	}
1221
1222	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1223	return (ENOSYS);
1224}
1225#endif	/*!__alpha__*/
1226