linux_socket.c revision 114216
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 * $FreeBSD: head/sys/compat/linux/linux_socket.c 114216 2003-04-29 13:36:06Z kan $
29 */
30
31/* XXX we use functions that might not exist. */
32#include "opt_compat.h"
33#include "opt_inet6.h"
34
35#ifndef COMPAT_43
36#error "Unable to compile Linux-emulator due to missing COMPAT_43 option!"
37#endif
38
39#include <sys/param.h>
40#include <sys/proc.h>
41#include <sys/systm.h>
42#include <sys/sysproto.h>
43#include <sys/fcntl.h>
44#include <sys/file.h>
45#include <sys/limits.h>
46#include <sys/malloc.h>
47#include <sys/socket.h>
48#include <sys/socketvar.h>
49#include <sys/syscallsubr.h>
50#include <sys/uio.h>
51#include <sys/syslog.h>
52
53#include <netinet/in.h>
54#include <netinet/in_systm.h>
55#include <netinet/ip.h>
56#ifdef INET6
57#include <netinet/ip6.h>
58#include <netinet6/ip6_var.h>
59#endif
60
61#include <machine/../linux/linux.h>
62#include <machine/../linux/linux_proto.h>
63#include <compat/linux/linux_socket.h>
64#include <compat/linux/linux_util.h>
65
66static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
67    struct malloc_type *);
68static int linux_to_bsd_domain(int);
69
70/*
71 * Reads a linux sockaddr and does any necessary translation.
72 * Linux sockaddrs don't have a length field, only a family.
73 */
74static int
75linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
76{
77	int osalen = len;
78
79	return (do_sa_get(sap, osa, &osalen, M_SONAME));
80}
81
82/*
83 * Copy the osockaddr structure pointed to by osa to kernel, adjust
84 * family and convert to sockaddr.
85 */
86static int
87do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
88    struct malloc_type *mtype)
89{
90	int error=0, bdom;
91	struct sockaddr *sa;
92	struct osockaddr *kosa;
93	int alloclen;
94#ifdef INET6
95	int oldv6size;
96	struct sockaddr_in6 *sin6;
97#endif
98
99	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
100		return (EINVAL);
101
102	alloclen = *osalen;
103#ifdef INET6
104	oldv6size = 0;
105	/*
106	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
107	 * if it's a v4-mapped address, so reserve the proper space
108	 * for it.
109	 */
110	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
111		alloclen = sizeof (struct sockaddr_in6);
112		oldv6size = 1;
113	}
114#endif
115
116	MALLOC(kosa, struct osockaddr *, alloclen, mtype, M_WAITOK);
117
118	if ((error = copyin(osa, kosa, *osalen)))
119		goto out;
120
121	bdom = linux_to_bsd_domain(kosa->sa_family);
122	if (bdom == -1) {
123		error = EINVAL;
124		goto out;
125	}
126
127#ifdef INET6
128	/*
129	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
130	 * which lacks the scope id compared with RFC2553 one. If we detect
131	 * the situation, reject the address and write a message to system log.
132	 *
133	 * Still accept addresses for which the scope id is not used.
134	 */
135	if (oldv6size && bdom == AF_INET6) {
136		sin6 = (struct sockaddr_in6 *)kosa;
137		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
138		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
139		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
140		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
141		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
142		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
143			sin6->sin6_scope_id = 0;
144		} else {
145			log(LOG_DEBUG,
146			    "obsolete pre-RFC2553 sockaddr_in6 rejected");
147			error = EINVAL;
148			goto out;
149		}
150	} else
151#endif
152	if (bdom == AF_INET)
153		alloclen = sizeof(struct sockaddr_in);
154
155	sa = (struct sockaddr *) kosa;
156	sa->sa_family = bdom;
157	sa->sa_len = alloclen;
158
159	*sap = sa;
160	*osalen = alloclen;
161	return (0);
162
163out:
164	FREE(kosa, mtype);
165	return (error);
166}
167
168static int
169linux_to_bsd_domain(int domain)
170{
171
172	switch (domain) {
173	case LINUX_AF_UNSPEC:
174		return (AF_UNSPEC);
175	case LINUX_AF_UNIX:
176		return (AF_LOCAL);
177	case LINUX_AF_INET:
178		return (AF_INET);
179	case LINUX_AF_INET6:
180		return (AF_INET6);
181	case LINUX_AF_AX25:
182		return (AF_CCITT);
183	case LINUX_AF_IPX:
184		return (AF_IPX);
185	case LINUX_AF_APPLETALK:
186		return (AF_APPLETALK);
187	}
188	return (-1);
189}
190
191#ifndef __alpha__
192static int
193bsd_to_linux_domain(int domain)
194{
195
196	switch (domain) {
197	case AF_UNSPEC:
198		return (LINUX_AF_UNSPEC);
199	case AF_LOCAL:
200		return (LINUX_AF_UNIX);
201	case AF_INET:
202		return (LINUX_AF_INET);
203	case AF_INET6:
204		return (LINUX_AF_INET6);
205	case AF_CCITT:
206		return (LINUX_AF_AX25);
207	case AF_IPX:
208		return (LINUX_AF_IPX);
209	case AF_APPLETALK:
210		return (LINUX_AF_APPLETALK);
211	}
212	return (-1);
213}
214
215static int
216linux_to_bsd_sockopt_level(int level)
217{
218
219	switch (level) {
220	case LINUX_SOL_SOCKET:
221		return (SOL_SOCKET);
222	}
223	return (level);
224}
225
226static int
227linux_to_bsd_ip_sockopt(int opt)
228{
229
230	switch (opt) {
231	case LINUX_IP_TOS:
232		return (IP_TOS);
233	case LINUX_IP_TTL:
234		return (IP_TTL);
235	case LINUX_IP_OPTIONS:
236		return (IP_OPTIONS);
237	case LINUX_IP_MULTICAST_IF:
238		return (IP_MULTICAST_IF);
239	case LINUX_IP_MULTICAST_TTL:
240		return (IP_MULTICAST_TTL);
241	case LINUX_IP_MULTICAST_LOOP:
242		return (IP_MULTICAST_LOOP);
243	case LINUX_IP_ADD_MEMBERSHIP:
244		return (IP_ADD_MEMBERSHIP);
245	case LINUX_IP_DROP_MEMBERSHIP:
246		return (IP_DROP_MEMBERSHIP);
247	case LINUX_IP_HDRINCL:
248		return (IP_HDRINCL);
249	}
250	return (-1);
251}
252
253static int
254linux_to_bsd_so_sockopt(int opt)
255{
256
257	switch (opt) {
258	case LINUX_SO_DEBUG:
259		return (SO_DEBUG);
260	case LINUX_SO_REUSEADDR:
261		return (SO_REUSEADDR);
262	case LINUX_SO_TYPE:
263		return (SO_TYPE);
264	case LINUX_SO_ERROR:
265		return (SO_ERROR);
266	case LINUX_SO_DONTROUTE:
267		return (SO_DONTROUTE);
268	case LINUX_SO_BROADCAST:
269		return (SO_BROADCAST);
270	case LINUX_SO_SNDBUF:
271		return (SO_SNDBUF);
272	case LINUX_SO_RCVBUF:
273		return (SO_RCVBUF);
274	case LINUX_SO_KEEPALIVE:
275		return (SO_KEEPALIVE);
276	case LINUX_SO_OOBINLINE:
277		return (SO_OOBINLINE);
278	case LINUX_SO_LINGER:
279		return (SO_LINGER);
280	}
281	return (-1);
282}
283
284static int
285linux_to_bsd_msg_flags(int flags)
286{
287	int ret_flags = 0;
288
289	if (flags & LINUX_MSG_OOB)
290		ret_flags |= MSG_OOB;
291	if (flags & LINUX_MSG_PEEK)
292		ret_flags |= MSG_PEEK;
293	if (flags & LINUX_MSG_DONTROUTE)
294		ret_flags |= MSG_DONTROUTE;
295	if (flags & LINUX_MSG_CTRUNC)
296		ret_flags |= MSG_CTRUNC;
297	if (flags & LINUX_MSG_TRUNC)
298		ret_flags |= MSG_TRUNC;
299	if (flags & LINUX_MSG_DONTWAIT)
300		ret_flags |= MSG_DONTWAIT;
301	if (flags & LINUX_MSG_EOR)
302		ret_flags |= MSG_EOR;
303	if (flags & LINUX_MSG_WAITALL)
304		ret_flags |= MSG_WAITALL;
305#if 0 /* not handled */
306	if (flags & LINUX_MSG_PROXY)
307		;
308	if (flags & LINUX_MSG_FIN)
309		;
310	if (flags & LINUX_MSG_SYN)
311		;
312	if (flags & LINUX_MSG_CONFIRM)
313		;
314	if (flags & LINUX_MSG_RST)
315		;
316	if (flags & LINUX_MSG_ERRQUEUE)
317		;
318	if (flags & LINUX_MSG_NOSIGNAL)
319		;
320#endif
321	return ret_flags;
322}
323
324/*
325 * Allocate stackgap and put the converted sockaddr structure
326 * there, address on stackgap returned in sap.
327 */
328static int
329linux_sa_get(caddr_t *sgp, struct sockaddr **sap,
330	     const struct osockaddr *osa, int *osalen)
331{
332	struct sockaddr *sa, *usa;
333	int alloclen, error;
334
335	alloclen = *osalen;
336	error = do_sa_get(&sa, osa, &alloclen, M_TEMP);
337	if (error)
338		return (error);
339
340	usa = (struct sockaddr *) stackgap_alloc(sgp, alloclen);
341	if (!usa) {
342		error = ENOMEM;
343		goto out;
344	}
345
346	if ((error = copyout(sa, usa, alloclen)))
347		goto out;
348
349	*sap = usa;
350	*osalen = alloclen;
351
352out:
353	FREE(sa, M_TEMP);
354	return (error);
355}
356
357static int
358linux_sa_put(struct osockaddr *osa)
359{
360	struct osockaddr sa;
361	int error, bdom;
362
363	/*
364	 * Only read/write the osockaddr family part, the rest is
365	 * not changed.
366	 */
367	error = copyin(osa, &sa, sizeof(sa.sa_family));
368	if (error)
369		return (error);
370
371	bdom = bsd_to_linux_domain(sa.sa_family);
372	if (bdom == -1)
373		return (EINVAL);
374
375	sa.sa_family = bdom;
376	error = copyout(&sa, osa, sizeof(sa.sa_family));
377	if (error)
378		return (error);
379
380	return (0);
381}
382
383/* Return 0 if IP_HDRINCL is set for the given socket. */
384static int
385linux_check_hdrincl(struct thread *td, caddr_t *sg, int s)
386{
387	struct getsockopt_args /* {
388		int s;
389		int level;
390		int name;
391		caddr_t val;
392		int *avalsize;
393	} */ bsd_args;
394	int error;
395	caddr_t val, valsize;
396	int size_val = sizeof val;
397	int optval;
398
399	val = stackgap_alloc(sg, sizeof(int));
400	valsize = stackgap_alloc(sg, sizeof(int));
401
402	if ((error = copyout(&size_val, valsize, sizeof(size_val))))
403		return (error);
404
405	bsd_args.s = s;
406	bsd_args.level = IPPROTO_IP;
407	bsd_args.name = IP_HDRINCL;
408	bsd_args.val = val;
409	bsd_args.avalsize = (int *)valsize;
410	if ((error = getsockopt(td, &bsd_args)))
411		return (error);
412
413	if ((error = copyin(val, &optval, sizeof(optval))))
414		return (error);
415
416	return (optval == 0);
417}
418
419/*
420 * Updated sendto() when IP_HDRINCL is set:
421 * tweak endian-dependent fields in the IP packet.
422 */
423static int
424linux_sendto_hdrincl(struct thread *td, caddr_t *sg, struct sendto_args *bsd_args)
425{
426/*
427 * linux_ip_copysize defines how many bytes we should copy
428 * from the beginning of the IP packet before we customize it for BSD.
429 * It should include all the fields we modify (ip_len and ip_off)
430 * and be as small as possible to minimize copying overhead.
431 */
432#define linux_ip_copysize	8
433
434	struct ip *packet;
435	struct msghdr *msg;
436	struct iovec *iov;
437
438	int error;
439	struct  sendmsg_args /* {
440		int s;
441		caddr_t msg;
442		int flags;
443	} */ sendmsg_args;
444
445	/* Check the packet isn't too small before we mess with it */
446	if (bsd_args->len < linux_ip_copysize)
447		return (EINVAL);
448
449	/*
450	 * Tweaking the user buffer in place would be bad manners.
451	 * We create a corrected IP header with just the needed length,
452	 * then use an iovec to glue it to the rest of the user packet
453	 * when calling sendmsg().
454	 */
455	packet = (struct ip *)stackgap_alloc(sg, linux_ip_copysize);
456	msg = (struct msghdr *)stackgap_alloc(sg, sizeof(*msg));
457	iov = (struct iovec *)stackgap_alloc(sg, sizeof(*iov)*2);
458
459	/* Make a copy of the beginning of the packet to be sent */
460	if ((error = copyin(bsd_args->buf, packet, linux_ip_copysize)))
461		return (error);
462
463	/* Convert fields from Linux to BSD raw IP socket format */
464	packet->ip_len = bsd_args->len;
465	packet->ip_off = ntohs(packet->ip_off);
466
467	/* Prepare the msghdr and iovec structures describing the new packet */
468	msg->msg_name = bsd_args->to;
469	msg->msg_namelen = bsd_args->tolen;
470	msg->msg_iov = iov;
471	msg->msg_iovlen = 2;
472	msg->msg_control = NULL;
473	msg->msg_controllen = 0;
474	msg->msg_flags = 0;
475	iov[0].iov_base = (char *)packet;
476	iov[0].iov_len = linux_ip_copysize;
477	iov[1].iov_base = (char *)(bsd_args->buf) + linux_ip_copysize;
478	iov[1].iov_len = bsd_args->len - linux_ip_copysize;
479
480	sendmsg_args.s = bsd_args->s;
481	sendmsg_args.msg = (caddr_t)msg;
482	sendmsg_args.flags = bsd_args->flags;
483	return (sendmsg(td, &sendmsg_args));
484}
485
486struct linux_socket_args {
487	int domain;
488	int type;
489	int protocol;
490};
491
492static int
493linux_socket(struct thread *td, struct linux_socket_args *args)
494{
495	struct linux_socket_args linux_args;
496	struct socket_args /* {
497		int domain;
498		int type;
499		int protocol;
500	} */ bsd_args;
501	struct setsockopt_args /* {
502		int s;
503		int level;
504		int name;
505		caddr_t val;
506		int valsize;
507	} */ bsd_setsockopt_args;
508	int error;
509	int retval_socket;
510
511	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
512		return (error);
513
514	bsd_args.protocol = linux_args.protocol;
515	bsd_args.type = linux_args.type;
516	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
517	if (bsd_args.domain == -1)
518		return (EINVAL);
519
520	retval_socket = socket(td, &bsd_args);
521	if (bsd_args.type == SOCK_RAW
522	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
523	    && bsd_args.domain == AF_INET
524	    && retval_socket >= 0) {
525		/* It's a raw IP socket: set the IP_HDRINCL option. */
526		caddr_t sg;
527		int *hdrincl;
528
529		sg = stackgap_init();
530		hdrincl = (int *)stackgap_alloc(&sg, sizeof(*hdrincl));
531		*hdrincl = 1;
532		bsd_setsockopt_args.s = td->td_retval[0];
533		bsd_setsockopt_args.level = IPPROTO_IP;
534		bsd_setsockopt_args.name = IP_HDRINCL;
535		bsd_setsockopt_args.val = (caddr_t)hdrincl;
536		bsd_setsockopt_args.valsize = sizeof(*hdrincl);
537		/* We ignore any error returned by setsockopt() */
538		setsockopt(td, &bsd_setsockopt_args);
539		/* Copy back the return value from socket() */
540		td->td_retval[0] = bsd_setsockopt_args.s;
541	}
542#ifdef INET6
543	/*
544	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
545	 * default and some apps depend on this. So, set V6ONLY to 0
546	 * for Linux apps if the sysctl value is set to 1.
547	 */
548	if (bsd_args.domain == PF_INET6 && retval_socket >= 0
549#ifndef KLD_MODULE
550	    /*
551	     * XXX: Avoid undefined symbol error with an IPv4 only
552	     * kernel.
553	     */
554	    && ip6_v6only
555#endif
556	    ) {
557		caddr_t sg;
558		int *v6only;
559
560		sg = stackgap_init();
561		v6only = (int *)stackgap_alloc(&sg, sizeof(*v6only));
562		*v6only = 0;
563		bsd_setsockopt_args.s = td->td_retval[0];
564		bsd_setsockopt_args.level = IPPROTO_IPV6;
565		bsd_setsockopt_args.name = IPV6_V6ONLY;
566		bsd_setsockopt_args.val = (caddr_t)v6only;
567		bsd_setsockopt_args.valsize = sizeof(*v6only);
568		/* We ignore any error returned by setsockopt() */
569		setsockopt(td, &bsd_setsockopt_args);
570		/* Copy back the return value from socket() */
571		td->td_retval[0] = bsd_setsockopt_args.s;
572	}
573#endif
574
575	return (retval_socket);
576}
577
578struct linux_bind_args {
579	int s;
580	struct osockaddr *name;
581	int namelen;
582};
583
584static int
585linux_bind(struct thread *td, struct linux_bind_args *args)
586{
587	struct linux_bind_args linux_args;
588	struct sockaddr *sa;
589	int error;
590
591	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
592		return (error);
593
594	error = linux_getsockaddr(&sa, linux_args.name, linux_args.namelen);
595	if (error)
596		return (error);
597
598	return (kern_bind(td, linux_args.s, sa));
599}
600
601struct linux_connect_args {
602	int s;
603	struct osockaddr * name;
604	int namelen;
605};
606int linux_connect(struct thread *, struct linux_connect_args *);
607#endif /* !__alpha__*/
608
609int
610linux_connect(struct thread *td, struct linux_connect_args *args)
611{
612	struct linux_connect_args linux_args;
613	struct socket *so;
614	struct sockaddr *sa;
615	u_int fflag;
616	int error;
617
618#ifdef __alpha__
619	bcopy(args, &linux_args, sizeof(linux_args));
620#else
621	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
622		return (error);
623#endif /* __alpha__ */
624
625	error = linux_getsockaddr(&sa, (struct osockaddr *)linux_args.name,
626	    linux_args.namelen);
627	if (error)
628		return (error);
629
630	error = kern_connect(td, linux_args.s, sa);
631	if (error != EISCONN)
632		return (error);
633
634	/*
635	 * Linux doesn't return EISCONN the first time it occurs,
636	 * when on a non-blocking socket. Instead it returns the
637	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
638	 */
639	if ((error = fgetsock(td, linux_args.s, &so, &fflag)) != 0)
640		return(error);
641	error = EISCONN;
642	if (fflag & FNONBLOCK) {
643		if (so->so_emuldata == 0)
644			error = so->so_error;
645		so->so_emuldata = (void *)1;
646	}
647	fputsock(so);
648	return (error);
649}
650
651#ifndef __alpha__
652
653struct linux_listen_args {
654	int s;
655	int backlog;
656};
657
658static int
659linux_listen(struct thread *td, struct linux_listen_args *args)
660{
661	struct linux_listen_args linux_args;
662	struct listen_args /* {
663		int s;
664		int backlog;
665	} */ bsd_args;
666	int error;
667
668	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
669		return (error);
670
671	bsd_args.s = linux_args.s;
672	bsd_args.backlog = linux_args.backlog;
673	return (listen(td, &bsd_args));
674}
675
676struct linux_accept_args {
677	int s;
678	struct osockaddr *addr;
679	int *namelen;
680};
681
682static int
683linux_accept(struct thread *td, struct linux_accept_args *args)
684{
685	struct linux_accept_args linux_args;
686	struct accept_args /* {
687		int s;
688		caddr_t name;
689		int *anamelen;
690	} */ bsd_args;
691	struct close_args /* {
692		int     fd;
693	} */ c_args;
694	struct fcntl_args /* {
695		int fd;
696		int cmd;
697		long arg;
698	} */ f_args;
699	int error;
700
701	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
702		return (error);
703
704	bsd_args.s = linux_args.s;
705	bsd_args.name = (caddr_t)linux_args.addr;
706	bsd_args.anamelen = linux_args.namelen;
707	error = oaccept(td, &bsd_args);
708	if (error)
709		return (error);
710	if (linux_args.addr) {
711		error = linux_sa_put(linux_args.addr);
712		if (error) {
713			c_args.fd = td->td_retval[0];
714			(void)close(td, &c_args);
715			return (error);
716		}
717	}
718
719	/*
720	 * linux appears not to copy flags from the parent socket to the
721	 * accepted one, so we must clear the flags in the new descriptor.
722	 * Ignore any errors, because we already have an open fd.
723	 */
724	f_args.fd = td->td_retval[0];
725	f_args.cmd = F_SETFL;
726	f_args.arg = 0;
727	(void)fcntl(td, &f_args);
728	td->td_retval[0] = f_args.fd;
729	return (0);
730}
731
732struct linux_getsockname_args {
733	int s;
734	struct osockaddr *addr;
735	int *namelen;
736};
737
738static int
739linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
740{
741	struct linux_getsockname_args linux_args;
742	struct getsockname_args /* {
743		int fdes;
744		caddr_t asa;
745		int *alen;
746	} */ bsd_args;
747	int error;
748
749	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
750		return (error);
751
752	bsd_args.fdes = linux_args.s;
753	bsd_args.asa = (caddr_t) linux_args.addr;
754	bsd_args.alen = linux_args.namelen;
755	error = ogetsockname(td, &bsd_args);
756	if (error)
757		return (error);
758	error = linux_sa_put(linux_args.addr);
759	if (error)
760		return (error);
761	return (0);
762}
763
764struct linux_getpeername_args {
765	int s;
766	struct osockaddr *addr;
767	int *namelen;
768};
769
770static int
771linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
772{
773	struct linux_getpeername_args linux_args;
774	struct ogetpeername_args /* {
775		int fdes;
776		caddr_t asa;
777		int *alen;
778	} */ bsd_args;
779	int error;
780
781	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
782		return (error);
783
784	bsd_args.fdes = linux_args.s;
785	bsd_args.asa = (caddr_t) linux_args.addr;
786	bsd_args.alen = linux_args.namelen;
787	error = ogetpeername(td, &bsd_args);
788	if (error)
789		return (error);
790	error = linux_sa_put(linux_args.addr);
791	if (error)
792		return (error);
793	return (0);
794}
795
796struct linux_socketpair_args {
797	int domain;
798	int type;
799	int protocol;
800	int *rsv;
801};
802
803static int
804linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
805{
806	struct linux_socketpair_args linux_args;
807	struct socketpair_args /* {
808		int domain;
809		int type;
810		int protocol;
811		int *rsv;
812	} */ bsd_args;
813	int error;
814
815	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
816		return (error);
817
818	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
819	if (bsd_args.domain == -1)
820		return (EINVAL);
821
822	bsd_args.type = linux_args.type;
823	bsd_args.protocol = linux_args.protocol;
824	bsd_args.rsv = linux_args.rsv;
825	return (socketpair(td, &bsd_args));
826}
827
828struct linux_send_args {
829	int s;
830	void *msg;
831	int len;
832	int flags;
833};
834
835static int
836linux_send(struct thread *td, struct linux_send_args *args)
837{
838	struct linux_send_args linux_args;
839	struct osend_args /* {
840		int s;
841		caddr_t buf;
842		int len;
843		int flags;
844	} */ bsd_args;
845	int error;
846
847	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
848		return (error);
849
850	bsd_args.s = linux_args.s;
851	bsd_args.buf = linux_args.msg;
852	bsd_args.len = linux_args.len;
853	bsd_args.flags = linux_args.flags;
854	return (osend(td, &bsd_args));
855}
856
857struct linux_recv_args {
858	int s;
859	void *msg;
860	int len;
861	int flags;
862};
863
864static int
865linux_recv(struct thread *td, struct linux_recv_args *args)
866{
867	struct linux_recv_args linux_args;
868	struct orecv_args /* {
869		int s;
870		caddr_t buf;
871		int len;
872		int flags;
873	} */ bsd_args;
874	int error;
875
876	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
877		return (error);
878
879	bsd_args.s = linux_args.s;
880	bsd_args.buf = linux_args.msg;
881	bsd_args.len = linux_args.len;
882	bsd_args.flags = linux_args.flags;
883	return (orecv(td, &bsd_args));
884}
885
886struct linux_sendto_args {
887	int s;
888	void *msg;
889	int len;
890	int flags;
891	caddr_t to;
892	int tolen;
893};
894
895static int
896linux_sendto(struct thread *td, struct linux_sendto_args *args)
897{
898	struct linux_sendto_args linux_args;
899	struct sendto_args /* {
900		int s;
901		caddr_t buf;
902		size_t len;
903		int flags;
904		caddr_t to;
905		int tolen;
906	} */ bsd_args;
907	caddr_t sg = stackgap_init();
908	struct sockaddr *to;
909	int tolen, error;
910
911	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
912		return (error);
913
914	tolen = linux_args.tolen;
915	if (linux_args.to) {
916		error = linux_sa_get(&sg, &to,
917		    (struct osockaddr *) linux_args.to, &tolen);
918		if (error)
919			return (error);
920	} else
921		to = NULL;
922
923	bsd_args.s = linux_args.s;
924	bsd_args.buf = linux_args.msg;
925	bsd_args.len = linux_args.len;
926	bsd_args.flags = linux_args.flags;
927	bsd_args.to = (caddr_t) to;
928	bsd_args.tolen = (unsigned int) tolen;
929
930	if (linux_check_hdrincl(td, &sg, linux_args.s) == 0)
931		/* IP_HDRINCL set, tweak the packet before sending */
932		return (linux_sendto_hdrincl(td, &sg, &bsd_args));
933
934	return (sendto(td, &bsd_args));
935}
936
937struct linux_recvfrom_args {
938	int s;
939	void *buf;
940	int len;
941	int flags;
942	caddr_t from;
943	int *fromlen;
944};
945
946static int
947linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
948{
949	struct linux_recvfrom_args linux_args;
950	struct recvfrom_args /* {
951		int s;
952		caddr_t buf;
953		size_t len;
954		int flags;
955		caddr_t from;
956		int *fromlenaddr;
957	} */ bsd_args;
958	int error;
959
960	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
961		return (error);
962
963	bsd_args.s = linux_args.s;
964	bsd_args.buf = linux_args.buf;
965	bsd_args.len = linux_args.len;
966	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
967	bsd_args.from = linux_args.from;
968	bsd_args.fromlenaddr = linux_args.fromlen;
969	error = orecvfrom(td, &bsd_args);
970	if (error)
971		return (error);
972	if (linux_args.from) {
973		error = linux_sa_put((struct osockaddr *) linux_args.from);
974		if (error)
975			return (error);
976	}
977	return (0);
978}
979
980struct linux_sendmsg_args {
981	int s;
982	const struct msghdr *msg;
983	int flags;
984};
985
986static int
987linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
988{
989	struct linux_sendmsg_args linux_args;
990	struct sendmsg_args /* {
991		int s;
992		const struct msghdr *msg;
993		int flags;
994	} */ bsd_args;
995	struct msghdr msg;
996	struct msghdr *nmsg = NULL;
997	int error;
998	int level;
999	caddr_t control;
1000
1001	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1002		return (error);
1003
1004	error = copyin(linux_args.msg, &msg, sizeof(msg));
1005	if (error)
1006		return (error);
1007
1008	if (msg.msg_name) {
1009		struct sockaddr *sa;
1010		caddr_t sg = stackgap_init();
1011
1012		nmsg = (struct msghdr *) stackgap_alloc(&sg,
1013		    sizeof(struct msghdr));
1014		if (!nmsg)
1015			return (ENOMEM);
1016
1017		error = linux_sa_get(&sg, &sa,
1018		    (struct osockaddr *) msg.msg_name, &msg.msg_namelen);
1019		if (error)
1020			return (error);
1021
1022		msg.msg_name = (struct sockaddr *) sa;
1023		error = copyout(&msg, nmsg, sizeof(struct msghdr));
1024		if (error)
1025			return (error);
1026	}
1027
1028	error = copyin(&linux_args.msg->msg_control, &control,
1029	    sizeof(caddr_t));
1030	if (error)
1031		return (error);
1032
1033	if (control == NULL)
1034		goto done;
1035
1036	error = copyin(&((struct cmsghdr*)control)->cmsg_level, &level,
1037	    sizeof(int));
1038	if (error)
1039		return (error);
1040
1041	if (level == 1) {
1042		/*
1043		 * Linux thinks that SOL_SOCKET is 1; we know
1044		 * that it's really 0xffff, of course.
1045		 */
1046		level = SOL_SOCKET;
1047		error = copyout(&level,
1048		    &((struct cmsghdr *)control)->cmsg_level, sizeof(int));
1049		if (error)
1050			return (error);
1051	}
1052done:
1053	bsd_args.s = linux_args.s;
1054	bsd_args.msg = (caddr_t)nmsg;
1055	bsd_args.flags = linux_args.flags;
1056	return (sendmsg(td, &bsd_args));
1057}
1058
1059struct linux_recvmsg_args {
1060	int s;
1061	struct msghdr *msg;
1062	int flags;
1063};
1064
1065static int
1066linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1067{
1068	struct linux_recvmsg_args linux_args;
1069	struct recvmsg_args /* {
1070		int	s;
1071		struct	msghdr *msg;
1072		int	flags;
1073	} */ bsd_args;
1074	struct msghdr msg;
1075	int error;
1076
1077	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1078		return (error);
1079
1080	bsd_args.s = linux_args.s;
1081	bsd_args.msg = linux_args.msg;
1082	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
1083	error = recvmsg(td, &bsd_args);
1084	if (error)
1085		return (error);
1086
1087	error = copyin(linux_args.msg, &msg, sizeof(msg));
1088	if (error)
1089		return (error);
1090	if (msg.msg_name && msg.msg_namelen > 2)
1091		error = linux_sa_put(msg.msg_name);
1092	return (error);
1093}
1094
1095struct linux_shutdown_args {
1096	int s;
1097	int how;
1098};
1099
1100static int
1101linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1102{
1103	struct linux_shutdown_args linux_args;
1104	struct shutdown_args /* {
1105		int s;
1106		int how;
1107	} */ bsd_args;
1108	int error;
1109
1110	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1111		return (error);
1112
1113	bsd_args.s = linux_args.s;
1114	bsd_args.how = linux_args.how;
1115	return (shutdown(td, &bsd_args));
1116}
1117
1118struct linux_setsockopt_args {
1119	int s;
1120	int level;
1121	int optname;
1122	void *optval;
1123	int optlen;
1124};
1125
1126static int
1127linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1128{
1129	struct linux_setsockopt_args linux_args;
1130	struct setsockopt_args /* {
1131		int s;
1132		int level;
1133		int name;
1134		caddr_t val;
1135		int valsize;
1136	} */ bsd_args;
1137	int error, name;
1138
1139	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1140		return (error);
1141
1142	bsd_args.s = linux_args.s;
1143	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1144	switch (bsd_args.level) {
1145	case SOL_SOCKET:
1146		name = linux_to_bsd_so_sockopt(linux_args.optname);
1147		break;
1148	case IPPROTO_IP:
1149		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1150		break;
1151	case IPPROTO_TCP:
1152		/* Linux TCP option values match BSD's */
1153		name = linux_args.optname;
1154		break;
1155	default:
1156		name = -1;
1157		break;
1158	}
1159	if (name == -1)
1160		return (EINVAL);
1161
1162	bsd_args.name = name;
1163	bsd_args.val = linux_args.optval;
1164	bsd_args.valsize = linux_args.optlen;
1165	return (setsockopt(td, &bsd_args));
1166}
1167
1168struct linux_getsockopt_args {
1169	int s;
1170	int level;
1171	int optname;
1172	void *optval;
1173	int *optlen;
1174};
1175
1176static int
1177linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1178{
1179	struct linux_getsockopt_args linux_args;
1180	struct getsockopt_args /* {
1181		int s;
1182		int level;
1183		int name;
1184		caddr_t val;
1185		int *avalsize;
1186	} */ bsd_args;
1187	int error, name;
1188
1189	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1190		return (error);
1191
1192	bsd_args.s = linux_args.s;
1193	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1194	switch (bsd_args.level) {
1195	case SOL_SOCKET:
1196		name = linux_to_bsd_so_sockopt(linux_args.optname);
1197		break;
1198	case IPPROTO_IP:
1199		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1200		break;
1201	case IPPROTO_TCP:
1202		/* Linux TCP option values match BSD's */
1203		name = linux_args.optname;
1204		break;
1205	default:
1206		name = -1;
1207		break;
1208	}
1209	if (name == -1)
1210		return (EINVAL);
1211
1212	bsd_args.name = name;
1213	bsd_args.val = linux_args.optval;
1214	bsd_args.avalsize = linux_args.optlen;
1215	return (getsockopt(td, &bsd_args));
1216}
1217
1218int
1219linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1220{
1221	void *arg = (void *)args->args;
1222
1223	switch (args->what) {
1224	case LINUX_SOCKET:
1225		return (linux_socket(td, arg));
1226	case LINUX_BIND:
1227		return (linux_bind(td, arg));
1228	case LINUX_CONNECT:
1229		return (linux_connect(td, arg));
1230	case LINUX_LISTEN:
1231		return (linux_listen(td, arg));
1232	case LINUX_ACCEPT:
1233		return (linux_accept(td, arg));
1234	case LINUX_GETSOCKNAME:
1235		return (linux_getsockname(td, arg));
1236	case LINUX_GETPEERNAME:
1237		return (linux_getpeername(td, arg));
1238	case LINUX_SOCKETPAIR:
1239		return (linux_socketpair(td, arg));
1240	case LINUX_SEND:
1241		return (linux_send(td, arg));
1242	case LINUX_RECV:
1243		return (linux_recv(td, arg));
1244	case LINUX_SENDTO:
1245		return (linux_sendto(td, arg));
1246	case LINUX_RECVFROM:
1247		return (linux_recvfrom(td, arg));
1248	case LINUX_SHUTDOWN:
1249		return (linux_shutdown(td, arg));
1250	case LINUX_SETSOCKOPT:
1251		return (linux_setsockopt(td, arg));
1252	case LINUX_GETSOCKOPT:
1253		return (linux_getsockopt(td, arg));
1254	case LINUX_SENDMSG:
1255		return (linux_sendmsg(td, arg));
1256	case LINUX_RECVMSG:
1257		return (linux_recvmsg(td, arg));
1258	}
1259
1260	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1261	return (ENOSYS);
1262}
1263#endif	/*!__alpha__*/
1264