linux_socket.c revision 111797
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 111797 2003-03-03 09:14:26Z des $
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/malloc.h>
46#include <sys/socket.h>
47#include <sys/socketvar.h>
48#include <sys/syscallsubr.h>
49#include <sys/uio.h>
50#include <sys/syslog.h>
51
52#include <netinet/in.h>
53#include <netinet/in_systm.h>
54#include <netinet/ip.h>
55#ifdef INET6
56#include <netinet/ip6.h>
57#include <netinet6/ip6_var.h>
58#endif
59
60#include <machine/../linux/linux.h>
61#include <machine/../linux/linux_proto.h>
62#include <compat/linux/linux_socket.h>
63#include <compat/linux/linux_util.h>
64
65#include <machine/limits.h>
66
67static int do_sa_get(struct sockaddr **, const struct osockaddr *, int *,
68    struct malloc_type *);
69static int linux_to_bsd_domain(int);
70
71/*
72 * Reads a linux sockaddr and does any necessary translation.
73 * Linux sockaddrs don't have a length field, only a family.
74 */
75static int
76linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int len)
77{
78	int osalen = len;
79
80	return (do_sa_get(sap, osa, &osalen, M_SONAME));
81}
82
83/*
84 * Copy the osockaddr structure pointed to by osa to kernel, adjust
85 * family and convert to sockaddr.
86 */
87static int
88do_sa_get(struct sockaddr **sap, const struct osockaddr *osa, int *osalen,
89    struct malloc_type *mtype)
90{
91	int error=0, bdom;
92	struct sockaddr *sa;
93	struct osockaddr *kosa;
94	int alloclen;
95#ifdef INET6
96	int oldv6size;
97	struct sockaddr_in6 *sin6;
98#endif
99
100	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa)
101		return (EINVAL);
102
103	alloclen = *osalen;
104#ifdef INET6
105	oldv6size = 0;
106	/*
107	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
108	 * if it's a v4-mapped address, so reserve the proper space
109	 * for it.
110	 */
111	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
112		alloclen = sizeof (struct sockaddr_in6);
113		oldv6size = 1;
114	}
115#endif
116
117	MALLOC(kosa, struct osockaddr *, alloclen, mtype, M_WAITOK);
118
119	if ((error = copyin(osa, kosa, *osalen)))
120		goto out;
121
122	bdom = linux_to_bsd_domain(kosa->sa_family);
123	if (bdom == -1) {
124		error = EINVAL;
125		goto out;
126	}
127
128#ifdef INET6
129	/*
130	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
131	 * which lacks the scope id compared with RFC2553 one. If we detect
132	 * the situation, reject the address and write a message to system log.
133	 *
134	 * Still accept addresses for which the scope id is not used.
135	 */
136	if (oldv6size && bdom == AF_INET6) {
137		sin6 = (struct sockaddr_in6 *)kosa;
138		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
139		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
140		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
141		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
142		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
143		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
144			sin6->sin6_scope_id = 0;
145		} else {
146			log(LOG_DEBUG,
147			    "obsolete pre-RFC2553 sockaddr_in6 rejected");
148			error = EINVAL;
149			goto out;
150		}
151	} else
152#endif
153	if (bdom == AF_INET)
154		alloclen = sizeof(struct sockaddr_in);
155
156	sa = (struct sockaddr *) kosa;
157	sa->sa_family = bdom;
158	sa->sa_len = alloclen;
159
160	*sap = sa;
161	*osalen = alloclen;
162	return (0);
163
164out:
165	FREE(kosa, mtype);
166	return (error);
167}
168
169static int
170linux_to_bsd_domain(int domain)
171{
172
173	switch (domain) {
174	case LINUX_AF_UNSPEC:
175		return (AF_UNSPEC);
176	case LINUX_AF_UNIX:
177		return (AF_LOCAL);
178	case LINUX_AF_INET:
179		return (AF_INET);
180	case LINUX_AF_INET6:
181		return (AF_INET6);
182	case LINUX_AF_AX25:
183		return (AF_CCITT);
184	case LINUX_AF_IPX:
185		return (AF_IPX);
186	case LINUX_AF_APPLETALK:
187		return (AF_APPLETALK);
188	}
189	return (-1);
190}
191
192#ifndef __alpha__
193static int
194bsd_to_linux_domain(int domain)
195{
196
197	switch (domain) {
198	case AF_UNSPEC:
199		return (LINUX_AF_UNSPEC);
200	case AF_LOCAL:
201		return (LINUX_AF_UNIX);
202	case AF_INET:
203		return (LINUX_AF_INET);
204	case AF_INET6:
205		return (LINUX_AF_INET6);
206	case AF_CCITT:
207		return (LINUX_AF_AX25);
208	case AF_IPX:
209		return (LINUX_AF_IPX);
210	case AF_APPLETALK:
211		return (LINUX_AF_APPLETALK);
212	}
213	return (-1);
214}
215
216static int
217linux_to_bsd_sockopt_level(int level)
218{
219
220	switch (level) {
221	case LINUX_SOL_SOCKET:
222		return (SOL_SOCKET);
223	}
224	return (level);
225}
226
227static int
228linux_to_bsd_ip_sockopt(int opt)
229{
230
231	switch (opt) {
232	case LINUX_IP_TOS:
233		return (IP_TOS);
234	case LINUX_IP_TTL:
235		return (IP_TTL);
236	case LINUX_IP_OPTIONS:
237		return (IP_OPTIONS);
238	case LINUX_IP_MULTICAST_IF:
239		return (IP_MULTICAST_IF);
240	case LINUX_IP_MULTICAST_TTL:
241		return (IP_MULTICAST_TTL);
242	case LINUX_IP_MULTICAST_LOOP:
243		return (IP_MULTICAST_LOOP);
244	case LINUX_IP_ADD_MEMBERSHIP:
245		return (IP_ADD_MEMBERSHIP);
246	case LINUX_IP_DROP_MEMBERSHIP:
247		return (IP_DROP_MEMBERSHIP);
248	case LINUX_IP_HDRINCL:
249		return (IP_HDRINCL);
250	}
251	return (-1);
252}
253
254static int
255linux_to_bsd_so_sockopt(int opt)
256{
257
258	switch (opt) {
259	case LINUX_SO_DEBUG:
260		return (SO_DEBUG);
261	case LINUX_SO_REUSEADDR:
262		return (SO_REUSEADDR);
263	case LINUX_SO_TYPE:
264		return (SO_TYPE);
265	case LINUX_SO_ERROR:
266		return (SO_ERROR);
267	case LINUX_SO_DONTROUTE:
268		return (SO_DONTROUTE);
269	case LINUX_SO_BROADCAST:
270		return (SO_BROADCAST);
271	case LINUX_SO_SNDBUF:
272		return (SO_SNDBUF);
273	case LINUX_SO_RCVBUF:
274		return (SO_RCVBUF);
275	case LINUX_SO_KEEPALIVE:
276		return (SO_KEEPALIVE);
277	case LINUX_SO_OOBINLINE:
278		return (SO_OOBINLINE);
279	case LINUX_SO_LINGER:
280		return (SO_LINGER);
281	}
282	return (-1);
283}
284
285static int
286linux_to_bsd_msg_flags(int flags)
287{
288	int ret_flags = 0;
289
290	if (flags & LINUX_MSG_OOB)
291		ret_flags |= MSG_OOB;
292	if (flags & LINUX_MSG_PEEK)
293		ret_flags |= MSG_PEEK;
294	if (flags & LINUX_MSG_DONTROUTE)
295		ret_flags |= MSG_DONTROUTE;
296	if (flags & LINUX_MSG_CTRUNC)
297		ret_flags |= MSG_CTRUNC;
298	if (flags & LINUX_MSG_TRUNC)
299		ret_flags |= MSG_TRUNC;
300	if (flags & LINUX_MSG_DONTWAIT)
301		ret_flags |= MSG_DONTWAIT;
302	if (flags & LINUX_MSG_EOR)
303		ret_flags |= MSG_EOR;
304	if (flags & LINUX_MSG_WAITALL)
305		ret_flags |= MSG_WAITALL;
306#if 0 /* not handled */
307	if (flags & LINUX_MSG_PROXY)
308		;
309	if (flags & LINUX_MSG_FIN)
310		;
311	if (flags & LINUX_MSG_SYN)
312		;
313	if (flags & LINUX_MSG_CONFIRM)
314		;
315	if (flags & LINUX_MSG_RST)
316		;
317	if (flags & LINUX_MSG_ERRQUEUE)
318		;
319	if (flags & LINUX_MSG_NOSIGNAL)
320		;
321#endif
322	return ret_flags;
323}
324
325/*
326 * Allocate stackgap and put the converted sockaddr structure
327 * there, address on stackgap returned in sap.
328 */
329static int
330linux_sa_get(caddr_t *sgp, struct sockaddr **sap,
331	     const struct osockaddr *osa, int *osalen)
332{
333	struct sockaddr *sa, *usa;
334	int alloclen, error;
335
336	alloclen = *osalen;
337	error = do_sa_get(&sa, osa, &alloclen, M_TEMP);
338	if (error)
339		return (error);
340
341	usa = (struct sockaddr *) stackgap_alloc(sgp, alloclen);
342	if (!usa) {
343		error = ENOMEM;
344		goto out;
345	}
346
347	if ((error = copyout(sa, usa, alloclen)))
348		goto out;
349
350	*sap = usa;
351	*osalen = alloclen;
352
353out:
354	FREE(sa, M_TEMP);
355	return (error);
356}
357
358static int
359linux_sa_put(struct osockaddr *osa)
360{
361	struct osockaddr sa;
362	int error, bdom;
363
364	/*
365	 * Only read/write the osockaddr family part, the rest is
366	 * not changed.
367	 */
368	error = copyin(osa, &sa, sizeof(sa.sa_family));
369	if (error)
370		return (error);
371
372	bdom = bsd_to_linux_domain(sa.sa_family);
373	if (bdom == -1)
374		return (EINVAL);
375
376	sa.sa_family = bdom;
377	error = copyout(&sa, osa, sizeof(sa.sa_family));
378	if (error)
379		return (error);
380
381	return (0);
382}
383
384/* Return 0 if IP_HDRINCL is set for the given socket. */
385static int
386linux_check_hdrincl(struct thread *td, caddr_t *sg, int s)
387{
388	struct getsockopt_args /* {
389		int s;
390		int level;
391		int name;
392		caddr_t val;
393		int *avalsize;
394	} */ bsd_args;
395	int error;
396	caddr_t val, valsize;
397	int size_val = sizeof val;
398	int optval;
399
400	val = stackgap_alloc(sg, sizeof(int));
401	valsize = stackgap_alloc(sg, sizeof(int));
402
403	if ((error = copyout(&size_val, valsize, sizeof(size_val))))
404		return (error);
405
406	bsd_args.s = s;
407	bsd_args.level = IPPROTO_IP;
408	bsd_args.name = IP_HDRINCL;
409	bsd_args.val = val;
410	bsd_args.avalsize = (int *)valsize;
411	if ((error = getsockopt(td, &bsd_args)))
412		return (error);
413
414	if ((error = copyin(val, &optval, sizeof(optval))))
415		return (error);
416
417	return (optval == 0);
418}
419
420/*
421 * Updated sendto() when IP_HDRINCL is set:
422 * tweak endian-dependent fields in the IP packet.
423 */
424static int
425linux_sendto_hdrincl(struct thread *td, caddr_t *sg, struct sendto_args *bsd_args)
426{
427/*
428 * linux_ip_copysize defines how many bytes we should copy
429 * from the beginning of the IP packet before we customize it for BSD.
430 * It should include all the fields we modify (ip_len and ip_off)
431 * and be as small as possible to minimize copying overhead.
432 */
433#define linux_ip_copysize	8
434
435	struct ip *packet;
436	struct msghdr *msg;
437	struct iovec *iov;
438
439	int error;
440	struct  sendmsg_args /* {
441		int s;
442		caddr_t msg;
443		int flags;
444	} */ sendmsg_args;
445
446	/* Check the packet isn't too small before we mess with it */
447	if (bsd_args->len < linux_ip_copysize)
448		return (EINVAL);
449
450	/*
451	 * Tweaking the user buffer in place would be bad manners.
452	 * We create a corrected IP header with just the needed length,
453	 * then use an iovec to glue it to the rest of the user packet
454	 * when calling sendmsg().
455	 */
456	packet = (struct ip *)stackgap_alloc(sg, linux_ip_copysize);
457	msg = (struct msghdr *)stackgap_alloc(sg, sizeof(*msg));
458	iov = (struct iovec *)stackgap_alloc(sg, sizeof(*iov)*2);
459
460	/* Make a copy of the beginning of the packet to be sent */
461	if ((error = copyin(bsd_args->buf, packet, linux_ip_copysize)))
462		return (error);
463
464	/* Convert fields from Linux to BSD raw IP socket format */
465	packet->ip_len = bsd_args->len;
466	packet->ip_off = ntohs(packet->ip_off);
467
468	/* Prepare the msghdr and iovec structures describing the new packet */
469	msg->msg_name = bsd_args->to;
470	msg->msg_namelen = bsd_args->tolen;
471	msg->msg_iov = iov;
472	msg->msg_iovlen = 2;
473	msg->msg_control = NULL;
474	msg->msg_controllen = 0;
475	msg->msg_flags = 0;
476	iov[0].iov_base = (char *)packet;
477	iov[0].iov_len = linux_ip_copysize;
478	iov[1].iov_base = (char *)(bsd_args->buf) + linux_ip_copysize;
479	iov[1].iov_len = bsd_args->len - linux_ip_copysize;
480
481	sendmsg_args.s = bsd_args->s;
482	sendmsg_args.msg = (caddr_t)msg;
483	sendmsg_args.flags = bsd_args->flags;
484	return (sendmsg(td, &sendmsg_args));
485}
486
487struct linux_socket_args {
488	int domain;
489	int type;
490	int protocol;
491};
492
493static int
494linux_socket(struct thread *td, struct linux_socket_args *args)
495{
496	struct linux_socket_args linux_args;
497	struct socket_args /* {
498		int domain;
499		int type;
500		int protocol;
501	} */ bsd_args;
502	struct setsockopt_args /* {
503		int s;
504		int level;
505		int name;
506		caddr_t val;
507		int valsize;
508	} */ bsd_setsockopt_args;
509	int error;
510	int retval_socket;
511
512	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
513		return (error);
514
515	bsd_args.protocol = linux_args.protocol;
516	bsd_args.type = linux_args.type;
517	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
518	if (bsd_args.domain == -1)
519		return (EINVAL);
520
521	retval_socket = socket(td, &bsd_args);
522	if (bsd_args.type == SOCK_RAW
523	    && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
524	    && bsd_args.domain == AF_INET
525	    && retval_socket >= 0) {
526		/* It's a raw IP socket: set the IP_HDRINCL option. */
527		caddr_t sg;
528		int *hdrincl;
529
530		sg = stackgap_init();
531		hdrincl = (int *)stackgap_alloc(&sg, sizeof(*hdrincl));
532		*hdrincl = 1;
533		bsd_setsockopt_args.s = td->td_retval[0];
534		bsd_setsockopt_args.level = IPPROTO_IP;
535		bsd_setsockopt_args.name = IP_HDRINCL;
536		bsd_setsockopt_args.val = (caddr_t)hdrincl;
537		bsd_setsockopt_args.valsize = sizeof(*hdrincl);
538		/* We ignore any error returned by setsockopt() */
539		setsockopt(td, &bsd_setsockopt_args);
540		/* Copy back the return value from socket() */
541		td->td_retval[0] = bsd_setsockopt_args.s;
542	}
543#ifdef INET6
544	/*
545	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
546	 * default and some apps depend on this. So, set V6ONLY to 0
547	 * for Linux apps if the sysctl value is set to 1.
548	 */
549	if (bsd_args.domain == PF_INET6 && retval_socket >= 0
550#ifndef KLD_MODULE
551	    /*
552	     * XXX: Avoid undefined symbol error with an IPv4 only
553	     * kernel.
554	     */
555	    && ip6_v6only
556#endif
557	    ) {
558		caddr_t sg;
559		int *v6only;
560
561		sg = stackgap_init();
562		v6only = (int *)stackgap_alloc(&sg, sizeof(*v6only));
563		*v6only = 0;
564		bsd_setsockopt_args.s = td->td_retval[0];
565		bsd_setsockopt_args.level = IPPROTO_IPV6;
566		bsd_setsockopt_args.name = IPV6_V6ONLY;
567		bsd_setsockopt_args.val = (caddr_t)v6only;
568		bsd_setsockopt_args.valsize = sizeof(*v6only);
569		/* We ignore any error returned by setsockopt() */
570		setsockopt(td, &bsd_setsockopt_args);
571		/* Copy back the return value from socket() */
572		td->td_retval[0] = bsd_setsockopt_args.s;
573	}
574#endif
575
576	return (retval_socket);
577}
578
579struct linux_bind_args {
580	int s;
581	struct osockaddr *name;
582	int namelen;
583};
584
585static int
586linux_bind(struct thread *td, struct linux_bind_args *args)
587{
588	struct linux_bind_args linux_args;
589	struct sockaddr *sa;
590	int error;
591
592	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
593		return (error);
594
595	error = linux_getsockaddr(&sa, linux_args.name, linux_args.namelen);
596	if (error)
597		return (error);
598
599	return (kern_bind(td, linux_args.s, sa));
600}
601
602struct linux_connect_args {
603	int s;
604	struct osockaddr * name;
605	int namelen;
606};
607int linux_connect(struct thread *, struct linux_connect_args *);
608#endif /* !__alpha__*/
609
610int
611linux_connect(struct thread *td, struct linux_connect_args *args)
612{
613	struct linux_connect_args linux_args;
614	struct socket *so;
615	struct sockaddr *sa;
616	u_int fflag;
617	int error;
618
619#ifdef __alpha__
620	bcopy(args, &linux_args, sizeof(linux_args));
621#else
622	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
623		return (error);
624#endif /* __alpha__ */
625
626	error = linux_getsockaddr(&sa, (struct osockaddr *)linux_args.name,
627	    linux_args.namelen);
628	if (error)
629		return (error);
630
631	error = kern_connect(td, linux_args.s, sa);
632	if (error != EISCONN)
633		return (error);
634
635	/*
636	 * Linux doesn't return EISCONN the first time it occurs,
637	 * when on a non-blocking socket. Instead it returns the
638	 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
639	 */
640	if ((error = fgetsock(td, linux_args.s, &so, &fflag)) != 0)
641		return(error);
642	error = EISCONN;
643	if (fflag & FNONBLOCK) {
644		if (so->so_emuldata == 0)
645			error = so->so_error;
646		so->so_emuldata = (void *)1;
647	}
648	fputsock(so);
649	return (error);
650}
651
652#ifndef __alpha__
653
654struct linux_listen_args {
655	int s;
656	int backlog;
657};
658
659static int
660linux_listen(struct thread *td, struct linux_listen_args *args)
661{
662	struct linux_listen_args linux_args;
663	struct listen_args /* {
664		int s;
665		int backlog;
666	} */ bsd_args;
667	int error;
668
669	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
670		return (error);
671
672	bsd_args.s = linux_args.s;
673	bsd_args.backlog = linux_args.backlog;
674	return (listen(td, &bsd_args));
675}
676
677struct linux_accept_args {
678	int s;
679	struct osockaddr *addr;
680	int *namelen;
681};
682
683static int
684linux_accept(struct thread *td, struct linux_accept_args *args)
685{
686	struct linux_accept_args linux_args;
687	struct accept_args /* {
688		int s;
689		caddr_t name;
690		int *anamelen;
691	} */ bsd_args;
692	struct close_args /* {
693		int     fd;
694	} */ c_args;
695	struct fcntl_args /* {
696		int fd;
697		int cmd;
698		long arg;
699	} */ f_args;
700	int error;
701
702	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
703		return (error);
704
705	bsd_args.s = linux_args.s;
706	bsd_args.name = (caddr_t)linux_args.addr;
707	bsd_args.anamelen = linux_args.namelen;
708	error = oaccept(td, &bsd_args);
709	if (error)
710		return (error);
711	if (linux_args.addr) {
712		error = linux_sa_put(linux_args.addr);
713		if (error) {
714			c_args.fd = td->td_retval[0];
715			(void)close(td, &c_args);
716			return (error);
717		}
718	}
719
720	/*
721	 * linux appears not to copy flags from the parent socket to the
722	 * accepted one, so we must clear the flags in the new descriptor.
723	 * Ignore any errors, because we already have an open fd.
724	 */
725	f_args.fd = td->td_retval[0];
726	f_args.cmd = F_SETFL;
727	f_args.arg = 0;
728	(void)fcntl(td, &f_args);
729	td->td_retval[0] = f_args.fd;
730	return (0);
731}
732
733struct linux_getsockname_args {
734	int s;
735	struct osockaddr *addr;
736	int *namelen;
737};
738
739static int
740linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
741{
742	struct linux_getsockname_args linux_args;
743	struct getsockname_args /* {
744		int fdes;
745		caddr_t asa;
746		int *alen;
747	} */ bsd_args;
748	int error;
749
750	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
751		return (error);
752
753	bsd_args.fdes = linux_args.s;
754	bsd_args.asa = (caddr_t) linux_args.addr;
755	bsd_args.alen = linux_args.namelen;
756	error = ogetsockname(td, &bsd_args);
757	if (error)
758		return (error);
759	error = linux_sa_put(linux_args.addr);
760	if (error)
761		return (error);
762	return (0);
763}
764
765struct linux_getpeername_args {
766	int s;
767	struct osockaddr *addr;
768	int *namelen;
769};
770
771static int
772linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
773{
774	struct linux_getpeername_args linux_args;
775	struct ogetpeername_args /* {
776		int fdes;
777		caddr_t asa;
778		int *alen;
779	} */ bsd_args;
780	int error;
781
782	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
783		return (error);
784
785	bsd_args.fdes = linux_args.s;
786	bsd_args.asa = (caddr_t) linux_args.addr;
787	bsd_args.alen = linux_args.namelen;
788	error = ogetpeername(td, &bsd_args);
789	if (error)
790		return (error);
791	error = linux_sa_put(linux_args.addr);
792	if (error)
793		return (error);
794	return (0);
795}
796
797struct linux_socketpair_args {
798	int domain;
799	int type;
800	int protocol;
801	int *rsv;
802};
803
804static int
805linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
806{
807	struct linux_socketpair_args linux_args;
808	struct socketpair_args /* {
809		int domain;
810		int type;
811		int protocol;
812		int *rsv;
813	} */ bsd_args;
814	int error;
815
816	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
817		return (error);
818
819	bsd_args.domain = linux_to_bsd_domain(linux_args.domain);
820	if (bsd_args.domain == -1)
821		return (EINVAL);
822
823	bsd_args.type = linux_args.type;
824	bsd_args.protocol = linux_args.protocol;
825	bsd_args.rsv = linux_args.rsv;
826	return (socketpair(td, &bsd_args));
827}
828
829struct linux_send_args {
830	int s;
831	void *msg;
832	int len;
833	int flags;
834};
835
836static int
837linux_send(struct thread *td, struct linux_send_args *args)
838{
839	struct linux_send_args linux_args;
840	struct osend_args /* {
841		int s;
842		caddr_t buf;
843		int len;
844		int flags;
845	} */ bsd_args;
846	int error;
847
848	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
849		return (error);
850
851	bsd_args.s = linux_args.s;
852	bsd_args.buf = linux_args.msg;
853	bsd_args.len = linux_args.len;
854	bsd_args.flags = linux_args.flags;
855	return (osend(td, &bsd_args));
856}
857
858struct linux_recv_args {
859	int s;
860	void *msg;
861	int len;
862	int flags;
863};
864
865static int
866linux_recv(struct thread *td, struct linux_recv_args *args)
867{
868	struct linux_recv_args linux_args;
869	struct orecv_args /* {
870		int s;
871		caddr_t buf;
872		int len;
873		int flags;
874	} */ bsd_args;
875	int error;
876
877	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
878		return (error);
879
880	bsd_args.s = linux_args.s;
881	bsd_args.buf = linux_args.msg;
882	bsd_args.len = linux_args.len;
883	bsd_args.flags = linux_args.flags;
884	return (orecv(td, &bsd_args));
885}
886
887struct linux_sendto_args {
888	int s;
889	void *msg;
890	int len;
891	int flags;
892	caddr_t to;
893	int tolen;
894};
895
896static int
897linux_sendto(struct thread *td, struct linux_sendto_args *args)
898{
899	struct linux_sendto_args linux_args;
900	struct sendto_args /* {
901		int s;
902		caddr_t buf;
903		size_t len;
904		int flags;
905		caddr_t to;
906		int tolen;
907	} */ bsd_args;
908	caddr_t sg = stackgap_init();
909	struct sockaddr *to;
910	int tolen, error;
911
912	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
913		return (error);
914
915	tolen = linux_args.tolen;
916	if (linux_args.to) {
917		error = linux_sa_get(&sg, &to,
918		    (struct osockaddr *) linux_args.to, &tolen);
919		if (error)
920			return (error);
921	} else
922		to = NULL;
923
924	bsd_args.s = linux_args.s;
925	bsd_args.buf = linux_args.msg;
926	bsd_args.len = linux_args.len;
927	bsd_args.flags = linux_args.flags;
928	bsd_args.to = (caddr_t) to;
929	bsd_args.tolen = (unsigned int) tolen;
930
931	if (linux_check_hdrincl(td, &sg, linux_args.s) == 0)
932		/* IP_HDRINCL set, tweak the packet before sending */
933		return (linux_sendto_hdrincl(td, &sg, &bsd_args));
934
935	return (sendto(td, &bsd_args));
936}
937
938struct linux_recvfrom_args {
939	int s;
940	void *buf;
941	int len;
942	int flags;
943	caddr_t from;
944	int *fromlen;
945};
946
947static int
948linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
949{
950	struct linux_recvfrom_args linux_args;
951	struct recvfrom_args /* {
952		int s;
953		caddr_t buf;
954		size_t len;
955		int flags;
956		caddr_t from;
957		int *fromlenaddr;
958	} */ bsd_args;
959	int error;
960
961	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
962		return (error);
963
964	bsd_args.s = linux_args.s;
965	bsd_args.buf = linux_args.buf;
966	bsd_args.len = linux_args.len;
967	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
968	bsd_args.from = linux_args.from;
969	bsd_args.fromlenaddr = linux_args.fromlen;
970	error = orecvfrom(td, &bsd_args);
971	if (error)
972		return (error);
973	if (linux_args.from) {
974		error = linux_sa_put((struct osockaddr *) linux_args.from);
975		if (error)
976			return (error);
977	}
978	return (0);
979}
980
981struct linux_sendmsg_args {
982	int s;
983	const struct msghdr *msg;
984	int flags;
985};
986
987static int
988linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
989{
990	struct linux_sendmsg_args linux_args;
991	struct sendmsg_args /* {
992		int s;
993		const struct msghdr *msg;
994		int flags;
995	} */ bsd_args;
996	struct msghdr msg;
997	struct msghdr *nmsg = NULL;
998	int error;
999	int level;
1000	caddr_t control;
1001
1002	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1003		return (error);
1004
1005	error = copyin(linux_args.msg, &msg, sizeof(msg));
1006	if (error)
1007		return (error);
1008
1009	if (msg.msg_name) {
1010		struct sockaddr *sa;
1011		caddr_t sg = stackgap_init();
1012
1013		nmsg = (struct msghdr *) stackgap_alloc(&sg,
1014		    sizeof(struct msghdr));
1015		if (!nmsg)
1016			return (ENOMEM);
1017
1018		error = linux_sa_get(&sg, &sa,
1019		    (struct osockaddr *) msg.msg_name, &msg.msg_namelen);
1020		if (error)
1021			return (error);
1022
1023		msg.msg_name = (struct sockaddr *) sa;
1024		error = copyout(&msg, nmsg, sizeof(struct msghdr));
1025		if (error)
1026			return (error);
1027	}
1028
1029	error = copyin(&linux_args.msg->msg_control, &control,
1030	    sizeof(caddr_t));
1031	if (error)
1032		return (error);
1033
1034	if (control == NULL)
1035		goto done;
1036
1037	error = copyin(&((struct cmsghdr*)control)->cmsg_level, &level,
1038	    sizeof(int));
1039	if (error)
1040		return (error);
1041
1042	if (level == 1) {
1043		/*
1044		 * Linux thinks that SOL_SOCKET is 1; we know
1045		 * that it's really 0xffff, of course.
1046		 */
1047		level = SOL_SOCKET;
1048		error = copyout(&level,
1049		    &((struct cmsghdr *)control)->cmsg_level, sizeof(int));
1050		if (error)
1051			return (error);
1052	}
1053done:
1054	bsd_args.s = linux_args.s;
1055	bsd_args.msg = (caddr_t)nmsg;
1056	bsd_args.flags = linux_args.flags;
1057	return (sendmsg(td, &bsd_args));
1058}
1059
1060struct linux_recvmsg_args {
1061	int s;
1062	struct msghdr *msg;
1063	int flags;
1064};
1065
1066static int
1067linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1068{
1069	struct linux_recvmsg_args linux_args;
1070	struct recvmsg_args /* {
1071		int	s;
1072		struct	msghdr *msg;
1073		int	flags;
1074	} */ bsd_args;
1075	struct msghdr msg;
1076	int error;
1077
1078	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1079		return (error);
1080
1081	bsd_args.s = linux_args.s;
1082	bsd_args.msg = linux_args.msg;
1083	bsd_args.flags = linux_to_bsd_msg_flags(linux_args.flags);
1084	error = recvmsg(td, &bsd_args);
1085	if (error)
1086		return (error);
1087
1088	error = copyin(linux_args.msg, &msg, sizeof(msg));
1089	if (error)
1090		return (error);
1091	if (msg.msg_name && msg.msg_namelen > 2)
1092		error = linux_sa_put(msg.msg_name);
1093	return (error);
1094}
1095
1096struct linux_shutdown_args {
1097	int s;
1098	int how;
1099};
1100
1101static int
1102linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1103{
1104	struct linux_shutdown_args linux_args;
1105	struct shutdown_args /* {
1106		int s;
1107		int how;
1108	} */ bsd_args;
1109	int error;
1110
1111	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1112		return (error);
1113
1114	bsd_args.s = linux_args.s;
1115	bsd_args.how = linux_args.how;
1116	return (shutdown(td, &bsd_args));
1117}
1118
1119struct linux_setsockopt_args {
1120	int s;
1121	int level;
1122	int optname;
1123	void *optval;
1124	int optlen;
1125};
1126
1127static int
1128linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1129{
1130	struct linux_setsockopt_args linux_args;
1131	struct setsockopt_args /* {
1132		int s;
1133		int level;
1134		int name;
1135		caddr_t val;
1136		int valsize;
1137	} */ bsd_args;
1138	int error, name;
1139
1140	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1141		return (error);
1142
1143	bsd_args.s = linux_args.s;
1144	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1145	switch (bsd_args.level) {
1146	case SOL_SOCKET:
1147		name = linux_to_bsd_so_sockopt(linux_args.optname);
1148		break;
1149	case IPPROTO_IP:
1150		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1151		break;
1152	case IPPROTO_TCP:
1153		/* Linux TCP option values match BSD's */
1154		name = linux_args.optname;
1155		break;
1156	default:
1157		name = -1;
1158		break;
1159	}
1160	if (name == -1)
1161		return (EINVAL);
1162
1163	bsd_args.name = name;
1164	bsd_args.val = linux_args.optval;
1165	bsd_args.valsize = linux_args.optlen;
1166	return (setsockopt(td, &bsd_args));
1167}
1168
1169struct linux_getsockopt_args {
1170	int s;
1171	int level;
1172	int optname;
1173	void *optval;
1174	int *optlen;
1175};
1176
1177static int
1178linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1179{
1180	struct linux_getsockopt_args linux_args;
1181	struct getsockopt_args /* {
1182		int s;
1183		int level;
1184		int name;
1185		caddr_t val;
1186		int *avalsize;
1187	} */ bsd_args;
1188	int error, name;
1189
1190	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
1191		return (error);
1192
1193	bsd_args.s = linux_args.s;
1194	bsd_args.level = linux_to_bsd_sockopt_level(linux_args.level);
1195	switch (bsd_args.level) {
1196	case SOL_SOCKET:
1197		name = linux_to_bsd_so_sockopt(linux_args.optname);
1198		break;
1199	case IPPROTO_IP:
1200		name = linux_to_bsd_ip_sockopt(linux_args.optname);
1201		break;
1202	case IPPROTO_TCP:
1203		/* Linux TCP option values match BSD's */
1204		name = linux_args.optname;
1205		break;
1206	default:
1207		name = -1;
1208		break;
1209	}
1210	if (name == -1)
1211		return (EINVAL);
1212
1213	bsd_args.name = name;
1214	bsd_args.val = linux_args.optval;
1215	bsd_args.avalsize = linux_args.optlen;
1216	return (getsockopt(td, &bsd_args));
1217}
1218
1219int
1220linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1221{
1222	void *arg = (void *)args->args;
1223
1224	switch (args->what) {
1225	case LINUX_SOCKET:
1226		return (linux_socket(td, arg));
1227	case LINUX_BIND:
1228		return (linux_bind(td, arg));
1229	case LINUX_CONNECT:
1230		return (linux_connect(td, arg));
1231	case LINUX_LISTEN:
1232		return (linux_listen(td, arg));
1233	case LINUX_ACCEPT:
1234		return (linux_accept(td, arg));
1235	case LINUX_GETSOCKNAME:
1236		return (linux_getsockname(td, arg));
1237	case LINUX_GETPEERNAME:
1238		return (linux_getpeername(td, arg));
1239	case LINUX_SOCKETPAIR:
1240		return (linux_socketpair(td, arg));
1241	case LINUX_SEND:
1242		return (linux_send(td, arg));
1243	case LINUX_RECV:
1244		return (linux_recv(td, arg));
1245	case LINUX_SENDTO:
1246		return (linux_sendto(td, arg));
1247	case LINUX_RECVFROM:
1248		return (linux_recvfrom(td, arg));
1249	case LINUX_SHUTDOWN:
1250		return (linux_shutdown(td, arg));
1251	case LINUX_SETSOCKOPT:
1252		return (linux_setsockopt(td, arg));
1253	case LINUX_GETSOCKOPT:
1254		return (linux_getsockopt(td, arg));
1255	case LINUX_SENDMSG:
1256		return (linux_sendmsg(td, arg));
1257	case LINUX_RECVMSG:
1258		return (linux_recvmsg(td, arg));
1259	}
1260
1261	uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1262	return (ENOSYS);
1263}
1264#endif	/*!__alpha__*/
1265