1 /*	$OpenBSD: linux_socket.c,v 1.33 2005/06/07 02:30:47 henning Exp $	*/
2 /*	$NetBSD: linux_socket.c,v 1.14 1996/04/05 00:01:50 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1995 Frank van der Linden
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *      This product includes software developed for the NetBSD Project
19  *      by Frank van der Linden
20  * 4. The name of the author may not be used to endorse or promote products
21  *    derived from this software without specific prior written permission
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
24  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
25  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
26  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
27  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
28  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
32  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33  */
34 
35 #include <sys/param.h>
36 #include <sys/kernel.h>
37 #include <sys/systm.h>
38 #include <sys/buf.h>
39 #include <sys/malloc.h>
40 #include <sys/ioctl.h>
41 #include <sys/tty.h>
42 #include <sys/file.h>
43 #include <sys/filedesc.h>
44 #include <sys/select.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <net/if.h>
48 #include <net/if_types.h>
49 #include <net/if_dl.h>
50 #include <netinet/in.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/ip.h>
53 #include <netinet/tcp.h>
54 #include <sys/mount.h>
55 #include <sys/proc.h>
56 #include <sys/vnode.h>
57 #include <sys/device.h>
58 
59 #include <sys/syscallargs.h>
60 
61 #include <compat/linux/linux_types.h>
62 #include <compat/linux/linux_util.h>
63 #include <compat/linux/linux_signal.h>
64 #include <compat/linux/linux_syscallargs.h>
65 #include <compat/linux/linux_ioctl.h>
66 #include <compat/linux/linux_socket.h>
67 #include <compat/linux/linux_socketcall.h>
68 #include <compat/linux/linux_sockio.h>
69 
70 /*
71  * All the calls in this file are entered via one common system
72  * call in Linux, represented here by linux_socketcall()
73  * Arguments for the various calls are on the user stack. A pointer
74  * to them is the only thing that is passed. It is up to the various
75  * calls to copy them in themselves. To make it look better, they
76  * are copied to structures.
77  */
78 
79 static int linux_to_bsd_domain (int);
80 static int bsd_to_linux_domain(int);
81 
82 int linux_socket(struct proc *, void *, register_t *);
83 int linux_bind(struct proc *, void *, register_t *);
84 int linux_connect(struct proc *, void *, register_t *);
85 int linux_listen(struct proc *, void *, register_t *);
86 int linux_accept(struct proc *, void *, register_t *);
87 int linux_getsockname(struct proc *, void *, register_t *);
88 int linux_getpeername(struct proc *, void *, register_t *);
89 int linux_socketpair(struct proc *, void *, register_t *);
90 int linux_send(struct proc *, void *, register_t *);
91 int linux_recv(struct proc *, void *, register_t *);
92 int linux_sendto(struct proc *, void *, register_t *);
93 int linux_recvfrom(struct proc *, void *, register_t *);
94 int linux_shutdown(struct proc *, void *, register_t *);
95 int linux_to_bsd_sopt_level(int);
96 int linux_to_bsd_so_sockopt(int);
97 int linux_to_bsd_ip_sockopt(int);
98 int linux_to_bsd_tcp_sockopt(int);
99 int linux_to_bsd_udp_sockopt(int);
100 int linux_setsockopt(struct proc *, void *, register_t *);
101 int linux_getsockopt(struct proc *, void *, register_t *);
102 int linux_recvmsg(struct proc *, void *, register_t *);
103 int linux_sendmsg(struct proc *, void *, register_t *);
104 
105 int linux_check_hdrincl(struct proc *, int, register_t *, caddr_t *);
106 int linux_sendto_hdrincl(struct proc *, struct sys_sendto_args *,
107     register_t *, caddr_t *);
108 
109 int linux_sa_get(struct proc *, caddr_t *, struct sockaddr **,
110     const struct osockaddr *, int *);
111 int linux_sa_put(struct osockaddr *);
112 
113 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
114 	AF_UNSPEC,
115 	AF_UNIX,
116 	AF_INET,
117 	-1,		/* LINUX_AF_AX25 */
118 	AF_IPX,
119 	AF_APPLETALK,
120 	-1,		/* LINUX_AF_NETROM */
121 	-1,		/* LINUX_AF_BRIDGE */
122 	-1,		/* LINUX_AF_ATMPVC */
123 	-1,		/* LINUX_AF_X25 */
124 	AF_INET6,
125 	-1,		/* LINUX_AF_ROSE */
126 	AF_DECnet,
127 	-1,		/* LINUX_AF_NETBEUI */
128 	-1,		/* LINUX_AF_SECURITY */
129 	-1,		/* pseudo_AF_KEY */
130 	AF_ROUTE,	/* LINUX_AF_NETLINK */
131 	-1,		/* LINUX_AF_PACKET */
132 	-1,		/* LINUX_AF_ASH */
133 	-1,		/* LINUX_AF_ECONET */
134 	-1,		/* LINUX_AF_ATMSVC */
135 	AF_SNA,
136 	/* rest up to LINUX_AF_MAX-1 is not allocated */
137 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
138 };
139 
140 static const int bsd_to_linux_domain_[AF_MAX] = {
141 	LINUX_AF_UNSPEC,
142 	LINUX_AF_UNIX,
143 	LINUX_AF_INET,
144 	-1,		/* AF_IMPLINK */
145 	-1,		/* AF_PUP */
146 	-1,		/* AF_CHAOS */
147 	-1,		/* AF_NS */
148 	-1,		/* AF_ISO */
149 	-1,		/* AF_ECMA */
150 	-1,		/* AF_DATAKIT */
151 	-1,		/* AF_CCITT */
152 	-1,		/* LINUX_AF_SNA */
153 	-1,		/* LINUX_AF_DECnet */
154 	-1,		/* AF_DLI */
155 	-1,		/* AF_LAT */
156 	-1,		/* AF_HYLINK */
157 	LINUX_AF_APPLETALK,
158 	-1,		/* LINUX_AF_NETLINK */
159 	-1,		/* AF_LINK */
160 	-1,		/* AF_XTP */
161 	-1,		/* AF_COIP */
162 	-1,		/* AF_CNT */
163 	-1,		/* pseudo_AF_RTIP */
164 	LINUX_AF_IPX,
165 	LINUX_AF_INET6,
166 	-1,		/* pseudo_AF_PIP */
167 	-1,		/* AF_ISDN */
168 	-1,		/* AF_NATM */
169 	-1,		/* AF_ARP */
170 	-1,		/* LINUX_pseudo_AF_KEY */
171 	-1,		/* pseudo_AF_HDRCMPLT */
172 };
173 
174 /*
175  * Convert between Linux and BSD socket domain values
176  */
177 static int
linux_to_bsd_domain(ldom)178 linux_to_bsd_domain(ldom)
179 	int ldom;
180 {
181 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
182 		return (-1);
183 
184 	return linux_to_bsd_domain_[ldom];
185 }
186 
187 /*
188  * Convert between BSD and Linux socket domain values
189  */
190 static int
bsd_to_linux_domain(bdom)191 bsd_to_linux_domain(bdom)
192 	int bdom;
193 {
194 	if (bdom < 0 || bdom >= AF_MAX)
195 		return (-1);
196 
197 	return bsd_to_linux_domain_[bdom];
198 }
199 
200 int
linux_socket(p,v,retval)201 linux_socket(p, v, retval)
202 	struct proc *p;
203 	void *v;
204 	register_t *retval;
205 {
206 	struct linux_socket_args /* {
207 		syscallarg(int)	domain;
208 		syscallarg(int)	type;
209 		syscallarg(int) protocol;
210 	} */ *uap = v;
211 	struct linux_socket_args lsa;
212 	struct sys_socket_args bsa;
213 	int error;
214 
215 	if ((error = copyin((caddr_t) uap, (caddr_t) &lsa, sizeof lsa)))
216 		return error;
217 
218 	SCARG(&bsa, protocol) = lsa.protocol;
219 	SCARG(&bsa, type) = lsa.type;
220 	SCARG(&bsa, domain) = linux_to_bsd_domain(lsa.domain);
221 	if (SCARG(&bsa, domain) == -1)
222 		return EINVAL;
223 	return sys_socket(p, &bsa, retval);
224 }
225 
226 int
linux_bind(p,v,retval)227 linux_bind(p, v, retval)
228 	struct proc *p;
229 	void *v;
230 	register_t *retval;
231 {
232 	struct linux_bind_args /* {
233 		syscallarg(int)	s;
234 		syscallarg(struct sockaddr *) name;
235 		syscallarg(int)	namelen;
236 	} */ *uap = v;
237 	struct linux_bind_args lba;
238 	struct sys_bind_args bba;
239 	int error;
240 	int namlen;
241 
242 	if ((error = copyin((caddr_t) uap, (caddr_t) &lba, sizeof lba)))
243 		return error;
244 
245 	SCARG(&bba, s) = lba.s;
246 	namlen = lba.namelen;
247 	if (lba.name) {
248 		struct sockaddr *sa;
249 		caddr_t sg = stackgap_init(p->p_emul);
250 
251 		error = linux_sa_get(p, &sg, &sa, lba.name, &namlen);
252 		if (error)
253 			return (error);
254 		SCARG(&bba, name) = sa;
255 	} else
256 		SCARG(&bba, name) = NULL;
257 	SCARG(&bba, namelen) = namlen;
258 
259 	return sys_bind(p, &bba, retval);
260 }
261 
262 int
linux_connect(p,v,retval)263 linux_connect(p, v, retval)
264 	struct proc *p;
265 	void *v;
266 	register_t *retval;
267 {
268 	struct linux_connect_args /* {
269 		syscallarg(int)	s;
270 		syscallarg(struct osockaddr *) name;
271 		syscallarg(int)	namelen;
272 	} */ *uap = v;
273 	struct linux_connect_args lca;
274 	struct sys_connect_args bca;
275 	struct sockaddr *sa;
276 	caddr_t sg = stackgap_init(p->p_emul);
277 	int namlen;
278 	int error;
279 
280 	if ((error = copyin((caddr_t) uap, (caddr_t) &lca, sizeof lca)))
281 		return error;
282 
283 	namlen = lca.namelen;
284 	error = linux_sa_get(p, &sg, &sa, lca.name, &namlen);
285 	if (error)
286 		return (error);
287 
288 	SCARG(&bca, s) = lca.s;
289 	SCARG(&bca, name) = sa;
290 	SCARG(&bca, namelen) = (unsigned int)namlen;
291 
292 	error = sys_connect(p, &bca, retval);
293 
294 	if (error == EISCONN) {
295 		struct sys_getsockopt_args bga;
296 #if 0
297 		struct sys_fcntl_args fca;
298 #endif
299 		void *status, *statusl;
300 		int stat, statl = sizeof stat;
301 
302 #if 0
303 		SCARG(&fca, fd) = lca.s;
304 		SCARG(&fca, cmd) = F_GETFL;
305 		SCARG(&fca, arg) = 0;
306 		if (sys_fcntl(p, &fca, retval) == -1 ||
307 		    (*retval & O_NONBLOCK) == 0)
308 			return error;
309 #endif
310 
311 		status = stackgap_alloc(&sg, sizeof stat);
312 		statusl = stackgap_alloc(&sg, sizeof statusl);
313 
314 		if ((error = copyout(&statl, statusl, sizeof statl)))
315 			return error;
316 
317 		SCARG(&bga, s) = lca.s;
318 		SCARG(&bga, level) = SOL_SOCKET;
319 		SCARG(&bga, name) = SO_ERROR;
320 		SCARG(&bga, val) = status;
321 		SCARG(&bga, avalsize) = statusl;
322 
323 		error = sys_getsockopt(p, &bga, retval);
324 		if (error)
325 			return error;
326 		if ((error = copyin(status, &stat, sizeof stat)))
327 			return error;
328 		return stat;
329 	}
330 	return error;
331 }
332 
333 int
linux_listen(p,v,retval)334 linux_listen(p, v, retval)
335 	struct proc *p;
336 	void *v;
337 	register_t *retval;
338 {
339 	struct linux_listen_args /* {
340 		syscallarg(int) s;
341 		syscallarg(int) backlog;
342 	} */ *uap = v;
343 	struct linux_listen_args lla;
344 	struct sys_listen_args bla;
345 	int error;
346 
347 	if ((error = copyin((caddr_t) uap, (caddr_t) &lla, sizeof lla)))
348 		return error;
349 
350 	SCARG(&bla, s) = lla.s;
351 	SCARG(&bla, backlog) = lla.backlog;
352 
353 	return sys_listen(p, &bla, retval);
354 }
355 
356 int
linux_accept(p,v,retval)357 linux_accept(p, v, retval)
358 	struct proc *p;
359 	void *v;
360 	register_t *retval;
361 {
362 	struct linux_accept_args /* {
363 		syscallarg(int) s;
364 		syscallarg(struct sockaddr *) addr;
365 		syscallarg(int *) namelen;
366 	} */ *uap = v;
367 	struct linux_accept_args laa;
368 	struct compat_43_sys_accept_args baa;
369 	struct sys_fcntl_args fca;
370 	int error;
371 
372 	if ((error = copyin((caddr_t) uap, (caddr_t) &laa, sizeof laa)))
373 		return error;
374 
375 	SCARG(&baa, s) = laa.s;
376 	SCARG(&baa, name) = (caddr_t) laa.addr;
377 	SCARG(&baa, anamelen) = laa.namelen;
378 
379 	error = compat_43_sys_accept(p, &baa, retval);
380 	if (error)
381 		return (error);
382 
383 	/*
384 	 * linux appears not to copy flags from the parent socket to the
385 	 * accepted one, so we must clear the flags in the new descriptor.
386 	 * Ignore any errors, because we already have an open fd.
387 	 */
388 	SCARG(&fca, fd) = *retval;
389 	SCARG(&fca, cmd) = F_SETFL;
390 	SCARG(&fca, arg) = 0;
391 	(void)sys_fcntl(p, &fca, retval);
392 	*retval = SCARG(&fca, fd);
393 	return (0);
394 }
395 
396 int
linux_getsockname(p,v,retval)397 linux_getsockname(p, v, retval)
398 	struct proc *p;
399 	void *v;
400 	register_t *retval;
401 {
402 	struct linux_getsockname_args /* {
403 		syscallarg(int) s;
404 		syscallarg(caddr_t) addr;
405 		syscallarg(int *) namelen;
406 	} */ *uap = v;
407 	struct linux_getsockname_args lga;
408 	struct sys_getsockname_args bga;
409 	int error;
410 
411 	if ((error = copyin((caddr_t) uap, (caddr_t) &lga, sizeof lga)))
412 		return error;
413 
414 	SCARG(&bga, fdes) = lga.s;
415 	SCARG(&bga, asa) = (struct sockaddr *) lga.addr;
416 	SCARG(&bga, alen) = lga.namelen;
417 
418 	error = sys_getsockname(p, &bga, retval);
419 	if (error)
420 		return (error);
421 
422 	if ((error = linux_sa_put((struct osockaddr *)lga.addr)))
423 		return (error);
424 
425 	return (0);
426 }
427 
428 int
linux_getpeername(p,v,retval)429 linux_getpeername(p, v, retval)
430 	struct proc *p;
431 	void *v;
432 	register_t *retval;
433 {
434 	struct linux_getpeername_args /* {
435 		syscallarg(int) s;
436 		syscallarg(struct sockaddr *) addr;
437 		syscallarg(int *) namelen;
438 	} */ *uap = v;
439 	struct linux_getpeername_args lga;
440 	struct sys_getpeername_args bga;
441 	int error;
442 
443 	if ((error = copyin((caddr_t) uap, (caddr_t) &lga, sizeof lga)))
444 		return error;
445 
446 	SCARG(&bga, fdes) = lga.s;
447 	SCARG(&bga, asa) = (struct sockaddr *) lga.addr;
448 	SCARG(&bga, alen) = lga.namelen;
449 
450 	error = sys_getpeername(p, &bga, retval);
451 	if (error)
452 		return (error);
453 
454 	if ((error = linux_sa_put((struct osockaddr *)lga.addr)))
455 		return (error);
456 
457 	return (0);
458 }
459 
460 int
linux_socketpair(p,v,retval)461 linux_socketpair(p, v, retval)
462 	struct proc *p;
463 	void *v;
464 	register_t *retval;
465 {
466 	struct linux_socketpair_args /* {
467 		syscallarg(int) domain;
468 		syscallarg(int) type;
469 		syscallarg(int) protocol;
470 		syscallarg(int *) rsv;
471 	} */ *uap = v;
472 	struct linux_socketpair_args lsa;
473 	struct sys_socketpair_args bsa;
474 	int error;
475 
476 	if ((error = copyin((caddr_t) uap, &lsa, sizeof lsa)))
477 		return error;
478 
479 	SCARG(&bsa, domain) = linux_to_bsd_domain(lsa.domain);
480 	if (SCARG(&bsa, domain) == -1)
481 		return EINVAL;
482 	SCARG(&bsa, type) = lsa.type;
483 	SCARG(&bsa, protocol) = lsa.protocol;
484 	SCARG(&bsa, rsv) = lsa.rsv;
485 
486 	return sys_socketpair(p, &bsa, retval);
487 }
488 
489 int
linux_send(p,v,retval)490 linux_send(p, v, retval)
491 	struct proc *p;
492 	void *v;
493 	register_t *retval;
494 {
495 	struct linux_send_args /* {
496 		syscallarg(int) s;
497 		syscallarg(void *) msg;
498 		syscallarg(int) len;
499 		syscallarg(int) flags;
500 	} */ *uap = v;
501 	struct linux_send_args lsa;
502 	struct compat_43_sys_send_args bsa;
503 	int error;
504 
505 	if ((error = copyin((caddr_t) uap, (caddr_t) &lsa, sizeof lsa)))
506 		return error;
507 
508 	SCARG(&bsa, s) = lsa.s;
509 	SCARG(&bsa, buf) = lsa.msg;
510 	SCARG(&bsa, len) = lsa.len;
511 	SCARG(&bsa, flags) = lsa.flags;
512 
513 	return compat_43_sys_send(p, &bsa, retval);
514 }
515 
516 int
linux_recv(p,v,retval)517 linux_recv(p, v, retval)
518 	struct proc *p;
519 	void *v;
520 	register_t *retval;
521 {
522 	struct linux_recv_args /* {
523 		syscallarg(int) s;
524 		syscallarg(void *) msg;
525 		syscallarg(int) len;
526 		syscallarg(int) flags;
527 	} */ *uap = v;
528 	struct linux_recv_args lra;
529 	struct compat_43_sys_recv_args bra;
530 	int error;
531 
532 	if ((error = copyin((caddr_t) uap, (caddr_t) &lra, sizeof lra)))
533 		return error;
534 
535 	SCARG(&bra, s) = lra.s;
536 	SCARG(&bra, buf) = lra.msg;
537 	SCARG(&bra, len) = lra.len;
538 	SCARG(&bra, flags) = lra.flags;
539 
540 	return compat_43_sys_recv(p, &bra, retval);
541 }
542 
543 int
linux_check_hdrincl(p,fd,retval,sgp)544 linux_check_hdrincl(p, fd, retval, sgp)
545 	struct proc *p;
546 	int fd;
547 	register_t *retval;
548 	caddr_t *sgp;
549 {
550 	struct sys_getsockopt_args /* {
551 		int s;
552 		int level;
553 		int name;
554 		caddr_t val;
555 		int *avalsize;
556 	} */ gsa;
557 	int error;
558 	caddr_t val;
559 	int *valsize;
560 	int size_val = sizeof val;
561 	int optval;
562 
563 	val = stackgap_alloc(sgp, sizeof(optval));
564 	valsize = stackgap_alloc(sgp, sizeof(size_val));
565 
566 	if ((error = copyout(&size_val, valsize, sizeof(size_val))))
567 		return (error);
568 	SCARG(&gsa, s) = fd;
569 	SCARG(&gsa, level) = IPPROTO_IP;
570 	SCARG(&gsa, name) = IP_HDRINCL;
571 	SCARG(&gsa, val) = val;
572 	SCARG(&gsa, avalsize) = valsize;
573 
574 	if ((error = sys_getsockopt(p, &gsa, retval)))
575 		return (error);
576 	if ((error = copyin(val, &optval, sizeof(optval))))
577 		return (error);
578 	return (optval == 0);
579 }
580 
581 /*
582  * linux_ip_copysize defines how many bytes we should copy
583  * from the beginning of the IP packet before we customize it for BSD.
584  * It should include all the fields we modify (ip_len and ip_off)
585  * and be as small as possible to minimize copying overhead.
586  */
587 #define linux_ip_copysize      8
588 
589 int
linux_sendto_hdrincl(p,bsa,retval,sgp)590 linux_sendto_hdrincl(p, bsa, retval, sgp)
591 	struct proc *p;
592 	struct sys_sendto_args *bsa;
593 	register_t *retval;
594 	caddr_t *sgp;
595 {
596 	struct sys_sendmsg_args ssa;
597 	struct ip *packet, rpacket;
598 	struct msghdr *msg, rmsg;
599 	struct iovec *iov, riov[2];
600 	int error;
601 
602 	/* Check the packet isn't too small before we mess with it */
603 	if (SCARG(bsa, len) < linux_ip_copysize)
604 		return EINVAL;
605 
606 	/*
607 	 * Tweaking the user buffer in place would be bad manners.
608 	 * We create a corrected IP header with just the needed length,
609 	 * then use an iovec to glue it to the rest of the user packet
610 	 * when calling sendmsg().
611 	 */
612 	packet = (struct ip *)stackgap_alloc(sgp, linux_ip_copysize);
613 	msg = (struct msghdr *)stackgap_alloc(sgp, sizeof(*msg));
614 	iov = (struct iovec *)stackgap_alloc(sgp, sizeof(*iov)*2);
615 
616 	/* Make a copy of the beginning of the packet to be sent */
617 	if ((error = copyin(SCARG(bsa, buf), (caddr_t)&rpacket,
618 	    linux_ip_copysize)))
619 		return error;
620 
621 	/* Convert fields from Linux to BSD raw IP socket format */
622 	rpacket.ip_len = SCARG(bsa, len);
623 	error = copyout(&rpacket, packet, linux_ip_copysize);
624 	if (error)
625 		return (error);
626 
627 	riov[0].iov_base = (char *)packet;
628 	riov[0].iov_len = linux_ip_copysize;
629 	riov[1].iov_base = (caddr_t)SCARG(bsa, buf) + linux_ip_copysize;
630 	riov[1].iov_len = SCARG(bsa, len) - linux_ip_copysize;
631 
632 	error = copyout(&riov[0], iov, sizeof(riov));
633 	if (error)
634 		return (error);
635 
636 	/* Prepare the msghdr and iovec structures describing the new packet */
637 	rmsg.msg_name = (void *)SCARG(bsa, to);
638 	rmsg.msg_namelen = SCARG(bsa, tolen);
639 	rmsg.msg_iov = iov;
640 	rmsg.msg_iovlen = 2;
641 	rmsg.msg_control = NULL;
642 	rmsg.msg_controllen = 0;
643 	rmsg.msg_flags = 0;
644 
645 	error = copyout(&riov[0], iov, sizeof(riov));
646 	if (error)
647 		return (error);
648 
649 	SCARG(&ssa, s) = SCARG(bsa, s);
650 	SCARG(&ssa, msg) = msg;
651 	SCARG(&ssa, flags) = SCARG(bsa, flags);
652 	return sys_sendmsg(p, &ssa, retval);
653 }
654 
655 int
linux_sendto(p,v,retval)656 linux_sendto(p, v, retval)
657 	struct proc *p;
658 	void *v;
659 	register_t *retval;
660 {
661 	struct linux_sendto_args /* {
662 		syscallarg(int) s;
663 		syscallarg(void *) msg;
664 		syscallarg(int) len;
665 		syscallarg(int) flags;
666 		syscallarg(osockaddr *) to;
667 		syscallarg(int) tolen;
668 	} */ *uap = v;
669 	struct linux_sendto_args lsa;
670 	struct sys_sendto_args bsa;
671 	int error;
672 	int tolen;
673 	caddr_t sg = stackgap_init(p->p_emul);
674 
675 	if ((error = copyin((caddr_t) uap, (caddr_t) &lsa, sizeof lsa)))
676 		return error;
677 
678 	SCARG(&bsa, s) = lsa.s;
679 	SCARG(&bsa, buf) = lsa.msg;
680 	SCARG(&bsa, len) = lsa.len;
681 	SCARG(&bsa, flags) = lsa.flags;
682 	tolen = lsa.tolen;
683 	if (lsa.to) {
684 		struct sockaddr *sa;
685 
686 		if ((error = linux_sa_get(p, &sg, &sa, lsa.to, &tolen)))
687 			return (error);
688 		SCARG(&bsa, to) = sa;
689 	} else
690 		SCARG(&bsa, to) = NULL;
691 	SCARG(&bsa, tolen) = tolen;
692 
693 	if (linux_check_hdrincl(p, lsa.s, retval, &sg) == 0)
694 		return linux_sendto_hdrincl(p, &bsa, retval, &sg);
695 	return sys_sendto(p, &bsa, retval);
696 }
697 
698 int
linux_recvfrom(p,v,retval)699 linux_recvfrom(p, v, retval)
700 	struct proc *p;
701 	void *v;
702 	register_t *retval;
703 {
704 	struct linux_recvfrom_args /* {
705 		syscallarg(int) s;
706 		syscallarg(void *) buf;
707 		syscallarg(int) len;
708 		syscallarg(int) flags;
709 		syscallarg(struct osockaddr *) from;
710 		syscallarg(int *) fromlen;
711 	} */ *uap = v;
712 	struct linux_recvfrom_args lra;
713 	struct sys_recvfrom_args bra;
714 	int error;
715 
716 	if ((error = copyin((caddr_t) uap, (caddr_t) &lra, sizeof lra)))
717 		return error;
718 
719 	SCARG(&bra, s) = lra.s;
720 	SCARG(&bra, buf) = lra.buf;
721 	SCARG(&bra, len) = lra.len;
722 	SCARG(&bra, flags) = lra.flags;
723 	SCARG(&bra, from) = (struct sockaddr *) lra.from;
724 	SCARG(&bra, fromlenaddr) = lra.fromlen;
725 
726 	if ((error = sys_recvfrom(p, &bra, retval)))
727 		return (error);
728 
729 	if (lra.from && (error = linux_sa_put(lra.from)))
730 		return (error);
731 
732 	return (0);
733 }
734 
735 int
linux_shutdown(p,v,retval)736 linux_shutdown(p, v, retval)
737 	struct proc *p;
738 	void *v;
739 	register_t *retval;
740 {
741 	struct linux_shutdown_args /* {
742 		syscallarg(int) s;
743 		syscallarg(int) how;
744 	} */ *uap = v;
745 	struct linux_shutdown_args lsa;
746 	struct sys_shutdown_args bsa;
747 	int error;
748 
749 	if ((error = copyin((caddr_t) uap, (caddr_t) &lsa, sizeof lsa)))
750 		return error;
751 
752 	SCARG(&bsa, s) = lsa.s;
753 	SCARG(&bsa, how) = lsa.how;
754 
755 	return sys_shutdown(p, &bsa, retval);
756 }
757 
758 /*
759  * Convert socket option level from Linux to OpenBSD value. Only SOL_SOCKET
760  * is different, the rest matches IPPROTO_* on both systems.
761  */
762 int
linux_to_bsd_sopt_level(llevel)763 linux_to_bsd_sopt_level(llevel)
764 	int llevel;
765 {
766 
767 	switch (llevel) {
768 	case LINUX_SOL_SOCKET:
769 		return SOL_SOCKET;
770 	case LINUX_SOL_IP:
771 		return IPPROTO_IP;
772 	case LINUX_SOL_TCP:
773 		return IPPROTO_TCP;
774 	case LINUX_SOL_UDP:
775 		return IPPROTO_UDP;
776 	default:
777 		return -1;
778 	}
779 }
780 
781 /*
782  * Convert Linux socket level socket option numbers to OpenBSD values.
783  */
784 int
linux_to_bsd_so_sockopt(lopt)785 linux_to_bsd_so_sockopt(lopt)
786 	int lopt;
787 {
788 
789 	switch (lopt) {
790 	case LINUX_SO_DEBUG:
791 		return SO_DEBUG;
792 	case LINUX_SO_REUSEADDR:
793 		/*
794 		 * Linux does not implement SO_REUSEPORT, but allows reuse
795 		 * of a host:port pair through SO_REUSEADDR even if the
796 		 * address is not a multicast-address.  Effectively, this
797 		 * means that we should use SO_REUSEPORT to allow Linux
798 		 * applications to not exit with EADDRINUSE.
799 		 */
800 		return SO_REUSEPORT;
801 	case LINUX_SO_TYPE:
802 		return SO_TYPE;
803 	case LINUX_SO_ERROR:
804 		return SO_ERROR;
805 	case LINUX_SO_DONTROUTE:
806 		return SO_DONTROUTE;
807 	case LINUX_SO_BROADCAST:
808 		return SO_BROADCAST;
809 	case LINUX_SO_SNDBUF:
810 		return SO_SNDBUF;
811 	case LINUX_SO_RCVBUF:
812 		return SO_RCVBUF;
813 	case LINUX_SO_KEEPALIVE:
814 		return SO_KEEPALIVE;
815 	case LINUX_SO_OOBINLINE:
816 		return SO_OOBINLINE;
817 	case LINUX_SO_LINGER:
818 		return SO_LINGER;
819 	case LINUX_SO_PRIORITY:
820 	case LINUX_SO_NO_CHECK:
821 	default:
822 		return -1;
823 	}
824 }
825 
826 /*
827  * Convert Linux IP level socket option number to OpenBSD values.
828  */
829 int
linux_to_bsd_ip_sockopt(lopt)830 linux_to_bsd_ip_sockopt(lopt)
831 	int lopt;
832 {
833 
834 	switch (lopt) {
835 	case LINUX_IP_TOS:
836 		return IP_TOS;
837 	case LINUX_IP_TTL:
838 		return IP_TTL;
839 	case LINUX_IP_MULTICAST_TTL:
840 		return IP_MULTICAST_TTL;
841 	case LINUX_IP_MULTICAST_LOOP:
842 		return IP_MULTICAST_LOOP;
843 	case LINUX_IP_MULTICAST_IF:
844 		return IP_MULTICAST_IF;
845 	case LINUX_IP_ADD_MEMBERSHIP:
846 		return IP_ADD_MEMBERSHIP;
847 	case LINUX_IP_DROP_MEMBERSHIP:
848 		return IP_DROP_MEMBERSHIP;
849 	case LINUX_IP_HDRINCL:
850 		return IP_HDRINCL;
851 	default:
852 		return -1;
853 	}
854 }
855 
856 /*
857  * Convert Linux TCP level socket option number to OpenBSD values.
858  */
859 int
linux_to_bsd_tcp_sockopt(lopt)860 linux_to_bsd_tcp_sockopt(lopt)
861 	int lopt;
862 {
863 
864 	switch (lopt) {
865 	case LINUX_TCP_NODELAY:
866 		return TCP_NODELAY;
867 	case LINUX_TCP_MAXSEG:
868 		return TCP_MAXSEG;
869 	default:
870 		return -1;
871 	}
872 }
873 
874 /*
875  * Convert Linux UDP level socket option number to OpenBSD values.
876  */
877 int
linux_to_bsd_udp_sockopt(lopt)878 linux_to_bsd_udp_sockopt(lopt)
879 	int lopt;
880 {
881 
882 	switch (lopt) {
883 	default:
884 		return -1;
885 	}
886 }
887 
888 /*
889  * Another reasonably straightforward function: setsockopt(2).
890  * The level and option numbers are converted; the values passed
891  * are not (yet) converted, the ones currently implemented don't
892  * need conversion, as they are the same on both systems.
893  */
894 int
linux_setsockopt(p,v,retval)895 linux_setsockopt(p, v, retval)
896 	struct proc *p;
897 	void *v;
898 	register_t *retval;
899 {
900 	struct linux_setsockopt_args /* {
901 		syscallarg(int) s;
902 		syscallarg(int) level;
903 		syscallarg(int) optname;
904 		syscallarg(void *) optval;
905 		syscallarg(int) optlen;
906 	} */ *uap = v;
907 	struct linux_setsockopt_args lsa;
908 	struct sys_setsockopt_args bsa;
909 	int error, name;
910 
911 	if ((error = copyin((caddr_t) uap, (caddr_t) &lsa, sizeof lsa)))
912 		return error;
913 
914 	SCARG(&bsa, s) = lsa.s;
915 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(lsa.level);
916 	SCARG(&bsa, val) = lsa.optval;
917 	SCARG(&bsa, valsize) = lsa.optlen;
918 
919 	switch (SCARG(&bsa, level)) {
920 	case SOL_SOCKET:
921 		name = linux_to_bsd_so_sockopt(lsa.optname);
922 		break;
923 	case IPPROTO_IP:
924 		name = linux_to_bsd_ip_sockopt(lsa.optname);
925 		break;
926 	case IPPROTO_TCP:
927 		name = linux_to_bsd_tcp_sockopt(lsa.optname);
928 		break;
929 	case IPPROTO_UDP:
930 		name = linux_to_bsd_udp_sockopt(lsa.optname);
931 		break;
932 	default:
933 		return EINVAL;
934 	}
935 
936 	if (name == -1)
937 		return EINVAL;
938 	SCARG(&bsa, name) = name;
939 
940 	return sys_setsockopt(p, &bsa, retval);
941 }
942 
943 /*
944  * getsockopt(2) is very much the same as setsockopt(2) (see above)
945  */
946 int
linux_getsockopt(p,v,retval)947 linux_getsockopt(p, v, retval)
948 	struct proc *p;
949 	void *v;
950 	register_t *retval;
951 {
952 	struct linux_getsockopt_args /* {
953 		syscallarg(int) s;
954 		syscallarg(int) level;
955 		syscallarg(int) optname;
956 		syscallarg(void *) optval;
957 		syscallarg(int) *optlen;
958 	} */ *uap = v;
959 	struct linux_getsockopt_args lga;
960 	struct sys_getsockopt_args bga;
961 	int error, name;
962 
963 	if ((error = copyin((caddr_t) uap, (caddr_t) &lga, sizeof lga)))
964 		return error;
965 
966 	SCARG(&bga, s) = lga.s;
967 	SCARG(&bga, level) = linux_to_bsd_sopt_level(lga.level);
968 	SCARG(&bga, val) = lga.optval;
969 	SCARG(&bga, avalsize) = lga.optlen;
970 
971 	switch (SCARG(&bga, level)) {
972 	case SOL_SOCKET:
973 		name = linux_to_bsd_so_sockopt(lga.optname);
974 		break;
975 	case IPPROTO_IP:
976 		name = linux_to_bsd_ip_sockopt(lga.optname);
977 		break;
978 	case IPPROTO_TCP:
979 		name = linux_to_bsd_tcp_sockopt(lga.optname);
980 		break;
981 	case IPPROTO_UDP:
982 		name = linux_to_bsd_udp_sockopt(lga.optname);
983 		break;
984 	default:
985 		return EINVAL;
986 	}
987 
988 	if (name == -1)
989 		return EINVAL;
990 	SCARG(&bga, name) = name;
991 
992 	return sys_getsockopt(p, &bga, retval);
993 }
994 
995 int
linux_recvmsg(p,v,retval)996 linux_recvmsg(p, v, retval)
997 	struct proc *p;
998 	void *v;
999 	register_t *retval;
1000 {
1001 	struct linux_recvmsg_args /* {
1002 		syscallarg(int) s;
1003 		syscallarg(caddr_t) msg;
1004 		syscallarg(int) flags;
1005 	} */ *uap = v;
1006 	struct linux_recvmsg_args lla;
1007 	struct sys_recvmsg_args bla;
1008 	struct msghdr msg;
1009 	int error;
1010 
1011 	if ((error = copyin((caddr_t) uap, (caddr_t) &lla, sizeof lla)))
1012 		return error;
1013 
1014 	SCARG(&bla, s) = lla.s;
1015 	SCARG(&bla, msg) = (struct msghdr *)lla.msg;
1016 	SCARG(&bla, flags) = lla.flags;
1017 
1018 	error = sys_recvmsg(p, &bla, retval);
1019 	if (error)
1020 		return (error);
1021 
1022 	error = copyin(lla.msg, (caddr_t)&msg, sizeof(msg));
1023 
1024 	if (!error && msg.msg_name && msg.msg_namelen > 2)
1025 		error = linux_sa_put(msg.msg_name);
1026 
1027 	return (error);
1028 }
1029 
1030 int
linux_sendmsg(p,v,retval)1031 linux_sendmsg(p, v, retval)
1032 	struct proc *p;
1033 	void *v;
1034 	register_t *retval;
1035 {
1036 	struct linux_sendmsg_args /* {
1037 		syscallarg(int) s;
1038 		syscallarg(struct msghdr *) msg;
1039 		syscallarg(int) flags;
1040 	} */ *uap = v;
1041 	struct linux_sendmsg_args lla;
1042 	struct sys_sendmsg_args bla;
1043 	struct msghdr msg, *nmsg = NULL;
1044 	int error;
1045 	caddr_t control;
1046 	int level;
1047 
1048 	if ((error = copyin((caddr_t) uap, (caddr_t) &lla, sizeof lla)))
1049 		return error;
1050 
1051 	if ((error = copyin(lla.msg, (caddr_t) &msg, sizeof(msg))))
1052 		return (error);
1053 
1054 	if (msg.msg_name) {
1055 		struct sockaddr *sa;
1056 		caddr_t sg = stackgap_init(p->p_emul);
1057 
1058 		nmsg = (struct msghdr *)stackgap_alloc(&sg,
1059 		    sizeof(struct msghdr));
1060 		if (!nmsg)
1061 			return (ENOMEM);
1062 
1063 		error = linux_sa_get(p, &sg, &sa,
1064 		    (struct osockaddr *)msg.msg_name, &msg.msg_namelen);
1065 		if (error)
1066 			return (error);
1067 
1068 		msg.msg_name = (struct sockaddr *)sa;
1069 		if ((error = copyout(&msg, nmsg, sizeof(struct msghdr))))
1070 			return (error);
1071 		lla.msg = nmsg;
1072 	}
1073 
1074 	SCARG(&bla, s) = lla.s;
1075 	SCARG(&bla, msg) = lla.msg;
1076 	SCARG(&bla, flags) = lla.flags;
1077 
1078 	error = copyin(lla.msg->msg_control, &control, sizeof(caddr_t));
1079 	if (error)
1080 		return error;
1081 	if (control == NULL)
1082 		goto done;
1083 	error = copyin(&((struct cmsghdr *)control)->cmsg_level,
1084 	    &level, sizeof(int));
1085 	if (error)
1086 		return error;
1087 	if (level == 1) {
1088 		/*
1089 		 * Linux thinks that SOL_SOCKET is 1; we know that it's really
1090 		 * 0xffff, of course.
1091 		 */
1092 		level = SOL_SOCKET;
1093 		/*
1094 		 * XXX should use stack gap!
1095 		 * We don't because the control header is variable length
1096 		 * up to 2048 bytes, and there's only 512 bytes of gap.
1097 		 */
1098 		error = copyout(&level, &((struct cmsghdr *)control)->
1099 		    cmsg_level, sizeof(int));
1100 		if (error)
1101 			return error;
1102 	}
1103 done:
1104 	error = sys_sendmsg(p, &bla, retval);
1105 	/* replace level, just in case somebody cares. */
1106 	if (level == SOL_SOCKET) {
1107 		level = 1;
1108 		/* don't worry about the error */
1109 		copyout(&level, &((struct cmsghdr *)control)->cmsg_level,
1110 		    sizeof(int));
1111 	}
1112 	return (error);
1113 }
1114 
1115 /*
1116  * Copy the osockaddr structure pointed to by osa to kernel, adjust
1117  * family and convert to sockaddr, allocate stackgap and put the
1118  * the converted structure there, address on stackgap returned in sap.
1119  */
1120 int
linux_sa_get(p,sgp,sap,osa,osalen)1121 linux_sa_get(p, sgp, sap, osa, osalen)
1122 	struct proc *p;
1123 	caddr_t *sgp;
1124 	struct sockaddr **sap;
1125 	const struct osockaddr *osa;
1126 	int *osalen;
1127 {
1128 	int error=0, bdom;
1129 	struct sockaddr *sa, *usa;
1130 	struct osockaddr *kosa;
1131 	int alloclen;
1132 #ifdef INET6
1133 	int oldv6size;
1134 	struct sockaddr_in6 *sin6;
1135 #endif
1136 
1137 	if (*osalen < 2 || *osalen > UCHAR_MAX || !osa) {
1138 		return (EINVAL);
1139 	}
1140 
1141 	alloclen = *osalen;
1142 #ifdef INET6
1143 	oldv6size = 0;
1144 	/*
1145 	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
1146 	 * if it's a v4-mapped address, so reserve the proper space
1147 	 * for it.
1148 	 */
1149 	if (alloclen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1150 		alloclen = sizeof (struct sockaddr_in6);
1151 		oldv6size = 1;
1152 	}
1153 #endif
1154 
1155 	kosa = (struct osockaddr *) malloc(alloclen, M_TEMP, M_WAITOK);
1156 
1157 	if ((error = copyin(osa, (caddr_t) kosa, *osalen))) {
1158 		goto out;
1159 	}
1160 
1161 	bdom = linux_to_bsd_domain(kosa->sa_family);
1162 	if (bdom == -1) {
1163 		error = EINVAL;
1164 		goto out;
1165 	}
1166 
1167 #ifdef INET6
1168 	/*
1169 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1170 	 * which lacks the scope id compared with RFC2553 one. If we detect
1171 	 * the situation, reject the address.
1172 	 *
1173 	 * Still accept addresses for which the scope id is not used.
1174 	 */
1175 	if (oldv6size && bdom == AF_INET6) {
1176 		sin6 = (struct sockaddr_in6 *)kosa;
1177 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
1178 		    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
1179 		     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
1180 		     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
1181 		     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
1182 		     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1183 			sin6->sin6_scope_id = 0;
1184 		} else {
1185 			error = EINVAL;
1186 			goto out;
1187 		}
1188 	} else
1189 #endif
1190 	if (bdom == AF_INET) {
1191 		alloclen = sizeof(struct sockaddr_in);
1192 	}
1193 
1194 	sa = (struct sockaddr *) kosa;
1195 	sa->sa_family = bdom;
1196 	sa->sa_len = alloclen;
1197 
1198 	usa = (struct sockaddr *) stackgap_alloc(sgp, alloclen);
1199 	if (!usa) {
1200 		error = ENOMEM;
1201 		goto out;
1202 	}
1203 
1204 	if ((error = copyout(sa, usa, alloclen))) {
1205 		goto out;
1206 	}
1207 
1208 	*sap = usa;
1209 
1210     out:
1211 	*osalen = alloclen;
1212 	free(kosa, M_TEMP);
1213 	return (error);
1214 }
1215 
1216 int
linux_sa_put(osa)1217 linux_sa_put(osa)
1218 	struct osockaddr *osa;
1219 {
1220 	struct sockaddr sa;
1221 	struct osockaddr *kosa;
1222 	int error, bdom, len;
1223 
1224 	/*
1225 	 * Only read/write the sockaddr family and length part, the rest is
1226 	 * not changed.
1227 	 */
1228 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1229 
1230 	error = copyin((caddr_t) osa, (caddr_t) &sa, len);
1231 	if (error)
1232 		return (error);
1233 
1234 	bdom = bsd_to_linux_domain(sa.sa_family);
1235 	if (bdom == -1)
1236 		return (EINVAL);
1237 
1238 	/* Note: we convert from sockaddr to osockaddr here, too */
1239 	kosa = (struct osockaddr *) &sa;
1240 	kosa->sa_family = bdom;
1241 	error = copyout(kosa, osa, len);
1242 	if (error)
1243 		return (error);
1244 
1245 	return (0);
1246 }
1247 
1248 /*
1249  * Entry point to all Linux socket calls. Just check which call to
1250  * make and take appropriate action.
1251  */
1252 int
linux_sys_socketcall(p,v,retval)1253 linux_sys_socketcall(p, v, retval)
1254 	struct proc *p;
1255 	void *v;
1256 	register_t *retval;
1257 {
1258 	struct linux_sys_socketcall_args /* {
1259 		syscallarg(int) what;
1260 		syscallarg(void *) args;
1261 	} */ *uap = v;
1262 
1263 	switch (SCARG(uap, what)) {
1264 	case LINUX_SYS_socket:
1265 		return linux_socket(p, SCARG(uap, args), retval);
1266 	case LINUX_SYS_bind:
1267 		return linux_bind(p, SCARG(uap, args), retval);
1268 	case LINUX_SYS_connect:
1269 		return linux_connect(p, SCARG(uap, args), retval);
1270 	case LINUX_SYS_listen:
1271 		return linux_listen(p, SCARG(uap, args), retval);
1272 	case LINUX_SYS_accept:
1273 		return linux_accept(p, SCARG(uap, args), retval);
1274 	case LINUX_SYS_getsockname:
1275 		return linux_getsockname(p, SCARG(uap, args), retval);
1276 	case LINUX_SYS_getpeername:
1277 		return linux_getpeername(p, SCARG(uap, args), retval);
1278 	case LINUX_SYS_socketpair:
1279 		return linux_socketpair(p, SCARG(uap, args), retval);
1280 	case LINUX_SYS_send:
1281 		return linux_send(p, SCARG(uap, args), retval);
1282 	case LINUX_SYS_recv:
1283 		return linux_recv(p, SCARG(uap, args), retval);
1284 	case LINUX_SYS_sendto:
1285 		return linux_sendto(p, SCARG(uap, args), retval);
1286 	case LINUX_SYS_recvfrom:
1287 		return linux_recvfrom(p, SCARG(uap, args), retval);
1288 	case LINUX_SYS_shutdown:
1289 		return linux_shutdown(p, SCARG(uap, args), retval);
1290 	case LINUX_SYS_setsockopt:
1291 		return linux_setsockopt(p, SCARG(uap, args), retval);
1292 	case LINUX_SYS_getsockopt:
1293 		return linux_getsockopt(p, SCARG(uap, args), retval);
1294 	case LINUX_SYS_sendmsg:
1295 		return linux_sendmsg(p, SCARG(uap, args), retval);
1296 	case LINUX_SYS_recvmsg:
1297 		return linux_recvmsg(p, SCARG(uap, args), retval);
1298 	default:
1299 		return ENOSYS;
1300 	}
1301 }
1302 
1303 int
linux_ioctl_socket(p,v,retval)1304 linux_ioctl_socket(p, v, retval)
1305 	register struct proc *p;
1306 	void *v;
1307 	register_t *retval;
1308 {
1309 	struct linux_sys_ioctl_args /* {
1310 		syscallarg(int) fd;
1311 		syscallarg(u_long) com;
1312 		syscallarg(caddr_t) data;
1313 	} */ *uap = v;
1314 	u_long com;
1315 	struct sys_ioctl_args ia;
1316 	struct file *fp;
1317 	struct filedesc *fdp;
1318 	struct vnode *vp;
1319 	int (*ioctlf)(struct file *, u_long, caddr_t, struct proc *);
1320 	struct ioctl_pt pt;
1321 	int error = 0, isdev = 0, dosys = 1;
1322 
1323 	fdp = p->p_fd;
1324 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1325 		return (EBADF);
1326 	FREF(fp);
1327 
1328 	if (fp->f_type == DTYPE_VNODE) {
1329 		vp = (struct vnode *)fp->f_data;
1330 		isdev = vp->v_type == VCHR;
1331 	}
1332 
1333 	/*
1334 	 * Don't try to interpret socket ioctl calls that are done
1335 	 * on a device filedescriptor, just pass them through, to
1336 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1337 	 * device will only handle these if it's prepared to do
1338 	 * so, to avoid unexpected things from happening.
1339 	 */
1340 	if (isdev) {
1341 		dosys = 0;
1342 		ioctlf = fp->f_ops->fo_ioctl;
1343 		pt.com = SCARG(uap, com);
1344 		pt.data = SCARG(uap, data);
1345 		error = ioctlf(fp, PTIOCLINUX, (caddr_t)&pt, p);
1346 		/*
1347 		 * XXX hack: if the function returns EJUSTRETURN,
1348 		 * it has stuffed a sysctl return value in pt.data.
1349 		 */
1350 		if (error == EJUSTRETURN) {
1351 			retval[0] = (register_t)pt.data;
1352 			error = 0;
1353 		}
1354 		goto out;
1355 	}
1356 
1357 	com = SCARG(uap, com);
1358 	retval[0] = 0;
1359 
1360 	switch (com) {
1361 	case LINUX_FIOSETOWN:
1362 		SCARG(&ia, com) = FIOSETOWN;
1363 		break;
1364 	case LINUX_SIOCSPGRP:
1365 		SCARG(&ia, com) = SIOCSPGRP;
1366 		break;
1367 	case LINUX_FIOGETOWN:
1368 		SCARG(&ia, com) = FIOGETOWN;
1369 		break;
1370 	case LINUX_SIOCGPGRP:
1371 		SCARG(&ia, com) = SIOCGPGRP;
1372 		break;
1373 	case LINUX_SIOCATMARK:
1374 		SCARG(&ia, com) = SIOCATMARK;
1375 		break;
1376 #if 0
1377 	case LINUX_SIOCGSTAMP:
1378 		SCARG(&ia, com) = SIOCGSTAMP;
1379 		break;
1380 #endif
1381 	case LINUX_SIOCGIFCONF:
1382 		SCARG(&ia, com) = OSIOCGIFCONF;
1383 		break;
1384 	case LINUX_SIOCGIFFLAGS:
1385 		SCARG(&ia, com) = SIOCGIFFLAGS;
1386 		break;
1387 	case LINUX_SIOCGIFADDR:
1388 		SCARG(&ia, com) = OSIOCGIFADDR;
1389 		break;
1390 	case LINUX_SIOCGIFDSTADDR:
1391 		SCARG(&ia, com) = OSIOCGIFDSTADDR;
1392 		break;
1393 	case LINUX_SIOCGIFBRDADDR:
1394 		SCARG(&ia, com) = OSIOCGIFBRDADDR;
1395 		break;
1396 	case LINUX_SIOCGIFNETMASK:
1397 		SCARG(&ia, com) = OSIOCGIFNETMASK;
1398 		break;
1399 	case LINUX_SIOCGIFMETRIC:
1400 		SCARG(&ia, com) = SIOCGIFMETRIC;
1401 		break;
1402 	case LINUX_SIOCGIFMTU:
1403 		SCARG(&ia, com) = SIOCGIFMTU;
1404 		break;
1405 	case LINUX_SIOCADDMULTI:
1406 		SCARG(&ia, com) = SIOCADDMULTI;
1407 		break;
1408 	case LINUX_SIOCDELMULTI:
1409 		SCARG(&ia, com) = SIOCDELMULTI;
1410 		break;
1411 	case LINUX_SIOCGIFHWADDR: {
1412 		struct linux_ifreq *ifr = (struct linux_ifreq *)SCARG(uap, data);
1413 		struct sockaddr_dl *sdl;
1414 		struct ifnet *ifp;
1415 		struct ifaddr *ifa;
1416 
1417 		/*
1418 		 * Note that we don't actually respect the name in the ifreq
1419 		 * structure, as Linux interface names are all different.
1420 		 */
1421 		for (ifp = ifnet.tqh_first; ifp != 0;
1422 		    ifp = ifp->if_list.tqe_next) {
1423 			if (ifp->if_type != IFT_ETHER)
1424 				continue;
1425 			for (ifa = ifp->if_addrlist.tqh_first; ifa;
1426 			    ifa = ifa->ifa_list.tqe_next) {
1427 				if ((sdl = (struct sockaddr_dl *)ifa->ifa_addr) &&
1428 				    (sdl->sdl_family == AF_LINK) &&
1429 				    (sdl->sdl_type == IFT_ETHER)) {
1430 					error = copyout(LLADDR(sdl),
1431 					    (caddr_t)&ifr->ifr_hwaddr.sa_data,
1432 					    LINUX_IFHWADDRLEN);
1433 					dosys = 0;
1434 					goto out;
1435 				}
1436 			}
1437 		}
1438 		error = ENOENT;
1439 		break;
1440 	    }
1441 	default:
1442 		error = EINVAL;
1443 	}
1444 
1445 out:
1446 	if (error == 0 && dosys) {
1447 		SCARG(&ia, fd) = SCARG(uap, fd);
1448 		SCARG(&ia, data) = SCARG(uap, data);
1449 		error = sys_ioctl(p, &ia, retval);
1450 	}
1451 
1452 	FRELE(fp);
1453 	return (error);
1454 }
1455