xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision 5bbd2a12505d72a8177929a37b5cee489d0a1cfd)
1 /*	$NetBSD: linux_socket.c,v 1.114 2012/06/20 15:03:18 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 1998, 2008 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Frank van der Linden and Eric Haszlakiewicz.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Functions in multiarch:
34  *	linux_sys_socketcall		: linux_socketcall.c
35  */
36 
37 #include <sys/cdefs.h>
38 __KERNEL_RCSID(0, "$NetBSD: linux_socket.c,v 1.114 2012/06/20 15:03:18 christos Exp $");
39 
40 #if defined(_KERNEL_OPT)
41 #include "opt_inet.h"
42 #endif /* defined(_KERNEL_OPT) */
43 
44 #include <sys/param.h>
45 #include <sys/kernel.h>
46 #include <sys/systm.h>
47 #include <sys/buf.h>
48 #include <sys/malloc.h>
49 #include <sys/ioctl.h>
50 #include <sys/tty.h>
51 #include <sys/file.h>
52 #include <sys/filedesc.h>
53 #include <sys/select.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/domain.h>
57 #include <net/if.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <netinet/in.h>
61 #include <netinet/tcp.h>
62 #include <sys/mount.h>
63 #include <sys/proc.h>
64 #include <sys/vnode.h>
65 #include <sys/device.h>
66 #include <sys/protosw.h>
67 #include <sys/mbuf.h>
68 #include <sys/syslog.h>
69 #include <sys/exec.h>
70 #include <sys/kauth.h>
71 #include <sys/syscallargs.h>
72 #include <sys/ktrace.h>
73 #include <sys/fcntl.h>
74 
75 #include <lib/libkern/libkern.h>
76 
77 #include <netinet/ip6.h>
78 #include <netinet6/ip6_var.h>
79 
80 #include <compat/sys/socket.h>
81 #include <compat/sys/sockio.h>
82 
83 #include <compat/linux/common/linux_types.h>
84 #include <compat/linux/common/linux_util.h>
85 #include <compat/linux/common/linux_signal.h>
86 #include <compat/linux/common/linux_ioctl.h>
87 #include <compat/linux/common/linux_socket.h>
88 #include <compat/linux/common/linux_fcntl.h>
89 #if !defined(__alpha__) && !defined(__amd64__)
90 #include <compat/linux/common/linux_socketcall.h>
91 #endif
92 #include <compat/linux/common/linux_sockio.h>
93 #include <compat/linux/common/linux_ipc.h>
94 #include <compat/linux/common/linux_sem.h>
95 
96 #include <compat/linux/linux_syscallargs.h>
97 
98 #ifdef DEBUG_LINUX
99 #define DPRINTF(a) uprintf a
100 #else
101 #define DPRINTF(a)
102 #endif
103 
104 /*
105  * The calls in this file are entered either via the linux_socketcall()
106  * interface or, on the Alpha, as individual syscalls.  The
107  * linux_socketcall function does any massaging of arguments so that all
108  * the calls in here need not think that they are anything other
109  * than a normal syscall.
110  */
111 
112 static int linux_to_bsd_domain(int);
113 static int bsd_to_linux_domain(int);
114 int linux_to_bsd_sopt_level(int);
115 int linux_to_bsd_so_sockopt(int);
116 int linux_to_bsd_ip_sockopt(int);
117 int linux_to_bsd_tcp_sockopt(int);
118 int linux_to_bsd_udp_sockopt(int);
119 int linux_getifname(struct lwp *, register_t *, void *);
120 int linux_getifconf(struct lwp *, register_t *, void *);
121 int linux_getifhwaddr(struct lwp *, register_t *, u_int, void *);
122 static int linux_get_sa(struct lwp *, int, struct mbuf **,
123 		const struct osockaddr *, unsigned int);
124 static int linux_sa_put(struct osockaddr *osa);
125 static int linux_to_bsd_msg_flags(int);
126 static int bsd_to_linux_msg_flags(int);
127 static void linux_to_bsd_msghdr(struct linux_msghdr *, struct msghdr *);
128 static void bsd_to_linux_msghdr(struct msghdr *, struct linux_msghdr *);
129 
130 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
131 	AF_UNSPEC,
132 	AF_UNIX,
133 	AF_INET,
134 	AF_CCITT,	/* LINUX_AF_AX25 */
135 	AF_IPX,
136 	AF_APPLETALK,
137 	-1,		/* LINUX_AF_NETROM */
138 	-1,		/* LINUX_AF_BRIDGE */
139 	-1,		/* LINUX_AF_ATMPVC */
140 	AF_CCITT,	/* LINUX_AF_X25 */
141 	AF_INET6,
142 	-1,		/* LINUX_AF_ROSE */
143 	AF_DECnet,
144 	-1,		/* LINUX_AF_NETBEUI */
145 	-1,		/* LINUX_AF_SECURITY */
146 	pseudo_AF_KEY,
147 	AF_ROUTE,	/* LINUX_AF_NETLINK */
148 	-1,		/* LINUX_AF_PACKET */
149 	-1,		/* LINUX_AF_ASH */
150 	-1,		/* LINUX_AF_ECONET */
151 	-1,		/* LINUX_AF_ATMSVC */
152 	AF_SNA,
153 	/* rest up to LINUX_AF_MAX-1 is not allocated */
154 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
155 };
156 
157 static const int bsd_to_linux_domain_[AF_MAX] = {
158 	LINUX_AF_UNSPEC,
159 	LINUX_AF_UNIX,
160 	LINUX_AF_INET,
161 	-1,		/* AF_IMPLINK */
162 	-1,		/* AF_PUP */
163 	-1,		/* AF_CHAOS */
164 	-1,		/* AF_NS */
165 	-1,		/* AF_ISO */
166 	-1,		/* AF_ECMA */
167 	-1,		/* AF_DATAKIT */
168 	LINUX_AF_AX25,	/* AF_CCITT */
169 	LINUX_AF_SNA,
170 	LINUX_AF_DECnet,
171 	-1,		/* AF_DLI */
172 	-1,		/* AF_LAT */
173 	-1,		/* AF_HYLINK */
174 	LINUX_AF_APPLETALK,
175 	LINUX_AF_NETLINK,
176 	-1,		/* AF_LINK */
177 	-1,		/* AF_XTP */
178 	-1,		/* AF_COIP */
179 	-1,		/* AF_CNT */
180 	-1,		/* pseudo_AF_RTIP */
181 	LINUX_AF_IPX,
182 	LINUX_AF_INET6,
183 	-1,		/* pseudo_AF_PIP */
184 	-1,		/* AF_ISDN */
185 	-1,		/* AF_NATM */
186 	-1,		/* AF_ARP */
187 	LINUX_pseudo_AF_KEY,
188 	-1,		/* pseudo_AF_HDRCMPLT */
189 };
190 
191 static const struct {
192 	int bfl;
193 	int lfl;
194 } bsd_to_linux_msg_flags_[] = {
195 	{MSG_OOB,		LINUX_MSG_OOB},
196 	{MSG_PEEK,		LINUX_MSG_PEEK},
197 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
198 	{MSG_EOR,		LINUX_MSG_EOR},
199 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
200 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
201 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
202 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
203 	{MSG_BCAST,		0},		/* not supported, clear */
204 	{MSG_MCAST,		0},		/* not supported, clear */
205 	{MSG_NOSIGNAL,		LINUX_MSG_NOSIGNAL},
206 	{-1, /* not supp */	LINUX_MSG_PROBE},
207 	{-1, /* not supp */	LINUX_MSG_FIN},
208 	{-1, /* not supp */	LINUX_MSG_SYN},
209 	{-1, /* not supp */	LINUX_MSG_CONFIRM},
210 	{-1, /* not supp */	LINUX_MSG_RST},
211 	{-1, /* not supp */	LINUX_MSG_ERRQUEUE},
212 	{-1, /* not supp */	LINUX_MSG_MORE},
213 };
214 
215 /*
216  * Convert between Linux and BSD socket domain values
217  */
218 static int
219 linux_to_bsd_domain(int ldom)
220 {
221 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
222 		return (-1);
223 
224 	return linux_to_bsd_domain_[ldom];
225 }
226 
227 /*
228  * Convert between BSD and Linux socket domain values
229  */
230 static int
231 bsd_to_linux_domain(int bdom)
232 {
233 	if (bdom < 0 || bdom >= AF_MAX)
234 		return (-1);
235 
236 	return bsd_to_linux_domain_[bdom];
237 }
238 
239 static int
240 linux_to_bsd_msg_flags(int lflag)
241 {
242 	int i, lfl, bfl;
243 	int bflag = 0;
244 
245 	if (lflag == 0)
246 		return (0);
247 
248 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
249 		bfl = bsd_to_linux_msg_flags_[i].bfl;
250 		lfl = bsd_to_linux_msg_flags_[i].lfl;
251 
252 		if (lfl == 0)
253 			continue;
254 
255 		if (lflag & lfl) {
256 			if (bfl < 0)
257 				return (-1);
258 
259 			bflag |= bfl;
260 		}
261 	}
262 
263 	return (bflag);
264 }
265 
266 static int
267 bsd_to_linux_msg_flags(int bflag)
268 {
269 	int i, lfl, bfl;
270 	int lflag = 0;
271 
272 	if (bflag == 0)
273 		return (0);
274 
275 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
276 		bfl = bsd_to_linux_msg_flags_[i].bfl;
277 		lfl = bsd_to_linux_msg_flags_[i].lfl;
278 
279 		if (bfl <= 0)
280 			continue;
281 
282 		if (bflag & bfl) {
283 			if (lfl < 0)
284 				return (-1);
285 
286 			lflag |= lfl;
287 		}
288 	}
289 
290 	return (lflag);
291 }
292 
293 int
294 linux_sys_socket(struct lwp *l, const struct linux_sys_socket_args *uap, register_t *retval)
295 {
296 	/* {
297 		syscallarg(int)	domain;
298 		syscallarg(int)	type;
299 		syscallarg(int) protocol;
300 	} */
301 	struct sys___socket30_args bsa;
302 	struct sys_fcntl_args fsa;
303 	register_t fretval[2];
304 	int error, flags;
305 
306 
307 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
308 	SCARG(&bsa, type) = SCARG(uap, type) & LINUX_SOCK_TYPE_MASK;
309 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
310 	if (SCARG(&bsa, domain) == -1)
311 		return EINVAL;
312 	/*
313 	 * Apparently linux uses this to talk to ISDN sockets. If we fail
314 	 * now programs seems to handle it, but if we don't we are going
315 	 * to fail when we bind and programs don't handle this well.
316 	 */
317 	if (SCARG(&bsa, domain) == AF_ROUTE && SCARG(&bsa, type) == SOCK_RAW)
318 		return ENOTSUP;
319 	flags = SCARG(uap, type) & ~LINUX_SOCK_TYPE_MASK;
320 	if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
321 		return EINVAL;
322 	error = sys___socket30(l, &bsa, retval);
323 
324 	/*
325 	 * Linux overloads the "type" parameter to include some
326 	 * fcntl flags to be set on the file descriptor.
327 	 * Process those if creating the socket succeeded.
328 	 */
329 
330 	if (!error && flags & LINUX_SOCK_CLOEXEC) {
331 		SCARG(&fsa, fd) = *retval;
332 		SCARG(&fsa, cmd) = F_SETFD;
333 		SCARG(&fsa, arg) = (void *)(uintptr_t)FD_CLOEXEC;
334 		(void) sys_fcntl(l, &fsa, fretval);
335 	}
336 	if (!error && flags & LINUX_SOCK_NONBLOCK) {
337 		SCARG(&fsa, fd) = *retval;
338 		SCARG(&fsa, cmd) = F_SETFL;
339 		SCARG(&fsa, arg) = (void *)(uintptr_t)O_NONBLOCK;
340 		error = sys_fcntl(l, &fsa, fretval);
341 		if (error) {
342 			struct sys_close_args csa;
343 
344 			SCARG(&csa, fd) = *retval;
345 			(void) sys_close(l, &csa, fretval);
346 		}
347 	}
348 
349 #ifdef INET6
350 	/*
351 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
352 	 * default and some apps depend on this. So, set V6ONLY to 0
353 	 * for Linux apps if the sysctl value is set to 1.
354 	 */
355 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
356 		struct socket *so;
357 
358 		if (fd_getsock(*retval, &so) == 0) {
359 			int val = 0;
360 
361 			/* ignore error */
362 			(void)so_setsockopt(l, so, IPPROTO_IPV6, IPV6_V6ONLY,
363 			    &val, sizeof(val));
364 
365 			fd_putfile(*retval);
366 		}
367 	}
368 #endif
369 
370 	return (error);
371 }
372 
373 int
374 linux_sys_socketpair(struct lwp *l, const struct linux_sys_socketpair_args *uap, register_t *retval)
375 {
376 	/* {
377 		syscallarg(int) domain;
378 		syscallarg(int) type;
379 		syscallarg(int) protocol;
380 		syscallarg(int *) rsv;
381 	} */
382 	struct sys_socketpair_args bsa;
383 
384 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
385 	if (SCARG(&bsa, domain) == -1)
386 		return EINVAL;
387 	SCARG(&bsa, type) = SCARG(uap, type);
388 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
389 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
390 
391 	return sys_socketpair(l, &bsa, retval);
392 }
393 
394 int
395 linux_sys_sendto(struct lwp *l, const struct linux_sys_sendto_args *uap, register_t *retval)
396 {
397 	/* {
398 		syscallarg(int)				s;
399 		syscallarg(void *)			msg;
400 		syscallarg(int)				len;
401 		syscallarg(int)				flags;
402 		syscallarg(struct osockaddr *)		to;
403 		syscallarg(int)				tolen;
404 	} */
405 	struct msghdr   msg;
406 	struct iovec    aiov;
407 	struct mbuf *nam;
408 	int bflags;
409 	int error;
410 
411 	/* Translate message flags.  */
412 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
413 	if (bflags < 0)
414 		/* Some supported flag */
415 		return EINVAL;
416 
417 	msg.msg_flags = 0;
418 	msg.msg_name = NULL;
419 	msg.msg_control = NULL;
420 
421 	if (SCARG(uap, tolen)) {
422 		/* Read in and convert the sockaddr */
423 		error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
424 		    SCARG(uap, tolen));
425 		if (error)
426 			return (error);
427 		msg.msg_flags |= MSG_NAMEMBUF;
428 		msg.msg_name = nam;
429 		msg.msg_namelen = SCARG(uap, tolen);
430 	}
431 
432 	msg.msg_iov = &aiov;
433 	msg.msg_iovlen = 1;
434 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
435 	aiov.iov_len = SCARG(uap, len);
436 
437 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
438 }
439 
440 static void
441 linux_to_bsd_msghdr(struct linux_msghdr *lmsg, struct msghdr *bmsg)
442 {
443 	bmsg->msg_name = lmsg->msg_name;
444 	bmsg->msg_namelen = lmsg->msg_namelen;
445 	bmsg->msg_iov = lmsg->msg_iov;
446 	bmsg->msg_iovlen = lmsg->msg_iovlen;
447 	bmsg->msg_control = lmsg->msg_control;
448 	bmsg->msg_controllen = lmsg->msg_controllen;
449 	bmsg->msg_flags = lmsg->msg_flags;
450 }
451 
452 static void
453 bsd_to_linux_msghdr(struct msghdr *bmsg, struct linux_msghdr *lmsg)
454 {
455 	lmsg->msg_name = bmsg->msg_name;
456 	lmsg->msg_namelen = bmsg->msg_namelen;
457 	lmsg->msg_iov = bmsg->msg_iov;
458 	lmsg->msg_iovlen = bmsg->msg_iovlen;
459 	lmsg->msg_control = bmsg->msg_control;
460 	lmsg->msg_controllen = bmsg->msg_controllen;
461 	lmsg->msg_flags = bmsg->msg_flags;
462 }
463 
464 int
465 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
466 {
467 	/* {
468 		syscallarg(int) s;
469 		syscallarg(struct linux_msghdr *) msg;
470 		syscallarg(u_int) flags;
471 	} */
472 	struct msghdr	msg;
473 	struct linux_msghdr lmsg;
474 	int		error;
475 	int		bflags;
476 	struct mbuf     *nam;
477 	u_int8_t	*control;
478 	struct mbuf     *ctl_mbuf = NULL;
479 
480 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
481 	if (error)
482 		return error;
483 	linux_to_bsd_msghdr(&lmsg, &msg);
484 
485 	msg.msg_flags = MSG_IOVUSRSPACE;
486 
487 	/*
488 	 * Translate message flags.
489 	 */
490 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
491 	if (bflags < 0)
492 		/* Some supported flag */
493 		return EINVAL;
494 
495 	if (lmsg.msg_name) {
496 		/* Read in and convert the sockaddr */
497 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
498 		    msg.msg_namelen);
499 		if (error)
500 			return (error);
501 		msg.msg_flags |= MSG_NAMEMBUF;
502 		msg.msg_name = nam;
503 	}
504 
505 	/*
506 	 * Handle cmsg if there is any.
507 	 */
508 	if (LINUX_CMSG_FIRSTHDR(&lmsg)) {
509 		struct linux_cmsghdr l_cmsg, *l_cc;
510 		struct cmsghdr *cmsg;
511 		ssize_t resid = msg.msg_controllen;
512 		size_t clen, cidx = 0, cspace;
513 
514 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
515 		clen = MLEN;
516 		control = mtod(ctl_mbuf, void *);
517 
518 		l_cc = LINUX_CMSG_FIRSTHDR(&lmsg);
519 		do {
520 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
521 			if (error)
522 				goto done;
523 
524 			/*
525 			 * Sanity check the control message length.
526 			 */
527 			if (l_cmsg.cmsg_len > resid
528 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
529 				error = EINVAL;
530 				goto done;
531 			}
532 
533 			/*
534 			 * Refuse unsupported control messages, and
535 			 * translate fields as appropriate.
536 			 */
537 			switch (l_cmsg.cmsg_level) {
538 			case LINUX_SOL_SOCKET:
539 				/* It only differs on some archs */
540 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
541 					l_cmsg.cmsg_level = SOL_SOCKET;
542 
543 				switch(l_cmsg.cmsg_type) {
544 				case LINUX_SCM_RIGHTS:
545 					/* Linux SCM_RIGHTS is same as NetBSD */
546 					break;
547 
548 				case LINUX_SCM_CREDENTIALS:
549 					/* no native equivalent, just drop it */
550 					m_free(ctl_mbuf);
551 					ctl_mbuf = NULL;
552 					msg.msg_control = NULL;
553 					msg.msg_controllen = 0;
554 					goto skipcmsg;
555 
556 				default:
557 					/* other types not supported */
558 					error = EINVAL;
559 					goto done;
560 				}
561 				break;
562 			default:
563 				/* pray and leave intact */
564 				break;
565 			}
566 
567 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
568 
569 			/* Check the buffer is big enough */
570 			if (__predict_false(cidx + cspace > clen)) {
571 				u_int8_t *nc;
572 
573 				clen = cidx + cspace;
574 				if (clen >= PAGE_SIZE) {
575 					error = EINVAL;
576 					goto done;
577 				}
578 				nc = realloc(clen <= MLEN ? NULL : control,
579 						clen, M_TEMP, M_WAITOK);
580 				if (!nc) {
581 					error = ENOMEM;
582 					goto done;
583 				}
584 				if (cidx <= MLEN)
585 					/* Old buffer was in mbuf... */
586 					memcpy(nc, control, cidx);
587 				control = nc;
588 			}
589 
590 			/* Copy header */
591 			cmsg = (void *)&control[cidx];
592 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
593 			cmsg->cmsg_level = l_cmsg.cmsg_level;
594 			cmsg->cmsg_type = l_cmsg.cmsg_type;
595 
596 			/* Zero area between header and data */
597 			memset(cmsg + 1, 0,
598 				CMSG_ALIGN(sizeof(*cmsg)) - sizeof(*cmsg));
599 
600 			/* Copyin the data */
601 			error = copyin(LINUX_CMSG_DATA(l_cc),
602 				CMSG_DATA(cmsg),
603 				l_cmsg.cmsg_len - sizeof(l_cmsg));
604 			if (error)
605 				goto done;
606 
607 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
608 			cidx += cspace;
609 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
610 
611 		/* If we allocated a buffer, attach to mbuf */
612 		if (cidx > MLEN) {
613 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
614 			ctl_mbuf->m_flags |= M_EXT_RW;
615 		}
616 		control = NULL;
617 		ctl_mbuf->m_len = cidx;
618 
619 		msg.msg_control = ctl_mbuf;
620 		msg.msg_flags |= MSG_CONTROLMBUF;
621 
622 		ktrkuser("mbcontrol", mtod(ctl_mbuf, void *),
623 		    msg.msg_controllen);
624 	}
625 
626 skipcmsg:
627 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
628 	/* Freed internally */
629 	ctl_mbuf = NULL;
630 
631 done:
632 	if (ctl_mbuf != NULL) {
633 		if (control != NULL && control != mtod(ctl_mbuf, void *))
634 			free(control, M_MBUF);
635 		m_free(ctl_mbuf);
636 	}
637 	return (error);
638 }
639 
640 int
641 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
642 {
643 	/* {
644 		syscallarg(int) s;
645 		syscallarg(void *) buf;
646 		syscallarg(int) len;
647 		syscallarg(int) flags;
648 		syscallarg(struct osockaddr *) from;
649 		syscallarg(int *) fromlenaddr;
650 	} */
651 	int		error;
652 	struct sys_recvfrom_args bra;
653 
654 	SCARG(&bra, s) = SCARG(uap, s);
655 	SCARG(&bra, buf) = SCARG(uap, buf);
656 	SCARG(&bra, len) = SCARG(uap, len);
657 	SCARG(&bra, flags) = SCARG(uap, flags);
658 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
659 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
660 
661 	if ((error = sys_recvfrom(l, &bra, retval)))
662 		return (error);
663 
664 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
665 		return (error);
666 
667 	return (0);
668 }
669 
670 static int
671 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
672 {
673 	int dlen, error = 0;
674 	struct cmsghdr *cmsg;
675 	struct linux_cmsghdr linux_cmsg;
676 	struct mbuf *m;
677 	char *q, *q_end;
678 
679 	if (mp->msg_controllen <= 0 || control == 0) {
680 		mp->msg_controllen = 0;
681 		free_control_mbuf(l, control, control);
682 		return 0;
683 	}
684 
685 	ktrkuser("msgcontrol", mtod(control, void *), mp->msg_controllen);
686 
687 	q = (char *)mp->msg_control;
688 	q_end = q + mp->msg_controllen;
689 
690 	for (m = control; m != NULL; ) {
691 		cmsg = mtod(m, struct cmsghdr *);
692 
693 		/*
694 		 * Fixup cmsg. We handle two things:
695 		 * 0. different sizeof cmsg_len.
696 		 * 1. different values for level/type on some archs
697 		 * 2. different alignment of CMSG_DATA on some archs
698 		 */
699 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
700 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
701 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
702 
703 		dlen = q_end - q;
704 		if (linux_cmsg.cmsg_len > dlen) {
705 			/* Not enough room for the parameter */
706 			dlen -= sizeof linux_cmsg;
707 			if (dlen <= 0)
708 				/* Discard if header wont fit */
709 				break;
710 			mp->msg_flags |= MSG_CTRUNC;
711 			if (linux_cmsg.cmsg_level == SOL_SOCKET
712 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
713 				/* Do not truncate me ... */
714 				break;
715 		} else
716 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
717 
718 		switch (linux_cmsg.cmsg_level) {
719 		case SOL_SOCKET:
720 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
721 			switch (linux_cmsg.cmsg_type) {
722 			case SCM_RIGHTS:
723 				/* Linux SCM_RIGHTS is same as NetBSD */
724 				break;
725 
726 			default:
727 				/* other types not supported */
728 				error = EINVAL;
729 				goto done;
730 			}
731 			/* machine dependent ! */
732 			break;
733 		default:
734 			/* pray and leave intact */
735 			break;
736 		}
737 
738 		/* There can be padding between the header and data... */
739 		error = copyout(&linux_cmsg, q, sizeof linux_cmsg);
740 		if (error != 0) {
741 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
742 			    dlen);
743 		}
744 		if (error != 0) {
745 			/* We must free all the SCM_RIGHTS */
746 			m = control;
747 			break;
748 		}
749 		m = m->m_next;
750 		if (m == NULL || q + LINUX_CMSG_SPACE(dlen) > q_end) {
751 			q += LINUX_CMSG_LEN(dlen);
752 			break;
753 		}
754 		q += LINUX_CMSG_SPACE(dlen);
755 	}
756 
757   done:
758 	free_control_mbuf(l, control, m);
759 
760 	mp->msg_controllen = q - (char *)mp->msg_control;
761 	return error;
762 }
763 
764 int
765 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
766 {
767 	/* {
768 		syscallarg(int) s;
769 		syscallarg(struct linux_msghdr *) msg;
770 		syscallarg(u_int) flags;
771 	} */
772 	struct msghdr	msg;
773 	struct linux_msghdr lmsg;
774 	int		error;
775 	struct mbuf	*from, *control;
776 
777 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
778 	if (error)
779 		return (error);
780 	linux_to_bsd_msghdr(&lmsg, &msg);
781 
782 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
783 	if (msg.msg_flags < 0) {
784 		/* Some unsupported flag */
785 		return (EINVAL);
786 	}
787 	msg.msg_flags |= MSG_IOVUSRSPACE;
788 
789 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
790 	    msg.msg_control != NULL ? &control : NULL, retval);
791 	if (error != 0)
792 		return error;
793 
794 	if (msg.msg_control != NULL)
795 		error = linux_copyout_msg_control(l, &msg, control);
796 
797 	if (error == 0 && from != 0) {
798 		mtod(from, struct osockaddr *)->sa_family =
799 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
800 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
801 			from);
802 	} else
803 		msg.msg_namelen = 0;
804 
805 	if (from != NULL)
806 		m_free(from);
807 
808 	if (error == 0) {
809 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
810 		if (msg.msg_flags < 0)
811 			/* Some flag unsupported by Linux */
812 			error = EINVAL;
813 		else {
814 			ktrkuser("msghdr", &msg, sizeof(msg));
815 			bsd_to_linux_msghdr(&msg, &lmsg);
816 			error = copyout(&lmsg, SCARG(uap, msg), sizeof(lmsg));
817 		}
818 	}
819 
820 	return (error);
821 }
822 
823 /*
824  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
825  * is different, the rest matches IPPROTO_* on both systems.
826  */
827 int
828 linux_to_bsd_sopt_level(int llevel)
829 {
830 
831 	switch (llevel) {
832 	case LINUX_SOL_SOCKET:
833 		return SOL_SOCKET;
834 	case LINUX_SOL_IP:
835 		return IPPROTO_IP;
836 	case LINUX_SOL_TCP:
837 		return IPPROTO_TCP;
838 	case LINUX_SOL_UDP:
839 		return IPPROTO_UDP;
840 	default:
841 		return -1;
842 	}
843 }
844 
845 /*
846  * Convert Linux socket level socket option numbers to NetBSD values.
847  */
848 int
849 linux_to_bsd_so_sockopt(int lopt)
850 {
851 
852 	switch (lopt) {
853 	case LINUX_SO_DEBUG:
854 		return SO_DEBUG;
855 	case LINUX_SO_REUSEADDR:
856 		/*
857 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
858 		 * host:port pair through SO_REUSEADDR even if the address is not a
859 		 * multicast-address.  Effectively, this means that we should use
860 		 * SO_REUSEPORT to allow Linux applications to not exit with
861 		 * EADDRINUSE
862 		 */
863 		return SO_REUSEPORT;
864 	case LINUX_SO_TYPE:
865 		return SO_TYPE;
866 	case LINUX_SO_ERROR:
867 		return SO_ERROR;
868 	case LINUX_SO_DONTROUTE:
869 		return SO_DONTROUTE;
870 	case LINUX_SO_BROADCAST:
871 		return SO_BROADCAST;
872 	case LINUX_SO_SNDBUF:
873 		return SO_SNDBUF;
874 	case LINUX_SO_RCVBUF:
875 		return SO_RCVBUF;
876 	case LINUX_SO_KEEPALIVE:
877 		return SO_KEEPALIVE;
878 	case LINUX_SO_OOBINLINE:
879 		return SO_OOBINLINE;
880 	case LINUX_SO_LINGER:
881 		return SO_LINGER;
882 	case LINUX_SO_PRIORITY:
883 	case LINUX_SO_NO_CHECK:
884 	default:
885 		return -1;
886 	}
887 }
888 
889 /*
890  * Convert Linux IP level socket option number to NetBSD values.
891  */
892 int
893 linux_to_bsd_ip_sockopt(int lopt)
894 {
895 
896 	switch (lopt) {
897 	case LINUX_IP_TOS:
898 		return IP_TOS;
899 	case LINUX_IP_TTL:
900 		return IP_TTL;
901 	case LINUX_IP_HDRINCL:
902 		return IP_HDRINCL;
903 	case LINUX_IP_MULTICAST_TTL:
904 		return IP_MULTICAST_TTL;
905 	case LINUX_IP_MULTICAST_LOOP:
906 		return IP_MULTICAST_LOOP;
907 	case LINUX_IP_MULTICAST_IF:
908 		return IP_MULTICAST_IF;
909 	case LINUX_IP_ADD_MEMBERSHIP:
910 		return IP_ADD_MEMBERSHIP;
911 	case LINUX_IP_DROP_MEMBERSHIP:
912 		return IP_DROP_MEMBERSHIP;
913 	default:
914 		return -1;
915 	}
916 }
917 
918 /*
919  * Convert Linux TCP level socket option number to NetBSD values.
920  */
921 int
922 linux_to_bsd_tcp_sockopt(int lopt)
923 {
924 
925 	switch (lopt) {
926 	case LINUX_TCP_NODELAY:
927 		return TCP_NODELAY;
928 	case LINUX_TCP_MAXSEG:
929 		return TCP_MAXSEG;
930 	default:
931 		return -1;
932 	}
933 }
934 
935 /*
936  * Convert Linux UDP level socket option number to NetBSD values.
937  */
938 int
939 linux_to_bsd_udp_sockopt(int lopt)
940 {
941 
942 	switch (lopt) {
943 	default:
944 		return -1;
945 	}
946 }
947 
948 /*
949  * Another reasonably straightforward function: setsockopt(2).
950  * The level and option numbers are converted; the values passed
951  * are not (yet) converted, the ones currently implemented don't
952  * need conversion, as they are the same on both systems.
953  */
954 int
955 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
956 {
957 	/* {
958 		syscallarg(int) s;
959 		syscallarg(int) level;
960 		syscallarg(int) optname;
961 		syscallarg(void *) optval;
962 		syscallarg(int) optlen;
963 	} */
964 	struct sys_setsockopt_args bsa;
965 	int name;
966 
967 	SCARG(&bsa, s) = SCARG(uap, s);
968 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
969 	SCARG(&bsa, val) = SCARG(uap, optval);
970 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
971 
972 	/*
973 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
974 	 * and returns EOPNOTSUPP for other levels
975 	 */
976 	if (SCARG(&bsa, level) != SOL_SOCKET) {
977 		struct socket *so;
978 		int error, family;
979 
980 		/* fd_getsock() will use the descriptor for us */
981 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
982 		    	return error;
983 		family = so->so_proto->pr_domain->dom_family;
984 		fd_putfile(SCARG(&bsa, s));
985 
986 		if (family == AF_LOCAL)
987 			return EOPNOTSUPP;
988 	}
989 
990 	switch (SCARG(&bsa, level)) {
991 	case SOL_SOCKET:
992 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
993 		break;
994 	case IPPROTO_IP:
995 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
996 		break;
997 	case IPPROTO_TCP:
998 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
999 		break;
1000 	case IPPROTO_UDP:
1001 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1002 		break;
1003 	default:
1004 		return EINVAL;
1005 	}
1006 
1007 	if (name == -1)
1008 		return EINVAL;
1009 	SCARG(&bsa, name) = name;
1010 
1011 	return sys_setsockopt(l, &bsa, retval);
1012 }
1013 
1014 /*
1015  * getsockopt(2) is very much the same as setsockopt(2) (see above)
1016  */
1017 int
1018 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
1019 {
1020 	/* {
1021 		syscallarg(int) s;
1022 		syscallarg(int) level;
1023 		syscallarg(int) optname;
1024 		syscallarg(void *) optval;
1025 		syscallarg(int *) optlen;
1026 	} */
1027 	struct sys_getsockopt_args bga;
1028 	int name;
1029 
1030 	SCARG(&bga, s) = SCARG(uap, s);
1031 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
1032 	SCARG(&bga, val) = SCARG(uap, optval);
1033 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
1034 
1035 	switch (SCARG(&bga, level)) {
1036 	case SOL_SOCKET:
1037 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1038 		break;
1039 	case IPPROTO_IP:
1040 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1041 		break;
1042 	case IPPROTO_TCP:
1043 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1044 		break;
1045 	case IPPROTO_UDP:
1046 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1047 		break;
1048 	default:
1049 		return EINVAL;
1050 	}
1051 
1052 	if (name == -1)
1053 		return EINVAL;
1054 	SCARG(&bga, name) = name;
1055 
1056 	return sys_getsockopt(l, &bga, retval);
1057 }
1058 
1059 int
1060 linux_getifname(struct lwp *l, register_t *retval, void *data)
1061 {
1062 	struct ifnet *ifp;
1063 	struct linux_ifreq ifr;
1064 	int error;
1065 
1066 	error = copyin(data, &ifr, sizeof(ifr));
1067 	if (error)
1068 		return error;
1069 
1070 	if (ifr.ifr_ifru.ifru_ifindex >= if_indexlim)
1071 		return ENODEV;
1072 
1073 	ifp = ifindex2ifnet[ifr.ifr_ifru.ifru_ifindex];
1074 	if (ifp == NULL)
1075 		return ENODEV;
1076 
1077 	strncpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name));
1078 
1079 	return copyout(&ifr, data, sizeof(ifr));
1080 }
1081 
1082 int
1083 linux_getifconf(struct lwp *l, register_t *retval, void *data)
1084 {
1085 	struct linux_ifreq ifr, *ifrp;
1086 	struct ifconf *ifc = data;
1087 	struct ifnet *ifp;
1088 	struct ifaddr *ifa;
1089 	struct sockaddr *sa;
1090 	struct osockaddr *osa;
1091 	int space, error = 0;
1092 	const int sz = (int)sizeof(ifr);
1093 
1094 	ifrp = (struct linux_ifreq *)ifc->ifc_req;
1095 	if (ifrp == NULL)
1096 		space = 0;
1097 	else
1098 		space = ifc->ifc_len;
1099 
1100 	IFNET_FOREACH(ifp) {
1101 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1102 		    sizeof(ifr.ifr_name));
1103 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
1104 			return ENAMETOOLONG;
1105 		if (IFADDR_EMPTY(ifp))
1106 			continue;
1107 		IFADDR_FOREACH(ifa, ifp) {
1108 			sa = ifa->ifa_addr;
1109 			if (sa->sa_family != AF_INET ||
1110 			    sa->sa_len > sizeof(*osa))
1111 				continue;
1112 			memcpy(&ifr.ifr_addr, sa, sa->sa_len);
1113 			osa = (struct osockaddr *)&ifr.ifr_addr;
1114 			osa->sa_family = sa->sa_family;
1115 			if (space >= sz) {
1116 				error = copyout(&ifr, ifrp, sz);
1117 				if (error != 0)
1118 					return error;
1119 				ifrp++;
1120 			}
1121 			space -= sz;
1122 		}
1123 	}
1124 
1125 	if (ifrp != NULL)
1126 		ifc->ifc_len -= space;
1127 	else
1128 		ifc->ifc_len = -space;
1129 
1130 	return 0;
1131 }
1132 
1133 int
1134 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1135     void *data)
1136 {
1137 	/* Not the full structure, just enough to map what we do here */
1138 	struct linux_ifreq lreq;
1139 	file_t *fp;
1140 	struct ifaddr *ifa;
1141 	struct ifnet *ifp;
1142 	struct sockaddr_dl *sadl;
1143 	int error, found;
1144 	int index, ifnum;
1145 
1146 	/*
1147 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1148 	 * SIOCGIFCONF, because the user buffer is not of the right
1149 	 * type to take those results.  We can't use kernel buffers to
1150 	 * receive the results, as the implementation of sys_ioctl()
1151 	 * and ifconf() [which implements SIOCGIFCONF] use
1152 	 * copyin()/copyout() which will fail on kernel addresses.
1153 	 *
1154 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1155 	 */
1156 
1157 	if ((fp = fd_getfile(fd)) == NULL)
1158 		return (EBADF);
1159 
1160 	KERNEL_LOCK(1, NULL);
1161 
1162 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1163 		error = EBADF;
1164 		goto out;
1165 	}
1166 
1167 	error = copyin(data, &lreq, sizeof(lreq));
1168 	if (error)
1169 		goto out;
1170 	lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0';		/* just in case */
1171 
1172 	/*
1173 	 * Try real interface name first, then fake "ethX"
1174 	 */
1175 	found = 0;
1176 	IFNET_FOREACH(ifp) {
1177 		if (found)
1178 			break;
1179 		if (strcmp(lreq.ifr_name, ifp->if_xname))
1180 			/* not this interface */
1181 			continue;
1182 		found=1;
1183 		if (IFADDR_EMPTY(ifp)) {
1184 			error = ENODEV;
1185 			goto out;
1186 		}
1187 		IFADDR_FOREACH(ifa, ifp) {
1188 			sadl = satosdl(ifa->ifa_addr);
1189 			/* only return ethernet addresses */
1190 			/* XXX what about FDDI, etc. ? */
1191 			if (sadl->sdl_family != AF_LINK ||
1192 			    sadl->sdl_type != IFT_ETHER)
1193 				continue;
1194 			memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl),
1195 			       MIN(sadl->sdl_alen,
1196 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1197 			lreq.ifr_hwaddr.sa_family =
1198 				sadl->sdl_family;
1199 			error = copyout(&lreq, data, sizeof(lreq));
1200 			goto out;
1201 		}
1202 	}
1203 
1204 	if (strncmp(lreq.ifr_name, "eth", 3) == 0) {
1205 		for (ifnum = 0, index = 3;
1206 		     index < LINUX_IFNAMSIZ && lreq.ifr_name[index] != '\0';
1207 		     index++) {
1208 			ifnum *= 10;
1209 			ifnum += lreq.ifr_name[index] - '0';
1210 		}
1211 
1212 		error = EINVAL;			/* in case we don't find one */
1213 		found = 0;
1214 		IFNET_FOREACH(ifp) {
1215 			if (found)
1216 				break;
1217 			memcpy(lreq.ifr_name, ifp->if_xname,
1218 			       MIN(LINUX_IFNAMSIZ, IFNAMSIZ));
1219 			IFADDR_FOREACH(ifa, ifp) {
1220 				sadl = satosdl(ifa->ifa_addr);
1221 				/* only return ethernet addresses */
1222 				/* XXX what about FDDI, etc. ? */
1223 				if (sadl->sdl_family != AF_LINK ||
1224 				    sadl->sdl_type != IFT_ETHER)
1225 					continue;
1226 				if (ifnum--)
1227 					/* not the reqested iface */
1228 					continue;
1229 				memcpy(&lreq.ifr_hwaddr.sa_data,
1230 				       CLLADDR(sadl),
1231 				       MIN(sadl->sdl_alen,
1232 					   sizeof(lreq.ifr_hwaddr.sa_data)));
1233 				lreq.ifr_hwaddr.sa_family =
1234 					sadl->sdl_family;
1235 				error = copyout(&lreq, data, sizeof(lreq));
1236 				found = 1;
1237 				break;
1238 			}
1239 		}
1240 	} else {
1241 		/* unknown interface, not even an "eth*" name */
1242 		error = ENODEV;
1243 	}
1244 
1245 out:
1246 	KERNEL_UNLOCK_ONE(NULL);
1247 	fd_putfile(fd);
1248 	return error;
1249 }
1250 
1251 int
1252 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1253 {
1254 	/* {
1255 		syscallarg(int) fd;
1256 		syscallarg(u_long) com;
1257 		syscallarg(void *) data;
1258 	} */
1259 	u_long com;
1260 	int error = 0, isdev = 0, dosys = 1;
1261 	struct sys_ioctl_args ia;
1262 	file_t *fp;
1263 	struct vnode *vp;
1264 	int (*ioctlf)(file_t *, u_long, void *);
1265 	struct ioctl_pt pt;
1266 
1267 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1268 		return (EBADF);
1269 
1270 	if (fp->f_type == DTYPE_VNODE) {
1271 		vp = (struct vnode *)fp->f_data;
1272 		isdev = vp->v_type == VCHR;
1273 	}
1274 
1275 	/*
1276 	 * Don't try to interpret socket ioctl calls that are done
1277 	 * on a device filedescriptor, just pass them through, to
1278 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1279 	 * device will only handle these if it's prepared to do
1280 	 * so, to avoid unexpected things from happening.
1281 	 */
1282 	if (isdev) {
1283 		dosys = 0;
1284 		ioctlf = fp->f_ops->fo_ioctl;
1285 		pt.com = SCARG(uap, com);
1286 		pt.data = SCARG(uap, data);
1287 		error = ioctlf(fp, PTIOCLINUX, &pt);
1288 		/*
1289 		 * XXX hack: if the function returns EJUSTRETURN,
1290 		 * it has stuffed a sysctl return value in pt.data.
1291 		 */
1292 		if (error == EJUSTRETURN) {
1293 			retval[0] = (register_t)pt.data;
1294 			error = 0;
1295 		}
1296 		goto out;
1297 	}
1298 
1299 	com = SCARG(uap, com);
1300 	retval[0] = 0;
1301 
1302 	switch (com) {
1303 	case LINUX_SIOCGIFNAME:
1304 		error = linux_getifname(l, retval, SCARG(uap, data));
1305 		dosys = 0;
1306 		break;
1307 	case LINUX_SIOCGIFCONF:
1308 		error = linux_getifconf(l, retval, SCARG(uap, data));
1309 		dosys = 0;
1310 		break;
1311 	case LINUX_SIOCGIFFLAGS:
1312 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1313 		break;
1314 	case LINUX_SIOCSIFFLAGS:
1315 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1316 		break;
1317 	case LINUX_SIOCGIFADDR:
1318 		SCARG(&ia, com) = OOSIOCGIFADDR;
1319 		break;
1320 	case LINUX_SIOCGIFDSTADDR:
1321 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1322 		break;
1323 	case LINUX_SIOCGIFBRDADDR:
1324 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1325 		break;
1326 	case LINUX_SIOCGIFNETMASK:
1327 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1328 		break;
1329 	case LINUX_SIOCGIFMTU:
1330 		SCARG(&ia, com) = OSIOCGIFMTU;
1331 		break;
1332 	case LINUX_SIOCADDMULTI:
1333 		SCARG(&ia, com) = OSIOCADDMULTI;
1334 		break;
1335 	case LINUX_SIOCDELMULTI:
1336 		SCARG(&ia, com) = OSIOCDELMULTI;
1337 		break;
1338 	case LINUX_SIOCGIFHWADDR:
1339 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1340 		    SCARG(uap, data));
1341 		dosys = 0;
1342 		break;
1343 	default:
1344 		error = EINVAL;
1345 	}
1346 
1347  out:
1348  	fd_putfile(SCARG(uap, fd));
1349 
1350 	if (error ==0 && dosys) {
1351 		SCARG(&ia, fd) = SCARG(uap, fd);
1352 		SCARG(&ia, data) = SCARG(uap, data);
1353 		error = sys_ioctl(curlwp, &ia, retval);
1354 	}
1355 
1356 	return error;
1357 }
1358 
1359 int
1360 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1361 {
1362 	/* {
1363 		syscallarg(int) s;
1364 		syscallarg(const struct sockaddr *) name;
1365 		syscallarg(int) namelen;
1366 	} */
1367 	int		error;
1368 	struct mbuf *nam;
1369 
1370 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1371 	    SCARG(uap, namelen));
1372 	if (error)
1373 		return (error);
1374 
1375 	error = do_sys_connect(l, SCARG(uap, s), nam);
1376 
1377 	if (error == EISCONN) {
1378 		struct socket *so;
1379 		int state, prflags;
1380 
1381 		/* fd_getsock() will use the descriptor for us */
1382 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1383 		    	return EISCONN;
1384 
1385 		solock(so);
1386 		state = so->so_state;
1387 		prflags = so->so_proto->pr_flags;
1388 		sounlock(so);
1389 		fd_putfile(SCARG(uap, s));
1390 		/*
1391 		 * We should only let this call succeed once per
1392 		 * non-blocking connect; however we don't have
1393 		 * a convenient place to keep that state..
1394 		 */
1395 		if ((state & (SS_ISCONNECTED|SS_NBIO)) ==
1396 		    (SS_ISCONNECTED|SS_NBIO) &&
1397 		    (prflags & PR_CONNREQUIRED))
1398 			return 0;
1399 	}
1400 
1401 	return (error);
1402 }
1403 
1404 int
1405 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1406 {
1407 	/* {
1408 		syscallarg(int) s;
1409 		syscallarg(const struct osockaddr *) name;
1410 		syscallarg(int) namelen;
1411 	} */
1412 	int		error;
1413 	struct mbuf     *nam;
1414 
1415 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1416 	    SCARG(uap, namelen));
1417 	if (error)
1418 		return (error);
1419 
1420 	return do_sys_bind(l, SCARG(uap, s), nam);
1421 }
1422 
1423 int
1424 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1425 {
1426 	/* {
1427 		syscallarg(int) fdes;
1428 		syscallarg(void *) asa;
1429 		syscallarg(int *) alen;
1430 	} */
1431 	int error;
1432 
1433 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1434 		return (error);
1435 
1436 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1437 		return (error);
1438 
1439 	return (0);
1440 }
1441 
1442 int
1443 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1444 {
1445 	/* {
1446 		syscallarg(int) fdes;
1447 		syscallarg(void *) asa;
1448 		syscallarg(int *) alen;
1449 	} */
1450 	int error;
1451 
1452 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1453 		return (error);
1454 
1455 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1456 		return (error);
1457 
1458 	return (0);
1459 }
1460 
1461 /*
1462  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1463  * family and convert to sockaddr.
1464  */
1465 static int
1466 linux_get_sa(struct lwp *l, int s, struct mbuf **mp,
1467     const struct osockaddr *osa, unsigned int salen)
1468 {
1469 	int error, bdom;
1470 	struct sockaddr *sa;
1471 	struct osockaddr *kosa;
1472 	struct mbuf *m;
1473 
1474 	if (salen == 1 || salen > UCHAR_MAX) {
1475 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1476 		return EINVAL;
1477 	}
1478 
1479 	/* We'll need the address in an mbuf later, so copy into one here */
1480 	m = m_get(M_WAIT, MT_SONAME);
1481 	if (salen > MLEN)
1482 		MEXTMALLOC(m, salen, M_WAITOK);
1483 
1484 	m->m_len = salen;
1485 
1486 	if (salen == 0) {
1487 		*mp = m;
1488 		return 0;
1489 	}
1490 
1491 	kosa = mtod(m, void *);
1492 	if ((error = copyin(osa, kosa, salen))) {
1493 		DPRINTF(("error %d copying osa %p len %d\n",
1494 				error, osa, salen));
1495 		goto bad;
1496 	}
1497 
1498 	ktrkuser("linux/sockaddr", kosa, salen);
1499 
1500 	bdom = linux_to_bsd_domain(kosa->sa_family);
1501 	if (bdom == -1) {
1502 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1503 		error = EINVAL;
1504 		goto bad;
1505 	}
1506 
1507 	/*
1508 	 * If the family is unspecified, use address family of the socket.
1509 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1510 	 */
1511 	if (bdom == AF_UNSPEC) {
1512 		struct socket *so;
1513 
1514 		/* fd_getsock() will use the descriptor for us */
1515 		if ((error = fd_getsock(s, &so)) != 0)
1516 			goto bad;
1517 
1518 		bdom = so->so_proto->pr_domain->dom_family;
1519 		fd_putfile(s);
1520 
1521 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1522 	}
1523 
1524 	/*
1525 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1526 	 * which lacks the scope id compared with RFC2553 one. If we detect
1527 	 * the situation, reject the address and write a message to system log.
1528 	 *
1529 	 * Still accept addresses for which the scope id is not used.
1530 	 */
1531 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1532 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1533 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1534 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1535 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1536 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1537 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1538 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1539 			struct proc *p = l->l_proc;
1540 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1541 
1542 			log(LOG_DEBUG,
1543 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1544 			    "sockaddr_in6 rejected",
1545 			    p->p_pid, p->p_comm, uid);
1546 			error = EINVAL;
1547 			goto bad;
1548 		}
1549 		salen = sizeof (struct sockaddr_in6);
1550 		sin6->sin6_scope_id = 0;
1551 	}
1552 
1553 	if (bdom == AF_INET)
1554 		salen = sizeof(struct sockaddr_in);
1555 
1556 	sa = (struct sockaddr *) kosa;
1557 	sa->sa_family = bdom;
1558 	sa->sa_len = salen;
1559 	m->m_len = salen;
1560 	ktrkuser("mbsoname", kosa, salen);
1561 
1562 #ifdef DEBUG_LINUX
1563 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1564 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1565 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1566 	DPRINTF(("\n"));
1567 #endif
1568 
1569 	*mp = m;
1570 	return 0;
1571 
1572     bad:
1573 	m_free(m);
1574 	return error;
1575 }
1576 
1577 static int
1578 linux_sa_put(struct osockaddr *osa)
1579 {
1580 	struct sockaddr sa;
1581 	struct osockaddr *kosa;
1582 	int error, bdom, len;
1583 
1584 	/*
1585 	 * Only read/write the sockaddr family and length part, the rest is
1586 	 * not changed.
1587 	 */
1588 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1589 
1590 	error = copyin(osa, &sa, len);
1591 	if (error)
1592 		return (error);
1593 
1594 	bdom = bsd_to_linux_domain(sa.sa_family);
1595 	if (bdom == -1)
1596 		return (EINVAL);
1597 
1598 	/* Note: we convert from sockaddr to osockaddr here, too */
1599 	kosa = (struct osockaddr *) &sa;
1600 	kosa->sa_family = bdom;
1601 	error = copyout(kosa, osa, len);
1602 	if (error)
1603 		return (error);
1604 
1605 	return (0);
1606 }
1607 
1608 #ifndef __amd64__
1609 int
1610 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1611 {
1612 	/* {
1613 		syscallarg(int) s;
1614 		syscallarg(void *) buf;
1615 		syscallarg(int) len;
1616 		syscallarg(int) flags;
1617 	} */
1618 	struct sys_recvfrom_args bra;
1619 
1620 
1621 	SCARG(&bra, s) = SCARG(uap, s);
1622 	SCARG(&bra, buf) = SCARG(uap, buf);
1623 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1624 	SCARG(&bra, flags) = SCARG(uap, flags);
1625 	SCARG(&bra, from) = NULL;
1626 	SCARG(&bra, fromlenaddr) = NULL;
1627 
1628 	return (sys_recvfrom(l, &bra, retval));
1629 }
1630 
1631 int
1632 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1633 {
1634 	/* {
1635 		syscallarg(int) s;
1636 		syscallarg(void *) buf;
1637 		syscallarg(int) len;
1638 		syscallarg(int) flags;
1639 	} */
1640 	struct sys_sendto_args bsa;
1641 
1642 	SCARG(&bsa, s)		= SCARG(uap, s);
1643 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1644 	SCARG(&bsa, len)	= SCARG(uap, len);
1645 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1646 	SCARG(&bsa, to)		= NULL;
1647 	SCARG(&bsa, tolen)	= 0;
1648 
1649 	return (sys_sendto(l, &bsa, retval));
1650 }
1651 #endif
1652 
1653 int
1654 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1655 {
1656 	/* {
1657 		syscallarg(int) s;
1658 		syscallarg(struct osockaddr *) name;
1659 		syscallarg(int *) anamelen;
1660 	} */
1661 	int error;
1662 	struct sys_accept_args baa;
1663 
1664 	SCARG(&baa, s)		= SCARG(uap, s);
1665 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1666 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1667 
1668 	if ((error = sys_accept(l, &baa, retval)))
1669 		return (error);
1670 
1671 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1672 		return (error);
1673 
1674 	return (0);
1675 }
1676