xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision aad9773e38ed2370a628a6416e098f9008fc10a7)
1 /*	$NetBSD: linux_socket.c,v 1.122 2014/11/26 09:53:53 ozaki-r 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 1.122 2014/11/26 09:53:53 ozaki-r 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/ioctl.h>
49 #include <sys/tty.h>
50 #include <sys/file.h>
51 #include <sys/filedesc.h>
52 #include <sys/select.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/domain.h>
56 #include <net/if.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <netinet/in.h>
60 #include <netinet/tcp.h>
61 #include <sys/mount.h>
62 #include <sys/proc.h>
63 #include <sys/vnode.h>
64 #include <sys/device.h>
65 #include <sys/protosw.h>
66 #include <sys/mbuf.h>
67 #include <sys/syslog.h>
68 #include <sys/exec.h>
69 #include <sys/kauth.h>
70 #include <sys/syscallargs.h>
71 #include <sys/ktrace.h>
72 
73 #include <lib/libkern/libkern.h>
74 
75 #include <netinet/ip6.h>
76 #include <netinet6/ip6_var.h>
77 
78 #include <compat/sys/socket.h>
79 #include <compat/sys/sockio.h>
80 
81 #include <compat/linux/common/linux_types.h>
82 #include <compat/linux/common/linux_util.h>
83 #include <compat/linux/common/linux_signal.h>
84 #include <compat/linux/common/linux_ioctl.h>
85 #include <compat/linux/common/linux_socket.h>
86 #include <compat/linux/common/linux_fcntl.h>
87 #if !defined(__alpha__) && !defined(__amd64__)
88 #include <compat/linux/common/linux_socketcall.h>
89 #endif
90 #include <compat/linux/common/linux_sockio.h>
91 #include <compat/linux/common/linux_ipc.h>
92 #include <compat/linux/common/linux_sem.h>
93 
94 #include <compat/linux/linux_syscallargs.h>
95 
96 #ifdef DEBUG_LINUX
97 #define DPRINTF(a) uprintf a
98 #else
99 #define DPRINTF(a)
100 #endif
101 
102 /*
103  * The calls in this file are entered either via the linux_socketcall()
104  * interface or, on the Alpha, as individual syscalls.  The
105  * linux_socketcall function does any massaging of arguments so that all
106  * the calls in here need not think that they are anything other
107  * than a normal syscall.
108  */
109 
110 static int linux_to_bsd_domain(int);
111 static int bsd_to_linux_domain(int);
112 static int linux_to_bsd_type(int);
113 int linux_to_bsd_sopt_level(int);
114 int linux_to_bsd_so_sockopt(int);
115 int linux_to_bsd_ip_sockopt(int);
116 int linux_to_bsd_ipv6_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_type(int ltype)
241 {
242 	int type, flags;
243 
244 	/* Real types are identical between Linux and NetBSD */
245 	type = ltype & LINUX_SOCK_TYPE_MASK;
246 
247 	/* But flags are not .. */
248 	flags = ltype & ~LINUX_SOCK_TYPE_MASK;
249 	if (flags & ~(LINUX_SOCK_CLOEXEC|LINUX_SOCK_NONBLOCK))
250 		return -1;
251 
252 	if (flags & LINUX_SOCK_CLOEXEC)
253 		type |= SOCK_CLOEXEC;
254 	if (flags & LINUX_SOCK_NONBLOCK)
255 		type |= SOCK_NONBLOCK;
256 
257 	return type;
258 }
259 
260 static int
261 linux_to_bsd_msg_flags(int lflag)
262 {
263 	int i, lfl, bfl;
264 	int bflag = 0;
265 
266 	if (lflag == 0)
267 		return (0);
268 
269 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
270 		bfl = bsd_to_linux_msg_flags_[i].bfl;
271 		lfl = bsd_to_linux_msg_flags_[i].lfl;
272 
273 		if (lfl == 0)
274 			continue;
275 
276 		if (lflag & lfl) {
277 			if (bfl < 0)
278 				return (-1);
279 
280 			bflag |= bfl;
281 		}
282 	}
283 
284 	return (bflag);
285 }
286 
287 static int
288 bsd_to_linux_msg_flags(int bflag)
289 {
290 	int i, lfl, bfl;
291 	int lflag = 0;
292 
293 	if (bflag == 0)
294 		return (0);
295 
296 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
297 		bfl = bsd_to_linux_msg_flags_[i].bfl;
298 		lfl = bsd_to_linux_msg_flags_[i].lfl;
299 
300 		if (bfl <= 0)
301 			continue;
302 
303 		if (bflag & bfl) {
304 			if (lfl < 0)
305 				return (-1);
306 
307 			lflag |= lfl;
308 		}
309 	}
310 
311 	return (lflag);
312 }
313 
314 int
315 linux_sys_socket(struct lwp *l, const struct linux_sys_socket_args *uap, register_t *retval)
316 {
317 	/* {
318 		syscallarg(int)	domain;
319 		syscallarg(int)	type;
320 		syscallarg(int) protocol;
321 	} */
322 	struct sys___socket30_args bsa;
323 	int error;
324 
325 
326 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
327 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
328 	if (SCARG(&bsa, domain) == -1)
329 		return EINVAL;
330 	SCARG(&bsa, type) = linux_to_bsd_type(SCARG(uap, type));
331 	if (SCARG(&bsa, type) == -1)
332 		return EINVAL;
333 	/*
334 	 * Apparently linux uses this to talk to ISDN sockets. If we fail
335 	 * now programs seems to handle it, but if we don't we are going
336 	 * to fail when we bind and programs don't handle this well.
337 	 */
338 	if (SCARG(&bsa, domain) == AF_ROUTE && SCARG(&bsa, type) == SOCK_RAW)
339 		return ENOTSUP;
340 	error = sys___socket30(l, &bsa, retval);
341 
342 #ifdef INET6
343 	/*
344 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
345 	 * default and some apps depend on this. So, set V6ONLY to 0
346 	 * for Linux apps if the sysctl value is set to 1.
347 	 */
348 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
349 		struct socket *so;
350 
351 		if (fd_getsock(*retval, &so) == 0) {
352 			int val = 0;
353 
354 			/* ignore error */
355 			(void)so_setsockopt(l, so, IPPROTO_IPV6, IPV6_V6ONLY,
356 			    &val, sizeof(val));
357 
358 			fd_putfile(*retval);
359 		}
360 	}
361 #endif
362 
363 	return (error);
364 }
365 
366 int
367 linux_sys_socketpair(struct lwp *l, const struct linux_sys_socketpair_args *uap, register_t *retval)
368 {
369 	/* {
370 		syscallarg(int) domain;
371 		syscallarg(int) type;
372 		syscallarg(int) protocol;
373 		syscallarg(int *) rsv;
374 	} */
375 	struct sys_socketpair_args bsa;
376 
377 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
378 	if (SCARG(&bsa, domain) == -1)
379 		return EINVAL;
380 	SCARG(&bsa, type) = linux_to_bsd_type(SCARG(uap, type));
381 	if (SCARG(&bsa, type) == -1)
382 		return EINVAL;
383 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
384 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
385 
386 	return sys_socketpair(l, &bsa, retval);
387 }
388 
389 int
390 linux_sys_sendto(struct lwp *l, const struct linux_sys_sendto_args *uap, register_t *retval)
391 {
392 	/* {
393 		syscallarg(int)				s;
394 		syscallarg(void *)			msg;
395 		syscallarg(int)				len;
396 		syscallarg(int)				flags;
397 		syscallarg(struct osockaddr *)		to;
398 		syscallarg(int)				tolen;
399 	} */
400 	struct msghdr   msg;
401 	struct iovec    aiov;
402 	struct mbuf *nam;
403 	int bflags;
404 	int error;
405 
406 	/* Translate message flags.  */
407 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
408 	if (bflags < 0)
409 		/* Some supported flag */
410 		return EINVAL;
411 
412 	msg.msg_flags = 0;
413 	msg.msg_name = NULL;
414 	msg.msg_control = NULL;
415 
416 	if (SCARG(uap, tolen)) {
417 		/* Read in and convert the sockaddr */
418 		error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
419 		    SCARG(uap, tolen));
420 		if (error)
421 			return (error);
422 		msg.msg_flags |= MSG_NAMEMBUF;
423 		msg.msg_name = nam;
424 		msg.msg_namelen = SCARG(uap, tolen);
425 	}
426 
427 	msg.msg_iov = &aiov;
428 	msg.msg_iovlen = 1;
429 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
430 	aiov.iov_len = SCARG(uap, len);
431 
432 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
433 }
434 
435 static void
436 linux_to_bsd_msghdr(struct linux_msghdr *lmsg, struct msghdr *bmsg)
437 {
438 	bmsg->msg_name = lmsg->msg_name;
439 	bmsg->msg_namelen = lmsg->msg_namelen;
440 	bmsg->msg_iov = lmsg->msg_iov;
441 	bmsg->msg_iovlen = lmsg->msg_iovlen;
442 	bmsg->msg_control = lmsg->msg_control;
443 	bmsg->msg_controllen = lmsg->msg_controllen;
444 	bmsg->msg_flags = lmsg->msg_flags;
445 }
446 
447 static void
448 bsd_to_linux_msghdr(struct msghdr *bmsg, struct linux_msghdr *lmsg)
449 {
450 	lmsg->msg_name = bmsg->msg_name;
451 	lmsg->msg_namelen = bmsg->msg_namelen;
452 	lmsg->msg_iov = bmsg->msg_iov;
453 	lmsg->msg_iovlen = bmsg->msg_iovlen;
454 	lmsg->msg_control = bmsg->msg_control;
455 	lmsg->msg_controllen = bmsg->msg_controllen;
456 	lmsg->msg_flags = bmsg->msg_flags;
457 }
458 
459 int
460 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
461 {
462 	/* {
463 		syscallarg(int) s;
464 		syscallarg(struct linux_msghdr *) msg;
465 		syscallarg(u_int) flags;
466 	} */
467 	struct msghdr	msg;
468 	struct linux_msghdr lmsg;
469 	int		error;
470 	int		bflags;
471 	struct mbuf     *nam;
472 	u_int8_t	*control;
473 	struct mbuf     *ctl_mbuf = NULL;
474 
475 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
476 	if (error)
477 		return error;
478 	linux_to_bsd_msghdr(&lmsg, &msg);
479 
480 	msg.msg_flags = MSG_IOVUSRSPACE;
481 
482 	/*
483 	 * Translate message flags.
484 	 */
485 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
486 	if (bflags < 0)
487 		/* Some supported flag */
488 		return EINVAL;
489 
490 	if (lmsg.msg_name) {
491 		/* Read in and convert the sockaddr */
492 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
493 		    msg.msg_namelen);
494 		if (error)
495 			return (error);
496 		msg.msg_flags |= MSG_NAMEMBUF;
497 		msg.msg_name = nam;
498 	}
499 
500 	/*
501 	 * Handle cmsg if there is any.
502 	 */
503 	if (LINUX_CMSG_FIRSTHDR(&lmsg)) {
504 		struct linux_cmsghdr l_cmsg, *l_cc;
505 		struct cmsghdr *cmsg;
506 		ssize_t resid = msg.msg_controllen;
507 		size_t clen, cidx = 0, cspace;
508 
509 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
510 		clen = MLEN;
511 		control = mtod(ctl_mbuf, void *);
512 
513 		l_cc = LINUX_CMSG_FIRSTHDR(&lmsg);
514 		do {
515 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
516 			if (error)
517 				goto done;
518 
519 			/*
520 			 * Sanity check the control message length.
521 			 */
522 			if (l_cmsg.cmsg_len > resid
523 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
524 				error = EINVAL;
525 				goto done;
526 			}
527 
528 			/*
529 			 * Refuse unsupported control messages, and
530 			 * translate fields as appropriate.
531 			 */
532 			switch (l_cmsg.cmsg_level) {
533 			case LINUX_SOL_SOCKET:
534 				/* It only differs on some archs */
535 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
536 					l_cmsg.cmsg_level = SOL_SOCKET;
537 
538 				switch(l_cmsg.cmsg_type) {
539 				case LINUX_SCM_RIGHTS:
540 					/* Linux SCM_RIGHTS is same as NetBSD */
541 					break;
542 
543 				case LINUX_SCM_CREDENTIALS:
544 					/* no native equivalent, just drop it */
545 					m_free(ctl_mbuf);
546 					ctl_mbuf = NULL;
547 					msg.msg_control = NULL;
548 					msg.msg_controllen = 0;
549 					goto skipcmsg;
550 
551 				default:
552 					/* other types not supported */
553 					error = EINVAL;
554 					goto done;
555 				}
556 				break;
557 			default:
558 				/* pray and leave intact */
559 				break;
560 			}
561 
562 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
563 
564 			/* Check the buffer is big enough */
565 			if (__predict_false(cidx + cspace > clen)) {
566 				u_int8_t *nc;
567 
568 				clen = cidx + cspace;
569 				if (clen >= PAGE_SIZE) {
570 					error = EINVAL;
571 					goto done;
572 				}
573 				nc = realloc(clen <= MLEN ? NULL : control,
574 						clen, M_TEMP, M_WAITOK);
575 				if (!nc) {
576 					error = ENOMEM;
577 					goto done;
578 				}
579 				if (cidx <= MLEN)
580 					/* Old buffer was in mbuf... */
581 					memcpy(nc, control, cidx);
582 				control = nc;
583 			}
584 
585 			/* Copy header */
586 			cmsg = (void *)&control[cidx];
587 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
588 			cmsg->cmsg_level = l_cmsg.cmsg_level;
589 			cmsg->cmsg_type = l_cmsg.cmsg_type;
590 
591 			/* Zero area between header and data */
592 			memset(cmsg + 1, 0,
593 				CMSG_ALIGN(sizeof(*cmsg)) - sizeof(*cmsg));
594 
595 			/* Copyin the data */
596 			error = copyin(LINUX_CMSG_DATA(l_cc),
597 				CMSG_DATA(cmsg),
598 				l_cmsg.cmsg_len - sizeof(l_cmsg));
599 			if (error)
600 				goto done;
601 
602 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
603 			cidx += cspace;
604 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
605 
606 		/* If we allocated a buffer, attach to mbuf */
607 		if (cidx > MLEN) {
608 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
609 			ctl_mbuf->m_flags |= M_EXT_RW;
610 		}
611 		control = NULL;
612 		ctl_mbuf->m_len = cidx;
613 
614 		msg.msg_control = ctl_mbuf;
615 		msg.msg_flags |= MSG_CONTROLMBUF;
616 
617 		ktrkuser("mbcontrol", mtod(ctl_mbuf, void *),
618 		    msg.msg_controllen);
619 	}
620 
621 skipcmsg:
622 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
623 	/* Freed internally */
624 	ctl_mbuf = NULL;
625 
626 done:
627 	if (ctl_mbuf != NULL) {
628 		if (control != NULL && control != mtod(ctl_mbuf, void *))
629 			free(control, M_MBUF);
630 		m_free(ctl_mbuf);
631 	}
632 	return (error);
633 }
634 
635 int
636 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
637 {
638 	/* {
639 		syscallarg(int) s;
640 		syscallarg(void *) buf;
641 		syscallarg(int) len;
642 		syscallarg(int) flags;
643 		syscallarg(struct osockaddr *) from;
644 		syscallarg(int *) fromlenaddr;
645 	} */
646 	int		error;
647 	struct sys_recvfrom_args bra;
648 
649 	SCARG(&bra, s) = SCARG(uap, s);
650 	SCARG(&bra, buf) = SCARG(uap, buf);
651 	SCARG(&bra, len) = SCARG(uap, len);
652 	SCARG(&bra, flags) = SCARG(uap, flags);
653 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
654 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
655 
656 	if ((error = sys_recvfrom(l, &bra, retval)))
657 		return (error);
658 
659 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
660 		return (error);
661 
662 	return (0);
663 }
664 
665 static int
666 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
667 {
668 	int dlen, error = 0;
669 	struct cmsghdr *cmsg;
670 	struct linux_cmsghdr linux_cmsg;
671 	struct mbuf *m;
672 	char *q, *q_end;
673 
674 	if (mp->msg_controllen <= 0 || control == 0) {
675 		mp->msg_controllen = 0;
676 		free_control_mbuf(l, control, control);
677 		return 0;
678 	}
679 
680 	ktrkuser("msgcontrol", mtod(control, void *), mp->msg_controllen);
681 
682 	q = (char *)mp->msg_control;
683 	q_end = q + mp->msg_controllen;
684 
685 	for (m = control; m != NULL; ) {
686 		cmsg = mtod(m, struct cmsghdr *);
687 
688 		/*
689 		 * Fixup cmsg. We handle two things:
690 		 * 0. different sizeof cmsg_len.
691 		 * 1. different values for level/type on some archs
692 		 * 2. different alignment of CMSG_DATA on some archs
693 		 */
694 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
695 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
696 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
697 
698 		dlen = q_end - q;
699 		if (linux_cmsg.cmsg_len > dlen) {
700 			/* Not enough room for the parameter */
701 			dlen -= sizeof linux_cmsg;
702 			if (dlen <= 0)
703 				/* Discard if header wont fit */
704 				break;
705 			mp->msg_flags |= MSG_CTRUNC;
706 			if (linux_cmsg.cmsg_level == SOL_SOCKET
707 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
708 				/* Do not truncate me ... */
709 				break;
710 		} else
711 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
712 
713 		switch (linux_cmsg.cmsg_level) {
714 		case SOL_SOCKET:
715 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
716 			switch (linux_cmsg.cmsg_type) {
717 			case SCM_RIGHTS:
718 				/* Linux SCM_RIGHTS is same as NetBSD */
719 				break;
720 
721 			default:
722 				/* other types not supported */
723 				error = EINVAL;
724 				goto done;
725 			}
726 			/* machine dependent ! */
727 			break;
728 		default:
729 			/* pray and leave intact */
730 			break;
731 		}
732 
733 		/* There can be padding between the header and data... */
734 		error = copyout(&linux_cmsg, q, sizeof linux_cmsg);
735 		if (error != 0) {
736 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
737 			    dlen);
738 		}
739 		if (error != 0) {
740 			/* We must free all the SCM_RIGHTS */
741 			m = control;
742 			break;
743 		}
744 		m = m->m_next;
745 		if (m == NULL || q + LINUX_CMSG_SPACE(dlen) > q_end) {
746 			q += LINUX_CMSG_LEN(dlen);
747 			break;
748 		}
749 		q += LINUX_CMSG_SPACE(dlen);
750 	}
751 
752   done:
753 	free_control_mbuf(l, control, m);
754 
755 	mp->msg_controllen = q - (char *)mp->msg_control;
756 	return error;
757 }
758 
759 int
760 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
761 {
762 	/* {
763 		syscallarg(int) s;
764 		syscallarg(struct linux_msghdr *) msg;
765 		syscallarg(u_int) flags;
766 	} */
767 	struct msghdr	msg;
768 	struct linux_msghdr lmsg;
769 	int		error;
770 	struct mbuf	*from, *control;
771 
772 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
773 	if (error)
774 		return (error);
775 	linux_to_bsd_msghdr(&lmsg, &msg);
776 
777 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
778 	if (msg.msg_flags < 0) {
779 		/* Some unsupported flag */
780 		return (EINVAL);
781 	}
782 	msg.msg_flags |= MSG_IOVUSRSPACE;
783 
784 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
785 	    msg.msg_control != NULL ? &control : NULL, retval);
786 	if (error != 0)
787 		return error;
788 
789 	if (msg.msg_control != NULL)
790 		error = linux_copyout_msg_control(l, &msg, control);
791 
792 	if (error == 0 && from != 0) {
793 		mtod(from, struct osockaddr *)->sa_family =
794 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
795 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
796 			from);
797 	} else
798 		msg.msg_namelen = 0;
799 
800 	if (from != NULL)
801 		m_free(from);
802 
803 	if (error == 0) {
804 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
805 		if (msg.msg_flags < 0)
806 			/* Some flag unsupported by Linux */
807 			error = EINVAL;
808 		else {
809 			ktrkuser("msghdr", &msg, sizeof(msg));
810 			bsd_to_linux_msghdr(&msg, &lmsg);
811 			error = copyout(&lmsg, SCARG(uap, msg), sizeof(lmsg));
812 		}
813 	}
814 
815 	return (error);
816 }
817 
818 /*
819  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
820  * is different, the rest matches IPPROTO_* on both systems.
821  */
822 int
823 linux_to_bsd_sopt_level(int llevel)
824 {
825 
826 	switch (llevel) {
827 	case LINUX_SOL_SOCKET:
828 		return SOL_SOCKET;
829 	case LINUX_SOL_IP:
830 		return IPPROTO_IP;
831 #ifdef INET6
832 	case LINUX_SOL_IPV6:
833 		return IPPROTO_IPV6;
834 #endif
835 	case LINUX_SOL_TCP:
836 		return IPPROTO_TCP;
837 	case LINUX_SOL_UDP:
838 		return IPPROTO_UDP;
839 	default:
840 		return -1;
841 	}
842 }
843 
844 /*
845  * Convert Linux socket level socket option numbers to NetBSD values.
846  */
847 int
848 linux_to_bsd_so_sockopt(int lopt)
849 {
850 
851 	switch (lopt) {
852 	case LINUX_SO_DEBUG:
853 		return SO_DEBUG;
854 	case LINUX_SO_REUSEADDR:
855 		/*
856 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
857 		 * host:port pair through SO_REUSEADDR even if the address is not a
858 		 * multicast-address.  Effectively, this means that we should use
859 		 * SO_REUSEPORT to allow Linux applications to not exit with
860 		 * EADDRINUSE
861 		 */
862 		return SO_REUSEPORT;
863 	case LINUX_SO_TYPE:
864 		return SO_TYPE;
865 	case LINUX_SO_ERROR:
866 		return SO_ERROR;
867 	case LINUX_SO_DONTROUTE:
868 		return SO_DONTROUTE;
869 	case LINUX_SO_BROADCAST:
870 		return SO_BROADCAST;
871 	case LINUX_SO_SNDBUF:
872 		return SO_SNDBUF;
873 	case LINUX_SO_RCVBUF:
874 		return SO_RCVBUF;
875 	case LINUX_SO_SNDLOWAT:
876 		return SO_SNDLOWAT;
877 	case LINUX_SO_RCVLOWAT:
878 		return SO_RCVLOWAT;
879 	case LINUX_SO_KEEPALIVE:
880 		return SO_KEEPALIVE;
881 	case LINUX_SO_OOBINLINE:
882 		return SO_OOBINLINE;
883 	case LINUX_SO_LINGER:
884 		return SO_LINGER;
885 	case LINUX_SO_ACCEPTCONN:
886 		return SO_ACCEPTCONN;
887 	case LINUX_SO_PRIORITY:
888 	case LINUX_SO_NO_CHECK:
889 	default:
890 		return -1;
891 	}
892 }
893 
894 /*
895  * Convert Linux IP level socket option number to NetBSD values.
896  */
897 int
898 linux_to_bsd_ip_sockopt(int lopt)
899 {
900 
901 	switch (lopt) {
902 	case LINUX_IP_TOS:
903 		return IP_TOS;
904 	case LINUX_IP_TTL:
905 		return IP_TTL;
906 	case LINUX_IP_HDRINCL:
907 		return IP_HDRINCL;
908 	case LINUX_IP_MULTICAST_TTL:
909 		return IP_MULTICAST_TTL;
910 	case LINUX_IP_MULTICAST_LOOP:
911 		return IP_MULTICAST_LOOP;
912 	case LINUX_IP_MULTICAST_IF:
913 		return IP_MULTICAST_IF;
914 	case LINUX_IP_ADD_MEMBERSHIP:
915 		return IP_ADD_MEMBERSHIP;
916 	case LINUX_IP_DROP_MEMBERSHIP:
917 		return IP_DROP_MEMBERSHIP;
918 	default:
919 		return -1;
920 	}
921 }
922 
923 /*
924  * Convert Linux IPV6 level socket option number to NetBSD values.
925  */
926 #ifdef INET6
927 int
928 linux_to_bsd_ipv6_sockopt(int lopt)
929 {
930 
931 	switch (lopt) {
932 	case LINUX_IPV6_V6ONLY:
933 		return IPV6_V6ONLY;
934 	default:
935 		return -1;
936 	}
937 }
938 #endif
939 
940 /*
941  * Convert Linux TCP level socket option number to NetBSD values.
942  */
943 int
944 linux_to_bsd_tcp_sockopt(int lopt)
945 {
946 
947 	switch (lopt) {
948 	case LINUX_TCP_NODELAY:
949 		return TCP_NODELAY;
950 	case LINUX_TCP_MAXSEG:
951 		return TCP_MAXSEG;
952 	default:
953 		return -1;
954 	}
955 }
956 
957 /*
958  * Convert Linux UDP level socket option number to NetBSD values.
959  */
960 int
961 linux_to_bsd_udp_sockopt(int lopt)
962 {
963 
964 	switch (lopt) {
965 	default:
966 		return -1;
967 	}
968 }
969 
970 /*
971  * Another reasonably straightforward function: setsockopt(2).
972  * The level and option numbers are converted; the values passed
973  * are not (yet) converted, the ones currently implemented don't
974  * need conversion, as they are the same on both systems.
975  */
976 int
977 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
978 {
979 	/* {
980 		syscallarg(int) s;
981 		syscallarg(int) level;
982 		syscallarg(int) optname;
983 		syscallarg(void *) optval;
984 		syscallarg(int) optlen;
985 	} */
986 	struct sys_setsockopt_args bsa;
987 	int name;
988 
989 	SCARG(&bsa, s) = SCARG(uap, s);
990 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
991 	SCARG(&bsa, val) = SCARG(uap, optval);
992 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
993 
994 	/*
995 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
996 	 * and returns EOPNOTSUPP for other levels
997 	 */
998 	if (SCARG(&bsa, level) != SOL_SOCKET) {
999 		struct socket *so;
1000 		int error, family;
1001 
1002 		/* fd_getsock() will use the descriptor for us */
1003 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
1004 		    	return error;
1005 		family = so->so_proto->pr_domain->dom_family;
1006 		fd_putfile(SCARG(&bsa, s));
1007 
1008 		if (family == AF_LOCAL)
1009 			return EOPNOTSUPP;
1010 	}
1011 
1012 	switch (SCARG(&bsa, level)) {
1013 	case SOL_SOCKET:
1014 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1015 		break;
1016 	case IPPROTO_IP:
1017 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1018 		break;
1019 #ifdef INET6
1020 	case IPPROTO_IPV6:
1021 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1022 		break;
1023 #endif
1024 	case IPPROTO_TCP:
1025 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1026 		break;
1027 	case IPPROTO_UDP:
1028 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1029 		break;
1030 	default:
1031 		return EINVAL;
1032 	}
1033 
1034 	if (name == -1)
1035 		return EINVAL;
1036 	SCARG(&bsa, name) = name;
1037 
1038 	return sys_setsockopt(l, &bsa, retval);
1039 }
1040 
1041 /*
1042  * getsockopt(2) is very much the same as setsockopt(2) (see above)
1043  */
1044 int
1045 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
1046 {
1047 	/* {
1048 		syscallarg(int) s;
1049 		syscallarg(int) level;
1050 		syscallarg(int) optname;
1051 		syscallarg(void *) optval;
1052 		syscallarg(int *) optlen;
1053 	} */
1054 	struct sys_getsockopt_args bga;
1055 	int name;
1056 
1057 	SCARG(&bga, s) = SCARG(uap, s);
1058 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
1059 	SCARG(&bga, val) = SCARG(uap, optval);
1060 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
1061 
1062 	switch (SCARG(&bga, level)) {
1063 	case SOL_SOCKET:
1064 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1065 		break;
1066 	case IPPROTO_IP:
1067 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1068 		break;
1069 #ifdef INET6
1070 	case IPPROTO_IPV6:
1071 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1072 		break;
1073 #endif
1074 	case IPPROTO_TCP:
1075 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1076 		break;
1077 	case IPPROTO_UDP:
1078 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1079 		break;
1080 	default:
1081 		return EINVAL;
1082 	}
1083 
1084 	if (name == -1)
1085 		return EINVAL;
1086 	SCARG(&bga, name) = name;
1087 
1088 	return sys_getsockopt(l, &bga, retval);
1089 }
1090 
1091 int
1092 linux_getifname(struct lwp *l, register_t *retval, void *data)
1093 {
1094 	struct ifnet *ifp;
1095 	struct linux_ifreq ifr;
1096 	int error;
1097 
1098 	error = copyin(data, &ifr, sizeof(ifr));
1099 	if (error)
1100 		return error;
1101 
1102 	ifp = if_byindex(ifr.ifr_ifru.ifru_ifindex);
1103 	if (ifp == NULL)
1104 		return ENODEV;
1105 
1106 	strncpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name));
1107 
1108 	return copyout(&ifr, data, sizeof(ifr));
1109 }
1110 
1111 int
1112 linux_getifconf(struct lwp *l, register_t *retval, void *data)
1113 {
1114 	struct linux_ifreq ifr, *ifrp = NULL;
1115 	struct linux_ifconf ifc;
1116 	struct ifnet *ifp;
1117 	struct ifaddr *ifa;
1118 	struct sockaddr *sa;
1119 	struct osockaddr *osa;
1120 	int space = 0, error = 0;
1121 	const int sz = (int)sizeof(ifr);
1122 	bool docopy;
1123 
1124 	error = copyin(data, &ifc, sizeof(ifc));
1125 	if (error)
1126 		return error;
1127 
1128 	docopy = ifc.ifc_req != NULL;
1129 	if (docopy) {
1130 		space = ifc.ifc_len;
1131 		ifrp = ifc.ifc_req;
1132 	}
1133 
1134 	IFNET_FOREACH(ifp) {
1135 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1136 		    sizeof(ifr.ifr_name));
1137 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
1138 			return ENAMETOOLONG;
1139 		if (IFADDR_EMPTY(ifp))
1140 			continue;
1141 		IFADDR_FOREACH(ifa, ifp) {
1142 			sa = ifa->ifa_addr;
1143 			if (sa->sa_family != AF_INET ||
1144 			    sa->sa_len > sizeof(*osa))
1145 				continue;
1146 			memcpy(&ifr.ifr_addr, sa, sa->sa_len);
1147 			osa = (struct osockaddr *)&ifr.ifr_addr;
1148 			osa->sa_family = sa->sa_family;
1149 			if (space >= sz) {
1150 				error = copyout(&ifr, ifrp, sz);
1151 				if (error != 0)
1152 					return error;
1153 				ifrp++;
1154 			}
1155 			space -= sz;
1156 		}
1157 	}
1158 
1159 	if (docopy)
1160 		ifc.ifc_len -= space;
1161 	else
1162 		ifc.ifc_len = -space;
1163 
1164 	return copyout(&ifc, data, sizeof(ifc));
1165 }
1166 
1167 int
1168 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1169     void *data)
1170 {
1171 	/* Not the full structure, just enough to map what we do here */
1172 	struct linux_ifreq lreq;
1173 	file_t *fp;
1174 	struct ifaddr *ifa;
1175 	struct ifnet *ifp;
1176 	struct sockaddr_dl *sadl;
1177 	int error, found;
1178 	int index, ifnum;
1179 
1180 	/*
1181 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1182 	 * SIOCGIFCONF, because the user buffer is not of the right
1183 	 * type to take those results.  We can't use kernel buffers to
1184 	 * receive the results, as the implementation of sys_ioctl()
1185 	 * and ifconf() [which implements SIOCGIFCONF] use
1186 	 * copyin()/copyout() which will fail on kernel addresses.
1187 	 *
1188 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1189 	 */
1190 
1191 	if ((fp = fd_getfile(fd)) == NULL)
1192 		return (EBADF);
1193 
1194 	KERNEL_LOCK(1, NULL);
1195 
1196 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1197 		error = EBADF;
1198 		goto out;
1199 	}
1200 
1201 	error = copyin(data, &lreq, sizeof(lreq));
1202 	if (error)
1203 		goto out;
1204 	lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0';		/* just in case */
1205 
1206 	/*
1207 	 * Try real interface name first, then fake "ethX"
1208 	 */
1209 	found = 0;
1210 	IFNET_FOREACH(ifp) {
1211 		if (found)
1212 			break;
1213 		if (strcmp(lreq.ifr_name, ifp->if_xname))
1214 			/* not this interface */
1215 			continue;
1216 		found=1;
1217 		if (IFADDR_EMPTY(ifp)) {
1218 			error = ENODEV;
1219 			goto out;
1220 		}
1221 		IFADDR_FOREACH(ifa, ifp) {
1222 			sadl = satosdl(ifa->ifa_addr);
1223 			/* only return ethernet addresses */
1224 			/* XXX what about FDDI, etc. ? */
1225 			if (sadl->sdl_family != AF_LINK ||
1226 			    sadl->sdl_type != IFT_ETHER)
1227 				continue;
1228 			memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl),
1229 			       MIN(sadl->sdl_alen,
1230 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1231 			lreq.ifr_hwaddr.sa_family =
1232 				sadl->sdl_family;
1233 			error = copyout(&lreq, data, sizeof(lreq));
1234 			goto out;
1235 		}
1236 	}
1237 
1238 	if (strncmp(lreq.ifr_name, "eth", 3) != 0) {
1239 		/* unknown interface, not even an "eth*" name */
1240 		error = ENODEV;
1241 		goto out;
1242 	}
1243 
1244 	for (ifnum = 0, index = 3;
1245 	     index < LINUX_IFNAMSIZ && lreq.ifr_name[index] != '\0';
1246 	     index++) {
1247 		ifnum *= 10;
1248 		ifnum += lreq.ifr_name[index] - '0';
1249 	}
1250 
1251 	error = EINVAL;			/* in case we don't find one */
1252 	found = 0;
1253 	IFNET_FOREACH(ifp) {
1254 		if (found)
1255 			break;
1256 		memcpy(lreq.ifr_name, ifp->if_xname,
1257 		       MIN(LINUX_IFNAMSIZ, IFNAMSIZ));
1258 		IFADDR_FOREACH(ifa, ifp) {
1259 			sadl = satosdl(ifa->ifa_addr);
1260 			/* only return ethernet addresses */
1261 			/* XXX what about FDDI, etc. ? */
1262 			if (sadl->sdl_family != AF_LINK ||
1263 			    sadl->sdl_type != IFT_ETHER)
1264 				continue;
1265 			if (ifnum--)
1266 				/* not the reqested iface */
1267 				continue;
1268 			memcpy(&lreq.ifr_hwaddr.sa_data,
1269 			       CLLADDR(sadl),
1270 			       MIN(sadl->sdl_alen,
1271 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1272 			lreq.ifr_hwaddr.sa_family =
1273 				sadl->sdl_family;
1274 			error = copyout(&lreq, data, sizeof(lreq));
1275 			found = 1;
1276 			break;
1277 		}
1278 	}
1279 
1280 out:
1281 	KERNEL_UNLOCK_ONE(NULL);
1282 	fd_putfile(fd);
1283 	return error;
1284 }
1285 
1286 int
1287 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1288 {
1289 	/* {
1290 		syscallarg(int) fd;
1291 		syscallarg(u_long) com;
1292 		syscallarg(void *) data;
1293 	} */
1294 	u_long com;
1295 	int error = 0, isdev = 0, dosys = 1;
1296 	struct sys_ioctl_args ia;
1297 	file_t *fp;
1298 	struct vnode *vp;
1299 	int (*ioctlf)(file_t *, u_long, void *);
1300 	struct ioctl_pt pt;
1301 
1302 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1303 		return (EBADF);
1304 
1305 	if (fp->f_type == DTYPE_VNODE) {
1306 		vp = (struct vnode *)fp->f_data;
1307 		isdev = vp->v_type == VCHR;
1308 	}
1309 
1310 	/*
1311 	 * Don't try to interpret socket ioctl calls that are done
1312 	 * on a device filedescriptor, just pass them through, to
1313 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1314 	 * device will only handle these if it's prepared to do
1315 	 * so, to avoid unexpected things from happening.
1316 	 */
1317 	if (isdev) {
1318 		dosys = 0;
1319 		ioctlf = fp->f_ops->fo_ioctl;
1320 		pt.com = SCARG(uap, com);
1321 		pt.data = SCARG(uap, data);
1322 		error = ioctlf(fp, PTIOCLINUX, &pt);
1323 		/*
1324 		 * XXX hack: if the function returns EJUSTRETURN,
1325 		 * it has stuffed a sysctl return value in pt.data.
1326 		 */
1327 		if (error == EJUSTRETURN) {
1328 			retval[0] = (register_t)pt.data;
1329 			error = 0;
1330 		}
1331 		goto out;
1332 	}
1333 
1334 	com = SCARG(uap, com);
1335 	retval[0] = 0;
1336 
1337 	switch (com) {
1338 	case LINUX_SIOCGIFNAME:
1339 		error = linux_getifname(l, retval, SCARG(uap, data));
1340 		dosys = 0;
1341 		break;
1342 	case LINUX_SIOCGIFCONF:
1343 		error = linux_getifconf(l, retval, SCARG(uap, data));
1344 		dosys = 0;
1345 		break;
1346 	case LINUX_SIOCGIFFLAGS:
1347 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1348 		break;
1349 	case LINUX_SIOCSIFFLAGS:
1350 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1351 		break;
1352 	case LINUX_SIOCGIFADDR:
1353 		SCARG(&ia, com) = OOSIOCGIFADDR;
1354 		break;
1355 	case LINUX_SIOCGIFDSTADDR:
1356 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1357 		break;
1358 	case LINUX_SIOCGIFBRDADDR:
1359 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1360 		break;
1361 	case LINUX_SIOCGIFNETMASK:
1362 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1363 		break;
1364 	case LINUX_SIOCGIFMTU:
1365 		SCARG(&ia, com) = OSIOCGIFMTU;
1366 		break;
1367 	case LINUX_SIOCADDMULTI:
1368 		SCARG(&ia, com) = OSIOCADDMULTI;
1369 		break;
1370 	case LINUX_SIOCDELMULTI:
1371 		SCARG(&ia, com) = OSIOCDELMULTI;
1372 		break;
1373 	case LINUX_SIOCGIFHWADDR:
1374 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1375 		    SCARG(uap, data));
1376 		dosys = 0;
1377 		break;
1378 	default:
1379 		error = EINVAL;
1380 	}
1381 
1382  out:
1383  	fd_putfile(SCARG(uap, fd));
1384 
1385 	if (error ==0 && dosys) {
1386 		SCARG(&ia, fd) = SCARG(uap, fd);
1387 		SCARG(&ia, data) = SCARG(uap, data);
1388 		error = sys_ioctl(curlwp, &ia, retval);
1389 	}
1390 
1391 	return error;
1392 }
1393 
1394 int
1395 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1396 {
1397 	/* {
1398 		syscallarg(int) s;
1399 		syscallarg(const struct sockaddr *) name;
1400 		syscallarg(int) namelen;
1401 	} */
1402 	int		error;
1403 	struct mbuf *nam;
1404 
1405 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1406 	    SCARG(uap, namelen));
1407 	if (error)
1408 		return (error);
1409 
1410 	error = do_sys_connect(l, SCARG(uap, s), nam);
1411 
1412 	if (error == EISCONN) {
1413 		struct socket *so;
1414 		int state, prflags;
1415 
1416 		/* fd_getsock() will use the descriptor for us */
1417 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1418 		    	return EISCONN;
1419 
1420 		solock(so);
1421 		state = so->so_state;
1422 		prflags = so->so_proto->pr_flags;
1423 		sounlock(so);
1424 		fd_putfile(SCARG(uap, s));
1425 		/*
1426 		 * We should only let this call succeed once per
1427 		 * non-blocking connect; however we don't have
1428 		 * a convenient place to keep that state..
1429 		 */
1430 		if ((state & (SS_ISCONNECTED|SS_NBIO)) ==
1431 		    (SS_ISCONNECTED|SS_NBIO) &&
1432 		    (prflags & PR_CONNREQUIRED))
1433 			return 0;
1434 	}
1435 
1436 	return (error);
1437 }
1438 
1439 int
1440 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1441 {
1442 	/* {
1443 		syscallarg(int) s;
1444 		syscallarg(const struct osockaddr *) name;
1445 		syscallarg(int) namelen;
1446 	} */
1447 	int		error;
1448 	struct mbuf     *nam;
1449 
1450 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1451 	    SCARG(uap, namelen));
1452 	if (error)
1453 		return (error);
1454 
1455 	return do_sys_bind(l, SCARG(uap, s), nam);
1456 }
1457 
1458 int
1459 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1460 {
1461 	/* {
1462 		syscallarg(int) fdes;
1463 		syscallarg(void *) asa;
1464 		syscallarg(int *) alen;
1465 	} */
1466 	int error;
1467 
1468 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1469 		return (error);
1470 
1471 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1472 		return (error);
1473 
1474 	return (0);
1475 }
1476 
1477 int
1478 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1479 {
1480 	/* {
1481 		syscallarg(int) fdes;
1482 		syscallarg(void *) asa;
1483 		syscallarg(int *) alen;
1484 	} */
1485 	int error;
1486 
1487 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1488 		return (error);
1489 
1490 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1491 		return (error);
1492 
1493 	return (0);
1494 }
1495 
1496 /*
1497  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1498  * family and convert to sockaddr.
1499  */
1500 static int
1501 linux_get_sa(struct lwp *l, int s, struct mbuf **mp,
1502     const struct osockaddr *osa, unsigned int salen)
1503 {
1504 	int error, bdom;
1505 	struct sockaddr *sa;
1506 	struct osockaddr *kosa;
1507 	struct mbuf *m;
1508 
1509 	if (salen == 1 || salen > UCHAR_MAX) {
1510 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1511 		return EINVAL;
1512 	}
1513 
1514 	/* We'll need the address in an mbuf later, so copy into one here */
1515 	m = m_get(M_WAIT, MT_SONAME);
1516 	if (salen > MLEN)
1517 		MEXTMALLOC(m, salen, M_WAITOK);
1518 
1519 	m->m_len = salen;
1520 
1521 	if (salen == 0) {
1522 		*mp = m;
1523 		return 0;
1524 	}
1525 
1526 	kosa = mtod(m, void *);
1527 	if ((error = copyin(osa, kosa, salen))) {
1528 		DPRINTF(("error %d copying osa %p len %d\n",
1529 				error, osa, salen));
1530 		goto bad;
1531 	}
1532 
1533 	ktrkuser("linux/sockaddr", kosa, salen);
1534 
1535 	bdom = linux_to_bsd_domain(kosa->sa_family);
1536 	if (bdom == -1) {
1537 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1538 		error = EINVAL;
1539 		goto bad;
1540 	}
1541 
1542 	/*
1543 	 * If the family is unspecified, use address family of the socket.
1544 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1545 	 */
1546 	if (bdom == AF_UNSPEC) {
1547 		struct socket *so;
1548 
1549 		/* fd_getsock() will use the descriptor for us */
1550 		if ((error = fd_getsock(s, &so)) != 0)
1551 			goto bad;
1552 
1553 		bdom = so->so_proto->pr_domain->dom_family;
1554 		fd_putfile(s);
1555 
1556 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1557 	}
1558 
1559 	/*
1560 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1561 	 * which lacks the scope id compared with RFC2553 one. If we detect
1562 	 * the situation, reject the address and write a message to system log.
1563 	 *
1564 	 * Still accept addresses for which the scope id is not used.
1565 	 */
1566 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1567 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1568 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1569 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1570 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1571 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1572 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1573 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1574 			struct proc *p = l->l_proc;
1575 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1576 
1577 			log(LOG_DEBUG,
1578 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1579 			    "sockaddr_in6 rejected",
1580 			    p->p_pid, p->p_comm, uid);
1581 			error = EINVAL;
1582 			goto bad;
1583 		}
1584 		salen = sizeof (struct sockaddr_in6);
1585 		sin6->sin6_scope_id = 0;
1586 	}
1587 
1588 	if (bdom == AF_INET)
1589 		salen = sizeof(struct sockaddr_in);
1590 
1591 	sa = (struct sockaddr *) kosa;
1592 	sa->sa_family = bdom;
1593 	sa->sa_len = salen;
1594 	m->m_len = salen;
1595 	ktrkuser("mbsoname", kosa, salen);
1596 
1597 #ifdef DEBUG_LINUX
1598 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1599 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1600 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1601 	DPRINTF(("\n"));
1602 #endif
1603 
1604 	*mp = m;
1605 	return 0;
1606 
1607     bad:
1608 	m_free(m);
1609 	return error;
1610 }
1611 
1612 static int
1613 linux_sa_put(struct osockaddr *osa)
1614 {
1615 	struct sockaddr sa;
1616 	struct osockaddr *kosa;
1617 	int error, bdom, len;
1618 
1619 	/*
1620 	 * Only read/write the sockaddr family and length part, the rest is
1621 	 * not changed.
1622 	 */
1623 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1624 
1625 	error = copyin(osa, &sa, len);
1626 	if (error)
1627 		return (error);
1628 
1629 	bdom = bsd_to_linux_domain(sa.sa_family);
1630 	if (bdom == -1)
1631 		return (EINVAL);
1632 
1633 	/* Note: we convert from sockaddr to osockaddr here, too */
1634 	kosa = (struct osockaddr *) &sa;
1635 	kosa->sa_family = bdom;
1636 	error = copyout(kosa, osa, len);
1637 	if (error)
1638 		return (error);
1639 
1640 	return (0);
1641 }
1642 
1643 #ifndef __amd64__
1644 int
1645 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1646 {
1647 	/* {
1648 		syscallarg(int) s;
1649 		syscallarg(void *) buf;
1650 		syscallarg(int) len;
1651 		syscallarg(int) flags;
1652 	} */
1653 	struct sys_recvfrom_args bra;
1654 
1655 
1656 	SCARG(&bra, s) = SCARG(uap, s);
1657 	SCARG(&bra, buf) = SCARG(uap, buf);
1658 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1659 	SCARG(&bra, flags) = SCARG(uap, flags);
1660 	SCARG(&bra, from) = NULL;
1661 	SCARG(&bra, fromlenaddr) = NULL;
1662 
1663 	return (sys_recvfrom(l, &bra, retval));
1664 }
1665 
1666 int
1667 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1668 {
1669 	/* {
1670 		syscallarg(int) s;
1671 		syscallarg(void *) buf;
1672 		syscallarg(int) len;
1673 		syscallarg(int) flags;
1674 	} */
1675 	struct sys_sendto_args bsa;
1676 
1677 	SCARG(&bsa, s)		= SCARG(uap, s);
1678 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1679 	SCARG(&bsa, len)	= SCARG(uap, len);
1680 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1681 	SCARG(&bsa, to)		= NULL;
1682 	SCARG(&bsa, tolen)	= 0;
1683 
1684 	return (sys_sendto(l, &bsa, retval));
1685 }
1686 #endif
1687 
1688 int
1689 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1690 {
1691 	/* {
1692 		syscallarg(int) s;
1693 		syscallarg(struct osockaddr *) name;
1694 		syscallarg(int *) anamelen;
1695 	} */
1696 	int error;
1697 	struct sys_accept_args baa;
1698 
1699 	SCARG(&baa, s)		= SCARG(uap, s);
1700 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1701 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1702 
1703 	if ((error = sys_accept(l, &baa, retval)))
1704 		return (error);
1705 
1706 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1707 		return (error);
1708 
1709 	return (0);
1710 }
1711