xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision 03dcb730d46d34d85c9f496c1f5a3a6a43f2b7b3)
1 /*	$NetBSD: linux_socket.c,v 1.138 2017/03/14 09:03:08 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 1.138 2017/03/14 09:03:08 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_sched.h>
86 #include <compat/linux/common/linux_socket.h>
87 #include <compat/linux/common/linux_fcntl.h>
88 #if !defined(__alpha__) && !defined(__amd64__)
89 #include <compat/linux/common/linux_socketcall.h>
90 #endif
91 #include <compat/linux/common/linux_sockio.h>
92 #include <compat/linux/common/linux_ipc.h>
93 #include <compat/linux/common/linux_sem.h>
94 
95 #include <compat/linux/linux_syscallargs.h>
96 
97 #ifdef DEBUG_LINUX
98 #define DPRINTF(a) uprintf a
99 #else
100 #define DPRINTF(a)
101 #endif
102 
103 /*
104  * The calls in this file are entered either via the linux_socketcall()
105  * interface or, on the Alpha, as individual syscalls.  The
106  * linux_socketcall function does any massaging of arguments so that all
107  * the calls in here need not think that they are anything other
108  * than a normal syscall.
109  */
110 
111 static int linux_to_bsd_domain(int);
112 static int bsd_to_linux_domain(int);
113 static int linux_to_bsd_type(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_ipv6_sockopt(int);
118 int linux_to_bsd_tcp_sockopt(int);
119 int linux_to_bsd_udp_sockopt(int);
120 int linux_getifname(struct lwp *, register_t *, void *);
121 int linux_getifconf(struct lwp *, register_t *, void *);
122 int linux_getifhwaddr(struct lwp *, register_t *, u_int, void *);
123 static int linux_get_sa(struct lwp *, int, struct sockaddr_big *,
124 		const struct osockaddr *, socklen_t);
125 static int linux_sa_put(struct osockaddr *osa);
126 static int linux_to_bsd_msg_flags(int);
127 static int bsd_to_linux_msg_flags(int);
128 static void linux_to_bsd_msghdr(const struct linux_msghdr *, struct msghdr *);
129 static void bsd_to_linux_msghdr(const struct msghdr *, struct linux_msghdr *);
130 
131 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
132 	AF_UNSPEC,
133 	AF_UNIX,
134 	AF_INET,
135 	AF_CCITT,	/* LINUX_AF_AX25 */
136 	AF_IPX,
137 	AF_APPLETALK,
138 	-1,		/* LINUX_AF_NETROM */
139 	-1,		/* LINUX_AF_BRIDGE */
140 	-1,		/* LINUX_AF_ATMPVC */
141 	AF_CCITT,	/* LINUX_AF_X25 */
142 	AF_INET6,
143 	-1,		/* LINUX_AF_ROSE */
144 	AF_DECnet,
145 	-1,		/* LINUX_AF_NETBEUI */
146 	-1,		/* LINUX_AF_SECURITY */
147 	pseudo_AF_KEY,
148 	AF_ROUTE,	/* LINUX_AF_NETLINK */
149 	-1,		/* LINUX_AF_PACKET */
150 	-1,		/* LINUX_AF_ASH */
151 	-1,		/* LINUX_AF_ECONET */
152 	-1,		/* LINUX_AF_ATMSVC */
153 	AF_SNA,
154 	/* rest up to LINUX_AF_MAX-1 is not allocated */
155 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
156 };
157 
158 static const int bsd_to_linux_domain_[AF_MAX] = {
159 	LINUX_AF_UNSPEC,
160 	LINUX_AF_UNIX,
161 	LINUX_AF_INET,
162 	-1,		/* AF_IMPLINK */
163 	-1,		/* AF_PUP */
164 	-1,		/* AF_CHAOS */
165 	-1,		/* AF_NS */
166 	-1,		/* AF_ISO */
167 	-1,		/* AF_ECMA */
168 	-1,		/* AF_DATAKIT */
169 	LINUX_AF_AX25,	/* AF_CCITT */
170 	LINUX_AF_SNA,
171 	LINUX_AF_DECnet,
172 	-1,		/* AF_DLI */
173 	-1,		/* AF_LAT */
174 	-1,		/* AF_HYLINK */
175 	LINUX_AF_APPLETALK,
176 	LINUX_AF_NETLINK,
177 	-1,		/* AF_LINK */
178 	-1,		/* AF_XTP */
179 	-1,		/* AF_COIP */
180 	-1,		/* AF_CNT */
181 	-1,		/* pseudo_AF_RTIP */
182 	LINUX_AF_IPX,
183 	LINUX_AF_INET6,
184 	-1,		/* pseudo_AF_PIP */
185 	-1,		/* AF_ISDN */
186 	-1,		/* AF_NATM */
187 	-1,		/* AF_ARP */
188 	LINUX_pseudo_AF_KEY,
189 	-1,		/* pseudo_AF_HDRCMPLT */
190 };
191 
192 static const struct {
193 	int bfl;
194 	int lfl;
195 } bsd_to_linux_msg_flags_[] = {
196 	{MSG_OOB,		LINUX_MSG_OOB},
197 	{MSG_PEEK,		LINUX_MSG_PEEK},
198 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
199 	{MSG_EOR,		LINUX_MSG_EOR},
200 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
201 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
202 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
203 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
204 	{MSG_BCAST,		0},		/* not supported, clear */
205 	{MSG_MCAST,		0},		/* not supported, clear */
206 	{MSG_NOSIGNAL,		LINUX_MSG_NOSIGNAL},
207 	{-1, /* not supp */	LINUX_MSG_PROBE},
208 	{-1, /* not supp */	LINUX_MSG_FIN},
209 	{-1, /* not supp */	LINUX_MSG_SYN},
210 	{-1, /* not supp */	LINUX_MSG_CONFIRM},
211 	{-1, /* not supp */	LINUX_MSG_RST},
212 	{-1, /* not supp */	LINUX_MSG_ERRQUEUE},
213 	{-1, /* not supp */	LINUX_MSG_MORE},
214 };
215 
216 /*
217  * Convert between Linux and BSD socket domain values
218  */
219 static int
220 linux_to_bsd_domain(int ldom)
221 {
222 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
223 		return (-1);
224 
225 	return linux_to_bsd_domain_[ldom];
226 }
227 
228 /*
229  * Convert between BSD and Linux socket domain values
230  */
231 static int
232 bsd_to_linux_domain(int bdom)
233 {
234 	if (bdom < 0 || bdom >= AF_MAX)
235 		return (-1);
236 
237 	return bsd_to_linux_domain_[bdom];
238 }
239 
240 static int
241 linux_to_bsd_type(int ltype)
242 {
243 	int type, flags;
244 
245 	/* Real types are identical between Linux and NetBSD */
246 	type = ltype & LINUX_SOCK_TYPE_MASK;
247 
248 	/* But flags are not .. */
249 	flags = ltype & ~LINUX_SOCK_TYPE_MASK;
250 	if (flags & ~(LINUX_SOCK_CLOEXEC|LINUX_SOCK_NONBLOCK))
251 		return -1;
252 
253 	if (flags & LINUX_SOCK_CLOEXEC)
254 		type |= SOCK_CLOEXEC;
255 	if (flags & LINUX_SOCK_NONBLOCK)
256 		type |= SOCK_NONBLOCK;
257 
258 	return type;
259 }
260 
261 static int
262 linux_to_bsd_msg_flags(int lflag)
263 {
264 	int i, lfl, bfl;
265 	int bflag = 0;
266 
267 	if (lflag == 0)
268 		return (0);
269 
270 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
271 		bfl = bsd_to_linux_msg_flags_[i].bfl;
272 		lfl = bsd_to_linux_msg_flags_[i].lfl;
273 
274 		if (lfl == 0)
275 			continue;
276 
277 		if (lflag & lfl) {
278 			if (bfl < 0)
279 				return (-1);
280 
281 			bflag |= bfl;
282 		}
283 	}
284 
285 	return (bflag);
286 }
287 
288 static int
289 bsd_to_linux_msg_flags(int bflag)
290 {
291 	int i, lfl, bfl;
292 	int lflag = 0;
293 
294 	if (bflag == 0)
295 		return (0);
296 
297 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
298 		bfl = bsd_to_linux_msg_flags_[i].bfl;
299 		lfl = bsd_to_linux_msg_flags_[i].lfl;
300 
301 		if (bfl <= 0)
302 			continue;
303 
304 		if (bflag & bfl) {
305 			if (lfl < 0)
306 				return (-1);
307 
308 			lflag |= lfl;
309 		}
310 	}
311 
312 	return (lflag);
313 }
314 
315 int
316 linux_sys_socket(struct lwp *l, const struct linux_sys_socket_args *uap, register_t *retval)
317 {
318 	/* {
319 		syscallarg(int)	domain;
320 		syscallarg(int)	type;
321 		syscallarg(int) protocol;
322 	} */
323 	struct sys___socket30_args bsa;
324 	int error;
325 
326 
327 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
328 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
329 	if (SCARG(&bsa, domain) == -1)
330 		return EINVAL;
331 	SCARG(&bsa, type) = linux_to_bsd_type(SCARG(uap, type));
332 	if (SCARG(&bsa, type) == -1)
333 		return EINVAL;
334 	/*
335 	 * Apparently linux uses this to talk to ISDN sockets. If we fail
336 	 * now programs seems to handle it, but if we don't we are going
337 	 * to fail when we bind and programs don't handle this well.
338 	 */
339 	if (SCARG(&bsa, domain) == AF_ROUTE && SCARG(&bsa, type) == SOCK_RAW)
340 		return ENOTSUP;
341 	error = sys___socket30(l, &bsa, retval);
342 
343 #ifdef INET6
344 	/*
345 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
346 	 * default and some apps depend on this. So, set V6ONLY to 0
347 	 * for Linux apps if the sysctl value is set to 1.
348 	 */
349 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
350 		struct socket *so;
351 
352 		if (fd_getsock(*retval, &so) == 0) {
353 			int val = 0;
354 
355 			/* ignore error */
356 			(void)so_setsockopt(l, so, IPPROTO_IPV6, IPV6_V6ONLY,
357 			    &val, sizeof(val));
358 
359 			fd_putfile(*retval);
360 		}
361 	}
362 #endif
363 
364 	return (error);
365 }
366 
367 int
368 linux_sys_socketpair(struct lwp *l, const struct linux_sys_socketpair_args *uap, register_t *retval)
369 {
370 	/* {
371 		syscallarg(int) domain;
372 		syscallarg(int) type;
373 		syscallarg(int) protocol;
374 		syscallarg(int *) rsv;
375 	} */
376 	struct sys_socketpair_args bsa;
377 
378 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
379 	if (SCARG(&bsa, domain) == -1)
380 		return EINVAL;
381 	SCARG(&bsa, type) = linux_to_bsd_type(SCARG(uap, type));
382 	if (SCARG(&bsa, type) == -1)
383 		return EINVAL;
384 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
385 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
386 
387 	return sys_socketpair(l, &bsa, retval);
388 }
389 
390 int
391 linux_sys_sendto(struct lwp *l, const struct linux_sys_sendto_args *uap, register_t *retval)
392 {
393 	/* {
394 		syscallarg(int)				s;
395 		syscallarg(void *)			msg;
396 		syscallarg(int)				len;
397 		syscallarg(int)				flags;
398 		syscallarg(struct osockaddr *)		to;
399 		syscallarg(int)				tolen;
400 	} */
401 	struct msghdr   msg;
402 	struct iovec    aiov;
403 	struct sockaddr_big nam;
404 	int bflags;
405 	int error;
406 
407 	/* Translate message flags.  */
408 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
409 	if (bflags < 0)
410 		/* Some supported flag */
411 		return EINVAL;
412 
413 	msg.msg_flags = 0;
414 	msg.msg_name = NULL;
415 	msg.msg_control = NULL;
416 
417 	if (SCARG(uap, tolen)) {
418 		/* Read in and convert the sockaddr */
419 		error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
420 		    SCARG(uap, tolen));
421 		if (error)
422 			return (error);
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,
433 	    NULL, 0, retval);
434 }
435 
436 static void
437 linux_to_bsd_msghdr(const struct linux_msghdr *lmsg, struct msghdr *bmsg)
438 {
439 	bmsg->msg_name = lmsg->msg_name;
440 	bmsg->msg_namelen = lmsg->msg_namelen;
441 	bmsg->msg_iov = lmsg->msg_iov;
442 	bmsg->msg_iovlen = lmsg->msg_iovlen;
443 	bmsg->msg_control = lmsg->msg_control;
444 	bmsg->msg_controllen = lmsg->msg_controllen;
445 	bmsg->msg_flags = lmsg->msg_flags;
446 }
447 
448 static void
449 bsd_to_linux_msghdr(const struct msghdr *bmsg, struct linux_msghdr *lmsg)
450 {
451 	lmsg->msg_name = bmsg->msg_name;
452 	lmsg->msg_namelen = bmsg->msg_namelen;
453 	lmsg->msg_iov = bmsg->msg_iov;
454 	lmsg->msg_iovlen = bmsg->msg_iovlen;
455 	lmsg->msg_control = bmsg->msg_control;
456 	lmsg->msg_controllen = bmsg->msg_controllen;
457 	lmsg->msg_flags = bmsg->msg_flags;
458 }
459 
460 int
461 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
462 {
463 	/* {
464 		syscallarg(int) s;
465 		syscallarg(struct linux_msghdr *) msg;
466 		syscallarg(u_int) flags;
467 	} */
468 	struct msghdr	msg;
469 	struct linux_msghdr lmsg;
470 	int		error;
471 	int		bflags;
472 	struct sockaddr_big nam;
473 	u_int8_t	*control;
474 	struct mbuf     *ctl_mbuf = NULL;
475 
476 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
477 	if (error)
478 		return error;
479 	linux_to_bsd_msghdr(&lmsg, &msg);
480 
481 	msg.msg_flags = MSG_IOVUSRSPACE;
482 
483 	/*
484 	 * Translate message flags.
485 	 */
486 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
487 	if (bflags < 0)
488 		/* Some supported flag */
489 		return EINVAL;
490 
491 	if (lmsg.msg_name) {
492 		/* Read in and convert the sockaddr */
493 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
494 		    msg.msg_namelen);
495 		if (error)
496 			return (error);
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,
623 	    NULL, 0, retval);
624 	/* Freed internally */
625 	ctl_mbuf = NULL;
626 
627 done:
628 	if (ctl_mbuf != NULL) {
629 		if (control != NULL && control != mtod(ctl_mbuf, void *))
630 			free(control, M_MBUF);
631 		m_free(ctl_mbuf);
632 	}
633 	return (error);
634 }
635 
636 int
637 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
638 {
639 	/* {
640 		syscallarg(int) s;
641 		syscallarg(void *) buf;
642 		syscallarg(int) len;
643 		syscallarg(int) flags;
644 		syscallarg(struct osockaddr *) from;
645 		syscallarg(int *) fromlenaddr;
646 	} */
647 	int		error;
648 	struct sys_recvfrom_args bra;
649 
650 	SCARG(&bra, s) = SCARG(uap, s);
651 	SCARG(&bra, buf) = SCARG(uap, buf);
652 	SCARG(&bra, len) = SCARG(uap, len);
653 	SCARG(&bra, flags) = SCARG(uap, flags);
654 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
655 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
656 
657 	if ((error = sys_recvfrom(l, &bra, retval)))
658 		return (error);
659 
660 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
661 		return (error);
662 
663 	return (0);
664 }
665 
666 static int
667 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
668 {
669 	int dlen, error = 0;
670 	struct cmsghdr *cmsg;
671 	struct linux_cmsghdr linux_cmsg;
672 	struct mbuf *m;
673 	char *q, *q_end;
674 
675 	if (mp->msg_controllen <= 0 || control == 0) {
676 		mp->msg_controllen = 0;
677 		free_control_mbuf(l, control, control);
678 		return 0;
679 	}
680 
681 	ktrkuser("msgcontrol", mtod(control, void *), mp->msg_controllen);
682 
683 	q = (char *)mp->msg_control;
684 	q_end = q + mp->msg_controllen;
685 
686 	for (m = control; m != NULL; ) {
687 		cmsg = mtod(m, struct cmsghdr *);
688 
689 		/*
690 		 * Fixup cmsg. We handle two things:
691 		 * 0. different sizeof cmsg_len.
692 		 * 1. different values for level/type on some archs
693 		 * 2. different alignment of CMSG_DATA on some archs
694 		 */
695 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
696 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
697 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
698 
699 		dlen = q_end - q;
700 		if (linux_cmsg.cmsg_len > dlen) {
701 			/* Not enough room for the parameter */
702 			dlen -= sizeof linux_cmsg;
703 			if (dlen <= 0)
704 				/* Discard if header wont fit */
705 				break;
706 			mp->msg_flags |= MSG_CTRUNC;
707 			if (linux_cmsg.cmsg_level == SOL_SOCKET
708 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
709 				/* Do not truncate me ... */
710 				break;
711 		} else
712 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
713 
714 		switch (linux_cmsg.cmsg_level) {
715 		case SOL_SOCKET:
716 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
717 			switch (linux_cmsg.cmsg_type) {
718 			case SCM_RIGHTS:
719 				/* Linux SCM_RIGHTS is same as NetBSD */
720 				break;
721 
722 			default:
723 				/* other types not supported */
724 				error = EINVAL;
725 				goto done;
726 			}
727 			/* machine dependent ! */
728 			break;
729 		default:
730 			/* pray and leave intact */
731 			break;
732 		}
733 
734 		/* There can be padding between the header and data... */
735 		error = copyout(&linux_cmsg, q, sizeof linux_cmsg);
736 		if (error != 0) {
737 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
738 			    dlen);
739 		}
740 		if (error != 0) {
741 			/* We must free all the SCM_RIGHTS */
742 			m = control;
743 			break;
744 		}
745 		m = m->m_next;
746 		if (m == NULL || q + LINUX_CMSG_SPACE(dlen) > q_end) {
747 			q += LINUX_CMSG_LEN(dlen);
748 			break;
749 		}
750 		q += LINUX_CMSG_SPACE(dlen);
751 	}
752 
753   done:
754 	free_control_mbuf(l, control, m);
755 
756 	mp->msg_controllen = q - (char *)mp->msg_control;
757 	return error;
758 }
759 
760 int
761 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
762 {
763 	/* {
764 		syscallarg(int) s;
765 		syscallarg(struct linux_msghdr *) msg;
766 		syscallarg(u_int) flags;
767 	} */
768 	struct msghdr	msg;
769 	struct linux_msghdr lmsg;
770 	int		error;
771 	struct mbuf	*from, *control;
772 
773 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
774 	if (error)
775 		return (error);
776 	linux_to_bsd_msghdr(&lmsg, &msg);
777 
778 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
779 	if (msg.msg_flags < 0) {
780 		/* Some unsupported flag */
781 		return (EINVAL);
782 	}
783 	msg.msg_flags |= MSG_IOVUSRSPACE;
784 
785 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, NULL, 0, &from,
786 	    msg.msg_control != NULL ? &control : NULL, retval);
787 	if (error != 0)
788 		return error;
789 
790 	if (msg.msg_control != NULL)
791 		error = linux_copyout_msg_control(l, &msg, control);
792 
793 	if (error == 0 && from != 0) {
794 		mtod(from, struct osockaddr *)->sa_family =
795 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
796 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
797 			from);
798 	} else
799 		msg.msg_namelen = 0;
800 
801 	if (from != NULL)
802 		m_free(from);
803 
804 	if (error == 0) {
805 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
806 		if (msg.msg_flags < 0)
807 			/* Some flag unsupported by Linux */
808 			error = EINVAL;
809 		else {
810 			ktrkuser("msghdr", &msg, sizeof(msg));
811 			bsd_to_linux_msghdr(&msg, &lmsg);
812 			error = copyout(&lmsg, SCARG(uap, msg), sizeof(lmsg));
813 		}
814 	}
815 
816 	return (error);
817 }
818 
819 /*
820  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
821  * is different, the rest matches IPPROTO_* on both systems.
822  */
823 int
824 linux_to_bsd_sopt_level(int llevel)
825 {
826 
827 	switch (llevel) {
828 	case LINUX_SOL_SOCKET:
829 		return SOL_SOCKET;
830 	case LINUX_SOL_IP:
831 		return IPPROTO_IP;
832 #ifdef INET6
833 	case LINUX_SOL_IPV6:
834 		return IPPROTO_IPV6;
835 #endif
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_SNDLOWAT:
877 		return SO_SNDLOWAT;
878 	case LINUX_SO_RCVLOWAT:
879 		return SO_RCVLOWAT;
880 	case LINUX_SO_KEEPALIVE:
881 		return SO_KEEPALIVE;
882 	case LINUX_SO_OOBINLINE:
883 		return SO_OOBINLINE;
884 	case LINUX_SO_LINGER:
885 		return SO_LINGER;
886 	case LINUX_SO_ACCEPTCONN:
887 		return SO_ACCEPTCONN;
888 	case LINUX_SO_PRIORITY:
889 	case LINUX_SO_NO_CHECK:
890 	default:
891 		return -1;
892 	}
893 }
894 
895 /*
896  * Convert Linux IP level socket option number to NetBSD values.
897  */
898 int
899 linux_to_bsd_ip_sockopt(int lopt)
900 {
901 
902 	switch (lopt) {
903 	case LINUX_IP_TOS:
904 		return IP_TOS;
905 	case LINUX_IP_TTL:
906 		return IP_TTL;
907 	case LINUX_IP_HDRINCL:
908 		return IP_HDRINCL;
909 	case LINUX_IP_MULTICAST_TTL:
910 		return IP_MULTICAST_TTL;
911 	case LINUX_IP_MULTICAST_LOOP:
912 		return IP_MULTICAST_LOOP;
913 	case LINUX_IP_MULTICAST_IF:
914 		return IP_MULTICAST_IF;
915 	case LINUX_IP_ADD_MEMBERSHIP:
916 		return IP_ADD_MEMBERSHIP;
917 	case LINUX_IP_DROP_MEMBERSHIP:
918 		return IP_DROP_MEMBERSHIP;
919 	default:
920 		return -1;
921 	}
922 }
923 
924 /*
925  * Convert Linux IPV6 level socket option number to NetBSD values.
926  */
927 #ifdef INET6
928 int
929 linux_to_bsd_ipv6_sockopt(int lopt)
930 {
931 
932 	switch (lopt) {
933 	case LINUX_IPV6_V6ONLY:
934 		return IPV6_V6ONLY;
935 	default:
936 		return -1;
937 	}
938 }
939 #endif
940 
941 /*
942  * Convert Linux TCP level socket option number to NetBSD values.
943  */
944 int
945 linux_to_bsd_tcp_sockopt(int lopt)
946 {
947 
948 	switch (lopt) {
949 	case LINUX_TCP_NODELAY:
950 		return TCP_NODELAY;
951 	case LINUX_TCP_MAXSEG:
952 		return TCP_MAXSEG;
953 	default:
954 		return -1;
955 	}
956 }
957 
958 /*
959  * Convert Linux UDP level socket option number to NetBSD values.
960  */
961 int
962 linux_to_bsd_udp_sockopt(int lopt)
963 {
964 
965 	switch (lopt) {
966 	default:
967 		return -1;
968 	}
969 }
970 
971 /*
972  * Another reasonably straightforward function: setsockopt(2).
973  * The level and option numbers are converted; the values passed
974  * are not (yet) converted, the ones currently implemented don't
975  * need conversion, as they are the same on both systems.
976  */
977 int
978 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
979 {
980 	/* {
981 		syscallarg(int) s;
982 		syscallarg(int) level;
983 		syscallarg(int) optname;
984 		syscallarg(void *) optval;
985 		syscallarg(int) optlen;
986 	} */
987 	struct sys_setsockopt_args bsa;
988 	int name;
989 
990 	SCARG(&bsa, s) = SCARG(uap, s);
991 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
992 	SCARG(&bsa, val) = SCARG(uap, optval);
993 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
994 
995 	/*
996 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
997 	 * and returns EOPNOTSUPP for other levels
998 	 */
999 	if (SCARG(&bsa, level) != SOL_SOCKET) {
1000 		struct socket *so;
1001 		int error, family;
1002 
1003 		/* fd_getsock() will use the descriptor for us */
1004 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
1005 		    	return error;
1006 		family = so->so_proto->pr_domain->dom_family;
1007 		fd_putfile(SCARG(&bsa, s));
1008 
1009 		if (family == AF_LOCAL)
1010 			return EOPNOTSUPP;
1011 	}
1012 
1013 	switch (SCARG(&bsa, level)) {
1014 	case SOL_SOCKET:
1015 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1016 		break;
1017 	case IPPROTO_IP:
1018 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1019 		break;
1020 #ifdef INET6
1021 	case IPPROTO_IPV6:
1022 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1023 		break;
1024 #endif
1025 	case IPPROTO_TCP:
1026 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1027 		break;
1028 	case IPPROTO_UDP:
1029 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1030 		break;
1031 	default:
1032 		return EINVAL;
1033 	}
1034 
1035 	if (name == -1)
1036 		return EINVAL;
1037 	SCARG(&bsa, name) = name;
1038 
1039 	return sys_setsockopt(l, &bsa, retval);
1040 }
1041 
1042 /*
1043  * getsockopt(2) is very much the same as setsockopt(2) (see above)
1044  */
1045 int
1046 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
1047 {
1048 	/* {
1049 		syscallarg(int) s;
1050 		syscallarg(int) level;
1051 		syscallarg(int) optname;
1052 		syscallarg(void *) optval;
1053 		syscallarg(int *) optlen;
1054 	} */
1055 	struct sys_getsockopt_args bga;
1056 	int name;
1057 
1058 	SCARG(&bga, s) = SCARG(uap, s);
1059 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
1060 	SCARG(&bga, val) = SCARG(uap, optval);
1061 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
1062 
1063 	switch (SCARG(&bga, level)) {
1064 	case SOL_SOCKET:
1065 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1066 		break;
1067 	case IPPROTO_IP:
1068 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1069 		break;
1070 #ifdef INET6
1071 	case IPPROTO_IPV6:
1072 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1073 		break;
1074 #endif
1075 	case IPPROTO_TCP:
1076 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1077 		break;
1078 	case IPPROTO_UDP:
1079 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1080 		break;
1081 	default:
1082 		return EINVAL;
1083 	}
1084 
1085 	if (name == -1)
1086 		return EINVAL;
1087 	SCARG(&bga, name) = name;
1088 
1089 	return sys_getsockopt(l, &bga, retval);
1090 }
1091 
1092 int
1093 linux_getifname(struct lwp *l, register_t *retval, void *data)
1094 {
1095 	struct ifnet *ifp;
1096 	struct linux_ifreq ifr;
1097 	int error;
1098 	int s;
1099 
1100 	error = copyin(data, &ifr, sizeof(ifr));
1101 	if (error)
1102 		return error;
1103 
1104 	s = pserialize_read_enter();
1105 	ifp = if_byindex(ifr.ifr_ifru.ifru_ifindex);
1106 	if (ifp == NULL) {
1107 		pserialize_read_exit(s);
1108 		return ENODEV;
1109 	}
1110 
1111 	strncpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name));
1112 	pserialize_read_exit(s);
1113 
1114 	return copyout(&ifr, data, sizeof(ifr));
1115 }
1116 
1117 int
1118 linux_getifconf(struct lwp *l, register_t *retval, void *data)
1119 {
1120 	struct linux_ifreq ifr, *ifrp = NULL;
1121 	struct linux_ifconf ifc;
1122 	struct ifnet *ifp;
1123 	struct sockaddr *sa;
1124 	struct osockaddr *osa;
1125 	int space = 0, error;
1126 	const int sz = (int)sizeof(ifr);
1127 	bool docopy;
1128 	int s;
1129 	int bound;
1130 	struct psref psref;
1131 
1132 	error = copyin(data, &ifc, sizeof(ifc));
1133 	if (error)
1134 		return error;
1135 
1136 	docopy = ifc.ifc_req != NULL;
1137 	if (docopy) {
1138 		space = ifc.ifc_len;
1139 		ifrp = ifc.ifc_req;
1140 	}
1141 
1142 	bound = curlwp_bind();
1143 	s = pserialize_read_enter();
1144 	IFNET_READER_FOREACH(ifp) {
1145 		struct ifaddr *ifa;
1146 		if_acquire(ifp, &psref);
1147 
1148 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1149 		    sizeof(ifr.ifr_name));
1150 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') {
1151 			error = ENAMETOOLONG;
1152 			goto release_exit;
1153 		}
1154 
1155 		IFADDR_READER_FOREACH(ifa, ifp) {
1156 			struct psref psref_ifa;
1157 			ifa_acquire(ifa, &psref_ifa);
1158 			pserialize_read_exit(s);
1159 
1160 			sa = ifa->ifa_addr;
1161 			if (sa->sa_family != AF_INET ||
1162 			    sa->sa_len > sizeof(*osa))
1163 				goto next;
1164 			memcpy(&ifr.ifr_addr, sa, sa->sa_len);
1165 			osa = (struct osockaddr *)&ifr.ifr_addr;
1166 			osa->sa_family = sa->sa_family;
1167 			if (space >= sz) {
1168 				error = copyout(&ifr, ifrp, sz);
1169 				if (error != 0) {
1170 					s = pserialize_read_enter();
1171 					ifa_release(ifa, &psref_ifa);
1172 					goto release_exit;
1173 				}
1174 				ifrp++;
1175 			}
1176 			space -= sz;
1177 		next:
1178 			s = pserialize_read_enter();
1179 			ifa_release(ifa, &psref_ifa);
1180 		}
1181 
1182 		if_release(ifp, &psref);
1183 	}
1184 	pserialize_read_exit(s);
1185 	curlwp_bindx(bound);
1186 
1187 	if (docopy)
1188 		ifc.ifc_len -= space;
1189 	else
1190 		ifc.ifc_len = -space;
1191 
1192 	return copyout(&ifc, data, sizeof(ifc));
1193 
1194 release_exit:
1195 	pserialize_read_exit(s);
1196 	if_release(ifp, &psref);
1197 	curlwp_bindx(bound);
1198 	return error;
1199 }
1200 
1201 int
1202 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1203     void *data)
1204 {
1205 	/* Not the full structure, just enough to map what we do here */
1206 	struct linux_ifreq lreq;
1207 	file_t *fp;
1208 	struct ifaddr *ifa;
1209 	struct ifnet *ifp;
1210 	struct sockaddr_dl *sadl;
1211 	int error, found;
1212 	int index, ifnum;
1213 	int s;
1214 
1215 	/*
1216 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1217 	 * SIOCGIFCONF, because the user buffer is not of the right
1218 	 * type to take those results.  We can't use kernel buffers to
1219 	 * receive the results, as the implementation of sys_ioctl()
1220 	 * and ifconf() [which implements SIOCGIFCONF] use
1221 	 * copyin()/copyout() which will fail on kernel addresses.
1222 	 *
1223 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1224 	 */
1225 
1226 	if ((fp = fd_getfile(fd)) == NULL)
1227 		return (EBADF);
1228 
1229 	KERNEL_LOCK(1, NULL);
1230 
1231 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1232 		error = EBADF;
1233 		goto out;
1234 	}
1235 
1236 	error = copyin(data, &lreq, sizeof(lreq));
1237 	if (error)
1238 		goto out;
1239 	lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0';		/* just in case */
1240 
1241 	/*
1242 	 * Try real interface name first, then fake "ethX"
1243 	 */
1244 	found = 0;
1245 	s = pserialize_read_enter();
1246 	IFNET_READER_FOREACH(ifp) {
1247 		if (found)
1248 			break;
1249 		if (strcmp(lreq.ifr_name, ifp->if_xname))
1250 			/* not this interface */
1251 			continue;
1252 
1253 		found=1;
1254 		if (IFADDR_READER_EMPTY(ifp)) {
1255 			pserialize_read_exit(s);
1256 			error = ENODEV;
1257 			goto out;
1258 		}
1259 		IFADDR_READER_FOREACH(ifa, ifp) {
1260 			sadl = satosdl(ifa->ifa_addr);
1261 			/* only return ethernet addresses */
1262 			/* XXX what about FDDI, etc. ? */
1263 			if (sadl->sdl_family != AF_LINK ||
1264 			    sadl->sdl_type != IFT_ETHER)
1265 				continue;
1266 			memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl),
1267 			       MIN(sadl->sdl_alen,
1268 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1269 			lreq.ifr_hwaddr.sa_family =
1270 				sadl->sdl_family;
1271 			pserialize_read_exit(s);
1272 
1273 			error = copyout(&lreq, data, sizeof(lreq));
1274 			goto out;
1275 		}
1276 	}
1277 	pserialize_read_exit(s);
1278 
1279 	if (strncmp(lreq.ifr_name, "eth", 3) != 0) {
1280 		/* unknown interface, not even an "eth*" name */
1281 		error = ENODEV;
1282 		goto out;
1283 	}
1284 
1285 	for (ifnum = 0, index = 3;
1286 	     index < LINUX_IFNAMSIZ && lreq.ifr_name[index] != '\0';
1287 	     index++) {
1288 		ifnum *= 10;
1289 		ifnum += lreq.ifr_name[index] - '0';
1290 	}
1291 
1292 	error = EINVAL;			/* in case we don't find one */
1293 	s = pserialize_read_enter();
1294 	IFNET_READER_FOREACH(ifp) {
1295 		memcpy(lreq.ifr_name, ifp->if_xname,
1296 		       MIN(LINUX_IFNAMSIZ, IFNAMSIZ));
1297 		IFADDR_READER_FOREACH(ifa, ifp) {
1298 			sadl = satosdl(ifa->ifa_addr);
1299 			/* only return ethernet addresses */
1300 			/* XXX what about FDDI, etc. ? */
1301 			if (sadl->sdl_family != AF_LINK ||
1302 			    sadl->sdl_type != IFT_ETHER)
1303 				continue;
1304 			if (ifnum--)
1305 				/* not the reqested iface */
1306 				continue;
1307 			memcpy(&lreq.ifr_hwaddr.sa_data,
1308 			       CLLADDR(sadl),
1309 			       MIN(sadl->sdl_alen,
1310 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1311 			lreq.ifr_hwaddr.sa_family =
1312 				sadl->sdl_family;
1313 			pserialize_read_exit(s);
1314 
1315 			error = copyout(&lreq, data, sizeof(lreq));
1316 			goto out;
1317 		}
1318 	}
1319 	pserialize_read_exit(s);
1320 
1321 out:
1322 	KERNEL_UNLOCK_ONE(NULL);
1323 	fd_putfile(fd);
1324 	return error;
1325 }
1326 
1327 int
1328 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1329 {
1330 	/* {
1331 		syscallarg(int) fd;
1332 		syscallarg(u_long) com;
1333 		syscallarg(void *) data;
1334 	} */
1335 	u_long com;
1336 	int error = 0, isdev = 0, dosys = 1;
1337 	struct sys_ioctl_args ia;
1338 	file_t *fp;
1339 	struct vnode *vp;
1340 	int (*ioctlf)(file_t *, u_long, void *);
1341 	struct ioctl_pt pt;
1342 
1343 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1344 		return (EBADF);
1345 
1346 	if (fp->f_type == DTYPE_VNODE) {
1347 		vp = (struct vnode *)fp->f_data;
1348 		isdev = vp->v_type == VCHR;
1349 	}
1350 
1351 	/*
1352 	 * Don't try to interpret socket ioctl calls that are done
1353 	 * on a device filedescriptor, just pass them through, to
1354 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1355 	 * device will only handle these if it's prepared to do
1356 	 * so, to avoid unexpected things from happening.
1357 	 */
1358 	if (isdev) {
1359 		dosys = 0;
1360 		ioctlf = fp->f_ops->fo_ioctl;
1361 		pt.com = SCARG(uap, com);
1362 		pt.data = SCARG(uap, data);
1363 		error = ioctlf(fp, PTIOCLINUX, &pt);
1364 		/*
1365 		 * XXX hack: if the function returns EJUSTRETURN,
1366 		 * it has stuffed a sysctl return value in pt.data.
1367 		 */
1368 		if (error == EJUSTRETURN) {
1369 			retval[0] = (register_t)pt.data;
1370 			error = 0;
1371 		}
1372 		goto out;
1373 	}
1374 
1375 	com = SCARG(uap, com);
1376 	retval[0] = 0;
1377 
1378 	switch (com) {
1379 	case LINUX_SIOCGIFNAME:
1380 		error = linux_getifname(l, retval, SCARG(uap, data));
1381 		dosys = 0;
1382 		break;
1383 	case LINUX_SIOCGIFCONF:
1384 		error = linux_getifconf(l, retval, SCARG(uap, data));
1385 		dosys = 0;
1386 		break;
1387 	case LINUX_SIOCGIFFLAGS:
1388 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1389 		break;
1390 	case LINUX_SIOCSIFFLAGS:
1391 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1392 		break;
1393 	case LINUX_SIOCGIFADDR:
1394 		SCARG(&ia, com) = OOSIOCGIFADDR;
1395 		break;
1396 	case LINUX_SIOCGIFDSTADDR:
1397 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1398 		break;
1399 	case LINUX_SIOCGIFBRDADDR:
1400 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1401 		break;
1402 	case LINUX_SIOCGIFNETMASK:
1403 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1404 		break;
1405 	case LINUX_SIOCGIFMTU:
1406 		SCARG(&ia, com) = OSIOCGIFMTU;
1407 		break;
1408 	case LINUX_SIOCADDMULTI:
1409 		SCARG(&ia, com) = OSIOCADDMULTI;
1410 		break;
1411 	case LINUX_SIOCDELMULTI:
1412 		SCARG(&ia, com) = OSIOCDELMULTI;
1413 		break;
1414 	case LINUX_SIOCGIFHWADDR:
1415 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1416 		    SCARG(uap, data));
1417 		dosys = 0;
1418 		break;
1419 	default:
1420 		error = EINVAL;
1421 	}
1422 
1423  out:
1424  	fd_putfile(SCARG(uap, fd));
1425 
1426 	if (error ==0 && dosys) {
1427 		SCARG(&ia, fd) = SCARG(uap, fd);
1428 		SCARG(&ia, data) = SCARG(uap, data);
1429 		error = sys_ioctl(curlwp, &ia, retval);
1430 	}
1431 
1432 	return error;
1433 }
1434 
1435 int
1436 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1437 {
1438 	/* {
1439 		syscallarg(int) s;
1440 		syscallarg(const struct sockaddr *) name;
1441 		syscallarg(int) namelen;
1442 	} */
1443 	int		error;
1444 	struct sockaddr_big sb;
1445 
1446 	error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name),
1447 	    SCARG(uap, namelen));
1448 	if (error)
1449 		return (error);
1450 
1451 	error = do_sys_connect(l, SCARG(uap, s), (struct sockaddr *)&sb);
1452 
1453 	if (error == EISCONN) {
1454 		struct socket *so;
1455 		int state, prflags;
1456 
1457 		/* fd_getsock() will use the descriptor for us */
1458 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1459 		    	return EISCONN;
1460 
1461 		solock(so);
1462 		state = so->so_state;
1463 		prflags = so->so_proto->pr_flags;
1464 		sounlock(so);
1465 		fd_putfile(SCARG(uap, s));
1466 		/*
1467 		 * We should only let this call succeed once per
1468 		 * non-blocking connect; however we don't have
1469 		 * a convenient place to keep that state..
1470 		 */
1471 		if ((state & (SS_ISCONNECTED|SS_NBIO)) ==
1472 		    (SS_ISCONNECTED|SS_NBIO) &&
1473 		    (prflags & PR_CONNREQUIRED))
1474 			return 0;
1475 	}
1476 
1477 	return (error);
1478 }
1479 
1480 int
1481 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1482 {
1483 	/* {
1484 		syscallarg(int) s;
1485 		syscallarg(const struct osockaddr *) name;
1486 		syscallarg(int) namelen;
1487 	} */
1488 	int		error;
1489 	struct sockaddr_big sb;
1490 
1491 	error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name),
1492 	    SCARG(uap, namelen));
1493 	if (error)
1494 		return (error);
1495 
1496 	return do_sys_bind(l, SCARG(uap, s), (struct sockaddr *)&sb);
1497 }
1498 
1499 int
1500 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1501 {
1502 	/* {
1503 		syscallarg(int) fdes;
1504 		syscallarg(void *) asa;
1505 		syscallarg(int *) alen;
1506 	} */
1507 	int error;
1508 
1509 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1510 		return (error);
1511 
1512 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1513 		return (error);
1514 
1515 	return (0);
1516 }
1517 
1518 int
1519 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1520 {
1521 	/* {
1522 		syscallarg(int) fdes;
1523 		syscallarg(void *) asa;
1524 		syscallarg(int *) alen;
1525 	} */
1526 	int error;
1527 
1528 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1529 		return (error);
1530 
1531 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1532 		return (error);
1533 
1534 	return (0);
1535 }
1536 
1537 /*
1538  * Copy the osockaddr structure pointed to by name to sb, adjust
1539  * family and convert to sockaddr.
1540  */
1541 static int
1542 linux_get_sa(struct lwp *l, int s, struct sockaddr_big *sb,
1543     const struct osockaddr *name, socklen_t namelen)
1544 {
1545 	int error, bdom;
1546 
1547 	if (namelen > UCHAR_MAX ||
1548 	    namelen <= offsetof(struct sockaddr_big, sb_data))
1549 		return EINVAL;
1550 
1551 	error = copyin(name, sb, namelen);
1552 	if (error)
1553 		return error;
1554 
1555 	bdom = linux_to_bsd_domain(sb->sb_family);
1556 	if (bdom == -1)
1557 		return EINVAL;
1558 
1559 	/*
1560 	 * If the family is unspecified, use address family of the socket.
1561 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1562 	 */
1563 	if (bdom == AF_UNSPEC) {
1564 		struct socket *so;
1565 
1566 		/* fd_getsock() will use the descriptor for us */
1567 		if ((error = fd_getsock(s, &so)) != 0)
1568 			return error;
1569 
1570 		bdom = so->so_proto->pr_domain->dom_family;
1571 		fd_putfile(s);
1572 	}
1573 
1574 	/*
1575 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1576 	 * which lacks the scope id compared with RFC2553 one. If we detect
1577 	 * the situation, reject the address and write a message to system log.
1578 	 *
1579 	 * Still accept addresses for which the scope id is not used.
1580 	 */
1581 	if (bdom == AF_INET6 &&
1582 	    namelen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
1583 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sb;
1584 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1585 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1586 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1587 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1588 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1589 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1590 			struct proc *p = l->l_proc;
1591 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1592 
1593 			log(LOG_DEBUG,
1594 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1595 			    "sockaddr_in6 rejected",
1596 			    p->p_pid, p->p_comm, uid);
1597 			return EINVAL;
1598 		}
1599 		namelen = sizeof(struct sockaddr_in6);
1600 		sin6->sin6_scope_id = 0;
1601 	}
1602 
1603 	if (bdom == AF_INET)
1604 		namelen = sizeof(struct sockaddr_in);
1605 
1606 	sb->sb_family = bdom;
1607 	sb->sb_len = namelen;
1608 	ktrkuser("mbsoname", sb, namelen);
1609 	return 0;
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 
1712 int
1713 linux_sys_accept4(struct lwp *l, const struct linux_sys_accept4_args *uap, register_t *retval)
1714 {
1715 	/* {
1716 		syscallarg(int) s;
1717 		syscallarg(struct osockaddr *) name;
1718 		syscallarg(int *) anamelen;
1719 		syscallarg(int) flags;
1720 	} */
1721 	int error, flags;
1722 	struct sockaddr_big name;
1723 
1724 	if ((flags = linux_to_bsd_type(SCARG(uap, flags))) == -1)
1725 		return EINVAL;
1726 
1727 	name.sb_len = UCHAR_MAX;
1728 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
1729 	    retval, NULL, flags, 0);
1730 	if (error != 0)
1731 		return error;
1732 
1733 	error = copyout_sockname_sb((struct sockaddr *)SCARG(uap, name),
1734 	    SCARG(uap, anamelen), MSG_LENUSRSPACE, &name);
1735 	if (error != 0) {
1736 		int fd = (int)*retval;
1737 		if (fd_getfile(fd) != NULL)
1738 			(void)fd_close(fd);
1739 		return error;
1740 	}
1741 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1742 		return error;
1743 
1744 	return 0;
1745 }
1746 
1747 int
1748 linux_sys_sendmmsg(struct lwp *l, const struct linux_sys_sendmmsg_args *uap,
1749     register_t *retval)
1750 {
1751 	/* {
1752 		syscallarg(int) s;
1753 		syscallarg(struct linux_mmsghdr *) msgvec;
1754 		syscallarg(unsigned int) vlen;
1755 		syscallarg(unsigned int) flags;
1756 	} */
1757 	struct linux_mmsghdr lmsg;
1758 	struct mmsghdr bmsg;
1759 	struct socket *so;
1760 	file_t *fp;
1761 	struct msghdr *msg = &bmsg.msg_hdr;
1762 	int error, s;
1763 	unsigned int vlen, flags, dg;
1764 
1765 	if ((flags = linux_to_bsd_msg_flags(SCARG(uap, flags))) == -1)
1766 		return EINVAL;
1767 
1768 	flags = (flags & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1769 
1770 	s = SCARG(uap, s);
1771 	if ((error = fd_getsock1(s, &so, &fp)) != 0)
1772 		return error;
1773 
1774 	vlen = SCARG(uap, vlen);
1775 	if (vlen > 1024)
1776 		vlen = 1024;
1777 
1778 	for (dg = 0; dg < vlen;) {
1779 		error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg));
1780 		if (error)
1781 			break;
1782 		linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr);
1783 
1784 		msg->msg_flags = flags;
1785 
1786 		error = do_sys_sendmsg_so(l, s, so, fp, msg, flags,
1787 		    &msg, sizeof(msg), retval);
1788 		if (error)
1789 			break;
1790 
1791 		ktrkuser("msghdr", msg, sizeof *msg);
1792 		lmsg.msg_len = *retval;
1793 		error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg));
1794 		if (error)
1795 			break;
1796 		dg++;
1797 
1798 	}
1799 
1800 	*retval = dg;
1801 	if (error)
1802 		so->so_error = error;
1803 
1804 	fd_putfile(s);
1805 
1806 	/*
1807 	 * If we succeeded at least once, return 0, hopefully so->so_error
1808 	 * will catch it next time.
1809 	 */
1810 	if (dg)
1811 		return 0;
1812 	return error;
1813 }
1814 
1815 int
1816 linux_sys_recvmmsg(struct lwp *l, const struct linux_sys_recvmmsg_args *uap,
1817     register_t *retval)
1818 {
1819 	/* {
1820 		syscallarg(int) s;
1821 		syscallarg(struct linux_mmsghdr *) msgvec;
1822 		syscallarg(unsigned int) vlen;
1823 		syscallarg(unsigned int) flags;
1824 		syscallarg(struct linux_timespec *) timeout;
1825 	} */
1826 	struct linux_mmsghdr lmsg;
1827 	struct mmsghdr bmsg;
1828 	struct socket *so;
1829 	struct msghdr *msg = &bmsg.msg_hdr;
1830 	int error, s;
1831 	struct mbuf *from, *control;
1832 	struct timespec ts, now;
1833 	struct linux_timespec lts;
1834 	unsigned int vlen, flags, dg;
1835 
1836 	if (SCARG(uap, timeout)) {
1837 		error = copyin(SCARG(uap, timeout), &lts, sizeof(lts));
1838 			return error;
1839 		ts.tv_sec = lts.tv_sec;
1840 		ts.tv_nsec = lts.tv_nsec;
1841 		getnanotime(&now);
1842 		timespecadd(&now, &ts, &ts);
1843 	}
1844 
1845 	s = SCARG(uap, s);
1846 	if ((error = fd_getsock(s, &so)) != 0)
1847 		return error;
1848 
1849 	vlen = SCARG(uap, vlen);
1850 	if (vlen > 1024)
1851 		vlen = 1024;
1852 
1853 	from = NULL;
1854 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1855 
1856 	for (dg = 0; dg < vlen;) {
1857 		error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg));
1858 		if (error)
1859 			break;
1860 		linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr);
1861 		msg->msg_flags = flags & ~MSG_WAITFORONE;
1862 
1863 		if (from != NULL) {
1864 			m_free(from);
1865 			from = NULL;
1866 		}
1867 
1868 		error = do_sys_recvmsg_so(l, s, so, msg, NULL, 0, &from,
1869 		    msg->msg_control != NULL ? &control : NULL, retval);
1870 		if (error) {
1871 			if (error == EAGAIN && dg > 0)
1872 				error = 0;
1873 			break;
1874 		}
1875 
1876 		if (msg->msg_control != NULL)
1877 			error = linux_copyout_msg_control(l, msg, control);
1878 		if (error)
1879 			break;
1880 
1881 		if (from != NULL) {
1882 			mtod(from, struct osockaddr *)->sa_family =
1883 			    bsd_to_linux_domain(mtod(from,
1884 			    struct sockaddr *)->sa_family);
1885 			error = copyout_sockname(msg->msg_name,
1886 			    &msg->msg_namelen, 0, from);
1887 			if (error)
1888 				break;
1889 		}
1890 
1891 
1892 		lmsg.msg_len = *retval;
1893 		ktrkuser("msghdr", msg, sizeof(*msg));
1894 		bsd_to_linux_msghdr(msg, &lmsg.msg_hdr);
1895 		error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg));
1896 		if (error)
1897 			break;
1898 
1899 		dg++;
1900 		if (msg->msg_flags & MSG_OOB)
1901 			break;
1902 
1903 		if (SCARG(uap, timeout)) {
1904 			getnanotime(&now);
1905 			timespecsub(&now, &ts, &now);
1906 			if (now.tv_sec > 0)
1907 				break;
1908 		}
1909 
1910 		if (flags & MSG_WAITFORONE)
1911 			flags |= MSG_DONTWAIT;
1912 
1913 	}
1914 
1915 	if (from != NULL)
1916 		m_free(from);
1917 
1918 	*retval = dg;
1919 	if (error)
1920 		so->so_error = error;
1921 
1922 	fd_putfile(s);
1923 
1924 	/*
1925 	 * If we succeeded at least once, return 0, hopefully so->so_error
1926 	 * will catch it next time.
1927 	 */
1928 	if (dg)
1929 		return 0;
1930 
1931 	return error;
1932 }
1933