xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: linux_socket.c,v 1.142 2018/05/10 01:32:24 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 1.142 2018/05/10 01:32:24 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 	struct mbuf *m;
405 	int bflags;
406 	int error;
407 
408 	/* Translate message flags.  */
409 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
410 	if (bflags < 0)
411 		/* Some supported flag */
412 		return EINVAL;
413 
414 	msg.msg_flags = 0;
415 	msg.msg_name = NULL;
416 	msg.msg_control = NULL;
417 
418 	if (SCARG(uap, tolen)) {
419 		/* Read in and convert the sockaddr */
420 		error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
421 		    SCARG(uap, tolen));
422 		if (error)
423 			return error;
424 		error = sockargs(&m, &nam, nam.sb_len, UIO_SYSSPACE, MT_SONAME);
425 		if (error)
426 			return error;
427 		msg.msg_flags |= MSG_NAMEMBUF;
428 		msg.msg_name = m;
429 		msg.msg_namelen = nam.sb_len;
430 	}
431 
432 	msg.msg_iov = &aiov;
433 	msg.msg_iovlen = 1;
434 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
435 	aiov.iov_len = SCARG(uap, len);
436 
437 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
438 }
439 
440 static void
441 linux_to_bsd_msghdr(const struct linux_msghdr *lmsg, struct msghdr *bmsg)
442 {
443 	bmsg->msg_name = lmsg->msg_name;
444 	bmsg->msg_namelen = lmsg->msg_namelen;
445 	bmsg->msg_iov = lmsg->msg_iov;
446 	bmsg->msg_iovlen = lmsg->msg_iovlen;
447 	bmsg->msg_control = lmsg->msg_control;
448 	bmsg->msg_controllen = lmsg->msg_controllen;
449 	bmsg->msg_flags = lmsg->msg_flags;
450 }
451 
452 static void
453 bsd_to_linux_msghdr(const struct msghdr *bmsg, struct linux_msghdr *lmsg)
454 {
455 	lmsg->msg_name = bmsg->msg_name;
456 	lmsg->msg_namelen = bmsg->msg_namelen;
457 	lmsg->msg_iov = bmsg->msg_iov;
458 	lmsg->msg_iovlen = bmsg->msg_iovlen;
459 	lmsg->msg_control = bmsg->msg_control;
460 	lmsg->msg_controllen = bmsg->msg_controllen;
461 	lmsg->msg_flags = bmsg->msg_flags;
462 }
463 
464 int
465 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
466 {
467 	/* {
468 		syscallarg(int) s;
469 		syscallarg(struct linux_msghdr *) msg;
470 		syscallarg(u_int) flags;
471 	} */
472 	struct msghdr	msg;
473 	struct linux_msghdr lmsg;
474 	int		error;
475 	int		bflags;
476 	struct sockaddr_big nam;
477 	u_int8_t	*control;
478 	struct mbuf     *ctl_mbuf = NULL;
479 
480 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
481 	if (error)
482 		return error;
483 	linux_to_bsd_msghdr(&lmsg, &msg);
484 
485 	msg.msg_flags = MSG_IOVUSRSPACE;
486 
487 	/*
488 	 * Translate message flags.
489 	 */
490 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
491 	if (bflags < 0)
492 		/* Some supported flag */
493 		return EINVAL;
494 
495 	if (lmsg.msg_name) {
496 		/* Read in and convert the sockaddr */
497 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
498 		    msg.msg_namelen);
499 		if (error)
500 			return (error);
501 		msg.msg_name = &nam;
502 	}
503 
504 	/*
505 	 * Handle cmsg if there is any.
506 	 */
507 	if (LINUX_CMSG_FIRSTHDR(&lmsg)) {
508 		struct linux_cmsghdr l_cmsg, *l_cc;
509 		struct cmsghdr *cmsg;
510 		ssize_t resid = msg.msg_controllen;
511 		size_t clen, cidx = 0, cspace;
512 
513 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
514 		clen = MLEN;
515 		control = mtod(ctl_mbuf, void *);
516 
517 		l_cc = LINUX_CMSG_FIRSTHDR(&lmsg);
518 		do {
519 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
520 			if (error)
521 				goto done;
522 
523 			/*
524 			 * Sanity check the control message length.
525 			 */
526 			if (l_cmsg.cmsg_len > resid
527 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
528 				error = EINVAL;
529 				goto done;
530 			}
531 
532 			/*
533 			 * Refuse unsupported control messages, and
534 			 * translate fields as appropriate.
535 			 */
536 			switch (l_cmsg.cmsg_level) {
537 			case LINUX_SOL_SOCKET:
538 				/* It only differs on some archs */
539 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
540 					l_cmsg.cmsg_level = SOL_SOCKET;
541 
542 				switch(l_cmsg.cmsg_type) {
543 				case LINUX_SCM_RIGHTS:
544 					/* Linux SCM_RIGHTS is same as NetBSD */
545 					break;
546 
547 				case LINUX_SCM_CREDENTIALS:
548 					/* no native equivalent, just drop it */
549 					m_free(ctl_mbuf);
550 					ctl_mbuf = NULL;
551 					msg.msg_control = NULL;
552 					msg.msg_controllen = 0;
553 					goto skipcmsg;
554 
555 				default:
556 					/* other types not supported */
557 					error = EINVAL;
558 					goto done;
559 				}
560 				break;
561 			default:
562 				/* pray and leave intact */
563 				break;
564 			}
565 
566 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
567 
568 			/* Check the buffer is big enough */
569 			if (__predict_false(cidx + cspace > clen)) {
570 				u_int8_t *nc;
571 
572 				clen = cidx + cspace;
573 				if (clen >= PAGE_SIZE) {
574 					error = EINVAL;
575 					goto done;
576 				}
577 				nc = realloc(clen <= MLEN ? NULL : control,
578 						clen, M_TEMP, M_WAITOK);
579 				if (!nc) {
580 					error = ENOMEM;
581 					goto done;
582 				}
583 				if (cidx <= MLEN)
584 					/* Old buffer was in mbuf... */
585 					memcpy(nc, control, cidx);
586 				control = nc;
587 			}
588 
589 			/* Copy header */
590 			cmsg = (void *)&control[cidx];
591 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
592 			cmsg->cmsg_level = l_cmsg.cmsg_level;
593 			cmsg->cmsg_type = l_cmsg.cmsg_type;
594 
595 			/* Zero area between header and data */
596 			memset(cmsg + 1, 0,
597 				CMSG_ALIGN(sizeof(*cmsg)) - sizeof(*cmsg));
598 
599 			/* Copyin the data */
600 			error = copyin(LINUX_CMSG_DATA(l_cc),
601 				CMSG_DATA(cmsg),
602 				l_cmsg.cmsg_len - sizeof(l_cmsg));
603 			if (error)
604 				goto done;
605 
606 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
607 			cidx += cspace;
608 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
609 
610 		/* If we allocated a buffer, attach to mbuf */
611 		if (cidx > MLEN) {
612 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
613 			ctl_mbuf->m_flags |= M_EXT_RW;
614 		}
615 		control = NULL;
616 		ctl_mbuf->m_len = cidx;
617 
618 		msg.msg_control = ctl_mbuf;
619 		msg.msg_flags |= MSG_CONTROLMBUF;
620 
621 		ktrkuser("mbcontrol", mtod(ctl_mbuf, void *),
622 		    msg.msg_controllen);
623 	}
624 
625 skipcmsg:
626 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
627 	/* Freed internally */
628 	ctl_mbuf = NULL;
629 
630 done:
631 	if (ctl_mbuf != NULL) {
632 		if (control != NULL && control != mtod(ctl_mbuf, void *))
633 			free(control, M_MBUF);
634 		m_free(ctl_mbuf);
635 	}
636 	return (error);
637 }
638 
639 int
640 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
641 {
642 	/* {
643 		syscallarg(int) s;
644 		syscallarg(void *) buf;
645 		syscallarg(int) len;
646 		syscallarg(int) flags;
647 		syscallarg(struct osockaddr *) from;
648 		syscallarg(int *) fromlenaddr;
649 	} */
650 	int		error;
651 	struct sys_recvfrom_args bra;
652 
653 	SCARG(&bra, s) = SCARG(uap, s);
654 	SCARG(&bra, buf) = SCARG(uap, buf);
655 	SCARG(&bra, len) = SCARG(uap, len);
656 	SCARG(&bra, flags) = SCARG(uap, flags);
657 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
658 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
659 
660 	if ((error = sys_recvfrom(l, &bra, retval)))
661 		return (error);
662 
663 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
664 		return (error);
665 
666 	return (0);
667 }
668 
669 static int
670 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
671 {
672 	int dlen, error = 0;
673 	struct cmsghdr *cmsg;
674 	struct linux_cmsghdr linux_cmsg;
675 	struct mbuf *m;
676 	char *q, *q_end;
677 
678 	if (mp->msg_controllen <= 0 || control == 0) {
679 		mp->msg_controllen = 0;
680 		free_control_mbuf(l, control, control);
681 		return 0;
682 	}
683 
684 	ktrkuser("msgcontrol", mtod(control, void *), mp->msg_controllen);
685 
686 	q = (char *)mp->msg_control;
687 	q_end = q + mp->msg_controllen;
688 
689 	for (m = control; m != NULL; ) {
690 		cmsg = mtod(m, struct cmsghdr *);
691 
692 		/*
693 		 * Fixup cmsg. We handle two things:
694 		 * 0. different sizeof cmsg_len.
695 		 * 1. different values for level/type on some archs
696 		 * 2. different alignment of CMSG_DATA on some archs
697 		 */
698 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
699 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
700 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
701 
702 		dlen = q_end - q;
703 		if (linux_cmsg.cmsg_len > dlen) {
704 			/* Not enough room for the parameter */
705 			dlen -= sizeof linux_cmsg;
706 			if (dlen <= 0)
707 				/* Discard if header wont fit */
708 				break;
709 			mp->msg_flags |= MSG_CTRUNC;
710 			if (linux_cmsg.cmsg_level == SOL_SOCKET
711 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
712 				/* Do not truncate me ... */
713 				break;
714 		} else
715 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
716 
717 		switch (linux_cmsg.cmsg_level) {
718 		case SOL_SOCKET:
719 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
720 			switch (linux_cmsg.cmsg_type) {
721 			case SCM_RIGHTS:
722 				/* Linux SCM_RIGHTS is same as NetBSD */
723 				break;
724 
725 			default:
726 				/* other types not supported */
727 				error = EINVAL;
728 				goto done;
729 			}
730 			/* machine dependent ! */
731 			break;
732 		default:
733 			/* pray and leave intact */
734 			break;
735 		}
736 
737 		/* There can be padding between the header and data... */
738 		error = copyout(&linux_cmsg, q, sizeof linux_cmsg);
739 		if (error != 0) {
740 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
741 			    dlen);
742 		}
743 		if (error != 0) {
744 			/* We must free all the SCM_RIGHTS */
745 			m = control;
746 			break;
747 		}
748 		m = m->m_next;
749 		if (m == NULL || q + LINUX_CMSG_SPACE(dlen) > q_end) {
750 			q += LINUX_CMSG_LEN(dlen);
751 			break;
752 		}
753 		q += LINUX_CMSG_SPACE(dlen);
754 	}
755 
756   done:
757 	free_control_mbuf(l, control, m);
758 
759 	mp->msg_controllen = q - (char *)mp->msg_control;
760 	return error;
761 }
762 
763 int
764 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
765 {
766 	/* {
767 		syscallarg(int) s;
768 		syscallarg(struct linux_msghdr *) msg;
769 		syscallarg(u_int) flags;
770 	} */
771 	struct msghdr	msg;
772 	struct linux_msghdr lmsg;
773 	int		error;
774 	struct mbuf	*from, *control;
775 
776 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
777 	if (error)
778 		return (error);
779 	linux_to_bsd_msghdr(&lmsg, &msg);
780 
781 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
782 	if (msg.msg_flags < 0) {
783 		/* Some unsupported flag */
784 		return (EINVAL);
785 	}
786 	msg.msg_flags |= MSG_IOVUSRSPACE;
787 
788 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
789 	    msg.msg_control != NULL ? &control : NULL, retval);
790 	if (error != 0)
791 		return error;
792 
793 	if (msg.msg_control != NULL)
794 		error = linux_copyout_msg_control(l, &msg, control);
795 
796 	if (error == 0 && from != 0) {
797 		mtod(from, struct osockaddr *)->sa_family =
798 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
799 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
800 			from);
801 	} else
802 		msg.msg_namelen = 0;
803 
804 	if (from != NULL)
805 		m_free(from);
806 
807 	if (error == 0) {
808 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
809 		if (msg.msg_flags < 0)
810 			/* Some flag unsupported by Linux */
811 			error = EINVAL;
812 		else {
813 			ktrkuser("msghdr", &msg, sizeof(msg));
814 			bsd_to_linux_msghdr(&msg, &lmsg);
815 			error = copyout(&lmsg, SCARG(uap, msg), sizeof(lmsg));
816 		}
817 	}
818 
819 	return (error);
820 }
821 
822 /*
823  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
824  * is different, the rest matches IPPROTO_* on both systems.
825  */
826 int
827 linux_to_bsd_sopt_level(int llevel)
828 {
829 
830 	switch (llevel) {
831 	case LINUX_SOL_SOCKET:
832 		return SOL_SOCKET;
833 	case LINUX_SOL_IP:
834 		return IPPROTO_IP;
835 #ifdef INET6
836 	case LINUX_SOL_IPV6:
837 		return IPPROTO_IPV6;
838 #endif
839 	case LINUX_SOL_TCP:
840 		return IPPROTO_TCP;
841 	case LINUX_SOL_UDP:
842 		return IPPROTO_UDP;
843 	default:
844 		return -1;
845 	}
846 }
847 
848 /*
849  * Convert Linux socket level socket option numbers to NetBSD values.
850  */
851 int
852 linux_to_bsd_so_sockopt(int lopt)
853 {
854 
855 	switch (lopt) {
856 	case LINUX_SO_DEBUG:
857 		return SO_DEBUG;
858 	case LINUX_SO_REUSEADDR:
859 		/*
860 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
861 		 * host:port pair through SO_REUSEADDR even if the address is not a
862 		 * multicast-address.  Effectively, this means that we should use
863 		 * SO_REUSEPORT to allow Linux applications to not exit with
864 		 * EADDRINUSE
865 		 */
866 		return SO_REUSEPORT;
867 	case LINUX_SO_TYPE:
868 		return SO_TYPE;
869 	case LINUX_SO_ERROR:
870 		return SO_ERROR;
871 	case LINUX_SO_DONTROUTE:
872 		return SO_DONTROUTE;
873 	case LINUX_SO_BROADCAST:
874 		return SO_BROADCAST;
875 	case LINUX_SO_SNDBUF:
876 		return SO_SNDBUF;
877 	case LINUX_SO_RCVBUF:
878 		return SO_RCVBUF;
879 	case LINUX_SO_SNDLOWAT:
880 		return SO_SNDLOWAT;
881 	case LINUX_SO_RCVLOWAT:
882 		return SO_RCVLOWAT;
883 	case LINUX_SO_KEEPALIVE:
884 		return SO_KEEPALIVE;
885 	case LINUX_SO_OOBINLINE:
886 		return SO_OOBINLINE;
887 	case LINUX_SO_LINGER:
888 		return SO_LINGER;
889 	case LINUX_SO_ACCEPTCONN:
890 		return SO_ACCEPTCONN;
891 	case LINUX_SO_PRIORITY:
892 	case LINUX_SO_NO_CHECK:
893 	default:
894 		return -1;
895 	}
896 }
897 
898 /*
899  * Convert Linux IP level socket option number to NetBSD values.
900  */
901 int
902 linux_to_bsd_ip_sockopt(int lopt)
903 {
904 
905 	switch (lopt) {
906 	case LINUX_IP_TOS:
907 		return IP_TOS;
908 	case LINUX_IP_TTL:
909 		return IP_TTL;
910 	case LINUX_IP_HDRINCL:
911 		return IP_HDRINCL;
912 	case LINUX_IP_MULTICAST_TTL:
913 		return IP_MULTICAST_TTL;
914 	case LINUX_IP_MULTICAST_LOOP:
915 		return IP_MULTICAST_LOOP;
916 	case LINUX_IP_MULTICAST_IF:
917 		return IP_MULTICAST_IF;
918 	case LINUX_IP_ADD_MEMBERSHIP:
919 		return IP_ADD_MEMBERSHIP;
920 	case LINUX_IP_DROP_MEMBERSHIP:
921 		return IP_DROP_MEMBERSHIP;
922 	default:
923 		return -1;
924 	}
925 }
926 
927 /*
928  * Convert Linux IPV6 level socket option number to NetBSD values.
929  */
930 #ifdef INET6
931 int
932 linux_to_bsd_ipv6_sockopt(int lopt)
933 {
934 
935 	switch (lopt) {
936 	case LINUX_IPV6_V6ONLY:
937 		return IPV6_V6ONLY;
938 	default:
939 		return -1;
940 	}
941 }
942 #endif
943 
944 /*
945  * Convert Linux TCP level socket option number to NetBSD values.
946  */
947 int
948 linux_to_bsd_tcp_sockopt(int lopt)
949 {
950 
951 	switch (lopt) {
952 	case LINUX_TCP_NODELAY:
953 		return TCP_NODELAY;
954 	case LINUX_TCP_MAXSEG:
955 		return TCP_MAXSEG;
956 	default:
957 		return -1;
958 	}
959 }
960 
961 /*
962  * Convert Linux UDP level socket option number to NetBSD values.
963  */
964 int
965 linux_to_bsd_udp_sockopt(int lopt)
966 {
967 
968 	switch (lopt) {
969 	default:
970 		return -1;
971 	}
972 }
973 
974 /*
975  * Another reasonably straightforward function: setsockopt(2).
976  * The level and option numbers are converted; the values passed
977  * are not (yet) converted, the ones currently implemented don't
978  * need conversion, as they are the same on both systems.
979  */
980 int
981 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
982 {
983 	/* {
984 		syscallarg(int) s;
985 		syscallarg(int) level;
986 		syscallarg(int) optname;
987 		syscallarg(void *) optval;
988 		syscallarg(int) optlen;
989 	} */
990 	struct sys_setsockopt_args bsa;
991 	int name;
992 
993 	SCARG(&bsa, s) = SCARG(uap, s);
994 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
995 	SCARG(&bsa, val) = SCARG(uap, optval);
996 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
997 
998 	/*
999 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
1000 	 * and returns EOPNOTSUPP for other levels
1001 	 */
1002 	if (SCARG(&bsa, level) != SOL_SOCKET) {
1003 		struct socket *so;
1004 		int error, family;
1005 
1006 		/* fd_getsock() will use the descriptor for us */
1007 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
1008 		    	return error;
1009 		family = so->so_proto->pr_domain->dom_family;
1010 		fd_putfile(SCARG(&bsa, s));
1011 
1012 		if (family == AF_LOCAL)
1013 			return EOPNOTSUPP;
1014 	}
1015 
1016 	switch (SCARG(&bsa, level)) {
1017 	case SOL_SOCKET:
1018 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1019 		break;
1020 	case IPPROTO_IP:
1021 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1022 		break;
1023 #ifdef INET6
1024 	case IPPROTO_IPV6:
1025 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1026 		break;
1027 #endif
1028 	case IPPROTO_TCP:
1029 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1030 		break;
1031 	case IPPROTO_UDP:
1032 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1033 		break;
1034 	default:
1035 		return EINVAL;
1036 	}
1037 
1038 	if (name == -1)
1039 		return EINVAL;
1040 	SCARG(&bsa, name) = name;
1041 
1042 	return sys_setsockopt(l, &bsa, retval);
1043 }
1044 
1045 /*
1046  * getsockopt(2) is very much the same as setsockopt(2) (see above)
1047  */
1048 int
1049 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
1050 {
1051 	/* {
1052 		syscallarg(int) s;
1053 		syscallarg(int) level;
1054 		syscallarg(int) optname;
1055 		syscallarg(void *) optval;
1056 		syscallarg(int *) optlen;
1057 	} */
1058 	struct sys_getsockopt_args bga;
1059 	int name;
1060 
1061 	SCARG(&bga, s) = SCARG(uap, s);
1062 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
1063 	SCARG(&bga, val) = SCARG(uap, optval);
1064 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
1065 
1066 	switch (SCARG(&bga, level)) {
1067 	case SOL_SOCKET:
1068 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1069 		break;
1070 	case IPPROTO_IP:
1071 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1072 		break;
1073 #ifdef INET6
1074 	case IPPROTO_IPV6:
1075 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1076 		break;
1077 #endif
1078 	case IPPROTO_TCP:
1079 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1080 		break;
1081 	case IPPROTO_UDP:
1082 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1083 		break;
1084 	default:
1085 		return EINVAL;
1086 	}
1087 
1088 	if (name == -1)
1089 		return EINVAL;
1090 	SCARG(&bga, name) = name;
1091 
1092 	return sys_getsockopt(l, &bga, retval);
1093 }
1094 
1095 int
1096 linux_getifname(struct lwp *l, register_t *retval, void *data)
1097 {
1098 	struct ifnet *ifp;
1099 	struct linux_ifreq ifr;
1100 	int error;
1101 	int s;
1102 
1103 	error = copyin(data, &ifr, sizeof(ifr));
1104 	if (error)
1105 		return error;
1106 
1107 	s = pserialize_read_enter();
1108 	ifp = if_byindex(ifr.ifr_ifru.ifru_ifindex);
1109 	if (ifp == NULL) {
1110 		pserialize_read_exit(s);
1111 		return ENODEV;
1112 	}
1113 
1114 	strncpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name));
1115 	pserialize_read_exit(s);
1116 
1117 	return copyout(&ifr, data, sizeof(ifr));
1118 }
1119 
1120 int
1121 linux_getifconf(struct lwp *l, register_t *retval, void *data)
1122 {
1123 	struct linux_ifreq ifr, *ifrp = NULL;
1124 	struct linux_ifconf ifc;
1125 	struct ifnet *ifp;
1126 	struct sockaddr *sa;
1127 	struct osockaddr *osa;
1128 	int space = 0, error;
1129 	const int sz = (int)sizeof(ifr);
1130 	bool docopy;
1131 	int s;
1132 	int bound;
1133 	struct psref psref;
1134 
1135 	error = copyin(data, &ifc, sizeof(ifc));
1136 	if (error)
1137 		return error;
1138 
1139 	docopy = ifc.ifc_req != NULL;
1140 	if (docopy) {
1141 		space = ifc.ifc_len;
1142 		ifrp = ifc.ifc_req;
1143 	}
1144 
1145 	bound = curlwp_bind();
1146 	s = pserialize_read_enter();
1147 	IFNET_READER_FOREACH(ifp) {
1148 		struct ifaddr *ifa;
1149 		if_acquire(ifp, &psref);
1150 		pserialize_read_exit(s);
1151 
1152 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1153 		    sizeof(ifr.ifr_name));
1154 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') {
1155 			error = ENAMETOOLONG;
1156 			goto release_exit;
1157 		}
1158 
1159 		s = pserialize_read_enter();
1160 		IFADDR_READER_FOREACH(ifa, ifp) {
1161 			struct psref psref_ifa;
1162 			ifa_acquire(ifa, &psref_ifa);
1163 			pserialize_read_exit(s);
1164 
1165 			sa = ifa->ifa_addr;
1166 			if (sa->sa_family != AF_INET ||
1167 			    sa->sa_len > sizeof(*osa))
1168 				goto next;
1169 			memcpy(&ifr.ifr_addr, sa, sa->sa_len);
1170 			osa = (struct osockaddr *)&ifr.ifr_addr;
1171 			osa->sa_family = sa->sa_family;
1172 			if (space >= sz) {
1173 				error = copyout(&ifr, ifrp, sz);
1174 				if (error != 0) {
1175 					ifa_release(ifa, &psref_ifa);
1176 					goto release_exit;
1177 				}
1178 				ifrp++;
1179 			}
1180 			space -= sz;
1181 		next:
1182 			s = pserialize_read_enter();
1183 			ifa_release(ifa, &psref_ifa);
1184 		}
1185 
1186 		KASSERT(pserialize_in_read_section());
1187 		if_release(ifp, &psref);
1188 	}
1189 	pserialize_read_exit(s);
1190 	curlwp_bindx(bound);
1191 
1192 	if (docopy)
1193 		ifc.ifc_len -= space;
1194 	else
1195 		ifc.ifc_len = -space;
1196 
1197 	return copyout(&ifc, data, sizeof(ifc));
1198 
1199 release_exit:
1200 	if_release(ifp, &psref);
1201 	curlwp_bindx(bound);
1202 	return error;
1203 }
1204 
1205 int
1206 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1207     void *data)
1208 {
1209 	/* Not the full structure, just enough to map what we do here */
1210 	struct linux_ifreq lreq;
1211 	file_t *fp;
1212 	struct ifaddr *ifa;
1213 	struct ifnet *ifp;
1214 	struct sockaddr_dl *sadl;
1215 	int error, found;
1216 	int index, ifnum;
1217 	int s;
1218 
1219 	/*
1220 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1221 	 * SIOCGIFCONF, because the user buffer is not of the right
1222 	 * type to take those results.  We can't use kernel buffers to
1223 	 * receive the results, as the implementation of sys_ioctl()
1224 	 * and ifconf() [which implements SIOCGIFCONF] use
1225 	 * copyin()/copyout() which will fail on kernel addresses.
1226 	 *
1227 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1228 	 */
1229 
1230 	if ((fp = fd_getfile(fd)) == NULL)
1231 		return (EBADF);
1232 
1233 	KERNEL_LOCK(1, NULL);
1234 
1235 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1236 		error = EBADF;
1237 		goto out;
1238 	}
1239 
1240 	error = copyin(data, &lreq, sizeof(lreq));
1241 	if (error)
1242 		goto out;
1243 	lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0';		/* just in case */
1244 
1245 	/*
1246 	 * Try real interface name first, then fake "ethX"
1247 	 */
1248 	found = 0;
1249 	s = pserialize_read_enter();
1250 	IFNET_READER_FOREACH(ifp) {
1251 		if (found)
1252 			break;
1253 		if (strcmp(lreq.ifr_name, ifp->if_xname))
1254 			/* not this interface */
1255 			continue;
1256 
1257 		found=1;
1258 		if (IFADDR_READER_EMPTY(ifp)) {
1259 			pserialize_read_exit(s);
1260 			error = ENODEV;
1261 			goto out;
1262 		}
1263 		IFADDR_READER_FOREACH(ifa, ifp) {
1264 			sadl = satosdl(ifa->ifa_addr);
1265 			/* only return ethernet addresses */
1266 			/* XXX what about FDDI, etc. ? */
1267 			if (sadl->sdl_family != AF_LINK ||
1268 			    sadl->sdl_type != IFT_ETHER)
1269 				continue;
1270 			memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl),
1271 			       MIN(sadl->sdl_alen,
1272 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1273 			lreq.ifr_hwaddr.sa_family =
1274 				sadl->sdl_family;
1275 			pserialize_read_exit(s);
1276 
1277 			error = copyout(&lreq, data, sizeof(lreq));
1278 			goto out;
1279 		}
1280 	}
1281 	pserialize_read_exit(s);
1282 
1283 	if (strncmp(lreq.ifr_name, "eth", 3) != 0) {
1284 		/* unknown interface, not even an "eth*" name */
1285 		error = ENODEV;
1286 		goto out;
1287 	}
1288 
1289 	for (ifnum = 0, index = 3;
1290 	     index < LINUX_IFNAMSIZ && lreq.ifr_name[index] != '\0';
1291 	     index++) {
1292 		ifnum *= 10;
1293 		ifnum += lreq.ifr_name[index] - '0';
1294 	}
1295 
1296 	error = EINVAL;			/* in case we don't find one */
1297 	s = pserialize_read_enter();
1298 	IFNET_READER_FOREACH(ifp) {
1299 		memcpy(lreq.ifr_name, ifp->if_xname,
1300 		       MIN(LINUX_IFNAMSIZ, IFNAMSIZ));
1301 		IFADDR_READER_FOREACH(ifa, ifp) {
1302 			sadl = satosdl(ifa->ifa_addr);
1303 			/* only return ethernet addresses */
1304 			/* XXX what about FDDI, etc. ? */
1305 			if (sadl->sdl_family != AF_LINK ||
1306 			    sadl->sdl_type != IFT_ETHER)
1307 				continue;
1308 			if (ifnum--)
1309 				/* not the reqested iface */
1310 				continue;
1311 			memcpy(&lreq.ifr_hwaddr.sa_data,
1312 			       CLLADDR(sadl),
1313 			       MIN(sadl->sdl_alen,
1314 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1315 			lreq.ifr_hwaddr.sa_family =
1316 				sadl->sdl_family;
1317 			pserialize_read_exit(s);
1318 
1319 			error = copyout(&lreq, data, sizeof(lreq));
1320 			goto out;
1321 		}
1322 	}
1323 	pserialize_read_exit(s);
1324 
1325 out:
1326 	KERNEL_UNLOCK_ONE(NULL);
1327 	fd_putfile(fd);
1328 	return error;
1329 }
1330 
1331 int
1332 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1333 {
1334 	/* {
1335 		syscallarg(int) fd;
1336 		syscallarg(u_long) com;
1337 		syscallarg(void *) data;
1338 	} */
1339 	u_long com;
1340 	int error = 0, isdev = 0, dosys = 1;
1341 	struct sys_ioctl_args ia;
1342 	file_t *fp;
1343 	struct vnode *vp;
1344 	int (*ioctlf)(file_t *, u_long, void *);
1345 	struct ioctl_pt pt;
1346 
1347 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1348 		return (EBADF);
1349 
1350 	if (fp->f_type == DTYPE_VNODE) {
1351 		vp = (struct vnode *)fp->f_data;
1352 		isdev = vp->v_type == VCHR;
1353 	}
1354 
1355 	/*
1356 	 * Don't try to interpret socket ioctl calls that are done
1357 	 * on a device filedescriptor, just pass them through, to
1358 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1359 	 * device will only handle these if it's prepared to do
1360 	 * so, to avoid unexpected things from happening.
1361 	 */
1362 	if (isdev) {
1363 		dosys = 0;
1364 		ioctlf = fp->f_ops->fo_ioctl;
1365 		pt.com = SCARG(uap, com);
1366 		pt.data = SCARG(uap, data);
1367 		error = ioctlf(fp, PTIOCLINUX, &pt);
1368 		/*
1369 		 * XXX hack: if the function returns EJUSTRETURN,
1370 		 * it has stuffed a sysctl return value in pt.data.
1371 		 */
1372 		if (error == EJUSTRETURN) {
1373 			retval[0] = (register_t)pt.data;
1374 			error = 0;
1375 		}
1376 		goto out;
1377 	}
1378 
1379 	com = SCARG(uap, com);
1380 	retval[0] = 0;
1381 
1382 	switch (com) {
1383 	case LINUX_SIOCGIFNAME:
1384 		error = linux_getifname(l, retval, SCARG(uap, data));
1385 		dosys = 0;
1386 		break;
1387 	case LINUX_SIOCGIFCONF:
1388 		error = linux_getifconf(l, retval, SCARG(uap, data));
1389 		dosys = 0;
1390 		break;
1391 	case LINUX_SIOCGIFFLAGS:
1392 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1393 		break;
1394 	case LINUX_SIOCSIFFLAGS:
1395 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1396 		break;
1397 	case LINUX_SIOCGIFADDR:
1398 		SCARG(&ia, com) = OOSIOCGIFADDR;
1399 		break;
1400 	case LINUX_SIOCGIFDSTADDR:
1401 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1402 		break;
1403 	case LINUX_SIOCGIFBRDADDR:
1404 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1405 		break;
1406 	case LINUX_SIOCGIFNETMASK:
1407 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1408 		break;
1409 	case LINUX_SIOCGIFMTU:
1410 		SCARG(&ia, com) = OSIOCGIFMTU;
1411 		break;
1412 	case LINUX_SIOCADDMULTI:
1413 		SCARG(&ia, com) = OSIOCADDMULTI;
1414 		break;
1415 	case LINUX_SIOCDELMULTI:
1416 		SCARG(&ia, com) = OSIOCDELMULTI;
1417 		break;
1418 	case LINUX_SIOCGIFHWADDR:
1419 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1420 		    SCARG(uap, data));
1421 		dosys = 0;
1422 		break;
1423 	default:
1424 		error = EINVAL;
1425 	}
1426 
1427  out:
1428  	fd_putfile(SCARG(uap, fd));
1429 
1430 	if (error ==0 && dosys) {
1431 		SCARG(&ia, fd) = SCARG(uap, fd);
1432 		SCARG(&ia, data) = SCARG(uap, data);
1433 		error = sys_ioctl(curlwp, &ia, retval);
1434 	}
1435 
1436 	return error;
1437 }
1438 
1439 int
1440 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1441 {
1442 	/* {
1443 		syscallarg(int) s;
1444 		syscallarg(const struct sockaddr *) name;
1445 		syscallarg(int) namelen;
1446 	} */
1447 	int		error;
1448 	struct sockaddr_big sb;
1449 
1450 	error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name),
1451 	    SCARG(uap, namelen));
1452 	if (error)
1453 		return (error);
1454 
1455 	error = do_sys_connect(l, SCARG(uap, s), (struct sockaddr *)&sb);
1456 
1457 	if (error == EISCONN) {
1458 		struct socket *so;
1459 		int state, prflags;
1460 
1461 		/* fd_getsock() will use the descriptor for us */
1462 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1463 		    	return EISCONN;
1464 
1465 		solock(so);
1466 		state = so->so_state;
1467 		prflags = so->so_proto->pr_flags;
1468 		sounlock(so);
1469 		fd_putfile(SCARG(uap, s));
1470 		/*
1471 		 * We should only let this call succeed once per
1472 		 * non-blocking connect; however we don't have
1473 		 * a convenient place to keep that state..
1474 		 */
1475 		if ((state & (SS_ISCONNECTED|SS_NBIO)) ==
1476 		    (SS_ISCONNECTED|SS_NBIO) &&
1477 		    (prflags & PR_CONNREQUIRED))
1478 			return 0;
1479 	}
1480 
1481 	return (error);
1482 }
1483 
1484 int
1485 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1486 {
1487 	/* {
1488 		syscallarg(int) s;
1489 		syscallarg(const struct osockaddr *) name;
1490 		syscallarg(int) namelen;
1491 	} */
1492 	int		error;
1493 	struct sockaddr_big sb;
1494 
1495 	error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name),
1496 	    SCARG(uap, namelen));
1497 	if (error)
1498 		return (error);
1499 
1500 	return do_sys_bind(l, SCARG(uap, s), (struct sockaddr *)&sb);
1501 }
1502 
1503 int
1504 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1505 {
1506 	/* {
1507 		syscallarg(int) fdes;
1508 		syscallarg(void *) asa;
1509 		syscallarg(int *) alen;
1510 	} */
1511 	int error;
1512 
1513 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1514 		return (error);
1515 
1516 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1517 		return (error);
1518 
1519 	return (0);
1520 }
1521 
1522 int
1523 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1524 {
1525 	/* {
1526 		syscallarg(int) fdes;
1527 		syscallarg(void *) asa;
1528 		syscallarg(int *) alen;
1529 	} */
1530 	int error;
1531 
1532 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1533 		return (error);
1534 
1535 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1536 		return (error);
1537 
1538 	return (0);
1539 }
1540 
1541 /*
1542  * Copy the osockaddr structure pointed to by name to sb, adjust
1543  * family and convert to sockaddr.
1544  */
1545 static int
1546 linux_get_sa(struct lwp *l, int s, struct sockaddr_big *sb,
1547     const struct osockaddr *name, socklen_t namelen)
1548 {
1549 	int error, bdom;
1550 
1551 	if (namelen > UCHAR_MAX ||
1552 	    namelen <= offsetof(struct sockaddr_big, sb_data))
1553 		return EINVAL;
1554 
1555 	error = copyin(name, sb, namelen);
1556 	if (error)
1557 		return error;
1558 
1559 	bdom = linux_to_bsd_domain(sb->sb_family);
1560 	if (bdom == -1)
1561 		return EINVAL;
1562 
1563 	/*
1564 	 * If the family is unspecified, use address family of the socket.
1565 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1566 	 */
1567 	if (bdom == AF_UNSPEC) {
1568 		struct socket *so;
1569 
1570 		/* fd_getsock() will use the descriptor for us */
1571 		if ((error = fd_getsock(s, &so)) != 0)
1572 			return error;
1573 
1574 		bdom = so->so_proto->pr_domain->dom_family;
1575 		fd_putfile(s);
1576 	}
1577 
1578 	/*
1579 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1580 	 * which lacks the scope id compared with RFC2553 one. If we detect
1581 	 * the situation, reject the address and write a message to system log.
1582 	 *
1583 	 * Still accept addresses for which the scope id is not used.
1584 	 */
1585 	if (bdom == AF_INET6 &&
1586 	    namelen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
1587 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sb;
1588 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1589 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1590 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1591 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1592 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1593 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1594 			struct proc *p = l->l_proc;
1595 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1596 
1597 			log(LOG_DEBUG,
1598 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1599 			    "sockaddr_in6 rejected",
1600 			    p->p_pid, p->p_comm, uid);
1601 			return EINVAL;
1602 		}
1603 		namelen = sizeof(struct sockaddr_in6);
1604 		sin6->sin6_scope_id = 0;
1605 	}
1606 
1607 	if (bdom == AF_INET)
1608 		namelen = sizeof(struct sockaddr_in);
1609 
1610 	sb->sb_family = bdom;
1611 	sb->sb_len = namelen;
1612 	ktrkuser("mbsoname", sb, namelen);
1613 	return 0;
1614 }
1615 
1616 static int
1617 linux_sa_put(struct osockaddr *osa)
1618 {
1619 	struct sockaddr sa;
1620 	struct osockaddr *kosa;
1621 	int error, bdom, len;
1622 
1623 	/*
1624 	 * Only read/write the sockaddr family and length part, the rest is
1625 	 * not changed.
1626 	 */
1627 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1628 
1629 	error = copyin(osa, &sa, len);
1630 	if (error)
1631 		return (error);
1632 
1633 	bdom = bsd_to_linux_domain(sa.sa_family);
1634 	if (bdom == -1)
1635 		return (EINVAL);
1636 
1637 	/* Note: we convert from sockaddr to osockaddr here, too */
1638 	kosa = (struct osockaddr *) &sa;
1639 	kosa->sa_family = bdom;
1640 	error = copyout(kosa, osa, len);
1641 	if (error)
1642 		return (error);
1643 
1644 	return (0);
1645 }
1646 
1647 #ifndef __amd64__
1648 int
1649 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1650 {
1651 	/* {
1652 		syscallarg(int) s;
1653 		syscallarg(void *) buf;
1654 		syscallarg(int) len;
1655 		syscallarg(int) flags;
1656 	} */
1657 	struct sys_recvfrom_args bra;
1658 
1659 
1660 	SCARG(&bra, s) = SCARG(uap, s);
1661 	SCARG(&bra, buf) = SCARG(uap, buf);
1662 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1663 	SCARG(&bra, flags) = SCARG(uap, flags);
1664 	SCARG(&bra, from) = NULL;
1665 	SCARG(&bra, fromlenaddr) = NULL;
1666 
1667 	return (sys_recvfrom(l, &bra, retval));
1668 }
1669 
1670 int
1671 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1672 {
1673 	/* {
1674 		syscallarg(int) s;
1675 		syscallarg(void *) buf;
1676 		syscallarg(int) len;
1677 		syscallarg(int) flags;
1678 	} */
1679 	struct sys_sendto_args bsa;
1680 
1681 	SCARG(&bsa, s)		= SCARG(uap, s);
1682 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1683 	SCARG(&bsa, len)	= SCARG(uap, len);
1684 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1685 	SCARG(&bsa, to)		= NULL;
1686 	SCARG(&bsa, tolen)	= 0;
1687 
1688 	return (sys_sendto(l, &bsa, retval));
1689 }
1690 #endif
1691 
1692 int
1693 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1694 {
1695 	/* {
1696 		syscallarg(int) s;
1697 		syscallarg(struct osockaddr *) name;
1698 		syscallarg(int *) anamelen;
1699 	} */
1700 	int error;
1701 	struct sys_accept_args baa;
1702 
1703 	SCARG(&baa, s)		= SCARG(uap, s);
1704 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1705 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1706 
1707 	if ((error = sys_accept(l, &baa, retval)))
1708 		return (error);
1709 
1710 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1711 		return (error);
1712 
1713 	return (0);
1714 }
1715 
1716 int
1717 linux_sys_accept4(struct lwp *l, const struct linux_sys_accept4_args *uap, register_t *retval)
1718 {
1719 	/* {
1720 		syscallarg(int) s;
1721 		syscallarg(struct osockaddr *) name;
1722 		syscallarg(int *) anamelen;
1723 		syscallarg(int) flags;
1724 	} */
1725 	int error, flags;
1726 	struct sockaddr_big name;
1727 
1728 	if ((flags = linux_to_bsd_type(SCARG(uap, flags))) == -1)
1729 		return EINVAL;
1730 
1731 	name.sb_len = UCHAR_MAX;
1732 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
1733 	    retval, NULL, flags, 0);
1734 	if (error != 0)
1735 		return error;
1736 
1737 	error = copyout_sockname_sb((struct sockaddr *)SCARG(uap, name),
1738 	    SCARG(uap, anamelen), MSG_LENUSRSPACE, &name);
1739 	if (error != 0) {
1740 		int fd = (int)*retval;
1741 		if (fd_getfile(fd) != NULL)
1742 			(void)fd_close(fd);
1743 		return error;
1744 	}
1745 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1746 		return error;
1747 
1748 	return 0;
1749 }
1750 
1751 int
1752 linux_sys_sendmmsg(struct lwp *l, const struct linux_sys_sendmmsg_args *uap,
1753     register_t *retval)
1754 {
1755 	/* {
1756 		syscallarg(int) s;
1757 		syscallarg(struct linux_mmsghdr *) msgvec;
1758 		syscallarg(unsigned int) vlen;
1759 		syscallarg(unsigned int) flags;
1760 	} */
1761 	struct linux_mmsghdr lmsg;
1762 	struct mmsghdr bmsg;
1763 	struct socket *so;
1764 	file_t *fp;
1765 	struct msghdr *msg = &bmsg.msg_hdr;
1766 	int error, s;
1767 	unsigned int vlen, flags, dg;
1768 
1769 	if ((flags = linux_to_bsd_msg_flags(SCARG(uap, flags))) == -1)
1770 		return EINVAL;
1771 
1772 	flags = (flags & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1773 
1774 	s = SCARG(uap, s);
1775 	if ((error = fd_getsock1(s, &so, &fp)) != 0)
1776 		return error;
1777 
1778 	vlen = SCARG(uap, vlen);
1779 	if (vlen > 1024)
1780 		vlen = 1024;
1781 
1782 	for (dg = 0; dg < vlen;) {
1783 		error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg));
1784 		if (error)
1785 			break;
1786 		linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr);
1787 
1788 		msg->msg_flags = flags;
1789 
1790 		error = do_sys_sendmsg_so(l, s, so, fp, msg, flags, retval);
1791 		if (error)
1792 			break;
1793 
1794 		ktrkuser("msghdr", msg, sizeof *msg);
1795 		lmsg.msg_len = *retval;
1796 		error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg));
1797 		if (error)
1798 			break;
1799 		dg++;
1800 
1801 	}
1802 
1803 	*retval = dg;
1804 	if (error)
1805 		so->so_error = error;
1806 
1807 	fd_putfile(s);
1808 
1809 	/*
1810 	 * If we succeeded at least once, return 0, hopefully so->so_error
1811 	 * will catch it next time.
1812 	 */
1813 	if (dg)
1814 		return 0;
1815 	return error;
1816 }
1817 
1818 int
1819 linux_sys_recvmmsg(struct lwp *l, const struct linux_sys_recvmmsg_args *uap,
1820     register_t *retval)
1821 {
1822 	/* {
1823 		syscallarg(int) s;
1824 		syscallarg(struct linux_mmsghdr *) msgvec;
1825 		syscallarg(unsigned int) vlen;
1826 		syscallarg(unsigned int) flags;
1827 		syscallarg(struct linux_timespec *) timeout;
1828 	} */
1829 	struct linux_mmsghdr lmsg;
1830 	struct mmsghdr bmsg;
1831 	struct socket *so;
1832 	struct msghdr *msg = &bmsg.msg_hdr;
1833 	int error, s;
1834 	struct mbuf *from, *control;
1835 	struct timespec ts, now;
1836 	struct linux_timespec lts;
1837 	unsigned int vlen, flags, dg;
1838 
1839 	if (SCARG(uap, timeout)) {
1840 		error = copyin(SCARG(uap, timeout), &lts, sizeof(lts));
1841 			return error;
1842 		ts.tv_sec = lts.tv_sec;
1843 		ts.tv_nsec = lts.tv_nsec;
1844 		getnanotime(&now);
1845 		timespecadd(&now, &ts, &ts);
1846 	}
1847 
1848 	s = SCARG(uap, s);
1849 	if ((error = fd_getsock(s, &so)) != 0)
1850 		return error;
1851 
1852 	vlen = SCARG(uap, vlen);
1853 	if (vlen > 1024)
1854 		vlen = 1024;
1855 
1856 	from = NULL;
1857 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1858 
1859 	for (dg = 0; dg < vlen;) {
1860 		error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg));
1861 		if (error)
1862 			break;
1863 		linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr);
1864 		msg->msg_flags = flags & ~MSG_WAITFORONE;
1865 
1866 		if (from != NULL) {
1867 			m_free(from);
1868 			from = NULL;
1869 		}
1870 
1871 		error = do_sys_recvmsg_so(l, s, so, msg, &from,
1872 		    msg->msg_control != NULL ? &control : NULL, retval);
1873 		if (error) {
1874 			if (error == EAGAIN && dg > 0)
1875 				error = 0;
1876 			break;
1877 		}
1878 
1879 		if (msg->msg_control != NULL)
1880 			error = linux_copyout_msg_control(l, msg, control);
1881 		if (error)
1882 			break;
1883 
1884 		if (from != NULL) {
1885 			mtod(from, struct osockaddr *)->sa_family =
1886 			    bsd_to_linux_domain(mtod(from,
1887 			    struct sockaddr *)->sa_family);
1888 			error = copyout_sockname(msg->msg_name,
1889 			    &msg->msg_namelen, 0, from);
1890 			if (error)
1891 				break;
1892 		}
1893 
1894 
1895 		lmsg.msg_len = *retval;
1896 		ktrkuser("msghdr", msg, sizeof(*msg));
1897 		bsd_to_linux_msghdr(msg, &lmsg.msg_hdr);
1898 		error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg));
1899 		if (error)
1900 			break;
1901 
1902 		dg++;
1903 		if (msg->msg_flags & MSG_OOB)
1904 			break;
1905 
1906 		if (SCARG(uap, timeout)) {
1907 			getnanotime(&now);
1908 			timespecsub(&now, &ts, &now);
1909 			if (now.tv_sec > 0)
1910 				break;
1911 		}
1912 
1913 		if (flags & MSG_WAITFORONE)
1914 			flags |= MSG_DONTWAIT;
1915 
1916 	}
1917 
1918 	if (from != NULL)
1919 		m_free(from);
1920 
1921 	*retval = dg;
1922 	if (error)
1923 		so->so_error = error;
1924 
1925 	fd_putfile(s);
1926 
1927 	/*
1928 	 * If we succeeded at least once, return 0, hopefully so->so_error
1929 	 * will catch it next time.
1930 	 */
1931 	if (dg)
1932 		return 0;
1933 
1934 	return error;
1935 }
1936