xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision cac8e449158efc7261bebc8657cbb0125a2cfdde)
1 /*	$NetBSD: linux_socket.c,v 1.97 2008/07/03 14:07:09 njoly 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 1.97 2008/07/03 14:07:09 njoly Exp $");
39 
40 #if defined(_KERNEL_OPT)
41 #include "opt_inet.h"
42 #endif /* defined(_KERNEL_OPT) */
43 
44 #include <sys/param.h>
45 #include <sys/kernel.h>
46 #include <sys/systm.h>
47 #include <sys/buf.h>
48 #include <sys/malloc.h>
49 #include <sys/ioctl.h>
50 #include <sys/tty.h>
51 #include <sys/file.h>
52 #include <sys/filedesc.h>
53 #include <sys/select.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/domain.h>
57 #include <net/if.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <netinet/in.h>
61 #include <netinet/tcp.h>
62 #include <sys/mount.h>
63 #include <sys/proc.h>
64 #include <sys/vnode.h>
65 #include <sys/device.h>
66 #include <sys/protosw.h>
67 #include <sys/mbuf.h>
68 #include <sys/syslog.h>
69 #include <sys/exec.h>
70 #include <sys/kauth.h>
71 #include <sys/syscallargs.h>
72 #include <sys/ktrace.h>
73 
74 #include <lib/libkern/libkern.h>
75 
76 #ifdef INET6
77 #include <netinet/ip6.h>
78 #include <netinet6/ip6_var.h>
79 #endif
80 
81 #include <compat/sys/socket.h>
82 #include <compat/sys/sockio.h>
83 
84 #include <compat/linux/common/linux_types.h>
85 #include <compat/linux/common/linux_util.h>
86 #include <compat/linux/common/linux_signal.h>
87 #include <compat/linux/common/linux_ioctl.h>
88 #include <compat/linux/common/linux_socket.h>
89 #if !defined(__alpha__) && !defined(__amd64__)
90 #include <compat/linux/common/linux_socketcall.h>
91 #endif
92 #include <compat/linux/common/linux_sockio.h>
93 #include <compat/linux/common/linux_ipc.h>
94 #include <compat/linux/common/linux_sem.h>
95 
96 #include <compat/linux/linux_syscallargs.h>
97 
98 #ifdef DEBUG_LINUX
99 #define DPRINTF(a) uprintf a
100 #else
101 #define DPRINTF(a)
102 #endif
103 
104 /*
105  * The calls in this file are entered either via the linux_socketcall()
106  * interface or, on the Alpha, as individual syscalls.  The
107  * linux_socketcall function does any massaging of arguments so that all
108  * the calls in here need not think that they are anything other
109  * than a normal syscall.
110  */
111 
112 static int linux_to_bsd_domain(int);
113 static int bsd_to_linux_domain(int);
114 int linux_to_bsd_sopt_level(int);
115 int linux_to_bsd_so_sockopt(int);
116 int linux_to_bsd_ip_sockopt(int);
117 int linux_to_bsd_tcp_sockopt(int);
118 int linux_to_bsd_udp_sockopt(int);
119 int linux_getifconf(struct lwp *, register_t *, void *);
120 int linux_getifhwaddr(struct lwp *, register_t *, u_int, void *);
121 static int linux_get_sa(struct lwp *, int, struct mbuf **,
122 		const struct osockaddr *, unsigned int);
123 static int linux_sa_put(struct osockaddr *osa);
124 static int linux_to_bsd_msg_flags(int);
125 static int bsd_to_linux_msg_flags(int);
126 
127 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
128 	AF_UNSPEC,
129 	AF_UNIX,
130 	AF_INET,
131 	AF_CCITT,	/* LINUX_AF_AX25 */
132 	AF_IPX,
133 	AF_APPLETALK,
134 	-1,		/* LINUX_AF_NETROM */
135 	-1,		/* LINUX_AF_BRIDGE */
136 	-1,		/* LINUX_AF_ATMPVC */
137 	AF_CCITT,	/* LINUX_AF_X25 */
138 	AF_INET6,
139 	-1,		/* LINUX_AF_ROSE */
140 	AF_DECnet,
141 	-1,		/* LINUX_AF_NETBEUI */
142 	-1,		/* LINUX_AF_SECURITY */
143 	pseudo_AF_KEY,
144 	AF_ROUTE,	/* LINUX_AF_NETLINK */
145 	-1,		/* LINUX_AF_PACKET */
146 	-1,		/* LINUX_AF_ASH */
147 	-1,		/* LINUX_AF_ECONET */
148 	-1,		/* LINUX_AF_ATMSVC */
149 	AF_SNA,
150 	/* rest up to LINUX_AF_MAX-1 is not allocated */
151 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
152 };
153 
154 static const int bsd_to_linux_domain_[AF_MAX] = {
155 	LINUX_AF_UNSPEC,
156 	LINUX_AF_UNIX,
157 	LINUX_AF_INET,
158 	-1,		/* AF_IMPLINK */
159 	-1,		/* AF_PUP */
160 	-1,		/* AF_CHAOS */
161 	-1,		/* AF_NS */
162 	-1,		/* AF_ISO */
163 	-1,		/* AF_ECMA */
164 	-1,		/* AF_DATAKIT */
165 	LINUX_AF_AX25,	/* AF_CCITT */
166 	LINUX_AF_SNA,
167 	LINUX_AF_DECnet,
168 	-1,		/* AF_DLI */
169 	-1,		/* AF_LAT */
170 	-1,		/* AF_HYLINK */
171 	LINUX_AF_APPLETALK,
172 	LINUX_AF_NETLINK,
173 	-1,		/* AF_LINK */
174 	-1,		/* AF_XTP */
175 	-1,		/* AF_COIP */
176 	-1,		/* AF_CNT */
177 	-1,		/* pseudo_AF_RTIP */
178 	LINUX_AF_IPX,
179 	LINUX_AF_INET6,
180 	-1,		/* pseudo_AF_PIP */
181 	-1,		/* AF_ISDN */
182 	-1,		/* AF_NATM */
183 	-1,		/* AF_ARP */
184 	LINUX_pseudo_AF_KEY,
185 	-1,		/* pseudo_AF_HDRCMPLT */
186 };
187 
188 static const struct {
189 	int bfl;
190 	int lfl;
191 } bsd_to_linux_msg_flags_[] = {
192 	{MSG_OOB,		LINUX_MSG_OOB},
193 	{MSG_PEEK,		LINUX_MSG_PEEK},
194 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
195 	{MSG_EOR,		LINUX_MSG_EOR},
196 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
197 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
198 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
199 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
200 	{MSG_BCAST,		0},		/* not supported, clear */
201 	{MSG_MCAST,		0},		/* not supported, clear */
202 	{-1, /* not supp */	LINUX_MSG_PROBE},
203 	{-1, /* not supp */	LINUX_MSG_FIN},
204 	{-1, /* not supp */	LINUX_MSG_SYN},
205 	{-1, /* not supp */	LINUX_MSG_CONFIRM},
206 	{-1, /* not supp */	LINUX_MSG_RST},
207 	{-1, /* not supp */	LINUX_MSG_ERRQUEUE},
208 	{-1, /* not supp */	LINUX_MSG_NOSIGNAL},
209 	{-1, /* not supp */	LINUX_MSG_MORE},
210 };
211 
212 /*
213  * Convert between Linux and BSD socket domain values
214  */
215 static int
216 linux_to_bsd_domain(int ldom)
217 {
218 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
219 		return (-1);
220 
221 	return linux_to_bsd_domain_[ldom];
222 }
223 
224 /*
225  * Convert between BSD and Linux socket domain values
226  */
227 static int
228 bsd_to_linux_domain(int bdom)
229 {
230 	if (bdom < 0 || bdom >= AF_MAX)
231 		return (-1);
232 
233 	return bsd_to_linux_domain_[bdom];
234 }
235 
236 static int
237 linux_to_bsd_msg_flags(int lflag)
238 {
239 	int i, lfl, bfl;
240 	int bflag = 0;
241 
242 	if (lflag == 0)
243 		return (0);
244 
245 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
246 		bfl = bsd_to_linux_msg_flags_[i].bfl;
247 		lfl = bsd_to_linux_msg_flags_[i].lfl;
248 
249 		if (lfl == 0)
250 			continue;
251 
252 		if (lflag & lfl) {
253 			if (bfl < 0)
254 				return (-1);
255 
256 			bflag |= bfl;
257 		}
258 	}
259 
260 	return (bflag);
261 }
262 
263 static int
264 bsd_to_linux_msg_flags(int bflag)
265 {
266 	int i, lfl, bfl;
267 	int lflag = 0;
268 
269 	if (bflag == 0)
270 		return (0);
271 
272 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
273 		bfl = bsd_to_linux_msg_flags_[i].bfl;
274 		lfl = bsd_to_linux_msg_flags_[i].lfl;
275 
276 		if (bfl <= 0)
277 			continue;
278 
279 		if (bflag & bfl) {
280 			if (lfl < 0)
281 				return (-1);
282 
283 			lflag |= lfl;
284 		}
285 	}
286 
287 	return (lflag);
288 }
289 
290 int
291 linux_sys_socket(struct lwp *l, const struct linux_sys_socket_args *uap, register_t *retval)
292 {
293 	/* {
294 		syscallarg(int)	domain;
295 		syscallarg(int)	type;
296 		syscallarg(int) protocol;
297 	} */
298 	struct sys___socket30_args bsa;
299 	int error;
300 
301 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
302 	SCARG(&bsa, type) = SCARG(uap, type);
303 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
304 	if (SCARG(&bsa, domain) == -1)
305 		return EINVAL;
306 	error = sys___socket30(l, &bsa, retval);
307 
308 #ifdef INET6
309 	/*
310 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
311 	 * default and some apps depend on this. So, set V6ONLY to 0
312 	 * for Linux apps if the sysctl value is set to 1.
313 	 */
314 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
315 		struct socket *so;
316 
317 		if (fd_getsock(*retval, &so) == 0) {
318 			struct mbuf *m;
319 
320 			m = m_get(M_WAIT, MT_SOOPTS);
321 			m->m_len = sizeof(int);
322 			*mtod(m, int *) = 0;
323 
324 			/* ignore error */
325 			(void) sosetopt(so, IPPROTO_IPV6, IPV6_V6ONLY, m);
326 
327 			fd_putfile(*retval);
328 		}
329 	}
330 #endif
331 
332 	return (error);
333 }
334 
335 int
336 linux_sys_socketpair(struct lwp *l, const struct linux_sys_socketpair_args *uap, register_t *retval)
337 {
338 	/* {
339 		syscallarg(int) domain;
340 		syscallarg(int) type;
341 		syscallarg(int) protocol;
342 		syscallarg(int *) rsv;
343 	} */
344 	struct sys_socketpair_args bsa;
345 
346 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
347 	if (SCARG(&bsa, domain) == -1)
348 		return EINVAL;
349 	SCARG(&bsa, type) = SCARG(uap, type);
350 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
351 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
352 
353 	return sys_socketpair(l, &bsa, retval);
354 }
355 
356 int
357 linux_sys_sendto(struct lwp *l, const struct linux_sys_sendto_args *uap, register_t *retval)
358 {
359 	/* {
360 		syscallarg(int)				s;
361 		syscallarg(void *)			msg;
362 		syscallarg(int)				len;
363 		syscallarg(int)				flags;
364 		syscallarg(struct osockaddr *)		to;
365 		syscallarg(int)				tolen;
366 	} */
367 	struct msghdr   msg;
368 	struct iovec    aiov;
369 	struct mbuf *nam;
370 	int bflags;
371 	int error;
372 
373 	/* Translate message flags.  */
374 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
375 	if (bflags < 0)
376 		/* Some supported flag */
377 		return EINVAL;
378 
379 	msg.msg_flags = 0;
380 	msg.msg_name = NULL;
381 	msg.msg_control = NULL;
382 
383 	if (SCARG(uap, tolen)) {
384 		/* Read in and convert the sockaddr */
385 		error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
386 		    SCARG(uap, tolen));
387 		if (error)
388 			return (error);
389 		msg.msg_flags |= MSG_NAMEMBUF;
390 		msg.msg_name = nam;
391 		msg.msg_namelen = SCARG(uap, tolen);
392 	}
393 
394 	msg.msg_iov = &aiov;
395 	msg.msg_iovlen = 1;
396 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
397 	aiov.iov_len = SCARG(uap, len);
398 
399 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
400 }
401 
402 int
403 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
404 {
405 	/* {
406 		syscallarg(int) s;
407 		syscallarg(struct msghdr *) msg;
408 		syscallarg(u_int) flags;
409 	} */
410 	struct msghdr	msg;
411 	int		error;
412 	int		bflags;
413 	struct mbuf     *nam;
414 	u_int8_t	*control;
415 	struct mbuf     *ctl_mbuf = NULL;
416 
417 	msg.msg_flags = MSG_IOVUSRSPACE;
418 
419 	/*
420 	 * Translate message flags.
421 	 */
422 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
423 	if (bflags < 0)
424 		/* Some supported flag */
425 		return EINVAL;
426 
427 	if (msg.msg_name) {
428 		/* Read in and convert the sockaddr */
429 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
430 		    msg.msg_namelen);
431 		if (error)
432 			return (error);
433 		msg.msg_flags |= MSG_NAMEMBUF;
434 		msg.msg_name = nam;
435 	}
436 
437 	/*
438 	 * Handle cmsg if there is any.
439 	 */
440 	if (CMSG_FIRSTHDR(&msg)) {
441 		struct linux_cmsghdr l_cmsg, *l_cc;
442 		struct cmsghdr *cmsg;
443 		ssize_t resid = msg.msg_controllen;
444 		size_t clen, cidx = 0, cspace;
445 
446 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
447 		clen = MLEN;
448 		control = mtod(ctl_mbuf, void *);
449 
450 		l_cc = LINUX_CMSG_FIRSTHDR(&msg);
451 		do {
452 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
453 			if (error)
454 				goto done;
455 
456 			/*
457 			 * Sanity check the control message length.
458 			 */
459 			if (l_cmsg.cmsg_len > resid
460 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
461 				error = EINVAL;
462 				goto done;
463 			}
464 
465 			/*
466 			 * Refuse unsupported control messages, and
467 			 * translate fields as appropriate.
468 			 */
469 			switch (l_cmsg.cmsg_level) {
470 			case LINUX_SOL_SOCKET:
471 				/* It only differs on some archs */
472 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
473 					l_cmsg.cmsg_level = SOL_SOCKET;
474 
475 				switch(l_cmsg.cmsg_type) {
476 				case LINUX_SCM_RIGHTS:
477 					/* Linux SCM_RIGHTS is same as NetBSD */
478 					break;
479 
480 				default:
481 					/* other types not supported */
482 					error = EINVAL;
483 					goto done;
484 				}
485 				break;
486 			default:
487 				/* pray and leave intact */
488 				break;
489 			}
490 
491 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
492 
493 			/* Check the buffer is big enough */
494 			if (__predict_false(cidx + cspace > clen)) {
495 				u_int8_t *nc;
496 
497 				clen = cidx + cspace;
498 				if (clen >= PAGE_SIZE) {
499 					error = EINVAL;
500 					goto done;
501 				}
502 				nc = realloc(clen <= MLEN ? NULL : control,
503 						clen, M_TEMP, M_WAITOK);
504 				if (!nc) {
505 					error = ENOMEM;
506 					goto done;
507 				}
508 				if (cidx <= MLEN)
509 					/* Old buffer was in mbuf... */
510 					memcpy(nc, control, cidx);
511 				control = nc;
512 			}
513 
514 			/* Copy header */
515 			cmsg = (void *)&control[cidx];
516 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
517 			cmsg->cmsg_level = l_cmsg.cmsg_level;
518 			cmsg->cmsg_type = l_cmsg.cmsg_type;
519 
520 			/* Zero are between header and data */
521 			memset(cmsg + 1, 0,
522 				CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg));
523 
524 			/* Copyin the data */
525 			error = copyin(LINUX_CMSG_DATA(l_cc),
526 				CMSG_DATA(control),
527 				l_cmsg.cmsg_len - sizeof(l_cmsg));
528 			if (error)
529 				goto done;
530 
531 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
532 			cidx += cspace;
533 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
534 
535 		/* If we allocated a buffer, attach to mbuf */
536 		if (cidx > MLEN) {
537 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
538 			ctl_mbuf->m_flags |= M_EXT_RW;
539 		}
540 		control = NULL;
541 		ctl_mbuf->m_len = cidx;
542 
543 		msg.msg_control = ctl_mbuf;
544 		msg.msg_flags |= MSG_CONTROLMBUF;
545 	}
546 
547 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
548 	/* Freed internally */
549 	ctl_mbuf = NULL;
550 
551 done:
552 	if (ctl_mbuf != NULL) {
553 		if (control != NULL && control != mtod(ctl_mbuf, void *))
554 			free(control, M_MBUF);
555 		m_free(ctl_mbuf);
556 	}
557 	return (error);
558 }
559 
560 int
561 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
562 {
563 	/* {
564 		syscallarg(int) s;
565 		syscallarg(void *) buf;
566 		syscallarg(int) len;
567 		syscallarg(int) flags;
568 		syscallarg(struct osockaddr *) from;
569 		syscallarg(int *) fromlenaddr;
570 	} */
571 	int		error;
572 	struct sys_recvfrom_args bra;
573 
574 	SCARG(&bra, s) = SCARG(uap, s);
575 	SCARG(&bra, buf) = SCARG(uap, buf);
576 	SCARG(&bra, len) = SCARG(uap, len);
577 	SCARG(&bra, flags) = SCARG(uap, flags);
578 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
579 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
580 
581 	if ((error = sys_recvfrom(l, &bra, retval)))
582 		return (error);
583 
584 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
585 		return (error);
586 
587 	return (0);
588 }
589 
590 static int
591 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
592 {
593 	int dlen, error = 0;
594 	struct cmsghdr *cmsg;
595 	struct linux_cmsghdr linux_cmsg;
596 	struct mbuf *m;
597 	char *q, *q_end;
598 
599 	if (mp->msg_controllen <= 0 || control == 0) {
600 		mp->msg_controllen = 0;
601 		free_control_mbuf(l, control, control);
602 		return 0;
603 	}
604 
605 	q = (char *)mp->msg_control;
606 	q_end = q + mp->msg_controllen;
607 
608 	for (m = control; m != NULL; ) {
609 		cmsg = mtod(m, struct cmsghdr *);
610 
611 		/*
612 		 * Fixup cmsg. We handle two things:
613 		 * 0. different sizeof cmsg_len.
614 		 * 1. different values for level/type on some archs
615 		 * 2. different alignment of CMSG_DATA on some archs
616 		 */
617 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
618 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
619 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
620 
621 		dlen = q_end - q;
622 		if (linux_cmsg.cmsg_len > dlen) {
623 			/* Not enough room for the parameter */
624 			dlen -= sizeof linux_cmsg;
625 			if (dlen <= 0)
626 				/* Discard if header wont fit */
627 				break;
628 			mp->msg_flags |= MSG_CTRUNC;
629 			if (linux_cmsg.cmsg_level == SOL_SOCKET
630 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
631 				/* Do not truncate me ... */
632 				break;
633 		} else
634 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
635 
636 		switch (linux_cmsg.cmsg_level) {
637 		case SOL_SOCKET:
638 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
639 			switch (linux_cmsg.cmsg_type) {
640 			case SCM_RIGHTS:
641 				/* Linux SCM_RIGHTS is same as NetBSD */
642 				break;
643 
644 			default:
645 				/* other types not supported */
646 				error = EINVAL;
647 				goto done;
648 			}
649 			/* machine dependant ! */
650 			break;
651 		default:
652 			/* pray and leave intact */
653 			break;
654 		}
655 
656 		/* There can be padding between the header and data... */
657 		error = copyout(&linux_cmsg, q, sizeof *cmsg);
658 		if (error != 0) {
659 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
660 			    dlen);
661 		}
662 		if (error != 0) {
663 			/* We must free all the SCM_RIGHTS */
664 			m = control;
665 			break;
666 		}
667 		m = m->m_next;
668 		if (m == NULL || q + LINUX_CMSG_ALIGN(dlen) > q_end) {
669 			q += dlen;
670 			break;
671 		}
672 		q += LINUX_CMSG_ALIGN(dlen);
673 	}
674 
675   done:
676 	free_control_mbuf(l, control, m);
677 
678 	mp->msg_controllen = q - (char *)mp->msg_control;
679 	return error;
680 }
681 
682 int
683 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
684 {
685 	/* {
686 		syscallarg(int) s;
687 		syscallarg(struct msghdr *) msg;
688 		syscallarg(u_int) flags;
689 	} */
690 	struct msghdr	msg;
691 	int		error;
692 	struct mbuf	*from, *control;
693 
694 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
695 	if (error)
696 		return (error);
697 
698 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
699 	if (msg.msg_flags < 0) {
700 		/* Some unsupported flag */
701 		return (EINVAL);
702 	}
703 	msg.msg_flags |= MSG_IOVUSRSPACE;
704 
705 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
706 	    msg.msg_control != NULL ? &control : NULL, retval);
707 	if (error != 0)
708 		return error;
709 
710 	if (msg.msg_control != NULL)
711 		error = linux_copyout_msg_control(l, &msg, control);
712 
713 	if (error == 0 && from != 0) {
714 		mtod(from, struct osockaddr *)->sa_family =
715 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
716 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
717 			from);
718 	} else
719 		msg.msg_namelen = 0;
720 
721 	if (from != NULL)
722 		m_free(from);
723 
724 	if (error == 0) {
725 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
726 		if (msg.msg_flags < 0)
727 			/* Some flag unsupported by Linux */
728 			error = EINVAL;
729 		else
730 			error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
731 	}
732 
733 	return (error);
734 }
735 
736 /*
737  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
738  * is different, the rest matches IPPROTO_* on both systems.
739  */
740 int
741 linux_to_bsd_sopt_level(int llevel)
742 {
743 
744 	switch (llevel) {
745 	case LINUX_SOL_SOCKET:
746 		return SOL_SOCKET;
747 	case LINUX_SOL_IP:
748 		return IPPROTO_IP;
749 	case LINUX_SOL_TCP:
750 		return IPPROTO_TCP;
751 	case LINUX_SOL_UDP:
752 		return IPPROTO_UDP;
753 	default:
754 		return -1;
755 	}
756 }
757 
758 /*
759  * Convert Linux socket level socket option numbers to NetBSD values.
760  */
761 int
762 linux_to_bsd_so_sockopt(int lopt)
763 {
764 
765 	switch (lopt) {
766 	case LINUX_SO_DEBUG:
767 		return SO_DEBUG;
768 	case LINUX_SO_REUSEADDR:
769 		/*
770 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
771 		 * host:port pair through SO_REUSEADDR even if the address is not a
772 		 * multicast-address.  Effectively, this means that we should use
773 		 * SO_REUSEPORT to allow Linux applications to not exit with
774 		 * EADDRINUSE
775 		 */
776 		return SO_REUSEPORT;
777 	case LINUX_SO_TYPE:
778 		return SO_TYPE;
779 	case LINUX_SO_ERROR:
780 		return SO_ERROR;
781 	case LINUX_SO_DONTROUTE:
782 		return SO_DONTROUTE;
783 	case LINUX_SO_BROADCAST:
784 		return SO_BROADCAST;
785 	case LINUX_SO_SNDBUF:
786 		return SO_SNDBUF;
787 	case LINUX_SO_RCVBUF:
788 		return SO_RCVBUF;
789 	case LINUX_SO_KEEPALIVE:
790 		return SO_KEEPALIVE;
791 	case LINUX_SO_OOBINLINE:
792 		return SO_OOBINLINE;
793 	case LINUX_SO_LINGER:
794 		return SO_LINGER;
795 	case LINUX_SO_PRIORITY:
796 	case LINUX_SO_NO_CHECK:
797 	default:
798 		return -1;
799 	}
800 }
801 
802 /*
803  * Convert Linux IP level socket option number to NetBSD values.
804  */
805 int
806 linux_to_bsd_ip_sockopt(int lopt)
807 {
808 
809 	switch (lopt) {
810 	case LINUX_IP_TOS:
811 		return IP_TOS;
812 	case LINUX_IP_TTL:
813 		return IP_TTL;
814 	case LINUX_IP_MULTICAST_TTL:
815 		return IP_MULTICAST_TTL;
816 	case LINUX_IP_MULTICAST_LOOP:
817 		return IP_MULTICAST_LOOP;
818 	case LINUX_IP_MULTICAST_IF:
819 		return IP_MULTICAST_IF;
820 	case LINUX_IP_ADD_MEMBERSHIP:
821 		return IP_ADD_MEMBERSHIP;
822 	case LINUX_IP_DROP_MEMBERSHIP:
823 		return IP_DROP_MEMBERSHIP;
824 	default:
825 		return -1;
826 	}
827 }
828 
829 /*
830  * Convert Linux TCP level socket option number to NetBSD values.
831  */
832 int
833 linux_to_bsd_tcp_sockopt(int lopt)
834 {
835 
836 	switch (lopt) {
837 	case LINUX_TCP_NODELAY:
838 		return TCP_NODELAY;
839 	case LINUX_TCP_MAXSEG:
840 		return TCP_MAXSEG;
841 	default:
842 		return -1;
843 	}
844 }
845 
846 /*
847  * Convert Linux UDP level socket option number to NetBSD values.
848  */
849 int
850 linux_to_bsd_udp_sockopt(int lopt)
851 {
852 
853 	switch (lopt) {
854 	default:
855 		return -1;
856 	}
857 }
858 
859 /*
860  * Another reasonably straightforward function: setsockopt(2).
861  * The level and option numbers are converted; the values passed
862  * are not (yet) converted, the ones currently implemented don't
863  * need conversion, as they are the same on both systems.
864  */
865 int
866 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
867 {
868 	/* {
869 		syscallarg(int) s;
870 		syscallarg(int) level;
871 		syscallarg(int) optname;
872 		syscallarg(void *) optval;
873 		syscallarg(int) optlen;
874 	} */
875 	struct sys_setsockopt_args bsa;
876 	int name;
877 
878 	SCARG(&bsa, s) = SCARG(uap, s);
879 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
880 	SCARG(&bsa, val) = SCARG(uap, optval);
881 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
882 
883 	/*
884 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
885 	 * and returns EOPNOTSUPP for other levels
886 	 */
887 	if (SCARG(&bsa, level) != SOL_SOCKET) {
888 		struct socket *so;
889 		int error, family;
890 
891 		/* fd_getsock() will use the descriptor for us */
892 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
893 		    	return error;
894 		family = so->so_proto->pr_domain->dom_family;
895 		fd_putfile(SCARG(&bsa, s));
896 
897 		if (family == AF_LOCAL)
898 			return EOPNOTSUPP;
899 	}
900 
901 	switch (SCARG(&bsa, level)) {
902 	case SOL_SOCKET:
903 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
904 		break;
905 	case IPPROTO_IP:
906 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
907 		break;
908 	case IPPROTO_TCP:
909 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
910 		break;
911 	case IPPROTO_UDP:
912 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
913 		break;
914 	default:
915 		return EINVAL;
916 	}
917 
918 	if (name == -1)
919 		return EINVAL;
920 	SCARG(&bsa, name) = name;
921 
922 	return sys_setsockopt(l, &bsa, retval);
923 }
924 
925 /*
926  * getsockopt(2) is very much the same as setsockopt(2) (see above)
927  */
928 int
929 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
930 {
931 	/* {
932 		syscallarg(int) s;
933 		syscallarg(int) level;
934 		syscallarg(int) optname;
935 		syscallarg(void *) optval;
936 		syscallarg(int *) optlen;
937 	} */
938 	struct sys_getsockopt_args bga;
939 	int name;
940 
941 	SCARG(&bga, s) = SCARG(uap, s);
942 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
943 	SCARG(&bga, val) = SCARG(uap, optval);
944 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
945 
946 	switch (SCARG(&bga, level)) {
947 	case SOL_SOCKET:
948 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
949 		break;
950 	case IPPROTO_IP:
951 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
952 		break;
953 	case IPPROTO_TCP:
954 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
955 		break;
956 	case IPPROTO_UDP:
957 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
958 		break;
959 	default:
960 		return EINVAL;
961 	}
962 
963 	if (name == -1)
964 		return EINVAL;
965 	SCARG(&bga, name) = name;
966 
967 	return sys_getsockopt(l, &bga, retval);
968 }
969 
970 int
971 linux_getifconf(struct lwp *l, register_t *retval, void *data)
972 {
973 	struct linux_ifreq ifr, *ifrp;
974 	struct ifconf *ifc = data;
975 	struct ifnet *ifp;
976 	struct ifaddr *ifa;
977 	struct sockaddr *sa;
978 	struct osockaddr *osa;
979 	int space, error = 0;
980 	const int sz = (int)sizeof(ifr);
981 
982 	ifrp = (struct linux_ifreq *)ifc->ifc_req;
983 	if (ifrp == NULL)
984 		space = 0;
985 	else
986 		space = ifc->ifc_len;
987 
988 	IFNET_FOREACH(ifp) {
989 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
990 		    sizeof(ifr.ifr_name));
991 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
992 			return ENAMETOOLONG;
993 		if (IFADDR_EMPTY(ifp))
994 			continue;
995 		IFADDR_FOREACH(ifa, ifp) {
996 			sa = ifa->ifa_addr;
997 			if (sa->sa_family != AF_INET ||
998 			    sa->sa_len > sizeof(*osa))
999 				continue;
1000 			memcpy(&ifr.ifr_addr, sa, sa->sa_len);
1001 			osa = (struct osockaddr *)&ifr.ifr_addr;
1002 			osa->sa_family = sa->sa_family;
1003 			if (space >= sz) {
1004 				error = copyout(&ifr, ifrp, sz);
1005 				if (error != 0)
1006 					return error;
1007 				ifrp++;
1008 			}
1009 			space -= sz;
1010 		}
1011 	}
1012 
1013 	if (ifrp != NULL)
1014 		ifc->ifc_len -= space;
1015 	else
1016 		ifc->ifc_len = -space;
1017 
1018 	return 0;
1019 }
1020 
1021 int
1022 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1023     void *data)
1024 {
1025 	/* Not the full structure, just enough to map what we do here */
1026 	struct linux_ifreq lreq;
1027 	file_t *fp;
1028 	struct ifaddr *ifa;
1029 	struct ifnet *ifp;
1030 	struct sockaddr_dl *sadl;
1031 	int error, found;
1032 	int index, ifnum;
1033 
1034 	/*
1035 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1036 	 * SIOCGIFCONF, because the user buffer is not of the right
1037 	 * type to take those results.  We can't use kernel buffers to
1038 	 * receive the results, as the implementation of sys_ioctl()
1039 	 * and ifconf() [which implements SIOCGIFCONF] use
1040 	 * copyin()/copyout() which will fail on kernel addresses.
1041 	 *
1042 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1043 	 */
1044 
1045 	if ((fp = fd_getfile(fd)) == NULL)
1046 		return (EBADF);
1047 
1048 	KERNEL_LOCK(1, NULL);
1049 
1050 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1051 		error = EBADF;
1052 		goto out;
1053 	}
1054 
1055 	error = copyin(data, &lreq, sizeof(lreq));
1056 	if (error)
1057 		goto out;
1058 	lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0';		/* just in case */
1059 
1060 	/*
1061 	 * Try real interface name first, then fake "ethX"
1062 	 */
1063 	found = 0;
1064 	IFNET_FOREACH(ifp) {
1065 		if (found)
1066 			break;
1067 		if (strcmp(lreq.ifr_name, ifp->if_xname))
1068 			/* not this interface */
1069 			continue;
1070 		found=1;
1071 		if (IFADDR_EMPTY(ifp)) {
1072 			error = ENODEV;
1073 			goto out;
1074 		}
1075 		IFADDR_FOREACH(ifa, ifp) {
1076 			sadl = satosdl(ifa->ifa_addr);
1077 			/* only return ethernet addresses */
1078 			/* XXX what about FDDI, etc. ? */
1079 			if (sadl->sdl_family != AF_LINK ||
1080 			    sadl->sdl_type != IFT_ETHER)
1081 				continue;
1082 			memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl),
1083 			       MIN(sadl->sdl_alen,
1084 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1085 			lreq.ifr_hwaddr.sa_family =
1086 				sadl->sdl_family;
1087 			error = copyout(&lreq, data, sizeof(lreq));
1088 			goto out;
1089 		}
1090 	}
1091 
1092 	if (strncmp(lreq.ifr_name, "eth", 3) == 0) {
1093 		for (ifnum = 0, index = 3;
1094 		     lreq.ifr_name[index] != '\0' && index < LINUX_IFNAMSIZ;
1095 		     index++) {
1096 			ifnum *= 10;
1097 			ifnum += lreq.ifr_name[index] - '0';
1098 		}
1099 
1100 		error = EINVAL;			/* in case we don't find one */
1101 		found = 0;
1102 		IFNET_FOREACH(ifp) {
1103 			if (found)
1104 				break;
1105 			memcpy(lreq.ifr_name, ifp->if_xname,
1106 			       MIN(LINUX_IFNAMSIZ, IFNAMSIZ));
1107 			IFADDR_FOREACH(ifa, ifp) {
1108 				sadl = satosdl(ifa->ifa_addr);
1109 				/* only return ethernet addresses */
1110 				/* XXX what about FDDI, etc. ? */
1111 				if (sadl->sdl_family != AF_LINK ||
1112 				    sadl->sdl_type != IFT_ETHER)
1113 					continue;
1114 				if (ifnum--)
1115 					/* not the reqested iface */
1116 					continue;
1117 				memcpy(&lreq.ifr_hwaddr.sa_data,
1118 				       CLLADDR(sadl),
1119 				       MIN(sadl->sdl_alen,
1120 					   sizeof(lreq.ifr_hwaddr.sa_data)));
1121 				lreq.ifr_hwaddr.sa_family =
1122 					sadl->sdl_family;
1123 				error = copyout(&lreq, data, sizeof(lreq));
1124 				found = 1;
1125 				break;
1126 			}
1127 		}
1128 	} else {
1129 		/* unknown interface, not even an "eth*" name */
1130 		error = ENODEV;
1131 	}
1132 
1133 out:
1134 	KERNEL_UNLOCK_ONE(NULL);
1135 	fd_putfile(fd);
1136 	return error;
1137 }
1138 
1139 int
1140 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1141 {
1142 	/* {
1143 		syscallarg(int) fd;
1144 		syscallarg(u_long) com;
1145 		syscallarg(void *) data;
1146 	} */
1147 	u_long com;
1148 	int error = 0, isdev = 0, dosys = 1;
1149 	struct sys_ioctl_args ia;
1150 	file_t *fp;
1151 	struct vnode *vp;
1152 	int (*ioctlf)(file_t *, u_long, void *);
1153 	struct ioctl_pt pt;
1154 
1155 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1156 		return (EBADF);
1157 
1158 	if (fp->f_type == DTYPE_VNODE) {
1159 		vp = (struct vnode *)fp->f_data;
1160 		isdev = vp->v_type == VCHR;
1161 	}
1162 
1163 	/*
1164 	 * Don't try to interpret socket ioctl calls that are done
1165 	 * on a device filedescriptor, just pass them through, to
1166 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1167 	 * device will only handle these if it's prepared to do
1168 	 * so, to avoid unexpected things from happening.
1169 	 */
1170 	if (isdev) {
1171 		dosys = 0;
1172 		ioctlf = fp->f_ops->fo_ioctl;
1173 		pt.com = SCARG(uap, com);
1174 		pt.data = SCARG(uap, data);
1175 		error = ioctlf(fp, PTIOCLINUX, &pt);
1176 		/*
1177 		 * XXX hack: if the function returns EJUSTRETURN,
1178 		 * it has stuffed a sysctl return value in pt.data.
1179 		 */
1180 		if (error == EJUSTRETURN) {
1181 			retval[0] = (register_t)pt.data;
1182 			error = 0;
1183 		}
1184 		goto out;
1185 	}
1186 
1187 	com = SCARG(uap, com);
1188 	retval[0] = 0;
1189 
1190 	switch (com) {
1191 	case LINUX_SIOCGIFCONF:
1192 		error = linux_getifconf(l, retval, SCARG(uap, data));
1193 		dosys = 0;
1194 		break;
1195 	case LINUX_SIOCGIFFLAGS:
1196 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1197 		break;
1198 	case LINUX_SIOCSIFFLAGS:
1199 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1200 		break;
1201 	case LINUX_SIOCGIFADDR:
1202 		SCARG(&ia, com) = OOSIOCGIFADDR;
1203 		break;
1204 	case LINUX_SIOCGIFDSTADDR:
1205 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1206 		break;
1207 	case LINUX_SIOCGIFBRDADDR:
1208 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1209 		break;
1210 	case LINUX_SIOCGIFNETMASK:
1211 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1212 		break;
1213 	case LINUX_SIOCADDMULTI:
1214 		SCARG(&ia, com) = OSIOCADDMULTI;
1215 		break;
1216 	case LINUX_SIOCDELMULTI:
1217 		SCARG(&ia, com) = OSIOCDELMULTI;
1218 		break;
1219 	case LINUX_SIOCGIFHWADDR:
1220 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1221 		    SCARG(uap, data));
1222 		dosys = 0;
1223 		break;
1224 	default:
1225 		error = EINVAL;
1226 	}
1227 
1228  out:
1229  	fd_putfile(SCARG(uap, fd));
1230 
1231 	if (error ==0 && dosys) {
1232 		SCARG(&ia, fd) = SCARG(uap, fd);
1233 		SCARG(&ia, data) = SCARG(uap, data);
1234 		error = sys_ioctl(curlwp, &ia, retval);
1235 	}
1236 
1237 	return error;
1238 }
1239 
1240 int
1241 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1242 {
1243 	/* {
1244 		syscallarg(int) s;
1245 		syscallarg(const struct sockaddr *) name;
1246 		syscallarg(int) namelen;
1247 	} */
1248 	int		error;
1249 	struct mbuf *nam;
1250 
1251 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1252 	    SCARG(uap, namelen));
1253 	if (error)
1254 		return (error);
1255 
1256 	error = do_sys_connect(l, SCARG(uap, s), nam);
1257 
1258 	if (error == EISCONN) {
1259 		struct socket *so;
1260 		int state, prflags, nbio;
1261 
1262 		/* fd_getsock() will use the descriptor for us */
1263 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1264 		    	return EISCONN;
1265 
1266 		solock(so);
1267 		state = so->so_state;
1268 		nbio = so->so_nbio;
1269 		prflags = so->so_proto->pr_flags;
1270 		sounlock(so);
1271 		fd_putfile(SCARG(uap, s));
1272 		/*
1273 		 * We should only let this call succeed once per
1274 		 * non-blocking connect; however we don't have
1275 		 * a convenient place to keep that state..
1276 		 */
1277 		if (nbio && (state & SS_ISCONNECTED) &&
1278 		    (prflags & PR_CONNREQUIRED))
1279 			return 0;
1280 	}
1281 
1282 	return (error);
1283 }
1284 
1285 int
1286 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1287 {
1288 	/* {
1289 		syscallarg(int) s;
1290 		syscallarg(const struct osockaddr *) name;
1291 		syscallarg(int) namelen;
1292 	} */
1293 	int		error;
1294 	struct mbuf     *nam;
1295 
1296 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1297 	    SCARG(uap, namelen));
1298 	if (error)
1299 		return (error);
1300 
1301 	return do_sys_bind(l, SCARG(uap, s), nam);
1302 }
1303 
1304 int
1305 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1306 {
1307 	/* {
1308 		syscallarg(int) fdes;
1309 		syscallarg(void *) asa;
1310 		syscallarg(int *) alen;
1311 	} */
1312 	int error;
1313 
1314 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1315 		return (error);
1316 
1317 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1318 		return (error);
1319 
1320 	return (0);
1321 }
1322 
1323 int
1324 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1325 {
1326 	/* {
1327 		syscallarg(int) fdes;
1328 		syscallarg(void *) asa;
1329 		syscallarg(int *) alen;
1330 	} */
1331 	int error;
1332 
1333 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1334 		return (error);
1335 
1336 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1337 		return (error);
1338 
1339 	return (0);
1340 }
1341 
1342 /*
1343  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1344  * family and convert to sockaddr.
1345  */
1346 static int
1347 linux_get_sa(struct lwp *l, int s, struct mbuf **mp,
1348     const struct osockaddr *osa, unsigned int salen)
1349 {
1350 	int error, bdom;
1351 	struct sockaddr *sa;
1352 	struct osockaddr *kosa;
1353 	struct mbuf *m;
1354 
1355 	if (salen == 1 || salen > UCHAR_MAX) {
1356 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1357 		return EINVAL;
1358 	}
1359 
1360 	/* We'll need the address in an mbuf later, so copy into one here */
1361 	m = m_get(M_WAIT, MT_SONAME);
1362 	if (salen > MLEN)
1363 		MEXTMALLOC(m, salen, M_WAITOK);
1364 
1365 	m->m_len = salen;
1366 
1367 	if (salen == 0) {
1368 		*mp = m;
1369 		return 0;
1370 	}
1371 
1372 	kosa = mtod(m, void *);
1373 	if ((error = copyin(osa, kosa, salen))) {
1374 		DPRINTF(("error %d copying osa %p len %d\n",
1375 				error, osa, salen));
1376 		goto bad;
1377 	}
1378 
1379 	ktrkuser("linux sockaddr", kosa, salen);
1380 
1381 	bdom = linux_to_bsd_domain(kosa->sa_family);
1382 	if (bdom == -1) {
1383 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1384 		error = EINVAL;
1385 		goto bad;
1386 	}
1387 
1388 	/*
1389 	 * If the family is unspecified, use address family of the socket.
1390 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1391 	 */
1392 	if (bdom == AF_UNSPEC) {
1393 		struct socket *so;
1394 
1395 		/* fd_getsock() will use the descriptor for us */
1396 		if ((error = fd_getsock(s, &so)) != 0)
1397 			goto bad;
1398 
1399 		bdom = so->so_proto->pr_domain->dom_family;
1400 		fd_putfile(s);
1401 
1402 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1403 	}
1404 
1405 #ifdef INET6
1406 	/*
1407 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1408 	 * which lacks the scope id compared with RFC2553 one. If we detect
1409 	 * the situation, reject the address and write a message to system log.
1410 	 *
1411 	 * Still accept addresses for which the scope id is not used.
1412 	 */
1413 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1414 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1415 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1416 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1417 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1418 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1419 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1420 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1421 			struct proc *p = l->l_proc;
1422 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1423 
1424 			log(LOG_DEBUG,
1425 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1426 			    "sockaddr_in6 rejected",
1427 			    p->p_pid, p->p_comm, uid);
1428 			error = EINVAL;
1429 			goto bad;
1430 		}
1431 		salen = sizeof (struct sockaddr_in6);
1432 		sin6->sin6_scope_id = 0;
1433 	}
1434 #endif
1435 
1436 	if (bdom == AF_INET)
1437 		salen = sizeof(struct sockaddr_in);
1438 
1439 	sa = (struct sockaddr *) kosa;
1440 	sa->sa_family = bdom;
1441 	sa->sa_len = salen;
1442 	m->m_len = salen;
1443 	ktrkuser("new sockaddr", kosa, salen);
1444 
1445 #ifdef DEBUG_LINUX
1446 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1447 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1448 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1449 	DPRINTF(("\n"));
1450 #endif
1451 
1452 	*mp = m;
1453 	return 0;
1454 
1455     bad:
1456 	m_free(m);
1457 	return error;
1458 }
1459 
1460 static int
1461 linux_sa_put(struct osockaddr *osa)
1462 {
1463 	struct sockaddr sa;
1464 	struct osockaddr *kosa;
1465 	int error, bdom, len;
1466 
1467 	/*
1468 	 * Only read/write the sockaddr family and length part, the rest is
1469 	 * not changed.
1470 	 */
1471 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1472 
1473 	error = copyin(osa, &sa, len);
1474 	if (error)
1475 		return (error);
1476 
1477 	bdom = bsd_to_linux_domain(sa.sa_family);
1478 	if (bdom == -1)
1479 		return (EINVAL);
1480 
1481 	/* Note: we convert from sockaddr to osockaddr here, too */
1482 	kosa = (struct osockaddr *) &sa;
1483 	kosa->sa_family = bdom;
1484 	error = copyout(kosa, osa, len);
1485 	if (error)
1486 		return (error);
1487 
1488 	return (0);
1489 }
1490 
1491 #ifndef __amd64__
1492 int
1493 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1494 {
1495 	/* {
1496 		syscallarg(int) s;
1497 		syscallarg(void *) buf;
1498 		syscallarg(int) len;
1499 		syscallarg(int) flags;
1500 	} */
1501 	struct sys_recvfrom_args bra;
1502 
1503 
1504 	SCARG(&bra, s) = SCARG(uap, s);
1505 	SCARG(&bra, buf) = SCARG(uap, buf);
1506 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1507 	SCARG(&bra, flags) = SCARG(uap, flags);
1508 	SCARG(&bra, from) = NULL;
1509 	SCARG(&bra, fromlenaddr) = NULL;
1510 
1511 	return (sys_recvfrom(l, &bra, retval));
1512 }
1513 
1514 int
1515 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1516 {
1517 	/* {
1518 		syscallarg(int) s;
1519 		syscallarg(void *) buf;
1520 		syscallarg(int) len;
1521 		syscallarg(int) flags;
1522 	} */
1523 	struct sys_sendto_args bsa;
1524 
1525 	SCARG(&bsa, s)		= SCARG(uap, s);
1526 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1527 	SCARG(&bsa, len)	= SCARG(uap, len);
1528 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1529 	SCARG(&bsa, to)		= NULL;
1530 	SCARG(&bsa, tolen)	= 0;
1531 
1532 	return (sys_sendto(l, &bsa, retval));
1533 }
1534 #endif
1535 
1536 int
1537 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1538 {
1539 	/* {
1540 		syscallarg(int) s;
1541 		syscallarg(struct osockaddr *) name;
1542 		syscallarg(int *) anamelen;
1543 	} */
1544 	int error;
1545 	struct sys_accept_args baa;
1546 
1547 	SCARG(&baa, s)		= SCARG(uap, s);
1548 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1549 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1550 
1551 	if ((error = sys_accept(l, &baa, retval)))
1552 		return (error);
1553 
1554 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1555 		return (error);
1556 
1557 	return (0);
1558 }
1559