xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision fff57c5525bbe431aee7bdb3983954f0627a42cb)
1 /*	$NetBSD: linux_socket.c,v 1.95 2008/06/11 13:02:00 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 1.95 2008/06/11 13:02:00 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_getifhwaddr(struct lwp *, register_t *, u_int, void *);
120 static int linux_get_sa(struct lwp *, int, struct mbuf **,
121 		const struct osockaddr *, unsigned int);
122 static int linux_sa_put(struct osockaddr *osa);
123 static int linux_to_bsd_msg_flags(int);
124 static int bsd_to_linux_msg_flags(int);
125 
126 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
127 	AF_UNSPEC,
128 	AF_UNIX,
129 	AF_INET,
130 	AF_CCITT,	/* LINUX_AF_AX25 */
131 	AF_IPX,
132 	AF_APPLETALK,
133 	-1,		/* LINUX_AF_NETROM */
134 	-1,		/* LINUX_AF_BRIDGE */
135 	-1,		/* LINUX_AF_ATMPVC */
136 	AF_CCITT,	/* LINUX_AF_X25 */
137 	AF_INET6,
138 	-1,		/* LINUX_AF_ROSE */
139 	AF_DECnet,
140 	-1,		/* LINUX_AF_NETBEUI */
141 	-1,		/* LINUX_AF_SECURITY */
142 	pseudo_AF_KEY,
143 	AF_ROUTE,	/* LINUX_AF_NETLINK */
144 	-1,		/* LINUX_AF_PACKET */
145 	-1,		/* LINUX_AF_ASH */
146 	-1,		/* LINUX_AF_ECONET */
147 	-1,		/* LINUX_AF_ATMSVC */
148 	AF_SNA,
149 	/* rest up to LINUX_AF_MAX-1 is not allocated */
150 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
151 };
152 
153 static const int bsd_to_linux_domain_[AF_MAX] = {
154 	LINUX_AF_UNSPEC,
155 	LINUX_AF_UNIX,
156 	LINUX_AF_INET,
157 	-1,		/* AF_IMPLINK */
158 	-1,		/* AF_PUP */
159 	-1,		/* AF_CHAOS */
160 	-1,		/* AF_NS */
161 	-1,		/* AF_ISO */
162 	-1,		/* AF_ECMA */
163 	-1,		/* AF_DATAKIT */
164 	LINUX_AF_AX25,	/* AF_CCITT */
165 	LINUX_AF_SNA,
166 	LINUX_AF_DECnet,
167 	-1,		/* AF_DLI */
168 	-1,		/* AF_LAT */
169 	-1,		/* AF_HYLINK */
170 	LINUX_AF_APPLETALK,
171 	LINUX_AF_NETLINK,
172 	-1,		/* AF_LINK */
173 	-1,		/* AF_XTP */
174 	-1,		/* AF_COIP */
175 	-1,		/* AF_CNT */
176 	-1,		/* pseudo_AF_RTIP */
177 	LINUX_AF_IPX,
178 	LINUX_AF_INET6,
179 	-1,		/* pseudo_AF_PIP */
180 	-1,		/* AF_ISDN */
181 	-1,		/* AF_NATM */
182 	-1,		/* AF_ARP */
183 	LINUX_pseudo_AF_KEY,
184 	-1,		/* pseudo_AF_HDRCMPLT */
185 };
186 
187 static const struct {
188 	int bfl;
189 	int lfl;
190 } bsd_to_linux_msg_flags_[] = {
191 	{MSG_OOB,		LINUX_MSG_OOB},
192 	{MSG_PEEK,		LINUX_MSG_PEEK},
193 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
194 	{MSG_EOR,		LINUX_MSG_EOR},
195 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
196 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
197 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
198 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
199 	{MSG_BCAST,		0},		/* not supported, clear */
200 	{MSG_MCAST,		0},		/* not supported, clear */
201 	{-1, /* not supp */	LINUX_MSG_PROBE},
202 	{-1, /* not supp */	LINUX_MSG_FIN},
203 	{-1, /* not supp */	LINUX_MSG_SYN},
204 	{-1, /* not supp */	LINUX_MSG_CONFIRM},
205 	{-1, /* not supp */	LINUX_MSG_RST},
206 	{-1, /* not supp */	LINUX_MSG_ERRQUEUE},
207 	{-1, /* not supp */	LINUX_MSG_NOSIGNAL},
208 	{-1, /* not supp */	LINUX_MSG_MORE},
209 };
210 
211 /*
212  * Convert between Linux and BSD socket domain values
213  */
214 static int
215 linux_to_bsd_domain(int ldom)
216 {
217 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
218 		return (-1);
219 
220 	return linux_to_bsd_domain_[ldom];
221 }
222 
223 /*
224  * Convert between BSD and Linux socket domain values
225  */
226 static int
227 bsd_to_linux_domain(int bdom)
228 {
229 	if (bdom < 0 || bdom >= AF_MAX)
230 		return (-1);
231 
232 	return bsd_to_linux_domain_[bdom];
233 }
234 
235 static int
236 linux_to_bsd_msg_flags(int lflag)
237 {
238 	int i, lfl, bfl;
239 	int bflag = 0;
240 
241 	if (lflag == 0)
242 		return (0);
243 
244 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
245 		bfl = bsd_to_linux_msg_flags_[i].bfl;
246 		lfl = bsd_to_linux_msg_flags_[i].lfl;
247 
248 		if (lfl == 0)
249 			continue;
250 
251 		if (lflag & lfl) {
252 			if (bfl < 0)
253 				return (-1);
254 
255 			bflag |= bfl;
256 		}
257 	}
258 
259 	return (bflag);
260 }
261 
262 static int
263 bsd_to_linux_msg_flags(int bflag)
264 {
265 	int i, lfl, bfl;
266 	int lflag = 0;
267 
268 	if (bflag == 0)
269 		return (0);
270 
271 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
272 		bfl = bsd_to_linux_msg_flags_[i].bfl;
273 		lfl = bsd_to_linux_msg_flags_[i].lfl;
274 
275 		if (bfl <= 0)
276 			continue;
277 
278 		if (bflag & bfl) {
279 			if (lfl < 0)
280 				return (-1);
281 
282 			lflag |= lfl;
283 		}
284 	}
285 
286 	return (lflag);
287 }
288 
289 int
290 linux_sys_socket(struct lwp *l, const struct linux_sys_socket_args *uap, register_t *retval)
291 {
292 	/* {
293 		syscallarg(int)	domain;
294 		syscallarg(int)	type;
295 		syscallarg(int) protocol;
296 	} */
297 	struct sys___socket30_args bsa;
298 	int error;
299 
300 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
301 	SCARG(&bsa, type) = SCARG(uap, type);
302 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
303 	if (SCARG(&bsa, domain) == -1)
304 		return EINVAL;
305 	error = sys___socket30(l, &bsa, retval);
306 
307 #ifdef INET6
308 	/*
309 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
310 	 * default and some apps depend on this. So, set V6ONLY to 0
311 	 * for Linux apps if the sysctl value is set to 1.
312 	 */
313 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
314 		struct socket *so;
315 
316 		if (fd_getsock(*retval, &so) == 0) {
317 			struct mbuf *m;
318 
319 			m = m_get(M_WAIT, MT_SOOPTS);
320 			m->m_len = sizeof(int);
321 			*mtod(m, int *) = 0;
322 
323 			/* ignore error */
324 			(void) sosetopt(so, IPPROTO_IPV6, IPV6_V6ONLY, m);
325 
326 			fd_putfile(*retval);
327 		}
328 	}
329 #endif
330 
331 	return (error);
332 }
333 
334 int
335 linux_sys_socketpair(struct lwp *l, const struct linux_sys_socketpair_args *uap, register_t *retval)
336 {
337 	/* {
338 		syscallarg(int) domain;
339 		syscallarg(int) type;
340 		syscallarg(int) protocol;
341 		syscallarg(int *) rsv;
342 	} */
343 	struct sys_socketpair_args bsa;
344 
345 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
346 	if (SCARG(&bsa, domain) == -1)
347 		return EINVAL;
348 	SCARG(&bsa, type) = SCARG(uap, type);
349 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
350 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
351 
352 	return sys_socketpair(l, &bsa, retval);
353 }
354 
355 int
356 linux_sys_sendto(struct lwp *l, const struct linux_sys_sendto_args *uap, register_t *retval)
357 {
358 	/* {
359 		syscallarg(int)				s;
360 		syscallarg(void *)			msg;
361 		syscallarg(int)				len;
362 		syscallarg(int)				flags;
363 		syscallarg(struct osockaddr *)		to;
364 		syscallarg(int)				tolen;
365 	} */
366 	struct msghdr   msg;
367 	struct iovec    aiov;
368 	struct mbuf *nam;
369 	int bflags;
370 	int error;
371 
372 	/* Translate message flags.  */
373 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
374 	if (bflags < 0)
375 		/* Some supported flag */
376 		return EINVAL;
377 
378 	msg.msg_flags = 0;
379 	msg.msg_name = NULL;
380 	msg.msg_control = NULL;
381 
382 	if (SCARG(uap, tolen)) {
383 		/* Read in and convert the sockaddr */
384 		error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
385 		    SCARG(uap, tolen));
386 		if (error)
387 			return (error);
388 		msg.msg_flags |= MSG_NAMEMBUF;
389 		msg.msg_name = nam;
390 		msg.msg_namelen = SCARG(uap, tolen);
391 	}
392 
393 	msg.msg_iov = &aiov;
394 	msg.msg_iovlen = 1;
395 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
396 	aiov.iov_len = SCARG(uap, len);
397 
398 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
399 }
400 
401 int
402 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
403 {
404 	/* {
405 		syscallarg(int) s;
406 		syscallarg(struct msghdr *) msg;
407 		syscallarg(u_int) flags;
408 	} */
409 	struct msghdr	msg;
410 	int		error;
411 	int		bflags;
412 	struct mbuf     *nam;
413 	u_int8_t	*control;
414 	struct mbuf     *ctl_mbuf = NULL;
415 
416 	msg.msg_flags = MSG_IOVUSRSPACE;
417 
418 	/*
419 	 * Translate message flags.
420 	 */
421 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
422 	if (bflags < 0)
423 		/* Some supported flag */
424 		return EINVAL;
425 
426 	if (msg.msg_name) {
427 		/* Read in and convert the sockaddr */
428 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
429 		    msg.msg_namelen);
430 		if (error)
431 			return (error);
432 		msg.msg_flags |= MSG_NAMEMBUF;
433 		msg.msg_name = nam;
434 	}
435 
436 	/*
437 	 * Handle cmsg if there is any.
438 	 */
439 	if (CMSG_FIRSTHDR(&msg)) {
440 		struct linux_cmsghdr l_cmsg, *l_cc;
441 		struct cmsghdr *cmsg;
442 		ssize_t resid = msg.msg_controllen;
443 		size_t clen, cidx = 0, cspace;
444 
445 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
446 		clen = MLEN;
447 		control = mtod(ctl_mbuf, void *);
448 
449 		l_cc = LINUX_CMSG_FIRSTHDR(&msg);
450 		do {
451 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
452 			if (error)
453 				goto done;
454 
455 			/*
456 			 * Sanity check the control message length.
457 			 */
458 			if (l_cmsg.cmsg_len > resid
459 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
460 				error = EINVAL;
461 				goto done;
462 			}
463 
464 			/*
465 			 * Refuse unsupported control messages, and
466 			 * translate fields as appropriate.
467 			 */
468 			switch (l_cmsg.cmsg_level) {
469 			case LINUX_SOL_SOCKET:
470 				/* It only differs on some archs */
471 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
472 					l_cmsg.cmsg_level = SOL_SOCKET;
473 
474 				switch(l_cmsg.cmsg_type) {
475 				case LINUX_SCM_RIGHTS:
476 					/* Linux SCM_RIGHTS is same as NetBSD */
477 					break;
478 
479 				default:
480 					/* other types not supported */
481 					error = EINVAL;
482 					goto done;
483 				}
484 				break;
485 			default:
486 				/* pray and leave intact */
487 				break;
488 			}
489 
490 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
491 
492 			/* Check the buffer is big enough */
493 			if (__predict_false(cidx + cspace > clen)) {
494 				u_int8_t *nc;
495 
496 				clen = cidx + cspace;
497 				if (clen >= PAGE_SIZE) {
498 					error = EINVAL;
499 					goto done;
500 				}
501 				nc = realloc(clen <= MLEN ? NULL : control,
502 						clen, M_TEMP, M_WAITOK);
503 				if (!nc) {
504 					error = ENOMEM;
505 					goto done;
506 				}
507 				if (cidx <= MLEN)
508 					/* Old buffer was in mbuf... */
509 					memcpy(nc, control, cidx);
510 				control = nc;
511 			}
512 
513 			/* Copy header */
514 			cmsg = (void *)&control[cidx];
515 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
516 			cmsg->cmsg_level = l_cmsg.cmsg_level;
517 			cmsg->cmsg_type = l_cmsg.cmsg_type;
518 
519 			/* Zero are between header and data */
520 			memset(cmsg + 1, 0,
521 				CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg));
522 
523 			/* Copyin the data */
524 			error = copyin(LINUX_CMSG_DATA(l_cc),
525 				CMSG_DATA(control),
526 				l_cmsg.cmsg_len - sizeof(l_cmsg));
527 			if (error)
528 				goto done;
529 
530 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
531 			cidx += cspace;
532 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
533 
534 		/* If we allocated a buffer, attach to mbuf */
535 		if (cidx > MLEN) {
536 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
537 			ctl_mbuf->m_flags |= M_EXT_RW;
538 		}
539 		control = NULL;
540 		ctl_mbuf->m_len = cidx;
541 
542 		msg.msg_control = ctl_mbuf;
543 		msg.msg_flags |= MSG_CONTROLMBUF;
544 	}
545 
546 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
547 	/* Freed internally */
548 	ctl_mbuf = NULL;
549 
550 done:
551 	if (ctl_mbuf != NULL) {
552 		if (control != NULL && control != mtod(ctl_mbuf, void *))
553 			free(control, M_MBUF);
554 		m_free(ctl_mbuf);
555 	}
556 	return (error);
557 }
558 
559 int
560 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
561 {
562 	/* {
563 		syscallarg(int) s;
564 		syscallarg(void *) buf;
565 		syscallarg(int) len;
566 		syscallarg(int) flags;
567 		syscallarg(struct osockaddr *) from;
568 		syscallarg(int *) fromlenaddr;
569 	} */
570 	int		error;
571 	struct sys_recvfrom_args bra;
572 
573 	SCARG(&bra, s) = SCARG(uap, s);
574 	SCARG(&bra, buf) = SCARG(uap, buf);
575 	SCARG(&bra, len) = SCARG(uap, len);
576 	SCARG(&bra, flags) = SCARG(uap, flags);
577 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
578 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
579 
580 	if ((error = sys_recvfrom(l, &bra, retval)))
581 		return (error);
582 
583 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
584 		return (error);
585 
586 	return (0);
587 }
588 
589 static int
590 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
591 {
592 	int dlen, error = 0;
593 	struct cmsghdr *cmsg;
594 	struct linux_cmsghdr linux_cmsg;
595 	struct mbuf *m;
596 	char *q, *q_end;
597 
598 	if (mp->msg_controllen <= 0 || control == 0) {
599 		mp->msg_controllen = 0;
600 		free_control_mbuf(l, control, control);
601 		return 0;
602 	}
603 
604 	q = (char *)mp->msg_control;
605 	q_end = q + mp->msg_controllen;
606 
607 	for (m = control; m != NULL; ) {
608 		cmsg = mtod(m, struct cmsghdr *);
609 
610 		/*
611 		 * Fixup cmsg. We handle two things:
612 		 * 0. different sizeof cmsg_len.
613 		 * 1. different values for level/type on some archs
614 		 * 2. different alignment of CMSG_DATA on some archs
615 		 */
616 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
617 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
618 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
619 
620 		dlen = q_end - q;
621 		if (linux_cmsg.cmsg_len > dlen) {
622 			/* Not enough room for the parameter */
623 			dlen -= sizeof linux_cmsg;
624 			if (dlen <= 0)
625 				/* Discard if header wont fit */
626 				break;
627 			mp->msg_flags |= MSG_CTRUNC;
628 			if (linux_cmsg.cmsg_level == SOL_SOCKET
629 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
630 				/* Do not truncate me ... */
631 				break;
632 		} else
633 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
634 
635 		switch (linux_cmsg.cmsg_level) {
636 		case SOL_SOCKET:
637 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
638 			switch (linux_cmsg.cmsg_type) {
639 			case SCM_RIGHTS:
640 				/* Linux SCM_RIGHTS is same as NetBSD */
641 				break;
642 
643 			default:
644 				/* other types not supported */
645 				error = EINVAL;
646 				goto done;
647 			}
648 			/* machine dependant ! */
649 			break;
650 		default:
651 			/* pray and leave intact */
652 			break;
653 		}
654 
655 		/* There can be padding between the header and data... */
656 		error = copyout(&linux_cmsg, q, sizeof *cmsg);
657 		if (error != 0) {
658 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
659 			    dlen);
660 		}
661 		if (error != 0) {
662 			/* We must free all the SCM_RIGHTS */
663 			m = control;
664 			break;
665 		}
666 		m = m->m_next;
667 		if (m == NULL || q + LINUX_CMSG_ALIGN(dlen) > q_end) {
668 			q += dlen;
669 			break;
670 		}
671 		q += LINUX_CMSG_ALIGN(dlen);
672 	}
673 
674   done:
675 	free_control_mbuf(l, control, m);
676 
677 	mp->msg_controllen = q - (char *)mp->msg_control;
678 	return error;
679 }
680 
681 int
682 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
683 {
684 	/* {
685 		syscallarg(int) s;
686 		syscallarg(struct msghdr *) msg;
687 		syscallarg(u_int) flags;
688 	} */
689 	struct msghdr	msg;
690 	int		error;
691 	struct mbuf	*from, *control;
692 
693 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
694 	if (error)
695 		return (error);
696 
697 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
698 	if (msg.msg_flags < 0) {
699 		/* Some unsupported flag */
700 		return (EINVAL);
701 	}
702 	msg.msg_flags |= MSG_IOVUSRSPACE;
703 
704 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
705 	    msg.msg_control != NULL ? &control : NULL, retval);
706 	if (error != 0)
707 		return error;
708 
709 	if (msg.msg_control != NULL)
710 		error = linux_copyout_msg_control(l, &msg, control);
711 
712 	if (error == 0 && from != 0) {
713 		mtod(from, struct osockaddr *)->sa_family =
714 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
715 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
716 			from);
717 	} else
718 		msg.msg_namelen = 0;
719 
720 	if (from != NULL)
721 		m_free(from);
722 
723 	if (error == 0) {
724 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
725 		if (msg.msg_flags < 0)
726 			/* Some flag unsupported by Linux */
727 			error = EINVAL;
728 		else
729 			error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
730 	}
731 
732 	return (error);
733 }
734 
735 /*
736  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
737  * is different, the rest matches IPPROTO_* on both systems.
738  */
739 int
740 linux_to_bsd_sopt_level(int llevel)
741 {
742 
743 	switch (llevel) {
744 	case LINUX_SOL_SOCKET:
745 		return SOL_SOCKET;
746 	case LINUX_SOL_IP:
747 		return IPPROTO_IP;
748 	case LINUX_SOL_TCP:
749 		return IPPROTO_TCP;
750 	case LINUX_SOL_UDP:
751 		return IPPROTO_UDP;
752 	default:
753 		return -1;
754 	}
755 }
756 
757 /*
758  * Convert Linux socket level socket option numbers to NetBSD values.
759  */
760 int
761 linux_to_bsd_so_sockopt(int lopt)
762 {
763 
764 	switch (lopt) {
765 	case LINUX_SO_DEBUG:
766 		return SO_DEBUG;
767 	case LINUX_SO_REUSEADDR:
768 		/*
769 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
770 		 * host:port pair through SO_REUSEADDR even if the address is not a
771 		 * multicast-address.  Effectively, this means that we should use
772 		 * SO_REUSEPORT to allow Linux applications to not exit with
773 		 * EADDRINUSE
774 		 */
775 		return SO_REUSEPORT;
776 	case LINUX_SO_TYPE:
777 		return SO_TYPE;
778 	case LINUX_SO_ERROR:
779 		return SO_ERROR;
780 	case LINUX_SO_DONTROUTE:
781 		return SO_DONTROUTE;
782 	case LINUX_SO_BROADCAST:
783 		return SO_BROADCAST;
784 	case LINUX_SO_SNDBUF:
785 		return SO_SNDBUF;
786 	case LINUX_SO_RCVBUF:
787 		return SO_RCVBUF;
788 	case LINUX_SO_KEEPALIVE:
789 		return SO_KEEPALIVE;
790 	case LINUX_SO_OOBINLINE:
791 		return SO_OOBINLINE;
792 	case LINUX_SO_LINGER:
793 		return SO_LINGER;
794 	case LINUX_SO_PRIORITY:
795 	case LINUX_SO_NO_CHECK:
796 	default:
797 		return -1;
798 	}
799 }
800 
801 /*
802  * Convert Linux IP level socket option number to NetBSD values.
803  */
804 int
805 linux_to_bsd_ip_sockopt(int lopt)
806 {
807 
808 	switch (lopt) {
809 	case LINUX_IP_TOS:
810 		return IP_TOS;
811 	case LINUX_IP_TTL:
812 		return IP_TTL;
813 	case LINUX_IP_MULTICAST_TTL:
814 		return IP_MULTICAST_TTL;
815 	case LINUX_IP_MULTICAST_LOOP:
816 		return IP_MULTICAST_LOOP;
817 	case LINUX_IP_MULTICAST_IF:
818 		return IP_MULTICAST_IF;
819 	case LINUX_IP_ADD_MEMBERSHIP:
820 		return IP_ADD_MEMBERSHIP;
821 	case LINUX_IP_DROP_MEMBERSHIP:
822 		return IP_DROP_MEMBERSHIP;
823 	default:
824 		return -1;
825 	}
826 }
827 
828 /*
829  * Convert Linux TCP level socket option number to NetBSD values.
830  */
831 int
832 linux_to_bsd_tcp_sockopt(int lopt)
833 {
834 
835 	switch (lopt) {
836 	case LINUX_TCP_NODELAY:
837 		return TCP_NODELAY;
838 	case LINUX_TCP_MAXSEG:
839 		return TCP_MAXSEG;
840 	default:
841 		return -1;
842 	}
843 }
844 
845 /*
846  * Convert Linux UDP level socket option number to NetBSD values.
847  */
848 int
849 linux_to_bsd_udp_sockopt(int lopt)
850 {
851 
852 	switch (lopt) {
853 	default:
854 		return -1;
855 	}
856 }
857 
858 /*
859  * Another reasonably straightforward function: setsockopt(2).
860  * The level and option numbers are converted; the values passed
861  * are not (yet) converted, the ones currently implemented don't
862  * need conversion, as they are the same on both systems.
863  */
864 int
865 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
866 {
867 	/* {
868 		syscallarg(int) s;
869 		syscallarg(int) level;
870 		syscallarg(int) optname;
871 		syscallarg(void *) optval;
872 		syscallarg(int) optlen;
873 	} */
874 	struct sys_setsockopt_args bsa;
875 	int name;
876 
877 	SCARG(&bsa, s) = SCARG(uap, s);
878 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
879 	SCARG(&bsa, val) = SCARG(uap, optval);
880 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
881 
882 	/*
883 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
884 	 * and returns EOPNOTSUPP for other levels
885 	 */
886 	if (SCARG(&bsa, level) != SOL_SOCKET) {
887 		struct socket *so;
888 		int error, family;
889 
890 		/* getsock() will use the descriptor for us */
891 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
892 		    	return error;
893 		family = so->so_proto->pr_domain->dom_family;
894 		fd_putfile(SCARG(&bsa, s));
895 
896 		if (family == AF_LOCAL)
897 			return EOPNOTSUPP;
898 	}
899 
900 	switch (SCARG(&bsa, level)) {
901 	case SOL_SOCKET:
902 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
903 		break;
904 	case IPPROTO_IP:
905 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
906 		break;
907 	case IPPROTO_TCP:
908 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
909 		break;
910 	case IPPROTO_UDP:
911 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
912 		break;
913 	default:
914 		return EINVAL;
915 	}
916 
917 	if (name == -1)
918 		return EINVAL;
919 	SCARG(&bsa, name) = name;
920 
921 	return sys_setsockopt(l, &bsa, retval);
922 }
923 
924 /*
925  * getsockopt(2) is very much the same as setsockopt(2) (see above)
926  */
927 int
928 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
929 {
930 	/* {
931 		syscallarg(int) s;
932 		syscallarg(int) level;
933 		syscallarg(int) optname;
934 		syscallarg(void *) optval;
935 		syscallarg(int *) optlen;
936 	} */
937 	struct sys_getsockopt_args bga;
938 	int name;
939 
940 	SCARG(&bga, s) = SCARG(uap, s);
941 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
942 	SCARG(&bga, val) = SCARG(uap, optval);
943 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
944 
945 	switch (SCARG(&bga, level)) {
946 	case SOL_SOCKET:
947 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
948 		break;
949 	case IPPROTO_IP:
950 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
951 		break;
952 	case IPPROTO_TCP:
953 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
954 		break;
955 	case IPPROTO_UDP:
956 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
957 		break;
958 	default:
959 		return EINVAL;
960 	}
961 
962 	if (name == -1)
963 		return EINVAL;
964 	SCARG(&bga, name) = name;
965 
966 	return sys_getsockopt(l, &bga, retval);
967 }
968 
969 #define IF_NAME_LEN 16
970 
971 int
972 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
973     void *data)
974 {
975 	/* Not the full structure, just enough to map what we do here */
976 	struct linux_ifreq {
977 		char if_name[IF_NAME_LEN];
978 		struct osockaddr hwaddr;
979 	} lreq;
980 	file_t *fp;
981 	struct ifaddr *ifa;
982 	struct ifnet *ifp;
983 	struct sockaddr_dl *sadl;
984 	int error, found;
985 	int index, ifnum;
986 
987 	/*
988 	 * We can't emulate this ioctl by calling sys_ioctl() to run
989 	 * SIOCGIFCONF, because the user buffer is not of the right
990 	 * type to take those results.  We can't use kernel buffers to
991 	 * receive the results, as the implementation of sys_ioctl()
992 	 * and ifconf() [which implements SIOCGIFCONF] use
993 	 * copyin()/copyout() which will fail on kernel addresses.
994 	 *
995 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
996 	 */
997 
998 	if ((fp = fd_getfile(fd)) == NULL)
999 		return (EBADF);
1000 
1001 	KERNEL_LOCK(1, NULL);
1002 
1003 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1004 		error = EBADF;
1005 		goto out;
1006 	}
1007 
1008 	error = copyin(data, &lreq, sizeof(lreq));
1009 	if (error)
1010 		goto out;
1011 	lreq.if_name[IF_NAME_LEN-1] = '\0';		/* just in case */
1012 
1013 	/*
1014 	 * Try real interface name first, then fake "ethX"
1015 	 */
1016 	found = 0;
1017 	IFNET_FOREACH(ifp) {
1018 		if (found)
1019 			break;
1020 		if (strcmp(lreq.if_name, ifp->if_xname))
1021 			/* not this interface */
1022 			continue;
1023 		found=1;
1024 		if (IFADDR_EMPTY(ifp)) {
1025 			error = ENODEV;
1026 			goto out;
1027 		}
1028 		IFADDR_FOREACH(ifa, ifp) {
1029 			sadl = satosdl(ifa->ifa_addr);
1030 			/* only return ethernet addresses */
1031 			/* XXX what about FDDI, etc. ? */
1032 			if (sadl->sdl_family != AF_LINK ||
1033 			    sadl->sdl_type != IFT_ETHER)
1034 				continue;
1035 			memcpy(&lreq.hwaddr.sa_data, CLLADDR(sadl),
1036 			       MIN(sadl->sdl_alen,
1037 				   sizeof(lreq.hwaddr.sa_data)));
1038 			lreq.hwaddr.sa_family =
1039 				sadl->sdl_family;
1040 			error = copyout(&lreq, data, sizeof(lreq));
1041 			goto out;
1042 		}
1043 	}
1044 
1045 	if (strncmp(lreq.if_name, "eth", 3) == 0) {
1046 		for (ifnum = 0, index = 3;
1047 		     lreq.if_name[index] != '\0' && index < IF_NAME_LEN;
1048 		     index++) {
1049 			ifnum *= 10;
1050 			ifnum += lreq.if_name[index] - '0';
1051 		}
1052 
1053 		error = EINVAL;			/* in case we don't find one */
1054 		found = 0;
1055 		IFNET_FOREACH(ifp) {
1056 			if (found)
1057 				break;
1058 			memcpy(lreq.if_name, ifp->if_xname,
1059 			       MIN(IF_NAME_LEN, IFNAMSIZ));
1060 			IFADDR_FOREACH(ifa, ifp) {
1061 				sadl = satosdl(ifa->ifa_addr);
1062 				/* only return ethernet addresses */
1063 				/* XXX what about FDDI, etc. ? */
1064 				if (sadl->sdl_family != AF_LINK ||
1065 				    sadl->sdl_type != IFT_ETHER)
1066 					continue;
1067 				if (ifnum--)
1068 					/* not the reqested iface */
1069 					continue;
1070 				memcpy(&lreq.hwaddr.sa_data,
1071 				       CLLADDR(sadl),
1072 				       MIN(sadl->sdl_alen,
1073 					   sizeof(lreq.hwaddr.sa_data)));
1074 				lreq.hwaddr.sa_family =
1075 					sadl->sdl_family;
1076 				error = copyout(&lreq, data, sizeof(lreq));
1077 				found = 1;
1078 				break;
1079 			}
1080 		}
1081 	} else {
1082 		/* unknown interface, not even an "eth*" name */
1083 		error = ENODEV;
1084 	}
1085 
1086 out:
1087 	KERNEL_UNLOCK_ONE(NULL);
1088 	fd_putfile(fd);
1089 	return error;
1090 }
1091 #undef IF_NAME_LEN
1092 
1093 int
1094 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1095 {
1096 	/* {
1097 		syscallarg(int) fd;
1098 		syscallarg(u_long) com;
1099 		syscallarg(void *) data;
1100 	} */
1101 	u_long com;
1102 	int error = 0, isdev = 0, dosys = 1;
1103 	struct sys_ioctl_args ia;
1104 	file_t *fp;
1105 	struct vnode *vp;
1106 	int (*ioctlf)(file_t *, u_long, void *);
1107 	struct ioctl_pt pt;
1108 
1109 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1110 		return (EBADF);
1111 
1112 	if (fp->f_type == DTYPE_VNODE) {
1113 		vp = (struct vnode *)fp->f_data;
1114 		isdev = vp->v_type == VCHR;
1115 	}
1116 
1117 	/*
1118 	 * Don't try to interpret socket ioctl calls that are done
1119 	 * on a device filedescriptor, just pass them through, to
1120 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1121 	 * device will only handle these if it's prepared to do
1122 	 * so, to avoid unexpected things from happening.
1123 	 */
1124 	if (isdev) {
1125 		dosys = 0;
1126 		ioctlf = fp->f_ops->fo_ioctl;
1127 		pt.com = SCARG(uap, com);
1128 		pt.data = SCARG(uap, data);
1129 		error = ioctlf(fp, PTIOCLINUX, &pt);
1130 		/*
1131 		 * XXX hack: if the function returns EJUSTRETURN,
1132 		 * it has stuffed a sysctl return value in pt.data.
1133 		 */
1134 		if (error == EJUSTRETURN) {
1135 			retval[0] = (register_t)pt.data;
1136 			error = 0;
1137 		}
1138 		goto out;
1139 	}
1140 
1141 	com = SCARG(uap, com);
1142 	retval[0] = 0;
1143 
1144 	switch (com) {
1145 	case LINUX_SIOCGIFCONF:
1146 		SCARG(&ia, com) = OOSIOCGIFCONF;
1147 		break;
1148 	case LINUX_SIOCGIFFLAGS:
1149 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1150 		break;
1151 	case LINUX_SIOCSIFFLAGS:
1152 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1153 		break;
1154 	case LINUX_SIOCGIFADDR:
1155 		SCARG(&ia, com) = OOSIOCGIFADDR;
1156 		break;
1157 	case LINUX_SIOCGIFDSTADDR:
1158 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1159 		break;
1160 	case LINUX_SIOCGIFBRDADDR:
1161 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1162 		break;
1163 	case LINUX_SIOCGIFNETMASK:
1164 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1165 		break;
1166 	case LINUX_SIOCADDMULTI:
1167 		SCARG(&ia, com) = OSIOCADDMULTI;
1168 		break;
1169 	case LINUX_SIOCDELMULTI:
1170 		SCARG(&ia, com) = OSIOCDELMULTI;
1171 		break;
1172 	case LINUX_SIOCGIFHWADDR:
1173 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1174 		    SCARG(uap, data));
1175 		dosys = 0;
1176 		break;
1177 	default:
1178 		error = EINVAL;
1179 	}
1180 
1181  out:
1182  	fd_putfile(SCARG(uap, fd));
1183 
1184 	if (error ==0 && dosys) {
1185 		SCARG(&ia, fd) = SCARG(uap, fd);
1186 		SCARG(&ia, data) = SCARG(uap, data);
1187 		error = sys_ioctl(curlwp, &ia, retval);
1188 	}
1189 
1190 	return error;
1191 }
1192 
1193 int
1194 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1195 {
1196 	/* {
1197 		syscallarg(int) s;
1198 		syscallarg(const struct sockaddr *) name;
1199 		syscallarg(int) namelen;
1200 	} */
1201 	int		error;
1202 	struct mbuf *nam;
1203 
1204 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1205 	    SCARG(uap, namelen));
1206 	if (error)
1207 		return (error);
1208 
1209 	error = do_sys_connect(l, SCARG(uap, s), nam);
1210 
1211 	if (error == EISCONN) {
1212 		struct socket *so;
1213 		int state, prflags, nbio;
1214 
1215 		/* getsock() will use the descriptor for us */
1216 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1217 		    	return EISCONN;
1218 
1219 		solock(so);
1220 		state = so->so_state;
1221 		nbio = so->so_nbio;
1222 		prflags = so->so_proto->pr_flags;
1223 		sounlock(so);
1224 		fd_putfile(SCARG(uap, s));
1225 		/*
1226 		 * We should only let this call succeed once per
1227 		 * non-blocking connect; however we don't have
1228 		 * a convenient place to keep that state..
1229 		 */
1230 		if (nbio && (state & SS_ISCONNECTED) &&
1231 		    (prflags & PR_CONNREQUIRED))
1232 			return 0;
1233 	}
1234 
1235 	return (error);
1236 }
1237 
1238 int
1239 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1240 {
1241 	/* {
1242 		syscallarg(int) s;
1243 		syscallarg(const struct osockaddr *) name;
1244 		syscallarg(int) namelen;
1245 	} */
1246 	int		error;
1247 	struct mbuf     *nam;
1248 
1249 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1250 	    SCARG(uap, namelen));
1251 	if (error)
1252 		return (error);
1253 
1254 	return do_sys_bind(l, SCARG(uap, s), nam);
1255 }
1256 
1257 int
1258 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1259 {
1260 	/* {
1261 		syscallarg(int) fdes;
1262 		syscallarg(void *) asa;
1263 		syscallarg(int *) alen;
1264 	} */
1265 	int error;
1266 
1267 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1268 		return (error);
1269 
1270 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1271 		return (error);
1272 
1273 	return (0);
1274 }
1275 
1276 int
1277 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1278 {
1279 	/* {
1280 		syscallarg(int) fdes;
1281 		syscallarg(void *) asa;
1282 		syscallarg(int *) alen;
1283 	} */
1284 	int error;
1285 
1286 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1287 		return (error);
1288 
1289 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1290 		return (error);
1291 
1292 	return (0);
1293 }
1294 
1295 /*
1296  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1297  * family and convert to sockaddr.
1298  */
1299 static int
1300 linux_get_sa(struct lwp *l, int s, struct mbuf **mp,
1301     const struct osockaddr *osa, unsigned int salen)
1302 {
1303 	int error, bdom;
1304 	struct sockaddr *sa;
1305 	struct osockaddr *kosa;
1306 	struct mbuf *m;
1307 
1308 	if (salen == 1 || salen > UCHAR_MAX) {
1309 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1310 		return EINVAL;
1311 	}
1312 
1313 	/* We'll need the address in an mbuf later, so copy into one here */
1314 	m = m_get(M_WAIT, MT_SONAME);
1315 	if (salen > MLEN)
1316 		MEXTMALLOC(m, salen, M_WAITOK);
1317 
1318 	m->m_len = salen;
1319 
1320 	if (salen == 0) {
1321 		*mp = m;
1322 		return 0;
1323 	}
1324 
1325 	kosa = mtod(m, void *);
1326 	if ((error = copyin(osa, kosa, salen))) {
1327 		DPRINTF(("error %d copying osa %p len %d\n",
1328 				error, osa, salen));
1329 		goto bad;
1330 	}
1331 
1332 	ktrkuser("linux sockaddr", kosa, salen);
1333 
1334 	bdom = linux_to_bsd_domain(kosa->sa_family);
1335 	if (bdom == -1) {
1336 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1337 		error = EINVAL;
1338 		goto bad;
1339 	}
1340 
1341 	/*
1342 	 * If the family is unspecified, use address family of the socket.
1343 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1344 	 */
1345 	if (bdom == AF_UNSPEC) {
1346 		struct socket *so;
1347 
1348 		/* getsock() will use the descriptor for us */
1349 		if ((error = fd_getsock(s, &so)) != 0)
1350 			goto bad;
1351 
1352 		bdom = so->so_proto->pr_domain->dom_family;
1353 		fd_putfile(s);
1354 
1355 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1356 	}
1357 
1358 #ifdef INET6
1359 	/*
1360 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1361 	 * which lacks the scope id compared with RFC2553 one. If we detect
1362 	 * the situation, reject the address and write a message to system log.
1363 	 *
1364 	 * Still accept addresses for which the scope id is not used.
1365 	 */
1366 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1367 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1368 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1369 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1370 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1371 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1372 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1373 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1374 			struct proc *p = l->l_proc;
1375 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1376 
1377 			log(LOG_DEBUG,
1378 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1379 			    "sockaddr_in6 rejected",
1380 			    p->p_pid, p->p_comm, uid);
1381 			error = EINVAL;
1382 			goto bad;
1383 		}
1384 		salen = sizeof (struct sockaddr_in6);
1385 		sin6->sin6_scope_id = 0;
1386 	}
1387 #endif
1388 
1389 	if (bdom == AF_INET)
1390 		salen = sizeof(struct sockaddr_in);
1391 
1392 	sa = (struct sockaddr *) kosa;
1393 	sa->sa_family = bdom;
1394 	sa->sa_len = salen;
1395 	m->m_len = salen;
1396 	ktrkuser("new sockaddr", kosa, salen);
1397 
1398 #ifdef DEBUG_LINUX
1399 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1400 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1401 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1402 	DPRINTF(("\n"));
1403 #endif
1404 
1405 	*mp = m;
1406 	return 0;
1407 
1408     bad:
1409 	m_free(m);
1410 	return error;
1411 }
1412 
1413 static int
1414 linux_sa_put(struct osockaddr *osa)
1415 {
1416 	struct sockaddr sa;
1417 	struct osockaddr *kosa;
1418 	int error, bdom, len;
1419 
1420 	/*
1421 	 * Only read/write the sockaddr family and length part, the rest is
1422 	 * not changed.
1423 	 */
1424 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1425 
1426 	error = copyin(osa, &sa, len);
1427 	if (error)
1428 		return (error);
1429 
1430 	bdom = bsd_to_linux_domain(sa.sa_family);
1431 	if (bdom == -1)
1432 		return (EINVAL);
1433 
1434 	/* Note: we convert from sockaddr to osockaddr here, too */
1435 	kosa = (struct osockaddr *) &sa;
1436 	kosa->sa_family = bdom;
1437 	error = copyout(kosa, osa, len);
1438 	if (error)
1439 		return (error);
1440 
1441 	return (0);
1442 }
1443 
1444 #ifndef __amd64__
1445 int
1446 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1447 {
1448 	/* {
1449 		syscallarg(int) s;
1450 		syscallarg(void *) buf;
1451 		syscallarg(int) len;
1452 		syscallarg(int) flags;
1453 	} */
1454 	struct sys_recvfrom_args bra;
1455 
1456 
1457 	SCARG(&bra, s) = SCARG(uap, s);
1458 	SCARG(&bra, buf) = SCARG(uap, buf);
1459 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1460 	SCARG(&bra, flags) = SCARG(uap, flags);
1461 	SCARG(&bra, from) = NULL;
1462 	SCARG(&bra, fromlenaddr) = NULL;
1463 
1464 	return (sys_recvfrom(l, &bra, retval));
1465 }
1466 
1467 int
1468 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1469 {
1470 	/* {
1471 		syscallarg(int) s;
1472 		syscallarg(void *) buf;
1473 		syscallarg(int) len;
1474 		syscallarg(int) flags;
1475 	} */
1476 	struct sys_sendto_args bsa;
1477 
1478 	SCARG(&bsa, s)		= SCARG(uap, s);
1479 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1480 	SCARG(&bsa, len)	= SCARG(uap, len);
1481 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1482 	SCARG(&bsa, to)		= NULL;
1483 	SCARG(&bsa, tolen)	= 0;
1484 
1485 	return (sys_sendto(l, &bsa, retval));
1486 }
1487 #endif
1488 
1489 int
1490 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1491 {
1492 	/* {
1493 		syscallarg(int) s;
1494 		syscallarg(struct osockaddr *) name;
1495 		syscallarg(int *) anamelen;
1496 	} */
1497 	int error;
1498 	struct sys_accept_args baa;
1499 
1500 	SCARG(&baa, s)		= SCARG(uap, s);
1501 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1502 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1503 
1504 	if ((error = sys_accept(l, &baa, retval)))
1505 		return (error);
1506 
1507 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1508 		return (error);
1509 
1510 	return (0);
1511 }
1512