xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision 7c3f385475147b6e1c4753f2bee961630e2dfc40)
1 /*	$NetBSD: linux_socket.c,v 1.89 2008/03/26 20:08:22 ad 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  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the NetBSD
21  *	Foundation, Inc. and its contributors.
22  * 4. Neither the name of The NetBSD Foundation nor the names of its
23  *    contributors may be used to endorse or promote products derived
24  *    from this software without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36  * POSSIBILITY OF SUCH DAMAGE.
37  */
38 
39 /*
40  * Functions in multiarch:
41  *	linux_sys_socketcall		: linux_socketcall.c
42  */
43 
44 #include <sys/cdefs.h>
45 __KERNEL_RCSID(0, "$NetBSD: linux_socket.c,v 1.89 2008/03/26 20:08:22 ad Exp $");
46 
47 #if defined(_KERNEL_OPT)
48 #include "opt_inet.h"
49 #endif /* defined(_KERNEL_OPT) */
50 
51 #include <sys/param.h>
52 #include <sys/kernel.h>
53 #include <sys/systm.h>
54 #include <sys/buf.h>
55 #include <sys/malloc.h>
56 #include <sys/ioctl.h>
57 #include <sys/tty.h>
58 #include <sys/file.h>
59 #include <sys/filedesc.h>
60 #include <sys/select.h>
61 #include <sys/socket.h>
62 #include <sys/socketvar.h>
63 #include <sys/domain.h>
64 #include <net/if.h>
65 #include <net/if_dl.h>
66 #include <net/if_types.h>
67 #include <netinet/in.h>
68 #include <netinet/tcp.h>
69 #include <sys/mount.h>
70 #include <sys/proc.h>
71 #include <sys/vnode.h>
72 #include <sys/device.h>
73 #include <sys/protosw.h>
74 #include <sys/mbuf.h>
75 #include <sys/syslog.h>
76 #include <sys/exec.h>
77 #include <sys/kauth.h>
78 #include <sys/syscallargs.h>
79 #include <sys/ktrace.h>
80 
81 #include <lib/libkern/libkern.h>
82 
83 #ifdef INET6
84 #include <netinet/ip6.h>
85 #include <netinet6/ip6_var.h>
86 #endif
87 
88 #include <compat/sys/socket.h>
89 #include <compat/sys/sockio.h>
90 
91 #include <compat/linux/common/linux_types.h>
92 #include <compat/linux/common/linux_util.h>
93 #include <compat/linux/common/linux_signal.h>
94 #include <compat/linux/common/linux_ioctl.h>
95 #include <compat/linux/common/linux_socket.h>
96 #if !defined(__alpha__) && !defined(__amd64__)
97 #include <compat/linux/common/linux_socketcall.h>
98 #endif
99 #include <compat/linux/common/linux_sockio.h>
100 #include <compat/linux/common/linux_ipc.h>
101 #include <compat/linux/common/linux_sem.h>
102 
103 #include <compat/linux/linux_syscallargs.h>
104 
105 #ifdef DEBUG_LINUX
106 #define DPRINTF(a) uprintf a
107 #else
108 #define DPRINTF(a)
109 #endif
110 
111 /*
112  * The calls in this file are entered either via the linux_socketcall()
113  * interface or, on the Alpha, as individual syscalls.  The
114  * linux_socketcall function does any massaging of arguments so that all
115  * the calls in here need not think that they are anything other
116  * than a normal syscall.
117  */
118 
119 static int linux_to_bsd_domain(int);
120 static int bsd_to_linux_domain(int);
121 int linux_to_bsd_sopt_level(int);
122 int linux_to_bsd_so_sockopt(int);
123 int linux_to_bsd_ip_sockopt(int);
124 int linux_to_bsd_tcp_sockopt(int);
125 int linux_to_bsd_udp_sockopt(int);
126 int linux_getifhwaddr(struct lwp *, register_t *, u_int, void *);
127 static int linux_get_sa(struct lwp *, int, struct mbuf **,
128 		const struct osockaddr *, int);
129 static int linux_sa_put(struct osockaddr *osa);
130 static int linux_to_bsd_msg_flags(int);
131 static int bsd_to_linux_msg_flags(int);
132 
133 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
134 	AF_UNSPEC,
135 	AF_UNIX,
136 	AF_INET,
137 	AF_CCITT,	/* LINUX_AF_AX25 */
138 	AF_IPX,
139 	AF_APPLETALK,
140 	-1,		/* LINUX_AF_NETROM */
141 	-1,		/* LINUX_AF_BRIDGE */
142 	-1,		/* LINUX_AF_ATMPVC */
143 	AF_CCITT,	/* LINUX_AF_X25 */
144 	AF_INET6,
145 	-1,		/* LINUX_AF_ROSE */
146 	AF_DECnet,
147 	-1,		/* LINUX_AF_NETBEUI */
148 	-1,		/* LINUX_AF_SECURITY */
149 	pseudo_AF_KEY,
150 	AF_ROUTE,	/* LINUX_AF_NETLINK */
151 	-1,		/* LINUX_AF_PACKET */
152 	-1,		/* LINUX_AF_ASH */
153 	-1,		/* LINUX_AF_ECONET */
154 	-1,		/* LINUX_AF_ATMSVC */
155 	AF_SNA,
156 	/* rest up to LINUX_AF_MAX-1 is not allocated */
157 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
158 };
159 
160 static const int bsd_to_linux_domain_[AF_MAX] = {
161 	LINUX_AF_UNSPEC,
162 	LINUX_AF_UNIX,
163 	LINUX_AF_INET,
164 	-1,		/* AF_IMPLINK */
165 	-1,		/* AF_PUP */
166 	-1,		/* AF_CHAOS */
167 	-1,		/* AF_NS */
168 	-1,		/* AF_ISO */
169 	-1,		/* AF_ECMA */
170 	-1,		/* AF_DATAKIT */
171 	LINUX_AF_AX25,	/* AF_CCITT */
172 	LINUX_AF_SNA,
173 	LINUX_AF_DECnet,
174 	-1,		/* AF_DLI */
175 	-1,		/* AF_LAT */
176 	-1,		/* AF_HYLINK */
177 	LINUX_AF_APPLETALK,
178 	LINUX_AF_NETLINK,
179 	-1,		/* AF_LINK */
180 	-1,		/* AF_XTP */
181 	-1,		/* AF_COIP */
182 	-1,		/* AF_CNT */
183 	-1,		/* pseudo_AF_RTIP */
184 	LINUX_AF_IPX,
185 	LINUX_AF_INET6,
186 	-1,		/* pseudo_AF_PIP */
187 	-1,		/* AF_ISDN */
188 	-1,		/* AF_NATM */
189 	-1,		/* AF_ARP */
190 	LINUX_pseudo_AF_KEY,
191 	-1,		/* pseudo_AF_HDRCMPLT */
192 };
193 
194 static const struct {
195 	int bfl;
196 	int lfl;
197 } bsd_to_linux_msg_flags_[] = {
198 	{MSG_OOB,		LINUX_MSG_OOB},
199 	{MSG_PEEK,		LINUX_MSG_PEEK},
200 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
201 	{MSG_EOR,		LINUX_MSG_EOR},
202 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
203 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
204 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
205 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
206 	{MSG_BCAST,		0},		/* not supported, clear */
207 	{MSG_MCAST,		0},		/* not supported, clear */
208 	{-1, /* not supp */	LINUX_MSG_PROBE},
209 	{-1, /* not supp */	LINUX_MSG_FIN},
210 	{-1, /* not supp */	LINUX_MSG_SYN},
211 	{-1, /* not supp */	LINUX_MSG_CONFIRM},
212 	{-1, /* not supp */	LINUX_MSG_RST},
213 	{-1, /* not supp */	LINUX_MSG_ERRQUEUE},
214 	{-1, /* not supp */	LINUX_MSG_NOSIGNAL},
215 	{-1, /* not supp */	LINUX_MSG_MORE},
216 };
217 
218 /*
219  * Convert between Linux and BSD socket domain values
220  */
221 static int
222 linux_to_bsd_domain(int ldom)
223 {
224 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
225 		return (-1);
226 
227 	return linux_to_bsd_domain_[ldom];
228 }
229 
230 /*
231  * Convert between BSD and Linux socket domain values
232  */
233 static int
234 bsd_to_linux_domain(int bdom)
235 {
236 	if (bdom < 0 || bdom >= AF_MAX)
237 		return (-1);
238 
239 	return bsd_to_linux_domain_[bdom];
240 }
241 
242 static int
243 linux_to_bsd_msg_flags(int lflag)
244 {
245 	int i, lfl, bfl;
246 	int bflag = 0;
247 
248 	if (lflag == 0)
249 		return (0);
250 
251 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
252 		bfl = bsd_to_linux_msg_flags_[i].bfl;
253 		lfl = bsd_to_linux_msg_flags_[i].lfl;
254 
255 		if (lfl == 0)
256 			continue;
257 
258 		if (lflag & lfl) {
259 			if (bfl < 0)
260 				return (-1);
261 
262 			bflag |= bfl;
263 		}
264 	}
265 
266 	return (bflag);
267 }
268 
269 static int
270 bsd_to_linux_msg_flags(int bflag)
271 {
272 	int i, lfl, bfl;
273 	int lflag = 0;
274 
275 	if (bflag == 0)
276 		return (0);
277 
278 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
279 		bfl = bsd_to_linux_msg_flags_[i].bfl;
280 		lfl = bsd_to_linux_msg_flags_[i].lfl;
281 
282 		if (bfl <= 0)
283 			continue;
284 
285 		if (bflag & bfl) {
286 			if (lfl < 0)
287 				return (-1);
288 
289 			lflag |= lfl;
290 		}
291 	}
292 
293 	return (lflag);
294 }
295 
296 int
297 linux_sys_socket(struct lwp *l, const struct linux_sys_socket_args *uap, register_t *retval)
298 {
299 	/* {
300 		syscallarg(int)	domain;
301 		syscallarg(int)	type;
302 		syscallarg(int) protocol;
303 	} */
304 	struct sys___socket30_args bsa;
305 	int error;
306 
307 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
308 	SCARG(&bsa, type) = SCARG(uap, type);
309 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
310 	if (SCARG(&bsa, domain) == -1)
311 		return EINVAL;
312 	error = sys___socket30(l, &bsa, retval);
313 
314 #ifdef INET6
315 	/*
316 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
317 	 * default and some apps depend on this. So, set V6ONLY to 0
318 	 * for Linux apps if the sysctl value is set to 1.
319 	 */
320 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
321 		struct socket *so;
322 
323 		if (fd_getsock(*retval, &so) == 0) {
324 			struct mbuf *m;
325 
326 			m = m_get(M_WAIT, MT_SOOPTS);
327 			m->m_len = sizeof(int);
328 			*mtod(m, int *) = 0;
329 
330 			/* ignore error */
331 			(void) sosetopt(so, IPPROTO_IPV6, IPV6_V6ONLY, m);
332 
333 			fd_putfile(*retval);
334 		}
335 	}
336 #endif
337 
338 	return (error);
339 }
340 
341 int
342 linux_sys_socketpair(struct lwp *l, const struct linux_sys_socketpair_args *uap, register_t *retval)
343 {
344 	/* {
345 		syscallarg(int) domain;
346 		syscallarg(int) type;
347 		syscallarg(int) protocol;
348 		syscallarg(int *) rsv;
349 	} */
350 	struct sys_socketpair_args bsa;
351 
352 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
353 	if (SCARG(&bsa, domain) == -1)
354 		return EINVAL;
355 	SCARG(&bsa, type) = SCARG(uap, type);
356 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
357 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
358 
359 	return sys_socketpair(l, &bsa, retval);
360 }
361 
362 int
363 linux_sys_sendto(struct lwp *l, const struct linux_sys_sendto_args *uap, register_t *retval)
364 {
365 	/* {
366 		syscallarg(int)				s;
367 		syscallarg(void *)			msg;
368 		syscallarg(int)				len;
369 		syscallarg(int)				flags;
370 		syscallarg(struct osockaddr *)		to;
371 		syscallarg(int)				tolen;
372 	} */
373 	struct msghdr   msg;
374 	struct iovec    aiov;
375 	struct mbuf *nam;
376 	int bflags;
377 	int error;
378 
379 	/* Translate message flags.  */
380 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
381 	if (bflags < 0)
382 		/* Some supported flag */
383 		return EINVAL;
384 
385 	/* Read in and convert the sockaddr */
386 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
387 	    SCARG(uap, tolen));
388 	if (error)
389 		return (error);
390 	msg.msg_flags = MSG_NAMEMBUF;
391 
392 	msg.msg_name = nam;
393 	msg.msg_namelen = SCARG(uap, tolen);
394 	msg.msg_iov = &aiov;
395 	msg.msg_iovlen = 1;
396 	msg.msg_control = 0;
397 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
398 	aiov.iov_len = SCARG(uap, len);
399 
400 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
401 }
402 
403 int
404 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
405 {
406 	/* {
407 		syscallarg(int) s;
408 		syscallarg(struct msghdr *) msg;
409 		syscallarg(u_int) flags;
410 	} */
411 	struct msghdr	msg;
412 	int		error;
413 	int		bflags;
414 	struct mbuf     *nam;
415 	u_int8_t	*control;
416 	struct mbuf     *ctl_mbuf = NULL;
417 
418 	msg.msg_flags = MSG_IOVUSRSPACE;
419 
420 	/*
421 	 * Translate message flags.
422 	 */
423 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
424 	if (bflags < 0)
425 		/* Some supported flag */
426 		return EINVAL;
427 
428 	if (msg.msg_name) {
429 		/* Read in and convert the sockaddr */
430 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
431 		    msg.msg_namelen);
432 		if (error)
433 			return (error);
434 		msg.msg_flags |= MSG_NAMEMBUF;
435 		msg.msg_name = nam;
436 	}
437 
438 	/*
439 	 * Handle cmsg if there is any.
440 	 */
441 	if (CMSG_FIRSTHDR(&msg)) {
442 		struct linux_cmsghdr l_cmsg, *l_cc;
443 		struct cmsghdr *cmsg;
444 		ssize_t resid = msg.msg_controllen;
445 		size_t clen, cidx = 0, cspace;
446 
447 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
448 		clen = MLEN;
449 		control = mtod(ctl_mbuf, void *);
450 
451 		l_cc = LINUX_CMSG_FIRSTHDR(&msg);
452 		do {
453 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
454 			if (error)
455 				goto done;
456 
457 			/*
458 			 * Sanity check the control message length.
459 			 */
460 			if (l_cmsg.cmsg_len > resid
461 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
462 				error = EINVAL;
463 				goto done;
464 			}
465 
466 			/*
467 			 * Refuse unsupported control messages, and
468 			 * translate fields as appropriate.
469 			 */
470 			switch (l_cmsg.cmsg_level) {
471 			case LINUX_SOL_SOCKET:
472 				/* It only differs on some archs */
473 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
474 					l_cmsg.cmsg_level = SOL_SOCKET;
475 
476 				switch(l_cmsg.cmsg_type) {
477 				case LINUX_SCM_RIGHTS:
478 					/* Linux SCM_RIGHTS is same as NetBSD */
479 					break;
480 
481 				default:
482 					/* other types not supported */
483 					error = EINVAL;
484 					goto done;
485 				}
486 				break;
487 			default:
488 				/* pray and leave intact */
489 				break;
490 			}
491 
492 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
493 
494 			/* Check the buffer is big enough */
495 			if (__predict_false(cidx + cspace > clen)) {
496 				u_int8_t *nc;
497 
498 				clen = cidx + cspace;
499 				if (clen >= PAGE_SIZE) {
500 					error = EINVAL;
501 					goto done;
502 				}
503 				nc = realloc(clen <= MLEN ? NULL : control,
504 						clen, M_TEMP, M_WAITOK);
505 				if (!nc) {
506 					error = ENOMEM;
507 					goto done;
508 				}
509 				if (cidx <= MLEN)
510 					/* Old buffer was in mbuf... */
511 					memcpy(nc, control, cidx);
512 				control = nc;
513 			}
514 
515 			/* Copy header */
516 			cmsg = (void *)&control[cidx];
517 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
518 			cmsg->cmsg_level = l_cmsg.cmsg_level;
519 			cmsg->cmsg_type = l_cmsg.cmsg_type;
520 
521 			/* Zero are between header and data */
522 			memset(cmsg + 1, 0,
523 				CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg));
524 
525 			/* Copyin the data */
526 			error = copyin(LINUX_CMSG_DATA(l_cc),
527 				CMSG_DATA(control),
528 				l_cmsg.cmsg_len - sizeof(l_cmsg));
529 			if (error)
530 				goto done;
531 
532 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
533 			cidx += cspace;
534 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
535 
536 		/* If we allocated a buffer, attach to mbuf */
537 		if (cidx > MLEN) {
538 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
539 			ctl_mbuf->m_flags |= M_EXT_RW;
540 		}
541 		control = NULL;
542 		ctl_mbuf->m_len = cidx;
543 
544 		msg.msg_control = ctl_mbuf;
545 		msg.msg_flags |= MSG_CONTROLMBUF;
546 	}
547 
548 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
549 	/* Freed internally */
550 	ctl_mbuf = NULL;
551 
552 done:
553 	if (ctl_mbuf != NULL) {
554 		if (control != NULL && control != mtod(ctl_mbuf, void *))
555 			free(control, M_MBUF);
556 		m_free(ctl_mbuf);
557 	}
558 	return (error);
559 }
560 
561 int
562 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
563 {
564 	/* {
565 		syscallarg(int) s;
566 		syscallarg(void *) buf;
567 		syscallarg(int) len;
568 		syscallarg(int) flags;
569 		syscallarg(struct osockaddr *) from;
570 		syscallarg(int *) fromlenaddr;
571 	} */
572 	int		error;
573 	struct sys_recvfrom_args bra;
574 
575 	SCARG(&bra, s) = SCARG(uap, s);
576 	SCARG(&bra, buf) = SCARG(uap, buf);
577 	SCARG(&bra, len) = SCARG(uap, len);
578 	SCARG(&bra, flags) = SCARG(uap, flags);
579 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
580 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
581 
582 	if ((error = sys_recvfrom(l, &bra, retval)))
583 		return (error);
584 
585 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
586 		return (error);
587 
588 	return (0);
589 }
590 
591 static int
592 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
593 {
594 	int dlen, error = 0;
595 	struct cmsghdr *cmsg;
596 	struct linux_cmsghdr linux_cmsg;
597 	struct mbuf *m;
598 	char *q, *q_end;
599 
600 	if (mp->msg_controllen <= 0 || control == 0) {
601 		mp->msg_controllen = 0;
602 		free_control_mbuf(l, control, control);
603 		return 0;
604 	}
605 
606 	q = (char *)mp->msg_control;
607 	q_end = q + mp->msg_controllen;
608 
609 	for (m = control; m != NULL; ) {
610 		cmsg = mtod(m, struct cmsghdr *);
611 
612 		/*
613 		 * Fixup cmsg. We handle two things:
614 		 * 0. different sizeof cmsg_len.
615 		 * 1. different values for level/type on some archs
616 		 * 2. different alignment of CMSG_DATA on some archs
617 		 */
618 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
619 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
620 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
621 
622 		dlen = q_end - q;
623 		if (linux_cmsg.cmsg_len > dlen) {
624 			/* Not enough room for the parameter */
625 			dlen -= sizeof linux_cmsg;
626 			if (dlen <= 0)
627 				/* Discard if header wont fit */
628 				break;
629 			mp->msg_flags |= MSG_CTRUNC;
630 			if (linux_cmsg.cmsg_level == SOL_SOCKET
631 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
632 				/* Do not truncate me ... */
633 				break;
634 		} else
635 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
636 
637 		switch (linux_cmsg.cmsg_level) {
638 		case SOL_SOCKET:
639 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
640 			switch (linux_cmsg.cmsg_type) {
641 			case SCM_RIGHTS:
642 				/* Linux SCM_RIGHTS is same as NetBSD */
643 				break;
644 
645 			default:
646 				/* other types not supported */
647 				error = EINVAL;
648 				goto done;
649 			}
650 			/* machine dependant ! */
651 			break;
652 		default:
653 			/* pray and leave intact */
654 			break;
655 		}
656 
657 		/* There can be padding between the header and data... */
658 		error = copyout(&linux_cmsg, q, sizeof *cmsg);
659 		if (error != 0) {
660 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
661 			    dlen);
662 		}
663 		if (error != 0) {
664 			/* We must free all the SCM_RIGHTS */
665 			m = control;
666 			break;
667 		}
668 		m = m->m_next;
669 		if (m == NULL || q + LINUX_CMSG_ALIGN(dlen) > q_end) {
670 			q += dlen;
671 			break;
672 		}
673 		q += LINUX_CMSG_ALIGN(dlen);
674 	}
675 
676   done:
677 	free_control_mbuf(l, control, m);
678 
679 	mp->msg_controllen = q - (char *)mp->msg_control;
680 	return error;
681 }
682 
683 int
684 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
685 {
686 	/* {
687 		syscallarg(int) s;
688 		syscallarg(struct msghdr *) msg;
689 		syscallarg(u_int) flags;
690 	} */
691 	struct msghdr	msg;
692 	int		error;
693 	struct mbuf	*from, *control;
694 
695 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
696 	if (error)
697 		return (error);
698 
699 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
700 	if (msg.msg_flags < 0) {
701 		/* Some unsupported flag */
702 		return (EINVAL);
703 	}
704 	msg.msg_flags |= MSG_IOVUSRSPACE;
705 
706 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
707 	    msg.msg_control != NULL ? &control : NULL, retval);
708 	if (error != 0)
709 		return error;
710 
711 	if (msg.msg_control != NULL)
712 		error = linux_copyout_msg_control(l, &msg, control);
713 
714 	if (error == 0 && from != 0) {
715 		mtod(from, struct osockaddr *)->sa_family =
716 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
717 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
718 			from);
719 	} else
720 		msg.msg_namelen = 0;
721 
722 	if (from != NULL)
723 		m_free(from);
724 
725 	if (error == 0) {
726 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
727 		if (msg.msg_flags < 0)
728 			/* Some flag unsupported by Linux */
729 			error = EINVAL;
730 		else
731 			error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
732 	}
733 
734 	return (error);
735 }
736 
737 /*
738  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
739  * is different, the rest matches IPPROTO_* on both systems.
740  */
741 int
742 linux_to_bsd_sopt_level(int llevel)
743 {
744 
745 	switch (llevel) {
746 	case LINUX_SOL_SOCKET:
747 		return SOL_SOCKET;
748 	case LINUX_SOL_IP:
749 		return IPPROTO_IP;
750 	case LINUX_SOL_TCP:
751 		return IPPROTO_TCP;
752 	case LINUX_SOL_UDP:
753 		return IPPROTO_UDP;
754 	default:
755 		return -1;
756 	}
757 }
758 
759 /*
760  * Convert Linux socket level socket option numbers to NetBSD values.
761  */
762 int
763 linux_to_bsd_so_sockopt(int lopt)
764 {
765 
766 	switch (lopt) {
767 	case LINUX_SO_DEBUG:
768 		return SO_DEBUG;
769 	case LINUX_SO_REUSEADDR:
770 		/*
771 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
772 		 * host:port pair through SO_REUSEADDR even if the address is not a
773 		 * multicast-address.  Effectively, this means that we should use
774 		 * SO_REUSEPORT to allow Linux applications to not exit with
775 		 * EADDRINUSE
776 		 */
777 		return SO_REUSEPORT;
778 	case LINUX_SO_TYPE:
779 		return SO_TYPE;
780 	case LINUX_SO_ERROR:
781 		return SO_ERROR;
782 	case LINUX_SO_DONTROUTE:
783 		return SO_DONTROUTE;
784 	case LINUX_SO_BROADCAST:
785 		return SO_BROADCAST;
786 	case LINUX_SO_SNDBUF:
787 		return SO_SNDBUF;
788 	case LINUX_SO_RCVBUF:
789 		return SO_RCVBUF;
790 	case LINUX_SO_KEEPALIVE:
791 		return SO_KEEPALIVE;
792 	case LINUX_SO_OOBINLINE:
793 		return SO_OOBINLINE;
794 	case LINUX_SO_LINGER:
795 		return SO_LINGER;
796 	case LINUX_SO_PRIORITY:
797 	case LINUX_SO_NO_CHECK:
798 	default:
799 		return -1;
800 	}
801 }
802 
803 /*
804  * Convert Linux IP level socket option number to NetBSD values.
805  */
806 int
807 linux_to_bsd_ip_sockopt(int lopt)
808 {
809 
810 	switch (lopt) {
811 	case LINUX_IP_TOS:
812 		return IP_TOS;
813 	case LINUX_IP_TTL:
814 		return IP_TTL;
815 	case LINUX_IP_MULTICAST_TTL:
816 		return IP_MULTICAST_TTL;
817 	case LINUX_IP_MULTICAST_LOOP:
818 		return IP_MULTICAST_LOOP;
819 	case LINUX_IP_MULTICAST_IF:
820 		return IP_MULTICAST_IF;
821 	case LINUX_IP_ADD_MEMBERSHIP:
822 		return IP_ADD_MEMBERSHIP;
823 	case LINUX_IP_DROP_MEMBERSHIP:
824 		return IP_DROP_MEMBERSHIP;
825 	default:
826 		return -1;
827 	}
828 }
829 
830 /*
831  * Convert Linux TCP level socket option number to NetBSD values.
832  */
833 int
834 linux_to_bsd_tcp_sockopt(int lopt)
835 {
836 
837 	switch (lopt) {
838 	case LINUX_TCP_NODELAY:
839 		return TCP_NODELAY;
840 	case LINUX_TCP_MAXSEG:
841 		return TCP_MAXSEG;
842 	default:
843 		return -1;
844 	}
845 }
846 
847 /*
848  * Convert Linux UDP level socket option number to NetBSD values.
849  */
850 int
851 linux_to_bsd_udp_sockopt(int lopt)
852 {
853 
854 	switch (lopt) {
855 	default:
856 		return -1;
857 	}
858 }
859 
860 /*
861  * Another reasonably straightforward function: setsockopt(2).
862  * The level and option numbers are converted; the values passed
863  * are not (yet) converted, the ones currently implemented don't
864  * need conversion, as they are the same on both systems.
865  */
866 int
867 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
868 {
869 	/* {
870 		syscallarg(int) s;
871 		syscallarg(int) level;
872 		syscallarg(int) optname;
873 		syscallarg(void *) optval;
874 		syscallarg(int) optlen;
875 	} */
876 	struct sys_setsockopt_args bsa;
877 	int name;
878 
879 	SCARG(&bsa, s) = SCARG(uap, s);
880 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
881 	SCARG(&bsa, val) = SCARG(uap, optval);
882 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
883 
884 	/*
885 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
886 	 * and returns EOPNOTSUPP for other levels
887 	 */
888 	if (SCARG(&bsa, level) != SOL_SOCKET) {
889 		struct socket *so;
890 		int error, family;
891 
892 		/* getsock() will use the descriptor for us */
893 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
894 		    	return error;
895 		family = so->so_proto->pr_domain->dom_family;
896 		fd_putfile(SCARG(&bsa, s));
897 
898 		if (family == AF_LOCAL)
899 			return EOPNOTSUPP;
900 	}
901 
902 	switch (SCARG(&bsa, level)) {
903 	case SOL_SOCKET:
904 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
905 		break;
906 	case IPPROTO_IP:
907 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
908 		break;
909 	case IPPROTO_TCP:
910 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
911 		break;
912 	case IPPROTO_UDP:
913 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
914 		break;
915 	default:
916 		return EINVAL;
917 	}
918 
919 	if (name == -1)
920 		return EINVAL;
921 	SCARG(&bsa, name) = name;
922 
923 	return sys_setsockopt(l, &bsa, retval);
924 }
925 
926 /*
927  * getsockopt(2) is very much the same as setsockopt(2) (see above)
928  */
929 int
930 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
931 {
932 	/* {
933 		syscallarg(int) s;
934 		syscallarg(int) level;
935 		syscallarg(int) optname;
936 		syscallarg(void *) optval;
937 		syscallarg(int *) optlen;
938 	} */
939 	struct sys_getsockopt_args bga;
940 	int name;
941 
942 	SCARG(&bga, s) = SCARG(uap, s);
943 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
944 	SCARG(&bga, val) = SCARG(uap, optval);
945 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
946 
947 	switch (SCARG(&bga, level)) {
948 	case SOL_SOCKET:
949 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
950 		break;
951 	case IPPROTO_IP:
952 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
953 		break;
954 	case IPPROTO_TCP:
955 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
956 		break;
957 	case IPPROTO_UDP:
958 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
959 		break;
960 	default:
961 		return EINVAL;
962 	}
963 
964 	if (name == -1)
965 		return EINVAL;
966 	SCARG(&bga, name) = name;
967 
968 	return sys_getsockopt(l, &bga, retval);
969 }
970 
971 #define IF_NAME_LEN 16
972 
973 int
974 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
975     void *data)
976 {
977 	/* Not the full structure, just enough to map what we do here */
978 	struct linux_ifreq {
979 		char if_name[IF_NAME_LEN];
980 		struct osockaddr hwaddr;
981 	} lreq;
982 	file_t *fp;
983 	struct ifaddr *ifa;
984 	struct ifnet *ifp;
985 	struct sockaddr_dl *sadl;
986 	int error, found;
987 	int index, ifnum;
988 
989 	/*
990 	 * We can't emulate this ioctl by calling sys_ioctl() to run
991 	 * SIOCGIFCONF, because the user buffer is not of the right
992 	 * type to take those results.  We can't use kernel buffers to
993 	 * receive the results, as the implementation of sys_ioctl()
994 	 * and ifconf() [which implements SIOCGIFCONF] use
995 	 * copyin()/copyout() which will fail on kernel addresses.
996 	 *
997 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
998 	 */
999 
1000 	if ((fp = fd_getfile(fd)) == NULL)
1001 		return (EBADF);
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 	fd_putfile(fd);
1088 	return error;
1089 }
1090 #undef IF_NAME_LEN
1091 
1092 int
1093 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1094 {
1095 	/* {
1096 		syscallarg(int) fd;
1097 		syscallarg(u_long) com;
1098 		syscallarg(void *) data;
1099 	} */
1100 	u_long com;
1101 	int error = 0, isdev = 0, dosys = 1;
1102 	struct sys_ioctl_args ia;
1103 	file_t *fp;
1104 	struct vnode *vp;
1105 	int (*ioctlf)(file_t *, u_long, void *);
1106 	struct ioctl_pt pt;
1107 
1108 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1109 		return (EBADF);
1110 
1111 	if (fp->f_type == DTYPE_VNODE) {
1112 		vp = (struct vnode *)fp->f_data;
1113 		isdev = vp->v_type == VCHR;
1114 	}
1115 
1116 	/*
1117 	 * Don't try to interpret socket ioctl calls that are done
1118 	 * on a device filedescriptor, just pass them through, to
1119 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1120 	 * device will only handle these if it's prepared to do
1121 	 * so, to avoid unexpected things from happening.
1122 	 */
1123 	if (isdev) {
1124 		dosys = 0;
1125 		ioctlf = fp->f_ops->fo_ioctl;
1126 		pt.com = SCARG(uap, com);
1127 		pt.data = SCARG(uap, data);
1128 		error = ioctlf(fp, PTIOCLINUX, &pt);
1129 		/*
1130 		 * XXX hack: if the function returns EJUSTRETURN,
1131 		 * it has stuffed a sysctl return value in pt.data.
1132 		 */
1133 		if (error == EJUSTRETURN) {
1134 			retval[0] = (register_t)pt.data;
1135 			error = 0;
1136 		}
1137 		goto out;
1138 	}
1139 
1140 	com = SCARG(uap, com);
1141 	retval[0] = 0;
1142 
1143 	switch (com) {
1144 	case LINUX_SIOCGIFCONF:
1145 		SCARG(&ia, com) = OOSIOCGIFCONF;
1146 		break;
1147 	case LINUX_SIOCGIFFLAGS:
1148 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1149 		break;
1150 	case LINUX_SIOCSIFFLAGS:
1151 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1152 		break;
1153 	case LINUX_SIOCGIFADDR:
1154 		SCARG(&ia, com) = OOSIOCGIFADDR;
1155 		break;
1156 	case LINUX_SIOCGIFDSTADDR:
1157 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1158 		break;
1159 	case LINUX_SIOCGIFBRDADDR:
1160 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1161 		break;
1162 	case LINUX_SIOCGIFNETMASK:
1163 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1164 		break;
1165 	case LINUX_SIOCADDMULTI:
1166 		SCARG(&ia, com) = OSIOCADDMULTI;
1167 		break;
1168 	case LINUX_SIOCDELMULTI:
1169 		SCARG(&ia, com) = OSIOCDELMULTI;
1170 		break;
1171 	case LINUX_SIOCGIFHWADDR:
1172 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1173 		    SCARG(uap, data));
1174 		dosys = 0;
1175 		break;
1176 	default:
1177 		error = EINVAL;
1178 	}
1179 
1180  out:
1181  	fd_putfile(SCARG(uap, fd));
1182 
1183 	if (error ==0 && dosys) {
1184 		SCARG(&ia, fd) = SCARG(uap, fd);
1185 		SCARG(&ia, data) = SCARG(uap, data);
1186 		error = sys_ioctl(curlwp, &ia, retval);
1187 	}
1188 
1189 	return error;
1190 }
1191 
1192 int
1193 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1194 {
1195 	/* {
1196 		syscallarg(int) s;
1197 		syscallarg(const struct sockaddr *) name;
1198 		syscallarg(int) namelen;
1199 	} */
1200 	int		error;
1201 	struct mbuf *nam;
1202 
1203 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1204 	    SCARG(uap, namelen));
1205 	if (error)
1206 		return (error);
1207 
1208 	error = do_sys_connect(l, SCARG(uap, s), nam);
1209 
1210 	if (error == EISCONN) {
1211 		struct socket *so;
1212 		int s, state, prflags, nbio;
1213 
1214 		/* getsock() will use the descriptor for us */
1215 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1216 		    	return EISCONN;
1217 
1218 		s = splsoftnet();
1219 		state = so->so_state;
1220 		nbio = so->so_nbio;
1221 		prflags = so->so_proto->pr_flags;
1222 		splx(s);
1223 		fd_putfile(SCARG(uap, s));
1224 		/*
1225 		 * We should only let this call succeed once per
1226 		 * non-blocking connect; however we don't have
1227 		 * a convenient place to keep that state..
1228 		 */
1229 		if (nbio && (state & SS_ISCONNECTED) &&
1230 		    (prflags & PR_CONNREQUIRED))
1231 			return 0;
1232 	}
1233 
1234 	return (error);
1235 }
1236 
1237 int
1238 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1239 {
1240 	/* {
1241 		syscallarg(int) s;
1242 		syscallarg(const struct osockaddr *) name;
1243 		syscallarg(int) namelen;
1244 	} */
1245 	int		error;
1246 	struct mbuf     *nam;
1247 
1248 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1249 	    SCARG(uap, namelen));
1250 	if (error)
1251 		return (error);
1252 
1253 	return do_sys_bind(l, SCARG(uap, s), nam);
1254 }
1255 
1256 int
1257 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1258 {
1259 	/* {
1260 		syscallarg(int) fdes;
1261 		syscallarg(void *) asa;
1262 		syscallarg(int *) alen;
1263 	} */
1264 	int error;
1265 
1266 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1267 		return (error);
1268 
1269 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1270 		return (error);
1271 
1272 	return (0);
1273 }
1274 
1275 int
1276 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1277 {
1278 	/* {
1279 		syscallarg(int) fdes;
1280 		syscallarg(void *) asa;
1281 		syscallarg(int *) alen;
1282 	} */
1283 	int error;
1284 
1285 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1286 		return (error);
1287 
1288 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1289 		return (error);
1290 
1291 	return (0);
1292 }
1293 
1294 /*
1295  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1296  * family and convert to sockaddr.
1297  */
1298 static int
1299 linux_get_sa(struct lwp *l, int s, struct mbuf **mp, const struct osockaddr *osa, int salen)
1300 {
1301 	int error, bdom;
1302 	struct sockaddr *sa;
1303 	struct osockaddr *kosa;
1304 	struct mbuf *m;
1305 
1306 	if (salen == 1 || salen > UCHAR_MAX) {
1307 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1308 		return EINVAL;
1309 	}
1310 
1311 	/* We'll need the address in an mbuf later, so copy into one here */
1312 	m = m_get(M_WAIT, MT_SONAME);
1313 	if (salen > MLEN)
1314 		MEXTMALLOC(m, salen, M_WAITOK);
1315 
1316 	m->m_len = salen;
1317 
1318 	if (salen == 0)
1319 		return 0;
1320 
1321 	kosa = mtod(m, void *);
1322 	if ((error = copyin(osa, kosa, salen))) {
1323 		DPRINTF(("error %d copying osa %p len %d\n",
1324 				error, osa, salen));
1325 		goto bad;
1326 	}
1327 
1328 	ktrkuser("linux sockaddr", kosa, salen);
1329 
1330 	bdom = linux_to_bsd_domain(kosa->sa_family);
1331 	if (bdom == -1) {
1332 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1333 		error = EINVAL;
1334 		goto bad;
1335 	}
1336 
1337 	/*
1338 	 * If the family is unspecified, use address family of the socket.
1339 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1340 	 */
1341 	if (bdom == AF_UNSPEC) {
1342 		struct socket *so;
1343 
1344 		/* getsock() will use the descriptor for us */
1345 		if ((error = fd_getsock(s, &so)) != 0)
1346 			goto bad;
1347 
1348 		bdom = so->so_proto->pr_domain->dom_family;
1349 		fd_putfile(s);
1350 
1351 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1352 	}
1353 
1354 #ifdef INET6
1355 	/*
1356 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1357 	 * which lacks the scope id compared with RFC2553 one. If we detect
1358 	 * the situation, reject the address and write a message to system log.
1359 	 *
1360 	 * Still accept addresses for which the scope id is not used.
1361 	 */
1362 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1363 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1364 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1365 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1366 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1367 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1368 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1369 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1370 			struct proc *p = l->l_proc;
1371 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1372 
1373 			log(LOG_DEBUG,
1374 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1375 			    "sockaddr_in6 rejected",
1376 			    p->p_pid, p->p_comm, uid);
1377 			error = EINVAL;
1378 			goto bad;
1379 		}
1380 		salen = sizeof (struct sockaddr_in6);
1381 		sin6->sin6_scope_id = 0;
1382 	}
1383 #endif
1384 
1385 	if (bdom == AF_INET)
1386 		salen = sizeof(struct sockaddr_in);
1387 
1388 	sa = (struct sockaddr *) kosa;
1389 	sa->sa_family = bdom;
1390 	sa->sa_len = salen;
1391 	m->m_len = salen;
1392 	ktrkuser("new sockaddr", kosa, salen);
1393 
1394 #ifdef DEBUG_LINUX
1395 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1396 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1397 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1398 	DPRINTF(("\n"));
1399 #endif
1400 
1401 	*mp = m;
1402 	return 0;
1403 
1404     bad:
1405 	m_free(m);
1406 	return error;
1407 }
1408 
1409 static int
1410 linux_sa_put(struct osockaddr *osa)
1411 {
1412 	struct sockaddr sa;
1413 	struct osockaddr *kosa;
1414 	int error, bdom, len;
1415 
1416 	/*
1417 	 * Only read/write the sockaddr family and length part, the rest is
1418 	 * not changed.
1419 	 */
1420 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1421 
1422 	error = copyin(osa, &sa, len);
1423 	if (error)
1424 		return (error);
1425 
1426 	bdom = bsd_to_linux_domain(sa.sa_family);
1427 	if (bdom == -1)
1428 		return (EINVAL);
1429 
1430 	/* Note: we convert from sockaddr to osockaddr here, too */
1431 	kosa = (struct osockaddr *) &sa;
1432 	kosa->sa_family = bdom;
1433 	error = copyout(kosa, osa, len);
1434 	if (error)
1435 		return (error);
1436 
1437 	return (0);
1438 }
1439 
1440 #ifndef __amd64__
1441 int
1442 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1443 {
1444 	/* {
1445 		syscallarg(int) s;
1446 		syscallarg(void *) buf;
1447 		syscallarg(int) len;
1448 		syscallarg(int) flags;
1449 	} */
1450 	struct sys_recvfrom_args bra;
1451 
1452 
1453 	SCARG(&bra, s) = SCARG(uap, s);
1454 	SCARG(&bra, buf) = SCARG(uap, buf);
1455 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1456 	SCARG(&bra, flags) = SCARG(uap, flags);
1457 	SCARG(&bra, from) = NULL;
1458 	SCARG(&bra, fromlenaddr) = NULL;
1459 
1460 	return (sys_recvfrom(l, &bra, retval));
1461 }
1462 
1463 int
1464 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1465 {
1466 	/* {
1467 		syscallarg(int) s;
1468 		syscallarg(void *) buf;
1469 		syscallarg(int) len;
1470 		syscallarg(int) flags;
1471 	} */
1472 	struct sys_sendto_args bsa;
1473 
1474 	SCARG(&bsa, s)		= SCARG(uap, s);
1475 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1476 	SCARG(&bsa, len)	= SCARG(uap, len);
1477 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1478 	SCARG(&bsa, to)		= NULL;
1479 	SCARG(&bsa, tolen)	= 0;
1480 
1481 	return (sys_sendto(l, &bsa, retval));
1482 }
1483 #endif
1484 
1485 int
1486 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1487 {
1488 	/* {
1489 		syscallarg(int) s;
1490 		syscallarg(struct osockaddr *) name;
1491 		syscallarg(int *) anamelen;
1492 	} */
1493 	int error;
1494 	struct sys_accept_args baa;
1495 
1496 	SCARG(&baa, s)		= SCARG(uap, s);
1497 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1498 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1499 
1500 	if ((error = sys_accept(l, &baa, retval)))
1501 		return (error);
1502 
1503 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1504 		return (error);
1505 
1506 	return (0);
1507 }
1508