xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision ce099b40997c43048fb78bd578195f81d2456523)
1 /*	$NetBSD: linux_socket.c,v 1.92 2008/04/28 20:23:44 martin 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 1.92 2008/04/28 20:23:44 martin 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 *, 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 	/* Read in and convert the sockaddr */
379 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
380 	    SCARG(uap, tolen));
381 	if (error)
382 		return (error);
383 	msg.msg_flags = MSG_NAMEMBUF;
384 
385 	msg.msg_name = nam;
386 	msg.msg_namelen = SCARG(uap, tolen);
387 	msg.msg_iov = &aiov;
388 	msg.msg_iovlen = 1;
389 	msg.msg_control = 0;
390 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
391 	aiov.iov_len = SCARG(uap, len);
392 
393 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
394 }
395 
396 int
397 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
398 {
399 	/* {
400 		syscallarg(int) s;
401 		syscallarg(struct msghdr *) msg;
402 		syscallarg(u_int) flags;
403 	} */
404 	struct msghdr	msg;
405 	int		error;
406 	int		bflags;
407 	struct mbuf     *nam;
408 	u_int8_t	*control;
409 	struct mbuf     *ctl_mbuf = NULL;
410 
411 	msg.msg_flags = MSG_IOVUSRSPACE;
412 
413 	/*
414 	 * Translate message flags.
415 	 */
416 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
417 	if (bflags < 0)
418 		/* Some supported flag */
419 		return EINVAL;
420 
421 	if (msg.msg_name) {
422 		/* Read in and convert the sockaddr */
423 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
424 		    msg.msg_namelen);
425 		if (error)
426 			return (error);
427 		msg.msg_flags |= MSG_NAMEMBUF;
428 		msg.msg_name = nam;
429 	}
430 
431 	/*
432 	 * Handle cmsg if there is any.
433 	 */
434 	if (CMSG_FIRSTHDR(&msg)) {
435 		struct linux_cmsghdr l_cmsg, *l_cc;
436 		struct cmsghdr *cmsg;
437 		ssize_t resid = msg.msg_controllen;
438 		size_t clen, cidx = 0, cspace;
439 
440 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
441 		clen = MLEN;
442 		control = mtod(ctl_mbuf, void *);
443 
444 		l_cc = LINUX_CMSG_FIRSTHDR(&msg);
445 		do {
446 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
447 			if (error)
448 				goto done;
449 
450 			/*
451 			 * Sanity check the control message length.
452 			 */
453 			if (l_cmsg.cmsg_len > resid
454 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
455 				error = EINVAL;
456 				goto done;
457 			}
458 
459 			/*
460 			 * Refuse unsupported control messages, and
461 			 * translate fields as appropriate.
462 			 */
463 			switch (l_cmsg.cmsg_level) {
464 			case LINUX_SOL_SOCKET:
465 				/* It only differs on some archs */
466 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
467 					l_cmsg.cmsg_level = SOL_SOCKET;
468 
469 				switch(l_cmsg.cmsg_type) {
470 				case LINUX_SCM_RIGHTS:
471 					/* Linux SCM_RIGHTS is same as NetBSD */
472 					break;
473 
474 				default:
475 					/* other types not supported */
476 					error = EINVAL;
477 					goto done;
478 				}
479 				break;
480 			default:
481 				/* pray and leave intact */
482 				break;
483 			}
484 
485 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
486 
487 			/* Check the buffer is big enough */
488 			if (__predict_false(cidx + cspace > clen)) {
489 				u_int8_t *nc;
490 
491 				clen = cidx + cspace;
492 				if (clen >= PAGE_SIZE) {
493 					error = EINVAL;
494 					goto done;
495 				}
496 				nc = realloc(clen <= MLEN ? NULL : control,
497 						clen, M_TEMP, M_WAITOK);
498 				if (!nc) {
499 					error = ENOMEM;
500 					goto done;
501 				}
502 				if (cidx <= MLEN)
503 					/* Old buffer was in mbuf... */
504 					memcpy(nc, control, cidx);
505 				control = nc;
506 			}
507 
508 			/* Copy header */
509 			cmsg = (void *)&control[cidx];
510 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
511 			cmsg->cmsg_level = l_cmsg.cmsg_level;
512 			cmsg->cmsg_type = l_cmsg.cmsg_type;
513 
514 			/* Zero are between header and data */
515 			memset(cmsg + 1, 0,
516 				CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg));
517 
518 			/* Copyin the data */
519 			error = copyin(LINUX_CMSG_DATA(l_cc),
520 				CMSG_DATA(control),
521 				l_cmsg.cmsg_len - sizeof(l_cmsg));
522 			if (error)
523 				goto done;
524 
525 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
526 			cidx += cspace;
527 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
528 
529 		/* If we allocated a buffer, attach to mbuf */
530 		if (cidx > MLEN) {
531 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
532 			ctl_mbuf->m_flags |= M_EXT_RW;
533 		}
534 		control = NULL;
535 		ctl_mbuf->m_len = cidx;
536 
537 		msg.msg_control = ctl_mbuf;
538 		msg.msg_flags |= MSG_CONTROLMBUF;
539 	}
540 
541 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
542 	/* Freed internally */
543 	ctl_mbuf = NULL;
544 
545 done:
546 	if (ctl_mbuf != NULL) {
547 		if (control != NULL && control != mtod(ctl_mbuf, void *))
548 			free(control, M_MBUF);
549 		m_free(ctl_mbuf);
550 	}
551 	return (error);
552 }
553 
554 int
555 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
556 {
557 	/* {
558 		syscallarg(int) s;
559 		syscallarg(void *) buf;
560 		syscallarg(int) len;
561 		syscallarg(int) flags;
562 		syscallarg(struct osockaddr *) from;
563 		syscallarg(int *) fromlenaddr;
564 	} */
565 	int		error;
566 	struct sys_recvfrom_args bra;
567 
568 	SCARG(&bra, s) = SCARG(uap, s);
569 	SCARG(&bra, buf) = SCARG(uap, buf);
570 	SCARG(&bra, len) = SCARG(uap, len);
571 	SCARG(&bra, flags) = SCARG(uap, flags);
572 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
573 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
574 
575 	if ((error = sys_recvfrom(l, &bra, retval)))
576 		return (error);
577 
578 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
579 		return (error);
580 
581 	return (0);
582 }
583 
584 static int
585 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
586 {
587 	int dlen, error = 0;
588 	struct cmsghdr *cmsg;
589 	struct linux_cmsghdr linux_cmsg;
590 	struct mbuf *m;
591 	char *q, *q_end;
592 
593 	if (mp->msg_controllen <= 0 || control == 0) {
594 		mp->msg_controllen = 0;
595 		free_control_mbuf(l, control, control);
596 		return 0;
597 	}
598 
599 	q = (char *)mp->msg_control;
600 	q_end = q + mp->msg_controllen;
601 
602 	for (m = control; m != NULL; ) {
603 		cmsg = mtod(m, struct cmsghdr *);
604 
605 		/*
606 		 * Fixup cmsg. We handle two things:
607 		 * 0. different sizeof cmsg_len.
608 		 * 1. different values for level/type on some archs
609 		 * 2. different alignment of CMSG_DATA on some archs
610 		 */
611 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
612 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
613 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
614 
615 		dlen = q_end - q;
616 		if (linux_cmsg.cmsg_len > dlen) {
617 			/* Not enough room for the parameter */
618 			dlen -= sizeof linux_cmsg;
619 			if (dlen <= 0)
620 				/* Discard if header wont fit */
621 				break;
622 			mp->msg_flags |= MSG_CTRUNC;
623 			if (linux_cmsg.cmsg_level == SOL_SOCKET
624 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
625 				/* Do not truncate me ... */
626 				break;
627 		} else
628 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
629 
630 		switch (linux_cmsg.cmsg_level) {
631 		case SOL_SOCKET:
632 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
633 			switch (linux_cmsg.cmsg_type) {
634 			case SCM_RIGHTS:
635 				/* Linux SCM_RIGHTS is same as NetBSD */
636 				break;
637 
638 			default:
639 				/* other types not supported */
640 				error = EINVAL;
641 				goto done;
642 			}
643 			/* machine dependant ! */
644 			break;
645 		default:
646 			/* pray and leave intact */
647 			break;
648 		}
649 
650 		/* There can be padding between the header and data... */
651 		error = copyout(&linux_cmsg, q, sizeof *cmsg);
652 		if (error != 0) {
653 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
654 			    dlen);
655 		}
656 		if (error != 0) {
657 			/* We must free all the SCM_RIGHTS */
658 			m = control;
659 			break;
660 		}
661 		m = m->m_next;
662 		if (m == NULL || q + LINUX_CMSG_ALIGN(dlen) > q_end) {
663 			q += dlen;
664 			break;
665 		}
666 		q += LINUX_CMSG_ALIGN(dlen);
667 	}
668 
669   done:
670 	free_control_mbuf(l, control, m);
671 
672 	mp->msg_controllen = q - (char *)mp->msg_control;
673 	return error;
674 }
675 
676 int
677 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
678 {
679 	/* {
680 		syscallarg(int) s;
681 		syscallarg(struct msghdr *) msg;
682 		syscallarg(u_int) flags;
683 	} */
684 	struct msghdr	msg;
685 	int		error;
686 	struct mbuf	*from, *control;
687 
688 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
689 	if (error)
690 		return (error);
691 
692 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
693 	if (msg.msg_flags < 0) {
694 		/* Some unsupported flag */
695 		return (EINVAL);
696 	}
697 	msg.msg_flags |= MSG_IOVUSRSPACE;
698 
699 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
700 	    msg.msg_control != NULL ? &control : NULL, retval);
701 	if (error != 0)
702 		return error;
703 
704 	if (msg.msg_control != NULL)
705 		error = linux_copyout_msg_control(l, &msg, control);
706 
707 	if (error == 0 && from != 0) {
708 		mtod(from, struct osockaddr *)->sa_family =
709 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
710 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
711 			from);
712 	} else
713 		msg.msg_namelen = 0;
714 
715 	if (from != NULL)
716 		m_free(from);
717 
718 	if (error == 0) {
719 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
720 		if (msg.msg_flags < 0)
721 			/* Some flag unsupported by Linux */
722 			error = EINVAL;
723 		else
724 			error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
725 	}
726 
727 	return (error);
728 }
729 
730 /*
731  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
732  * is different, the rest matches IPPROTO_* on both systems.
733  */
734 int
735 linux_to_bsd_sopt_level(int llevel)
736 {
737 
738 	switch (llevel) {
739 	case LINUX_SOL_SOCKET:
740 		return SOL_SOCKET;
741 	case LINUX_SOL_IP:
742 		return IPPROTO_IP;
743 	case LINUX_SOL_TCP:
744 		return IPPROTO_TCP;
745 	case LINUX_SOL_UDP:
746 		return IPPROTO_UDP;
747 	default:
748 		return -1;
749 	}
750 }
751 
752 /*
753  * Convert Linux socket level socket option numbers to NetBSD values.
754  */
755 int
756 linux_to_bsd_so_sockopt(int lopt)
757 {
758 
759 	switch (lopt) {
760 	case LINUX_SO_DEBUG:
761 		return SO_DEBUG;
762 	case LINUX_SO_REUSEADDR:
763 		/*
764 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
765 		 * host:port pair through SO_REUSEADDR even if the address is not a
766 		 * multicast-address.  Effectively, this means that we should use
767 		 * SO_REUSEPORT to allow Linux applications to not exit with
768 		 * EADDRINUSE
769 		 */
770 		return SO_REUSEPORT;
771 	case LINUX_SO_TYPE:
772 		return SO_TYPE;
773 	case LINUX_SO_ERROR:
774 		return SO_ERROR;
775 	case LINUX_SO_DONTROUTE:
776 		return SO_DONTROUTE;
777 	case LINUX_SO_BROADCAST:
778 		return SO_BROADCAST;
779 	case LINUX_SO_SNDBUF:
780 		return SO_SNDBUF;
781 	case LINUX_SO_RCVBUF:
782 		return SO_RCVBUF;
783 	case LINUX_SO_KEEPALIVE:
784 		return SO_KEEPALIVE;
785 	case LINUX_SO_OOBINLINE:
786 		return SO_OOBINLINE;
787 	case LINUX_SO_LINGER:
788 		return SO_LINGER;
789 	case LINUX_SO_PRIORITY:
790 	case LINUX_SO_NO_CHECK:
791 	default:
792 		return -1;
793 	}
794 }
795 
796 /*
797  * Convert Linux IP level socket option number to NetBSD values.
798  */
799 int
800 linux_to_bsd_ip_sockopt(int lopt)
801 {
802 
803 	switch (lopt) {
804 	case LINUX_IP_TOS:
805 		return IP_TOS;
806 	case LINUX_IP_TTL:
807 		return IP_TTL;
808 	case LINUX_IP_MULTICAST_TTL:
809 		return IP_MULTICAST_TTL;
810 	case LINUX_IP_MULTICAST_LOOP:
811 		return IP_MULTICAST_LOOP;
812 	case LINUX_IP_MULTICAST_IF:
813 		return IP_MULTICAST_IF;
814 	case LINUX_IP_ADD_MEMBERSHIP:
815 		return IP_ADD_MEMBERSHIP;
816 	case LINUX_IP_DROP_MEMBERSHIP:
817 		return IP_DROP_MEMBERSHIP;
818 	default:
819 		return -1;
820 	}
821 }
822 
823 /*
824  * Convert Linux TCP level socket option number to NetBSD values.
825  */
826 int
827 linux_to_bsd_tcp_sockopt(int lopt)
828 {
829 
830 	switch (lopt) {
831 	case LINUX_TCP_NODELAY:
832 		return TCP_NODELAY;
833 	case LINUX_TCP_MAXSEG:
834 		return TCP_MAXSEG;
835 	default:
836 		return -1;
837 	}
838 }
839 
840 /*
841  * Convert Linux UDP level socket option number to NetBSD values.
842  */
843 int
844 linux_to_bsd_udp_sockopt(int lopt)
845 {
846 
847 	switch (lopt) {
848 	default:
849 		return -1;
850 	}
851 }
852 
853 /*
854  * Another reasonably straightforward function: setsockopt(2).
855  * The level and option numbers are converted; the values passed
856  * are not (yet) converted, the ones currently implemented don't
857  * need conversion, as they are the same on both systems.
858  */
859 int
860 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
861 {
862 	/* {
863 		syscallarg(int) s;
864 		syscallarg(int) level;
865 		syscallarg(int) optname;
866 		syscallarg(void *) optval;
867 		syscallarg(int) optlen;
868 	} */
869 	struct sys_setsockopt_args bsa;
870 	int name;
871 
872 	SCARG(&bsa, s) = SCARG(uap, s);
873 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
874 	SCARG(&bsa, val) = SCARG(uap, optval);
875 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
876 
877 	/*
878 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
879 	 * and returns EOPNOTSUPP for other levels
880 	 */
881 	if (SCARG(&bsa, level) != SOL_SOCKET) {
882 		struct socket *so;
883 		int error, family;
884 
885 		/* getsock() will use the descriptor for us */
886 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
887 		    	return error;
888 		family = so->so_proto->pr_domain->dom_family;
889 		fd_putfile(SCARG(&bsa, s));
890 
891 		if (family == AF_LOCAL)
892 			return EOPNOTSUPP;
893 	}
894 
895 	switch (SCARG(&bsa, level)) {
896 	case SOL_SOCKET:
897 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
898 		break;
899 	case IPPROTO_IP:
900 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
901 		break;
902 	case IPPROTO_TCP:
903 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
904 		break;
905 	case IPPROTO_UDP:
906 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
907 		break;
908 	default:
909 		return EINVAL;
910 	}
911 
912 	if (name == -1)
913 		return EINVAL;
914 	SCARG(&bsa, name) = name;
915 
916 	return sys_setsockopt(l, &bsa, retval);
917 }
918 
919 /*
920  * getsockopt(2) is very much the same as setsockopt(2) (see above)
921  */
922 int
923 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
924 {
925 	/* {
926 		syscallarg(int) s;
927 		syscallarg(int) level;
928 		syscallarg(int) optname;
929 		syscallarg(void *) optval;
930 		syscallarg(int *) optlen;
931 	} */
932 	struct sys_getsockopt_args bga;
933 	int name;
934 
935 	SCARG(&bga, s) = SCARG(uap, s);
936 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
937 	SCARG(&bga, val) = SCARG(uap, optval);
938 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
939 
940 	switch (SCARG(&bga, level)) {
941 	case SOL_SOCKET:
942 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
943 		break;
944 	case IPPROTO_IP:
945 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
946 		break;
947 	case IPPROTO_TCP:
948 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
949 		break;
950 	case IPPROTO_UDP:
951 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
952 		break;
953 	default:
954 		return EINVAL;
955 	}
956 
957 	if (name == -1)
958 		return EINVAL;
959 	SCARG(&bga, name) = name;
960 
961 	return sys_getsockopt(l, &bga, retval);
962 }
963 
964 #define IF_NAME_LEN 16
965 
966 int
967 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
968     void *data)
969 {
970 	/* Not the full structure, just enough to map what we do here */
971 	struct linux_ifreq {
972 		char if_name[IF_NAME_LEN];
973 		struct osockaddr hwaddr;
974 	} lreq;
975 	file_t *fp;
976 	struct ifaddr *ifa;
977 	struct ifnet *ifp;
978 	struct sockaddr_dl *sadl;
979 	int error, found;
980 	int index, ifnum;
981 
982 	/*
983 	 * We can't emulate this ioctl by calling sys_ioctl() to run
984 	 * SIOCGIFCONF, because the user buffer is not of the right
985 	 * type to take those results.  We can't use kernel buffers to
986 	 * receive the results, as the implementation of sys_ioctl()
987 	 * and ifconf() [which implements SIOCGIFCONF] use
988 	 * copyin()/copyout() which will fail on kernel addresses.
989 	 *
990 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
991 	 */
992 
993 	if ((fp = fd_getfile(fd)) == NULL)
994 		return (EBADF);
995 
996 	KERNEL_LOCK(1, NULL);
997 
998 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
999 		error = EBADF;
1000 		goto out;
1001 	}
1002 
1003 	error = copyin(data, &lreq, sizeof(lreq));
1004 	if (error)
1005 		goto out;
1006 	lreq.if_name[IF_NAME_LEN-1] = '\0';		/* just in case */
1007 
1008 	/*
1009 	 * Try real interface name first, then fake "ethX"
1010 	 */
1011 	found = 0;
1012 	IFNET_FOREACH(ifp) {
1013 		if (found)
1014 			break;
1015 		if (strcmp(lreq.if_name, ifp->if_xname))
1016 			/* not this interface */
1017 			continue;
1018 		found=1;
1019 		if (IFADDR_EMPTY(ifp)) {
1020 			error = ENODEV;
1021 			goto out;
1022 		}
1023 		IFADDR_FOREACH(ifa, ifp) {
1024 			sadl = satosdl(ifa->ifa_addr);
1025 			/* only return ethernet addresses */
1026 			/* XXX what about FDDI, etc. ? */
1027 			if (sadl->sdl_family != AF_LINK ||
1028 			    sadl->sdl_type != IFT_ETHER)
1029 				continue;
1030 			memcpy(&lreq.hwaddr.sa_data, CLLADDR(sadl),
1031 			       MIN(sadl->sdl_alen,
1032 				   sizeof(lreq.hwaddr.sa_data)));
1033 			lreq.hwaddr.sa_family =
1034 				sadl->sdl_family;
1035 			error = copyout(&lreq, data, sizeof(lreq));
1036 			goto out;
1037 		}
1038 	}
1039 
1040 	if (strncmp(lreq.if_name, "eth", 3) == 0) {
1041 		for (ifnum = 0, index = 3;
1042 		     lreq.if_name[index] != '\0' && index < IF_NAME_LEN;
1043 		     index++) {
1044 			ifnum *= 10;
1045 			ifnum += lreq.if_name[index] - '0';
1046 		}
1047 
1048 		error = EINVAL;			/* in case we don't find one */
1049 		found = 0;
1050 		IFNET_FOREACH(ifp) {
1051 			if (found)
1052 				break;
1053 			memcpy(lreq.if_name, ifp->if_xname,
1054 			       MIN(IF_NAME_LEN, IFNAMSIZ));
1055 			IFADDR_FOREACH(ifa, ifp) {
1056 				sadl = satosdl(ifa->ifa_addr);
1057 				/* only return ethernet addresses */
1058 				/* XXX what about FDDI, etc. ? */
1059 				if (sadl->sdl_family != AF_LINK ||
1060 				    sadl->sdl_type != IFT_ETHER)
1061 					continue;
1062 				if (ifnum--)
1063 					/* not the reqested iface */
1064 					continue;
1065 				memcpy(&lreq.hwaddr.sa_data,
1066 				       CLLADDR(sadl),
1067 				       MIN(sadl->sdl_alen,
1068 					   sizeof(lreq.hwaddr.sa_data)));
1069 				lreq.hwaddr.sa_family =
1070 					sadl->sdl_family;
1071 				error = copyout(&lreq, data, sizeof(lreq));
1072 				found = 1;
1073 				break;
1074 			}
1075 		}
1076 	} else {
1077 		/* unknown interface, not even an "eth*" name */
1078 		error = ENODEV;
1079 	}
1080 
1081 out:
1082 	KERNEL_UNLOCK_ONE(NULL);
1083 	fd_putfile(fd);
1084 	return error;
1085 }
1086 #undef IF_NAME_LEN
1087 
1088 int
1089 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1090 {
1091 	/* {
1092 		syscallarg(int) fd;
1093 		syscallarg(u_long) com;
1094 		syscallarg(void *) data;
1095 	} */
1096 	u_long com;
1097 	int error = 0, isdev = 0, dosys = 1;
1098 	struct sys_ioctl_args ia;
1099 	file_t *fp;
1100 	struct vnode *vp;
1101 	int (*ioctlf)(file_t *, u_long, void *);
1102 	struct ioctl_pt pt;
1103 
1104 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1105 		return (EBADF);
1106 
1107 	if (fp->f_type == DTYPE_VNODE) {
1108 		vp = (struct vnode *)fp->f_data;
1109 		isdev = vp->v_type == VCHR;
1110 	}
1111 
1112 	/*
1113 	 * Don't try to interpret socket ioctl calls that are done
1114 	 * on a device filedescriptor, just pass them through, to
1115 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1116 	 * device will only handle these if it's prepared to do
1117 	 * so, to avoid unexpected things from happening.
1118 	 */
1119 	if (isdev) {
1120 		dosys = 0;
1121 		ioctlf = fp->f_ops->fo_ioctl;
1122 		pt.com = SCARG(uap, com);
1123 		pt.data = SCARG(uap, data);
1124 		error = ioctlf(fp, PTIOCLINUX, &pt);
1125 		/*
1126 		 * XXX hack: if the function returns EJUSTRETURN,
1127 		 * it has stuffed a sysctl return value in pt.data.
1128 		 */
1129 		if (error == EJUSTRETURN) {
1130 			retval[0] = (register_t)pt.data;
1131 			error = 0;
1132 		}
1133 		goto out;
1134 	}
1135 
1136 	com = SCARG(uap, com);
1137 	retval[0] = 0;
1138 
1139 	switch (com) {
1140 	case LINUX_SIOCGIFCONF:
1141 		SCARG(&ia, com) = OOSIOCGIFCONF;
1142 		break;
1143 	case LINUX_SIOCGIFFLAGS:
1144 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1145 		break;
1146 	case LINUX_SIOCSIFFLAGS:
1147 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1148 		break;
1149 	case LINUX_SIOCGIFADDR:
1150 		SCARG(&ia, com) = OOSIOCGIFADDR;
1151 		break;
1152 	case LINUX_SIOCGIFDSTADDR:
1153 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1154 		break;
1155 	case LINUX_SIOCGIFBRDADDR:
1156 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1157 		break;
1158 	case LINUX_SIOCGIFNETMASK:
1159 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1160 		break;
1161 	case LINUX_SIOCADDMULTI:
1162 		SCARG(&ia, com) = OSIOCADDMULTI;
1163 		break;
1164 	case LINUX_SIOCDELMULTI:
1165 		SCARG(&ia, com) = OSIOCDELMULTI;
1166 		break;
1167 	case LINUX_SIOCGIFHWADDR:
1168 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1169 		    SCARG(uap, data));
1170 		dosys = 0;
1171 		break;
1172 	default:
1173 		error = EINVAL;
1174 	}
1175 
1176  out:
1177  	fd_putfile(SCARG(uap, fd));
1178 
1179 	if (error ==0 && dosys) {
1180 		SCARG(&ia, fd) = SCARG(uap, fd);
1181 		SCARG(&ia, data) = SCARG(uap, data);
1182 		error = sys_ioctl(curlwp, &ia, retval);
1183 	}
1184 
1185 	return error;
1186 }
1187 
1188 int
1189 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1190 {
1191 	/* {
1192 		syscallarg(int) s;
1193 		syscallarg(const struct sockaddr *) name;
1194 		syscallarg(int) namelen;
1195 	} */
1196 	int		error;
1197 	struct mbuf *nam;
1198 
1199 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1200 	    SCARG(uap, namelen));
1201 	if (error)
1202 		return (error);
1203 
1204 	error = do_sys_connect(l, SCARG(uap, s), nam);
1205 
1206 	if (error == EISCONN) {
1207 		struct socket *so;
1208 		int state, prflags, nbio;
1209 
1210 		/* getsock() will use the descriptor for us */
1211 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1212 		    	return EISCONN;
1213 
1214 		solock(so);
1215 		state = so->so_state;
1216 		nbio = so->so_nbio;
1217 		prflags = so->so_proto->pr_flags;
1218 		sounlock(so);
1219 		fd_putfile(SCARG(uap, s));
1220 		/*
1221 		 * We should only let this call succeed once per
1222 		 * non-blocking connect; however we don't have
1223 		 * a convenient place to keep that state..
1224 		 */
1225 		if (nbio && (state & SS_ISCONNECTED) &&
1226 		    (prflags & PR_CONNREQUIRED))
1227 			return 0;
1228 	}
1229 
1230 	return (error);
1231 }
1232 
1233 int
1234 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1235 {
1236 	/* {
1237 		syscallarg(int) s;
1238 		syscallarg(const struct osockaddr *) name;
1239 		syscallarg(int) namelen;
1240 	} */
1241 	int		error;
1242 	struct mbuf     *nam;
1243 
1244 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1245 	    SCARG(uap, namelen));
1246 	if (error)
1247 		return (error);
1248 
1249 	return do_sys_bind(l, SCARG(uap, s), nam);
1250 }
1251 
1252 int
1253 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1254 {
1255 	/* {
1256 		syscallarg(int) fdes;
1257 		syscallarg(void *) asa;
1258 		syscallarg(int *) alen;
1259 	} */
1260 	int error;
1261 
1262 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1263 		return (error);
1264 
1265 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1266 		return (error);
1267 
1268 	return (0);
1269 }
1270 
1271 int
1272 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1273 {
1274 	/* {
1275 		syscallarg(int) fdes;
1276 		syscallarg(void *) asa;
1277 		syscallarg(int *) alen;
1278 	} */
1279 	int error;
1280 
1281 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1282 		return (error);
1283 
1284 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1285 		return (error);
1286 
1287 	return (0);
1288 }
1289 
1290 /*
1291  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1292  * family and convert to sockaddr.
1293  */
1294 static int
1295 linux_get_sa(struct lwp *l, int s, struct mbuf **mp, const struct osockaddr *osa, int salen)
1296 {
1297 	int error, bdom;
1298 	struct sockaddr *sa;
1299 	struct osockaddr *kosa;
1300 	struct mbuf *m;
1301 
1302 	if (salen == 1 || salen > UCHAR_MAX) {
1303 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1304 		return EINVAL;
1305 	}
1306 
1307 	/* We'll need the address in an mbuf later, so copy into one here */
1308 	m = m_get(M_WAIT, MT_SONAME);
1309 	if (salen > MLEN)
1310 		MEXTMALLOC(m, salen, M_WAITOK);
1311 
1312 	m->m_len = salen;
1313 
1314 	if (salen == 0)
1315 		return 0;
1316 
1317 	kosa = mtod(m, void *);
1318 	if ((error = copyin(osa, kosa, salen))) {
1319 		DPRINTF(("error %d copying osa %p len %d\n",
1320 				error, osa, salen));
1321 		goto bad;
1322 	}
1323 
1324 	ktrkuser("linux sockaddr", kosa, salen);
1325 
1326 	bdom = linux_to_bsd_domain(kosa->sa_family);
1327 	if (bdom == -1) {
1328 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1329 		error = EINVAL;
1330 		goto bad;
1331 	}
1332 
1333 	/*
1334 	 * If the family is unspecified, use address family of the socket.
1335 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1336 	 */
1337 	if (bdom == AF_UNSPEC) {
1338 		struct socket *so;
1339 
1340 		/* getsock() will use the descriptor for us */
1341 		if ((error = fd_getsock(s, &so)) != 0)
1342 			goto bad;
1343 
1344 		bdom = so->so_proto->pr_domain->dom_family;
1345 		fd_putfile(s);
1346 
1347 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1348 	}
1349 
1350 #ifdef INET6
1351 	/*
1352 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1353 	 * which lacks the scope id compared with RFC2553 one. If we detect
1354 	 * the situation, reject the address and write a message to system log.
1355 	 *
1356 	 * Still accept addresses for which the scope id is not used.
1357 	 */
1358 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1359 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1360 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1361 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1362 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1363 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1364 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1365 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1366 			struct proc *p = l->l_proc;
1367 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1368 
1369 			log(LOG_DEBUG,
1370 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1371 			    "sockaddr_in6 rejected",
1372 			    p->p_pid, p->p_comm, uid);
1373 			error = EINVAL;
1374 			goto bad;
1375 		}
1376 		salen = sizeof (struct sockaddr_in6);
1377 		sin6->sin6_scope_id = 0;
1378 	}
1379 #endif
1380 
1381 	if (bdom == AF_INET)
1382 		salen = sizeof(struct sockaddr_in);
1383 
1384 	sa = (struct sockaddr *) kosa;
1385 	sa->sa_family = bdom;
1386 	sa->sa_len = salen;
1387 	m->m_len = salen;
1388 	ktrkuser("new sockaddr", kosa, salen);
1389 
1390 #ifdef DEBUG_LINUX
1391 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1392 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1393 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1394 	DPRINTF(("\n"));
1395 #endif
1396 
1397 	*mp = m;
1398 	return 0;
1399 
1400     bad:
1401 	m_free(m);
1402 	return error;
1403 }
1404 
1405 static int
1406 linux_sa_put(struct osockaddr *osa)
1407 {
1408 	struct sockaddr sa;
1409 	struct osockaddr *kosa;
1410 	int error, bdom, len;
1411 
1412 	/*
1413 	 * Only read/write the sockaddr family and length part, the rest is
1414 	 * not changed.
1415 	 */
1416 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1417 
1418 	error = copyin(osa, &sa, len);
1419 	if (error)
1420 		return (error);
1421 
1422 	bdom = bsd_to_linux_domain(sa.sa_family);
1423 	if (bdom == -1)
1424 		return (EINVAL);
1425 
1426 	/* Note: we convert from sockaddr to osockaddr here, too */
1427 	kosa = (struct osockaddr *) &sa;
1428 	kosa->sa_family = bdom;
1429 	error = copyout(kosa, osa, len);
1430 	if (error)
1431 		return (error);
1432 
1433 	return (0);
1434 }
1435 
1436 #ifndef __amd64__
1437 int
1438 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1439 {
1440 	/* {
1441 		syscallarg(int) s;
1442 		syscallarg(void *) buf;
1443 		syscallarg(int) len;
1444 		syscallarg(int) flags;
1445 	} */
1446 	struct sys_recvfrom_args bra;
1447 
1448 
1449 	SCARG(&bra, s) = SCARG(uap, s);
1450 	SCARG(&bra, buf) = SCARG(uap, buf);
1451 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1452 	SCARG(&bra, flags) = SCARG(uap, flags);
1453 	SCARG(&bra, from) = NULL;
1454 	SCARG(&bra, fromlenaddr) = NULL;
1455 
1456 	return (sys_recvfrom(l, &bra, retval));
1457 }
1458 
1459 int
1460 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1461 {
1462 	/* {
1463 		syscallarg(int) s;
1464 		syscallarg(void *) buf;
1465 		syscallarg(int) len;
1466 		syscallarg(int) flags;
1467 	} */
1468 	struct sys_sendto_args bsa;
1469 
1470 	SCARG(&bsa, s)		= SCARG(uap, s);
1471 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1472 	SCARG(&bsa, len)	= SCARG(uap, len);
1473 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1474 	SCARG(&bsa, to)		= NULL;
1475 	SCARG(&bsa, tolen)	= 0;
1476 
1477 	return (sys_sendto(l, &bsa, retval));
1478 }
1479 #endif
1480 
1481 int
1482 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1483 {
1484 	/* {
1485 		syscallarg(int) s;
1486 		syscallarg(struct osockaddr *) name;
1487 		syscallarg(int *) anamelen;
1488 	} */
1489 	int error;
1490 	struct sys_accept_args baa;
1491 
1492 	SCARG(&baa, s)		= SCARG(uap, s);
1493 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1494 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1495 
1496 	if ((error = sys_accept(l, &baa, retval)))
1497 		return (error);
1498 
1499 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1500 		return (error);
1501 
1502 	return (0);
1503 }
1504