xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision 60dbe74596d99c974a4936eec2674e13fae0a482)
1 /*	$NetBSD: linux_socket.c,v 1.79 2007/08/29 22:33:42 dyoung Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 1998 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 1.79 2007/08/29 22:33:42 dyoung 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 
79 #include <sys/syscallargs.h>
80 #include <sys/ktrace.h>
81 
82 #include <lib/libkern/libkern.h>
83 
84 #ifdef INET6
85 #include <netinet/ip6.h>
86 #include <netinet6/ip6_var.h>
87 #endif
88 
89 #include <compat/sys/socket.h>
90 #include <compat/sys/sockio.h>
91 
92 #include <compat/linux/common/linux_types.h>
93 #include <compat/linux/common/linux_util.h>
94 #include <compat/linux/common/linux_signal.h>
95 #include <compat/linux/common/linux_ioctl.h>
96 #include <compat/linux/common/linux_socket.h>
97 #if !defined(__alpha__) && !defined(__amd64__)
98 #include <compat/linux/common/linux_socketcall.h>
99 #endif
100 #include <compat/linux/common/linux_sockio.h>
101 
102 #include <compat/linux/linux_syscallargs.h>
103 
104 #ifdef DEBUG_LINUX
105 #define DPRINTF(a) uprintf a
106 #else
107 #define DPRINTF(a)
108 #endif
109 
110 /*
111  * The calls in this file are entered either via the linux_socketcall()
112  * interface or, on the Alpha, as individual syscalls.  The
113  * linux_socketcall function does any massaging of arguments so that all
114  * the calls in here need not think that they are anything other
115  * than a normal syscall.
116  */
117 
118 static int linux_to_bsd_domain __P((int));
119 static int bsd_to_linux_domain __P((int));
120 int linux_to_bsd_sopt_level __P((int));
121 int linux_to_bsd_so_sockopt __P((int));
122 int linux_to_bsd_ip_sockopt __P((int));
123 int linux_to_bsd_tcp_sockopt __P((int));
124 int linux_to_bsd_udp_sockopt __P((int));
125 int linux_getifhwaddr __P((struct lwp *, register_t *, u_int, void *));
126 static int linux_get_sa(struct lwp *, int, struct mbuf **,
127 		const struct osockaddr *, int);
128 static int linux_sa_put __P((struct osockaddr *osa));
129 static int linux_to_bsd_msg_flags __P((int));
130 static int bsd_to_linux_msg_flags __P((int));
131 
132 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
133 	AF_UNSPEC,
134 	AF_UNIX,
135 	AF_INET,
136 	AF_CCITT,	/* LINUX_AF_AX25 */
137 	AF_IPX,
138 	AF_APPLETALK,
139 	-1,		/* LINUX_AF_NETROM */
140 	-1,		/* LINUX_AF_BRIDGE */
141 	-1,		/* LINUX_AF_ATMPVC */
142 	AF_CCITT,	/* LINUX_AF_X25 */
143 	AF_INET6,
144 	-1,		/* LINUX_AF_ROSE */
145 	AF_DECnet,
146 	-1,		/* LINUX_AF_NETBEUI */
147 	-1,		/* LINUX_AF_SECURITY */
148 	pseudo_AF_KEY,
149 	AF_ROUTE,	/* LINUX_AF_NETLINK */
150 	-1,		/* LINUX_AF_PACKET */
151 	-1,		/* LINUX_AF_ASH */
152 	-1,		/* LINUX_AF_ECONET */
153 	-1,		/* LINUX_AF_ATMSVC */
154 	AF_SNA,
155 	/* rest up to LINUX_AF_MAX-1 is not allocated */
156 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
157 };
158 
159 static const int bsd_to_linux_domain_[AF_MAX] = {
160 	LINUX_AF_UNSPEC,
161 	LINUX_AF_UNIX,
162 	LINUX_AF_INET,
163 	-1,		/* AF_IMPLINK */
164 	-1,		/* AF_PUP */
165 	-1,		/* AF_CHAOS */
166 	-1,		/* AF_NS */
167 	-1,		/* AF_ISO */
168 	-1,		/* AF_ECMA */
169 	-1,		/* AF_DATAKIT */
170 	LINUX_AF_AX25,	/* AF_CCITT */
171 	LINUX_AF_SNA,
172 	LINUX_AF_DECnet,
173 	-1,		/* AF_DLI */
174 	-1,		/* AF_LAT */
175 	-1,		/* AF_HYLINK */
176 	LINUX_AF_APPLETALK,
177 	LINUX_AF_NETLINK,
178 	-1,		/* AF_LINK */
179 	-1,		/* AF_XTP */
180 	-1,		/* AF_COIP */
181 	-1,		/* AF_CNT */
182 	-1,		/* pseudo_AF_RTIP */
183 	LINUX_AF_IPX,
184 	LINUX_AF_INET6,
185 	-1,		/* pseudo_AF_PIP */
186 	-1,		/* AF_ISDN */
187 	-1,		/* AF_NATM */
188 	-1,		/* AF_ARP */
189 	LINUX_pseudo_AF_KEY,
190 	-1,		/* pseudo_AF_HDRCMPLT */
191 };
192 
193 static const struct {
194 	int bfl;
195 	int lfl;
196 } bsd_to_linux_msg_flags_[] = {
197 	{MSG_OOB,		LINUX_MSG_OOB},
198 	{MSG_PEEK,		LINUX_MSG_PEEK},
199 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
200 	{MSG_EOR,		LINUX_MSG_EOR},
201 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
202 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
203 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
204 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
205 	{MSG_BCAST,		0},		/* not supported, clear */
206 	{MSG_MCAST,		0},		/* not supported, clear */
207 	{-1, /* not supp */	LINUX_MSG_PROBE},
208 	{-1, /* not supp */	LINUX_MSG_FIN},
209 	{-1, /* not supp */	LINUX_MSG_SYN},
210 	{-1, /* not supp */	LINUX_MSG_CONFIRM},
211 	{-1, /* not supp */	LINUX_MSG_RST},
212 	{-1, /* not supp */	LINUX_MSG_ERRQUEUE},
213 	{-1, /* not supp */	LINUX_MSG_NOSIGNAL},
214 	{-1, /* not supp */	LINUX_MSG_MORE},
215 };
216 
217 /*
218  * Convert between Linux and BSD socket domain values
219  */
220 static int
221 linux_to_bsd_domain(ldom)
222 	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(bdom)
235 	int bdom;
236 {
237 	if (bdom < 0 || bdom >= AF_MAX)
238 		return (-1);
239 
240 	return bsd_to_linux_domain_[bdom];
241 }
242 
243 static int
244 linux_to_bsd_msg_flags(lflag)
245 	int lflag;
246 {
247 	int i, lfl, bfl;
248 	int bflag = 0;
249 
250 	if (lflag == 0)
251 		return (0);
252 
253 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
254 		bfl = bsd_to_linux_msg_flags_[i].bfl;
255 		lfl = bsd_to_linux_msg_flags_[i].lfl;
256 
257 		if (lfl == 0)
258 			continue;
259 
260 		if (lflag & lfl) {
261 			if (bfl < 0)
262 				return (-1);
263 
264 			bflag |= bfl;
265 		}
266 	}
267 
268 	return (bflag);
269 }
270 
271 static int
272 bsd_to_linux_msg_flags(bflag)
273 	int bflag;
274 {
275 	int i, lfl, bfl;
276 	int lflag = 0;
277 
278 	if (bflag == 0)
279 		return (0);
280 
281 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
282 		bfl = bsd_to_linux_msg_flags_[i].bfl;
283 		lfl = bsd_to_linux_msg_flags_[i].lfl;
284 
285 		if (bfl <= 0)
286 			continue;
287 
288 		if (bflag & bfl) {
289 			if (lfl < 0)
290 				return (-1);
291 
292 			lflag |= lfl;
293 		}
294 	}
295 
296 	return (lflag);
297 }
298 
299 int
300 linux_sys_socket(l, v, retval)
301 	struct lwp *l;
302 	void *v;
303 	register_t *retval;
304 {
305 	struct linux_sys_socket_args /* {
306 		syscallarg(int)	domain;
307 		syscallarg(int)	type;
308 		syscallarg(int) protocol;
309 	} */ *uap = v;
310 	struct compat_30_sys_socket_args bsa;
311 	int error;
312 
313 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
314 	SCARG(&bsa, type) = SCARG(uap, type);
315 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
316 	if (SCARG(&bsa, domain) == -1)
317 		return EINVAL;
318 	error = sys___socket30(l, &bsa, retval);
319 
320 #ifdef INET6
321 	/*
322 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
323 	 * default and some apps depend on this. So, set V6ONLY to 0
324 	 * for Linux apps if the sysctl value is set to 1.
325 	 */
326 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
327 		struct proc *p = l->l_proc;
328 		struct file *fp;
329 
330 		if (getsock(p->p_fd, *retval, &fp) == 0) {
331 			struct mbuf *m;
332 
333 			m = m_get(M_WAIT, MT_SOOPTS);
334 			m->m_len = sizeof(int);
335 			*mtod(m, int *) = 0;
336 
337 			/* ignore error */
338 			(void) sosetopt((struct socket *)fp->f_data,
339 				IPPROTO_IPV6, IPV6_V6ONLY, m);
340 
341 			FILE_UNUSE(fp, l);
342 		}
343 	}
344 #endif
345 
346 	return (error);
347 }
348 
349 int
350 linux_sys_socketpair(l, v, retval)
351 	struct lwp *l;
352 	void *v;
353 	register_t *retval;
354 {
355 	struct linux_sys_socketpair_args /* {
356 		syscallarg(int) domain;
357 		syscallarg(int) type;
358 		syscallarg(int) protocol;
359 		syscallarg(int *) rsv;
360 	} */ *uap = v;
361 	struct sys_socketpair_args bsa;
362 
363 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
364 	if (SCARG(&bsa, domain) == -1)
365 		return EINVAL;
366 	SCARG(&bsa, type) = SCARG(uap, type);
367 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
368 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
369 
370 	return sys_socketpair(l, &bsa, retval);
371 }
372 
373 int
374 linux_sys_sendto(l, v, retval)
375 	struct lwp *l;
376 	void *v;
377 	register_t *retval;
378 {
379 	struct linux_sys_sendto_args /* {
380 		syscallarg(int)				s;
381 		syscallarg(void *)			msg;
382 		syscallarg(int)				len;
383 		syscallarg(int)				flags;
384 		syscallarg(struct osockaddr *)		to;
385 		syscallarg(int)				tolen;
386 	} */ *uap = v;
387 	struct msghdr   msg;
388 	struct iovec    aiov;
389 	struct mbuf *nam;
390 	int bflags;
391 	int error;
392 
393 	/* Translate message flags.  */
394 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
395 	if (bflags < 0)
396 		/* Some supported flag */
397 		return EINVAL;
398 
399 	/* Read in and convert the sockaddr */
400 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
401 	    SCARG(uap, tolen));
402 	if (error)
403 		return (error);
404 	msg.msg_flags = MSG_NAMEMBUF;
405 
406 	msg.msg_name = nam;
407 	msg.msg_namelen = SCARG(uap, tolen);
408 	msg.msg_iov = &aiov;
409 	msg.msg_iovlen = 1;
410 	msg.msg_control = 0;
411 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
412 	aiov.iov_len = SCARG(uap, len);
413 
414 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
415 }
416 
417 int
418 linux_sys_sendmsg(l, v, retval)
419 	struct lwp *l;
420 	void *v;
421 	register_t *retval;
422 {
423 	struct linux_sys_sendmsg_args /* {
424 		syscallarg(int) s;
425 		syscallarg(struct msghdr *) msg;
426 		syscallarg(u_int) flags;
427 	} */ *uap = v;
428 	struct msghdr	msg;
429 	int		error;
430 	int		bflags;
431 	struct mbuf     *nam;
432 	u_int8_t	*control;
433 	struct mbuf     *ctl_mbuf = NULL;
434 
435 	msg.msg_flags = MSG_IOVUSRSPACE;
436 
437 	/*
438 	 * Translate message flags.
439 	 */
440 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
441 	if (bflags < 0)
442 		/* Some supported flag */
443 		return EINVAL;
444 
445 	if (msg.msg_name) {
446 		/* Read in and convert the sockaddr */
447 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
448 		    msg.msg_namelen);
449 		if (error)
450 			return (error);
451 		msg.msg_flags |= MSG_NAMEMBUF;
452 		msg.msg_name = nam;
453 	}
454 
455 	/*
456 	 * Handle cmsg if there is any.
457 	 */
458 	if (CMSG_FIRSTHDR(&msg)) {
459 		struct linux_cmsghdr l_cmsg, *l_cc;
460 		struct cmsghdr *cmsg;
461 		ssize_t resid = msg.msg_controllen;
462 		size_t clen, cidx = 0, cspace;
463 
464 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
465 		clen = MLEN;
466 		control = mtod(ctl_mbuf, void *);
467 
468 		l_cc = LINUX_CMSG_FIRSTHDR(&msg);
469 		do {
470 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
471 			if (error)
472 				goto done;
473 
474 			/*
475 			 * Sanity check the control message length.
476 			 */
477 			if (l_cmsg.cmsg_len > resid
478 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
479 				error = EINVAL;
480 				goto done;
481 			}
482 
483 			/*
484 			 * Refuse unsupported control messages, and
485 			 * translate fields as appropriate.
486 			 */
487 			switch (l_cmsg.cmsg_level) {
488 			case LINUX_SOL_SOCKET:
489 				/* It only differs on some archs */
490 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
491 					l_cmsg.cmsg_level = SOL_SOCKET;
492 
493 				switch(l_cmsg.cmsg_type) {
494 				case LINUX_SCM_RIGHTS:
495 					/* Linux SCM_RIGHTS is same as NetBSD */
496 					break;
497 
498 				default:
499 					/* other types not supported */
500 					error = EINVAL;
501 					goto done;
502 				}
503 				break;
504 			default:
505 				/* pray and leave intact */
506 				break;
507 			}
508 
509 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
510 
511 			/* Check the buffer is big enough */
512 			if (__predict_false(cidx + cspace > clen)) {
513 				u_int8_t *nc;
514 
515 				clen = cidx + cspace;
516 				if (clen >= PAGE_SIZE) {
517 					error = EINVAL;
518 					goto done;
519 				}
520 				nc = realloc(clen <= MLEN ? NULL : control,
521 						clen, M_TEMP, M_WAITOK);
522 				if (!nc) {
523 					error = ENOMEM;
524 					goto done;
525 				}
526 				if (cidx <= MLEN)
527 					/* Old buffer was in mbuf... */
528 					memcpy(nc, control, cidx);
529 				control = nc;
530 			}
531 
532 			/* Copy header */
533 			cmsg = (void *)&control[cidx];
534 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
535 			cmsg->cmsg_level = l_cmsg.cmsg_level;
536 			cmsg->cmsg_type = l_cmsg.cmsg_type;
537 
538 			/* Zero are between header and data */
539 			memset(cmsg + 1, 0,
540 				CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg));
541 
542 			/* Copyin the data */
543 			error = copyin(LINUX_CMSG_DATA(l_cc),
544 				CMSG_DATA(control),
545 				l_cmsg.cmsg_len - sizeof(l_cmsg));
546 			if (error)
547 				goto done;
548 
549 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
550 			cidx += cspace;
551 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
552 
553 		/* If we allocated a buffer, attach to mbuf */
554 		if (cidx > MLEN) {
555 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
556 			ctl_mbuf->m_flags |= M_EXT_RW;
557 		}
558 		control = NULL;
559 		ctl_mbuf->m_len = cidx;
560 
561 		msg.msg_control = ctl_mbuf;
562 		msg.msg_flags |= MSG_CONTROLMBUF;
563 	}
564 
565 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
566 	/* Freed internally */
567 	ctl_mbuf = NULL;
568 
569 done:
570 	if (ctl_mbuf != NULL) {
571 		if (control != NULL && control != mtod(ctl_mbuf, void *))
572 			free(control, M_MBUF);
573 		m_free(ctl_mbuf);
574 	}
575 	return (error);
576 }
577 
578 int
579 linux_sys_recvfrom(l, v, retval)
580 	struct lwp *l;
581 	void *v;
582 	register_t *retval;
583 {
584 	struct linux_sys_recvfrom_args /* {
585 		syscallarg(int) s;
586 		syscallarg(void *) buf;
587 		syscallarg(int) len;
588 		syscallarg(int) flags;
589 		syscallarg(struct osockaddr *) from;
590 		syscallarg(int *) fromlenaddr;
591 	} */ *uap = v;
592 	int		error;
593 	struct sys_recvfrom_args bra;
594 
595 	SCARG(&bra, s) = SCARG(uap, s);
596 	SCARG(&bra, buf) = SCARG(uap, buf);
597 	SCARG(&bra, len) = SCARG(uap, len);
598 	SCARG(&bra, flags) = SCARG(uap, flags);
599 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
600 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
601 
602 	if ((error = sys_recvfrom(l, &bra, retval)))
603 		return (error);
604 
605 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
606 		return (error);
607 
608 	return (0);
609 }
610 
611 static int
612 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
613 {
614 	int dlen, error = 0;
615 	struct cmsghdr *cmsg;
616 	struct linux_cmsghdr linux_cmsg;
617 	struct mbuf *m;
618 	char *q, *q_end;
619 
620 	if (mp->msg_controllen <= 0 || control == 0) {
621 		mp->msg_controllen = 0;
622 		free_control_mbuf(l, control, control);
623 		return 0;
624 	}
625 
626 	q = (char *)mp->msg_control;
627 	q_end = q + mp->msg_controllen;
628 
629 	for (m = control; m != NULL; ) {
630 		cmsg = mtod(m, struct cmsghdr *);
631 
632 		/*
633 		 * Fixup cmsg. We handle two things:
634 		 * 0. different sizeof cmsg_len.
635 		 * 1. different values for level/type on some archs
636 		 * 2. different alignment of CMSG_DATA on some archs
637 		 */
638 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
639 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
640 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
641 
642 		dlen = q_end - q;
643 		if (linux_cmsg.cmsg_len > dlen) {
644 			/* Not enough room for the parameter */
645 			dlen -= sizeof linux_cmsg;
646 			if (dlen <= 0)
647 				/* Discard if header wont fit */
648 				break;
649 			mp->msg_flags |= MSG_CTRUNC;
650 			if (linux_cmsg.cmsg_level == SOL_SOCKET
651 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
652 				/* Do not truncate me ... */
653 				break;
654 		} else
655 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
656 
657 		switch (linux_cmsg.cmsg_level) {
658 		case SOL_SOCKET:
659 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
660 			switch (linux_cmsg.cmsg_type) {
661 			case SCM_RIGHTS:
662 				/* Linux SCM_RIGHTS is same as NetBSD */
663 				break;
664 
665 			default:
666 				/* other types not supported */
667 				error = EINVAL;
668 				goto done;
669 			}
670 			/* machine dependant ! */
671 			break;
672 		default:
673 			/* pray and leave intact */
674 			break;
675 		}
676 
677 		/* There can be padding between the header and data... */
678 		error = copyout(&linux_cmsg, q, sizeof *cmsg);
679 		if (error != 0) {
680 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
681 			    dlen);
682 		}
683 		if (error != 0) {
684 			/* We must free all the SCM_RIGHTS */
685 			m = control;
686 			break;
687 		}
688 		m = m->m_next;
689 		if (m == NULL || q + LINUX_CMSG_ALIGN(dlen) > q_end) {
690 			q += dlen;
691 			break;
692 		}
693 		q += LINUX_CMSG_ALIGN(dlen);
694 	}
695 
696   done:
697 	free_control_mbuf(l, control, m);
698 
699 	mp->msg_controllen = q - (char *)mp->msg_control;
700 	return error;
701 }
702 
703 int
704 linux_sys_recvmsg(l, v, retval)
705 	struct lwp *l;
706 	void *v;
707 	register_t *retval;
708 {
709 	struct linux_sys_recvmsg_args /* {
710 		syscallarg(int) s;
711 		syscallarg(struct msghdr *) msg;
712 		syscallarg(u_int) flags;
713 	} */ *uap = v;
714 	struct msghdr	msg;
715 	int		error;
716 	struct mbuf	*from, *control;
717 
718 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
719 	if (error)
720 		return (error);
721 
722 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
723 	if (msg.msg_flags < 0) {
724 		/* Some unsupported flag */
725 		return (EINVAL);
726 	}
727 	msg.msg_flags |= MSG_IOVUSRSPACE;
728 
729 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
730 	    msg.msg_control != NULL ? &control : NULL, retval);
731 	if (error != 0)
732 		return error;
733 
734 	if (msg.msg_control != NULL)
735 		error = linux_copyout_msg_control(l, &msg, control);
736 
737 	if (error == 0 && from != 0) {
738 		mtod(from, struct osockaddr *)->sa_family =
739 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
740 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
741 			from);
742 	} else
743 		msg.msg_namelen = 0;
744 
745 	if (from != NULL)
746 		m_free(from);
747 
748 	if (error == 0) {
749 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
750 		if (msg.msg_flags < 0)
751 			/* Some flag unsupported by Linux */
752 			error = EINVAL;
753 		else
754 			error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
755 	}
756 
757 	return (error);
758 }
759 
760 /*
761  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
762  * is different, the rest matches IPPROTO_* on both systems.
763  */
764 int
765 linux_to_bsd_sopt_level(llevel)
766 	int llevel;
767 {
768 
769 	switch (llevel) {
770 	case LINUX_SOL_SOCKET:
771 		return SOL_SOCKET;
772 	case LINUX_SOL_IP:
773 		return IPPROTO_IP;
774 	case LINUX_SOL_TCP:
775 		return IPPROTO_TCP;
776 	case LINUX_SOL_UDP:
777 		return IPPROTO_UDP;
778 	default:
779 		return -1;
780 	}
781 }
782 
783 /*
784  * Convert Linux socket level socket option numbers to NetBSD values.
785  */
786 int
787 linux_to_bsd_so_sockopt(lopt)
788 	int lopt;
789 {
790 
791 	switch (lopt) {
792 	case LINUX_SO_DEBUG:
793 		return SO_DEBUG;
794 	case LINUX_SO_REUSEADDR:
795 		/*
796 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
797 		 * host:port pair through SO_REUSEADDR even if the address is not a
798 		 * multicast-address.  Effectively, this means that we should use
799 		 * SO_REUSEPORT to allow Linux applications to not exit with
800 		 * EADDRINUSE
801 		 */
802 		return SO_REUSEPORT;
803 	case LINUX_SO_TYPE:
804 		return SO_TYPE;
805 	case LINUX_SO_ERROR:
806 		return SO_ERROR;
807 	case LINUX_SO_DONTROUTE:
808 		return SO_DONTROUTE;
809 	case LINUX_SO_BROADCAST:
810 		return SO_BROADCAST;
811 	case LINUX_SO_SNDBUF:
812 		return SO_SNDBUF;
813 	case LINUX_SO_RCVBUF:
814 		return SO_RCVBUF;
815 	case LINUX_SO_KEEPALIVE:
816 		return SO_KEEPALIVE;
817 	case LINUX_SO_OOBINLINE:
818 		return SO_OOBINLINE;
819 	case LINUX_SO_LINGER:
820 		return SO_LINGER;
821 	case LINUX_SO_PRIORITY:
822 	case LINUX_SO_NO_CHECK:
823 	default:
824 		return -1;
825 	}
826 }
827 
828 /*
829  * Convert Linux IP level socket option number to NetBSD values.
830  */
831 int
832 linux_to_bsd_ip_sockopt(lopt)
833 	int lopt;
834 {
835 
836 	switch (lopt) {
837 	case LINUX_IP_TOS:
838 		return IP_TOS;
839 	case LINUX_IP_TTL:
840 		return IP_TTL;
841 	case LINUX_IP_MULTICAST_TTL:
842 		return IP_MULTICAST_TTL;
843 	case LINUX_IP_MULTICAST_LOOP:
844 		return IP_MULTICAST_LOOP;
845 	case LINUX_IP_MULTICAST_IF:
846 		return IP_MULTICAST_IF;
847 	case LINUX_IP_ADD_MEMBERSHIP:
848 		return IP_ADD_MEMBERSHIP;
849 	case LINUX_IP_DROP_MEMBERSHIP:
850 		return IP_DROP_MEMBERSHIP;
851 	default:
852 		return -1;
853 	}
854 }
855 
856 /*
857  * Convert Linux TCP level socket option number to NetBSD values.
858  */
859 int
860 linux_to_bsd_tcp_sockopt(lopt)
861 	int lopt;
862 {
863 
864 	switch (lopt) {
865 	case LINUX_TCP_NODELAY:
866 		return TCP_NODELAY;
867 	case LINUX_TCP_MAXSEG:
868 		return TCP_MAXSEG;
869 	default:
870 		return -1;
871 	}
872 }
873 
874 /*
875  * Convert Linux UDP level socket option number to NetBSD values.
876  */
877 int
878 linux_to_bsd_udp_sockopt(lopt)
879 	int lopt;
880 {
881 
882 	switch (lopt) {
883 	default:
884 		return -1;
885 	}
886 }
887 
888 /*
889  * Another reasonably straightforward function: setsockopt(2).
890  * The level and option numbers are converted; the values passed
891  * are not (yet) converted, the ones currently implemented don't
892  * need conversion, as they are the same on both systems.
893  */
894 int
895 linux_sys_setsockopt(l, v, retval)
896 	struct lwp *l;
897 	void *v;
898 	register_t *retval;
899 {
900 	struct linux_sys_setsockopt_args /* {
901 		syscallarg(int) s;
902 		syscallarg(int) level;
903 		syscallarg(int) optname;
904 		syscallarg(void *) optval;
905 		syscallarg(int) optlen;
906 	} */ *uap = v;
907 	struct proc *p = l->l_proc;
908 	struct sys_setsockopt_args bsa;
909 	int name;
910 
911 	SCARG(&bsa, s) = SCARG(uap, s);
912 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
913 	SCARG(&bsa, val) = SCARG(uap, optval);
914 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
915 
916 	/*
917 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
918 	 * and returns EOPNOTSUPP for other levels
919 	 */
920 	if (SCARG(&bsa, level) != SOL_SOCKET) {
921 		struct file *fp;
922 		struct socket *so;
923 		int error, s, family;
924 
925 		/* getsock() will use the descriptor for us */
926 	    	if ((error = getsock(p->p_fd, SCARG(&bsa, s), &fp)) != 0)
927 		    	return error;
928 
929 		s = splsoftnet();
930 		so = (struct socket *)fp->f_data;
931 		family = so->so_proto->pr_domain->dom_family;
932 		splx(s);
933 		FILE_UNUSE(fp, l);
934 
935 		if (family == AF_LOCAL)
936 			return EOPNOTSUPP;
937 	}
938 
939 	switch (SCARG(&bsa, level)) {
940 	case SOL_SOCKET:
941 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
942 		break;
943 	case IPPROTO_IP:
944 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
945 		break;
946 	case IPPROTO_TCP:
947 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
948 		break;
949 	case IPPROTO_UDP:
950 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
951 		break;
952 	default:
953 		return EINVAL;
954 	}
955 
956 	if (name == -1)
957 		return EINVAL;
958 	SCARG(&bsa, name) = name;
959 
960 	return sys_setsockopt(l, &bsa, retval);
961 }
962 
963 /*
964  * getsockopt(2) is very much the same as setsockopt(2) (see above)
965  */
966 int
967 linux_sys_getsockopt(l, v, retval)
968 	struct lwp *l;
969 	void *v;
970 	register_t *retval;
971 {
972 	struct linux_sys_getsockopt_args /* {
973 		syscallarg(int) s;
974 		syscallarg(int) level;
975 		syscallarg(int) optname;
976 		syscallarg(void *) optval;
977 		syscallarg(int *) optlen;
978 	} */ *uap = v;
979 	struct sys_getsockopt_args bga;
980 	int name;
981 
982 	SCARG(&bga, s) = SCARG(uap, s);
983 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
984 	SCARG(&bga, val) = SCARG(uap, optval);
985 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
986 
987 	switch (SCARG(&bga, level)) {
988 	case SOL_SOCKET:
989 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
990 		break;
991 	case IPPROTO_IP:
992 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
993 		break;
994 	case IPPROTO_TCP:
995 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
996 		break;
997 	case IPPROTO_UDP:
998 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
999 		break;
1000 	default:
1001 		return EINVAL;
1002 	}
1003 
1004 	if (name == -1)
1005 		return EINVAL;
1006 	SCARG(&bga, name) = name;
1007 
1008 	return sys_getsockopt(l, &bga, retval);
1009 }
1010 
1011 #define IF_NAME_LEN 16
1012 
1013 int
1014 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1015     void *data)
1016 {
1017 	/* Not the full structure, just enough to map what we do here */
1018 	struct linux_ifreq {
1019 		char if_name[IF_NAME_LEN];
1020 		struct osockaddr hwaddr;
1021 	} lreq;
1022 	struct proc *p = l->l_proc;
1023 	struct filedesc *fdp;
1024 	struct file *fp;
1025 	struct ifaddr *ifa;
1026 	struct ifnet *ifp;
1027 	struct sockaddr_dl *sadl;
1028 	int error, found;
1029 	int index, ifnum;
1030 
1031 	/*
1032 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1033 	 * SIOCGIFCONF, because the user buffer is not of the right
1034 	 * type to take those results.  We can't use kernel buffers to
1035 	 * receive the results, as the implementation of sys_ioctl()
1036 	 * and ifconf() [which implements SIOCGIFCONF] use
1037 	 * copyin()/copyout() which will fail on kernel addresses.
1038 	 *
1039 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1040 	 */
1041 
1042 	fdp = p->p_fd;
1043 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1044 		return (EBADF);
1045 
1046 	FILE_USE(fp);
1047 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1048 		error = EBADF;
1049 		goto out;
1050 	}
1051 
1052 	error = copyin(data, &lreq, sizeof(lreq));
1053 	if (error)
1054 		goto out;
1055 	lreq.if_name[IF_NAME_LEN-1] = '\0';		/* just in case */
1056 
1057 	/*
1058 	 * Try real interface name first, then fake "ethX"
1059 	 */
1060 	for (ifp = ifnet.tqh_first, found = 0;
1061 	     ifp != 0 && !found;
1062 	     ifp = ifp->if_list.tqe_next) {
1063 		if (strcmp(lreq.if_name, ifp->if_xname))
1064 			/* not this interface */
1065 			continue;
1066 		found=1;
1067 		if ((ifa = ifp->if_addrlist.tqh_first) != 0) {
1068 			for (; ifa != 0; ifa = ifa->ifa_list.tqe_next) {
1069 				sadl = satosdl(ifa->ifa_addr);
1070 				/* only return ethernet addresses */
1071 				/* XXX what about FDDI, etc. ? */
1072 				if (sadl->sdl_family != AF_LINK ||
1073 				    sadl->sdl_type != IFT_ETHER)
1074 					continue;
1075 				memcpy(&lreq.hwaddr.sa_data, CLLADDR(sadl),
1076 				       MIN(sadl->sdl_alen,
1077 					   sizeof(lreq.hwaddr.sa_data)));
1078 				lreq.hwaddr.sa_family =
1079 					sadl->sdl_family;
1080 				error = copyout(&lreq, data, sizeof(lreq));
1081 				goto out;
1082 			}
1083 		} else {
1084 			error = ENODEV;
1085 			goto out;
1086 		}
1087 	}
1088 
1089 	if (strncmp(lreq.if_name, "eth", 3) == 0) {
1090 		for (ifnum = 0, index = 3;
1091 		     lreq.if_name[index] != '\0' && index < IF_NAME_LEN;
1092 		     index++) {
1093 			ifnum *= 10;
1094 			ifnum += lreq.if_name[index] - '0';
1095 		}
1096 
1097 		error = EINVAL;			/* in case we don't find one */
1098 		for (ifp = ifnet.tqh_first, found = 0;
1099 		     ifp != 0 && !found;
1100 		     ifp = ifp->if_list.tqe_next) {
1101 			memcpy(lreq.if_name, ifp->if_xname,
1102 			       MIN(IF_NAME_LEN, IFNAMSIZ));
1103 			if ((ifa = ifp->if_addrlist.tqh_first) == 0)
1104 				/* no addresses on this interface */
1105 				continue;
1106 			else
1107 				for (; ifa != 0; ifa = ifa->ifa_list.tqe_next) {
1108 					sadl = satosdl(ifa->ifa_addr);
1109 					/* only return ethernet addresses */
1110 					/* XXX what about FDDI, etc. ? */
1111 					if (sadl->sdl_family != AF_LINK ||
1112 					    sadl->sdl_type != IFT_ETHER)
1113 						continue;
1114 					if (ifnum--)
1115 						/* not the reqested iface */
1116 						continue;
1117 					memcpy(&lreq.hwaddr.sa_data,
1118 					       CLLADDR(sadl),
1119 					       MIN(sadl->sdl_alen,
1120 						   sizeof(lreq.hwaddr.sa_data)));
1121 					lreq.hwaddr.sa_family =
1122 						sadl->sdl_family;
1123 					error = copyout(&lreq, data, sizeof(lreq));
1124 					found = 1;
1125 					break;
1126 				}
1127 		}
1128 	} else {
1129 		/* unknown interface, not even an "eth*" name */
1130 		error = ENODEV;
1131 	}
1132 
1133 out:
1134 	FILE_UNUSE(fp, l);
1135 	return error;
1136 }
1137 #undef IF_NAME_LEN
1138 
1139 int
1140 linux_ioctl_socket(l, uap, retval)
1141 	struct lwp *l;
1142 	struct linux_sys_ioctl_args /* {
1143 		syscallarg(int) fd;
1144 		syscallarg(u_long) com;
1145 		syscallarg(void *) data;
1146 	} */ *uap;
1147 	register_t *retval;
1148 {
1149 	struct proc *p = l->l_proc;
1150 	u_long com;
1151 	int error = 0, isdev = 0, dosys = 1;
1152 	struct sys_ioctl_args ia;
1153 	struct file *fp;
1154 	struct filedesc *fdp;
1155 	struct vnode *vp;
1156 	int (*ioctlf)(struct file *, u_long, void *, struct lwp *);
1157 	struct ioctl_pt pt;
1158 
1159 	fdp = p->p_fd;
1160 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1161 		return (EBADF);
1162 
1163 	FILE_USE(fp);
1164 
1165 	if (fp->f_type == DTYPE_VNODE) {
1166 		vp = (struct vnode *)fp->f_data;
1167 		isdev = vp->v_type == VCHR;
1168 	}
1169 
1170 	/*
1171 	 * Don't try to interpret socket ioctl calls that are done
1172 	 * on a device filedescriptor, just pass them through, to
1173 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1174 	 * device will only handle these if it's prepared to do
1175 	 * so, to avoid unexpected things from happening.
1176 	 */
1177 	if (isdev) {
1178 		dosys = 0;
1179 		ioctlf = fp->f_ops->fo_ioctl;
1180 		pt.com = SCARG(uap, com);
1181 		pt.data = SCARG(uap, data);
1182 		error = ioctlf(fp, PTIOCLINUX, (void *)&pt, l);
1183 		/*
1184 		 * XXX hack: if the function returns EJUSTRETURN,
1185 		 * it has stuffed a sysctl return value in pt.data.
1186 		 */
1187 		if (error == EJUSTRETURN) {
1188 			retval[0] = (register_t)pt.data;
1189 			error = 0;
1190 		}
1191 		goto out;
1192 	}
1193 
1194 	com = SCARG(uap, com);
1195 	retval[0] = 0;
1196 
1197 	switch (com) {
1198 	case LINUX_SIOCGIFCONF:
1199 		SCARG(&ia, com) = OOSIOCGIFCONF;
1200 		break;
1201 	case LINUX_SIOCGIFFLAGS:
1202 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1203 		break;
1204 	case LINUX_SIOCSIFFLAGS:
1205 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1206 		break;
1207 	case LINUX_SIOCGIFADDR:
1208 		SCARG(&ia, com) = OOSIOCGIFADDR;
1209 		break;
1210 	case LINUX_SIOCGIFDSTADDR:
1211 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1212 		break;
1213 	case LINUX_SIOCGIFBRDADDR:
1214 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1215 		break;
1216 	case LINUX_SIOCGIFNETMASK:
1217 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1218 		break;
1219 	case LINUX_SIOCADDMULTI:
1220 		SCARG(&ia, com) = OSIOCADDMULTI;
1221 		break;
1222 	case LINUX_SIOCDELMULTI:
1223 		SCARG(&ia, com) = OSIOCDELMULTI;
1224 		break;
1225 	case LINUX_SIOCGIFHWADDR:
1226 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1227 					 SCARG(uap, data));
1228 		dosys = 0;
1229 		break;
1230 	default:
1231 		error = EINVAL;
1232 	}
1233 
1234 out:
1235 	FILE_UNUSE(fp, l);
1236 
1237 	if (error ==0 && dosys) {
1238 		SCARG(&ia, fd) = SCARG(uap, fd);
1239 		SCARG(&ia, data) = SCARG(uap, data);
1240 		/* XXX NJWLWP */
1241 		error = sys_ioctl(curlwp, &ia, retval);
1242 	}
1243 
1244 	return error;
1245 }
1246 
1247 int
1248 linux_sys_connect(l, v, retval)
1249 	struct lwp *l;
1250 	void *v;
1251 	register_t *retval;
1252 {
1253 	struct linux_sys_connect_args /* {
1254 		syscallarg(int) s;
1255 		syscallarg(const struct sockaddr *) name;
1256 		syscallarg(int) namelen;
1257 	} */ *uap = v;
1258 	int		error;
1259 	struct mbuf *nam;
1260 
1261 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1262 	    SCARG(uap, namelen));
1263 	if (error)
1264 		return (error);
1265 
1266 	error = do_sys_connect(l, SCARG(uap, s), nam);
1267 
1268 	if (error == EISCONN) {
1269 		struct file *fp;
1270 		struct socket *so;
1271 		int s, state, prflags;
1272 
1273 		/* getsock() will use the descriptor for us */
1274 	    	if (getsock(l->l_proc->p_fd, SCARG(uap, s), &fp) != 0)
1275 		    	return EISCONN;
1276 
1277 		s = splsoftnet();
1278 		so = (struct socket *)fp->f_data;
1279 		state = so->so_state;
1280 		prflags = so->so_proto->pr_flags;
1281 		splx(s);
1282 		FILE_UNUSE(fp, l);
1283 		/*
1284 		 * We should only let this call succeed once per
1285 		 * non-blocking connect; however we don't have
1286 		 * a convenient place to keep that state..
1287 		 */
1288 		if ((state & SS_NBIO) && (state & SS_ISCONNECTED) &&
1289 		    (prflags & PR_CONNREQUIRED))
1290 			return 0;
1291 	}
1292 
1293 	return (error);
1294 }
1295 
1296 int
1297 linux_sys_bind(l, v, retval)
1298 	struct lwp *l;
1299 	void *v;
1300 	register_t *retval;
1301 {
1302 	struct linux_sys_bind_args /* {
1303 		syscallarg(int) s;
1304 		syscallarg(const struct osockaddr *) name;
1305 		syscallarg(int) namelen;
1306 	} */ *uap = v;
1307 	int		error;
1308 	struct mbuf     *nam;
1309 
1310 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1311 	    SCARG(uap, namelen));
1312 	if (error)
1313 		return (error);
1314 
1315 	return do_sys_bind(l, SCARG(uap, s), nam);
1316 }
1317 
1318 int
1319 linux_sys_getsockname(l, v, retval)
1320 	struct lwp *l;
1321 	void *v;
1322 	register_t *retval;
1323 {
1324 	struct linux_sys_getsockname_args /* {
1325 		syscallarg(int) fdes;
1326 		syscallarg(void *) asa;
1327 		syscallarg(int *) alen;
1328 	} */ *uap = v;
1329 	int error;
1330 
1331 	if ((error = sys_getsockname(l, uap, retval)) != 0)
1332 		return (error);
1333 
1334 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1335 		return (error);
1336 
1337 	return (0);
1338 }
1339 
1340 int
1341 linux_sys_getpeername(l, v, retval)
1342 	struct lwp *l;
1343 	void *v;
1344 	register_t *retval;
1345 {
1346 	struct sys_getpeername_args /* {
1347 		syscallarg(int) fdes;
1348 		syscallarg(void *) asa;
1349 		syscallarg(int *) alen;
1350 	} */ *uap = v;
1351 	int error;
1352 
1353 	if ((error = sys_getpeername(l, uap, retval)) != 0)
1354 		return (error);
1355 
1356 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1357 		return (error);
1358 
1359 	return (0);
1360 }
1361 
1362 /*
1363  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1364  * family and convert to sockaddr.
1365  */
1366 static int
1367 linux_get_sa(struct lwp *l, int s, struct mbuf **mp, const struct osockaddr *osa, int salen)
1368 {
1369 	int error, bdom;
1370 	struct sockaddr *sa;
1371 	struct osockaddr *kosa;
1372 	struct mbuf *m;
1373 
1374 	if (salen == 1 || salen > UCHAR_MAX) {
1375 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1376 		return EINVAL;
1377 	}
1378 
1379 	/* We'll need the address in an mbuf later, so copy into one here */
1380 	m = m_get(M_WAIT, MT_SONAME);
1381 	if (salen > MLEN)
1382 		MEXTMALLOC(m, salen, M_WAITOK);
1383 
1384 	m->m_len = salen;
1385 
1386 	if (salen == 0)
1387 		return 0;
1388 
1389 	kosa = mtod(m, void *);
1390 	if ((error = copyin(osa, kosa, salen))) {
1391 		DPRINTF(("error %d copying osa %p len %d\n",
1392 				error, osa, salen));
1393 		goto bad;
1394 	}
1395 
1396 	ktrkuser("linux sockaddr", kosa, salen);
1397 
1398 	bdom = linux_to_bsd_domain(kosa->sa_family);
1399 	if (bdom == -1) {
1400 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1401 		error = EINVAL;
1402 		goto bad;
1403 	}
1404 
1405 	/*
1406 	 * If the family is unspecified, use address family of the socket.
1407 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1408 	 */
1409 	if (bdom == AF_UNSPEC) {
1410 		struct file *fp;
1411 		struct socket *so;
1412 
1413 		/* getsock() will use the descriptor for us */
1414 		if ((error = getsock(l->l_proc->p_fd, s, &fp)) != 0)
1415 			goto bad;
1416 
1417 		so = (struct socket *)fp->f_data;
1418 		bdom = so->so_proto->pr_domain->dom_family;
1419 
1420 		FILE_UNUSE(fp, l);
1421 
1422 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1423 	}
1424 
1425 #ifdef INET6
1426 	/*
1427 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1428 	 * which lacks the scope id compared with RFC2553 one. If we detect
1429 	 * the situation, reject the address and write a message to system log.
1430 	 *
1431 	 * Still accept addresses for which the scope id is not used.
1432 	 */
1433 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1434 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1435 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1436 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1437 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1438 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1439 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1440 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1441 			struct proc *p = l->l_proc;
1442 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1443 
1444 			log(LOG_DEBUG,
1445 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1446 			    "sockaddr_in6 rejected",
1447 			    p->p_pid, p->p_comm, uid);
1448 			error = EINVAL;
1449 			goto bad;
1450 		}
1451 		salen = sizeof (struct sockaddr_in6);
1452 		sin6->sin6_scope_id = 0;
1453 	}
1454 #endif
1455 
1456 	if (bdom == AF_INET)
1457 		salen = sizeof(struct sockaddr_in);
1458 
1459 	sa = (struct sockaddr *) kosa;
1460 	sa->sa_family = bdom;
1461 	sa->sa_len = salen;
1462 	m->m_len = salen;
1463 	ktrkuser("new sockaddr", kosa, salen);
1464 
1465 #ifdef DEBUG_LINUX
1466 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1467 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1468 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1469 	DPRINTF(("\n"));
1470 #endif
1471 
1472 	*mp = m;
1473 	return 0;
1474 
1475     bad:
1476 	m_free(m);
1477 	return error;
1478 }
1479 
1480 static int
1481 linux_sa_put(osa)
1482 	struct osockaddr *osa;
1483 {
1484 	struct sockaddr sa;
1485 	struct osockaddr *kosa;
1486 	int error, bdom, len;
1487 
1488 	/*
1489 	 * Only read/write the sockaddr family and length part, the rest is
1490 	 * not changed.
1491 	 */
1492 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1493 
1494 	error = copyin(osa, &sa, len);
1495 	if (error)
1496 		return (error);
1497 
1498 	bdom = bsd_to_linux_domain(sa.sa_family);
1499 	if (bdom == -1)
1500 		return (EINVAL);
1501 
1502 	/* Note: we convert from sockaddr to osockaddr here, too */
1503 	kosa = (struct osockaddr *) &sa;
1504 	kosa->sa_family = bdom;
1505 	error = copyout(kosa, osa, len);
1506 	if (error)
1507 		return (error);
1508 
1509 	return (0);
1510 }
1511 
1512 #ifndef __amd64__
1513 int
1514 linux_sys_recv(l, v, retval)
1515 	struct lwp *l;
1516 	void *v;
1517 	register_t *retval;
1518 {
1519 	struct linux_sys_recv_args /* {
1520 		syscallarg(int) s;
1521 		syscallarg(void *) buf;
1522 		syscallarg(int) len;
1523 		syscallarg(int) flags;
1524 	} */ *uap = v;
1525 	struct sys_recvfrom_args bra;
1526 
1527 
1528 	SCARG(&bra, s) = SCARG(uap, s);
1529 	SCARG(&bra, buf) = SCARG(uap, buf);
1530 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1531 	SCARG(&bra, flags) = SCARG(uap, flags);
1532 	SCARG(&bra, from) = NULL;
1533 	SCARG(&bra, fromlenaddr) = NULL;
1534 
1535 	return (sys_recvfrom(l, &bra, retval));
1536 }
1537 
1538 int
1539 linux_sys_send(l, v, retval)
1540 	struct lwp *l;
1541 	void *v;
1542 	register_t *retval;
1543 {
1544 	struct linux_sys_send_args /* {
1545 		syscallarg(int) s;
1546 		syscallarg(void *) buf;
1547 		syscallarg(int) len;
1548 		syscallarg(int) flags;
1549 	} */ *uap = v;
1550 	struct sys_sendto_args bsa;
1551 
1552 	SCARG(&bsa, s)		= SCARG(uap, s);
1553 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1554 	SCARG(&bsa, len)	= SCARG(uap, len);
1555 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1556 	SCARG(&bsa, to)		= NULL;
1557 	SCARG(&bsa, tolen)	= 0;
1558 
1559 	return (sys_sendto(l, &bsa, retval));
1560 }
1561 #endif
1562 
1563 int
1564 linux_sys_accept(l, v, retval)
1565 	struct lwp *l;
1566 	void *v;
1567 	register_t *retval;
1568 {
1569 	struct linux_sys_accept_args /* {
1570 		syscallarg(int) s;
1571 		syscallarg(struct osockaddr *) name;
1572 		syscallarg(int *) anamelen;
1573 	} */ *uap = v;
1574 	int error;
1575 	struct sys_accept_args baa;
1576 
1577 	SCARG(&baa, s)		= SCARG(uap, s);
1578 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1579 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1580 
1581 	if ((error = sys_accept(l, &baa, retval)))
1582 		return (error);
1583 
1584 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1585 		return (error);
1586 
1587 	return (0);
1588 }
1589