xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: linux_socket.c,v 1.84 2007/12/08 18:36:10 dsl 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 1.84 2007/12/08 18:36:10 dsl 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 #include <compat/linux/common/linux_ipc.h>
102 #include <compat/linux/common/linux_sem.h>
103 
104 #include <compat/linux/linux_syscallargs.h>
105 
106 #ifdef DEBUG_LINUX
107 #define DPRINTF(a) uprintf a
108 #else
109 #define DPRINTF(a)
110 #endif
111 
112 /*
113  * The calls in this file are entered either via the linux_socketcall()
114  * interface or, on the Alpha, as individual syscalls.  The
115  * linux_socketcall function does any massaging of arguments so that all
116  * the calls in here need not think that they are anything other
117  * than a normal syscall.
118  */
119 
120 static int linux_to_bsd_domain(int);
121 static int bsd_to_linux_domain(int);
122 int linux_to_bsd_sopt_level(int);
123 int linux_to_bsd_so_sockopt(int);
124 int linux_to_bsd_ip_sockopt(int);
125 int linux_to_bsd_tcp_sockopt(int);
126 int linux_to_bsd_udp_sockopt(int);
127 int linux_getifhwaddr(struct lwp *, register_t *, u_int, void *);
128 static int linux_get_sa(struct lwp *, int, struct mbuf **,
129 		const struct osockaddr *, int);
130 static int linux_sa_put(struct osockaddr *osa);
131 static int linux_to_bsd_msg_flags(int);
132 static int bsd_to_linux_msg_flags(int);
133 
134 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
135 	AF_UNSPEC,
136 	AF_UNIX,
137 	AF_INET,
138 	AF_CCITT,	/* LINUX_AF_AX25 */
139 	AF_IPX,
140 	AF_APPLETALK,
141 	-1,		/* LINUX_AF_NETROM */
142 	-1,		/* LINUX_AF_BRIDGE */
143 	-1,		/* LINUX_AF_ATMPVC */
144 	AF_CCITT,	/* LINUX_AF_X25 */
145 	AF_INET6,
146 	-1,		/* LINUX_AF_ROSE */
147 	AF_DECnet,
148 	-1,		/* LINUX_AF_NETBEUI */
149 	-1,		/* LINUX_AF_SECURITY */
150 	pseudo_AF_KEY,
151 	AF_ROUTE,	/* LINUX_AF_NETLINK */
152 	-1,		/* LINUX_AF_PACKET */
153 	-1,		/* LINUX_AF_ASH */
154 	-1,		/* LINUX_AF_ECONET */
155 	-1,		/* LINUX_AF_ATMSVC */
156 	AF_SNA,
157 	/* rest up to LINUX_AF_MAX-1 is not allocated */
158 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
159 };
160 
161 static const int bsd_to_linux_domain_[AF_MAX] = {
162 	LINUX_AF_UNSPEC,
163 	LINUX_AF_UNIX,
164 	LINUX_AF_INET,
165 	-1,		/* AF_IMPLINK */
166 	-1,		/* AF_PUP */
167 	-1,		/* AF_CHAOS */
168 	-1,		/* AF_NS */
169 	-1,		/* AF_ISO */
170 	-1,		/* AF_ECMA */
171 	-1,		/* AF_DATAKIT */
172 	LINUX_AF_AX25,	/* AF_CCITT */
173 	LINUX_AF_SNA,
174 	LINUX_AF_DECnet,
175 	-1,		/* AF_DLI */
176 	-1,		/* AF_LAT */
177 	-1,		/* AF_HYLINK */
178 	LINUX_AF_APPLETALK,
179 	LINUX_AF_NETLINK,
180 	-1,		/* AF_LINK */
181 	-1,		/* AF_XTP */
182 	-1,		/* AF_COIP */
183 	-1,		/* AF_CNT */
184 	-1,		/* pseudo_AF_RTIP */
185 	LINUX_AF_IPX,
186 	LINUX_AF_INET6,
187 	-1,		/* pseudo_AF_PIP */
188 	-1,		/* AF_ISDN */
189 	-1,		/* AF_NATM */
190 	-1,		/* AF_ARP */
191 	LINUX_pseudo_AF_KEY,
192 	-1,		/* pseudo_AF_HDRCMPLT */
193 };
194 
195 static const struct {
196 	int bfl;
197 	int lfl;
198 } bsd_to_linux_msg_flags_[] = {
199 	{MSG_OOB,		LINUX_MSG_OOB},
200 	{MSG_PEEK,		LINUX_MSG_PEEK},
201 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
202 	{MSG_EOR,		LINUX_MSG_EOR},
203 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
204 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
205 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
206 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
207 	{MSG_BCAST,		0},		/* not supported, clear */
208 	{MSG_MCAST,		0},		/* not supported, clear */
209 	{-1, /* not supp */	LINUX_MSG_PROBE},
210 	{-1, /* not supp */	LINUX_MSG_FIN},
211 	{-1, /* not supp */	LINUX_MSG_SYN},
212 	{-1, /* not supp */	LINUX_MSG_CONFIRM},
213 	{-1, /* not supp */	LINUX_MSG_RST},
214 	{-1, /* not supp */	LINUX_MSG_ERRQUEUE},
215 	{-1, /* not supp */	LINUX_MSG_NOSIGNAL},
216 	{-1, /* not supp */	LINUX_MSG_MORE},
217 };
218 
219 /*
220  * Convert between Linux and BSD socket domain values
221  */
222 static int
223 linux_to_bsd_domain(int ldom)
224 {
225 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
226 		return (-1);
227 
228 	return linux_to_bsd_domain_[ldom];
229 }
230 
231 /*
232  * Convert between BSD and Linux socket domain values
233  */
234 static int
235 bsd_to_linux_domain(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(int lflag)
245 {
246 	int i, lfl, bfl;
247 	int bflag = 0;
248 
249 	if (lflag == 0)
250 		return (0);
251 
252 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
253 		bfl = bsd_to_linux_msg_flags_[i].bfl;
254 		lfl = bsd_to_linux_msg_flags_[i].lfl;
255 
256 		if (lfl == 0)
257 			continue;
258 
259 		if (lflag & lfl) {
260 			if (bfl < 0)
261 				return (-1);
262 
263 			bflag |= bfl;
264 		}
265 	}
266 
267 	return (bflag);
268 }
269 
270 static int
271 bsd_to_linux_msg_flags(int bflag)
272 {
273 	int i, lfl, bfl;
274 	int lflag = 0;
275 
276 	if (bflag == 0)
277 		return (0);
278 
279 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
280 		bfl = bsd_to_linux_msg_flags_[i].bfl;
281 		lfl = bsd_to_linux_msg_flags_[i].lfl;
282 
283 		if (bfl <= 0)
284 			continue;
285 
286 		if (bflag & bfl) {
287 			if (lfl < 0)
288 				return (-1);
289 
290 			lflag |= lfl;
291 		}
292 	}
293 
294 	return (lflag);
295 }
296 
297 int
298 linux_sys_socket(struct lwp *l, void *v, register_t *retval)
299 {
300 	struct linux_sys_socket_args /* {
301 		syscallarg(int)	domain;
302 		syscallarg(int)	type;
303 		syscallarg(int) protocol;
304 	} */ *uap = v;
305 	struct compat_30_sys_socket_args bsa;
306 	int error;
307 
308 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
309 	SCARG(&bsa, type) = SCARG(uap, type);
310 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
311 	if (SCARG(&bsa, domain) == -1)
312 		return EINVAL;
313 	error = sys___socket30(l, &bsa, retval);
314 
315 #ifdef INET6
316 	/*
317 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
318 	 * default and some apps depend on this. So, set V6ONLY to 0
319 	 * for Linux apps if the sysctl value is set to 1.
320 	 */
321 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
322 		struct proc *p = l->l_proc;
323 		struct file *fp;
324 
325 		if (getsock(p->p_fd, *retval, &fp) == 0) {
326 			struct mbuf *m;
327 
328 			m = m_get(M_WAIT, MT_SOOPTS);
329 			m->m_len = sizeof(int);
330 			*mtod(m, int *) = 0;
331 
332 			/* ignore error */
333 			(void) sosetopt((struct socket *)fp->f_data,
334 				IPPROTO_IPV6, IPV6_V6ONLY, m);
335 
336 			FILE_UNUSE(fp, l);
337 		}
338 	}
339 #endif
340 
341 	return (error);
342 }
343 
344 int
345 linux_sys_socketpair(struct lwp *l, void *v, register_t *retval)
346 {
347 	struct linux_sys_socketpair_args /* {
348 		syscallarg(int) domain;
349 		syscallarg(int) type;
350 		syscallarg(int) protocol;
351 		syscallarg(int *) rsv;
352 	} */ *uap = v;
353 	struct sys_socketpair_args bsa;
354 
355 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
356 	if (SCARG(&bsa, domain) == -1)
357 		return EINVAL;
358 	SCARG(&bsa, type) = SCARG(uap, type);
359 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
360 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
361 
362 	return sys_socketpair(l, &bsa, retval);
363 }
364 
365 int
366 linux_sys_sendto(struct lwp *l, void *v, register_t *retval)
367 {
368 	struct linux_sys_sendto_args /* {
369 		syscallarg(int)				s;
370 		syscallarg(void *)			msg;
371 		syscallarg(int)				len;
372 		syscallarg(int)				flags;
373 		syscallarg(struct osockaddr *)		to;
374 		syscallarg(int)				tolen;
375 	} */ *uap = v;
376 	struct msghdr   msg;
377 	struct iovec    aiov;
378 	struct mbuf *nam;
379 	int bflags;
380 	int error;
381 
382 	/* Translate message flags.  */
383 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
384 	if (bflags < 0)
385 		/* Some supported flag */
386 		return EINVAL;
387 
388 	/* Read in and convert the sockaddr */
389 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
390 	    SCARG(uap, tolen));
391 	if (error)
392 		return (error);
393 	msg.msg_flags = MSG_NAMEMBUF;
394 
395 	msg.msg_name = nam;
396 	msg.msg_namelen = SCARG(uap, tolen);
397 	msg.msg_iov = &aiov;
398 	msg.msg_iovlen = 1;
399 	msg.msg_control = 0;
400 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
401 	aiov.iov_len = SCARG(uap, len);
402 
403 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
404 }
405 
406 int
407 linux_sys_sendmsg(struct lwp *l, void *v, register_t *retval)
408 {
409 	struct linux_sys_sendmsg_args /* {
410 		syscallarg(int) s;
411 		syscallarg(struct msghdr *) msg;
412 		syscallarg(u_int) flags;
413 	} */ *uap = v;
414 	struct msghdr	msg;
415 	int		error;
416 	int		bflags;
417 	struct mbuf     *nam;
418 	u_int8_t	*control;
419 	struct mbuf     *ctl_mbuf = NULL;
420 
421 	msg.msg_flags = MSG_IOVUSRSPACE;
422 
423 	/*
424 	 * Translate message flags.
425 	 */
426 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
427 	if (bflags < 0)
428 		/* Some supported flag */
429 		return EINVAL;
430 
431 	if (msg.msg_name) {
432 		/* Read in and convert the sockaddr */
433 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
434 		    msg.msg_namelen);
435 		if (error)
436 			return (error);
437 		msg.msg_flags |= MSG_NAMEMBUF;
438 		msg.msg_name = nam;
439 	}
440 
441 	/*
442 	 * Handle cmsg if there is any.
443 	 */
444 	if (CMSG_FIRSTHDR(&msg)) {
445 		struct linux_cmsghdr l_cmsg, *l_cc;
446 		struct cmsghdr *cmsg;
447 		ssize_t resid = msg.msg_controllen;
448 		size_t clen, cidx = 0, cspace;
449 
450 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
451 		clen = MLEN;
452 		control = mtod(ctl_mbuf, void *);
453 
454 		l_cc = LINUX_CMSG_FIRSTHDR(&msg);
455 		do {
456 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
457 			if (error)
458 				goto done;
459 
460 			/*
461 			 * Sanity check the control message length.
462 			 */
463 			if (l_cmsg.cmsg_len > resid
464 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
465 				error = EINVAL;
466 				goto done;
467 			}
468 
469 			/*
470 			 * Refuse unsupported control messages, and
471 			 * translate fields as appropriate.
472 			 */
473 			switch (l_cmsg.cmsg_level) {
474 			case LINUX_SOL_SOCKET:
475 				/* It only differs on some archs */
476 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
477 					l_cmsg.cmsg_level = SOL_SOCKET;
478 
479 				switch(l_cmsg.cmsg_type) {
480 				case LINUX_SCM_RIGHTS:
481 					/* Linux SCM_RIGHTS is same as NetBSD */
482 					break;
483 
484 				default:
485 					/* other types not supported */
486 					error = EINVAL;
487 					goto done;
488 				}
489 				break;
490 			default:
491 				/* pray and leave intact */
492 				break;
493 			}
494 
495 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
496 
497 			/* Check the buffer is big enough */
498 			if (__predict_false(cidx + cspace > clen)) {
499 				u_int8_t *nc;
500 
501 				clen = cidx + cspace;
502 				if (clen >= PAGE_SIZE) {
503 					error = EINVAL;
504 					goto done;
505 				}
506 				nc = realloc(clen <= MLEN ? NULL : control,
507 						clen, M_TEMP, M_WAITOK);
508 				if (!nc) {
509 					error = ENOMEM;
510 					goto done;
511 				}
512 				if (cidx <= MLEN)
513 					/* Old buffer was in mbuf... */
514 					memcpy(nc, control, cidx);
515 				control = nc;
516 			}
517 
518 			/* Copy header */
519 			cmsg = (void *)&control[cidx];
520 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
521 			cmsg->cmsg_level = l_cmsg.cmsg_level;
522 			cmsg->cmsg_type = l_cmsg.cmsg_type;
523 
524 			/* Zero are between header and data */
525 			memset(cmsg + 1, 0,
526 				CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg));
527 
528 			/* Copyin the data */
529 			error = copyin(LINUX_CMSG_DATA(l_cc),
530 				CMSG_DATA(control),
531 				l_cmsg.cmsg_len - sizeof(l_cmsg));
532 			if (error)
533 				goto done;
534 
535 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
536 			cidx += cspace;
537 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
538 
539 		/* If we allocated a buffer, attach to mbuf */
540 		if (cidx > MLEN) {
541 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
542 			ctl_mbuf->m_flags |= M_EXT_RW;
543 		}
544 		control = NULL;
545 		ctl_mbuf->m_len = cidx;
546 
547 		msg.msg_control = ctl_mbuf;
548 		msg.msg_flags |= MSG_CONTROLMBUF;
549 	}
550 
551 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
552 	/* Freed internally */
553 	ctl_mbuf = NULL;
554 
555 done:
556 	if (ctl_mbuf != NULL) {
557 		if (control != NULL && control != mtod(ctl_mbuf, void *))
558 			free(control, M_MBUF);
559 		m_free(ctl_mbuf);
560 	}
561 	return (error);
562 }
563 
564 int
565 linux_sys_recvfrom(struct lwp *l, void *v, register_t *retval)
566 {
567 	struct linux_sys_recvfrom_args /* {
568 		syscallarg(int) s;
569 		syscallarg(void *) buf;
570 		syscallarg(int) len;
571 		syscallarg(int) flags;
572 		syscallarg(struct osockaddr *) from;
573 		syscallarg(int *) fromlenaddr;
574 	} */ *uap = v;
575 	int		error;
576 	struct sys_recvfrom_args bra;
577 
578 	SCARG(&bra, s) = SCARG(uap, s);
579 	SCARG(&bra, buf) = SCARG(uap, buf);
580 	SCARG(&bra, len) = SCARG(uap, len);
581 	SCARG(&bra, flags) = SCARG(uap, flags);
582 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
583 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
584 
585 	if ((error = sys_recvfrom(l, &bra, retval)))
586 		return (error);
587 
588 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
589 		return (error);
590 
591 	return (0);
592 }
593 
594 static int
595 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
596 {
597 	int dlen, error = 0;
598 	struct cmsghdr *cmsg;
599 	struct linux_cmsghdr linux_cmsg;
600 	struct mbuf *m;
601 	char *q, *q_end;
602 
603 	if (mp->msg_controllen <= 0 || control == 0) {
604 		mp->msg_controllen = 0;
605 		free_control_mbuf(l, control, control);
606 		return 0;
607 	}
608 
609 	q = (char *)mp->msg_control;
610 	q_end = q + mp->msg_controllen;
611 
612 	for (m = control; m != NULL; ) {
613 		cmsg = mtod(m, struct cmsghdr *);
614 
615 		/*
616 		 * Fixup cmsg. We handle two things:
617 		 * 0. different sizeof cmsg_len.
618 		 * 1. different values for level/type on some archs
619 		 * 2. different alignment of CMSG_DATA on some archs
620 		 */
621 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
622 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
623 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
624 
625 		dlen = q_end - q;
626 		if (linux_cmsg.cmsg_len > dlen) {
627 			/* Not enough room for the parameter */
628 			dlen -= sizeof linux_cmsg;
629 			if (dlen <= 0)
630 				/* Discard if header wont fit */
631 				break;
632 			mp->msg_flags |= MSG_CTRUNC;
633 			if (linux_cmsg.cmsg_level == SOL_SOCKET
634 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
635 				/* Do not truncate me ... */
636 				break;
637 		} else
638 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
639 
640 		switch (linux_cmsg.cmsg_level) {
641 		case SOL_SOCKET:
642 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
643 			switch (linux_cmsg.cmsg_type) {
644 			case SCM_RIGHTS:
645 				/* Linux SCM_RIGHTS is same as NetBSD */
646 				break;
647 
648 			default:
649 				/* other types not supported */
650 				error = EINVAL;
651 				goto done;
652 			}
653 			/* machine dependant ! */
654 			break;
655 		default:
656 			/* pray and leave intact */
657 			break;
658 		}
659 
660 		/* There can be padding between the header and data... */
661 		error = copyout(&linux_cmsg, q, sizeof *cmsg);
662 		if (error != 0) {
663 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
664 			    dlen);
665 		}
666 		if (error != 0) {
667 			/* We must free all the SCM_RIGHTS */
668 			m = control;
669 			break;
670 		}
671 		m = m->m_next;
672 		if (m == NULL || q + LINUX_CMSG_ALIGN(dlen) > q_end) {
673 			q += dlen;
674 			break;
675 		}
676 		q += LINUX_CMSG_ALIGN(dlen);
677 	}
678 
679   done:
680 	free_control_mbuf(l, control, m);
681 
682 	mp->msg_controllen = q - (char *)mp->msg_control;
683 	return error;
684 }
685 
686 int
687 linux_sys_recvmsg(struct lwp *l, void *v, register_t *retval)
688 {
689 	struct linux_sys_recvmsg_args /* {
690 		syscallarg(int) s;
691 		syscallarg(struct msghdr *) msg;
692 		syscallarg(u_int) flags;
693 	} */ *uap = v;
694 	struct msghdr	msg;
695 	int		error;
696 	struct mbuf	*from, *control;
697 
698 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
699 	if (error)
700 		return (error);
701 
702 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
703 	if (msg.msg_flags < 0) {
704 		/* Some unsupported flag */
705 		return (EINVAL);
706 	}
707 	msg.msg_flags |= MSG_IOVUSRSPACE;
708 
709 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
710 	    msg.msg_control != NULL ? &control : NULL, retval);
711 	if (error != 0)
712 		return error;
713 
714 	if (msg.msg_control != NULL)
715 		error = linux_copyout_msg_control(l, &msg, control);
716 
717 	if (error == 0 && from != 0) {
718 		mtod(from, struct osockaddr *)->sa_family =
719 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
720 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
721 			from);
722 	} else
723 		msg.msg_namelen = 0;
724 
725 	if (from != NULL)
726 		m_free(from);
727 
728 	if (error == 0) {
729 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
730 		if (msg.msg_flags < 0)
731 			/* Some flag unsupported by Linux */
732 			error = EINVAL;
733 		else
734 			error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
735 	}
736 
737 	return (error);
738 }
739 
740 /*
741  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
742  * is different, the rest matches IPPROTO_* on both systems.
743  */
744 int
745 linux_to_bsd_sopt_level(int llevel)
746 {
747 
748 	switch (llevel) {
749 	case LINUX_SOL_SOCKET:
750 		return SOL_SOCKET;
751 	case LINUX_SOL_IP:
752 		return IPPROTO_IP;
753 	case LINUX_SOL_TCP:
754 		return IPPROTO_TCP;
755 	case LINUX_SOL_UDP:
756 		return IPPROTO_UDP;
757 	default:
758 		return -1;
759 	}
760 }
761 
762 /*
763  * Convert Linux socket level socket option numbers to NetBSD values.
764  */
765 int
766 linux_to_bsd_so_sockopt(int lopt)
767 {
768 
769 	switch (lopt) {
770 	case LINUX_SO_DEBUG:
771 		return SO_DEBUG;
772 	case LINUX_SO_REUSEADDR:
773 		/*
774 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
775 		 * host:port pair through SO_REUSEADDR even if the address is not a
776 		 * multicast-address.  Effectively, this means that we should use
777 		 * SO_REUSEPORT to allow Linux applications to not exit with
778 		 * EADDRINUSE
779 		 */
780 		return SO_REUSEPORT;
781 	case LINUX_SO_TYPE:
782 		return SO_TYPE;
783 	case LINUX_SO_ERROR:
784 		return SO_ERROR;
785 	case LINUX_SO_DONTROUTE:
786 		return SO_DONTROUTE;
787 	case LINUX_SO_BROADCAST:
788 		return SO_BROADCAST;
789 	case LINUX_SO_SNDBUF:
790 		return SO_SNDBUF;
791 	case LINUX_SO_RCVBUF:
792 		return SO_RCVBUF;
793 	case LINUX_SO_KEEPALIVE:
794 		return SO_KEEPALIVE;
795 	case LINUX_SO_OOBINLINE:
796 		return SO_OOBINLINE;
797 	case LINUX_SO_LINGER:
798 		return SO_LINGER;
799 	case LINUX_SO_PRIORITY:
800 	case LINUX_SO_NO_CHECK:
801 	default:
802 		return -1;
803 	}
804 }
805 
806 /*
807  * Convert Linux IP level socket option number to NetBSD values.
808  */
809 int
810 linux_to_bsd_ip_sockopt(int lopt)
811 {
812 
813 	switch (lopt) {
814 	case LINUX_IP_TOS:
815 		return IP_TOS;
816 	case LINUX_IP_TTL:
817 		return IP_TTL;
818 	case LINUX_IP_MULTICAST_TTL:
819 		return IP_MULTICAST_TTL;
820 	case LINUX_IP_MULTICAST_LOOP:
821 		return IP_MULTICAST_LOOP;
822 	case LINUX_IP_MULTICAST_IF:
823 		return IP_MULTICAST_IF;
824 	case LINUX_IP_ADD_MEMBERSHIP:
825 		return IP_ADD_MEMBERSHIP;
826 	case LINUX_IP_DROP_MEMBERSHIP:
827 		return IP_DROP_MEMBERSHIP;
828 	default:
829 		return -1;
830 	}
831 }
832 
833 /*
834  * Convert Linux TCP level socket option number to NetBSD values.
835  */
836 int
837 linux_to_bsd_tcp_sockopt(int lopt)
838 {
839 
840 	switch (lopt) {
841 	case LINUX_TCP_NODELAY:
842 		return TCP_NODELAY;
843 	case LINUX_TCP_MAXSEG:
844 		return TCP_MAXSEG;
845 	default:
846 		return -1;
847 	}
848 }
849 
850 /*
851  * Convert Linux UDP level socket option number to NetBSD values.
852  */
853 int
854 linux_to_bsd_udp_sockopt(int lopt)
855 {
856 
857 	switch (lopt) {
858 	default:
859 		return -1;
860 	}
861 }
862 
863 /*
864  * Another reasonably straightforward function: setsockopt(2).
865  * The level and option numbers are converted; the values passed
866  * are not (yet) converted, the ones currently implemented don't
867  * need conversion, as they are the same on both systems.
868  */
869 int
870 linux_sys_setsockopt(struct lwp *l, void *v, register_t *retval)
871 {
872 	struct linux_sys_setsockopt_args /* {
873 		syscallarg(int) s;
874 		syscallarg(int) level;
875 		syscallarg(int) optname;
876 		syscallarg(void *) optval;
877 		syscallarg(int) optlen;
878 	} */ *uap = v;
879 	struct proc *p = l->l_proc;
880 	struct sys_setsockopt_args bsa;
881 	int name;
882 
883 	SCARG(&bsa, s) = SCARG(uap, s);
884 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
885 	SCARG(&bsa, val) = SCARG(uap, optval);
886 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
887 
888 	/*
889 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
890 	 * and returns EOPNOTSUPP for other levels
891 	 */
892 	if (SCARG(&bsa, level) != SOL_SOCKET) {
893 		struct file *fp;
894 		struct socket *so;
895 		int error, s, family;
896 
897 		/* getsock() will use the descriptor for us */
898 	    	if ((error = getsock(p->p_fd, SCARG(&bsa, s), &fp)) != 0)
899 		    	return error;
900 
901 		s = splsoftnet();
902 		so = (struct socket *)fp->f_data;
903 		family = so->so_proto->pr_domain->dom_family;
904 		splx(s);
905 		FILE_UNUSE(fp, l);
906 
907 		if (family == AF_LOCAL)
908 			return EOPNOTSUPP;
909 	}
910 
911 	switch (SCARG(&bsa, level)) {
912 	case SOL_SOCKET:
913 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
914 		break;
915 	case IPPROTO_IP:
916 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
917 		break;
918 	case IPPROTO_TCP:
919 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
920 		break;
921 	case IPPROTO_UDP:
922 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
923 		break;
924 	default:
925 		return EINVAL;
926 	}
927 
928 	if (name == -1)
929 		return EINVAL;
930 	SCARG(&bsa, name) = name;
931 
932 	return sys_setsockopt(l, &bsa, retval);
933 }
934 
935 /*
936  * getsockopt(2) is very much the same as setsockopt(2) (see above)
937  */
938 int
939 linux_sys_getsockopt(struct lwp *l, void *v, register_t *retval)
940 {
941 	struct linux_sys_getsockopt_args /* {
942 		syscallarg(int) s;
943 		syscallarg(int) level;
944 		syscallarg(int) optname;
945 		syscallarg(void *) optval;
946 		syscallarg(int *) optlen;
947 	} */ *uap = v;
948 	struct sys_getsockopt_args bga;
949 	int name;
950 
951 	SCARG(&bga, s) = SCARG(uap, s);
952 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
953 	SCARG(&bga, val) = SCARG(uap, optval);
954 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
955 
956 	switch (SCARG(&bga, level)) {
957 	case SOL_SOCKET:
958 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
959 		break;
960 	case IPPROTO_IP:
961 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
962 		break;
963 	case IPPROTO_TCP:
964 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
965 		break;
966 	case IPPROTO_UDP:
967 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
968 		break;
969 	default:
970 		return EINVAL;
971 	}
972 
973 	if (name == -1)
974 		return EINVAL;
975 	SCARG(&bga, name) = name;
976 
977 	return sys_getsockopt(l, &bga, retval);
978 }
979 
980 #define IF_NAME_LEN 16
981 
982 int
983 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
984     void *data)
985 {
986 	/* Not the full structure, just enough to map what we do here */
987 	struct linux_ifreq {
988 		char if_name[IF_NAME_LEN];
989 		struct osockaddr hwaddr;
990 	} lreq;
991 	struct proc *p = l->l_proc;
992 	struct filedesc *fdp;
993 	struct file *fp;
994 	struct ifaddr *ifa;
995 	struct ifnet *ifp;
996 	struct sockaddr_dl *sadl;
997 	int error, found;
998 	int index, ifnum;
999 
1000 	/*
1001 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1002 	 * SIOCGIFCONF, because the user buffer is not of the right
1003 	 * type to take those results.  We can't use kernel buffers to
1004 	 * receive the results, as the implementation of sys_ioctl()
1005 	 * and ifconf() [which implements SIOCGIFCONF] use
1006 	 * copyin()/copyout() which will fail on kernel addresses.
1007 	 *
1008 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1009 	 */
1010 
1011 	fdp = p->p_fd;
1012 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1013 		return (EBADF);
1014 
1015 	FILE_USE(fp);
1016 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1017 		error = EBADF;
1018 		goto out;
1019 	}
1020 
1021 	error = copyin(data, &lreq, sizeof(lreq));
1022 	if (error)
1023 		goto out;
1024 	lreq.if_name[IF_NAME_LEN-1] = '\0';		/* just in case */
1025 
1026 	/*
1027 	 * Try real interface name first, then fake "ethX"
1028 	 */
1029 	found = 0;
1030 	IFNET_FOREACH(ifp) {
1031 		if (found)
1032 			break;
1033 		if (strcmp(lreq.if_name, ifp->if_xname))
1034 			/* not this interface */
1035 			continue;
1036 		found=1;
1037 		if (IFADDR_EMPTY(ifp)) {
1038 			error = ENODEV;
1039 			goto out;
1040 		}
1041 		IFADDR_FOREACH(ifa, ifp) {
1042 			sadl = satosdl(ifa->ifa_addr);
1043 			/* only return ethernet addresses */
1044 			/* XXX what about FDDI, etc. ? */
1045 			if (sadl->sdl_family != AF_LINK ||
1046 			    sadl->sdl_type != IFT_ETHER)
1047 				continue;
1048 			memcpy(&lreq.hwaddr.sa_data, CLLADDR(sadl),
1049 			       MIN(sadl->sdl_alen,
1050 				   sizeof(lreq.hwaddr.sa_data)));
1051 			lreq.hwaddr.sa_family =
1052 				sadl->sdl_family;
1053 			error = copyout(&lreq, data, sizeof(lreq));
1054 			goto out;
1055 		}
1056 	}
1057 
1058 	if (strncmp(lreq.if_name, "eth", 3) == 0) {
1059 		for (ifnum = 0, index = 3;
1060 		     lreq.if_name[index] != '\0' && index < IF_NAME_LEN;
1061 		     index++) {
1062 			ifnum *= 10;
1063 			ifnum += lreq.if_name[index] - '0';
1064 		}
1065 
1066 		error = EINVAL;			/* in case we don't find one */
1067 		found = 0;
1068 		IFNET_FOREACH(ifp) {
1069 			if (found)
1070 				break;
1071 			memcpy(lreq.if_name, ifp->if_xname,
1072 			       MIN(IF_NAME_LEN, IFNAMSIZ));
1073 			IFADDR_FOREACH(ifa, ifp) {
1074 				sadl = satosdl(ifa->ifa_addr);
1075 				/* only return ethernet addresses */
1076 				/* XXX what about FDDI, etc. ? */
1077 				if (sadl->sdl_family != AF_LINK ||
1078 				    sadl->sdl_type != IFT_ETHER)
1079 					continue;
1080 				if (ifnum--)
1081 					/* not the reqested iface */
1082 					continue;
1083 				memcpy(&lreq.hwaddr.sa_data,
1084 				       CLLADDR(sadl),
1085 				       MIN(sadl->sdl_alen,
1086 					   sizeof(lreq.hwaddr.sa_data)));
1087 				lreq.hwaddr.sa_family =
1088 					sadl->sdl_family;
1089 				error = copyout(&lreq, data, sizeof(lreq));
1090 				found = 1;
1091 				break;
1092 			}
1093 		}
1094 	} else {
1095 		/* unknown interface, not even an "eth*" name */
1096 		error = ENODEV;
1097 	}
1098 
1099 out:
1100 	FILE_UNUSE(fp, l);
1101 	return error;
1102 }
1103 #undef IF_NAME_LEN
1104 
1105 int
1106 linux_ioctl_socket(l, uap, retval)
1107 	struct lwp *l;
1108 	struct linux_sys_ioctl_args /* {
1109 		syscallarg(int) fd;
1110 		syscallarg(u_long) com;
1111 		syscallarg(void *) data;
1112 	} */ *uap;
1113 	register_t *retval;
1114 {
1115 	struct proc *p = l->l_proc;
1116 	u_long com;
1117 	int error = 0, isdev = 0, dosys = 1;
1118 	struct sys_ioctl_args ia;
1119 	struct file *fp;
1120 	struct filedesc *fdp;
1121 	struct vnode *vp;
1122 	int (*ioctlf)(struct file *, u_long, void *, struct lwp *);
1123 	struct ioctl_pt pt;
1124 
1125 	fdp = p->p_fd;
1126 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1127 		return (EBADF);
1128 
1129 	FILE_USE(fp);
1130 
1131 	if (fp->f_type == DTYPE_VNODE) {
1132 		vp = (struct vnode *)fp->f_data;
1133 		isdev = vp->v_type == VCHR;
1134 	}
1135 
1136 	/*
1137 	 * Don't try to interpret socket ioctl calls that are done
1138 	 * on a device filedescriptor, just pass them through, to
1139 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1140 	 * device will only handle these if it's prepared to do
1141 	 * so, to avoid unexpected things from happening.
1142 	 */
1143 	if (isdev) {
1144 		dosys = 0;
1145 		ioctlf = fp->f_ops->fo_ioctl;
1146 		pt.com = SCARG(uap, com);
1147 		pt.data = SCARG(uap, data);
1148 		error = ioctlf(fp, PTIOCLINUX, (void *)&pt, l);
1149 		/*
1150 		 * XXX hack: if the function returns EJUSTRETURN,
1151 		 * it has stuffed a sysctl return value in pt.data.
1152 		 */
1153 		if (error == EJUSTRETURN) {
1154 			retval[0] = (register_t)pt.data;
1155 			error = 0;
1156 		}
1157 		goto out;
1158 	}
1159 
1160 	com = SCARG(uap, com);
1161 	retval[0] = 0;
1162 
1163 	switch (com) {
1164 	case LINUX_SIOCGIFCONF:
1165 		SCARG(&ia, com) = OOSIOCGIFCONF;
1166 		break;
1167 	case LINUX_SIOCGIFFLAGS:
1168 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1169 		break;
1170 	case LINUX_SIOCSIFFLAGS:
1171 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1172 		break;
1173 	case LINUX_SIOCGIFADDR:
1174 		SCARG(&ia, com) = OOSIOCGIFADDR;
1175 		break;
1176 	case LINUX_SIOCGIFDSTADDR:
1177 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1178 		break;
1179 	case LINUX_SIOCGIFBRDADDR:
1180 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1181 		break;
1182 	case LINUX_SIOCGIFNETMASK:
1183 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1184 		break;
1185 	case LINUX_SIOCADDMULTI:
1186 		SCARG(&ia, com) = OSIOCADDMULTI;
1187 		break;
1188 	case LINUX_SIOCDELMULTI:
1189 		SCARG(&ia, com) = OSIOCDELMULTI;
1190 		break;
1191 	case LINUX_SIOCGIFHWADDR:
1192 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1193 					 SCARG(uap, data));
1194 		dosys = 0;
1195 		break;
1196 	default:
1197 		error = EINVAL;
1198 	}
1199 
1200 out:
1201 	FILE_UNUSE(fp, l);
1202 
1203 	if (error ==0 && dosys) {
1204 		SCARG(&ia, fd) = SCARG(uap, fd);
1205 		SCARG(&ia, data) = SCARG(uap, data);
1206 		/* XXX NJWLWP */
1207 		error = sys_ioctl(curlwp, &ia, retval);
1208 	}
1209 
1210 	return error;
1211 }
1212 
1213 int
1214 linux_sys_connect(struct lwp *l, void *v, register_t *retval)
1215 {
1216 	struct linux_sys_connect_args /* {
1217 		syscallarg(int) s;
1218 		syscallarg(const struct sockaddr *) name;
1219 		syscallarg(int) namelen;
1220 	} */ *uap = v;
1221 	int		error;
1222 	struct mbuf *nam;
1223 
1224 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1225 	    SCARG(uap, namelen));
1226 	if (error)
1227 		return (error);
1228 
1229 	error = do_sys_connect(l, SCARG(uap, s), nam);
1230 
1231 	if (error == EISCONN) {
1232 		struct file *fp;
1233 		struct socket *so;
1234 		int s, state, prflags;
1235 
1236 		/* getsock() will use the descriptor for us */
1237 	    	if (getsock(l->l_proc->p_fd, SCARG(uap, s), &fp) != 0)
1238 		    	return EISCONN;
1239 
1240 		s = splsoftnet();
1241 		so = (struct socket *)fp->f_data;
1242 		state = so->so_state;
1243 		prflags = so->so_proto->pr_flags;
1244 		splx(s);
1245 		FILE_UNUSE(fp, l);
1246 		/*
1247 		 * We should only let this call succeed once per
1248 		 * non-blocking connect; however we don't have
1249 		 * a convenient place to keep that state..
1250 		 */
1251 		if ((state & SS_NBIO) && (state & SS_ISCONNECTED) &&
1252 		    (prflags & PR_CONNREQUIRED))
1253 			return 0;
1254 	}
1255 
1256 	return (error);
1257 }
1258 
1259 int
1260 linux_sys_bind(struct lwp *l, void *v, register_t *retval)
1261 {
1262 	struct linux_sys_bind_args /* {
1263 		syscallarg(int) s;
1264 		syscallarg(const struct osockaddr *) name;
1265 		syscallarg(int) namelen;
1266 	} */ *uap = v;
1267 	int		error;
1268 	struct mbuf     *nam;
1269 
1270 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1271 	    SCARG(uap, namelen));
1272 	if (error)
1273 		return (error);
1274 
1275 	return do_sys_bind(l, SCARG(uap, s), nam);
1276 }
1277 
1278 int
1279 linux_sys_getsockname(struct lwp *l, void *v, register_t *retval)
1280 {
1281 	struct linux_sys_getsockname_args /* {
1282 		syscallarg(int) fdes;
1283 		syscallarg(void *) asa;
1284 		syscallarg(int *) alen;
1285 	} */ *uap = v;
1286 	int error;
1287 
1288 	if ((error = sys_getsockname(l, uap, retval)) != 0)
1289 		return (error);
1290 
1291 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1292 		return (error);
1293 
1294 	return (0);
1295 }
1296 
1297 int
1298 linux_sys_getpeername(struct lwp *l, void *v, register_t *retval)
1299 {
1300 	struct sys_getpeername_args /* {
1301 		syscallarg(int) fdes;
1302 		syscallarg(void *) asa;
1303 		syscallarg(int *) alen;
1304 	} */ *uap = v;
1305 	int error;
1306 
1307 	if ((error = sys_getpeername(l, uap, retval)) != 0)
1308 		return (error);
1309 
1310 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1311 		return (error);
1312 
1313 	return (0);
1314 }
1315 
1316 /*
1317  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1318  * family and convert to sockaddr.
1319  */
1320 static int
1321 linux_get_sa(struct lwp *l, int s, struct mbuf **mp, const struct osockaddr *osa, int salen)
1322 {
1323 	int error, bdom;
1324 	struct sockaddr *sa;
1325 	struct osockaddr *kosa;
1326 	struct mbuf *m;
1327 
1328 	if (salen == 1 || salen > UCHAR_MAX) {
1329 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1330 		return EINVAL;
1331 	}
1332 
1333 	/* We'll need the address in an mbuf later, so copy into one here */
1334 	m = m_get(M_WAIT, MT_SONAME);
1335 	if (salen > MLEN)
1336 		MEXTMALLOC(m, salen, M_WAITOK);
1337 
1338 	m->m_len = salen;
1339 
1340 	if (salen == 0)
1341 		return 0;
1342 
1343 	kosa = mtod(m, void *);
1344 	if ((error = copyin(osa, kosa, salen))) {
1345 		DPRINTF(("error %d copying osa %p len %d\n",
1346 				error, osa, salen));
1347 		goto bad;
1348 	}
1349 
1350 	ktrkuser("linux sockaddr", kosa, salen);
1351 
1352 	bdom = linux_to_bsd_domain(kosa->sa_family);
1353 	if (bdom == -1) {
1354 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1355 		error = EINVAL;
1356 		goto bad;
1357 	}
1358 
1359 	/*
1360 	 * If the family is unspecified, use address family of the socket.
1361 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1362 	 */
1363 	if (bdom == AF_UNSPEC) {
1364 		struct file *fp;
1365 		struct socket *so;
1366 
1367 		/* getsock() will use the descriptor for us */
1368 		if ((error = getsock(l->l_proc->p_fd, s, &fp)) != 0)
1369 			goto bad;
1370 
1371 		so = (struct socket *)fp->f_data;
1372 		bdom = so->so_proto->pr_domain->dom_family;
1373 
1374 		FILE_UNUSE(fp, l);
1375 
1376 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1377 	}
1378 
1379 #ifdef INET6
1380 	/*
1381 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1382 	 * which lacks the scope id compared with RFC2553 one. If we detect
1383 	 * the situation, reject the address and write a message to system log.
1384 	 *
1385 	 * Still accept addresses for which the scope id is not used.
1386 	 */
1387 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1388 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1389 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1390 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1391 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1392 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1393 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1394 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1395 			struct proc *p = l->l_proc;
1396 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1397 
1398 			log(LOG_DEBUG,
1399 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1400 			    "sockaddr_in6 rejected",
1401 			    p->p_pid, p->p_comm, uid);
1402 			error = EINVAL;
1403 			goto bad;
1404 		}
1405 		salen = sizeof (struct sockaddr_in6);
1406 		sin6->sin6_scope_id = 0;
1407 	}
1408 #endif
1409 
1410 	if (bdom == AF_INET)
1411 		salen = sizeof(struct sockaddr_in);
1412 
1413 	sa = (struct sockaddr *) kosa;
1414 	sa->sa_family = bdom;
1415 	sa->sa_len = salen;
1416 	m->m_len = salen;
1417 	ktrkuser("new sockaddr", kosa, salen);
1418 
1419 #ifdef DEBUG_LINUX
1420 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1421 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1422 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1423 	DPRINTF(("\n"));
1424 #endif
1425 
1426 	*mp = m;
1427 	return 0;
1428 
1429     bad:
1430 	m_free(m);
1431 	return error;
1432 }
1433 
1434 static int
1435 linux_sa_put(struct osockaddr *osa)
1436 {
1437 	struct sockaddr sa;
1438 	struct osockaddr *kosa;
1439 	int error, bdom, len;
1440 
1441 	/*
1442 	 * Only read/write the sockaddr family and length part, the rest is
1443 	 * not changed.
1444 	 */
1445 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1446 
1447 	error = copyin(osa, &sa, len);
1448 	if (error)
1449 		return (error);
1450 
1451 	bdom = bsd_to_linux_domain(sa.sa_family);
1452 	if (bdom == -1)
1453 		return (EINVAL);
1454 
1455 	/* Note: we convert from sockaddr to osockaddr here, too */
1456 	kosa = (struct osockaddr *) &sa;
1457 	kosa->sa_family = bdom;
1458 	error = copyout(kosa, osa, len);
1459 	if (error)
1460 		return (error);
1461 
1462 	return (0);
1463 }
1464 
1465 #ifndef __amd64__
1466 int
1467 linux_sys_recv(struct lwp *l, void *v, register_t *retval)
1468 {
1469 	struct linux_sys_recv_args /* {
1470 		syscallarg(int) s;
1471 		syscallarg(void *) buf;
1472 		syscallarg(int) len;
1473 		syscallarg(int) flags;
1474 	} */ *uap = v;
1475 	struct sys_recvfrom_args bra;
1476 
1477 
1478 	SCARG(&bra, s) = SCARG(uap, s);
1479 	SCARG(&bra, buf) = SCARG(uap, buf);
1480 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1481 	SCARG(&bra, flags) = SCARG(uap, flags);
1482 	SCARG(&bra, from) = NULL;
1483 	SCARG(&bra, fromlenaddr) = NULL;
1484 
1485 	return (sys_recvfrom(l, &bra, retval));
1486 }
1487 
1488 int
1489 linux_sys_send(struct lwp *l, void *v, register_t *retval)
1490 {
1491 	struct linux_sys_send_args /* {
1492 		syscallarg(int) s;
1493 		syscallarg(void *) buf;
1494 		syscallarg(int) len;
1495 		syscallarg(int) flags;
1496 	} */ *uap = v;
1497 	struct sys_sendto_args bsa;
1498 
1499 	SCARG(&bsa, s)		= SCARG(uap, s);
1500 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1501 	SCARG(&bsa, len)	= SCARG(uap, len);
1502 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1503 	SCARG(&bsa, to)		= NULL;
1504 	SCARG(&bsa, tolen)	= 0;
1505 
1506 	return (sys_sendto(l, &bsa, retval));
1507 }
1508 #endif
1509 
1510 int
1511 linux_sys_accept(struct lwp *l, void *v, register_t *retval)
1512 {
1513 	struct linux_sys_accept_args /* {
1514 		syscallarg(int) s;
1515 		syscallarg(struct osockaddr *) name;
1516 		syscallarg(int *) anamelen;
1517 	} */ *uap = v;
1518 	int error;
1519 	struct sys_accept_args baa;
1520 
1521 	SCARG(&baa, s)		= SCARG(uap, s);
1522 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1523 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1524 
1525 	if ((error = sys_accept(l, &baa, retval)))
1526 		return (error);
1527 
1528 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1529 		return (error);
1530 
1531 	return (0);
1532 }
1533