xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision 4d6fc14bc9b0c5bf3e30be318c143ee82cadd108)
1 /*	$NetBSD: linux_socket.c,v 1.152 2020/11/03 22:08:44 christos 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 1.152 2020/11/03 22:08:44 christos 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/ioctl.h>
49 #include <sys/tty.h>
50 #include <sys/file.h>
51 #include <sys/filedesc.h>
52 #include <sys/select.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/domain.h>
56 #include <net/if.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <netinet/in.h>
60 #include <netinet/tcp.h>
61 #include <sys/mount.h>
62 #include <sys/proc.h>
63 #include <sys/vnode.h>
64 #include <sys/device.h>
65 #include <sys/protosw.h>
66 #include <sys/mbuf.h>
67 #include <sys/syslog.h>
68 #include <sys/exec.h>
69 #include <sys/kauth.h>
70 #include <sys/syscallargs.h>
71 #include <sys/ktrace.h>
72 
73 #include <lib/libkern/libkern.h>
74 
75 #include <netinet/ip6.h>
76 #include <netinet6/ip6_var.h>
77 
78 #include <compat/sys/socket.h>
79 #include <compat/sys/sockio.h>
80 
81 #include <compat/linux/common/linux_types.h>
82 #include <compat/linux/common/linux_util.h>
83 #include <compat/linux/common/linux_signal.h>
84 #include <compat/linux/common/linux_ioctl.h>
85 #include <compat/linux/common/linux_sched.h>
86 #include <compat/linux/common/linux_socket.h>
87 #include <compat/linux/common/linux_fcntl.h>
88 #if !defined(__alpha__) && !defined(__amd64__)
89 #include <compat/linux/common/linux_socketcall.h>
90 #endif
91 #include <compat/linux/common/linux_sockio.h>
92 #include <compat/linux/common/linux_ipc.h>
93 #include <compat/linux/common/linux_sem.h>
94 
95 #include <compat/linux/linux_syscallargs.h>
96 
97 #ifdef DEBUG_LINUX
98 #define DPRINTF(a) uprintf a
99 #else
100 #define DPRINTF(a)
101 #endif
102 
103 /*
104  * The calls in this file are entered either via the linux_socketcall()
105  * interface or, on the Alpha, as individual syscalls.  The
106  * linux_socketcall function does any massaging of arguments so that all
107  * the calls in here need not think that they are anything other
108  * than a normal syscall.
109  */
110 
111 static int linux_to_bsd_domain(int);
112 static int bsd_to_linux_domain(int);
113 static int linux_to_bsd_type(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_ipv6_sockopt(int);
118 int linux_to_bsd_tcp_sockopt(int);
119 int linux_to_bsd_udp_sockopt(int);
120 int linux_getifname(struct lwp *, register_t *, void *);
121 int linux_getifconf(struct lwp *, register_t *, void *);
122 int linux_getifhwaddr(struct lwp *, register_t *, u_int, void *);
123 static int linux_get_sa(struct lwp *, int, struct sockaddr_big *,
124 		const struct osockaddr *, socklen_t);
125 static int linux_sa_put(struct osockaddr *osa);
126 static int linux_to_bsd_msg_flags(int);
127 static int bsd_to_linux_msg_flags(int);
128 static void linux_to_bsd_msghdr(const struct linux_msghdr *, struct msghdr *);
129 static void bsd_to_linux_msghdr(const struct msghdr *, struct linux_msghdr *);
130 
131 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
132 	AF_UNSPEC,
133 	AF_UNIX,
134 	AF_INET,
135 	AF_CCITT,	/* LINUX_AF_AX25 */
136 	AF_IPX,
137 	AF_APPLETALK,
138 	-1,		/* LINUX_AF_NETROM */
139 	-1,		/* LINUX_AF_BRIDGE */
140 	-1,		/* LINUX_AF_ATMPVC */
141 	AF_CCITT,	/* LINUX_AF_X25 */
142 	AF_INET6,
143 	-1,		/* LINUX_AF_ROSE */
144 	AF_DECnet,
145 	-1,		/* LINUX_AF_NETBEUI */
146 	-1,		/* LINUX_AF_SECURITY */
147 	pseudo_AF_KEY,
148 	AF_ROUTE,	/* LINUX_AF_NETLINK */
149 	-1,		/* LINUX_AF_PACKET */
150 	-1,		/* LINUX_AF_ASH */
151 	-1,		/* LINUX_AF_ECONET */
152 	-1,		/* LINUX_AF_ATMSVC */
153 	AF_SNA,
154 	/* rest up to LINUX_AF_MAX-1 is not allocated */
155 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
156 };
157 
158 static const int bsd_to_linux_domain_[AF_MAX] = {
159 	LINUX_AF_UNSPEC,
160 	LINUX_AF_UNIX,
161 	LINUX_AF_INET,
162 	-1,		/* AF_IMPLINK */
163 	-1,		/* AF_PUP */
164 	-1,		/* AF_CHAOS */
165 	-1,		/* AF_NS */
166 	-1,		/* AF_ISO */
167 	-1,		/* AF_ECMA */
168 	-1,		/* AF_DATAKIT */
169 	LINUX_AF_AX25,	/* AF_CCITT */
170 	LINUX_AF_SNA,
171 	LINUX_AF_DECnet,
172 	-1,		/* AF_DLI */
173 	-1,		/* AF_LAT */
174 	-1,		/* AF_HYLINK */
175 	LINUX_AF_APPLETALK,
176 	LINUX_AF_NETLINK,
177 	-1,		/* AF_LINK */
178 	-1,		/* AF_XTP */
179 	-1,		/* AF_COIP */
180 	-1,		/* AF_CNT */
181 	-1,		/* pseudo_AF_RTIP */
182 	LINUX_AF_IPX,
183 	LINUX_AF_INET6,
184 	-1,		/* pseudo_AF_PIP */
185 	-1,		/* AF_ISDN */
186 	-1,		/* AF_NATM */
187 	-1,		/* AF_ARP */
188 	LINUX_pseudo_AF_KEY,
189 	-1,		/* pseudo_AF_HDRCMPLT */
190 };
191 
192 static const struct {
193 	int bfl;
194 	int lfl;
195 } bsd_to_linux_msg_flags_[] = {
196 	{MSG_OOB,		LINUX_MSG_OOB},
197 	{MSG_PEEK,		LINUX_MSG_PEEK},
198 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
199 	{MSG_EOR,		LINUX_MSG_EOR},
200 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
201 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
202 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
203 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
204 	{MSG_BCAST,		0},		/* not supported, clear */
205 	{MSG_MCAST,		0},		/* not supported, clear */
206 	{MSG_NOSIGNAL,		LINUX_MSG_NOSIGNAL},
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_MORE},
214 };
215 
216 /*
217  * Convert between Linux and BSD socket domain values
218  */
219 static int
220 linux_to_bsd_domain(int ldom)
221 {
222 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
223 		return (-1);
224 
225 	return linux_to_bsd_domain_[ldom];
226 }
227 
228 /*
229  * Convert between BSD and Linux socket domain values
230  */
231 static int
232 bsd_to_linux_domain(int bdom)
233 {
234 	if (bdom < 0 || bdom >= AF_MAX)
235 		return (-1);
236 
237 	return bsd_to_linux_domain_[bdom];
238 }
239 
240 static int
241 linux_to_bsd_type(int ltype)
242 {
243 	int type, flags;
244 
245 	/* Real types are identical between Linux and NetBSD */
246 	type = ltype & LINUX_SOCK_TYPE_MASK;
247 
248 	/* But flags are not .. */
249 	flags = ltype & ~LINUX_SOCK_TYPE_MASK;
250 	if (flags & ~(LINUX_SOCK_CLOEXEC|LINUX_SOCK_NONBLOCK))
251 		return -1;
252 
253 	if (flags & LINUX_SOCK_CLOEXEC)
254 		type |= SOCK_CLOEXEC;
255 	if (flags & LINUX_SOCK_NONBLOCK)
256 		type |= SOCK_NONBLOCK;
257 
258 	return type;
259 }
260 
261 static int
262 linux_to_bsd_msg_flags(int lflag)
263 {
264 	int i, lfl, bfl;
265 	int bflag = 0;
266 
267 	if (lflag == 0)
268 		return (0);
269 
270 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
271 		bfl = bsd_to_linux_msg_flags_[i].bfl;
272 		lfl = bsd_to_linux_msg_flags_[i].lfl;
273 
274 		if (lfl == 0)
275 			continue;
276 
277 		if (lflag & lfl) {
278 			if (bfl < 0)
279 				return (-1);
280 
281 			bflag |= bfl;
282 		}
283 	}
284 
285 	return (bflag);
286 }
287 
288 static int
289 bsd_to_linux_msg_flags(int bflag)
290 {
291 	int i, lfl, bfl;
292 	int lflag = 0;
293 
294 	if (bflag == 0)
295 		return (0);
296 
297 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
298 		bfl = bsd_to_linux_msg_flags_[i].bfl;
299 		lfl = bsd_to_linux_msg_flags_[i].lfl;
300 
301 		if (bfl <= 0)
302 			continue;
303 
304 		if (bflag & bfl) {
305 			if (lfl < 0)
306 				return (-1);
307 
308 			lflag |= lfl;
309 		}
310 	}
311 
312 	return (lflag);
313 }
314 
315 int
316 linux_sys_socket(struct lwp *l, const struct linux_sys_socket_args *uap, register_t *retval)
317 {
318 	/* {
319 		syscallarg(int)	domain;
320 		syscallarg(int)	type;
321 		syscallarg(int) protocol;
322 	} */
323 	struct sys___socket30_args bsa;
324 	int error;
325 
326 
327 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
328 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
329 	if (SCARG(&bsa, domain) == -1)
330 		return EINVAL;
331 	SCARG(&bsa, type) = linux_to_bsd_type(SCARG(uap, type));
332 	if (SCARG(&bsa, type) == -1)
333 		return EINVAL;
334 	/*
335 	 * Apparently linux uses this to talk to ISDN sockets. If we fail
336 	 * now programs seems to handle it, but if we don't we are going
337 	 * to fail when we bind and programs don't handle this well.
338 	 */
339 	if (SCARG(&bsa, domain) == AF_ROUTE && SCARG(&bsa, type) == SOCK_RAW)
340 		return ENOTSUP;
341 	error = sys___socket30(l, &bsa, retval);
342 
343 #ifdef INET6
344 	/*
345 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
346 	 * default and some apps depend on this. So, set V6ONLY to 0
347 	 * for Linux apps if the sysctl value is set to 1.
348 	 */
349 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
350 		struct socket *so;
351 
352 		if (fd_getsock(*retval, &so) == 0) {
353 			int val = 0;
354 
355 			/* ignore error */
356 			(void)so_setsockopt(l, so, IPPROTO_IPV6, IPV6_V6ONLY,
357 			    &val, sizeof(val));
358 
359 			fd_putfile(*retval);
360 		}
361 	}
362 #endif
363 
364 	return (error);
365 }
366 
367 int
368 linux_sys_socketpair(struct lwp *l, const struct linux_sys_socketpair_args *uap, register_t *retval)
369 {
370 	/* {
371 		syscallarg(int) domain;
372 		syscallarg(int) type;
373 		syscallarg(int) protocol;
374 		syscallarg(int *) rsv;
375 	} */
376 	struct sys_socketpair_args bsa;
377 
378 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
379 	if (SCARG(&bsa, domain) == -1)
380 		return EINVAL;
381 	SCARG(&bsa, type) = linux_to_bsd_type(SCARG(uap, type));
382 	if (SCARG(&bsa, type) == -1)
383 		return EINVAL;
384 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
385 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
386 
387 	return sys_socketpair(l, &bsa, retval);
388 }
389 
390 int
391 linux_sys_sendto(struct lwp *l, const struct linux_sys_sendto_args *uap, register_t *retval)
392 {
393 	/* {
394 		syscallarg(int)				s;
395 		syscallarg(void *)			msg;
396 		syscallarg(int)				len;
397 		syscallarg(int)				flags;
398 		syscallarg(struct osockaddr *)		to;
399 		syscallarg(int)				tolen;
400 	} */
401 	struct msghdr   msg;
402 	struct iovec    aiov;
403 	struct sockaddr_big nam;
404 	struct mbuf *m;
405 	int bflags;
406 	int error;
407 
408 	/* Translate message flags.  */
409 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
410 	if (bflags < 0)
411 		/* Some supported flag */
412 		return EINVAL;
413 
414 	msg.msg_flags = 0;
415 	msg.msg_name = NULL;
416 	msg.msg_control = NULL;
417 
418 	if (SCARG(uap, tolen)) {
419 		/* Read in and convert the sockaddr */
420 		error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
421 		    SCARG(uap, tolen));
422 		if (error)
423 			return error;
424 		error = sockargs(&m, &nam, nam.sb_len, UIO_SYSSPACE, MT_SONAME);
425 		if (error)
426 			return error;
427 		msg.msg_flags |= MSG_NAMEMBUF;
428 		msg.msg_name = m;
429 		msg.msg_namelen = nam.sb_len;
430 	}
431 
432 	msg.msg_iov = &aiov;
433 	msg.msg_iovlen = 1;
434 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
435 	aiov.iov_len = SCARG(uap, len);
436 
437 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
438 }
439 
440 static void
441 linux_to_bsd_msghdr(const struct linux_msghdr *lmsg, struct msghdr *bmsg)
442 {
443 	bmsg->msg_name = lmsg->msg_name;
444 	bmsg->msg_namelen = lmsg->msg_namelen;
445 	bmsg->msg_iov = lmsg->msg_iov;
446 	bmsg->msg_iovlen = lmsg->msg_iovlen;
447 	bmsg->msg_control = lmsg->msg_control;
448 	bmsg->msg_controllen = lmsg->msg_controllen;
449 	bmsg->msg_flags = lmsg->msg_flags;
450 }
451 
452 static void
453 bsd_to_linux_msghdr(const struct msghdr *bmsg, struct linux_msghdr *lmsg)
454 {
455 	lmsg->msg_name = bmsg->msg_name;
456 	lmsg->msg_namelen = bmsg->msg_namelen;
457 	lmsg->msg_iov = bmsg->msg_iov;
458 	lmsg->msg_iovlen = bmsg->msg_iovlen;
459 	lmsg->msg_control = bmsg->msg_control;
460 	lmsg->msg_controllen = bmsg->msg_controllen;
461 	lmsg->msg_flags = bmsg->msg_flags;
462 }
463 
464 int
465 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
466 {
467 	/* {
468 		syscallarg(int) s;
469 		syscallarg(struct linux_msghdr *) msg;
470 		syscallarg(u_int) flags;
471 	} */
472 	struct msghdr	msg;
473 	struct linux_msghdr lmsg;
474 	int		error;
475 	int		bflags;
476 	struct sockaddr_big nam;
477 	u_int8_t	*control;
478 	struct mbuf     *ctl_mbuf = NULL;
479 
480 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
481 	if (error)
482 		return error;
483 	linux_to_bsd_msghdr(&lmsg, &msg);
484 
485 	msg.msg_flags = MSG_IOVUSRSPACE;
486 
487 	/*
488 	 * Translate message flags.
489 	 */
490 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
491 	if (bflags < 0)
492 		/* Some supported flag */
493 		return EINVAL;
494 
495 	if (lmsg.msg_name) {
496 		/* Read in and convert the sockaddr */
497 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
498 		    msg.msg_namelen);
499 		if (error)
500 			return (error);
501 		msg.msg_name = &nam;
502 	}
503 
504 	/*
505 	 * Handle cmsg if there is any.
506 	 */
507 	if (LINUX_CMSG_FIRSTHDR(&lmsg)) {
508 		struct linux_cmsghdr l_cmsg, *l_cc;
509 		struct cmsghdr *cmsg;
510 		ssize_t resid = msg.msg_controllen;
511 		size_t clen, cidx = 0, cspace;
512 
513 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
514 		clen = MLEN;
515 		control = mtod(ctl_mbuf, void *);
516 
517 		l_cc = LINUX_CMSG_FIRSTHDR(&lmsg);
518 		do {
519 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
520 			if (error)
521 				goto done;
522 
523 			/*
524 			 * Sanity check the control message length.
525 			 */
526 			if (l_cmsg.cmsg_len > resid
527 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
528 				error = EINVAL;
529 				goto done;
530 			}
531 
532 			/*
533 			 * Refuse unsupported control messages, and
534 			 * translate fields as appropriate.
535 			 */
536 			switch (l_cmsg.cmsg_level) {
537 			case LINUX_SOL_SOCKET:
538 				/* It only differs on some archs */
539 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
540 					l_cmsg.cmsg_level = SOL_SOCKET;
541 
542 				switch(l_cmsg.cmsg_type) {
543 				case LINUX_SCM_RIGHTS:
544 					/* Linux SCM_RIGHTS is same as NetBSD */
545 					break;
546 
547 				case LINUX_SCM_CREDENTIALS:
548 					/* no native equivalent, just drop it */
549 					if (control != mtod(ctl_mbuf, void *))
550 						free(control, M_MBUF);
551 					m_free(ctl_mbuf);
552 					ctl_mbuf = NULL;
553 					msg.msg_control = NULL;
554 					msg.msg_controllen = 0;
555 					goto skipcmsg;
556 
557 				default:
558 					/* other types not supported */
559 					error = EINVAL;
560 					goto done;
561 				}
562 				break;
563 			default:
564 				/* pray and leave intact */
565 				break;
566 			}
567 
568 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
569 
570 			/* Check the buffer is big enough */
571 			if (__predict_false(cidx + cspace > clen)) {
572 				u_int8_t *nc;
573 				size_t nclen;
574 
575 				nclen = cidx + cspace;
576 				if (nclen >= PAGE_SIZE) {
577 					error = EINVAL;
578 					goto done;
579 				}
580 				nc = realloc(clen <= MLEN ? NULL : control,
581 						nclen, M_TEMP, M_WAITOK);
582 				if (!nc) {
583 					error = ENOMEM;
584 					goto done;
585 				}
586 				if (cidx <= MLEN)
587 					/* Old buffer was in mbuf... */
588 					memcpy(nc, control, cidx);
589 				control = nc;
590 				clen = nclen;
591 			}
592 
593 			/* Copy header */
594 			cmsg = (void *)&control[cidx];
595 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
596 			cmsg->cmsg_level = l_cmsg.cmsg_level;
597 			cmsg->cmsg_type = l_cmsg.cmsg_type;
598 
599 			/* Zero area between header and data */
600 			memset(cmsg + 1, 0,
601 				CMSG_ALIGN(sizeof(*cmsg)) - sizeof(*cmsg));
602 
603 			/* Copyin the data */
604 			error = copyin(LINUX_CMSG_DATA(l_cc),
605 				CMSG_DATA(cmsg),
606 				l_cmsg.cmsg_len - sizeof(l_cmsg));
607 			if (error)
608 				goto done;
609 
610 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
611 			cidx += cspace;
612 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc, &l_cmsg)) && resid > 0);
613 
614 		/* If we allocated a buffer, attach to mbuf */
615 		if (cidx > MLEN) {
616 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
617 			ctl_mbuf->m_flags |= M_EXT_RW;
618 		}
619 		control = NULL;
620 		ctl_mbuf->m_len = cidx;
621 
622 		msg.msg_control = ctl_mbuf;
623 		msg.msg_flags |= MSG_CONTROLMBUF;
624 
625 		ktrkuser("mbcontrol", mtod(ctl_mbuf, void *),
626 		    msg.msg_controllen);
627 	}
628 
629 skipcmsg:
630 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
631 	/* Freed internally */
632 	ctl_mbuf = NULL;
633 
634 done:
635 	if (ctl_mbuf != NULL) {
636 		if (control != NULL && control != mtod(ctl_mbuf, void *))
637 			free(control, M_MBUF);
638 		m_free(ctl_mbuf);
639 	}
640 	return (error);
641 }
642 
643 int
644 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
645 {
646 	/* {
647 		syscallarg(int) s;
648 		syscallarg(void *) buf;
649 		syscallarg(int) len;
650 		syscallarg(int) flags;
651 		syscallarg(struct osockaddr *) from;
652 		syscallarg(int *) fromlenaddr;
653 	} */
654 	int		error;
655 	struct sys_recvfrom_args bra;
656 
657 	SCARG(&bra, s) = SCARG(uap, s);
658 	SCARG(&bra, buf) = SCARG(uap, buf);
659 	SCARG(&bra, len) = SCARG(uap, len);
660 	SCARG(&bra, flags) = SCARG(uap, flags);
661 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
662 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
663 
664 	if ((error = sys_recvfrom(l, &bra, retval)))
665 		return (error);
666 
667 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
668 		return (error);
669 
670 	return (0);
671 }
672 
673 static int
674 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
675 {
676 	int dlen, error = 0;
677 	struct cmsghdr *cmsg;
678 	struct linux_cmsghdr linux_cmsg;
679 	struct mbuf *m;
680 	char *q, *q_end;
681 
682 	if (mp->msg_controllen <= 0 || control == 0) {
683 		mp->msg_controllen = 0;
684 		free_control_mbuf(l, control, control);
685 		return 0;
686 	}
687 
688 	ktrkuser("msgcontrol", mtod(control, void *), mp->msg_controllen);
689 
690 	q = (char *)mp->msg_control;
691 	q_end = q + mp->msg_controllen;
692 
693 	for (m = control; m != NULL; ) {
694 		cmsg = mtod(m, struct cmsghdr *);
695 
696 		/*
697 		 * Fixup cmsg. We handle two things:
698 		 * 0. different sizeof cmsg_len.
699 		 * 1. different values for level/type on some archs
700 		 * 2. different alignment of CMSG_DATA on some archs
701 		 */
702 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
703 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
704 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
705 
706 		dlen = q_end - q;
707 		if (linux_cmsg.cmsg_len > dlen) {
708 			/* Not enough room for the parameter */
709 			dlen -= sizeof linux_cmsg;
710 			if (dlen <= 0)
711 				/* Discard if header wont fit */
712 				break;
713 			mp->msg_flags |= MSG_CTRUNC;
714 			if (linux_cmsg.cmsg_level == SOL_SOCKET
715 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
716 				/* Do not truncate me ... */
717 				break;
718 		} else
719 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
720 
721 		switch (linux_cmsg.cmsg_level) {
722 		case SOL_SOCKET:
723 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
724 			switch (linux_cmsg.cmsg_type) {
725 			case SCM_RIGHTS:
726 				/* Linux SCM_RIGHTS is same as NetBSD */
727 				break;
728 
729 			default:
730 				/* other types not supported */
731 				error = EINVAL;
732 				goto done;
733 			}
734 			/* machine dependent ! */
735 			break;
736 		default:
737 			/* pray and leave intact */
738 			break;
739 		}
740 
741 		/* There can be padding between the header and data... */
742 		error = copyout(&linux_cmsg, q, sizeof linux_cmsg);
743 		if (error != 0) {
744 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
745 			    dlen);
746 		}
747 		if (error != 0) {
748 			/* We must free all the SCM_RIGHTS */
749 			m = control;
750 			break;
751 		}
752 		m = m->m_next;
753 		if (m == NULL || q + LINUX_CMSG_SPACE(dlen) > q_end) {
754 			q += LINUX_CMSG_LEN(dlen);
755 			break;
756 		}
757 		q += LINUX_CMSG_SPACE(dlen);
758 	}
759 
760   done:
761 	free_control_mbuf(l, control, m);
762 
763 	mp->msg_controllen = q - (char *)mp->msg_control;
764 	return error;
765 }
766 
767 int
768 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
769 {
770 	/* {
771 		syscallarg(int) s;
772 		syscallarg(struct linux_msghdr *) msg;
773 		syscallarg(u_int) flags;
774 	} */
775 	struct msghdr	msg;
776 	struct linux_msghdr lmsg;
777 	int		error;
778 	struct mbuf	*from, *control;
779 
780 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
781 	if (error)
782 		return (error);
783 	linux_to_bsd_msghdr(&lmsg, &msg);
784 
785 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
786 	if (msg.msg_flags < 0) {
787 		/* Some unsupported flag */
788 		return (EINVAL);
789 	}
790 	msg.msg_flags |= MSG_IOVUSRSPACE;
791 
792 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
793 	    msg.msg_control != NULL ? &control : NULL, retval);
794 	if (error != 0)
795 		return error;
796 
797 	if (msg.msg_control != NULL)
798 		error = linux_copyout_msg_control(l, &msg, control);
799 
800 	if (error == 0 && from != 0) {
801 		mtod(from, struct osockaddr *)->sa_family =
802 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
803 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
804 			from);
805 	} else
806 		msg.msg_namelen = 0;
807 
808 	if (from != NULL)
809 		m_free(from);
810 
811 	if (error == 0) {
812 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
813 		if (msg.msg_flags < 0)
814 			/* Some flag unsupported by Linux */
815 			error = EINVAL;
816 		else {
817 			ktrkuser("msghdr", &msg, sizeof(msg));
818 			bsd_to_linux_msghdr(&msg, &lmsg);
819 			error = copyout(&lmsg, SCARG(uap, msg), sizeof(lmsg));
820 		}
821 	}
822 
823 	return (error);
824 }
825 
826 /*
827  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
828  * is different, the rest matches IPPROTO_* on both systems.
829  */
830 int
831 linux_to_bsd_sopt_level(int llevel)
832 {
833 
834 	switch (llevel) {
835 	case LINUX_SOL_SOCKET:
836 		return SOL_SOCKET;
837 	case LINUX_SOL_IP:
838 		return IPPROTO_IP;
839 #ifdef INET6
840 	case LINUX_SOL_IPV6:
841 		return IPPROTO_IPV6;
842 #endif
843 	case LINUX_SOL_TCP:
844 		return IPPROTO_TCP;
845 	case LINUX_SOL_UDP:
846 		return IPPROTO_UDP;
847 	default:
848 		return -1;
849 	}
850 }
851 
852 /*
853  * Convert Linux socket level socket option numbers to NetBSD values.
854  */
855 int
856 linux_to_bsd_so_sockopt(int lopt)
857 {
858 
859 	switch (lopt) {
860 	case LINUX_SO_DEBUG:
861 		return SO_DEBUG;
862 	case LINUX_SO_REUSEADDR:
863 		/*
864 		 * Linux does not implement SO_REUSEPORT, but allows reuse of
865 		 * a host:port pair through SO_REUSEADDR even if the address
866 		 * is not a multicast-address. Effectively, this means that we
867 		 * should use SO_REUSEPORT to allow Linux applications to not
868 		 * exit with EADDRINUSE
869 		 */
870 		return SO_REUSEPORT;
871 	case LINUX_SO_TYPE:
872 		return SO_TYPE;
873 	case LINUX_SO_ERROR:
874 		return SO_ERROR;
875 	case LINUX_SO_DONTROUTE:
876 		return SO_DONTROUTE;
877 	case LINUX_SO_BROADCAST:
878 		return SO_BROADCAST;
879 	case LINUX_SO_SNDBUF:
880 		return SO_SNDBUF;
881 	case LINUX_SO_RCVBUF:
882 		return SO_RCVBUF;
883 	case LINUX_SO_KEEPALIVE:
884 		return SO_KEEPALIVE;
885 	case LINUX_SO_OOBINLINE:
886 		return SO_OOBINLINE;
887 	case LINUX_SO_NO_CHECK:
888 	case LINUX_SO_PRIORITY:
889 		return -1;
890 	case LINUX_SO_LINGER:
891 		return SO_LINGER;
892 	case LINUX_SO_BSDCOMPAT:
893 	case LINUX_SO_PASSCRED:
894 	case LINUX_SO_PEERCRED:
895 		return -1;
896 	case LINUX_SO_RCVLOWAT:
897 		return SO_RCVLOWAT;
898 	case LINUX_SO_SNDLOWAT:
899 		return SO_SNDLOWAT;
900 	case LINUX_SO_RCVTIMEO:
901 		return SO_RCVTIMEO;
902 	case LINUX_SO_SNDTIMEO:
903 		return SO_SNDTIMEO;
904 	case LINUX_SO_SECURITY_AUTHENTICATION:
905 	case LINUX_SO_SECURITY_ENCRYPTION_TRANSPORT:
906 	case LINUX_SO_SECURITY_ENCRYPTION_NETWORK:
907 	case LINUX_SO_BINDTODEVICE:
908 	case LINUX_SO_ATTACH_FILTER:
909 	case LINUX_SO_DETACH_FILTER:
910 	case LINUX_SO_PEERNAME:
911 		return -1;
912 	case LINUX_SO_TIMESTAMP:
913 		return SO_TIMESTAMP;
914 	case LINUX_SO_ACCEPTCONN:
915 	case LINUX_SO_PEERSEC:
916 	case LINUX_SO_SNDBUFFORCE:
917 	case LINUX_SO_RCVBUFFORCE:
918 	case LINUX_SO_PASSSEC:
919 	case LINUX_SO_TIMESTAMPNS:
920 	case LINUX_SO_MARK:
921 	case LINUX_SO_TIMESTAMPING:
922 	case LINUX_SO_PROTOCOL:
923 	case LINUX_SO_DOMAIN:
924 	case LINUX_SO_RXQ_OVFL:
925 	case LINUX_SO_WIFI_STATUS:
926 	case LINUX_SO_PEEK_OFF:
927 	case LINUX_SO_NOFCS:
928 	default:
929 		return -1;
930 	}
931 }
932 
933 /*
934  * Convert Linux IP level socket option number to NetBSD values.
935  */
936 int
937 linux_to_bsd_ip_sockopt(int lopt)
938 {
939 
940 	switch (lopt) {
941 	case LINUX_IP_TOS:
942 		return IP_TOS;
943 	case LINUX_IP_TTL:
944 		return IP_TTL;
945 	case LINUX_IP_HDRINCL:
946 		return IP_HDRINCL;
947 	case LINUX_IP_MULTICAST_TTL:
948 		return IP_MULTICAST_TTL;
949 	case LINUX_IP_MULTICAST_LOOP:
950 		return IP_MULTICAST_LOOP;
951 	case LINUX_IP_MULTICAST_IF:
952 		return IP_MULTICAST_IF;
953 	case LINUX_IP_ADD_MEMBERSHIP:
954 		return IP_ADD_MEMBERSHIP;
955 	case LINUX_IP_DROP_MEMBERSHIP:
956 		return IP_DROP_MEMBERSHIP;
957 	default:
958 		return -1;
959 	}
960 }
961 
962 /*
963  * Convert Linux IPV6 level socket option number to NetBSD values.
964  */
965 #ifdef INET6
966 int
967 linux_to_bsd_ipv6_sockopt(int lopt)
968 {
969 
970 	switch (lopt) {
971 	case LINUX_IPV6_V6ONLY:
972 		return IPV6_V6ONLY;
973 	default:
974 		return -1;
975 	}
976 }
977 #endif
978 
979 /*
980  * Convert Linux TCP level socket option number to NetBSD values.
981  */
982 int
983 linux_to_bsd_tcp_sockopt(int lopt)
984 {
985 
986 	switch (lopt) {
987 	case LINUX_TCP_NODELAY:
988 		return TCP_NODELAY;
989 	case LINUX_TCP_MAXSEG:
990 		return TCP_MAXSEG;
991 	default:
992 		return -1;
993 	}
994 }
995 
996 /*
997  * Convert Linux UDP level socket option number to NetBSD values.
998  */
999 int
1000 linux_to_bsd_udp_sockopt(int lopt)
1001 {
1002 
1003 	switch (lopt) {
1004 	default:
1005 		return -1;
1006 	}
1007 }
1008 
1009 /*
1010  * Another reasonably straightforward function: setsockopt(2).
1011  * The level and option numbers are converted; the values passed
1012  * are not (yet) converted, the ones currently implemented don't
1013  * need conversion, as they are the same on both systems.
1014  */
1015 int
1016 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
1017 {
1018 	/* {
1019 		syscallarg(int) s;
1020 		syscallarg(int) level;
1021 		syscallarg(int) optname;
1022 		syscallarg(void *) optval;
1023 		syscallarg(int) optlen;
1024 	} */
1025 	struct sys_setsockopt_args bsa;
1026 	int name;
1027 
1028 	SCARG(&bsa, s) = SCARG(uap, s);
1029 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
1030 	SCARG(&bsa, val) = SCARG(uap, optval);
1031 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
1032 
1033 	/*
1034 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
1035 	 * and returns EOPNOTSUPP for other levels
1036 	 */
1037 	if (SCARG(&bsa, level) != SOL_SOCKET) {
1038 		struct socket *so;
1039 		int error, family;
1040 
1041 		/* fd_getsock() will use the descriptor for us */
1042 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
1043 		    	return error;
1044 		family = so->so_proto->pr_domain->dom_family;
1045 		fd_putfile(SCARG(&bsa, s));
1046 
1047 		if (family == AF_LOCAL)
1048 			return EOPNOTSUPP;
1049 	}
1050 
1051 	switch (SCARG(&bsa, level)) {
1052 	case SOL_SOCKET:
1053 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1054 		break;
1055 	case IPPROTO_IP:
1056 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1057 		break;
1058 #ifdef INET6
1059 	case IPPROTO_IPV6:
1060 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1061 		break;
1062 #endif
1063 	case IPPROTO_TCP:
1064 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1065 		break;
1066 	case IPPROTO_UDP:
1067 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1068 		break;
1069 	default:
1070 		return EINVAL;
1071 	}
1072 
1073 	if (name == -1)
1074 		return EINVAL;
1075 	SCARG(&bsa, name) = name;
1076 
1077 	return sys_setsockopt(l, &bsa, retval);
1078 }
1079 
1080 /*
1081  * getsockopt(2) is very much the same as setsockopt(2) (see above)
1082  */
1083 int
1084 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
1085 {
1086 	/* {
1087 		syscallarg(int) s;
1088 		syscallarg(int) level;
1089 		syscallarg(int) optname;
1090 		syscallarg(void *) optval;
1091 		syscallarg(int *) optlen;
1092 	} */
1093 	struct sys_getsockopt_args bga;
1094 	int name;
1095 
1096 	SCARG(&bga, s) = SCARG(uap, s);
1097 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
1098 	SCARG(&bga, val) = SCARG(uap, optval);
1099 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
1100 
1101 	switch (SCARG(&bga, level)) {
1102 	case SOL_SOCKET:
1103 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1104 		break;
1105 	case IPPROTO_IP:
1106 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1107 		break;
1108 #ifdef INET6
1109 	case IPPROTO_IPV6:
1110 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1111 		break;
1112 #endif
1113 	case IPPROTO_TCP:
1114 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1115 		break;
1116 	case IPPROTO_UDP:
1117 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1118 		break;
1119 	default:
1120 		return EINVAL;
1121 	}
1122 
1123 	if (name == -1)
1124 		return EINVAL;
1125 	SCARG(&bga, name) = name;
1126 
1127 	return sys_getsockopt(l, &bga, retval);
1128 }
1129 
1130 int
1131 linux_getifname(struct lwp *l, register_t *retval, void *data)
1132 {
1133 	struct ifnet *ifp;
1134 	struct linux_ifreq ifr;
1135 	int error;
1136 	int s;
1137 
1138 	error = copyin(data, &ifr, sizeof(ifr));
1139 	if (error)
1140 		return error;
1141 
1142 	s = pserialize_read_enter();
1143 	ifp = if_byindex(ifr.ifr_ifru.ifru_ifindex);
1144 	if (ifp == NULL) {
1145 		pserialize_read_exit(s);
1146 		return ENODEV;
1147 	}
1148 
1149 	strncpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name));
1150 	pserialize_read_exit(s);
1151 
1152 	return copyout(&ifr, data, sizeof(ifr));
1153 }
1154 
1155 int
1156 linux_getifconf(struct lwp *l, register_t *retval, void *data)
1157 {
1158 	struct linux_ifreq ifr, *ifrp = NULL;
1159 	struct linux_ifconf ifc;
1160 	struct ifnet *ifp;
1161 	struct sockaddr *sa;
1162 	struct osockaddr *osa;
1163 	int space = 0, error;
1164 	const int sz = (int)sizeof(ifr);
1165 	bool docopy;
1166 	int s;
1167 	int bound;
1168 	struct psref psref;
1169 
1170 	error = copyin(data, &ifc, sizeof(ifc));
1171 	if (error)
1172 		return error;
1173 
1174 	docopy = ifc.ifc_req != NULL;
1175 	if (docopy) {
1176 		if (ifc.ifc_len < 0)
1177 			return EINVAL;
1178 
1179 		space = ifc.ifc_len;
1180 		ifrp = ifc.ifc_req;
1181 	}
1182 	memset(&ifr, 0, sizeof(ifr));
1183 
1184 	bound = curlwp_bind();
1185 	s = pserialize_read_enter();
1186 	IFNET_READER_FOREACH(ifp) {
1187 		struct ifaddr *ifa;
1188 		if_acquire(ifp, &psref);
1189 		pserialize_read_exit(s);
1190 
1191 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1192 		    sizeof(ifr.ifr_name));
1193 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') {
1194 			error = ENAMETOOLONG;
1195 			goto release_exit;
1196 		}
1197 
1198 		s = pserialize_read_enter();
1199 		IFADDR_READER_FOREACH(ifa, ifp) {
1200 			struct psref psref_ifa;
1201 			ifa_acquire(ifa, &psref_ifa);
1202 			pserialize_read_exit(s);
1203 
1204 			sa = ifa->ifa_addr;
1205 			if (sa->sa_family != AF_INET ||
1206 			    sa->sa_len > sizeof(*osa))
1207 				goto next;
1208 			memcpy(&ifr.ifr_addr, sa, sa->sa_len);
1209 			osa = (struct osockaddr *)&ifr.ifr_addr;
1210 			osa->sa_family = sa->sa_family;
1211 			if (space >= sz) {
1212 				error = copyout(&ifr, ifrp, sz);
1213 				if (error != 0) {
1214 					ifa_release(ifa, &psref_ifa);
1215 					goto release_exit;
1216 				}
1217 				ifrp++;
1218 			}
1219 			space -= sz;
1220 		next:
1221 			s = pserialize_read_enter();
1222 			ifa_release(ifa, &psref_ifa);
1223 		}
1224 
1225 		KASSERT(pserialize_in_read_section());
1226 		if_release(ifp, &psref);
1227 	}
1228 	pserialize_read_exit(s);
1229 	curlwp_bindx(bound);
1230 
1231 	if (docopy)
1232 		ifc.ifc_len -= space;
1233 	else
1234 		ifc.ifc_len = -space;
1235 
1236 	return copyout(&ifc, data, sizeof(ifc));
1237 
1238 release_exit:
1239 	if_release(ifp, &psref);
1240 	curlwp_bindx(bound);
1241 	return error;
1242 }
1243 
1244 int
1245 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1246     void *data)
1247 {
1248 	/* Not the full structure, just enough to map what we do here */
1249 	struct linux_ifreq lreq;
1250 	file_t *fp;
1251 	struct ifaddr *ifa;
1252 	struct ifnet *ifp;
1253 	struct sockaddr_dl *sadl;
1254 	int error, found;
1255 	int index, ifnum;
1256 	int s;
1257 
1258 	/*
1259 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1260 	 * SIOCGIFCONF, because the user buffer is not of the right
1261 	 * type to take those results.  We can't use kernel buffers to
1262 	 * receive the results, as the implementation of sys_ioctl()
1263 	 * and ifconf() [which implements SIOCGIFCONF] use
1264 	 * copyin()/copyout() which will fail on kernel addresses.
1265 	 *
1266 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1267 	 */
1268 
1269 	if ((fp = fd_getfile(fd)) == NULL)
1270 		return (EBADF);
1271 
1272 	KERNEL_LOCK(1, NULL);
1273 
1274 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1275 		error = EBADF;
1276 		goto out;
1277 	}
1278 
1279 	error = copyin(data, &lreq, sizeof(lreq));
1280 	if (error)
1281 		goto out;
1282 	lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0';		/* just in case */
1283 
1284 	/*
1285 	 * Try real interface name first, then fake "ethX"
1286 	 */
1287 	found = 0;
1288 	s = pserialize_read_enter();
1289 	IFNET_READER_FOREACH(ifp) {
1290 		if (found)
1291 			break;
1292 		if (strcmp(lreq.ifr_name, ifp->if_xname))
1293 			/* not this interface */
1294 			continue;
1295 
1296 		found=1;
1297 		if (IFADDR_READER_EMPTY(ifp)) {
1298 			pserialize_read_exit(s);
1299 			error = ENODEV;
1300 			goto out;
1301 		}
1302 		IFADDR_READER_FOREACH(ifa, ifp) {
1303 			sadl = satosdl(ifa->ifa_addr);
1304 			/* only return ethernet addresses */
1305 			/* XXX what about FDDI, etc. ? */
1306 			if (sadl->sdl_family != AF_LINK ||
1307 			    sadl->sdl_type != IFT_ETHER)
1308 				continue;
1309 			memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl),
1310 			       MIN(sadl->sdl_alen,
1311 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1312 			lreq.ifr_hwaddr.sa_family =
1313 				sadl->sdl_family;
1314 			pserialize_read_exit(s);
1315 
1316 			error = copyout(&lreq, data, sizeof(lreq));
1317 			goto out;
1318 		}
1319 	}
1320 	pserialize_read_exit(s);
1321 
1322 	if (strncmp(lreq.ifr_name, "eth", 3) != 0) {
1323 		/* unknown interface, not even an "eth*" name */
1324 		error = ENODEV;
1325 		goto out;
1326 	}
1327 
1328 	for (ifnum = 0, index = 3;
1329 	     index < LINUX_IFNAMSIZ && lreq.ifr_name[index] != '\0';
1330 	     index++) {
1331 		ifnum *= 10;
1332 		ifnum += lreq.ifr_name[index] - '0';
1333 	}
1334 
1335 	error = EINVAL;			/* in case we don't find one */
1336 	s = pserialize_read_enter();
1337 	IFNET_READER_FOREACH(ifp) {
1338 		memcpy(lreq.ifr_name, ifp->if_xname,
1339 		       MIN(LINUX_IFNAMSIZ, IFNAMSIZ));
1340 		IFADDR_READER_FOREACH(ifa, ifp) {
1341 			sadl = satosdl(ifa->ifa_addr);
1342 			/* only return ethernet addresses */
1343 			/* XXX what about FDDI, etc. ? */
1344 			if (sadl->sdl_family != AF_LINK ||
1345 			    sadl->sdl_type != IFT_ETHER)
1346 				continue;
1347 			if (ifnum--)
1348 				/* not the reqested iface */
1349 				continue;
1350 			memcpy(&lreq.ifr_hwaddr.sa_data,
1351 			       CLLADDR(sadl),
1352 			       MIN(sadl->sdl_alen,
1353 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1354 			lreq.ifr_hwaddr.sa_family =
1355 				sadl->sdl_family;
1356 			pserialize_read_exit(s);
1357 
1358 			error = copyout(&lreq, data, sizeof(lreq));
1359 			goto out;
1360 		}
1361 	}
1362 	pserialize_read_exit(s);
1363 
1364 out:
1365 	KERNEL_UNLOCK_ONE(NULL);
1366 	fd_putfile(fd);
1367 	return error;
1368 }
1369 
1370 int
1371 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1372 {
1373 	/* {
1374 		syscallarg(int) fd;
1375 		syscallarg(u_long) com;
1376 		syscallarg(void *) data;
1377 	} */
1378 	u_long com;
1379 	int error = 0, isdev = 0, dosys = 1;
1380 	struct sys_ioctl_args ia;
1381 	file_t *fp;
1382 	struct vnode *vp;
1383 	int (*ioctlf)(file_t *, u_long, void *);
1384 	struct ioctl_pt pt;
1385 
1386 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1387 		return (EBADF);
1388 
1389 	if (fp->f_type == DTYPE_VNODE) {
1390 		vp = (struct vnode *)fp->f_data;
1391 		isdev = vp->v_type == VCHR;
1392 	}
1393 
1394 	/*
1395 	 * Don't try to interpret socket ioctl calls that are done
1396 	 * on a device filedescriptor, just pass them through, to
1397 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1398 	 * device will only handle these if it's prepared to do
1399 	 * so, to avoid unexpected things from happening.
1400 	 */
1401 	if (isdev) {
1402 		dosys = 0;
1403 		ioctlf = fp->f_ops->fo_ioctl;
1404 		pt.com = SCARG(uap, com);
1405 		pt.data = SCARG(uap, data);
1406 		error = ioctlf(fp, PTIOCLINUX, &pt);
1407 		/*
1408 		 * XXX hack: if the function returns EJUSTRETURN,
1409 		 * it has stuffed a sysctl return value in pt.data.
1410 		 */
1411 		if (error == EJUSTRETURN) {
1412 			retval[0] = (register_t)pt.data;
1413 			error = 0;
1414 		}
1415 		goto out;
1416 	}
1417 
1418 	com = SCARG(uap, com);
1419 	retval[0] = 0;
1420 
1421 	switch (com) {
1422 	case LINUX_SIOCGIFNAME:
1423 		error = linux_getifname(l, retval, SCARG(uap, data));
1424 		dosys = 0;
1425 		break;
1426 	case LINUX_SIOCGIFCONF:
1427 		error = linux_getifconf(l, retval, SCARG(uap, data));
1428 		dosys = 0;
1429 		break;
1430 	case LINUX_SIOCGIFFLAGS:
1431 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1432 		break;
1433 	case LINUX_SIOCSIFFLAGS:
1434 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1435 		break;
1436 	case LINUX_SIOCGIFADDR:
1437 		SCARG(&ia, com) = OOSIOCGIFADDR;
1438 		break;
1439 	case LINUX_SIOCGIFDSTADDR:
1440 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1441 		break;
1442 	case LINUX_SIOCGIFBRDADDR:
1443 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1444 		break;
1445 	case LINUX_SIOCGIFNETMASK:
1446 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1447 		break;
1448 	case LINUX_SIOCGIFMTU:
1449 		SCARG(&ia, com) = OSIOCGIFMTU;
1450 		break;
1451 	case LINUX_SIOCADDMULTI:
1452 		SCARG(&ia, com) = OSIOCADDMULTI;
1453 		break;
1454 	case LINUX_SIOCDELMULTI:
1455 		SCARG(&ia, com) = OSIOCDELMULTI;
1456 		break;
1457 	case LINUX_SIOCGIFHWADDR:
1458 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1459 		    SCARG(uap, data));
1460 		dosys = 0;
1461 		break;
1462 	default:
1463 		error = EINVAL;
1464 	}
1465 
1466  out:
1467  	fd_putfile(SCARG(uap, fd));
1468 
1469 	if (error ==0 && dosys) {
1470 		SCARG(&ia, fd) = SCARG(uap, fd);
1471 		SCARG(&ia, data) = SCARG(uap, data);
1472 		error = sys_ioctl(curlwp, &ia, retval);
1473 	}
1474 
1475 	return error;
1476 }
1477 
1478 int
1479 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1480 {
1481 	/* {
1482 		syscallarg(int) s;
1483 		syscallarg(const struct sockaddr *) name;
1484 		syscallarg(int) namelen;
1485 	} */
1486 	int		error;
1487 	struct sockaddr_big sb;
1488 
1489 	error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name),
1490 	    SCARG(uap, namelen));
1491 	if (error)
1492 		return (error);
1493 
1494 	error = do_sys_connect(l, SCARG(uap, s), (struct sockaddr *)&sb);
1495 
1496 	if (error == EISCONN) {
1497 		struct socket *so;
1498 		int state, prflags;
1499 
1500 		/* fd_getsock() will use the descriptor for us */
1501 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1502 		    	return EISCONN;
1503 
1504 		solock(so);
1505 		state = so->so_state;
1506 		prflags = so->so_proto->pr_flags;
1507 		sounlock(so);
1508 		fd_putfile(SCARG(uap, s));
1509 		/*
1510 		 * We should only let this call succeed once per
1511 		 * non-blocking connect; however we don't have
1512 		 * a convenient place to keep that state..
1513 		 */
1514 		if ((state & (SS_ISCONNECTED|SS_NBIO)) ==
1515 		    (SS_ISCONNECTED|SS_NBIO) &&
1516 		    (prflags & PR_CONNREQUIRED))
1517 			return 0;
1518 	}
1519 
1520 	return (error);
1521 }
1522 
1523 int
1524 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1525 {
1526 	/* {
1527 		syscallarg(int) s;
1528 		syscallarg(const struct osockaddr *) name;
1529 		syscallarg(int) namelen;
1530 	} */
1531 	int		error;
1532 	struct sockaddr_big sb;
1533 
1534 	error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name),
1535 	    SCARG(uap, namelen));
1536 	if (error)
1537 		return (error);
1538 
1539 	return do_sys_bind(l, SCARG(uap, s), (struct sockaddr *)&sb);
1540 }
1541 
1542 int
1543 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1544 {
1545 	/* {
1546 		syscallarg(int) fdes;
1547 		syscallarg(void *) asa;
1548 		syscallarg(int *) alen;
1549 	} */
1550 	int error;
1551 
1552 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1553 		return (error);
1554 
1555 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1556 		return (error);
1557 
1558 	return (0);
1559 }
1560 
1561 int
1562 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1563 {
1564 	/* {
1565 		syscallarg(int) fdes;
1566 		syscallarg(void *) asa;
1567 		syscallarg(int *) alen;
1568 	} */
1569 	int error;
1570 
1571 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1572 		return (error);
1573 
1574 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1575 		return (error);
1576 
1577 	return (0);
1578 }
1579 
1580 /*
1581  * Copy the osockaddr structure pointed to by name to sb, adjust
1582  * family and convert to sockaddr.
1583  */
1584 static int
1585 linux_get_sa(struct lwp *l, int s, struct sockaddr_big *sb,
1586     const struct osockaddr *name, socklen_t namelen)
1587 {
1588 	int error, bdom;
1589 
1590 	if (namelen > UCHAR_MAX ||
1591 	    namelen <= offsetof(struct sockaddr_big, sb_data))
1592 		return EINVAL;
1593 
1594 	error = copyin(name, sb, namelen);
1595 	if (error)
1596 		return error;
1597 
1598 	bdom = linux_to_bsd_domain(sb->sb_family);
1599 	if (bdom == -1)
1600 		return EINVAL;
1601 
1602 	/*
1603 	 * If the family is unspecified, use address family of the socket.
1604 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1605 	 */
1606 	if (bdom == AF_UNSPEC) {
1607 		struct socket *so;
1608 
1609 		/* fd_getsock() will use the descriptor for us */
1610 		if ((error = fd_getsock(s, &so)) != 0)
1611 			return error;
1612 
1613 		bdom = so->so_proto->pr_domain->dom_family;
1614 		fd_putfile(s);
1615 	}
1616 
1617 	/*
1618 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1619 	 * which lacks the scope id compared with RFC2553 one. If we detect
1620 	 * the situation, reject the address and write a message to system log.
1621 	 *
1622 	 * Still accept addresses for which the scope id is not used.
1623 	 */
1624 	if (bdom == AF_INET6 &&
1625 	    namelen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
1626 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sb;
1627 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1628 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1629 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1630 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1631 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1632 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1633 			struct proc *p = l->l_proc;
1634 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1635 
1636 			log(LOG_DEBUG,
1637 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1638 			    "sockaddr_in6 rejected",
1639 			    p->p_pid, p->p_comm, uid);
1640 			return EINVAL;
1641 		}
1642 		namelen = sizeof(struct sockaddr_in6);
1643 		sin6->sin6_scope_id = 0;
1644 	}
1645 
1646 	/*
1647 	 * Linux is less strict than NetBSD and permits namelen to be larger
1648 	 * than valid struct sockaddr_in*.  If this is the case, truncate
1649 	 * the value to the correct size, so that NetBSD networking does not
1650 	 * return an error.
1651 	 */
1652 	switch (bdom) {
1653 	case AF_INET:
1654 		namelen = MIN(namelen, sizeof(struct sockaddr_in));
1655 		break;
1656 	case AF_INET6:
1657 		namelen = MIN(namelen, sizeof(struct sockaddr_in6));
1658 		break;
1659 	}
1660 
1661 	sb->sb_family = bdom;
1662 	sb->sb_len = namelen;
1663 	ktrkuser("mbsoname", sb, namelen);
1664 	return 0;
1665 }
1666 
1667 static int
1668 linux_sa_put(struct osockaddr *osa)
1669 {
1670 	struct sockaddr sa;
1671 	struct osockaddr *kosa;
1672 	int error, bdom, len;
1673 
1674 	/*
1675 	 * Only read/write the sockaddr family and length part, the rest is
1676 	 * not changed.
1677 	 */
1678 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1679 
1680 	error = copyin(osa, &sa, len);
1681 	if (error)
1682 		return (error);
1683 
1684 	bdom = bsd_to_linux_domain(sa.sa_family);
1685 	if (bdom == -1)
1686 		return (EINVAL);
1687 
1688 	/* Note: we convert from sockaddr to osockaddr here, too */
1689 	kosa = (struct osockaddr *) &sa;
1690 	kosa->sa_family = bdom;
1691 	error = copyout(kosa, osa, len);
1692 	if (error)
1693 		return (error);
1694 
1695 	return (0);
1696 }
1697 
1698 #ifndef __amd64__
1699 int
1700 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1701 {
1702 	/* {
1703 		syscallarg(int) s;
1704 		syscallarg(void *) buf;
1705 		syscallarg(int) len;
1706 		syscallarg(int) flags;
1707 	} */
1708 	struct sys_recvfrom_args bra;
1709 
1710 
1711 	SCARG(&bra, s) = SCARG(uap, s);
1712 	SCARG(&bra, buf) = SCARG(uap, buf);
1713 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1714 	SCARG(&bra, flags) = SCARG(uap, flags);
1715 	SCARG(&bra, from) = NULL;
1716 	SCARG(&bra, fromlenaddr) = NULL;
1717 
1718 	return (sys_recvfrom(l, &bra, retval));
1719 }
1720 
1721 int
1722 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1723 {
1724 	/* {
1725 		syscallarg(int) s;
1726 		syscallarg(void *) buf;
1727 		syscallarg(int) len;
1728 		syscallarg(int) flags;
1729 	} */
1730 	struct sys_sendto_args bsa;
1731 
1732 	SCARG(&bsa, s)		= SCARG(uap, s);
1733 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1734 	SCARG(&bsa, len)	= SCARG(uap, len);
1735 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1736 	SCARG(&bsa, to)		= NULL;
1737 	SCARG(&bsa, tolen)	= 0;
1738 
1739 	return (sys_sendto(l, &bsa, retval));
1740 }
1741 #endif
1742 
1743 int
1744 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1745 {
1746 	/* {
1747 		syscallarg(int) s;
1748 		syscallarg(struct osockaddr *) name;
1749 		syscallarg(int *) anamelen;
1750 	} */
1751 	int error;
1752 	struct sys_accept_args baa;
1753 
1754 	SCARG(&baa, s)		= SCARG(uap, s);
1755 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1756 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1757 
1758 	if ((error = sys_accept(l, &baa, retval)))
1759 		return (error);
1760 
1761 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1762 		return (error);
1763 
1764 	return (0);
1765 }
1766 
1767 int
1768 linux_sys_accept4(struct lwp *l, const struct linux_sys_accept4_args *uap, register_t *retval)
1769 {
1770 	/* {
1771 		syscallarg(int) s;
1772 		syscallarg(struct osockaddr *) name;
1773 		syscallarg(int *) anamelen;
1774 		syscallarg(int) flags;
1775 	} */
1776 	int error, flags;
1777 	struct sockaddr_big name;
1778 
1779 	if ((flags = linux_to_bsd_type(SCARG(uap, flags))) == -1)
1780 		return EINVAL;
1781 
1782 	name.sb_len = UCHAR_MAX;
1783 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
1784 	    retval, NULL, flags, 0);
1785 	if (error != 0)
1786 		return error;
1787 
1788 	error = copyout_sockname_sb((struct sockaddr *)SCARG(uap, name),
1789 	    SCARG(uap, anamelen), MSG_LENUSRSPACE, &name);
1790 	if (error != 0) {
1791 		int fd = (int)*retval;
1792 		if (fd_getfile(fd) != NULL)
1793 			(void)fd_close(fd);
1794 		return error;
1795 	}
1796 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1797 		return error;
1798 
1799 	return 0;
1800 }
1801 
1802 int
1803 linux_sys_sendmmsg(struct lwp *l, const struct linux_sys_sendmmsg_args *uap,
1804     register_t *retval)
1805 {
1806 	/* {
1807 		syscallarg(int) s;
1808 		syscallarg(struct linux_mmsghdr *) msgvec;
1809 		syscallarg(unsigned int) vlen;
1810 		syscallarg(unsigned int) flags;
1811 	} */
1812 	struct linux_mmsghdr lmsg;
1813 	struct mmsghdr bmsg;
1814 	struct socket *so;
1815 	file_t *fp;
1816 	struct msghdr *msg = &bmsg.msg_hdr;
1817 	int error, s;
1818 	unsigned int vlen, flags, dg;
1819 
1820 	if ((flags = linux_to_bsd_msg_flags(SCARG(uap, flags))) == -1)
1821 		return EINVAL;
1822 
1823 	flags = (flags & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1824 
1825 	s = SCARG(uap, s);
1826 	if ((error = fd_getsock1(s, &so, &fp)) != 0)
1827 		return error;
1828 
1829 	vlen = SCARG(uap, vlen);
1830 	if (vlen > 1024)
1831 		vlen = 1024;
1832 
1833 	for (dg = 0; dg < vlen;) {
1834 		error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg));
1835 		if (error)
1836 			break;
1837 		linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr);
1838 
1839 		msg->msg_flags = flags;
1840 
1841 		error = do_sys_sendmsg_so(l, s, so, fp, msg, flags, retval);
1842 		if (error)
1843 			break;
1844 
1845 		ktrkuser("msghdr", msg, sizeof *msg);
1846 		lmsg.msg_len = *retval;
1847 		error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg));
1848 		if (error)
1849 			break;
1850 		dg++;
1851 
1852 	}
1853 
1854 	*retval = dg;
1855 
1856 	fd_putfile(s);
1857 
1858 	/*
1859 	 * If we succeeded at least once, return 0.
1860 	 */
1861 	if (dg)
1862 		return 0;
1863 	return error;
1864 }
1865 
1866 int
1867 linux_sys_recvmmsg(struct lwp *l, const struct linux_sys_recvmmsg_args *uap,
1868     register_t *retval)
1869 {
1870 	/* {
1871 		syscallarg(int) s;
1872 		syscallarg(struct linux_mmsghdr *) msgvec;
1873 		syscallarg(unsigned int) vlen;
1874 		syscallarg(unsigned int) flags;
1875 		syscallarg(struct linux_timespec *) timeout;
1876 	} */
1877 	struct linux_mmsghdr lmsg;
1878 	struct mmsghdr bmsg;
1879 	struct socket *so;
1880 	struct msghdr *msg = &bmsg.msg_hdr;
1881 	int error, s;
1882 	struct mbuf *from, *control;
1883 	struct timespec ts = {0}, now;
1884 	struct linux_timespec lts;
1885 	unsigned int vlen, flags, dg;
1886 
1887 	if (SCARG(uap, timeout)) {
1888 		error = copyin(SCARG(uap, timeout), &lts, sizeof(lts));
1889 			return error;
1890 		ts.tv_sec = lts.tv_sec;
1891 		ts.tv_nsec = lts.tv_nsec;
1892 		getnanotime(&now);
1893 		timespecadd(&now, &ts, &ts);
1894 	}
1895 
1896 	s = SCARG(uap, s);
1897 	if ((error = fd_getsock(s, &so)) != 0)
1898 		return error;
1899 
1900 	/*
1901 	 * If so->so_rerror holds a deferred error return it now.
1902 	 */
1903 	if (so->so_rerror) {
1904 		error = so->so_rerror;
1905 		so->so_rerror = 0;
1906 		fd_putfile(s);
1907 		return error;
1908 	}
1909 
1910 	vlen = SCARG(uap, vlen);
1911 	if (vlen > 1024)
1912 		vlen = 1024;
1913 
1914 	from = NULL;
1915 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1916 
1917 	for (dg = 0; dg < vlen;) {
1918 		error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg));
1919 		if (error)
1920 			break;
1921 		linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr);
1922 		msg->msg_flags = flags & ~MSG_WAITFORONE;
1923 
1924 		if (from != NULL) {
1925 			m_free(from);
1926 			from = NULL;
1927 		}
1928 
1929 		error = do_sys_recvmsg_so(l, s, so, msg, &from,
1930 		    msg->msg_control != NULL ? &control : NULL, retval);
1931 		if (error) {
1932 			if (error == EAGAIN && dg > 0)
1933 				error = 0;
1934 			break;
1935 		}
1936 
1937 		if (msg->msg_control != NULL)
1938 			error = linux_copyout_msg_control(l, msg, control);
1939 		if (error)
1940 			break;
1941 
1942 		if (from != NULL) {
1943 			mtod(from, struct osockaddr *)->sa_family =
1944 			    bsd_to_linux_domain(mtod(from,
1945 			    struct sockaddr *)->sa_family);
1946 			error = copyout_sockname(msg->msg_name,
1947 			    &msg->msg_namelen, 0, from);
1948 			if (error)
1949 				break;
1950 		}
1951 
1952 
1953 		lmsg.msg_len = *retval;
1954 		ktrkuser("msghdr", msg, sizeof(*msg));
1955 		bsd_to_linux_msghdr(msg, &lmsg.msg_hdr);
1956 		error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg));
1957 		if (error)
1958 			break;
1959 
1960 		dg++;
1961 		if (msg->msg_flags & MSG_OOB)
1962 			break;
1963 
1964 		if (SCARG(uap, timeout)) {
1965 			getnanotime(&now);
1966 			timespecsub(&now, &ts, &now);
1967 			if (now.tv_sec > 0)
1968 				break;
1969 		}
1970 
1971 		if (flags & MSG_WAITFORONE)
1972 			flags |= MSG_DONTWAIT;
1973 
1974 	}
1975 
1976 	if (from != NULL)
1977 		m_free(from);
1978 
1979 	*retval = dg;
1980 
1981 	/*
1982 	 * If we succeeded at least once, return 0, hopefully so->so_rerror
1983 	 * will catch it next time.
1984 	 */
1985 	if (error && dg > 0) {
1986 		so->so_rerror = error;
1987 		error = 0;
1988 	}
1989 
1990 	fd_putfile(s);
1991 
1992 	return error;
1993 }
1994