xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision 88fcb00c0357f2d7c1774f86a352637bfda96184)
1 /*	$NetBSD: linux_socket.c,v 1.109 2011/06/30 20:09:39 wiz 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 1.109 2011/06/30 20:09:39 wiz Exp $");
39 
40 #if defined(_KERNEL_OPT)
41 #include "opt_inet.h"
42 #endif /* defined(_KERNEL_OPT) */
43 
44 #include <sys/param.h>
45 #include <sys/kernel.h>
46 #include <sys/systm.h>
47 #include <sys/buf.h>
48 #include <sys/malloc.h>
49 #include <sys/ioctl.h>
50 #include <sys/tty.h>
51 #include <sys/file.h>
52 #include <sys/filedesc.h>
53 #include <sys/select.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/domain.h>
57 #include <net/if.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <netinet/in.h>
61 #include <netinet/tcp.h>
62 #include <sys/mount.h>
63 #include <sys/proc.h>
64 #include <sys/vnode.h>
65 #include <sys/device.h>
66 #include <sys/protosw.h>
67 #include <sys/mbuf.h>
68 #include <sys/syslog.h>
69 #include <sys/exec.h>
70 #include <sys/kauth.h>
71 #include <sys/syscallargs.h>
72 #include <sys/ktrace.h>
73 #include <sys/fcntl.h>
74 
75 #include <lib/libkern/libkern.h>
76 
77 #include <netinet/ip6.h>
78 #include <netinet6/ip6_var.h>
79 
80 #include <compat/sys/socket.h>
81 #include <compat/sys/sockio.h>
82 
83 #include <compat/linux/common/linux_types.h>
84 #include <compat/linux/common/linux_util.h>
85 #include <compat/linux/common/linux_signal.h>
86 #include <compat/linux/common/linux_ioctl.h>
87 #include <compat/linux/common/linux_socket.h>
88 #include <compat/linux/common/linux_fcntl.h>
89 #if !defined(__alpha__) && !defined(__amd64__)
90 #include <compat/linux/common/linux_socketcall.h>
91 #endif
92 #include <compat/linux/common/linux_sockio.h>
93 #include <compat/linux/common/linux_ipc.h>
94 #include <compat/linux/common/linux_sem.h>
95 
96 #include <compat/linux/linux_syscallargs.h>
97 
98 #ifdef DEBUG_LINUX
99 #define DPRINTF(a) uprintf a
100 #else
101 #define DPRINTF(a)
102 #endif
103 
104 /*
105  * The calls in this file are entered either via the linux_socketcall()
106  * interface or, on the Alpha, as individual syscalls.  The
107  * linux_socketcall function does any massaging of arguments so that all
108  * the calls in here need not think that they are anything other
109  * than a normal syscall.
110  */
111 
112 static int linux_to_bsd_domain(int);
113 static int bsd_to_linux_domain(int);
114 int linux_to_bsd_sopt_level(int);
115 int linux_to_bsd_so_sockopt(int);
116 int linux_to_bsd_ip_sockopt(int);
117 int linux_to_bsd_tcp_sockopt(int);
118 int linux_to_bsd_udp_sockopt(int);
119 int linux_getifname(struct lwp *, register_t *, void *);
120 int linux_getifconf(struct lwp *, register_t *, void *);
121 int linux_getifhwaddr(struct lwp *, register_t *, u_int, void *);
122 static int linux_get_sa(struct lwp *, int, struct mbuf **,
123 		const struct osockaddr *, unsigned int);
124 static int linux_sa_put(struct osockaddr *osa);
125 static int linux_to_bsd_msg_flags(int);
126 static int bsd_to_linux_msg_flags(int);
127 static void linux_to_bsd_msghdr(struct linux_msghdr *, struct msghdr *);
128 static void bsd_to_linux_msghdr(struct msghdr *, struct linux_msghdr *);
129 
130 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
131 	AF_UNSPEC,
132 	AF_UNIX,
133 	AF_INET,
134 	AF_CCITT,	/* LINUX_AF_AX25 */
135 	AF_IPX,
136 	AF_APPLETALK,
137 	-1,		/* LINUX_AF_NETROM */
138 	-1,		/* LINUX_AF_BRIDGE */
139 	-1,		/* LINUX_AF_ATMPVC */
140 	AF_CCITT,	/* LINUX_AF_X25 */
141 	AF_INET6,
142 	-1,		/* LINUX_AF_ROSE */
143 	AF_DECnet,
144 	-1,		/* LINUX_AF_NETBEUI */
145 	-1,		/* LINUX_AF_SECURITY */
146 	pseudo_AF_KEY,
147 	AF_ROUTE,	/* LINUX_AF_NETLINK */
148 	-1,		/* LINUX_AF_PACKET */
149 	-1,		/* LINUX_AF_ASH */
150 	-1,		/* LINUX_AF_ECONET */
151 	-1,		/* LINUX_AF_ATMSVC */
152 	AF_SNA,
153 	/* rest up to LINUX_AF_MAX-1 is not allocated */
154 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
155 };
156 
157 static const int bsd_to_linux_domain_[AF_MAX] = {
158 	LINUX_AF_UNSPEC,
159 	LINUX_AF_UNIX,
160 	LINUX_AF_INET,
161 	-1,		/* AF_IMPLINK */
162 	-1,		/* AF_PUP */
163 	-1,		/* AF_CHAOS */
164 	-1,		/* AF_NS */
165 	-1,		/* AF_ISO */
166 	-1,		/* AF_ECMA */
167 	-1,		/* AF_DATAKIT */
168 	LINUX_AF_AX25,	/* AF_CCITT */
169 	LINUX_AF_SNA,
170 	LINUX_AF_DECnet,
171 	-1,		/* AF_DLI */
172 	-1,		/* AF_LAT */
173 	-1,		/* AF_HYLINK */
174 	LINUX_AF_APPLETALK,
175 	LINUX_AF_NETLINK,
176 	-1,		/* AF_LINK */
177 	-1,		/* AF_XTP */
178 	-1,		/* AF_COIP */
179 	-1,		/* AF_CNT */
180 	-1,		/* pseudo_AF_RTIP */
181 	LINUX_AF_IPX,
182 	LINUX_AF_INET6,
183 	-1,		/* pseudo_AF_PIP */
184 	-1,		/* AF_ISDN */
185 	-1,		/* AF_NATM */
186 	-1,		/* AF_ARP */
187 	LINUX_pseudo_AF_KEY,
188 	-1,		/* pseudo_AF_HDRCMPLT */
189 };
190 
191 static const struct {
192 	int bfl;
193 	int lfl;
194 } bsd_to_linux_msg_flags_[] = {
195 	{MSG_OOB,		LINUX_MSG_OOB},
196 	{MSG_PEEK,		LINUX_MSG_PEEK},
197 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
198 	{MSG_EOR,		LINUX_MSG_EOR},
199 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
200 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
201 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
202 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
203 	{MSG_BCAST,		0},		/* not supported, clear */
204 	{MSG_MCAST,		0},		/* not supported, clear */
205 	{MSG_NOSIGNAL,		LINUX_MSG_NOSIGNAL},
206 	{-1, /* not supp */	LINUX_MSG_PROBE},
207 	{-1, /* not supp */	LINUX_MSG_FIN},
208 	{-1, /* not supp */	LINUX_MSG_SYN},
209 	{-1, /* not supp */	LINUX_MSG_CONFIRM},
210 	{-1, /* not supp */	LINUX_MSG_RST},
211 	{-1, /* not supp */	LINUX_MSG_ERRQUEUE},
212 	{-1, /* not supp */	LINUX_MSG_MORE},
213 };
214 
215 /*
216  * Convert between Linux and BSD socket domain values
217  */
218 static int
219 linux_to_bsd_domain(int ldom)
220 {
221 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
222 		return (-1);
223 
224 	return linux_to_bsd_domain_[ldom];
225 }
226 
227 /*
228  * Convert between BSD and Linux socket domain values
229  */
230 static int
231 bsd_to_linux_domain(int bdom)
232 {
233 	if (bdom < 0 || bdom >= AF_MAX)
234 		return (-1);
235 
236 	return bsd_to_linux_domain_[bdom];
237 }
238 
239 static int
240 linux_to_bsd_msg_flags(int lflag)
241 {
242 	int i, lfl, bfl;
243 	int bflag = 0;
244 
245 	if (lflag == 0)
246 		return (0);
247 
248 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
249 		bfl = bsd_to_linux_msg_flags_[i].bfl;
250 		lfl = bsd_to_linux_msg_flags_[i].lfl;
251 
252 		if (lfl == 0)
253 			continue;
254 
255 		if (lflag & lfl) {
256 			if (bfl < 0)
257 				return (-1);
258 
259 			bflag |= bfl;
260 		}
261 	}
262 
263 	return (bflag);
264 }
265 
266 static int
267 bsd_to_linux_msg_flags(int bflag)
268 {
269 	int i, lfl, bfl;
270 	int lflag = 0;
271 
272 	if (bflag == 0)
273 		return (0);
274 
275 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
276 		bfl = bsd_to_linux_msg_flags_[i].bfl;
277 		lfl = bsd_to_linux_msg_flags_[i].lfl;
278 
279 		if (bfl <= 0)
280 			continue;
281 
282 		if (bflag & bfl) {
283 			if (lfl < 0)
284 				return (-1);
285 
286 			lflag |= lfl;
287 		}
288 	}
289 
290 	return (lflag);
291 }
292 
293 int
294 linux_sys_socket(struct lwp *l, const struct linux_sys_socket_args *uap, register_t *retval)
295 {
296 	/* {
297 		syscallarg(int)	domain;
298 		syscallarg(int)	type;
299 		syscallarg(int) protocol;
300 	} */
301 	struct sys___socket30_args bsa;
302 	struct sys_fcntl_args fsa;
303 	register_t fretval[2];
304 	int error, flags;
305 
306 
307 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
308 	SCARG(&bsa, type) = SCARG(uap, type) & LINUX_SOCK_TYPE_MASK;
309 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
310 	if (SCARG(&bsa, domain) == -1)
311 		return EINVAL;
312 	flags = SCARG(uap, type) & ~LINUX_SOCK_TYPE_MASK;
313 	if (flags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
314 		return EINVAL;
315 	error = sys___socket30(l, &bsa, retval);
316 
317 	/*
318 	 * Linux overloads the "type" parameter to include some
319 	 * fcntl flags to be set on the file descriptor.
320 	 * Process those if creating the socket succeeded.
321 	 */
322 
323 	if (!error && flags & LINUX_SOCK_CLOEXEC) {
324 		SCARG(&fsa, fd) = *retval;
325 		SCARG(&fsa, cmd) = F_SETFD;
326 		SCARG(&fsa, arg) = (void *)(uintptr_t)FD_CLOEXEC;
327 		(void) sys_fcntl(l, &fsa, fretval);
328 	}
329 	if (!error && flags & LINUX_SOCK_NONBLOCK) {
330 		SCARG(&fsa, fd) = *retval;
331 		SCARG(&fsa, cmd) = F_SETFL;
332 		SCARG(&fsa, arg) = (void *)(uintptr_t)O_NONBLOCK;
333 		error = sys_fcntl(l, &fsa, fretval);
334 		if (error) {
335 			struct sys_close_args csa;
336 
337 			SCARG(&csa, fd) = *retval;
338 			(void) sys_close(l, &csa, fretval);
339 		}
340 	}
341 
342 #ifdef INET6
343 	/*
344 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
345 	 * default and some apps depend on this. So, set V6ONLY to 0
346 	 * for Linux apps if the sysctl value is set to 1.
347 	 */
348 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
349 		struct socket *so;
350 
351 		if (fd_getsock(*retval, &so) == 0) {
352 			int val = 0;
353 
354 			/* ignore error */
355 			(void)so_setsockopt(l, so, IPPROTO_IPV6, IPV6_V6ONLY,
356 			    &val, sizeof(val));
357 
358 			fd_putfile(*retval);
359 		}
360 	}
361 #endif
362 
363 	return (error);
364 }
365 
366 int
367 linux_sys_socketpair(struct lwp *l, const struct linux_sys_socketpair_args *uap, register_t *retval)
368 {
369 	/* {
370 		syscallarg(int) domain;
371 		syscallarg(int) type;
372 		syscallarg(int) protocol;
373 		syscallarg(int *) rsv;
374 	} */
375 	struct sys_socketpair_args bsa;
376 
377 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
378 	if (SCARG(&bsa, domain) == -1)
379 		return EINVAL;
380 	SCARG(&bsa, type) = SCARG(uap, type);
381 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
382 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
383 
384 	return sys_socketpair(l, &bsa, retval);
385 }
386 
387 int
388 linux_sys_sendto(struct lwp *l, const struct linux_sys_sendto_args *uap, register_t *retval)
389 {
390 	/* {
391 		syscallarg(int)				s;
392 		syscallarg(void *)			msg;
393 		syscallarg(int)				len;
394 		syscallarg(int)				flags;
395 		syscallarg(struct osockaddr *)		to;
396 		syscallarg(int)				tolen;
397 	} */
398 	struct msghdr   msg;
399 	struct iovec    aiov;
400 	struct mbuf *nam;
401 	int bflags;
402 	int error;
403 
404 	/* Translate message flags.  */
405 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
406 	if (bflags < 0)
407 		/* Some supported flag */
408 		return EINVAL;
409 
410 	msg.msg_flags = 0;
411 	msg.msg_name = NULL;
412 	msg.msg_control = NULL;
413 
414 	if (SCARG(uap, tolen)) {
415 		/* Read in and convert the sockaddr */
416 		error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
417 		    SCARG(uap, tolen));
418 		if (error)
419 			return (error);
420 		msg.msg_flags |= MSG_NAMEMBUF;
421 		msg.msg_name = nam;
422 		msg.msg_namelen = SCARG(uap, tolen);
423 	}
424 
425 	msg.msg_iov = &aiov;
426 	msg.msg_iovlen = 1;
427 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
428 	aiov.iov_len = SCARG(uap, len);
429 
430 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
431 }
432 
433 static void
434 linux_to_bsd_msghdr(struct linux_msghdr *lmsg, struct msghdr *bmsg)
435 {
436 	bmsg->msg_name = lmsg->msg_name;
437 	bmsg->msg_namelen = lmsg->msg_namelen;
438 	bmsg->msg_iov = lmsg->msg_iov;
439 	bmsg->msg_iovlen = lmsg->msg_iovlen;
440 	bmsg->msg_control = lmsg->msg_control;
441 	bmsg->msg_controllen = lmsg->msg_controllen;
442 	bmsg->msg_flags = lmsg->msg_flags;
443 }
444 
445 static void
446 bsd_to_linux_msghdr(struct msghdr *bmsg, struct linux_msghdr *lmsg)
447 {
448 	lmsg->msg_name = bmsg->msg_name;
449 	lmsg->msg_namelen = bmsg->msg_namelen;
450 	lmsg->msg_iov = bmsg->msg_iov;
451 	lmsg->msg_iovlen = bmsg->msg_iovlen;
452 	lmsg->msg_control = bmsg->msg_control;
453 	lmsg->msg_controllen = bmsg->msg_controllen;
454 	lmsg->msg_flags = bmsg->msg_flags;
455 }
456 
457 int
458 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
459 {
460 	/* {
461 		syscallarg(int) s;
462 		syscallarg(struct linux_msghdr *) msg;
463 		syscallarg(u_int) flags;
464 	} */
465 	struct msghdr	msg;
466 	struct linux_msghdr lmsg;
467 	int		error;
468 	int		bflags;
469 	struct mbuf     *nam;
470 	u_int8_t	*control;
471 	struct mbuf     *ctl_mbuf = NULL;
472 
473 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
474 	if (error)
475 		return error;
476 	linux_to_bsd_msghdr(&lmsg, &msg);
477 
478 	msg.msg_flags = MSG_IOVUSRSPACE;
479 
480 	/*
481 	 * Translate message flags.
482 	 */
483 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
484 	if (bflags < 0)
485 		/* Some supported flag */
486 		return EINVAL;
487 
488 	if (lmsg.msg_name) {
489 		/* Read in and convert the sockaddr */
490 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
491 		    msg.msg_namelen);
492 		if (error)
493 			return (error);
494 		msg.msg_flags |= MSG_NAMEMBUF;
495 		msg.msg_name = nam;
496 	}
497 
498 	/*
499 	 * Handle cmsg if there is any.
500 	 */
501 	if (LINUX_CMSG_FIRSTHDR(&lmsg)) {
502 		struct linux_cmsghdr l_cmsg, *l_cc;
503 		struct cmsghdr *cmsg;
504 		ssize_t resid = msg.msg_controllen;
505 		size_t clen, cidx = 0, cspace;
506 
507 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
508 		clen = MLEN;
509 		control = mtod(ctl_mbuf, void *);
510 
511 		l_cc = LINUX_CMSG_FIRSTHDR(&lmsg);
512 		do {
513 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
514 			if (error)
515 				goto done;
516 
517 			/*
518 			 * Sanity check the control message length.
519 			 */
520 			if (l_cmsg.cmsg_len > resid
521 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
522 				error = EINVAL;
523 				goto done;
524 			}
525 
526 			/*
527 			 * Refuse unsupported control messages, and
528 			 * translate fields as appropriate.
529 			 */
530 			switch (l_cmsg.cmsg_level) {
531 			case LINUX_SOL_SOCKET:
532 				/* It only differs on some archs */
533 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
534 					l_cmsg.cmsg_level = SOL_SOCKET;
535 
536 				switch(l_cmsg.cmsg_type) {
537 				case LINUX_SCM_RIGHTS:
538 					/* Linux SCM_RIGHTS is same as NetBSD */
539 					break;
540 
541 				case LINUX_SCM_CREDENTIALS:
542 					/* no native equivalent, just drop it */
543 					m_free(ctl_mbuf);
544 					ctl_mbuf = NULL;
545 					msg.msg_control = NULL;
546 					msg.msg_controllen = 0;
547 					goto skipcmsg;
548 
549 				default:
550 					/* other types not supported */
551 					error = EINVAL;
552 					goto done;
553 				}
554 				break;
555 			default:
556 				/* pray and leave intact */
557 				break;
558 			}
559 
560 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
561 
562 			/* Check the buffer is big enough */
563 			if (__predict_false(cidx + cspace > clen)) {
564 				u_int8_t *nc;
565 
566 				clen = cidx + cspace;
567 				if (clen >= PAGE_SIZE) {
568 					error = EINVAL;
569 					goto done;
570 				}
571 				nc = realloc(clen <= MLEN ? NULL : control,
572 						clen, M_TEMP, M_WAITOK);
573 				if (!nc) {
574 					error = ENOMEM;
575 					goto done;
576 				}
577 				if (cidx <= MLEN)
578 					/* Old buffer was in mbuf... */
579 					memcpy(nc, control, cidx);
580 				control = nc;
581 			}
582 
583 			/* Copy header */
584 			cmsg = (void *)&control[cidx];
585 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
586 			cmsg->cmsg_level = l_cmsg.cmsg_level;
587 			cmsg->cmsg_type = l_cmsg.cmsg_type;
588 
589 			/* Zero area between header and data */
590 			memset(cmsg + 1, 0,
591 				CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg));
592 
593 			/* Copyin the data */
594 			error = copyin(LINUX_CMSG_DATA(l_cc),
595 				CMSG_DATA(control),
596 				l_cmsg.cmsg_len - sizeof(l_cmsg));
597 			if (error)
598 				goto done;
599 
600 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
601 			cidx += cspace;
602 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
603 
604 		/* If we allocated a buffer, attach to mbuf */
605 		if (cidx > MLEN) {
606 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
607 			ctl_mbuf->m_flags |= M_EXT_RW;
608 		}
609 		control = NULL;
610 		ctl_mbuf->m_len = cidx;
611 
612 		msg.msg_control = ctl_mbuf;
613 		msg.msg_flags |= MSG_CONTROLMBUF;
614 
615 		ktrkuser("msgcontrol", mtod(ctl_mbuf, void *),
616 		    msg.msg_controllen);
617 	}
618 
619 skipcmsg:
620 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
621 	/* Freed internally */
622 	ctl_mbuf = NULL;
623 
624 done:
625 	if (ctl_mbuf != NULL) {
626 		if (control != NULL && control != mtod(ctl_mbuf, void *))
627 			free(control, M_MBUF);
628 		m_free(ctl_mbuf);
629 	}
630 	return (error);
631 }
632 
633 int
634 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
635 {
636 	/* {
637 		syscallarg(int) s;
638 		syscallarg(void *) buf;
639 		syscallarg(int) len;
640 		syscallarg(int) flags;
641 		syscallarg(struct osockaddr *) from;
642 		syscallarg(int *) fromlenaddr;
643 	} */
644 	int		error;
645 	struct sys_recvfrom_args bra;
646 
647 	SCARG(&bra, s) = SCARG(uap, s);
648 	SCARG(&bra, buf) = SCARG(uap, buf);
649 	SCARG(&bra, len) = SCARG(uap, len);
650 	SCARG(&bra, flags) = SCARG(uap, flags);
651 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
652 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
653 
654 	if ((error = sys_recvfrom(l, &bra, retval)))
655 		return (error);
656 
657 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
658 		return (error);
659 
660 	return (0);
661 }
662 
663 static int
664 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
665 {
666 	int dlen, error = 0;
667 	struct cmsghdr *cmsg;
668 	struct linux_cmsghdr linux_cmsg;
669 	struct mbuf *m;
670 	char *q, *q_end;
671 
672 	if (mp->msg_controllen <= 0 || control == 0) {
673 		mp->msg_controllen = 0;
674 		free_control_mbuf(l, control, control);
675 		return 0;
676 	}
677 
678 	ktrkuser("msgcontrol", mtod(control, void *), mp->msg_controllen);
679 
680 	q = (char *)mp->msg_control;
681 	q_end = q + mp->msg_controllen;
682 
683 	for (m = control; m != NULL; ) {
684 		cmsg = mtod(m, struct cmsghdr *);
685 
686 		/*
687 		 * Fixup cmsg. We handle two things:
688 		 * 0. different sizeof cmsg_len.
689 		 * 1. different values for level/type on some archs
690 		 * 2. different alignment of CMSG_DATA on some archs
691 		 */
692 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
693 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
694 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
695 
696 		dlen = q_end - q;
697 		if (linux_cmsg.cmsg_len > dlen) {
698 			/* Not enough room for the parameter */
699 			dlen -= sizeof linux_cmsg;
700 			if (dlen <= 0)
701 				/* Discard if header wont fit */
702 				break;
703 			mp->msg_flags |= MSG_CTRUNC;
704 			if (linux_cmsg.cmsg_level == SOL_SOCKET
705 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
706 				/* Do not truncate me ... */
707 				break;
708 		} else
709 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
710 
711 		switch (linux_cmsg.cmsg_level) {
712 		case SOL_SOCKET:
713 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
714 			switch (linux_cmsg.cmsg_type) {
715 			case SCM_RIGHTS:
716 				/* Linux SCM_RIGHTS is same as NetBSD */
717 				break;
718 
719 			default:
720 				/* other types not supported */
721 				error = EINVAL;
722 				goto done;
723 			}
724 			/* machine dependent ! */
725 			break;
726 		default:
727 			/* pray and leave intact */
728 			break;
729 		}
730 
731 		/* There can be padding between the header and data... */
732 		error = copyout(&linux_cmsg, q, sizeof linux_cmsg);
733 		if (error != 0) {
734 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
735 			    dlen);
736 		}
737 		if (error != 0) {
738 			/* We must free all the SCM_RIGHTS */
739 			m = control;
740 			break;
741 		}
742 		m = m->m_next;
743 		if (m == NULL || q + LINUX_CMSG_SPACE(dlen) > q_end) {
744 			q += LINUX_CMSG_LEN(dlen);
745 			break;
746 		}
747 		q += LINUX_CMSG_SPACE(dlen);
748 	}
749 
750   done:
751 	free_control_mbuf(l, control, m);
752 
753 	mp->msg_controllen = q - (char *)mp->msg_control;
754 	return error;
755 }
756 
757 int
758 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
759 {
760 	/* {
761 		syscallarg(int) s;
762 		syscallarg(struct linux_msghdr *) msg;
763 		syscallarg(u_int) flags;
764 	} */
765 	struct msghdr	msg;
766 	struct linux_msghdr lmsg;
767 	int		error;
768 	struct mbuf	*from, *control;
769 
770 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
771 	if (error)
772 		return (error);
773 	linux_to_bsd_msghdr(&lmsg, &msg);
774 
775 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
776 	if (msg.msg_flags < 0) {
777 		/* Some unsupported flag */
778 		return (EINVAL);
779 	}
780 	msg.msg_flags |= MSG_IOVUSRSPACE;
781 
782 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
783 	    msg.msg_control != NULL ? &control : NULL, retval);
784 	if (error != 0)
785 		return error;
786 
787 	if (msg.msg_control != NULL)
788 		error = linux_copyout_msg_control(l, &msg, control);
789 
790 	if (error == 0 && from != 0) {
791 		mtod(from, struct osockaddr *)->sa_family =
792 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
793 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
794 			from);
795 	} else
796 		msg.msg_namelen = 0;
797 
798 	if (from != NULL)
799 		m_free(from);
800 
801 	if (error == 0) {
802 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
803 		if (msg.msg_flags < 0)
804 			/* Some flag unsupported by Linux */
805 			error = EINVAL;
806 		else {
807 			ktrkuser("msghdr", &msg, sizeof(msg));
808 			bsd_to_linux_msghdr(&msg, &lmsg);
809 			error = copyout(&lmsg, SCARG(uap, msg), sizeof(lmsg));
810 		}
811 	}
812 
813 	return (error);
814 }
815 
816 /*
817  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
818  * is different, the rest matches IPPROTO_* on both systems.
819  */
820 int
821 linux_to_bsd_sopt_level(int llevel)
822 {
823 
824 	switch (llevel) {
825 	case LINUX_SOL_SOCKET:
826 		return SOL_SOCKET;
827 	case LINUX_SOL_IP:
828 		return IPPROTO_IP;
829 	case LINUX_SOL_TCP:
830 		return IPPROTO_TCP;
831 	case LINUX_SOL_UDP:
832 		return IPPROTO_UDP;
833 	default:
834 		return -1;
835 	}
836 }
837 
838 /*
839  * Convert Linux socket level socket option numbers to NetBSD values.
840  */
841 int
842 linux_to_bsd_so_sockopt(int lopt)
843 {
844 
845 	switch (lopt) {
846 	case LINUX_SO_DEBUG:
847 		return SO_DEBUG;
848 	case LINUX_SO_REUSEADDR:
849 		/*
850 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
851 		 * host:port pair through SO_REUSEADDR even if the address is not a
852 		 * multicast-address.  Effectively, this means that we should use
853 		 * SO_REUSEPORT to allow Linux applications to not exit with
854 		 * EADDRINUSE
855 		 */
856 		return SO_REUSEPORT;
857 	case LINUX_SO_TYPE:
858 		return SO_TYPE;
859 	case LINUX_SO_ERROR:
860 		return SO_ERROR;
861 	case LINUX_SO_DONTROUTE:
862 		return SO_DONTROUTE;
863 	case LINUX_SO_BROADCAST:
864 		return SO_BROADCAST;
865 	case LINUX_SO_SNDBUF:
866 		return SO_SNDBUF;
867 	case LINUX_SO_RCVBUF:
868 		return SO_RCVBUF;
869 	case LINUX_SO_KEEPALIVE:
870 		return SO_KEEPALIVE;
871 	case LINUX_SO_OOBINLINE:
872 		return SO_OOBINLINE;
873 	case LINUX_SO_LINGER:
874 		return SO_LINGER;
875 	case LINUX_SO_PRIORITY:
876 	case LINUX_SO_NO_CHECK:
877 	default:
878 		return -1;
879 	}
880 }
881 
882 /*
883  * Convert Linux IP level socket option number to NetBSD values.
884  */
885 int
886 linux_to_bsd_ip_sockopt(int lopt)
887 {
888 
889 	switch (lopt) {
890 	case LINUX_IP_TOS:
891 		return IP_TOS;
892 	case LINUX_IP_TTL:
893 		return IP_TTL;
894 	case LINUX_IP_HDRINCL:
895 		return IP_HDRINCL;
896 	case LINUX_IP_MULTICAST_TTL:
897 		return IP_MULTICAST_TTL;
898 	case LINUX_IP_MULTICAST_LOOP:
899 		return IP_MULTICAST_LOOP;
900 	case LINUX_IP_MULTICAST_IF:
901 		return IP_MULTICAST_IF;
902 	case LINUX_IP_ADD_MEMBERSHIP:
903 		return IP_ADD_MEMBERSHIP;
904 	case LINUX_IP_DROP_MEMBERSHIP:
905 		return IP_DROP_MEMBERSHIP;
906 	default:
907 		return -1;
908 	}
909 }
910 
911 /*
912  * Convert Linux TCP level socket option number to NetBSD values.
913  */
914 int
915 linux_to_bsd_tcp_sockopt(int lopt)
916 {
917 
918 	switch (lopt) {
919 	case LINUX_TCP_NODELAY:
920 		return TCP_NODELAY;
921 	case LINUX_TCP_MAXSEG:
922 		return TCP_MAXSEG;
923 	default:
924 		return -1;
925 	}
926 }
927 
928 /*
929  * Convert Linux UDP level socket option number to NetBSD values.
930  */
931 int
932 linux_to_bsd_udp_sockopt(int lopt)
933 {
934 
935 	switch (lopt) {
936 	default:
937 		return -1;
938 	}
939 }
940 
941 /*
942  * Another reasonably straightforward function: setsockopt(2).
943  * The level and option numbers are converted; the values passed
944  * are not (yet) converted, the ones currently implemented don't
945  * need conversion, as they are the same on both systems.
946  */
947 int
948 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
949 {
950 	/* {
951 		syscallarg(int) s;
952 		syscallarg(int) level;
953 		syscallarg(int) optname;
954 		syscallarg(void *) optval;
955 		syscallarg(int) optlen;
956 	} */
957 	struct sys_setsockopt_args bsa;
958 	int name;
959 
960 	SCARG(&bsa, s) = SCARG(uap, s);
961 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
962 	SCARG(&bsa, val) = SCARG(uap, optval);
963 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
964 
965 	/*
966 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
967 	 * and returns EOPNOTSUPP for other levels
968 	 */
969 	if (SCARG(&bsa, level) != SOL_SOCKET) {
970 		struct socket *so;
971 		int error, family;
972 
973 		/* fd_getsock() will use the descriptor for us */
974 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
975 		    	return error;
976 		family = so->so_proto->pr_domain->dom_family;
977 		fd_putfile(SCARG(&bsa, s));
978 
979 		if (family == AF_LOCAL)
980 			return EOPNOTSUPP;
981 	}
982 
983 	switch (SCARG(&bsa, level)) {
984 	case SOL_SOCKET:
985 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
986 		break;
987 	case IPPROTO_IP:
988 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
989 		break;
990 	case IPPROTO_TCP:
991 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
992 		break;
993 	case IPPROTO_UDP:
994 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
995 		break;
996 	default:
997 		return EINVAL;
998 	}
999 
1000 	if (name == -1)
1001 		return EINVAL;
1002 	SCARG(&bsa, name) = name;
1003 
1004 	return sys_setsockopt(l, &bsa, retval);
1005 }
1006 
1007 /*
1008  * getsockopt(2) is very much the same as setsockopt(2) (see above)
1009  */
1010 int
1011 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
1012 {
1013 	/* {
1014 		syscallarg(int) s;
1015 		syscallarg(int) level;
1016 		syscallarg(int) optname;
1017 		syscallarg(void *) optval;
1018 		syscallarg(int *) optlen;
1019 	} */
1020 	struct sys_getsockopt_args bga;
1021 	int name;
1022 
1023 	SCARG(&bga, s) = SCARG(uap, s);
1024 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
1025 	SCARG(&bga, val) = SCARG(uap, optval);
1026 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
1027 
1028 	switch (SCARG(&bga, level)) {
1029 	case SOL_SOCKET:
1030 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1031 		break;
1032 	case IPPROTO_IP:
1033 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1034 		break;
1035 	case IPPROTO_TCP:
1036 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1037 		break;
1038 	case IPPROTO_UDP:
1039 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1040 		break;
1041 	default:
1042 		return EINVAL;
1043 	}
1044 
1045 	if (name == -1)
1046 		return EINVAL;
1047 	SCARG(&bga, name) = name;
1048 
1049 	return sys_getsockopt(l, &bga, retval);
1050 }
1051 
1052 int
1053 linux_getifname(struct lwp *l, register_t *retval, void *data)
1054 {
1055 	struct ifnet *ifp;
1056 	struct linux_ifreq ifr;
1057 	int error;
1058 
1059 	error = copyin(data, &ifr, sizeof(ifr));
1060 	if (error)
1061 		return error;
1062 
1063 	if (ifr.ifr_ifru.ifru_ifindex >= if_indexlim)
1064 		return ENODEV;
1065 
1066 	ifp = ifindex2ifnet[ifr.ifr_ifru.ifru_ifindex];
1067 	if (ifp == NULL)
1068 		return ENODEV;
1069 
1070 	strncpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name));
1071 
1072 	return copyout(&ifr, data, sizeof(ifr));
1073 }
1074 
1075 int
1076 linux_getifconf(struct lwp *l, register_t *retval, void *data)
1077 {
1078 	struct linux_ifreq ifr, *ifrp;
1079 	struct ifconf *ifc = data;
1080 	struct ifnet *ifp;
1081 	struct ifaddr *ifa;
1082 	struct sockaddr *sa;
1083 	struct osockaddr *osa;
1084 	int space, error = 0;
1085 	const int sz = (int)sizeof(ifr);
1086 
1087 	ifrp = (struct linux_ifreq *)ifc->ifc_req;
1088 	if (ifrp == NULL)
1089 		space = 0;
1090 	else
1091 		space = ifc->ifc_len;
1092 
1093 	IFNET_FOREACH(ifp) {
1094 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1095 		    sizeof(ifr.ifr_name));
1096 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
1097 			return ENAMETOOLONG;
1098 		if (IFADDR_EMPTY(ifp))
1099 			continue;
1100 		IFADDR_FOREACH(ifa, ifp) {
1101 			sa = ifa->ifa_addr;
1102 			if (sa->sa_family != AF_INET ||
1103 			    sa->sa_len > sizeof(*osa))
1104 				continue;
1105 			memcpy(&ifr.ifr_addr, sa, sa->sa_len);
1106 			osa = (struct osockaddr *)&ifr.ifr_addr;
1107 			osa->sa_family = sa->sa_family;
1108 			if (space >= sz) {
1109 				error = copyout(&ifr, ifrp, sz);
1110 				if (error != 0)
1111 					return error;
1112 				ifrp++;
1113 			}
1114 			space -= sz;
1115 		}
1116 	}
1117 
1118 	if (ifrp != NULL)
1119 		ifc->ifc_len -= space;
1120 	else
1121 		ifc->ifc_len = -space;
1122 
1123 	return 0;
1124 }
1125 
1126 int
1127 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1128     void *data)
1129 {
1130 	/* Not the full structure, just enough to map what we do here */
1131 	struct linux_ifreq lreq;
1132 	file_t *fp;
1133 	struct ifaddr *ifa;
1134 	struct ifnet *ifp;
1135 	struct sockaddr_dl *sadl;
1136 	int error, found;
1137 	int index, ifnum;
1138 
1139 	/*
1140 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1141 	 * SIOCGIFCONF, because the user buffer is not of the right
1142 	 * type to take those results.  We can't use kernel buffers to
1143 	 * receive the results, as the implementation of sys_ioctl()
1144 	 * and ifconf() [which implements SIOCGIFCONF] use
1145 	 * copyin()/copyout() which will fail on kernel addresses.
1146 	 *
1147 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1148 	 */
1149 
1150 	if ((fp = fd_getfile(fd)) == NULL)
1151 		return (EBADF);
1152 
1153 	KERNEL_LOCK(1, NULL);
1154 
1155 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1156 		error = EBADF;
1157 		goto out;
1158 	}
1159 
1160 	error = copyin(data, &lreq, sizeof(lreq));
1161 	if (error)
1162 		goto out;
1163 	lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0';		/* just in case */
1164 
1165 	/*
1166 	 * Try real interface name first, then fake "ethX"
1167 	 */
1168 	found = 0;
1169 	IFNET_FOREACH(ifp) {
1170 		if (found)
1171 			break;
1172 		if (strcmp(lreq.ifr_name, ifp->if_xname))
1173 			/* not this interface */
1174 			continue;
1175 		found=1;
1176 		if (IFADDR_EMPTY(ifp)) {
1177 			error = ENODEV;
1178 			goto out;
1179 		}
1180 		IFADDR_FOREACH(ifa, ifp) {
1181 			sadl = satosdl(ifa->ifa_addr);
1182 			/* only return ethernet addresses */
1183 			/* XXX what about FDDI, etc. ? */
1184 			if (sadl->sdl_family != AF_LINK ||
1185 			    sadl->sdl_type != IFT_ETHER)
1186 				continue;
1187 			memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl),
1188 			       MIN(sadl->sdl_alen,
1189 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1190 			lreq.ifr_hwaddr.sa_family =
1191 				sadl->sdl_family;
1192 			error = copyout(&lreq, data, sizeof(lreq));
1193 			goto out;
1194 		}
1195 	}
1196 
1197 	if (strncmp(lreq.ifr_name, "eth", 3) == 0) {
1198 		for (ifnum = 0, index = 3;
1199 		     lreq.ifr_name[index] != '\0' && index < LINUX_IFNAMSIZ;
1200 		     index++) {
1201 			ifnum *= 10;
1202 			ifnum += lreq.ifr_name[index] - '0';
1203 		}
1204 
1205 		error = EINVAL;			/* in case we don't find one */
1206 		found = 0;
1207 		IFNET_FOREACH(ifp) {
1208 			if (found)
1209 				break;
1210 			memcpy(lreq.ifr_name, ifp->if_xname,
1211 			       MIN(LINUX_IFNAMSIZ, IFNAMSIZ));
1212 			IFADDR_FOREACH(ifa, ifp) {
1213 				sadl = satosdl(ifa->ifa_addr);
1214 				/* only return ethernet addresses */
1215 				/* XXX what about FDDI, etc. ? */
1216 				if (sadl->sdl_family != AF_LINK ||
1217 				    sadl->sdl_type != IFT_ETHER)
1218 					continue;
1219 				if (ifnum--)
1220 					/* not the reqested iface */
1221 					continue;
1222 				memcpy(&lreq.ifr_hwaddr.sa_data,
1223 				       CLLADDR(sadl),
1224 				       MIN(sadl->sdl_alen,
1225 					   sizeof(lreq.ifr_hwaddr.sa_data)));
1226 				lreq.ifr_hwaddr.sa_family =
1227 					sadl->sdl_family;
1228 				error = copyout(&lreq, data, sizeof(lreq));
1229 				found = 1;
1230 				break;
1231 			}
1232 		}
1233 	} else {
1234 		/* unknown interface, not even an "eth*" name */
1235 		error = ENODEV;
1236 	}
1237 
1238 out:
1239 	KERNEL_UNLOCK_ONE(NULL);
1240 	fd_putfile(fd);
1241 	return error;
1242 }
1243 
1244 int
1245 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1246 {
1247 	/* {
1248 		syscallarg(int) fd;
1249 		syscallarg(u_long) com;
1250 		syscallarg(void *) data;
1251 	} */
1252 	u_long com;
1253 	int error = 0, isdev = 0, dosys = 1;
1254 	struct sys_ioctl_args ia;
1255 	file_t *fp;
1256 	struct vnode *vp;
1257 	int (*ioctlf)(file_t *, u_long, void *);
1258 	struct ioctl_pt pt;
1259 
1260 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1261 		return (EBADF);
1262 
1263 	if (fp->f_type == DTYPE_VNODE) {
1264 		vp = (struct vnode *)fp->f_data;
1265 		isdev = vp->v_type == VCHR;
1266 	}
1267 
1268 	/*
1269 	 * Don't try to interpret socket ioctl calls that are done
1270 	 * on a device filedescriptor, just pass them through, to
1271 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1272 	 * device will only handle these if it's prepared to do
1273 	 * so, to avoid unexpected things from happening.
1274 	 */
1275 	if (isdev) {
1276 		dosys = 0;
1277 		ioctlf = fp->f_ops->fo_ioctl;
1278 		pt.com = SCARG(uap, com);
1279 		pt.data = SCARG(uap, data);
1280 		error = ioctlf(fp, PTIOCLINUX, &pt);
1281 		/*
1282 		 * XXX hack: if the function returns EJUSTRETURN,
1283 		 * it has stuffed a sysctl return value in pt.data.
1284 		 */
1285 		if (error == EJUSTRETURN) {
1286 			retval[0] = (register_t)pt.data;
1287 			error = 0;
1288 		}
1289 		goto out;
1290 	}
1291 
1292 	com = SCARG(uap, com);
1293 	retval[0] = 0;
1294 
1295 	switch (com) {
1296 	case LINUX_SIOCGIFNAME:
1297 		error = linux_getifname(l, retval, SCARG(uap, data));
1298 		dosys = 0;
1299 		break;
1300 	case LINUX_SIOCGIFCONF:
1301 		error = linux_getifconf(l, retval, SCARG(uap, data));
1302 		dosys = 0;
1303 		break;
1304 	case LINUX_SIOCGIFFLAGS:
1305 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1306 		break;
1307 	case LINUX_SIOCSIFFLAGS:
1308 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1309 		break;
1310 	case LINUX_SIOCGIFADDR:
1311 		SCARG(&ia, com) = OOSIOCGIFADDR;
1312 		break;
1313 	case LINUX_SIOCGIFDSTADDR:
1314 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1315 		break;
1316 	case LINUX_SIOCGIFBRDADDR:
1317 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1318 		break;
1319 	case LINUX_SIOCGIFNETMASK:
1320 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1321 		break;
1322 	case LINUX_SIOCGIFMTU:
1323 		SCARG(&ia, com) = OSIOCGIFMTU;
1324 		break;
1325 	case LINUX_SIOCADDMULTI:
1326 		SCARG(&ia, com) = OSIOCADDMULTI;
1327 		break;
1328 	case LINUX_SIOCDELMULTI:
1329 		SCARG(&ia, com) = OSIOCDELMULTI;
1330 		break;
1331 	case LINUX_SIOCGIFHWADDR:
1332 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1333 		    SCARG(uap, data));
1334 		dosys = 0;
1335 		break;
1336 	default:
1337 		error = EINVAL;
1338 	}
1339 
1340  out:
1341  	fd_putfile(SCARG(uap, fd));
1342 
1343 	if (error ==0 && dosys) {
1344 		SCARG(&ia, fd) = SCARG(uap, fd);
1345 		SCARG(&ia, data) = SCARG(uap, data);
1346 		error = sys_ioctl(curlwp, &ia, retval);
1347 	}
1348 
1349 	return error;
1350 }
1351 
1352 int
1353 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1354 {
1355 	/* {
1356 		syscallarg(int) s;
1357 		syscallarg(const struct sockaddr *) name;
1358 		syscallarg(int) namelen;
1359 	} */
1360 	int		error;
1361 	struct mbuf *nam;
1362 
1363 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1364 	    SCARG(uap, namelen));
1365 	if (error)
1366 		return (error);
1367 
1368 	error = do_sys_connect(l, SCARG(uap, s), nam);
1369 
1370 	if (error == EISCONN) {
1371 		struct socket *so;
1372 		int state, prflags, nbio;
1373 
1374 		/* fd_getsock() will use the descriptor for us */
1375 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1376 		    	return EISCONN;
1377 
1378 		solock(so);
1379 		state = so->so_state;
1380 		nbio = so->so_nbio;
1381 		prflags = so->so_proto->pr_flags;
1382 		sounlock(so);
1383 		fd_putfile(SCARG(uap, s));
1384 		/*
1385 		 * We should only let this call succeed once per
1386 		 * non-blocking connect; however we don't have
1387 		 * a convenient place to keep that state..
1388 		 */
1389 		if (nbio && (state & SS_ISCONNECTED) &&
1390 		    (prflags & PR_CONNREQUIRED))
1391 			return 0;
1392 	}
1393 
1394 	return (error);
1395 }
1396 
1397 int
1398 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1399 {
1400 	/* {
1401 		syscallarg(int) s;
1402 		syscallarg(const struct osockaddr *) name;
1403 		syscallarg(int) namelen;
1404 	} */
1405 	int		error;
1406 	struct mbuf     *nam;
1407 
1408 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1409 	    SCARG(uap, namelen));
1410 	if (error)
1411 		return (error);
1412 
1413 	return do_sys_bind(l, SCARG(uap, s), nam);
1414 }
1415 
1416 int
1417 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1418 {
1419 	/* {
1420 		syscallarg(int) fdes;
1421 		syscallarg(void *) asa;
1422 		syscallarg(int *) alen;
1423 	} */
1424 	int error;
1425 
1426 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1427 		return (error);
1428 
1429 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1430 		return (error);
1431 
1432 	return (0);
1433 }
1434 
1435 int
1436 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1437 {
1438 	/* {
1439 		syscallarg(int) fdes;
1440 		syscallarg(void *) asa;
1441 		syscallarg(int *) alen;
1442 	} */
1443 	int error;
1444 
1445 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1446 		return (error);
1447 
1448 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1449 		return (error);
1450 
1451 	return (0);
1452 }
1453 
1454 /*
1455  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1456  * family and convert to sockaddr.
1457  */
1458 static int
1459 linux_get_sa(struct lwp *l, int s, struct mbuf **mp,
1460     const struct osockaddr *osa, unsigned int salen)
1461 {
1462 	int error, bdom;
1463 	struct sockaddr *sa;
1464 	struct osockaddr *kosa;
1465 	struct mbuf *m;
1466 
1467 	if (salen == 1 || salen > UCHAR_MAX) {
1468 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1469 		return EINVAL;
1470 	}
1471 
1472 	/* We'll need the address in an mbuf later, so copy into one here */
1473 	m = m_get(M_WAIT, MT_SONAME);
1474 	if (salen > MLEN)
1475 		MEXTMALLOC(m, salen, M_WAITOK);
1476 
1477 	m->m_len = salen;
1478 
1479 	if (salen == 0) {
1480 		*mp = m;
1481 		return 0;
1482 	}
1483 
1484 	kosa = mtod(m, void *);
1485 	if ((error = copyin(osa, kosa, salen))) {
1486 		DPRINTF(("error %d copying osa %p len %d\n",
1487 				error, osa, salen));
1488 		goto bad;
1489 	}
1490 
1491 	ktrkuser("linux sockaddr", kosa, salen);
1492 
1493 	bdom = linux_to_bsd_domain(kosa->sa_family);
1494 	if (bdom == -1) {
1495 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1496 		error = EINVAL;
1497 		goto bad;
1498 	}
1499 
1500 	/*
1501 	 * If the family is unspecified, use address family of the socket.
1502 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1503 	 */
1504 	if (bdom == AF_UNSPEC) {
1505 		struct socket *so;
1506 
1507 		/* fd_getsock() will use the descriptor for us */
1508 		if ((error = fd_getsock(s, &so)) != 0)
1509 			goto bad;
1510 
1511 		bdom = so->so_proto->pr_domain->dom_family;
1512 		fd_putfile(s);
1513 
1514 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1515 	}
1516 
1517 	/*
1518 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1519 	 * which lacks the scope id compared with RFC2553 one. If we detect
1520 	 * the situation, reject the address and write a message to system log.
1521 	 *
1522 	 * Still accept addresses for which the scope id is not used.
1523 	 */
1524 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1525 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1526 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1527 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1528 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1529 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1530 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1531 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1532 			struct proc *p = l->l_proc;
1533 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1534 
1535 			log(LOG_DEBUG,
1536 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1537 			    "sockaddr_in6 rejected",
1538 			    p->p_pid, p->p_comm, uid);
1539 			error = EINVAL;
1540 			goto bad;
1541 		}
1542 		salen = sizeof (struct sockaddr_in6);
1543 		sin6->sin6_scope_id = 0;
1544 	}
1545 
1546 	if (bdom == AF_INET)
1547 		salen = sizeof(struct sockaddr_in);
1548 
1549 	sa = (struct sockaddr *) kosa;
1550 	sa->sa_family = bdom;
1551 	sa->sa_len = salen;
1552 	m->m_len = salen;
1553 	ktrkuser("new sockaddr", kosa, salen);
1554 
1555 #ifdef DEBUG_LINUX
1556 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1557 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1558 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1559 	DPRINTF(("\n"));
1560 #endif
1561 
1562 	*mp = m;
1563 	return 0;
1564 
1565     bad:
1566 	m_free(m);
1567 	return error;
1568 }
1569 
1570 static int
1571 linux_sa_put(struct osockaddr *osa)
1572 {
1573 	struct sockaddr sa;
1574 	struct osockaddr *kosa;
1575 	int error, bdom, len;
1576 
1577 	/*
1578 	 * Only read/write the sockaddr family and length part, the rest is
1579 	 * not changed.
1580 	 */
1581 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1582 
1583 	error = copyin(osa, &sa, len);
1584 	if (error)
1585 		return (error);
1586 
1587 	bdom = bsd_to_linux_domain(sa.sa_family);
1588 	if (bdom == -1)
1589 		return (EINVAL);
1590 
1591 	/* Note: we convert from sockaddr to osockaddr here, too */
1592 	kosa = (struct osockaddr *) &sa;
1593 	kosa->sa_family = bdom;
1594 	error = copyout(kosa, osa, len);
1595 	if (error)
1596 		return (error);
1597 
1598 	return (0);
1599 }
1600 
1601 #ifndef __amd64__
1602 int
1603 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1604 {
1605 	/* {
1606 		syscallarg(int) s;
1607 		syscallarg(void *) buf;
1608 		syscallarg(int) len;
1609 		syscallarg(int) flags;
1610 	} */
1611 	struct sys_recvfrom_args bra;
1612 
1613 
1614 	SCARG(&bra, s) = SCARG(uap, s);
1615 	SCARG(&bra, buf) = SCARG(uap, buf);
1616 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1617 	SCARG(&bra, flags) = SCARG(uap, flags);
1618 	SCARG(&bra, from) = NULL;
1619 	SCARG(&bra, fromlenaddr) = NULL;
1620 
1621 	return (sys_recvfrom(l, &bra, retval));
1622 }
1623 
1624 int
1625 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1626 {
1627 	/* {
1628 		syscallarg(int) s;
1629 		syscallarg(void *) buf;
1630 		syscallarg(int) len;
1631 		syscallarg(int) flags;
1632 	} */
1633 	struct sys_sendto_args bsa;
1634 
1635 	SCARG(&bsa, s)		= SCARG(uap, s);
1636 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1637 	SCARG(&bsa, len)	= SCARG(uap, len);
1638 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1639 	SCARG(&bsa, to)		= NULL;
1640 	SCARG(&bsa, tolen)	= 0;
1641 
1642 	return (sys_sendto(l, &bsa, retval));
1643 }
1644 #endif
1645 
1646 int
1647 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1648 {
1649 	/* {
1650 		syscallarg(int) s;
1651 		syscallarg(struct osockaddr *) name;
1652 		syscallarg(int *) anamelen;
1653 	} */
1654 	int error;
1655 	struct sys_accept_args baa;
1656 
1657 	SCARG(&baa, s)		= SCARG(uap, s);
1658 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1659 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1660 
1661 	if ((error = sys_accept(l, &baa, retval)))
1662 		return (error);
1663 
1664 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1665 		return (error);
1666 
1667 	return (0);
1668 }
1669