xref: /netbsd-src/sys/compat/linux/common/linux_socket.c (revision 0df165c04d0a9ca1adde9ed2b890344c937954a6)
1 /*	$NetBSD: linux_socket.c,v 1.80 2007/10/19 18:52:12 njoly Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 1998 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Frank van der Linden and Eric Haszlakiewicz.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the NetBSD
21  *	Foundation, Inc. and its contributors.
22  * 4. Neither the name of The NetBSD Foundation nor the names of its
23  *    contributors may be used to endorse or promote products derived
24  *    from this software without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36  * POSSIBILITY OF SUCH DAMAGE.
37  */
38 
39 /*
40  * Functions in multiarch:
41  *	linux_sys_socketcall		: linux_socketcall.c
42  */
43 
44 #include <sys/cdefs.h>
45 __KERNEL_RCSID(0, "$NetBSD: linux_socket.c,v 1.80 2007/10/19 18:52:12 njoly Exp $");
46 
47 #if defined(_KERNEL_OPT)
48 #include "opt_inet.h"
49 #endif /* defined(_KERNEL_OPT) */
50 
51 #include <sys/param.h>
52 #include <sys/kernel.h>
53 #include <sys/systm.h>
54 #include <sys/buf.h>
55 #include <sys/malloc.h>
56 #include <sys/ioctl.h>
57 #include <sys/tty.h>
58 #include <sys/file.h>
59 #include <sys/filedesc.h>
60 #include <sys/select.h>
61 #include <sys/socket.h>
62 #include <sys/socketvar.h>
63 #include <sys/domain.h>
64 #include <net/if.h>
65 #include <net/if_dl.h>
66 #include <net/if_types.h>
67 #include <netinet/in.h>
68 #include <netinet/tcp.h>
69 #include <sys/mount.h>
70 #include <sys/proc.h>
71 #include <sys/vnode.h>
72 #include <sys/device.h>
73 #include <sys/protosw.h>
74 #include <sys/mbuf.h>
75 #include <sys/syslog.h>
76 #include <sys/exec.h>
77 #include <sys/kauth.h>
78 
79 #include <sys/syscallargs.h>
80 #include <sys/ktrace.h>
81 
82 #include <lib/libkern/libkern.h>
83 
84 #ifdef INET6
85 #include <netinet/ip6.h>
86 #include <netinet6/ip6_var.h>
87 #endif
88 
89 #include <compat/sys/socket.h>
90 #include <compat/sys/sockio.h>
91 
92 #include <compat/linux/common/linux_types.h>
93 #include <compat/linux/common/linux_util.h>
94 #include <compat/linux/common/linux_signal.h>
95 #include <compat/linux/common/linux_ioctl.h>
96 #include <compat/linux/common/linux_socket.h>
97 #if !defined(__alpha__) && !defined(__amd64__)
98 #include <compat/linux/common/linux_socketcall.h>
99 #endif
100 #include <compat/linux/common/linux_sockio.h>
101 #include <compat/linux/common/linux_ipc.h>
102 #include <compat/linux/common/linux_sem.h>
103 
104 #include <compat/linux/linux_syscallargs.h>
105 
106 #ifdef DEBUG_LINUX
107 #define DPRINTF(a) uprintf a
108 #else
109 #define DPRINTF(a)
110 #endif
111 
112 /*
113  * The calls in this file are entered either via the linux_socketcall()
114  * interface or, on the Alpha, as individual syscalls.  The
115  * linux_socketcall function does any massaging of arguments so that all
116  * the calls in here need not think that they are anything other
117  * than a normal syscall.
118  */
119 
120 static int linux_to_bsd_domain __P((int));
121 static int bsd_to_linux_domain __P((int));
122 int linux_to_bsd_sopt_level __P((int));
123 int linux_to_bsd_so_sockopt __P((int));
124 int linux_to_bsd_ip_sockopt __P((int));
125 int linux_to_bsd_tcp_sockopt __P((int));
126 int linux_to_bsd_udp_sockopt __P((int));
127 int linux_getifhwaddr __P((struct lwp *, register_t *, u_int, void *));
128 static int linux_get_sa(struct lwp *, int, struct mbuf **,
129 		const struct osockaddr *, int);
130 static int linux_sa_put __P((struct osockaddr *osa));
131 static int linux_to_bsd_msg_flags __P((int));
132 static int bsd_to_linux_msg_flags __P((int));
133 
134 static const int linux_to_bsd_domain_[LINUX_AF_MAX] = {
135 	AF_UNSPEC,
136 	AF_UNIX,
137 	AF_INET,
138 	AF_CCITT,	/* LINUX_AF_AX25 */
139 	AF_IPX,
140 	AF_APPLETALK,
141 	-1,		/* LINUX_AF_NETROM */
142 	-1,		/* LINUX_AF_BRIDGE */
143 	-1,		/* LINUX_AF_ATMPVC */
144 	AF_CCITT,	/* LINUX_AF_X25 */
145 	AF_INET6,
146 	-1,		/* LINUX_AF_ROSE */
147 	AF_DECnet,
148 	-1,		/* LINUX_AF_NETBEUI */
149 	-1,		/* LINUX_AF_SECURITY */
150 	pseudo_AF_KEY,
151 	AF_ROUTE,	/* LINUX_AF_NETLINK */
152 	-1,		/* LINUX_AF_PACKET */
153 	-1,		/* LINUX_AF_ASH */
154 	-1,		/* LINUX_AF_ECONET */
155 	-1,		/* LINUX_AF_ATMSVC */
156 	AF_SNA,
157 	/* rest up to LINUX_AF_MAX-1 is not allocated */
158 	-1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
159 };
160 
161 static const int bsd_to_linux_domain_[AF_MAX] = {
162 	LINUX_AF_UNSPEC,
163 	LINUX_AF_UNIX,
164 	LINUX_AF_INET,
165 	-1,		/* AF_IMPLINK */
166 	-1,		/* AF_PUP */
167 	-1,		/* AF_CHAOS */
168 	-1,		/* AF_NS */
169 	-1,		/* AF_ISO */
170 	-1,		/* AF_ECMA */
171 	-1,		/* AF_DATAKIT */
172 	LINUX_AF_AX25,	/* AF_CCITT */
173 	LINUX_AF_SNA,
174 	LINUX_AF_DECnet,
175 	-1,		/* AF_DLI */
176 	-1,		/* AF_LAT */
177 	-1,		/* AF_HYLINK */
178 	LINUX_AF_APPLETALK,
179 	LINUX_AF_NETLINK,
180 	-1,		/* AF_LINK */
181 	-1,		/* AF_XTP */
182 	-1,		/* AF_COIP */
183 	-1,		/* AF_CNT */
184 	-1,		/* pseudo_AF_RTIP */
185 	LINUX_AF_IPX,
186 	LINUX_AF_INET6,
187 	-1,		/* pseudo_AF_PIP */
188 	-1,		/* AF_ISDN */
189 	-1,		/* AF_NATM */
190 	-1,		/* AF_ARP */
191 	LINUX_pseudo_AF_KEY,
192 	-1,		/* pseudo_AF_HDRCMPLT */
193 };
194 
195 static const struct {
196 	int bfl;
197 	int lfl;
198 } bsd_to_linux_msg_flags_[] = {
199 	{MSG_OOB,		LINUX_MSG_OOB},
200 	{MSG_PEEK,		LINUX_MSG_PEEK},
201 	{MSG_DONTROUTE,		LINUX_MSG_DONTROUTE},
202 	{MSG_EOR,		LINUX_MSG_EOR},
203 	{MSG_TRUNC,		LINUX_MSG_TRUNC},
204 	{MSG_CTRUNC,		LINUX_MSG_CTRUNC},
205 	{MSG_WAITALL,		LINUX_MSG_WAITALL},
206 	{MSG_DONTWAIT,		LINUX_MSG_DONTWAIT},
207 	{MSG_BCAST,		0},		/* not supported, clear */
208 	{MSG_MCAST,		0},		/* not supported, clear */
209 	{-1, /* not supp */	LINUX_MSG_PROBE},
210 	{-1, /* not supp */	LINUX_MSG_FIN},
211 	{-1, /* not supp */	LINUX_MSG_SYN},
212 	{-1, /* not supp */	LINUX_MSG_CONFIRM},
213 	{-1, /* not supp */	LINUX_MSG_RST},
214 	{-1, /* not supp */	LINUX_MSG_ERRQUEUE},
215 	{-1, /* not supp */	LINUX_MSG_NOSIGNAL},
216 	{-1, /* not supp */	LINUX_MSG_MORE},
217 };
218 
219 /*
220  * Convert between Linux and BSD socket domain values
221  */
222 static int
223 linux_to_bsd_domain(ldom)
224 	int ldom;
225 {
226 	if (ldom < 0 || ldom >= LINUX_AF_MAX)
227 		return (-1);
228 
229 	return linux_to_bsd_domain_[ldom];
230 }
231 
232 /*
233  * Convert between BSD and Linux socket domain values
234  */
235 static int
236 bsd_to_linux_domain(bdom)
237 	int bdom;
238 {
239 	if (bdom < 0 || bdom >= AF_MAX)
240 		return (-1);
241 
242 	return bsd_to_linux_domain_[bdom];
243 }
244 
245 static int
246 linux_to_bsd_msg_flags(lflag)
247 	int lflag;
248 {
249 	int i, lfl, bfl;
250 	int bflag = 0;
251 
252 	if (lflag == 0)
253 		return (0);
254 
255 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
256 		bfl = bsd_to_linux_msg_flags_[i].bfl;
257 		lfl = bsd_to_linux_msg_flags_[i].lfl;
258 
259 		if (lfl == 0)
260 			continue;
261 
262 		if (lflag & lfl) {
263 			if (bfl < 0)
264 				return (-1);
265 
266 			bflag |= bfl;
267 		}
268 	}
269 
270 	return (bflag);
271 }
272 
273 static int
274 bsd_to_linux_msg_flags(bflag)
275 	int bflag;
276 {
277 	int i, lfl, bfl;
278 	int lflag = 0;
279 
280 	if (bflag == 0)
281 		return (0);
282 
283 	for(i = 0; i < __arraycount(bsd_to_linux_msg_flags_); i++) {
284 		bfl = bsd_to_linux_msg_flags_[i].bfl;
285 		lfl = bsd_to_linux_msg_flags_[i].lfl;
286 
287 		if (bfl <= 0)
288 			continue;
289 
290 		if (bflag & bfl) {
291 			if (lfl < 0)
292 				return (-1);
293 
294 			lflag |= lfl;
295 		}
296 	}
297 
298 	return (lflag);
299 }
300 
301 int
302 linux_sys_socket(l, v, retval)
303 	struct lwp *l;
304 	void *v;
305 	register_t *retval;
306 {
307 	struct linux_sys_socket_args /* {
308 		syscallarg(int)	domain;
309 		syscallarg(int)	type;
310 		syscallarg(int) protocol;
311 	} */ *uap = v;
312 	struct compat_30_sys_socket_args bsa;
313 	int error;
314 
315 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
316 	SCARG(&bsa, type) = SCARG(uap, type);
317 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
318 	if (SCARG(&bsa, domain) == -1)
319 		return EINVAL;
320 	error = sys___socket30(l, &bsa, retval);
321 
322 #ifdef INET6
323 	/*
324 	 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by
325 	 * default and some apps depend on this. So, set V6ONLY to 0
326 	 * for Linux apps if the sysctl value is set to 1.
327 	 */
328 	if (!error && ip6_v6only && SCARG(&bsa, domain) == PF_INET6) {
329 		struct proc *p = l->l_proc;
330 		struct file *fp;
331 
332 		if (getsock(p->p_fd, *retval, &fp) == 0) {
333 			struct mbuf *m;
334 
335 			m = m_get(M_WAIT, MT_SOOPTS);
336 			m->m_len = sizeof(int);
337 			*mtod(m, int *) = 0;
338 
339 			/* ignore error */
340 			(void) sosetopt((struct socket *)fp->f_data,
341 				IPPROTO_IPV6, IPV6_V6ONLY, m);
342 
343 			FILE_UNUSE(fp, l);
344 		}
345 	}
346 #endif
347 
348 	return (error);
349 }
350 
351 int
352 linux_sys_socketpair(l, v, retval)
353 	struct lwp *l;
354 	void *v;
355 	register_t *retval;
356 {
357 	struct linux_sys_socketpair_args /* {
358 		syscallarg(int) domain;
359 		syscallarg(int) type;
360 		syscallarg(int) protocol;
361 		syscallarg(int *) rsv;
362 	} */ *uap = v;
363 	struct sys_socketpair_args bsa;
364 
365 	SCARG(&bsa, domain) = linux_to_bsd_domain(SCARG(uap, domain));
366 	if (SCARG(&bsa, domain) == -1)
367 		return EINVAL;
368 	SCARG(&bsa, type) = SCARG(uap, type);
369 	SCARG(&bsa, protocol) = SCARG(uap, protocol);
370 	SCARG(&bsa, rsv) = SCARG(uap, rsv);
371 
372 	return sys_socketpair(l, &bsa, retval);
373 }
374 
375 int
376 linux_sys_sendto(l, v, retval)
377 	struct lwp *l;
378 	void *v;
379 	register_t *retval;
380 {
381 	struct linux_sys_sendto_args /* {
382 		syscallarg(int)				s;
383 		syscallarg(void *)			msg;
384 		syscallarg(int)				len;
385 		syscallarg(int)				flags;
386 		syscallarg(struct osockaddr *)		to;
387 		syscallarg(int)				tolen;
388 	} */ *uap = v;
389 	struct msghdr   msg;
390 	struct iovec    aiov;
391 	struct mbuf *nam;
392 	int bflags;
393 	int error;
394 
395 	/* Translate message flags.  */
396 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
397 	if (bflags < 0)
398 		/* Some supported flag */
399 		return EINVAL;
400 
401 	/* Read in and convert the sockaddr */
402 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, to),
403 	    SCARG(uap, tolen));
404 	if (error)
405 		return (error);
406 	msg.msg_flags = MSG_NAMEMBUF;
407 
408 	msg.msg_name = nam;
409 	msg.msg_namelen = SCARG(uap, tolen);
410 	msg.msg_iov = &aiov;
411 	msg.msg_iovlen = 1;
412 	msg.msg_control = 0;
413 	aiov.iov_base = __UNCONST(SCARG(uap, msg));
414 	aiov.iov_len = SCARG(uap, len);
415 
416 	return do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
417 }
418 
419 int
420 linux_sys_sendmsg(l, v, retval)
421 	struct lwp *l;
422 	void *v;
423 	register_t *retval;
424 {
425 	struct linux_sys_sendmsg_args /* {
426 		syscallarg(int) s;
427 		syscallarg(struct msghdr *) msg;
428 		syscallarg(u_int) flags;
429 	} */ *uap = v;
430 	struct msghdr	msg;
431 	int		error;
432 	int		bflags;
433 	struct mbuf     *nam;
434 	u_int8_t	*control;
435 	struct mbuf     *ctl_mbuf = NULL;
436 
437 	msg.msg_flags = MSG_IOVUSRSPACE;
438 
439 	/*
440 	 * Translate message flags.
441 	 */
442 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
443 	if (bflags < 0)
444 		/* Some supported flag */
445 		return EINVAL;
446 
447 	if (msg.msg_name) {
448 		/* Read in and convert the sockaddr */
449 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
450 		    msg.msg_namelen);
451 		if (error)
452 			return (error);
453 		msg.msg_flags |= MSG_NAMEMBUF;
454 		msg.msg_name = nam;
455 	}
456 
457 	/*
458 	 * Handle cmsg if there is any.
459 	 */
460 	if (CMSG_FIRSTHDR(&msg)) {
461 		struct linux_cmsghdr l_cmsg, *l_cc;
462 		struct cmsghdr *cmsg;
463 		ssize_t resid = msg.msg_controllen;
464 		size_t clen, cidx = 0, cspace;
465 
466 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
467 		clen = MLEN;
468 		control = mtod(ctl_mbuf, void *);
469 
470 		l_cc = LINUX_CMSG_FIRSTHDR(&msg);
471 		do {
472 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
473 			if (error)
474 				goto done;
475 
476 			/*
477 			 * Sanity check the control message length.
478 			 */
479 			if (l_cmsg.cmsg_len > resid
480 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
481 				error = EINVAL;
482 				goto done;
483 			}
484 
485 			/*
486 			 * Refuse unsupported control messages, and
487 			 * translate fields as appropriate.
488 			 */
489 			switch (l_cmsg.cmsg_level) {
490 			case LINUX_SOL_SOCKET:
491 				/* It only differs on some archs */
492 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
493 					l_cmsg.cmsg_level = SOL_SOCKET;
494 
495 				switch(l_cmsg.cmsg_type) {
496 				case LINUX_SCM_RIGHTS:
497 					/* Linux SCM_RIGHTS is same as NetBSD */
498 					break;
499 
500 				default:
501 					/* other types not supported */
502 					error = EINVAL;
503 					goto done;
504 				}
505 				break;
506 			default:
507 				/* pray and leave intact */
508 				break;
509 			}
510 
511 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
512 
513 			/* Check the buffer is big enough */
514 			if (__predict_false(cidx + cspace > clen)) {
515 				u_int8_t *nc;
516 
517 				clen = cidx + cspace;
518 				if (clen >= PAGE_SIZE) {
519 					error = EINVAL;
520 					goto done;
521 				}
522 				nc = realloc(clen <= MLEN ? NULL : control,
523 						clen, M_TEMP, M_WAITOK);
524 				if (!nc) {
525 					error = ENOMEM;
526 					goto done;
527 				}
528 				if (cidx <= MLEN)
529 					/* Old buffer was in mbuf... */
530 					memcpy(nc, control, cidx);
531 				control = nc;
532 			}
533 
534 			/* Copy header */
535 			cmsg = (void *)&control[cidx];
536 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
537 			cmsg->cmsg_level = l_cmsg.cmsg_level;
538 			cmsg->cmsg_type = l_cmsg.cmsg_type;
539 
540 			/* Zero are between header and data */
541 			memset(cmsg + 1, 0,
542 				CMSG_ALIGN(sizeof(cmsg)) - sizeof(cmsg));
543 
544 			/* Copyin the data */
545 			error = copyin(LINUX_CMSG_DATA(l_cc),
546 				CMSG_DATA(control),
547 				l_cmsg.cmsg_len - sizeof(l_cmsg));
548 			if (error)
549 				goto done;
550 
551 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
552 			cidx += cspace;
553 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc)) && resid > 0);
554 
555 		/* If we allocated a buffer, attach to mbuf */
556 		if (cidx > MLEN) {
557 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
558 			ctl_mbuf->m_flags |= M_EXT_RW;
559 		}
560 		control = NULL;
561 		ctl_mbuf->m_len = cidx;
562 
563 		msg.msg_control = ctl_mbuf;
564 		msg.msg_flags |= MSG_CONTROLMBUF;
565 	}
566 
567 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
568 	/* Freed internally */
569 	ctl_mbuf = NULL;
570 
571 done:
572 	if (ctl_mbuf != NULL) {
573 		if (control != NULL && control != mtod(ctl_mbuf, void *))
574 			free(control, M_MBUF);
575 		m_free(ctl_mbuf);
576 	}
577 	return (error);
578 }
579 
580 int
581 linux_sys_recvfrom(l, v, retval)
582 	struct lwp *l;
583 	void *v;
584 	register_t *retval;
585 {
586 	struct linux_sys_recvfrom_args /* {
587 		syscallarg(int) s;
588 		syscallarg(void *) buf;
589 		syscallarg(int) len;
590 		syscallarg(int) flags;
591 		syscallarg(struct osockaddr *) from;
592 		syscallarg(int *) fromlenaddr;
593 	} */ *uap = v;
594 	int		error;
595 	struct sys_recvfrom_args bra;
596 
597 	SCARG(&bra, s) = SCARG(uap, s);
598 	SCARG(&bra, buf) = SCARG(uap, buf);
599 	SCARG(&bra, len) = SCARG(uap, len);
600 	SCARG(&bra, flags) = SCARG(uap, flags);
601 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
602 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
603 
604 	if ((error = sys_recvfrom(l, &bra, retval)))
605 		return (error);
606 
607 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
608 		return (error);
609 
610 	return (0);
611 }
612 
613 static int
614 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
615 {
616 	int dlen, error = 0;
617 	struct cmsghdr *cmsg;
618 	struct linux_cmsghdr linux_cmsg;
619 	struct mbuf *m;
620 	char *q, *q_end;
621 
622 	if (mp->msg_controllen <= 0 || control == 0) {
623 		mp->msg_controllen = 0;
624 		free_control_mbuf(l, control, control);
625 		return 0;
626 	}
627 
628 	q = (char *)mp->msg_control;
629 	q_end = q + mp->msg_controllen;
630 
631 	for (m = control; m != NULL; ) {
632 		cmsg = mtod(m, struct cmsghdr *);
633 
634 		/*
635 		 * Fixup cmsg. We handle two things:
636 		 * 0. different sizeof cmsg_len.
637 		 * 1. different values for level/type on some archs
638 		 * 2. different alignment of CMSG_DATA on some archs
639 		 */
640 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
641 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
642 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
643 
644 		dlen = q_end - q;
645 		if (linux_cmsg.cmsg_len > dlen) {
646 			/* Not enough room for the parameter */
647 			dlen -= sizeof linux_cmsg;
648 			if (dlen <= 0)
649 				/* Discard if header wont fit */
650 				break;
651 			mp->msg_flags |= MSG_CTRUNC;
652 			if (linux_cmsg.cmsg_level == SOL_SOCKET
653 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
654 				/* Do not truncate me ... */
655 				break;
656 		} else
657 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
658 
659 		switch (linux_cmsg.cmsg_level) {
660 		case SOL_SOCKET:
661 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
662 			switch (linux_cmsg.cmsg_type) {
663 			case SCM_RIGHTS:
664 				/* Linux SCM_RIGHTS is same as NetBSD */
665 				break;
666 
667 			default:
668 				/* other types not supported */
669 				error = EINVAL;
670 				goto done;
671 			}
672 			/* machine dependant ! */
673 			break;
674 		default:
675 			/* pray and leave intact */
676 			break;
677 		}
678 
679 		/* There can be padding between the header and data... */
680 		error = copyout(&linux_cmsg, q, sizeof *cmsg);
681 		if (error != 0) {
682 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
683 			    dlen);
684 		}
685 		if (error != 0) {
686 			/* We must free all the SCM_RIGHTS */
687 			m = control;
688 			break;
689 		}
690 		m = m->m_next;
691 		if (m == NULL || q + LINUX_CMSG_ALIGN(dlen) > q_end) {
692 			q += dlen;
693 			break;
694 		}
695 		q += LINUX_CMSG_ALIGN(dlen);
696 	}
697 
698   done:
699 	free_control_mbuf(l, control, m);
700 
701 	mp->msg_controllen = q - (char *)mp->msg_control;
702 	return error;
703 }
704 
705 int
706 linux_sys_recvmsg(l, v, retval)
707 	struct lwp *l;
708 	void *v;
709 	register_t *retval;
710 {
711 	struct linux_sys_recvmsg_args /* {
712 		syscallarg(int) s;
713 		syscallarg(struct msghdr *) msg;
714 		syscallarg(u_int) flags;
715 	} */ *uap = v;
716 	struct msghdr	msg;
717 	int		error;
718 	struct mbuf	*from, *control;
719 
720 	error = copyin(SCARG(uap, msg), &msg, sizeof(msg));
721 	if (error)
722 		return (error);
723 
724 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
725 	if (msg.msg_flags < 0) {
726 		/* Some unsupported flag */
727 		return (EINVAL);
728 	}
729 	msg.msg_flags |= MSG_IOVUSRSPACE;
730 
731 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
732 	    msg.msg_control != NULL ? &control : NULL, retval);
733 	if (error != 0)
734 		return error;
735 
736 	if (msg.msg_control != NULL)
737 		error = linux_copyout_msg_control(l, &msg, control);
738 
739 	if (error == 0 && from != 0) {
740 		mtod(from, struct osockaddr *)->sa_family =
741 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
742 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
743 			from);
744 	} else
745 		msg.msg_namelen = 0;
746 
747 	if (from != NULL)
748 		m_free(from);
749 
750 	if (error == 0) {
751 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
752 		if (msg.msg_flags < 0)
753 			/* Some flag unsupported by Linux */
754 			error = EINVAL;
755 		else
756 			error = copyout(&msg, SCARG(uap, msg), sizeof(msg));
757 	}
758 
759 	return (error);
760 }
761 
762 /*
763  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
764  * is different, the rest matches IPPROTO_* on both systems.
765  */
766 int
767 linux_to_bsd_sopt_level(llevel)
768 	int llevel;
769 {
770 
771 	switch (llevel) {
772 	case LINUX_SOL_SOCKET:
773 		return SOL_SOCKET;
774 	case LINUX_SOL_IP:
775 		return IPPROTO_IP;
776 	case LINUX_SOL_TCP:
777 		return IPPROTO_TCP;
778 	case LINUX_SOL_UDP:
779 		return IPPROTO_UDP;
780 	default:
781 		return -1;
782 	}
783 }
784 
785 /*
786  * Convert Linux socket level socket option numbers to NetBSD values.
787  */
788 int
789 linux_to_bsd_so_sockopt(lopt)
790 	int lopt;
791 {
792 
793 	switch (lopt) {
794 	case LINUX_SO_DEBUG:
795 		return SO_DEBUG;
796 	case LINUX_SO_REUSEADDR:
797 		/*
798 		 * Linux does not implement SO_REUSEPORT, but allows reuse of a
799 		 * host:port pair through SO_REUSEADDR even if the address is not a
800 		 * multicast-address.  Effectively, this means that we should use
801 		 * SO_REUSEPORT to allow Linux applications to not exit with
802 		 * EADDRINUSE
803 		 */
804 		return SO_REUSEPORT;
805 	case LINUX_SO_TYPE:
806 		return SO_TYPE;
807 	case LINUX_SO_ERROR:
808 		return SO_ERROR;
809 	case LINUX_SO_DONTROUTE:
810 		return SO_DONTROUTE;
811 	case LINUX_SO_BROADCAST:
812 		return SO_BROADCAST;
813 	case LINUX_SO_SNDBUF:
814 		return SO_SNDBUF;
815 	case LINUX_SO_RCVBUF:
816 		return SO_RCVBUF;
817 	case LINUX_SO_KEEPALIVE:
818 		return SO_KEEPALIVE;
819 	case LINUX_SO_OOBINLINE:
820 		return SO_OOBINLINE;
821 	case LINUX_SO_LINGER:
822 		return SO_LINGER;
823 	case LINUX_SO_PRIORITY:
824 	case LINUX_SO_NO_CHECK:
825 	default:
826 		return -1;
827 	}
828 }
829 
830 /*
831  * Convert Linux IP level socket option number to NetBSD values.
832  */
833 int
834 linux_to_bsd_ip_sockopt(lopt)
835 	int lopt;
836 {
837 
838 	switch (lopt) {
839 	case LINUX_IP_TOS:
840 		return IP_TOS;
841 	case LINUX_IP_TTL:
842 		return IP_TTL;
843 	case LINUX_IP_MULTICAST_TTL:
844 		return IP_MULTICAST_TTL;
845 	case LINUX_IP_MULTICAST_LOOP:
846 		return IP_MULTICAST_LOOP;
847 	case LINUX_IP_MULTICAST_IF:
848 		return IP_MULTICAST_IF;
849 	case LINUX_IP_ADD_MEMBERSHIP:
850 		return IP_ADD_MEMBERSHIP;
851 	case LINUX_IP_DROP_MEMBERSHIP:
852 		return IP_DROP_MEMBERSHIP;
853 	default:
854 		return -1;
855 	}
856 }
857 
858 /*
859  * Convert Linux TCP level socket option number to NetBSD values.
860  */
861 int
862 linux_to_bsd_tcp_sockopt(lopt)
863 	int lopt;
864 {
865 
866 	switch (lopt) {
867 	case LINUX_TCP_NODELAY:
868 		return TCP_NODELAY;
869 	case LINUX_TCP_MAXSEG:
870 		return TCP_MAXSEG;
871 	default:
872 		return -1;
873 	}
874 }
875 
876 /*
877  * Convert Linux UDP level socket option number to NetBSD values.
878  */
879 int
880 linux_to_bsd_udp_sockopt(lopt)
881 	int lopt;
882 {
883 
884 	switch (lopt) {
885 	default:
886 		return -1;
887 	}
888 }
889 
890 /*
891  * Another reasonably straightforward function: setsockopt(2).
892  * The level and option numbers are converted; the values passed
893  * are not (yet) converted, the ones currently implemented don't
894  * need conversion, as they are the same on both systems.
895  */
896 int
897 linux_sys_setsockopt(l, v, retval)
898 	struct lwp *l;
899 	void *v;
900 	register_t *retval;
901 {
902 	struct linux_sys_setsockopt_args /* {
903 		syscallarg(int) s;
904 		syscallarg(int) level;
905 		syscallarg(int) optname;
906 		syscallarg(void *) optval;
907 		syscallarg(int) optlen;
908 	} */ *uap = v;
909 	struct proc *p = l->l_proc;
910 	struct sys_setsockopt_args bsa;
911 	int name;
912 
913 	SCARG(&bsa, s) = SCARG(uap, s);
914 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
915 	SCARG(&bsa, val) = SCARG(uap, optval);
916 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
917 
918 	/*
919 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
920 	 * and returns EOPNOTSUPP for other levels
921 	 */
922 	if (SCARG(&bsa, level) != SOL_SOCKET) {
923 		struct file *fp;
924 		struct socket *so;
925 		int error, s, family;
926 
927 		/* getsock() will use the descriptor for us */
928 	    	if ((error = getsock(p->p_fd, SCARG(&bsa, s), &fp)) != 0)
929 		    	return error;
930 
931 		s = splsoftnet();
932 		so = (struct socket *)fp->f_data;
933 		family = so->so_proto->pr_domain->dom_family;
934 		splx(s);
935 		FILE_UNUSE(fp, l);
936 
937 		if (family == AF_LOCAL)
938 			return EOPNOTSUPP;
939 	}
940 
941 	switch (SCARG(&bsa, level)) {
942 	case SOL_SOCKET:
943 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
944 		break;
945 	case IPPROTO_IP:
946 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
947 		break;
948 	case IPPROTO_TCP:
949 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
950 		break;
951 	case IPPROTO_UDP:
952 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
953 		break;
954 	default:
955 		return EINVAL;
956 	}
957 
958 	if (name == -1)
959 		return EINVAL;
960 	SCARG(&bsa, name) = name;
961 
962 	return sys_setsockopt(l, &bsa, retval);
963 }
964 
965 /*
966  * getsockopt(2) is very much the same as setsockopt(2) (see above)
967  */
968 int
969 linux_sys_getsockopt(l, v, retval)
970 	struct lwp *l;
971 	void *v;
972 	register_t *retval;
973 {
974 	struct linux_sys_getsockopt_args /* {
975 		syscallarg(int) s;
976 		syscallarg(int) level;
977 		syscallarg(int) optname;
978 		syscallarg(void *) optval;
979 		syscallarg(int *) optlen;
980 	} */ *uap = v;
981 	struct sys_getsockopt_args bga;
982 	int name;
983 
984 	SCARG(&bga, s) = SCARG(uap, s);
985 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
986 	SCARG(&bga, val) = SCARG(uap, optval);
987 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
988 
989 	switch (SCARG(&bga, level)) {
990 	case SOL_SOCKET:
991 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
992 		break;
993 	case IPPROTO_IP:
994 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
995 		break;
996 	case IPPROTO_TCP:
997 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
998 		break;
999 	case IPPROTO_UDP:
1000 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1001 		break;
1002 	default:
1003 		return EINVAL;
1004 	}
1005 
1006 	if (name == -1)
1007 		return EINVAL;
1008 	SCARG(&bga, name) = name;
1009 
1010 	return sys_getsockopt(l, &bga, retval);
1011 }
1012 
1013 #define IF_NAME_LEN 16
1014 
1015 int
1016 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1017     void *data)
1018 {
1019 	/* Not the full structure, just enough to map what we do here */
1020 	struct linux_ifreq {
1021 		char if_name[IF_NAME_LEN];
1022 		struct osockaddr hwaddr;
1023 	} lreq;
1024 	struct proc *p = l->l_proc;
1025 	struct filedesc *fdp;
1026 	struct file *fp;
1027 	struct ifaddr *ifa;
1028 	struct ifnet *ifp;
1029 	struct sockaddr_dl *sadl;
1030 	int error, found;
1031 	int index, ifnum;
1032 
1033 	/*
1034 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1035 	 * SIOCGIFCONF, because the user buffer is not of the right
1036 	 * type to take those results.  We can't use kernel buffers to
1037 	 * receive the results, as the implementation of sys_ioctl()
1038 	 * and ifconf() [which implements SIOCGIFCONF] use
1039 	 * copyin()/copyout() which will fail on kernel addresses.
1040 	 *
1041 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1042 	 */
1043 
1044 	fdp = p->p_fd;
1045 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1046 		return (EBADF);
1047 
1048 	FILE_USE(fp);
1049 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1050 		error = EBADF;
1051 		goto out;
1052 	}
1053 
1054 	error = copyin(data, &lreq, sizeof(lreq));
1055 	if (error)
1056 		goto out;
1057 	lreq.if_name[IF_NAME_LEN-1] = '\0';		/* just in case */
1058 
1059 	/*
1060 	 * Try real interface name first, then fake "ethX"
1061 	 */
1062 	for (ifp = ifnet.tqh_first, found = 0;
1063 	     ifp != 0 && !found;
1064 	     ifp = ifp->if_list.tqe_next) {
1065 		if (strcmp(lreq.if_name, ifp->if_xname))
1066 			/* not this interface */
1067 			continue;
1068 		found=1;
1069 		if ((ifa = ifp->if_addrlist.tqh_first) != 0) {
1070 			for (; ifa != 0; ifa = ifa->ifa_list.tqe_next) {
1071 				sadl = satosdl(ifa->ifa_addr);
1072 				/* only return ethernet addresses */
1073 				/* XXX what about FDDI, etc. ? */
1074 				if (sadl->sdl_family != AF_LINK ||
1075 				    sadl->sdl_type != IFT_ETHER)
1076 					continue;
1077 				memcpy(&lreq.hwaddr.sa_data, CLLADDR(sadl),
1078 				       MIN(sadl->sdl_alen,
1079 					   sizeof(lreq.hwaddr.sa_data)));
1080 				lreq.hwaddr.sa_family =
1081 					sadl->sdl_family;
1082 				error = copyout(&lreq, data, sizeof(lreq));
1083 				goto out;
1084 			}
1085 		} else {
1086 			error = ENODEV;
1087 			goto out;
1088 		}
1089 	}
1090 
1091 	if (strncmp(lreq.if_name, "eth", 3) == 0) {
1092 		for (ifnum = 0, index = 3;
1093 		     lreq.if_name[index] != '\0' && index < IF_NAME_LEN;
1094 		     index++) {
1095 			ifnum *= 10;
1096 			ifnum += lreq.if_name[index] - '0';
1097 		}
1098 
1099 		error = EINVAL;			/* in case we don't find one */
1100 		for (ifp = ifnet.tqh_first, found = 0;
1101 		     ifp != 0 && !found;
1102 		     ifp = ifp->if_list.tqe_next) {
1103 			memcpy(lreq.if_name, ifp->if_xname,
1104 			       MIN(IF_NAME_LEN, IFNAMSIZ));
1105 			if ((ifa = ifp->if_addrlist.tqh_first) == 0)
1106 				/* no addresses on this interface */
1107 				continue;
1108 			else
1109 				for (; ifa != 0; ifa = ifa->ifa_list.tqe_next) {
1110 					sadl = satosdl(ifa->ifa_addr);
1111 					/* only return ethernet addresses */
1112 					/* XXX what about FDDI, etc. ? */
1113 					if (sadl->sdl_family != AF_LINK ||
1114 					    sadl->sdl_type != IFT_ETHER)
1115 						continue;
1116 					if (ifnum--)
1117 						/* not the reqested iface */
1118 						continue;
1119 					memcpy(&lreq.hwaddr.sa_data,
1120 					       CLLADDR(sadl),
1121 					       MIN(sadl->sdl_alen,
1122 						   sizeof(lreq.hwaddr.sa_data)));
1123 					lreq.hwaddr.sa_family =
1124 						sadl->sdl_family;
1125 					error = copyout(&lreq, data, sizeof(lreq));
1126 					found = 1;
1127 					break;
1128 				}
1129 		}
1130 	} else {
1131 		/* unknown interface, not even an "eth*" name */
1132 		error = ENODEV;
1133 	}
1134 
1135 out:
1136 	FILE_UNUSE(fp, l);
1137 	return error;
1138 }
1139 #undef IF_NAME_LEN
1140 
1141 int
1142 linux_ioctl_socket(l, uap, retval)
1143 	struct lwp *l;
1144 	struct linux_sys_ioctl_args /* {
1145 		syscallarg(int) fd;
1146 		syscallarg(u_long) com;
1147 		syscallarg(void *) data;
1148 	} */ *uap;
1149 	register_t *retval;
1150 {
1151 	struct proc *p = l->l_proc;
1152 	u_long com;
1153 	int error = 0, isdev = 0, dosys = 1;
1154 	struct sys_ioctl_args ia;
1155 	struct file *fp;
1156 	struct filedesc *fdp;
1157 	struct vnode *vp;
1158 	int (*ioctlf)(struct file *, u_long, void *, struct lwp *);
1159 	struct ioctl_pt pt;
1160 
1161 	fdp = p->p_fd;
1162 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1163 		return (EBADF);
1164 
1165 	FILE_USE(fp);
1166 
1167 	if (fp->f_type == DTYPE_VNODE) {
1168 		vp = (struct vnode *)fp->f_data;
1169 		isdev = vp->v_type == VCHR;
1170 	}
1171 
1172 	/*
1173 	 * Don't try to interpret socket ioctl calls that are done
1174 	 * on a device filedescriptor, just pass them through, to
1175 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1176 	 * device will only handle these if it's prepared to do
1177 	 * so, to avoid unexpected things from happening.
1178 	 */
1179 	if (isdev) {
1180 		dosys = 0;
1181 		ioctlf = fp->f_ops->fo_ioctl;
1182 		pt.com = SCARG(uap, com);
1183 		pt.data = SCARG(uap, data);
1184 		error = ioctlf(fp, PTIOCLINUX, (void *)&pt, l);
1185 		/*
1186 		 * XXX hack: if the function returns EJUSTRETURN,
1187 		 * it has stuffed a sysctl return value in pt.data.
1188 		 */
1189 		if (error == EJUSTRETURN) {
1190 			retval[0] = (register_t)pt.data;
1191 			error = 0;
1192 		}
1193 		goto out;
1194 	}
1195 
1196 	com = SCARG(uap, com);
1197 	retval[0] = 0;
1198 
1199 	switch (com) {
1200 	case LINUX_SIOCGIFCONF:
1201 		SCARG(&ia, com) = OOSIOCGIFCONF;
1202 		break;
1203 	case LINUX_SIOCGIFFLAGS:
1204 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1205 		break;
1206 	case LINUX_SIOCSIFFLAGS:
1207 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1208 		break;
1209 	case LINUX_SIOCGIFADDR:
1210 		SCARG(&ia, com) = OOSIOCGIFADDR;
1211 		break;
1212 	case LINUX_SIOCGIFDSTADDR:
1213 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1214 		break;
1215 	case LINUX_SIOCGIFBRDADDR:
1216 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1217 		break;
1218 	case LINUX_SIOCGIFNETMASK:
1219 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1220 		break;
1221 	case LINUX_SIOCADDMULTI:
1222 		SCARG(&ia, com) = OSIOCADDMULTI;
1223 		break;
1224 	case LINUX_SIOCDELMULTI:
1225 		SCARG(&ia, com) = OSIOCDELMULTI;
1226 		break;
1227 	case LINUX_SIOCGIFHWADDR:
1228 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1229 					 SCARG(uap, data));
1230 		dosys = 0;
1231 		break;
1232 	default:
1233 		error = EINVAL;
1234 	}
1235 
1236 out:
1237 	FILE_UNUSE(fp, l);
1238 
1239 	if (error ==0 && dosys) {
1240 		SCARG(&ia, fd) = SCARG(uap, fd);
1241 		SCARG(&ia, data) = SCARG(uap, data);
1242 		/* XXX NJWLWP */
1243 		error = sys_ioctl(curlwp, &ia, retval);
1244 	}
1245 
1246 	return error;
1247 }
1248 
1249 int
1250 linux_sys_connect(l, v, retval)
1251 	struct lwp *l;
1252 	void *v;
1253 	register_t *retval;
1254 {
1255 	struct linux_sys_connect_args /* {
1256 		syscallarg(int) s;
1257 		syscallarg(const struct sockaddr *) name;
1258 		syscallarg(int) namelen;
1259 	} */ *uap = v;
1260 	int		error;
1261 	struct mbuf *nam;
1262 
1263 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1264 	    SCARG(uap, namelen));
1265 	if (error)
1266 		return (error);
1267 
1268 	error = do_sys_connect(l, SCARG(uap, s), nam);
1269 
1270 	if (error == EISCONN) {
1271 		struct file *fp;
1272 		struct socket *so;
1273 		int s, state, prflags;
1274 
1275 		/* getsock() will use the descriptor for us */
1276 	    	if (getsock(l->l_proc->p_fd, SCARG(uap, s), &fp) != 0)
1277 		    	return EISCONN;
1278 
1279 		s = splsoftnet();
1280 		so = (struct socket *)fp->f_data;
1281 		state = so->so_state;
1282 		prflags = so->so_proto->pr_flags;
1283 		splx(s);
1284 		FILE_UNUSE(fp, l);
1285 		/*
1286 		 * We should only let this call succeed once per
1287 		 * non-blocking connect; however we don't have
1288 		 * a convenient place to keep that state..
1289 		 */
1290 		if ((state & SS_NBIO) && (state & SS_ISCONNECTED) &&
1291 		    (prflags & PR_CONNREQUIRED))
1292 			return 0;
1293 	}
1294 
1295 	return (error);
1296 }
1297 
1298 int
1299 linux_sys_bind(l, v, retval)
1300 	struct lwp *l;
1301 	void *v;
1302 	register_t *retval;
1303 {
1304 	struct linux_sys_bind_args /* {
1305 		syscallarg(int) s;
1306 		syscallarg(const struct osockaddr *) name;
1307 		syscallarg(int) namelen;
1308 	} */ *uap = v;
1309 	int		error;
1310 	struct mbuf     *nam;
1311 
1312 	error = linux_get_sa(l, SCARG(uap, s), &nam, SCARG(uap, name),
1313 	    SCARG(uap, namelen));
1314 	if (error)
1315 		return (error);
1316 
1317 	return do_sys_bind(l, SCARG(uap, s), nam);
1318 }
1319 
1320 int
1321 linux_sys_getsockname(l, v, retval)
1322 	struct lwp *l;
1323 	void *v;
1324 	register_t *retval;
1325 {
1326 	struct linux_sys_getsockname_args /* {
1327 		syscallarg(int) fdes;
1328 		syscallarg(void *) asa;
1329 		syscallarg(int *) alen;
1330 	} */ *uap = v;
1331 	int error;
1332 
1333 	if ((error = sys_getsockname(l, uap, retval)) != 0)
1334 		return (error);
1335 
1336 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1337 		return (error);
1338 
1339 	return (0);
1340 }
1341 
1342 int
1343 linux_sys_getpeername(l, v, retval)
1344 	struct lwp *l;
1345 	void *v;
1346 	register_t *retval;
1347 {
1348 	struct sys_getpeername_args /* {
1349 		syscallarg(int) fdes;
1350 		syscallarg(void *) asa;
1351 		syscallarg(int *) alen;
1352 	} */ *uap = v;
1353 	int error;
1354 
1355 	if ((error = sys_getpeername(l, uap, retval)) != 0)
1356 		return (error);
1357 
1358 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1359 		return (error);
1360 
1361 	return (0);
1362 }
1363 
1364 /*
1365  * Copy the osockaddr structure pointed to by osa to mbuf, adjust
1366  * family and convert to sockaddr.
1367  */
1368 static int
1369 linux_get_sa(struct lwp *l, int s, struct mbuf **mp, const struct osockaddr *osa, int salen)
1370 {
1371 	int error, bdom;
1372 	struct sockaddr *sa;
1373 	struct osockaddr *kosa;
1374 	struct mbuf *m;
1375 
1376 	if (salen == 1 || salen > UCHAR_MAX) {
1377 		DPRINTF(("bad osa=%p salen=%d\n", osa, salen));
1378 		return EINVAL;
1379 	}
1380 
1381 	/* We'll need the address in an mbuf later, so copy into one here */
1382 	m = m_get(M_WAIT, MT_SONAME);
1383 	if (salen > MLEN)
1384 		MEXTMALLOC(m, salen, M_WAITOK);
1385 
1386 	m->m_len = salen;
1387 
1388 	if (salen == 0)
1389 		return 0;
1390 
1391 	kosa = mtod(m, void *);
1392 	if ((error = copyin(osa, kosa, salen))) {
1393 		DPRINTF(("error %d copying osa %p len %d\n",
1394 				error, osa, salen));
1395 		goto bad;
1396 	}
1397 
1398 	ktrkuser("linux sockaddr", kosa, salen);
1399 
1400 	bdom = linux_to_bsd_domain(kosa->sa_family);
1401 	if (bdom == -1) {
1402 		DPRINTF(("bad linux family=%d\n", kosa->sa_family));
1403 		error = EINVAL;
1404 		goto bad;
1405 	}
1406 
1407 	/*
1408 	 * If the family is unspecified, use address family of the socket.
1409 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1410 	 */
1411 	if (bdom == AF_UNSPEC) {
1412 		struct file *fp;
1413 		struct socket *so;
1414 
1415 		/* getsock() will use the descriptor for us */
1416 		if ((error = getsock(l->l_proc->p_fd, s, &fp)) != 0)
1417 			goto bad;
1418 
1419 		so = (struct socket *)fp->f_data;
1420 		bdom = so->so_proto->pr_domain->dom_family;
1421 
1422 		FILE_UNUSE(fp, l);
1423 
1424 		DPRINTF(("AF_UNSPEC family adjusted to %d\n", bdom));
1425 	}
1426 
1427 #ifdef INET6
1428 	/*
1429 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1430 	 * which lacks the scope id compared with RFC2553 one. If we detect
1431 	 * the situation, reject the address and write a message to system log.
1432 	 *
1433 	 * Still accept addresses for which the scope id is not used.
1434 	 */
1435 	if (bdom == AF_INET6 && salen == sizeof (struct sockaddr_in6) - sizeof (u_int32_t)) {
1436 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)kosa;
1437 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1438 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1439 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1440 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1441 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1442 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1443 			struct proc *p = l->l_proc;
1444 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1445 
1446 			log(LOG_DEBUG,
1447 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1448 			    "sockaddr_in6 rejected",
1449 			    p->p_pid, p->p_comm, uid);
1450 			error = EINVAL;
1451 			goto bad;
1452 		}
1453 		salen = sizeof (struct sockaddr_in6);
1454 		sin6->sin6_scope_id = 0;
1455 	}
1456 #endif
1457 
1458 	if (bdom == AF_INET)
1459 		salen = sizeof(struct sockaddr_in);
1460 
1461 	sa = (struct sockaddr *) kosa;
1462 	sa->sa_family = bdom;
1463 	sa->sa_len = salen;
1464 	m->m_len = salen;
1465 	ktrkuser("new sockaddr", kosa, salen);
1466 
1467 #ifdef DEBUG_LINUX
1468 	DPRINTF(("family %d, len = %d [ ", sa->sa_family, sa->sa_len));
1469 	for (bdom = 0; bdom < sizeof(sa->sa_data); bdom++)
1470 	    DPRINTF(("%02x ", (unsigned char) sa->sa_data[bdom]));
1471 	DPRINTF(("\n"));
1472 #endif
1473 
1474 	*mp = m;
1475 	return 0;
1476 
1477     bad:
1478 	m_free(m);
1479 	return error;
1480 }
1481 
1482 static int
1483 linux_sa_put(osa)
1484 	struct osockaddr *osa;
1485 {
1486 	struct sockaddr sa;
1487 	struct osockaddr *kosa;
1488 	int error, bdom, len;
1489 
1490 	/*
1491 	 * Only read/write the sockaddr family and length part, the rest is
1492 	 * not changed.
1493 	 */
1494 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1495 
1496 	error = copyin(osa, &sa, len);
1497 	if (error)
1498 		return (error);
1499 
1500 	bdom = bsd_to_linux_domain(sa.sa_family);
1501 	if (bdom == -1)
1502 		return (EINVAL);
1503 
1504 	/* Note: we convert from sockaddr to osockaddr here, too */
1505 	kosa = (struct osockaddr *) &sa;
1506 	kosa->sa_family = bdom;
1507 	error = copyout(kosa, osa, len);
1508 	if (error)
1509 		return (error);
1510 
1511 	return (0);
1512 }
1513 
1514 #ifndef __amd64__
1515 int
1516 linux_sys_recv(l, v, retval)
1517 	struct lwp *l;
1518 	void *v;
1519 	register_t *retval;
1520 {
1521 	struct linux_sys_recv_args /* {
1522 		syscallarg(int) s;
1523 		syscallarg(void *) buf;
1524 		syscallarg(int) len;
1525 		syscallarg(int) flags;
1526 	} */ *uap = v;
1527 	struct sys_recvfrom_args bra;
1528 
1529 
1530 	SCARG(&bra, s) = SCARG(uap, s);
1531 	SCARG(&bra, buf) = SCARG(uap, buf);
1532 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1533 	SCARG(&bra, flags) = SCARG(uap, flags);
1534 	SCARG(&bra, from) = NULL;
1535 	SCARG(&bra, fromlenaddr) = NULL;
1536 
1537 	return (sys_recvfrom(l, &bra, retval));
1538 }
1539 
1540 int
1541 linux_sys_send(l, v, retval)
1542 	struct lwp *l;
1543 	void *v;
1544 	register_t *retval;
1545 {
1546 	struct linux_sys_send_args /* {
1547 		syscallarg(int) s;
1548 		syscallarg(void *) buf;
1549 		syscallarg(int) len;
1550 		syscallarg(int) flags;
1551 	} */ *uap = v;
1552 	struct sys_sendto_args bsa;
1553 
1554 	SCARG(&bsa, s)		= SCARG(uap, s);
1555 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1556 	SCARG(&bsa, len)	= SCARG(uap, len);
1557 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1558 	SCARG(&bsa, to)		= NULL;
1559 	SCARG(&bsa, tolen)	= 0;
1560 
1561 	return (sys_sendto(l, &bsa, retval));
1562 }
1563 #endif
1564 
1565 int
1566 linux_sys_accept(l, v, retval)
1567 	struct lwp *l;
1568 	void *v;
1569 	register_t *retval;
1570 {
1571 	struct linux_sys_accept_args /* {
1572 		syscallarg(int) s;
1573 		syscallarg(struct osockaddr *) name;
1574 		syscallarg(int *) anamelen;
1575 	} */ *uap = v;
1576 	int error;
1577 	struct sys_accept_args baa;
1578 
1579 	SCARG(&baa, s)		= SCARG(uap, s);
1580 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1581 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1582 
1583 	if ((error = sys_accept(l, &baa, retval)))
1584 		return (error);
1585 
1586 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1587 		return (error);
1588 
1589 	return (0);
1590 }
1591