xref: /netbsd/sys/compat/linux/common/linux_socket.c (revision deeec4a8)
1 /*	$NetBSD: linux_socket.c,v 1.155 2022/12/24 15:23:02 andvar 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 1.155 2022/12/24 15:23:02 andvar 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(__aarch64__) && !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
linux_to_bsd_domain(int ldom)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
bsd_to_linux_domain(int bdom)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
linux_to_bsd_type(int ltype)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
linux_to_bsd_msg_flags(int lflag)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
bsd_to_linux_msg_flags(int bflag)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
linux_sys_socket(struct lwp * l,const struct linux_sys_socket_args * uap,register_t * retval)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
linux_sys_socketpair(struct lwp * l,const struct linux_sys_socketpair_args * uap,register_t * retval)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
linux_sys_sendto(struct lwp * l,const struct linux_sys_sendto_args * uap,register_t * retval)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
linux_to_bsd_msghdr(const struct linux_msghdr * lmsg,struct msghdr * bmsg)441 linux_to_bsd_msghdr(const struct linux_msghdr *lmsg, struct msghdr *bmsg)
442 {
443 	memset(bmsg, 0, sizeof(*bmsg));
444 	bmsg->msg_name = lmsg->msg_name;
445 	bmsg->msg_namelen = lmsg->msg_namelen;
446 	bmsg->msg_iov = lmsg->msg_iov;
447 	bmsg->msg_iovlen = lmsg->msg_iovlen;
448 	bmsg->msg_control = lmsg->msg_control;
449 	bmsg->msg_controllen = lmsg->msg_controllen;
450 	bmsg->msg_flags = lmsg->msg_flags;
451 }
452 
453 static void
bsd_to_linux_msghdr(const struct msghdr * bmsg,struct linux_msghdr * lmsg)454 bsd_to_linux_msghdr(const struct msghdr *bmsg, struct linux_msghdr *lmsg)
455 {
456 	memset(lmsg, 0, sizeof(*lmsg));
457 	lmsg->msg_name = bmsg->msg_name;
458 	lmsg->msg_namelen = bmsg->msg_namelen;
459 	lmsg->msg_iov = bmsg->msg_iov;
460 	lmsg->msg_iovlen = bmsg->msg_iovlen;
461 	lmsg->msg_control = bmsg->msg_control;
462 	lmsg->msg_controllen = bmsg->msg_controllen;
463 	lmsg->msg_flags = bmsg->msg_flags;
464 }
465 
466 int
linux_sys_sendmsg(struct lwp * l,const struct linux_sys_sendmsg_args * uap,register_t * retval)467 linux_sys_sendmsg(struct lwp *l, const struct linux_sys_sendmsg_args *uap, register_t *retval)
468 {
469 	/* {
470 		syscallarg(int) s;
471 		syscallarg(struct linux_msghdr *) msg;
472 		syscallarg(u_int) flags;
473 	} */
474 	struct msghdr	msg;
475 	struct linux_msghdr lmsg;
476 	int		error;
477 	int		bflags;
478 	struct sockaddr_big nam;
479 	u_int8_t	*control;
480 	struct mbuf     *ctl_mbuf = NULL;
481 
482 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
483 	if (error)
484 		return error;
485 	linux_to_bsd_msghdr(&lmsg, &msg);
486 
487 	msg.msg_flags = MSG_IOVUSRSPACE;
488 
489 	/*
490 	 * Translate message flags.
491 	 */
492 	bflags = linux_to_bsd_msg_flags(SCARG(uap, flags));
493 	if (bflags < 0)
494 		/* Some supported flag */
495 		return EINVAL;
496 
497 	if (lmsg.msg_name) {
498 		/* Read in and convert the sockaddr */
499 		error = linux_get_sa(l, SCARG(uap, s), &nam, msg.msg_name,
500 		    msg.msg_namelen);
501 		if (error)
502 			return (error);
503 		msg.msg_name = &nam;
504 	}
505 
506 	/*
507 	 * Handle cmsg if there is any.
508 	 */
509 	if (LINUX_CMSG_FIRSTHDR(&lmsg)) {
510 		struct linux_cmsghdr l_cmsg, *l_cc;
511 		struct cmsghdr *cmsg;
512 		ssize_t resid = msg.msg_controllen;
513 		size_t clen, cidx = 0, cspace;
514 
515 		ctl_mbuf = m_get(M_WAIT, MT_CONTROL);
516 		clen = MLEN;
517 		control = mtod(ctl_mbuf, void *);
518 
519 		l_cc = LINUX_CMSG_FIRSTHDR(&lmsg);
520 		do {
521 			error = copyin(l_cc, &l_cmsg, sizeof(l_cmsg));
522 			if (error)
523 				goto done;
524 
525 			/*
526 			 * Sanity check the control message length.
527 			 */
528 			if (l_cmsg.cmsg_len > resid
529 			    || l_cmsg.cmsg_len < sizeof l_cmsg) {
530 				error = EINVAL;
531 				goto done;
532 			}
533 
534 			/*
535 			 * Refuse unsupported control messages, and
536 			 * translate fields as appropriate.
537 			 */
538 			switch (l_cmsg.cmsg_level) {
539 			case LINUX_SOL_SOCKET:
540 				/* It only differs on some archs */
541 				if (LINUX_SOL_SOCKET != SOL_SOCKET)
542 					l_cmsg.cmsg_level = SOL_SOCKET;
543 
544 				switch(l_cmsg.cmsg_type) {
545 				case LINUX_SCM_RIGHTS:
546 					/* Linux SCM_RIGHTS is same as NetBSD */
547 					break;
548 
549 				case LINUX_SCM_CREDENTIALS:
550 					/* no native equivalent, just drop it */
551 					if (control != mtod(ctl_mbuf, void *))
552 						free(control, M_MBUF);
553 					m_free(ctl_mbuf);
554 					ctl_mbuf = NULL;
555 					msg.msg_control = NULL;
556 					msg.msg_controllen = 0;
557 					goto skipcmsg;
558 
559 				default:
560 					/* other types not supported */
561 					error = EINVAL;
562 					goto done;
563 				}
564 				break;
565 			default:
566 				/* pray and leave intact */
567 				break;
568 			}
569 
570 			cspace = CMSG_SPACE(l_cmsg.cmsg_len - sizeof(l_cmsg));
571 
572 			/* Check the buffer is big enough */
573 			if (__predict_false(cidx + cspace > clen)) {
574 				u_int8_t *nc;
575 				size_t nclen;
576 
577 				nclen = cidx + cspace;
578 				if (nclen >= PAGE_SIZE) {
579 					error = EINVAL;
580 					goto done;
581 				}
582 				nc = realloc(clen <= MLEN ? NULL : control,
583 						nclen, M_TEMP, M_WAITOK);
584 				if (!nc) {
585 					error = ENOMEM;
586 					goto done;
587 				}
588 				if (cidx <= MLEN)
589 					/* Old buffer was in mbuf... */
590 					memcpy(nc, control, cidx);
591 				control = nc;
592 				clen = nclen;
593 			}
594 
595 			/* Copy header */
596 			cmsg = (void *)&control[cidx];
597 			cmsg->cmsg_len = l_cmsg.cmsg_len + LINUX_CMSG_ALIGN_DELTA;
598 			cmsg->cmsg_level = l_cmsg.cmsg_level;
599 			cmsg->cmsg_type = l_cmsg.cmsg_type;
600 
601 			/* Zero area between header and data */
602 			memset(cmsg + 1, 0,
603 				CMSG_ALIGN(sizeof(*cmsg)) - sizeof(*cmsg));
604 
605 			/* Copyin the data */
606 			error = copyin(LINUX_CMSG_DATA(l_cc),
607 				CMSG_DATA(cmsg),
608 				l_cmsg.cmsg_len - sizeof(l_cmsg));
609 			if (error)
610 				goto done;
611 
612 			resid -= LINUX_CMSG_ALIGN(l_cmsg.cmsg_len);
613 			cidx += cspace;
614 		} while ((l_cc = LINUX_CMSG_NXTHDR(&msg, l_cc, &l_cmsg)) && resid > 0);
615 
616 		/* If we allocated a buffer, attach to mbuf */
617 		if (cidx > MLEN) {
618 			MEXTADD(ctl_mbuf, control, clen, M_MBUF, NULL, NULL);
619 			ctl_mbuf->m_flags |= M_EXT_RW;
620 		}
621 		control = NULL;
622 		ctl_mbuf->m_len = cidx;
623 
624 		msg.msg_control = ctl_mbuf;
625 		msg.msg_flags |= MSG_CONTROLMBUF;
626 
627 		ktrkuser("mbcontrol", mtod(ctl_mbuf, void *),
628 		    msg.msg_controllen);
629 	}
630 
631 skipcmsg:
632 	error = do_sys_sendmsg(l, SCARG(uap, s), &msg, bflags, retval);
633 	/* Freed internally */
634 	ctl_mbuf = NULL;
635 
636 done:
637 	if (ctl_mbuf != NULL) {
638 		if (control != NULL && control != mtod(ctl_mbuf, void *))
639 			free(control, M_MBUF);
640 		m_free(ctl_mbuf);
641 	}
642 	return (error);
643 }
644 
645 int
linux_sys_recvfrom(struct lwp * l,const struct linux_sys_recvfrom_args * uap,register_t * retval)646 linux_sys_recvfrom(struct lwp *l, const struct linux_sys_recvfrom_args *uap, register_t *retval)
647 {
648 	/* {
649 		syscallarg(int) s;
650 		syscallarg(void *) buf;
651 		syscallarg(int) len;
652 		syscallarg(int) flags;
653 		syscallarg(struct osockaddr *) from;
654 		syscallarg(int *) fromlenaddr;
655 	} */
656 	int		error;
657 	struct sys_recvfrom_args bra;
658 
659 	SCARG(&bra, s) = SCARG(uap, s);
660 	SCARG(&bra, buf) = SCARG(uap, buf);
661 	SCARG(&bra, len) = SCARG(uap, len);
662 	SCARG(&bra, flags) = SCARG(uap, flags);
663 	SCARG(&bra, from) = (struct sockaddr *) SCARG(uap, from);
664 	SCARG(&bra, fromlenaddr) = (socklen_t *)SCARG(uap, fromlenaddr);
665 
666 	if ((error = sys_recvfrom(l, &bra, retval)))
667 		return (error);
668 
669 	if (SCARG(uap, from) && (error = linux_sa_put(SCARG(uap, from))))
670 		return (error);
671 
672 	return (0);
673 }
674 
675 static int
linux_copyout_msg_control(struct lwp * l,struct msghdr * mp,struct mbuf * control)676 linux_copyout_msg_control(struct lwp *l, struct msghdr *mp, struct mbuf *control)
677 {
678 	int dlen, error = 0;
679 	struct cmsghdr *cmsg;
680 	struct linux_cmsghdr linux_cmsg;
681 	struct mbuf *m;
682 	char *q, *q_end;
683 
684 	if (mp->msg_controllen <= 0 || control == 0) {
685 		mp->msg_controllen = 0;
686 		free_control_mbuf(l, control, control);
687 		return 0;
688 	}
689 
690 	ktrkuser("msgcontrol", mtod(control, void *), mp->msg_controllen);
691 
692 	q = (char *)mp->msg_control;
693 	q_end = q + mp->msg_controllen;
694 
695 	for (m = control; m != NULL; ) {
696 		cmsg = mtod(m, struct cmsghdr *);
697 
698 		/*
699 		 * Fixup cmsg. We handle two things:
700 		 * 0. different sizeof cmsg_len.
701 		 * 1. different values for level/type on some archs
702 		 * 2. different alignment of CMSG_DATA on some archs
703 		 */
704 		memset(&linux_cmsg, 0, sizeof(linux_cmsg));
705 		linux_cmsg.cmsg_len = cmsg->cmsg_len - LINUX_CMSG_ALIGN_DELTA;
706 		linux_cmsg.cmsg_level = cmsg->cmsg_level;
707 		linux_cmsg.cmsg_type = cmsg->cmsg_type;
708 
709 		dlen = q_end - q;
710 		if (linux_cmsg.cmsg_len > dlen) {
711 			/* Not enough room for the parameter */
712 			dlen -= sizeof linux_cmsg;
713 			if (dlen <= 0)
714 				/* Discard if header wont fit */
715 				break;
716 			mp->msg_flags |= MSG_CTRUNC;
717 			if (linux_cmsg.cmsg_level == SOL_SOCKET
718 			    && linux_cmsg.cmsg_type == SCM_RIGHTS)
719 				/* Do not truncate me ... */
720 				break;
721 		} else
722 			dlen = linux_cmsg.cmsg_len - sizeof linux_cmsg;
723 
724 		switch (linux_cmsg.cmsg_level) {
725 		case SOL_SOCKET:
726 			linux_cmsg.cmsg_level = LINUX_SOL_SOCKET;
727 			switch (linux_cmsg.cmsg_type) {
728 			case SCM_RIGHTS:
729 				/* Linux SCM_RIGHTS is same as NetBSD */
730 				break;
731 
732 			default:
733 				/* other types not supported */
734 				error = EINVAL;
735 				goto done;
736 			}
737 			/* machine dependent ! */
738 			break;
739 		default:
740 			/* pray and leave intact */
741 			break;
742 		}
743 
744 		/* There can be padding between the header and data... */
745 		error = copyout(&linux_cmsg, q, sizeof linux_cmsg);
746 		if (error != 0) {
747 			error = copyout(CCMSG_DATA(cmsg), q + sizeof linux_cmsg,
748 			    dlen);
749 		}
750 		if (error != 0) {
751 			/* We must free all the SCM_RIGHTS */
752 			m = control;
753 			break;
754 		}
755 		m = m->m_next;
756 		if (m == NULL || q + LINUX_CMSG_SPACE(dlen) > q_end) {
757 			q += LINUX_CMSG_LEN(dlen);
758 			break;
759 		}
760 		q += LINUX_CMSG_SPACE(dlen);
761 	}
762 
763   done:
764 	free_control_mbuf(l, control, m);
765 
766 	mp->msg_controllen = q - (char *)mp->msg_control;
767 	return error;
768 }
769 
770 int
linux_sys_recvmsg(struct lwp * l,const struct linux_sys_recvmsg_args * uap,register_t * retval)771 linux_sys_recvmsg(struct lwp *l, const struct linux_sys_recvmsg_args *uap, register_t *retval)
772 {
773 	/* {
774 		syscallarg(int) s;
775 		syscallarg(struct linux_msghdr *) msg;
776 		syscallarg(u_int) flags;
777 	} */
778 	struct msghdr	msg;
779 	struct linux_msghdr lmsg;
780 	int		error;
781 	struct mbuf	*from, *control;
782 
783 	error = copyin(SCARG(uap, msg), &lmsg, sizeof(lmsg));
784 	if (error)
785 		return (error);
786 	linux_to_bsd_msghdr(&lmsg, &msg);
787 
788 	msg.msg_flags = linux_to_bsd_msg_flags(SCARG(uap, flags));
789 	if (msg.msg_flags < 0) {
790 		/* Some unsupported flag */
791 		return (EINVAL);
792 	}
793 	msg.msg_flags |= MSG_IOVUSRSPACE;
794 
795 	error = do_sys_recvmsg(l, SCARG(uap, s), &msg, &from,
796 	    msg.msg_control != NULL ? &control : NULL, retval);
797 	if (error != 0)
798 		return error;
799 
800 	if (msg.msg_control != NULL)
801 		error = linux_copyout_msg_control(l, &msg, control);
802 
803 	if (error == 0 && from != 0) {
804 		mtod(from, struct osockaddr *)->sa_family =
805 		    bsd_to_linux_domain(mtod(from, struct sockaddr *)->sa_family);
806 		error = copyout_sockname(msg.msg_name, &msg.msg_namelen, 0,
807 			from);
808 	} else
809 		msg.msg_namelen = 0;
810 
811 	if (from != NULL)
812 		m_free(from);
813 
814 	if (error == 0) {
815 		msg.msg_flags = bsd_to_linux_msg_flags(msg.msg_flags);
816 		if (msg.msg_flags < 0)
817 			/* Some flag unsupported by Linux */
818 			error = EINVAL;
819 		else {
820 			ktrkuser("msghdr", &msg, sizeof(msg));
821 			bsd_to_linux_msghdr(&msg, &lmsg);
822 			error = copyout(&lmsg, SCARG(uap, msg), sizeof(lmsg));
823 		}
824 	}
825 
826 	return (error);
827 }
828 
829 /*
830  * Convert socket option level from Linux to NetBSD value. Only SOL_SOCKET
831  * is different, the rest matches IPPROTO_* on both systems.
832  */
833 int
linux_to_bsd_sopt_level(int llevel)834 linux_to_bsd_sopt_level(int llevel)
835 {
836 
837 	switch (llevel) {
838 	case LINUX_SOL_SOCKET:
839 		return SOL_SOCKET;
840 	case LINUX_SOL_IP:
841 		return IPPROTO_IP;
842 #ifdef INET6
843 	case LINUX_SOL_IPV6:
844 		return IPPROTO_IPV6;
845 #endif
846 	case LINUX_SOL_TCP:
847 		return IPPROTO_TCP;
848 	case LINUX_SOL_UDP:
849 		return IPPROTO_UDP;
850 	default:
851 		return -1;
852 	}
853 }
854 
855 /*
856  * Convert Linux socket level socket option numbers to NetBSD values.
857  */
858 int
linux_to_bsd_so_sockopt(int lopt)859 linux_to_bsd_so_sockopt(int lopt)
860 {
861 
862 	switch (lopt) {
863 	case LINUX_SO_DEBUG:
864 		return SO_DEBUG;
865 	case LINUX_SO_REUSEADDR:
866 		/*
867 		 * Linux does not implement SO_REUSEPORT, but allows reuse of
868 		 * a host:port pair through SO_REUSEADDR even if the address
869 		 * is not a multicast-address. Effectively, this means that we
870 		 * should use SO_REUSEPORT to allow Linux applications to not
871 		 * exit with EADDRINUSE
872 		 */
873 		return SO_REUSEPORT;
874 	case LINUX_SO_TYPE:
875 		return SO_TYPE;
876 	case LINUX_SO_ERROR:
877 		return SO_ERROR;
878 	case LINUX_SO_DONTROUTE:
879 		return SO_DONTROUTE;
880 	case LINUX_SO_BROADCAST:
881 		return SO_BROADCAST;
882 	case LINUX_SO_SNDBUF:
883 		return SO_SNDBUF;
884 	case LINUX_SO_RCVBUF:
885 		return SO_RCVBUF;
886 	case LINUX_SO_KEEPALIVE:
887 		return SO_KEEPALIVE;
888 	case LINUX_SO_OOBINLINE:
889 		return SO_OOBINLINE;
890 	case LINUX_SO_NO_CHECK:
891 	case LINUX_SO_PRIORITY:
892 		return -1;
893 	case LINUX_SO_LINGER:
894 		return SO_LINGER;
895 	case LINUX_SO_BSDCOMPAT:
896 	case LINUX_SO_PASSCRED:
897 	case LINUX_SO_PEERCRED:
898 		return -1;
899 	case LINUX_SO_RCVLOWAT:
900 		return SO_RCVLOWAT;
901 	case LINUX_SO_SNDLOWAT:
902 		return SO_SNDLOWAT;
903 	case LINUX_SO_RCVTIMEO:
904 		return SO_RCVTIMEO;
905 	case LINUX_SO_SNDTIMEO:
906 		return SO_SNDTIMEO;
907 	case LINUX_SO_SECURITY_AUTHENTICATION:
908 	case LINUX_SO_SECURITY_ENCRYPTION_TRANSPORT:
909 	case LINUX_SO_SECURITY_ENCRYPTION_NETWORK:
910 	case LINUX_SO_BINDTODEVICE:
911 	case LINUX_SO_ATTACH_FILTER:
912 	case LINUX_SO_DETACH_FILTER:
913 	case LINUX_SO_PEERNAME:
914 		return -1;
915 	case LINUX_SO_TIMESTAMP:
916 		return SO_TIMESTAMP;
917 	case LINUX_SO_ACCEPTCONN:
918 	case LINUX_SO_PEERSEC:
919 	case LINUX_SO_SNDBUFFORCE:
920 	case LINUX_SO_RCVBUFFORCE:
921 	case LINUX_SO_PASSSEC:
922 	case LINUX_SO_TIMESTAMPNS:
923 	case LINUX_SO_MARK:
924 	case LINUX_SO_TIMESTAMPING:
925 	case LINUX_SO_PROTOCOL:
926 	case LINUX_SO_DOMAIN:
927 	case LINUX_SO_RXQ_OVFL:
928 	case LINUX_SO_WIFI_STATUS:
929 	case LINUX_SO_PEEK_OFF:
930 	case LINUX_SO_NOFCS:
931 	default:
932 		return -1;
933 	}
934 }
935 
936 /*
937  * Convert Linux IP level socket option number to NetBSD values.
938  */
939 int
linux_to_bsd_ip_sockopt(int lopt)940 linux_to_bsd_ip_sockopt(int lopt)
941 {
942 
943 	switch (lopt) {
944 	case LINUX_IP_TOS:
945 		return IP_TOS;
946 	case LINUX_IP_TTL:
947 		return IP_TTL;
948 	case LINUX_IP_HDRINCL:
949 		return IP_HDRINCL;
950 	case LINUX_IP_MULTICAST_TTL:
951 		return IP_MULTICAST_TTL;
952 	case LINUX_IP_MULTICAST_LOOP:
953 		return IP_MULTICAST_LOOP;
954 	case LINUX_IP_MULTICAST_IF:
955 		return IP_MULTICAST_IF;
956 	case LINUX_IP_ADD_MEMBERSHIP:
957 		return IP_ADD_MEMBERSHIP;
958 	case LINUX_IP_DROP_MEMBERSHIP:
959 		return IP_DROP_MEMBERSHIP;
960 	case LINUX_IP_RECVERR:
961 		return -2;	/* ignored */
962 	default:
963 		return -1;
964 	}
965 }
966 
967 /*
968  * Convert Linux IPV6 level socket option number to NetBSD values.
969  */
970 #ifdef INET6
971 int
linux_to_bsd_ipv6_sockopt(int lopt)972 linux_to_bsd_ipv6_sockopt(int lopt)
973 {
974 
975 	switch (lopt) {
976 	case LINUX_IPV6_V6ONLY:
977 		return IPV6_V6ONLY;
978 	default:
979 		return -1;
980 	}
981 }
982 #endif
983 
984 /*
985  * Convert Linux TCP level socket option number to NetBSD values.
986  */
987 int
linux_to_bsd_tcp_sockopt(int lopt)988 linux_to_bsd_tcp_sockopt(int lopt)
989 {
990 
991 	switch (lopt) {
992 	case LINUX_TCP_NODELAY:
993 		return TCP_NODELAY;
994 	case LINUX_TCP_MAXSEG:
995 		return TCP_MAXSEG;
996 	default:
997 		return -1;
998 	}
999 }
1000 
1001 /*
1002  * Convert Linux UDP level socket option number to NetBSD values.
1003  */
1004 int
linux_to_bsd_udp_sockopt(int lopt)1005 linux_to_bsd_udp_sockopt(int lopt)
1006 {
1007 
1008 	switch (lopt) {
1009 	default:
1010 		return -1;
1011 	}
1012 }
1013 
1014 /*
1015  * Another reasonably straightforward function: setsockopt(2).
1016  * The level and option numbers are converted; the values passed
1017  * are not (yet) converted, the ones currently implemented don't
1018  * need conversion, as they are the same on both systems.
1019  */
1020 int
linux_sys_setsockopt(struct lwp * l,const struct linux_sys_setsockopt_args * uap,register_t * retval)1021 linux_sys_setsockopt(struct lwp *l, const struct linux_sys_setsockopt_args *uap, register_t *retval)
1022 {
1023 	/* {
1024 		syscallarg(int) s;
1025 		syscallarg(int) level;
1026 		syscallarg(int) optname;
1027 		syscallarg(void *) optval;
1028 		syscallarg(int) optlen;
1029 	} */
1030 	struct sys_setsockopt_args bsa;
1031 	int name;
1032 
1033 	SCARG(&bsa, s) = SCARG(uap, s);
1034 	SCARG(&bsa, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
1035 	SCARG(&bsa, val) = SCARG(uap, optval);
1036 	SCARG(&bsa, valsize) = SCARG(uap, optlen);
1037 
1038 	/*
1039 	 * Linux supports only SOL_SOCKET for AF_LOCAL domain sockets
1040 	 * and returns EOPNOTSUPP for other levels
1041 	 */
1042 	if (SCARG(&bsa, level) != SOL_SOCKET) {
1043 		struct socket *so;
1044 		int error, family;
1045 
1046 		/* fd_getsock() will use the descriptor for us */
1047 	    	if ((error = fd_getsock(SCARG(&bsa, s), &so)) != 0)
1048 		    	return error;
1049 		family = so->so_proto->pr_domain->dom_family;
1050 		fd_putfile(SCARG(&bsa, s));
1051 
1052 		if (family == AF_LOCAL)
1053 			return EOPNOTSUPP;
1054 	}
1055 
1056 	switch (SCARG(&bsa, level)) {
1057 	case SOL_SOCKET:
1058 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1059 		break;
1060 	case IPPROTO_IP:
1061 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1062 		break;
1063 #ifdef INET6
1064 	case IPPROTO_IPV6:
1065 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1066 		break;
1067 #endif
1068 	case IPPROTO_TCP:
1069 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1070 		break;
1071 	case IPPROTO_UDP:
1072 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1073 		break;
1074 	default:
1075 		return EINVAL;
1076 	}
1077 
1078 	if (name == -1)
1079 		return EINVAL;
1080 	if (name == -2)
1081 		return 0;
1082 	SCARG(&bsa, name) = name;
1083 
1084 	return sys_setsockopt(l, &bsa, retval);
1085 }
1086 
1087 /*
1088  * getsockopt(2) is very much the same as setsockopt(2) (see above)
1089  */
1090 int
linux_sys_getsockopt(struct lwp * l,const struct linux_sys_getsockopt_args * uap,register_t * retval)1091 linux_sys_getsockopt(struct lwp *l, const struct linux_sys_getsockopt_args *uap, register_t *retval)
1092 {
1093 	/* {
1094 		syscallarg(int) s;
1095 		syscallarg(int) level;
1096 		syscallarg(int) optname;
1097 		syscallarg(void *) optval;
1098 		syscallarg(int *) optlen;
1099 	} */
1100 	struct sys_getsockopt_args bga;
1101 	int name;
1102 
1103 	SCARG(&bga, s) = SCARG(uap, s);
1104 	SCARG(&bga, level) = linux_to_bsd_sopt_level(SCARG(uap, level));
1105 	SCARG(&bga, val) = SCARG(uap, optval);
1106 	SCARG(&bga, avalsize) = (socklen_t *)SCARG(uap, optlen);
1107 
1108 	switch (SCARG(&bga, level)) {
1109 	case SOL_SOCKET:
1110 		name = linux_to_bsd_so_sockopt(SCARG(uap, optname));
1111 		break;
1112 	case IPPROTO_IP:
1113 		name = linux_to_bsd_ip_sockopt(SCARG(uap, optname));
1114 		break;
1115 #ifdef INET6
1116 	case IPPROTO_IPV6:
1117 		name = linux_to_bsd_ipv6_sockopt(SCARG(uap, optname));
1118 		break;
1119 #endif
1120 	case IPPROTO_TCP:
1121 		name = linux_to_bsd_tcp_sockopt(SCARG(uap, optname));
1122 		break;
1123 	case IPPROTO_UDP:
1124 		name = linux_to_bsd_udp_sockopt(SCARG(uap, optname));
1125 		break;
1126 	default:
1127 		return EINVAL;
1128 	}
1129 
1130 	if (name == -1)
1131 		return EINVAL;
1132 	SCARG(&bga, name) = name;
1133 
1134 	return sys_getsockopt(l, &bga, retval);
1135 }
1136 
1137 int
linux_getifname(struct lwp * l,register_t * retval,void * data)1138 linux_getifname(struct lwp *l, register_t *retval, void *data)
1139 {
1140 	struct ifnet *ifp;
1141 	struct linux_ifreq ifr;
1142 	int error;
1143 	int s;
1144 
1145 	error = copyin(data, &ifr, sizeof(ifr));
1146 	if (error)
1147 		return error;
1148 
1149 	s = pserialize_read_enter();
1150 	ifp = if_byindex(ifr.ifr_ifru.ifru_ifindex);
1151 	if (ifp == NULL) {
1152 		pserialize_read_exit(s);
1153 		return ENODEV;
1154 	}
1155 
1156 	strncpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name));
1157 	pserialize_read_exit(s);
1158 
1159 	return copyout(&ifr, data, sizeof(ifr));
1160 }
1161 
1162 int
linux_getifconf(struct lwp * l,register_t * retval,void * data)1163 linux_getifconf(struct lwp *l, register_t *retval, void *data)
1164 {
1165 	struct linux_ifreq ifr, *ifrp = NULL;
1166 	struct linux_ifconf ifc;
1167 	struct ifnet *ifp;
1168 	struct sockaddr *sa;
1169 	struct osockaddr *osa;
1170 	int space = 0, error;
1171 	const int sz = (int)sizeof(ifr);
1172 	bool docopy;
1173 	int s;
1174 	int bound;
1175 	struct psref psref;
1176 
1177 	error = copyin(data, &ifc, sizeof(ifc));
1178 	if (error)
1179 		return error;
1180 
1181 	docopy = ifc.ifc_req != NULL;
1182 	if (docopy) {
1183 		if (ifc.ifc_len < 0)
1184 			return EINVAL;
1185 
1186 		space = ifc.ifc_len;
1187 		ifrp = ifc.ifc_req;
1188 	}
1189 	memset(&ifr, 0, sizeof(ifr));
1190 
1191 	bound = curlwp_bind();
1192 	s = pserialize_read_enter();
1193 	IFNET_READER_FOREACH(ifp) {
1194 		struct ifaddr *ifa;
1195 		if_acquire(ifp, &psref);
1196 		pserialize_read_exit(s);
1197 
1198 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
1199 		    sizeof(ifr.ifr_name));
1200 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') {
1201 			error = ENAMETOOLONG;
1202 			goto release_exit;
1203 		}
1204 
1205 		s = pserialize_read_enter();
1206 		IFADDR_READER_FOREACH(ifa, ifp) {
1207 			struct psref psref_ifa;
1208 			ifa_acquire(ifa, &psref_ifa);
1209 			pserialize_read_exit(s);
1210 
1211 			sa = ifa->ifa_addr;
1212 			if (sa->sa_family != AF_INET ||
1213 			    sa->sa_len > sizeof(*osa))
1214 				goto next;
1215 			memcpy(&ifr.ifr_addr, sa, sa->sa_len);
1216 			osa = (struct osockaddr *)&ifr.ifr_addr;
1217 			osa->sa_family = sa->sa_family;
1218 			if (space >= sz) {
1219 				error = copyout(&ifr, ifrp, sz);
1220 				if (error != 0) {
1221 					ifa_release(ifa, &psref_ifa);
1222 					goto release_exit;
1223 				}
1224 				ifrp++;
1225 			}
1226 			space -= sz;
1227 		next:
1228 			s = pserialize_read_enter();
1229 			ifa_release(ifa, &psref_ifa);
1230 		}
1231 
1232 		KASSERT(pserialize_in_read_section());
1233 		if_release(ifp, &psref);
1234 	}
1235 	pserialize_read_exit(s);
1236 	curlwp_bindx(bound);
1237 
1238 	if (docopy)
1239 		ifc.ifc_len -= space;
1240 	else
1241 		ifc.ifc_len = -space;
1242 
1243 	return copyout(&ifc, data, sizeof(ifc));
1244 
1245 release_exit:
1246 	if_release(ifp, &psref);
1247 	curlwp_bindx(bound);
1248 	return error;
1249 }
1250 
1251 int
linux_getifhwaddr(struct lwp * l,register_t * retval,u_int fd,void * data)1252 linux_getifhwaddr(struct lwp *l, register_t *retval, u_int fd,
1253     void *data)
1254 {
1255 	/* Not the full structure, just enough to map what we do here */
1256 	struct linux_ifreq lreq;
1257 	file_t *fp;
1258 	struct ifaddr *ifa;
1259 	struct ifnet *ifp;
1260 	struct sockaddr_dl *sadl;
1261 	int error, found;
1262 	int index, ifnum;
1263 	int s;
1264 
1265 	/*
1266 	 * We can't emulate this ioctl by calling sys_ioctl() to run
1267 	 * SIOCGIFCONF, because the user buffer is not of the right
1268 	 * type to take those results.  We can't use kernel buffers to
1269 	 * receive the results, as the implementation of sys_ioctl()
1270 	 * and ifconf() [which implements SIOCGIFCONF] use
1271 	 * copyin()/copyout() which will fail on kernel addresses.
1272 	 *
1273 	 * So, we must duplicate code from sys_ioctl() and ifconf().  Ugh.
1274 	 */
1275 
1276 	if ((fp = fd_getfile(fd)) == NULL)
1277 		return (EBADF);
1278 
1279 	KERNEL_LOCK(1, NULL);
1280 
1281 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
1282 		error = EBADF;
1283 		goto out;
1284 	}
1285 
1286 	error = copyin(data, &lreq, sizeof(lreq));
1287 	if (error)
1288 		goto out;
1289 	lreq.ifr_name[LINUX_IFNAMSIZ-1] = '\0';		/* just in case */
1290 
1291 	/*
1292 	 * Try real interface name first, then fake "ethX"
1293 	 */
1294 	found = 0;
1295 	s = pserialize_read_enter();
1296 	IFNET_READER_FOREACH(ifp) {
1297 		if (found)
1298 			break;
1299 		if (strcmp(lreq.ifr_name, ifp->if_xname))
1300 			/* not this interface */
1301 			continue;
1302 
1303 		found=1;
1304 		if (IFADDR_READER_EMPTY(ifp)) {
1305 			pserialize_read_exit(s);
1306 			error = ENODEV;
1307 			goto out;
1308 		}
1309 		IFADDR_READER_FOREACH(ifa, ifp) {
1310 			sadl = satosdl(ifa->ifa_addr);
1311 			/* only return ethernet addresses */
1312 			/* XXX what about FDDI, etc. ? */
1313 			if (sadl->sdl_family != AF_LINK ||
1314 			    sadl->sdl_type != IFT_ETHER)
1315 				continue;
1316 			memcpy(&lreq.ifr_hwaddr.sa_data, CLLADDR(sadl),
1317 			       MIN(sadl->sdl_alen,
1318 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1319 			lreq.ifr_hwaddr.sa_family =
1320 				sadl->sdl_family;
1321 			pserialize_read_exit(s);
1322 
1323 			error = copyout(&lreq, data, sizeof(lreq));
1324 			goto out;
1325 		}
1326 	}
1327 	pserialize_read_exit(s);
1328 
1329 	if (strncmp(lreq.ifr_name, "eth", 3) != 0) {
1330 		/* unknown interface, not even an "eth*" name */
1331 		error = ENODEV;
1332 		goto out;
1333 	}
1334 
1335 	for (ifnum = 0, index = 3;
1336 	     index < LINUX_IFNAMSIZ && lreq.ifr_name[index] != '\0';
1337 	     index++) {
1338 		ifnum *= 10;
1339 		ifnum += lreq.ifr_name[index] - '0';
1340 	}
1341 
1342 	error = EINVAL;			/* in case we don't find one */
1343 	s = pserialize_read_enter();
1344 	IFNET_READER_FOREACH(ifp) {
1345 		memcpy(lreq.ifr_name, ifp->if_xname,
1346 		       MIN(LINUX_IFNAMSIZ, IFNAMSIZ));
1347 		IFADDR_READER_FOREACH(ifa, ifp) {
1348 			sadl = satosdl(ifa->ifa_addr);
1349 			/* only return ethernet addresses */
1350 			/* XXX what about FDDI, etc. ? */
1351 			if (sadl->sdl_family != AF_LINK ||
1352 			    sadl->sdl_type != IFT_ETHER)
1353 				continue;
1354 			if (ifnum--)
1355 				/* not the requested iface */
1356 				continue;
1357 			memcpy(&lreq.ifr_hwaddr.sa_data,
1358 			       CLLADDR(sadl),
1359 			       MIN(sadl->sdl_alen,
1360 				   sizeof(lreq.ifr_hwaddr.sa_data)));
1361 			lreq.ifr_hwaddr.sa_family =
1362 				sadl->sdl_family;
1363 			pserialize_read_exit(s);
1364 
1365 			error = copyout(&lreq, data, sizeof(lreq));
1366 			goto out;
1367 		}
1368 	}
1369 	pserialize_read_exit(s);
1370 
1371 out:
1372 	KERNEL_UNLOCK_ONE(NULL);
1373 	fd_putfile(fd);
1374 	return error;
1375 }
1376 
1377 int
linux_ioctl_socket(struct lwp * l,const struct linux_sys_ioctl_args * uap,register_t * retval)1378 linux_ioctl_socket(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
1379 {
1380 	/* {
1381 		syscallarg(int) fd;
1382 		syscallarg(u_long) com;
1383 		syscallarg(void *) data;
1384 	} */
1385 	u_long com;
1386 	int error = 0, isdev = 0, dosys = 1;
1387 	struct sys_ioctl_args ia;
1388 	file_t *fp;
1389 	struct vnode *vp;
1390 	int (*ioctlf)(file_t *, u_long, void *);
1391 	struct ioctl_pt pt;
1392 
1393 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
1394 		return (EBADF);
1395 
1396 	if (fp->f_type == DTYPE_VNODE) {
1397 		vp = (struct vnode *)fp->f_data;
1398 		isdev = vp->v_type == VCHR;
1399 	}
1400 
1401 	/*
1402 	 * Don't try to interpret socket ioctl calls that are done
1403 	 * on a device filedescriptor, just pass them through, to
1404 	 * emulate Linux behaviour. Use PTIOCLINUX so that the
1405 	 * device will only handle these if it's prepared to do
1406 	 * so, to avoid unexpected things from happening.
1407 	 */
1408 	if (isdev) {
1409 		dosys = 0;
1410 		ioctlf = fp->f_ops->fo_ioctl;
1411 		pt.com = SCARG(uap, com);
1412 		pt.data = SCARG(uap, data);
1413 		error = ioctlf(fp, PTIOCLINUX, &pt);
1414 		/*
1415 		 * XXX hack: if the function returns EJUSTRETURN,
1416 		 * it has stuffed a sysctl return value in pt.data.
1417 		 */
1418 		if (error == EJUSTRETURN) {
1419 			retval[0] = (register_t)pt.data;
1420 			error = 0;
1421 		}
1422 		goto out;
1423 	}
1424 
1425 	com = SCARG(uap, com);
1426 	retval[0] = 0;
1427 
1428 	switch (com) {
1429 	case LINUX_SIOCGIFNAME:
1430 		error = linux_getifname(l, retval, SCARG(uap, data));
1431 		dosys = 0;
1432 		break;
1433 	case LINUX_SIOCGIFCONF:
1434 		error = linux_getifconf(l, retval, SCARG(uap, data));
1435 		dosys = 0;
1436 		break;
1437 	case LINUX_SIOCGIFFLAGS:
1438 		SCARG(&ia, com) = OSIOCGIFFLAGS;
1439 		break;
1440 	case LINUX_SIOCSIFFLAGS:
1441 		SCARG(&ia, com) = OSIOCSIFFLAGS;
1442 		break;
1443 	case LINUX_SIOCGIFADDR:
1444 		SCARG(&ia, com) = OOSIOCGIFADDR;
1445 		break;
1446 	case LINUX_SIOCGIFDSTADDR:
1447 		SCARG(&ia, com) = OOSIOCGIFDSTADDR;
1448 		break;
1449 	case LINUX_SIOCGIFBRDADDR:
1450 		SCARG(&ia, com) = OOSIOCGIFBRDADDR;
1451 		break;
1452 	case LINUX_SIOCGIFNETMASK:
1453 		SCARG(&ia, com) = OOSIOCGIFNETMASK;
1454 		break;
1455 	case LINUX_SIOCGIFMTU:
1456 		SCARG(&ia, com) = OSIOCGIFMTU;
1457 		break;
1458 	case LINUX_SIOCADDMULTI:
1459 		SCARG(&ia, com) = OSIOCADDMULTI;
1460 		break;
1461 	case LINUX_SIOCDELMULTI:
1462 		SCARG(&ia, com) = OSIOCDELMULTI;
1463 		break;
1464 	case LINUX_SIOCGIFHWADDR:
1465 		error = linux_getifhwaddr(l, retval, SCARG(uap, fd),
1466 		    SCARG(uap, data));
1467 		dosys = 0;
1468 		break;
1469 	default:
1470 		error = EINVAL;
1471 	}
1472 
1473  out:
1474  	fd_putfile(SCARG(uap, fd));
1475 
1476 	if (error ==0 && dosys) {
1477 		SCARG(&ia, fd) = SCARG(uap, fd);
1478 		SCARG(&ia, data) = SCARG(uap, data);
1479 		error = sys_ioctl(curlwp, &ia, retval);
1480 	}
1481 
1482 	return error;
1483 }
1484 
1485 int
linux_sys_connect(struct lwp * l,const struct linux_sys_connect_args * uap,register_t * retval)1486 linux_sys_connect(struct lwp *l, const struct linux_sys_connect_args *uap, register_t *retval)
1487 {
1488 	/* {
1489 		syscallarg(int) s;
1490 		syscallarg(const struct sockaddr *) name;
1491 		syscallarg(int) namelen;
1492 	} */
1493 	int		error;
1494 	struct sockaddr_big sb;
1495 
1496 	error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name),
1497 	    SCARG(uap, namelen));
1498 	if (error)
1499 		return (error);
1500 
1501 	error = do_sys_connect(l, SCARG(uap, s), (struct sockaddr *)&sb);
1502 
1503 	if (error == EISCONN) {
1504 		struct socket *so;
1505 		int state, prflags;
1506 
1507 		/* fd_getsock() will use the descriptor for us */
1508 	    	if (fd_getsock(SCARG(uap, s), &so) != 0)
1509 		    	return EISCONN;
1510 
1511 		solock(so);
1512 		state = so->so_state;
1513 		prflags = so->so_proto->pr_flags;
1514 		sounlock(so);
1515 		fd_putfile(SCARG(uap, s));
1516 		/*
1517 		 * We should only let this call succeed once per
1518 		 * non-blocking connect; however we don't have
1519 		 * a convenient place to keep that state..
1520 		 */
1521 		if ((state & (SS_ISCONNECTED|SS_NBIO)) ==
1522 		    (SS_ISCONNECTED|SS_NBIO) &&
1523 		    (prflags & PR_CONNREQUIRED))
1524 			return 0;
1525 	}
1526 
1527 	return (error);
1528 }
1529 
1530 int
linux_sys_bind(struct lwp * l,const struct linux_sys_bind_args * uap,register_t * retval)1531 linux_sys_bind(struct lwp *l, const struct linux_sys_bind_args *uap, register_t *retval)
1532 {
1533 	/* {
1534 		syscallarg(int) s;
1535 		syscallarg(const struct osockaddr *) name;
1536 		syscallarg(int) namelen;
1537 	} */
1538 	int		error;
1539 	struct sockaddr_big sb;
1540 
1541 	error = linux_get_sa(l, SCARG(uap, s), &sb, SCARG(uap, name),
1542 	    SCARG(uap, namelen));
1543 	if (error)
1544 		return (error);
1545 
1546 	return do_sys_bind(l, SCARG(uap, s), (struct sockaddr *)&sb);
1547 }
1548 
1549 int
linux_sys_getsockname(struct lwp * l,const struct linux_sys_getsockname_args * uap,register_t * retval)1550 linux_sys_getsockname(struct lwp *l, const struct linux_sys_getsockname_args *uap, register_t *retval)
1551 {
1552 	/* {
1553 		syscallarg(int) fdes;
1554 		syscallarg(void *) asa;
1555 		syscallarg(int *) alen;
1556 	} */
1557 	int error;
1558 
1559 	if ((error = sys_getsockname(l, (const void *)uap, retval)) != 0)
1560 		return (error);
1561 
1562 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1563 		return (error);
1564 
1565 	return (0);
1566 }
1567 
1568 int
linux_sys_getpeername(struct lwp * l,const struct linux_sys_getpeername_args * uap,register_t * retval)1569 linux_sys_getpeername(struct lwp *l, const struct linux_sys_getpeername_args *uap, register_t *retval)
1570 {
1571 	/* {
1572 		syscallarg(int) fdes;
1573 		syscallarg(void *) asa;
1574 		syscallarg(int *) alen;
1575 	} */
1576 	int error;
1577 
1578 	if ((error = sys_getpeername(l, (const void *)uap, retval)) != 0)
1579 		return (error);
1580 
1581 	if ((error = linux_sa_put((struct osockaddr *)SCARG(uap, asa))))
1582 		return (error);
1583 
1584 	return (0);
1585 }
1586 
1587 /*
1588  * Copy the osockaddr structure pointed to by name to sb, adjust
1589  * family and convert to sockaddr.
1590  */
1591 static int
linux_get_sa(struct lwp * l,int s,struct sockaddr_big * sb,const struct osockaddr * name,socklen_t namelen)1592 linux_get_sa(struct lwp *l, int s, struct sockaddr_big *sb,
1593     const struct osockaddr *name, socklen_t namelen)
1594 {
1595 	int error, bdom;
1596 
1597 	if (namelen > UCHAR_MAX ||
1598 	    namelen <= offsetof(struct sockaddr_big, sb_data))
1599 		return EINVAL;
1600 
1601 	error = copyin(name, sb, namelen);
1602 	if (error)
1603 		return error;
1604 
1605 	bdom = linux_to_bsd_domain(sb->sb_family);
1606 	if (bdom == -1)
1607 		return EINVAL;
1608 
1609 	/*
1610 	 * If the family is unspecified, use address family of the socket.
1611 	 * This avoid triggering strict family checks in netinet/in_pcb.c et.al.
1612 	 */
1613 	if (bdom == AF_UNSPEC) {
1614 		struct socket *so;
1615 
1616 		/* fd_getsock() will use the descriptor for us */
1617 		if ((error = fd_getsock(s, &so)) != 0)
1618 			return error;
1619 
1620 		bdom = so->so_proto->pr_domain->dom_family;
1621 		fd_putfile(s);
1622 	}
1623 
1624 	/*
1625 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
1626 	 * which lacks the scope id compared with RFC2553 one. If we detect
1627 	 * the situation, reject the address and write a message to system log.
1628 	 *
1629 	 * Still accept addresses for which the scope id is not used.
1630 	 */
1631 	if (bdom == AF_INET6 &&
1632 	    namelen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
1633 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sb;
1634 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) &&
1635 		    (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) ||
1636 		     IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) ||
1637 		     IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) ||
1638 		     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
1639 		     IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
1640 			struct proc *p = l->l_proc;
1641 			int uid = l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1;
1642 
1643 			log(LOG_DEBUG,
1644 			    "pid %d (%s), uid %d: obsolete pre-RFC2553 "
1645 			    "sockaddr_in6 rejected",
1646 			    p->p_pid, p->p_comm, uid);
1647 			return EINVAL;
1648 		}
1649 		namelen = sizeof(struct sockaddr_in6);
1650 		sin6->sin6_scope_id = 0;
1651 	}
1652 
1653 	/*
1654 	 * Linux is less strict than NetBSD and permits namelen to be larger
1655 	 * than valid struct sockaddr_in*.  If this is the case, truncate
1656 	 * the value to the correct size, so that NetBSD networking does not
1657 	 * return an error.
1658 	 */
1659 	switch (bdom) {
1660 	case AF_INET:
1661 		namelen = MIN(namelen, sizeof(struct sockaddr_in));
1662 		break;
1663 	case AF_INET6:
1664 		namelen = MIN(namelen, sizeof(struct sockaddr_in6));
1665 		break;
1666 	}
1667 
1668 	sb->sb_family = bdom;
1669 	sb->sb_len = namelen;
1670 	ktrkuser("mbsoname", sb, namelen);
1671 	return 0;
1672 }
1673 
1674 static int
linux_sa_put(struct osockaddr * osa)1675 linux_sa_put(struct osockaddr *osa)
1676 {
1677 	struct sockaddr sa;
1678 	struct osockaddr *kosa;
1679 	int error, bdom, len;
1680 
1681 	/*
1682 	 * Only read/write the sockaddr family and length part, the rest is
1683 	 * not changed.
1684 	 */
1685 	len = sizeof(sa.sa_len) + sizeof(sa.sa_family);
1686 
1687 	error = copyin(osa, &sa, len);
1688 	if (error)
1689 		return (error);
1690 
1691 	bdom = bsd_to_linux_domain(sa.sa_family);
1692 	if (bdom == -1)
1693 		return (EINVAL);
1694 
1695 	/* Note: we convert from sockaddr to osockaddr here, too */
1696 	kosa = (struct osockaddr *) &sa;
1697 	kosa->sa_family = bdom;
1698 	error = copyout(kosa, osa, len);
1699 	if (error)
1700 		return (error);
1701 
1702 	return (0);
1703 }
1704 
1705 #if !defined(__aarch64__) && !defined(__amd64__)
1706 int
linux_sys_recv(struct lwp * l,const struct linux_sys_recv_args * uap,register_t * retval)1707 linux_sys_recv(struct lwp *l, const struct linux_sys_recv_args *uap, register_t *retval)
1708 {
1709 	/* {
1710 		syscallarg(int) s;
1711 		syscallarg(void *) buf;
1712 		syscallarg(int) len;
1713 		syscallarg(int) flags;
1714 	} */
1715 	struct sys_recvfrom_args bra;
1716 
1717 
1718 	SCARG(&bra, s) = SCARG(uap, s);
1719 	SCARG(&bra, buf) = SCARG(uap, buf);
1720 	SCARG(&bra, len) = (size_t) SCARG(uap, len);
1721 	SCARG(&bra, flags) = SCARG(uap, flags);
1722 	SCARG(&bra, from) = NULL;
1723 	SCARG(&bra, fromlenaddr) = NULL;
1724 
1725 	return (sys_recvfrom(l, &bra, retval));
1726 }
1727 
1728 int
linux_sys_send(struct lwp * l,const struct linux_sys_send_args * uap,register_t * retval)1729 linux_sys_send(struct lwp *l, const struct linux_sys_send_args *uap, register_t *retval)
1730 {
1731 	/* {
1732 		syscallarg(int) s;
1733 		syscallarg(void *) buf;
1734 		syscallarg(int) len;
1735 		syscallarg(int) flags;
1736 	} */
1737 	struct sys_sendto_args bsa;
1738 
1739 	SCARG(&bsa, s)		= SCARG(uap, s);
1740 	SCARG(&bsa, buf)	= SCARG(uap, buf);
1741 	SCARG(&bsa, len)	= SCARG(uap, len);
1742 	SCARG(&bsa, flags)	= SCARG(uap, flags);
1743 	SCARG(&bsa, to)		= NULL;
1744 	SCARG(&bsa, tolen)	= 0;
1745 
1746 	return (sys_sendto(l, &bsa, retval));
1747 }
1748 #endif
1749 
1750 int
linux_sys_accept(struct lwp * l,const struct linux_sys_accept_args * uap,register_t * retval)1751 linux_sys_accept(struct lwp *l, const struct linux_sys_accept_args *uap, register_t *retval)
1752 {
1753 	/* {
1754 		syscallarg(int) s;
1755 		syscallarg(struct osockaddr *) name;
1756 		syscallarg(int *) anamelen;
1757 	} */
1758 	int error;
1759 	struct sys_accept_args baa;
1760 
1761 	SCARG(&baa, s)		= SCARG(uap, s);
1762 	SCARG(&baa, name)	= (struct sockaddr *) SCARG(uap, name);
1763 	SCARG(&baa, anamelen)	= (unsigned int *) SCARG(uap, anamelen);
1764 
1765 	if ((error = sys_accept(l, &baa, retval)))
1766 		return (error);
1767 
1768 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1769 		return (error);
1770 
1771 	return (0);
1772 }
1773 
1774 int
linux_sys_accept4(struct lwp * l,const struct linux_sys_accept4_args * uap,register_t * retval)1775 linux_sys_accept4(struct lwp *l, const struct linux_sys_accept4_args *uap, register_t *retval)
1776 {
1777 	/* {
1778 		syscallarg(int) s;
1779 		syscallarg(struct osockaddr *) name;
1780 		syscallarg(int *) anamelen;
1781 		syscallarg(int) flags;
1782 	} */
1783 	int error, flags;
1784 	struct sockaddr_big name;
1785 
1786 	if ((flags = linux_to_bsd_type(SCARG(uap, flags))) == -1)
1787 		return EINVAL;
1788 
1789 	name.sb_len = UCHAR_MAX;
1790 	error = do_sys_accept(l, SCARG(uap, s), (struct sockaddr *)&name,
1791 	    retval, NULL, flags, 0);
1792 	if (error != 0)
1793 		return error;
1794 
1795 	error = copyout_sockname_sb((struct sockaddr *)SCARG(uap, name),
1796 	    SCARG(uap, anamelen), MSG_LENUSRSPACE, &name);
1797 	if (error != 0) {
1798 		int fd = (int)*retval;
1799 		if (fd_getfile(fd) != NULL)
1800 			(void)fd_close(fd);
1801 		return error;
1802 	}
1803 	if (SCARG(uap, name) && (error = linux_sa_put(SCARG(uap, name))))
1804 		return error;
1805 
1806 	return 0;
1807 }
1808 
1809 int
linux_sys_sendmmsg(struct lwp * l,const struct linux_sys_sendmmsg_args * uap,register_t * retval)1810 linux_sys_sendmmsg(struct lwp *l, const struct linux_sys_sendmmsg_args *uap,
1811     register_t *retval)
1812 {
1813 	/* {
1814 		syscallarg(int) s;
1815 		syscallarg(struct linux_mmsghdr *) msgvec;
1816 		syscallarg(unsigned int) vlen;
1817 		syscallarg(unsigned int) flags;
1818 	} */
1819 	struct linux_mmsghdr lmsg;
1820 	struct mmsghdr bmsg;
1821 	struct socket *so;
1822 	file_t *fp;
1823 	struct msghdr *msg = &bmsg.msg_hdr;
1824 	int error, s;
1825 	unsigned int vlen, flags, dg;
1826 
1827 	if ((flags = linux_to_bsd_msg_flags(SCARG(uap, flags))) == -1)
1828 		return EINVAL;
1829 
1830 	flags = (flags & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1831 
1832 	s = SCARG(uap, s);
1833 	if ((error = fd_getsock1(s, &so, &fp)) != 0)
1834 		return error;
1835 
1836 	vlen = SCARG(uap, vlen);
1837 	if (vlen > 1024)
1838 		vlen = 1024;
1839 
1840 	for (dg = 0; dg < vlen;) {
1841 		error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg));
1842 		if (error)
1843 			break;
1844 		linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr);
1845 
1846 		msg->msg_flags = flags;
1847 
1848 		error = do_sys_sendmsg_so(l, s, so, fp, msg, flags, retval);
1849 		if (error)
1850 			break;
1851 
1852 		ktrkuser("msghdr", msg, sizeof *msg);
1853 		lmsg.msg_len = *retval;
1854 		error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg));
1855 		if (error)
1856 			break;
1857 		dg++;
1858 
1859 	}
1860 
1861 	*retval = dg;
1862 
1863 	fd_putfile(s);
1864 
1865 	/*
1866 	 * If we succeeded at least once, return 0.
1867 	 */
1868 	if (dg)
1869 		return 0;
1870 	return error;
1871 }
1872 
1873 int
linux_sys_recvmmsg(struct lwp * l,const struct linux_sys_recvmmsg_args * uap,register_t * retval)1874 linux_sys_recvmmsg(struct lwp *l, const struct linux_sys_recvmmsg_args *uap,
1875     register_t *retval)
1876 {
1877 	/* {
1878 		syscallarg(int) s;
1879 		syscallarg(struct linux_mmsghdr *) msgvec;
1880 		syscallarg(unsigned int) vlen;
1881 		syscallarg(unsigned int) flags;
1882 		syscallarg(struct linux_timespec *) timeout;
1883 	} */
1884 	struct linux_mmsghdr lmsg;
1885 	struct mmsghdr bmsg;
1886 	struct socket *so;
1887 	struct msghdr *msg = &bmsg.msg_hdr;
1888 	int error, s;
1889 	struct mbuf *from, *control;
1890 	struct timespec ts = {0}, now;
1891 	struct linux_timespec lts;
1892 	unsigned int vlen, flags, dg;
1893 
1894 	if (SCARG(uap, timeout)) {
1895 		error = copyin(SCARG(uap, timeout), &lts, sizeof(lts));
1896 			return error;
1897 		ts.tv_sec = lts.tv_sec;
1898 		ts.tv_nsec = lts.tv_nsec;
1899 		getnanotime(&now);
1900 		timespecadd(&now, &ts, &ts);
1901 	}
1902 
1903 	s = SCARG(uap, s);
1904 	if ((error = fd_getsock(s, &so)) != 0)
1905 		return error;
1906 
1907 	/*
1908 	 * If so->so_rerror holds a deferred error return it now.
1909 	 */
1910 	if (so->so_rerror) {
1911 		error = so->so_rerror;
1912 		so->so_rerror = 0;
1913 		fd_putfile(s);
1914 		return error;
1915 	}
1916 
1917 	vlen = SCARG(uap, vlen);
1918 	if (vlen > 1024)
1919 		vlen = 1024;
1920 
1921 	from = NULL;
1922 	flags = (SCARG(uap, flags) & MSG_USERFLAGS) | MSG_IOVUSRSPACE;
1923 
1924 	for (dg = 0; dg < vlen;) {
1925 		error = copyin(SCARG(uap, msgvec) + dg, &lmsg, sizeof(lmsg));
1926 		if (error)
1927 			break;
1928 		linux_to_bsd_msghdr(&lmsg.msg_hdr, &bmsg.msg_hdr);
1929 		msg->msg_flags = flags & ~MSG_WAITFORONE;
1930 
1931 		if (from != NULL) {
1932 			m_free(from);
1933 			from = NULL;
1934 		}
1935 
1936 		error = do_sys_recvmsg_so(l, s, so, msg, &from,
1937 		    msg->msg_control != NULL ? &control : NULL, retval);
1938 		if (error) {
1939 			if (error == EAGAIN && dg > 0)
1940 				error = 0;
1941 			break;
1942 		}
1943 
1944 		if (msg->msg_control != NULL)
1945 			error = linux_copyout_msg_control(l, msg, control);
1946 		if (error)
1947 			break;
1948 
1949 		if (from != NULL) {
1950 			mtod(from, struct osockaddr *)->sa_family =
1951 			    bsd_to_linux_domain(mtod(from,
1952 			    struct sockaddr *)->sa_family);
1953 			error = copyout_sockname(msg->msg_name,
1954 			    &msg->msg_namelen, 0, from);
1955 			if (error)
1956 				break;
1957 		}
1958 
1959 
1960 		lmsg.msg_len = *retval;
1961 		ktrkuser("msghdr", msg, sizeof(*msg));
1962 		bsd_to_linux_msghdr(msg, &lmsg.msg_hdr);
1963 		error = copyout(&lmsg, SCARG(uap, msgvec) + dg, sizeof(lmsg));
1964 		if (error)
1965 			break;
1966 
1967 		dg++;
1968 		if (msg->msg_flags & MSG_OOB)
1969 			break;
1970 
1971 		if (SCARG(uap, timeout)) {
1972 			getnanotime(&now);
1973 			timespecsub(&now, &ts, &now);
1974 			if (now.tv_sec > 0)
1975 				break;
1976 		}
1977 
1978 		if (flags & MSG_WAITFORONE)
1979 			flags |= MSG_DONTWAIT;
1980 
1981 	}
1982 
1983 	if (from != NULL)
1984 		m_free(from);
1985 
1986 	*retval = dg;
1987 
1988 	/*
1989 	 * If we succeeded at least once, return 0, hopefully so->so_rerror
1990 	 * will catch it next time.
1991 	 */
1992 	if (error && dg > 0) {
1993 		so->so_rerror = error;
1994 		error = 0;
1995 	}
1996 
1997 	fd_putfile(s);
1998 
1999 	return error;
2000 }
2001