xref: /netbsd-src/external/bsd/ntp/dist/ntpd/ntp_io.c (revision a5847cc334d9a7029f6352b847e9e8d71a0f9e0c)
1 /*	$NetBSD: ntp_io.c,v 1.6 2011/09/14 16:18:29 christos Exp $	*/
2 
3 /*
4  * ntp_io.c - input/output routines for ntpd.	The socket-opening code
5  *		   was shamelessly stolen from ntpd.
6  */
7 
8 #ifdef HAVE_CONFIG_H
9 # include <config.h>
10 #endif
11 
12 #include <stdio.h>
13 #include <signal.h>
14 #ifdef HAVE_SYS_PARAM_H
15 # include <sys/param.h>
16 #endif
17 #ifdef HAVE_SYS_IOCTL_H
18 # include <sys/ioctl.h>
19 #endif
20 #ifdef HAVE_SYS_SOCKIO_H	/* UXPV: SIOC* #defines (Frank Vance <fvance@waii.com>) */
21 # include <sys/sockio.h>
22 #endif
23 #ifdef HAVE_SYS_UIO_H
24 # include <sys/uio.h>
25 #endif
26 
27 #include "ntp_machine.h"
28 #include "ntpd.h"
29 #include "ntp_io.h"
30 #include "iosignal.h"
31 #include "ntp_lists.h"
32 #include "ntp_refclock.h"
33 #include "ntp_stdlib.h"
34 #include "ntp_request.h"
35 #include "ntp.h"
36 #include "ntp_unixtime.h"
37 #include "ntp_assert.h"
38 #include "ntpd-opts.h"
39 
40 /* Don't include ISC's version of IPv6 variables and structures */
41 #define ISC_IPV6_H 1
42 #include <isc/mem.h>
43 #include <isc/interfaceiter.h>
44 #include <isc/netaddr.h>
45 #include <isc/result.h>
46 #include <isc/sockaddr.h>
47 
48 #ifdef SIM
49 #include "ntpsim.h"
50 #endif
51 
52 #ifdef HAS_ROUTING_SOCKET
53 # include <net/route.h>
54 # ifdef HAVE_RTNETLINK
55 #  include <linux/rtnetlink.h>
56 # endif
57 #endif
58 
59 
60 /*
61  * setsockopt does not always have the same arg declaration
62  * across all platforms. If it's not defined we make it empty
63  */
64 
65 #ifndef SETSOCKOPT_ARG_CAST
66 #define SETSOCKOPT_ARG_CAST
67 #endif
68 
69 extern int listen_to_virtual_ips;
70 
71 /*
72  * NIC rule entry
73  */
74 typedef struct nic_rule_tag nic_rule;
75 
76 struct nic_rule_tag {
77 	nic_rule *	next;
78 	nic_rule_action	action;
79 	nic_rule_match	match_type;
80 	char *		if_name;
81 	isc_netaddr_t	netaddr;
82 	int		prefixlen;
83 };
84 
85 /*
86  * NIC rule listhead.  Entries are added at the head so that the first
87  * match in the list is the last matching rule specified.
88  */
89 nic_rule *nic_rule_list;
90 
91 
92 #if defined(SO_TIMESTAMP) && defined(SCM_TIMESTAMP)
93 #if defined(CMSG_FIRSTHDR)
94 #define HAVE_TIMESTAMP
95 #define USE_TIMESTAMP_CMSG
96 #ifndef TIMESTAMP_CTLMSGBUF_SIZE
97 #define TIMESTAMP_CTLMSGBUF_SIZE 1536 /* moderate default */
98 #endif
99 #else
100 /* fill in for old/other timestamp interfaces */
101 #endif
102 #endif
103 
104 #if defined(SYS_WINNT)
105 #include <transmitbuff.h>
106 #include <isc/win32os.h>
107 /*
108  * Windows C runtime ioctl() can't deal properly with sockets,
109  * map to ioctlsocket for this source file.
110  */
111 #define ioctl(fd, opt, val)  ioctlsocket((fd), (opt), (u_long *)(val))
112 #endif  /* SYS_WINNT */
113 
114 /*
115  * We do asynchronous input using the SIGIO facility.  A number of
116  * recvbuf buffers are preallocated for input.	In the signal
117  * handler we poll to see which sockets are ready and read the
118  * packets from them into the recvbuf's along with a time stamp and
119  * an indication of the source host and the interface it was received
120  * through.  This allows us to get as accurate receive time stamps
121  * as possible independent of other processing going on.
122  *
123  * We watch the number of recvbufs available to the signal handler
124  * and allocate more when this number drops below the low water
125  * mark.  If the signal handler should run out of buffers in the
126  * interim it will drop incoming frames, the idea being that it is
127  * better to drop a packet than to be inaccurate.
128  */
129 
130 
131 /*
132  * Other statistics of possible interest
133  */
134 volatile u_long packets_dropped;	/* total number of packets dropped on reception */
135 volatile u_long packets_ignored;	/* packets received on wild card interface */
136 volatile u_long packets_received;	/* total number of packets received */
137 	 u_long packets_sent;		/* total number of packets sent */
138 	 u_long packets_notsent;	/* total number of packets which couldn't be sent */
139 
140 volatile u_long handler_calls;	/* number of calls to interrupt handler */
141 volatile u_long handler_pkts;	/* number of pkts received by handler */
142 u_long io_timereset;		/* time counters were reset */
143 
144 /*
145  * Interface stuff
146  */
147 struct interface *any_interface;	/* default ipv4 interface */
148 struct interface *any6_interface;	/* default ipv6 interface */
149 struct interface *loopback_interface;	/* loopback ipv4 interface */
150 
151 isc_boolean_t broadcast_client_enabled;	/* is broadcast client enabled */
152 int ninterfaces;			/* Total number of interfaces */
153 
154 int disable_dynamic_updates;		/* scan interfaces once only */
155 
156 #ifdef REFCLOCK
157 /*
158  * Refclock stuff.	We keep a chain of structures with data concerning
159  * the guys we are doing I/O for.
160  */
161 static	struct refclockio *refio;
162 #endif /* REFCLOCK */
163 
164 #if defined(HAVE_IPTOS_SUPPORT)
165 /* set IP_TOS to minimize packet delay */
166 # if defined(IPTOS_PREC_INTERNETCONTROL)
167 	unsigned int qos = IPTOS_PREC_INTERNETCONTROL;
168 # else
169 	 unsigned int qos = IPTOS_LOWDELAY;
170 # endif
171 #endif
172 
173 /*
174  * File descriptor masks etc. for call to select
175  * Not needed for I/O Completion Ports
176  */
177 fd_set activefds;
178 int maxactivefd;
179 /*
180  * bit alternating value to detect verified interfaces during an update cycle
181  */
182 static  u_short		sys_interphase = 0;
183 
184 static struct interface *new_interface	(struct interface *);
185 static void		add_interface	(struct interface *);
186 static int		update_interfaces(u_short, interface_receiver_t, void *);
187 static void		remove_interface(struct interface *);
188 static struct interface *create_interface(u_short, struct interface *);
189 
190 static int	move_fd			(SOCKET);
191 static int	is_wildcard_addr	(sockaddr_u *);
192 static int	is_wildcard_netaddr	(const isc_netaddr_t *);
193 
194 /*
195  * Multicast functions
196  */
197 static	isc_boolean_t	addr_ismulticast	(sockaddr_u *);
198 /*
199  * Not all platforms support multicast
200  */
201 #ifdef MCAST
202 static	isc_boolean_t	socket_multicast_enable	(struct interface *, int, sockaddr_u *);
203 static	isc_boolean_t	socket_multicast_disable(struct interface *, sockaddr_u *);
204 #endif
205 
206 #ifdef DEBUG
207 static void interface_dump	(struct interface *);
208 static void sockaddr_dump	(sockaddr_u *psau);
209 static void print_interface	(struct interface *, const char *, const char *);
210 #define DPRINT_INTERFACE(level, args) do { if (debug >= (level)) { print_interface args; } } while (0)
211 #else
212 #define DPRINT_INTERFACE(level, args) do {} while (0)
213 #endif
214 
215 typedef struct vsock vsock_t;
216 enum desc_type { FD_TYPE_SOCKET, FD_TYPE_FILE };
217 
218 struct vsock {
219 	vsock_t	*	link;
220 	SOCKET		fd;
221 	enum desc_type	type;
222 };
223 
224 vsock_t	*fd_list;
225 
226 #if !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET)
227 /*
228  * async notification processing (e. g. routing sockets)
229  */
230 /*
231  * support for receiving data on fd that is not a refclock or a socket
232  * like e. g. routing sockets
233  */
234 struct asyncio_reader {
235 	struct asyncio_reader *link;		    /* the list this is being kept in */
236 	SOCKET fd;				    /* fd to be read */
237 	void  *data;				    /* possibly local data */
238 	void (*receiver)(struct asyncio_reader *);  /* input handler */
239 };
240 
241 struct asyncio_reader *asyncio_reader_list;
242 
243 static void delete_asyncio_reader (struct asyncio_reader *);
244 static struct asyncio_reader *new_asyncio_reader (void);
245 static void add_asyncio_reader (struct asyncio_reader *, enum desc_type);
246 static void remove_asyncio_reader (struct asyncio_reader *);
247 
248 #endif /* !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET) */
249 
250 static void init_async_notifications (void);
251 
252 static	int create_sockets	(u_short);
253 static	SOCKET	open_socket	(sockaddr_u *, int, int, struct interface *);
254 static	char *	fdbits		(int, fd_set *);
255 static	void	set_reuseaddr	(int);
256 static	isc_boolean_t	socket_broadcast_enable	 (struct interface *, SOCKET, sockaddr_u *);
257 static	isc_boolean_t	socket_broadcast_disable (struct interface *, sockaddr_u *);
258 
259 typedef struct remaddr remaddr_t;
260 
261 struct remaddr {
262 	remaddr_t *		link;
263 	sockaddr_u		addr;
264 	struct interface *	interface;
265 };
266 
267 remaddr_t *		remoteaddr_list;
268 
269 struct interface *	inter_list;
270 
271 static struct interface *wildipv4 = NULL;
272 static struct interface *wildipv6 = NULL;
273 
274 static void		add_fd_to_list		(SOCKET,
275 						 enum desc_type);
276 static struct interface *find_addr_in_list	(sockaddr_u *);
277 static struct interface *find_samenet_addr_in_list(sockaddr_u *);
278 static struct interface *find_flagged_addr_in_list(sockaddr_u *, int);
279 static void		delete_addr_from_list	(sockaddr_u *);
280 static void		delete_interface_from_list(struct interface *);
281 static void		close_and_delete_fd_from_list(SOCKET);
282 static void		add_addr_to_list	(sockaddr_u *,
283 						 struct interface *);
284 static void		create_wildcards	(u_short);
285 #ifdef DEBUG
286 static const char *	action_text		(nic_rule_action);
287 #endif
288 static nic_rule_action	interface_action(char *, isc_netaddr_t *,
289 					 isc_uint32_t);
290 static void		convert_isc_if	(isc_interface_t *,
291 					 struct interface *, u_short);
292 static struct interface *getinterface	(sockaddr_u *, int);
293 static struct interface *getsamenetinterface	(sockaddr_u *, int);
294 static struct interface *findlocalinterface	(sockaddr_u *, int, int);
295 static struct interface *findlocalcastinterface	(sockaddr_u *);
296 
297 /*
298  * Routines to read the ntp packets
299  */
300 #if !defined(HAVE_IO_COMPLETION_PORT)
301 static inline int     read_network_packet	(SOCKET, struct interface *, l_fp);
302 static inline int     read_refclock_packet	(SOCKET, struct refclockio *, l_fp);
303 #endif
304 
305 
306 #ifdef SYS_WINNT
307 /*
308  * Windows 2000 systems incorrectly cause UDP sockets using WASRecvFrom
309  * to not work correctly, returning a WSACONNRESET error when a WSASendTo
310  * fails with an "ICMP port unreachable" response and preventing the
311  * socket from using the WSARecvFrom in subsequent operations.
312  * The function below fixes this, but requires that Windows 2000
313  * Service Pack 2 or later be installed on the system.  NT 4.0
314  * systems are not affected by this and work correctly.
315  * See Microsoft Knowledge Base Article Q263823 for details of this.
316  */
317 void
318 connection_reset_fix(
319 	SOCKET		fd,
320 	sockaddr_u *	addr
321 	)
322 {
323 	DWORD dw;
324 	BOOL  bNewBehavior = FALSE;
325 	DWORD status;
326 
327 	/*
328 	 * disable bad behavior using IOCTL: SIO_UDP_CONNRESET
329 	 * NT 4.0 has no problem
330 	 */
331 	if (isc_win32os_majorversion() >= 5) {
332 		status = WSAIoctl(fd, SIO_UDP_CONNRESET, &bNewBehavior,
333 				  sizeof(bNewBehavior), NULL, 0,
334 				  &dw, NULL, NULL);
335 		if (SOCKET_ERROR == status)
336 			msyslog(LOG_ERR,
337 				"connection_reset_fix() failed for address %s: %m",
338 				stoa(addr));
339 	}
340 }
341 #endif
342 
343 /*
344  * on Unix systems the stdio library typically
345  * makes use of file descriptors in the lower
346  * integer range.  stdio usually will make use
347  * of the file descriptors in the range of
348  * [0..FOPEN_MAX)
349  * in order to keep this range clean, for socket
350  * file descriptors we attempt to move them above
351  * FOPEN_MAX. This is not as easy as it sounds as
352  * FOPEN_MAX changes from implementation to implementation
353  * and may exceed to current file decriptor limits.
354  * We are using following strategy:
355  * - keep a current socket fd boundary initialized with
356  *   max(0, min(getdtablesize() - FD_CHUNK, FOPEN_MAX))
357  * - attempt to move the descriptor to the boundary or
358  *   above.
359  *   - if that fails and boundary > 0 set boundary
360  *     to min(0, socket_fd_boundary - FD_CHUNK)
361  *     -> retry
362  *     if failure and boundary == 0 return old fd
363  *   - on success close old fd return new fd
364  *
365  * effects:
366  *   - fds will be moved above the socket fd boundary
367  *     if at all possible.
368  *   - the socket boundary will be reduced until
369  *     allocation is possible or 0 is reached - at this
370  *     point the algrithm will be disabled
371  */
372 static int
373 move_fd(
374 	SOCKET fd
375 	)
376 {
377 #if !defined(SYS_WINNT) && defined(F_DUPFD)
378 #ifndef FD_CHUNK
379 #define FD_CHUNK	10
380 #endif
381 /*
382  * number of fds we would like to have for
383  * stdio FILE* available.
384  * we can pick a "low" number as our use of
385  * FILE* is limited to log files and temporarily
386  * to data and config files. Except for log files
387  * we don't keep the other FILE* open beyond the
388  * scope of the function that opened it.
389  */
390 #ifndef FD_PREFERRED_SOCKBOUNDARY
391 #define FD_PREFERRED_SOCKBOUNDARY 48
392 #endif
393 
394 #ifndef HAVE_GETDTABLESIZE
395 /*
396  * if we have no idea about the max fd value set up things
397  * so we will start at FOPEN_MAX
398  */
399 #define getdtablesize() (FOPEN_MAX+FD_CHUNK)
400 #endif
401 
402 #ifndef FOPEN_MAX
403 #define FOPEN_MAX	20	/* assume that for the lack of anything better */
404 #endif
405 	static SOCKET socket_boundary = -1;
406 	SOCKET newfd;
407 
408 	NTP_REQUIRE((int)fd >= 0);
409 
410 	/*
411 	 * check whether boundary has be set up
412 	 * already
413 	 */
414 	if (socket_boundary == -1) {
415 		socket_boundary = max(0, min(getdtablesize() - FD_CHUNK,
416 					     min(FOPEN_MAX, FD_PREFERRED_SOCKBOUNDARY)));
417 #ifdef DEBUG
418 		msyslog(LOG_DEBUG,
419 			"ntp_io: estimated max descriptors: %d, initial socket boundary: %d",
420 			getdtablesize(), socket_boundary);
421 #endif
422 	}
423 
424 	/*
425 	 * Leave a space for stdio to work in. potentially moving the
426 	 * socket_boundary lower until allocation succeeds.
427 	 */
428 	do {
429 		if (fd >= 0 && fd < socket_boundary) {
430 			/* inside reserved range: attempt to move fd */
431 			newfd = fcntl(fd, F_DUPFD, socket_boundary);
432 
433 			if (newfd != -1) {
434 				/* success: drop the old one - return the new one */
435 				close(fd);
436 				return newfd;
437 			}
438 		} else {
439 			/* outside reserved range: no work - return the original one */
440 			return fd;
441 		}
442 		socket_boundary = max(0, socket_boundary - FD_CHUNK);
443 #ifdef DEBUG
444 		msyslog(LOG_DEBUG,
445 			"ntp_io: selecting new socket boundary: %d",
446 			socket_boundary);
447 #endif
448 	} while (socket_boundary > 0);
449 #else
450 	NTP_REQUIRE((int)fd >= 0);
451 #endif /* !defined(SYS_WINNT) && defined(F_DUPFD) */
452 	return fd;
453 }
454 
455 #ifdef DEBUG_TIMING
456 /*
457  * collect timing information for various processing
458  * paths. currently we only pass then on to the file
459  * for later processing. this could also do histogram
460  * based analysis in other to reduce the load (and skew)
461  * dur to the file output
462  */
463 void
464 collect_timing(struct recvbuf *rb, const char *tag, int count, l_fp *dts)
465 {
466 	char buf[256];
467 
468 	snprintf(buf, sizeof(buf), "%s %d %s %s",
469 		 (rb != NULL)
470 		     ? ((rb->dstadr != NULL)
471 			    ? stoa(&rb->recv_srcadr)
472 			    : "-REFCLOCK-")
473 		     : "-",
474 		 count, lfptoa(dts, 9), tag);
475 	record_timing_stats(buf);
476 }
477 #endif
478 
479 /*
480  * About dynamic interfaces, sockets, reception and more...
481  *
482  * the code solves following tasks:
483  *
484  *   - keep a current list of active interfaces in order
485  *     to bind to to the interface address on NTP_PORT so that
486  *     all wild and specific bindings for NTP_PORT are taken by ntpd
487  *     to avoid other daemons messing with the time or sockets.
488  *   - all interfaces keep a list of peers that are referencing
489  *     the interface in order to quickly re-assign the peers to
490  *     new interface in case an interface is deleted (=> gone from system or
491  *     down)
492  *   - have a preconfigured socket ready with the right local address
493  *     for transmission and reception
494  *   - have an address list for all destination addresses used within ntpd
495  *     to find the "right" preconfigured socket.
496  *   - facilitate updating the internal interface list with respect to
497  *     the current kernel state
498  *
499  * special issues:
500  *
501  *   - mapping of multicast addresses to the interface affected is not always
502  *     one to one - especially on hosts with multiple interfaces
503  *     the code here currently allocates a separate interface entry for those
504  *     multicast addresses
505  *     iff it is able to bind to a *new* socket with the multicast address (flags |= MCASTIF)
506  *     in case of failure the multicast address is bound to an existing interface.
507  *   - on some systems it is perfectly legal to assign the same address to
508  *     multiple interfaces. Therefore this code does not keep a list of interfaces
509  *     but a list of interfaces that represent a unique address as determined by the kernel
510  *     by the procedure in findlocalinterface. Thus it is perfectly legal to see only
511  *     one representative of a group of real interfaces if they share the same address.
512  *
513  * Frank Kardel 20050910
514  */
515 
516 /*
517  * init_io - initialize I/O data structures and call socket creation routine
518  */
519 void
520 init_io(void)
521 {
522 	/*
523 	 * Init buffer free list and stat counters
524 	 */
525 	init_recvbuff(RECV_INIT);
526 
527 #ifdef SYS_WINNT
528 	init_io_completion_port();
529 #endif /* SYS_WINNT */
530 
531 #if defined(HAVE_SIGNALED_IO)
532 	(void) set_signal();
533 #endif
534 }
535 
536 
537 /*
538  * io_open_sockets - call socket creation routine
539  */
540 void
541 io_open_sockets(void)
542 {
543 	static int already_opened;
544 
545 	if (already_opened || HAVE_OPT( SAVECONFIGQUIT ))
546 		return;
547 
548 	already_opened = 1;
549 
550 	/*
551 	 * Create the sockets
552 	 */
553 	BLOCKIO();
554 	create_sockets(NTP_PORT);
555 	UNBLOCKIO();
556 
557 	init_async_notifications();
558 
559 	DPRINTF(3, ("io_open_sockets: maxactivefd %d\n", maxactivefd));
560 }
561 
562 
563 #ifdef DEBUG
564 /*
565  * function to dump the contents of the interface structure
566  * for debugging use only.
567  */
568 void
569 interface_dump(struct interface *itf)
570 {
571 	printf("Dumping interface: %p\n", itf);
572 	printf("fd = %d\n", itf->fd);
573 	printf("bfd = %d\n", itf->bfd);
574 	printf("sin = %s,\n", stoa(&itf->sin));
575 	sockaddr_dump(&itf->sin);
576 	printf("bcast = %s,\n", stoa(&itf->bcast));
577 	sockaddr_dump(&itf->bcast);
578 	printf("mask = %s,\n", stoa(&itf->mask));
579 	sockaddr_dump(&itf->mask);
580 	printf("name = %s\n", itf->name);
581 	printf("flags = 0x%08x\n", itf->flags);
582 	printf("last_ttl = %d\n", itf->last_ttl);
583 	printf("addr_refid = %08x\n", itf->addr_refid);
584 	printf("num_mcast = %d\n", itf->num_mcast);
585 	printf("received = %ld\n", itf->received);
586 	printf("sent = %ld\n", itf->sent);
587 	printf("notsent = %ld\n", itf->notsent);
588 	printf("scopeid = %u\n", itf->scopeid);
589 	printf("peercnt = %u\n", itf->peercnt);
590 	printf("phase = %u\n", itf->phase);
591 }
592 
593 /*
594  * sockaddr_dump - hex dump the start of a sockaddr_u
595  */
596 static void
597 sockaddr_dump(sockaddr_u *psau)
598 {
599 	/* Limit the size of the sockaddr_storage hex dump */
600 	const int maxsize = min(32, sizeof(psau->sas));
601 	u_char *	cp;
602 	int		i;
603 
604 	cp = (u_char *)&psau->sas;
605 
606 	for(i = 0; i < maxsize; i++)
607 	{
608 		printf("%02x", *cp++);
609 		if (!((i + 1) % 4))
610 			printf(" ");
611 	}
612 	printf("\n");
613 }
614 
615 /*
616  * print_interface - helper to output debug information
617  */
618 static void
619 print_interface(struct interface *iface, const char *pfx, const char *sfx)
620 {
621 	printf("%sinterface #%d: fd=%d, bfd=%d, name=%s, flags=0x%x, scope=%d, sin=%s",
622 	       pfx,
623 	       iface->ifnum,
624 	       iface->fd,
625 	       iface->bfd,
626 	       iface->name,
627 	       iface->flags,
628 	       iface->scopeid,
629 	       stoa(&iface->sin));
630 	if (AF_INET == iface->family) {
631 		if (iface->flags & INT_BROADCAST)
632 			printf(", bcast=%s", stoa(&iface->bcast));
633 		printf(", mask=%s", stoa(&iface->mask));
634 	}
635 	printf(", %s:%s",
636 	       (iface->ignore_packets)
637 		   ? "Disabled"
638 		   : "Enabled",
639 	       sfx);
640 	if (debug > 4)	/* in-depth debugging only */
641 		interface_dump(iface);
642 }
643 #endif
644 
645 #if !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET)
646 /*
647  * create an asyncio_reader structure
648  */
649 static struct asyncio_reader *
650 new_asyncio_reader(void)
651 {
652 	struct asyncio_reader *reader;
653 
654 	reader = emalloc(sizeof(*reader));
655 
656 	memset(reader, 0, sizeof(*reader));
657 	reader->fd = INVALID_SOCKET;
658 	return reader;
659 }
660 
661 /*
662  * delete a reader
663  */
664 static void
665 delete_asyncio_reader(
666 	struct asyncio_reader *reader
667 	)
668 {
669 	free(reader);
670 }
671 
672 /*
673  * add asynchio_reader
674  */
675 static void
676 add_asyncio_reader(
677 	struct asyncio_reader *	reader,
678 	enum desc_type		type)
679 {
680 	LINK_SLIST(asyncio_reader_list, reader, link);
681 	add_fd_to_list(reader->fd, type);
682 }
683 
684 /*
685  * remove asynchio_reader
686  */
687 static void
688 remove_asyncio_reader(
689 	struct asyncio_reader *reader
690 	)
691 {
692 	struct asyncio_reader *unlinked;
693 
694 	UNLINK_SLIST(unlinked, asyncio_reader_list, reader, link,
695 	    struct asyncio_reader);
696 
697 	if (reader->fd != INVALID_SOCKET)
698 		close_and_delete_fd_from_list(reader->fd);
699 
700 	reader->fd = INVALID_SOCKET;
701 }
702 #endif /* !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET) */
703 
704 /*
705  * Code to tell if we have an IP address
706  * If we have then return the sockaddr structure
707  * and set the return value
708  * see the bind9/getaddresses.c for details
709  */
710 isc_boolean_t
711 is_ip_address(
712 	const char *	host,
713 	isc_netaddr_t *	addr
714 	)
715 {
716 	struct in_addr in4;
717 	struct in6_addr in6;
718 	char tmpbuf[128];
719 	char *pch;
720 
721 	NTP_REQUIRE(host != NULL);
722 	NTP_REQUIRE(addr != NULL);
723 
724 	/*
725 	 * Try IPv4, then IPv6.  In order to handle the extended format
726 	 * for IPv6 scoped addresses (address%scope_ID), we'll use a local
727 	 * working buffer of 128 bytes.  The length is an ad-hoc value, but
728 	 * should be enough for this purpose; the buffer can contain a string
729 	 * of at least 80 bytes for scope_ID in addition to any IPv6 numeric
730 	 * addresses (up to 46 bytes), the delimiter character and the
731 	 * terminating NULL character.
732 	 */
733 	if (inet_pton(AF_INET, host, &in4) == 1) {
734 		isc_netaddr_fromin(addr, &in4);
735 		return (ISC_TRUE);
736 	} else if (sizeof(tmpbuf) > strlen(host)) {
737 		if ('[' == host[0]) {
738 			strncpy(tmpbuf, &host[1], sizeof(tmpbuf));
739 			pch = strchr(tmpbuf, ']');
740 			if (pch != NULL)
741 				*pch = '\0';
742 		} else
743 			strncpy(tmpbuf, host, sizeof(tmpbuf));
744 		pch = strchr(tmpbuf, '%');
745 		if (pch != NULL)
746 			*pch = '\0';
747 
748 		if (inet_pton(AF_INET6, tmpbuf, &in6) == 1) {
749 			isc_netaddr_fromin6(addr, &in6);
750 			return (ISC_TRUE);
751 		}
752 	}
753 	/*
754 	 * If we got here it was not an IP address
755 	 */
756 	return (ISC_FALSE);
757 }
758 
759 
760 /*
761  * interface list enumerator - visitor pattern
762  */
763 void
764 interface_enumerate(
765 	interface_receiver_t	receiver,
766 	void *			data
767 	)
768 {
769 	interface_info_t ifi;
770 
771 	ifi.action = IFS_EXISTS;
772 
773 	for (ifi.interface = inter_list;
774 	     ifi.interface != NULL;
775 	     ifi.interface = ifi.interface->link)
776 		(*receiver)(data, &ifi);
777 }
778 
779 /*
780  * do standard initialization of interface structure
781  */
782 static void
783 init_interface(
784 	struct interface *iface
785 	)
786 {
787 	memset(iface, 0, sizeof(*iface));
788 	iface->fd = INVALID_SOCKET;
789 	iface->bfd = INVALID_SOCKET;
790 	iface->phase = sys_interphase;
791 }
792 
793 
794 /*
795  * create new interface structure initialize from
796  * template structure or via standard initialization
797  * function
798  */
799 static struct interface *
800 new_interface(
801 	struct interface *interface
802 	)
803 {
804 	static u_int		sys_ifnum = 0;
805 	struct interface *	iface;
806 
807 	iface = emalloc(sizeof(*iface));
808 
809 	if (NULL == interface)
810 		init_interface(iface);
811 	else				/* use the template */
812 		memcpy(iface, interface, sizeof(*iface));
813 
814 	/* count every new instance of an interface in the system */
815 	iface->ifnum = sys_ifnum++;
816 	iface->starttime = current_time;
817 
818 	return iface;
819 }
820 
821 
822 /*
823  * return interface storage into free memory pool
824  */
825 static inline void
826 delete_interface(
827 	struct interface *interface
828 	)
829 {
830 	free(interface);
831 }
832 
833 
834 /*
835  * link interface into list of known interfaces
836  */
837 static void
838 add_interface(
839 	struct interface *interface
840 	)
841 {
842 	/*
843 	 * Calculate the address hash
844 	 */
845 	interface->addr_refid = addr2refid(&interface->sin);
846 
847 	LINK_SLIST(inter_list, interface, link);
848 	ninterfaces++;
849 }
850 
851 
852 /*
853  * remove interface from known interface list and clean up
854  * associated resources
855  */
856 static void
857 remove_interface(
858 	struct interface *iface
859 	)
860 {
861 	struct interface *unlinked;
862 	sockaddr_u resmask;
863 
864 	UNLINK_SLIST(unlinked, inter_list, iface, link, struct
865 	    interface);
866 
867 	delete_interface_from_list(iface);
868 
869 	if (iface->fd != INVALID_SOCKET) {
870 		msyslog(LOG_INFO,
871 			"Deleting interface #%d %s, %s#%d, interface stats: received=%ld, sent=%ld, dropped=%ld, active_time=%ld secs",
872 			iface->ifnum,
873 			iface->name,
874 			stoa(&iface->sin),
875 			SRCPORT(&iface->sin),
876 			iface->received,
877 			iface->sent,
878 			iface->notsent,
879 			current_time - iface->starttime);
880 
881 		close_and_delete_fd_from_list(iface->fd);
882 	}
883 
884 	if (iface->bfd != INVALID_SOCKET) {
885 		msyslog(LOG_INFO,
886 			"Deleting broadcast address %s#%d from interface #%d %s",
887 			stoa(&iface->bcast),
888 			SRCPORT(&iface->bcast),
889 			iface->ifnum,
890 			iface->name);
891 
892 		close_and_delete_fd_from_list(iface->bfd);
893 	}
894 
895 	ninterfaces--;
896 	ntp_monclearinterface(iface);
897 
898 	/* remove restrict interface entry */
899 	SET_HOSTMASK(&resmask, AF(&iface->sin));
900 	hack_restrict(RESTRICT_REMOVEIF, &iface->sin, &resmask,
901 		      RESM_NTPONLY | RESM_INTERFACE, RES_IGNORE);
902 }
903 
904 
905 static void
906 list_if_listening(
907 	struct interface *	iface
908 	)
909 {
910 	msyslog(LOG_INFO, "%s on %d %s %s UDP %d",
911 		(iface->ignore_packets)
912 		    ? "Listen and drop"
913 		    : "Listen normally",
914 		iface->ifnum,
915 		iface->name,
916 		stoa(&iface->sin),
917 		SRCPORT(&iface->sin));
918 }
919 
920 
921 static void
922 create_wildcards(
923 	u_short	port
924 	)
925 {
926 	int			v4wild;
927 #ifdef INCLUDE_IPV6_SUPPORT
928 	int                     v6wild;
929 #endif
930 	sockaddr_u		wildaddr;
931 	isc_netaddr_t		wnaddr;
932 	nic_rule_action		action;
933 	struct interface *	wildif;
934 
935 	/*
936 	 * silence "potentially uninitialized" warnings from VC9
937 	 * failing to follow the logic.  Ideally action could remain
938 	 * uninitialized, and the memset be the first statement under
939 	 * the first if (v4wild).
940 	 */
941 	action = ACTION_LISTEN;
942 	memset(&wildaddr, 0, sizeof(wildaddr));
943 
944 	/*
945 	 * create pseudo-interface with wildcard IPv4 address
946 	 */
947 	v4wild = ipv4_works;
948 	if (v4wild) {
949 		/* set wildaddr to the v4 wildcard address 0.0.0.0 */
950 		AF(&wildaddr) = AF_INET;
951 		SET_ADDR4(&wildaddr, INADDR_ANY);
952 		SET_PORT(&wildaddr, port);
953 
954 		/* make an libisc-friendly copy */
955 		isc_netaddr_fromin(&wnaddr, &wildaddr.sa4.sin_addr);
956 
957 		/* check for interface/nic rules affecting the wildcard */
958 		action = interface_action(NULL, &wnaddr, 0);
959 		v4wild = (ACTION_IGNORE != action);
960 	}
961 	if (v4wild) {
962 		wildif = new_interface(NULL);
963 
964 		strncpy(wildif->name, "v4wildcard", sizeof(wildif->name));
965 		memcpy(&wildif->sin, &wildaddr, sizeof(wildif->sin));
966 		wildif->family = AF_INET;
967 		AF(&wildif->mask) = AF_INET;
968 		SET_ONESMASK(&wildif->mask);
969 
970 		wildif->flags = INT_BROADCAST | INT_UP | INT_WILDCARD;
971 		wildif->ignore_packets = (ACTION_DROP == action);
972 #if defined(MCAST)
973 		/*
974 		 * enable multicast reception on the broadcast socket
975 		 */
976 		AF(&wildif->bcast) = AF_INET;
977 		SET_ADDR4(&wildif->bcast, INADDR_ANY);
978 		SET_PORT(&wildif->bcast, port);
979 #endif /* MCAST */
980 		wildif->fd = open_socket(&wildif->sin, 0, 1, wildif);
981 
982 		if (wildif->fd != INVALID_SOCKET) {
983 			wildipv4 = wildif;
984 			any_interface = wildif;
985 
986 			add_addr_to_list(&wildif->sin, wildif);
987 			add_interface(wildif);
988 			list_if_listening(wildif);
989 		} else {
990 			msyslog(LOG_ERR,
991 				"unable to bind to wildcard address %s - another process may be running - EXITING",
992 				stoa(&wildif->sin));
993 			exit(1);
994 		}
995 		DPRINT_INTERFACE(2, (wildif, "created ", "\n"));
996 	}
997 
998 #ifdef INCLUDE_IPV6_SUPPORT
999 	/*
1000 	 * create pseudo-interface with wildcard IPv6 address
1001 	 */
1002 	v6wild = ipv6_works;
1003 	if (v6wild) {
1004 		/* set wildaddr to the v6 wildcard address :: */
1005 		memset(&wildaddr, 0, sizeof(wildaddr));
1006 		AF(&wildaddr) = AF_INET6;
1007 		SET_ADDR6N(&wildaddr, in6addr_any);
1008 		SET_PORT(&wildaddr, port);
1009 		SET_SCOPE(&wildaddr, 0);
1010 
1011 		/* make an libisc-friendly copy */
1012 		isc_netaddr_fromin(&wnaddr, &wildaddr.sa4.sin_addr);
1013 
1014 		/* check for interface/nic rules affecting the wildcard */
1015 		action = interface_action(NULL, &wnaddr, 0);
1016 		v6wild = (ACTION_IGNORE != action);
1017 	}
1018 	if (v6wild) {
1019 		wildif = new_interface(NULL);
1020 
1021 		strncpy(wildif->name, "v6wildcard", sizeof(wildif->name));
1022 		memcpy(&wildif->sin, &wildaddr, sizeof(wildif->sin));
1023 		wildif->family = AF_INET6;
1024 		AF(&wildif->mask) = AF_INET6;
1025 		SET_ONESMASK(&wildif->mask);
1026 
1027 		wildif->flags = INT_UP | INT_WILDCARD;
1028 		wildif->ignore_packets = (ACTION_DROP == action);
1029 
1030 		wildif->fd = open_socket(&wildif->sin, 0, 1, wildif);
1031 
1032 		if (wildif->fd != INVALID_SOCKET) {
1033 			wildipv6 = wildif;
1034 			any6_interface = wildif;
1035 			add_addr_to_list(&wildif->sin, wildif);
1036 			add_interface(wildif);
1037 			list_if_listening(wildif);
1038 		} else {
1039 			msyslog(LOG_ERR,
1040 				"unable to bind to wildcard address %s - another process may be running - EXITING",
1041 				stoa(&wildif->sin));
1042 			exit(1);
1043 		}
1044 		DPRINT_INTERFACE(2, (wildif, "created ", "\n"));
1045 	}
1046 #endif
1047 }
1048 
1049 
1050 /*
1051  * add_nic_rule() -- insert a rule entry at the head of nic_rule_list.
1052  */
1053 void
1054 add_nic_rule(
1055 	nic_rule_match	match_type,
1056 	const char *	if_name,	/* interface name or numeric address */
1057 	int		prefixlen,
1058 	nic_rule_action	action
1059 	)
1060 {
1061 	nic_rule *	rule;
1062 	isc_boolean_t	is_ip;
1063 
1064 	rule = emalloc(sizeof(*rule));
1065 	memset(rule, 0, sizeof(*rule));
1066 	rule->match_type = match_type;
1067 	rule->prefixlen = prefixlen;
1068 	rule->action = action;
1069 
1070 	if (MATCH_IFNAME == match_type) {
1071 		NTP_REQUIRE(NULL != if_name);
1072 		rule->if_name = estrdup(if_name);
1073 	} else if (MATCH_IFADDR == match_type) {
1074 		NTP_REQUIRE(NULL != if_name);
1075 		/* set rule->netaddr */
1076 		is_ip = is_ip_address(if_name, &rule->netaddr);
1077 		NTP_REQUIRE(is_ip);
1078 	} else
1079 		NTP_REQUIRE(NULL == if_name);
1080 
1081 	LINK_SLIST(nic_rule_list, rule, next);
1082 }
1083 
1084 
1085 #ifdef DEBUG
1086 static const char *
1087 action_text(
1088 	nic_rule_action	action
1089 	)
1090 {
1091 	const char *t;
1092 
1093 	switch (action) {
1094 
1095 	default:
1096 		t = "ERROR";	/* quiet uninit warning */
1097 		DPRINTF(1, ("fatal: unknown nic_rule_action %d\n",
1098 			    action));
1099 		NTP_ENSURE(0);
1100 		break;
1101 
1102 	case ACTION_LISTEN:
1103 		t = "listen";
1104 		break;
1105 
1106 	case ACTION_IGNORE:
1107 		t = "ignore";
1108 		break;
1109 
1110 	case ACTION_DROP:
1111 		t = "drop";
1112 		break;
1113 	}
1114 
1115 	return t;
1116 }
1117 #endif	/* DEBUG */
1118 
1119 
1120 static nic_rule_action
1121 interface_action(
1122 	char *		if_name,
1123 	isc_netaddr_t *	if_netaddr,
1124 	isc_uint32_t	if_flags
1125 	)
1126 {
1127 	nic_rule *rule;
1128 	int isloopback;
1129 	int iswildcard;
1130 
1131 	DPRINTF(4, ("interface_action: interface %s ",
1132 		    (if_name != NULL) ? if_name : "wildcard"));
1133 
1134 	iswildcard = is_wildcard_netaddr(if_netaddr);
1135 
1136 	/*
1137 	 * Always listen on 127.0.0.1 - required by ntp_intres
1138 	 */
1139 	if (if_flags & INTERFACE_F_LOOPBACK) {
1140 		isloopback = 1;
1141 		if (AF_INET == if_netaddr->family) {
1142 			DPRINTF(4, ("IPv4 loopback - listen\n"));
1143 			return ACTION_LISTEN;
1144 		}
1145 	} else
1146 		isloopback = 0;
1147 
1148 	/*
1149 	 * Find any matching NIC rule from --interface / -I or ntp.conf
1150 	 * interface/nic rules.
1151 	 */
1152 	for (rule = nic_rule_list; rule != NULL; rule = rule->next) {
1153 
1154 		switch (rule->match_type) {
1155 
1156 		case MATCH_ALL:
1157 			/* loopback and wildcard excluded from "all" */
1158 			if (isloopback || iswildcard)
1159 				break;
1160 			DPRINTF(4, ("nic all %s\n",
1161 			    action_text(rule->action)));
1162 			return rule->action;
1163 
1164 		case MATCH_IPV4:
1165 			if (AF_INET == if_netaddr->family) {
1166 				DPRINTF(4, ("nic ipv4 %s\n",
1167 				    action_text(rule->action)));
1168 				return rule->action;
1169 			}
1170 			break;
1171 
1172 		case MATCH_IPV6:
1173 			if (AF_INET6 == if_netaddr->family) {
1174 				DPRINTF(4, ("nic ipv6 %s\n",
1175 				    action_text(rule->action)));
1176 				return rule->action;
1177 			}
1178 			break;
1179 
1180 		case MATCH_WILDCARD:
1181 			if (iswildcard) {
1182 				DPRINTF(4, ("nic wildcard %s\n",
1183 				    action_text(rule->action)));
1184 				return rule->action;
1185 			}
1186 			break;
1187 
1188 		case MATCH_IFADDR:
1189 			if (rule->prefixlen != -1) {
1190 				if (isc_netaddr_eqprefix(if_netaddr,
1191 				    &rule->netaddr, rule->prefixlen)) {
1192 
1193 					DPRINTF(4, ("subnet address match - %s\n",
1194 					    action_text(rule->action)));
1195 					return rule->action;
1196 				}
1197 			} else
1198 				if (isc_netaddr_equal(if_netaddr,
1199 				    &rule->netaddr)) {
1200 
1201 					DPRINTF(4, ("address match - %s\n",
1202 					    action_text(rule->action)));
1203 					return rule->action;
1204 				}
1205 			break;
1206 
1207 		case MATCH_IFNAME:
1208 			if (if_name != NULL
1209 			    && !strcasecmp(if_name, rule->if_name)) {
1210 
1211 				DPRINTF(4, ("interface name match - %s\n",
1212 				    action_text(rule->action)));
1213 				return rule->action;
1214 			}
1215 			break;
1216 		}
1217 	}
1218 
1219 	/*
1220 	 * Unless explicitly disabled such as with "nic ignore ::1"
1221 	 * listen on loopback addresses.  Since ntpq and ntpdc query
1222 	 * "localhost" by default, which typically resolves to ::1 and
1223 	 * 127.0.0.1, it's useful to default to listening on both.
1224 	 */
1225 	if (isloopback) {
1226 		DPRINTF(4, ("default loopback listen\n"));
1227 		return ACTION_LISTEN;
1228 	}
1229 
1230 	/*
1231 	 * Treat wildcard addresses specially.  If there is no explicit
1232 	 * "nic ... wildcard" or "nic ... 0.0.0.0" or "nic ... ::" rule
1233 	 * default to drop.
1234 	 */
1235 	if (iswildcard) {
1236 		DPRINTF(4, ("default wildcard drop\n"));
1237 		return ACTION_DROP;
1238 	}
1239 
1240 	/*
1241 	 * Check for "virtual IP" (colon in the interface name) after
1242 	 * the rules so that "ntpd --interface eth0:1 -novirtualips"
1243 	 * does indeed listen on eth0:1's addresses.
1244 	 */
1245 	if (!listen_to_virtual_ips && if_name != NULL
1246 	    && (strchr(if_name, ':') != NULL)) {
1247 
1248 		DPRINTF(4, ("virtual ip - ignore\n"));
1249 		return ACTION_IGNORE;
1250 	}
1251 
1252 	/*
1253 	 * If there are no --interface/-I command-line options and no
1254 	 * interface/nic rules in ntp.conf, the default action is to
1255 	 * listen.  In the presence of rules from either, the default
1256 	 * is to ignore.  This implements ntpd's traditional listen-
1257 	 * every default with no interface listen configuration, and
1258 	 * ensures a single -I eth0 or "nic listen eth0" means do not
1259 	 * listen on any other addresses.
1260 	 */
1261 	if (NULL == nic_rule_list) {
1262 		DPRINTF(4, ("default listen\n"));
1263 		return ACTION_LISTEN;
1264 	}
1265 
1266 	DPRINTF(4, ("implicit ignore\n"));
1267 	return ACTION_IGNORE;
1268 }
1269 
1270 
1271 static void
1272 convert_isc_if(
1273 	isc_interface_t *isc_if,
1274 	struct interface *itf,
1275 	u_short port
1276 	)
1277 {
1278 	strncpy(itf->name, isc_if->name, sizeof(itf->name));
1279 	itf->name[sizeof(itf->name) - 1] = 0; /* strncpy may not */
1280 	itf->family = (u_short)isc_if->af;
1281 	AF(&itf->sin) = itf->family;
1282 	AF(&itf->mask) = itf->family;
1283 	AF(&itf->bcast) = itf->family;
1284 	SET_PORT(&itf->sin, port);
1285 	SET_PORT(&itf->mask, port);
1286 	SET_PORT(&itf->bcast, port);
1287 
1288 	if (IS_IPV4(&itf->sin)) {
1289 		NSRCADR(&itf->sin) = isc_if->address.type.in.s_addr;
1290 		NSRCADR(&itf->mask) = isc_if->netmask.type.in.s_addr;
1291 
1292 		if (isc_if->flags & INTERFACE_F_BROADCAST) {
1293 			itf->flags |= INT_BROADCAST;
1294 			NSRCADR(&itf->bcast) =
1295 			    isc_if->broadcast.type.in.s_addr;
1296 		}
1297 	}
1298 #ifdef INCLUDE_IPV6_SUPPORT
1299 	else if (IS_IPV6(&itf->sin)) {
1300 		SET_ADDR6N(&itf->sin, isc_if->address.type.in6);
1301 		SET_ADDR6N(&itf->mask, isc_if->netmask.type.in6);
1302 
1303 		itf->scopeid = isc_netaddr_getzone(&isc_if->address);
1304 		SET_SCOPE(&itf->sin, itf->scopeid);
1305 	}
1306 #endif /* INCLUDE_IPV6_SUPPORT */
1307 
1308 
1309 	/* Process the rest of the flags */
1310 
1311 	itf->flags |=
1312 		  ((INTERFACE_F_UP & isc_if->flags)
1313 			 ? INT_UP : 0)
1314 		| ((INTERFACE_F_LOOPBACK & isc_if->flags)
1315 			 ? INT_LOOPBACK : 0)
1316 		| ((INTERFACE_F_POINTTOPOINT & isc_if->flags)
1317 			 ? INT_PPP : 0)
1318 		| ((INTERFACE_F_MULTICAST & isc_if->flags)
1319 			 ? INT_MULTICAST : 0)
1320 		;
1321 }
1322 
1323 
1324 /*
1325  * refresh_interface
1326  *
1327  * some OSes have been observed to keep
1328  * cached routes even when more specific routes
1329  * become available.
1330  * this can be mitigated by re-binding
1331  * the socket.
1332  */
1333 static int
1334 refresh_interface(
1335 	struct interface * interface
1336 	)
1337 {
1338 #ifdef  OS_MISSES_SPECIFIC_ROUTE_UPDATES
1339 	if (interface->fd != INVALID_SOCKET) {
1340 		close_and_delete_fd_from_list(interface->fd);
1341 		interface->fd = open_socket(&interface->sin,
1342 					    0, 0, interface);
1343 		 /*
1344 		  * reset TTL indication so TTL is is set again
1345 		  * next time around
1346 		  */
1347 		interface->last_ttl = 0;
1348 		return (interface->fd != INVALID_SOCKET);
1349 	} else
1350 		return 0;	/* invalid sockets are not refreshable */
1351 #else /* !OS_MISSES_SPECIFIC_ROUTE_UPDATES */
1352 	return (interface->fd != INVALID_SOCKET);
1353 #endif /* !OS_MISSES_SPECIFIC_ROUTE_UPDATES */
1354 }
1355 
1356 /*
1357  * interface_update - externally callable update function
1358  */
1359 void
1360 interface_update(
1361 	interface_receiver_t	receiver,
1362 	void *			data)
1363 {
1364 	int new_interface_found;
1365 
1366 	if (disable_dynamic_updates)
1367 		return;
1368 
1369 	BLOCKIO();
1370 	new_interface_found = update_interfaces(NTP_PORT, receiver, data);
1371 	UNBLOCKIO();
1372 
1373 	if (!new_interface_found)
1374 		return;
1375 
1376 #ifdef DEBUG
1377 	msyslog(LOG_DEBUG, "new interface(s) found: waking up resolver");
1378 #endif
1379 #ifdef SYS_WINNT
1380 	/* wake up the resolver thread */
1381 	if (ResolverEventHandle != NULL)
1382 		SetEvent(ResolverEventHandle);
1383 #else
1384 	/* write any single byte to the pipe to wake up the resolver process */
1385 	write( resolver_pipe_fd[1], &new_interface_found, 1 );
1386 #endif
1387 }
1388 
1389 
1390 /*
1391  * sau_from_netaddr() - convert network address on-wire formats.
1392  * Convert from libisc's isc_netaddr_t to NTP's sockaddr_u
1393  */
1394 void
1395 sau_from_netaddr(
1396 	sockaddr_u *psau,
1397 	const isc_netaddr_t *pna
1398 	)
1399 {
1400 	memset(psau, 0, sizeof(*psau));
1401 	AF(psau) = (u_short)pna->family;
1402 	switch (pna->family) {
1403 
1404 	case AF_INET:
1405 		memcpy(&psau->sa4.sin_addr, &pna->type.in,
1406 		       sizeof(psau->sa4.sin_addr));
1407 		break;
1408 
1409 	case AF_INET6:
1410 		memcpy(&psau->sa6.sin6_addr, &pna->type.in6,
1411 		       sizeof(psau->sa6.sin6_addr));
1412 		break;
1413 	}
1414 }
1415 
1416 
1417 static int
1418 is_wildcard_addr(
1419 	sockaddr_u *psau
1420 	)
1421 {
1422 	if (IS_IPV4(psau) && !NSRCADR(psau))
1423 		return 1;
1424 
1425 #ifdef INCLUDE_IPV6_SUPPORT
1426 	if (IS_IPV6(psau) && S_ADDR6_EQ(psau, &in6addr_any))
1427 		return 1;
1428 #endif
1429 
1430 	return 0;
1431 }
1432 
1433 
1434 static int
1435 is_wildcard_netaddr(
1436 	const isc_netaddr_t *pna
1437 	)
1438 {
1439 	sockaddr_u sau;
1440 
1441 	sau_from_netaddr(&sau, pna);
1442 
1443 	return is_wildcard_addr(&sau);
1444 }
1445 
1446 
1447 #ifdef OS_NEEDS_REUSEADDR_FOR_IFADDRBIND
1448 /*
1449  * enable/disable re-use of wildcard address socket
1450  */
1451 static void
1452 set_wildcard_reuse(
1453 	u_short	family,
1454 	int	on
1455 	)
1456 {
1457 	struct interface *any;
1458 	SOCKET fd = INVALID_SOCKET;
1459 
1460 	any = ANY_INTERFACE_BYFAM(family);
1461 	if (any != NULL)
1462 		fd = any->fd;
1463 
1464 	if (fd != INVALID_SOCKET) {
1465 		if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
1466 			       (char *)&on, sizeof(on)))
1467 			msyslog(LOG_ERR,
1468 				"set_wildcard_reuse: setsockopt(SO_REUSEADDR, %s) failed: %m",
1469 				on ? "on" : "off");
1470 
1471 		DPRINTF(4, ("set SO_REUSEADDR to %s on %s\n",
1472 			    on ? "on" : "off",
1473 			    stoa(&any->sin)));
1474 	}
1475 }
1476 #endif /* OS_NEEDS_REUSEADDR_FOR_IFADDRBIND */
1477 
1478 /*
1479  * update_interface strategy
1480  *
1481  * toggle configuration phase
1482  *
1483  * Phase 1:
1484  * forall currently existing interfaces
1485  *   if address is known:
1486  *	drop socket - rebind again
1487  *
1488  *   if address is NOT known:
1489  *	attempt to create a new interface entry
1490  *
1491  * Phase 2:
1492  * forall currently known non MCAST and WILDCARD interfaces
1493  *   if interface does not match configuration phase (not seen in phase 1):
1494  *	remove interface from known interface list
1495  *	forall peers associated with this interface
1496  *         disconnect peer from this interface
1497  *
1498  * Phase 3:
1499  *   attempt to re-assign interfaces to peers
1500  *
1501  */
1502 
1503 static int
1504 update_interfaces(
1505 	u_short			port,
1506 	interface_receiver_t	receiver,
1507 	void *			data
1508 	)
1509 {
1510 	isc_mem_t *		mctx = (void *)-1;
1511 	interface_info_t	ifi;
1512 	isc_interfaceiter_t *	iter;
1513 	isc_result_t		result;
1514 	isc_interface_t		isc_if;
1515 	int			new_interface_found;
1516 	unsigned int		family;
1517 	struct interface	interface;
1518 	struct interface *	iface;
1519 	struct interface *	next;
1520 
1521 	DPRINTF(3, ("update_interfaces(%d)\n", port));
1522 
1523 	/*
1524 	 * phase one - scan interfaces
1525 	 * - create those that are not found
1526 	 * - update those that are found
1527 	 */
1528 
1529 	new_interface_found = 0;
1530 	iter = NULL;
1531 	result = isc_interfaceiter_create(mctx, &iter);
1532 
1533 	if (result != ISC_R_SUCCESS)
1534 		return 0;
1535 
1536 	/*
1537 	 * Toggle system interface scan phase to find untouched
1538 	 * interfaces to be deleted.
1539 	 */
1540 	sys_interphase ^= 0x1;
1541 
1542 	for (result = isc_interfaceiter_first(iter);
1543 	     ISC_R_SUCCESS == result;
1544 	     result = isc_interfaceiter_next(iter)) {
1545 
1546 		result = isc_interfaceiter_current(iter, &isc_if);
1547 
1548 		if (result != ISC_R_SUCCESS)
1549 			break;
1550 
1551 		/* See if we have a valid family to use */
1552 		family = isc_if.address.family;
1553 		if (AF_INET != family && AF_INET6 != family)
1554 			continue;
1555 		if (AF_INET == family && !ipv4_works)
1556 			continue;
1557 		if (AF_INET6 == family && !ipv6_works)
1558 			continue;
1559 
1560 		/* create prototype */
1561 		init_interface(&interface);
1562 
1563 		convert_isc_if(&isc_if, &interface, port);
1564 
1565 		/*
1566 		 * Check if and how we are going to use the interface.
1567 		 */
1568 		switch (interface_action(isc_if.name, &isc_if.address,
1569 					 isc_if.flags)) {
1570 
1571 		case ACTION_IGNORE:
1572 			continue;
1573 
1574 		case ACTION_LISTEN:
1575 			interface.ignore_packets = ISC_FALSE;
1576 			break;
1577 
1578 		case ACTION_DROP:
1579 			interface.ignore_packets = ISC_TRUE;
1580 			break;
1581 		}
1582 
1583 		DPRINT_INTERFACE(4, (&interface, "examining ", "\n"));
1584 
1585 		 /* interfaces must be UP to be usable */
1586 		if (!(interface.flags & INT_UP)) {
1587 			DPRINTF(4, ("skipping interface %s (%s) - DOWN\n",
1588 				    interface.name, stoa(&interface.sin)));
1589 			continue;
1590 		}
1591 
1592 		/*
1593 		 * skip any interfaces UP and bound to a wildcard
1594 		 * address - some dhcp clients produce that in the
1595 		 * wild
1596 		 */
1597 		if (is_wildcard_addr(&interface.sin))
1598 			continue;
1599 
1600 		/*
1601 		 * map to local *address* in order to map all duplicate
1602 		 * interfaces to an interface structure with the
1603 		 * appropriate socket.  Our name space is (ip-address),
1604 		 * NOT (interface name, ip-address).
1605 		 */
1606 		iface = getinterface(&interface.sin, INT_WILDCARD);
1607 
1608 		if (iface != NULL && refresh_interface(iface)) {
1609 			/*
1610 			 * found existing and up to date interface -
1611 			 * mark present.
1612 			 */
1613 			if (iface->phase != sys_interphase) {
1614 				/*
1615 				 * On a new round we reset the name so
1616 				 * the interface name shows up again if
1617 				 * this address is no longer shared.
1618 				 * The same reasoning goes for the
1619 				 * ignore_packets flag.
1620 				 */
1621 				strncpy(iface->name, interface.name,
1622 					sizeof(iface->name));
1623 				iface->ignore_packets =
1624 					interface.ignore_packets;
1625 			} else
1626 				/* name collision - rename interface */
1627 				strncpy(iface->name, "*multiple*",
1628 					sizeof(iface->name));
1629 
1630 			DPRINT_INTERFACE(4, (iface, "updating ",
1631 					     " present\n"));
1632 
1633 			if (iface->ignore_packets !=
1634 			    interface.ignore_packets) {
1635 				/*
1636 				 * We have conflicting configurations
1637 				 * for the interface address. This is
1638 				 * caused by using -I <interfacename>
1639 				 * for an interface that shares its
1640 				 * address with other interfaces. We
1641 				 * can not disambiguate incoming
1642 				 * packets delivered to this socket
1643 				 * without extra syscalls/features.
1644 				 * These are not (commonly) available.
1645 				 * Note this is a more unusual
1646 				 * configuration where several
1647 				 * interfaces share an address but
1648 				 * filtering via interface name is
1649 				 * attempted.  We resolve the
1650 				 * configuration conflict by disabling
1651 				 * the processing of received packets.
1652 				 * This leads to no service on the
1653 				 * interface address where the conflict
1654 				 * occurs.
1655 				 */
1656 				msyslog(LOG_ERR,
1657 					"WARNING: conflicting enable configuration for interfaces %s and %s for address %s - unsupported configuration - address DISABLED",
1658 					interface.name, iface->name,
1659 					stoa(&interface.sin));
1660 
1661 				iface->ignore_packets = ISC_TRUE;
1662 			}
1663 
1664 			iface->phase = sys_interphase;
1665 
1666 			ifi.action = IFS_EXISTS;
1667 			ifi.interface = iface;
1668 			if (receiver != NULL)
1669 				(*receiver)(data, &ifi);
1670 		} else {
1671 			/*
1672 			 * This is new or refreshing failed - add to
1673 			 * our interface list.  If refreshing failed we
1674 			 * will delete the interface structure in phase
1675 			 * 2 as the interface was not marked current.
1676 			 * We can bind to the address as the refresh
1677 			 * code already closed the offending socket
1678 			 */
1679 			iface = create_interface(port, &interface);
1680 
1681 			if (iface != NULL) {
1682 				ifi.action = IFS_CREATED;
1683 				ifi.interface = iface;
1684 				if (receiver != NULL)
1685 					(*receiver)(data, &ifi);
1686 
1687 				new_interface_found = 1;
1688 
1689 				DPRINT_INTERFACE(3,
1690 					(iface, "updating ",
1691 					 " new - created\n"));
1692 			} else {
1693 				DPRINT_INTERFACE(3,
1694 					(&interface, "updating ",
1695 					 " new - creation FAILED"));
1696 
1697 				msyslog(LOG_INFO,
1698 					"failed to init interface for address %s",
1699 					stoa(&interface.sin));
1700 				continue;
1701 			}
1702 		}
1703 	}
1704 
1705 	isc_interfaceiter_destroy(&iter);
1706 
1707 	/*
1708 	 * phase 2 - delete gone interfaces - reassigning peers to
1709 	 * other interfaces
1710 	 */
1711 	iface = inter_list;
1712 
1713 	while (iface != NULL) {
1714 		next = iface->link;
1715 
1716 		if (!(iface->flags & (INT_WILDCARD | INT_MCASTIF))) {
1717 			/*
1718 			 * if phase does not match sys_phase this
1719 			 * interface was not enumerated during the last
1720 			 * interface scan - so it is gone and will be
1721 			 * deleted here unless it is solely an MCAST or
1722 			 * WILDCARD interface.
1723 			 */
1724 			if (iface->phase != sys_interphase) {
1725 				DPRINT_INTERFACE(3,
1726 					(iface, "updating ",
1727 					 "GONE - deleting\n"));
1728 				remove_interface(iface);
1729 
1730 				ifi.action = IFS_DELETED;
1731 				ifi.interface = iface;
1732 				if (receiver != NULL)
1733 					(*receiver)(data, &ifi);
1734 
1735 				/*
1736 				 * disconnect peers from deleted
1737 				 * interface
1738 				 */
1739 				while (iface->peers != NULL)
1740 					set_peerdstadr(iface->peers, NULL);
1741 
1742 				/*
1743 				 * update globals in case we lose
1744 				 * a loopback interface
1745 				 */
1746 				if (iface == loopback_interface)
1747 					loopback_interface = NULL;
1748 
1749 				delete_interface(iface);
1750 			}
1751 		}
1752 		iface = next;
1753 	}
1754 
1755 	/*
1756 	 * phase 3 - re-configure as the world has changed if necessary
1757 	 */
1758 	refresh_all_peerinterfaces();
1759 	return new_interface_found;
1760 }
1761 
1762 
1763 /*
1764  * create_sockets - create a socket for each interface plus a default
1765  *			socket for when we don't know where to send
1766  */
1767 static int
1768 create_sockets(
1769 	u_short port
1770 	)
1771 {
1772 #ifndef HAVE_IO_COMPLETION_PORT
1773 	/*
1774 	 * I/O Completion Ports don't care about the select and FD_SET
1775 	 */
1776 	maxactivefd = 0;
1777 	FD_ZERO(&activefds);
1778 #endif
1779 
1780 	DPRINTF(2, ("create_sockets(%d)\n", port));
1781 
1782 	create_wildcards(port);
1783 
1784 	update_interfaces(port, NULL, NULL);
1785 
1786 	if (sys_bclient)
1787 		io_setbclient();
1788 
1789 	/*
1790 	 * Now that we have opened all the sockets, turn off the reuse
1791 	 * flag for security.
1792 	 */
1793 	set_reuseaddr(0);
1794 
1795 	DPRINTF(2, ("create_sockets: Total interfaces = %d\n", ninterfaces));
1796 
1797 	return ninterfaces;
1798 }
1799 
1800 /*
1801  * create_interface - create a new interface for a given prototype
1802  *		      binding the socket.
1803  */
1804 static struct interface *
1805 create_interface(
1806 	u_short			port,
1807 	struct interface *	protot
1808 	)
1809 {
1810 	sockaddr_u resmask;
1811 	struct interface *iface;
1812 
1813 	DPRINTF(2, ("create_interface(%s#%d)\n", stoa(&protot->sin),
1814 		    port));
1815 
1816 	/* build an interface */
1817 	iface = new_interface(protot);
1818 
1819 	/*
1820 	 * create socket
1821 	 */
1822 	iface->fd = open_socket(&iface->sin, 0, 0, iface);
1823 
1824 	if (iface->fd != INVALID_SOCKET)
1825 		list_if_listening(iface);
1826 
1827 	if ((INT_BROADCAST & iface->flags)
1828 	    && iface->bfd != INVALID_SOCKET)
1829 		msyslog(LOG_INFO, "Listening on broadcast address %s#%d",
1830 			stoa((&iface->bcast)), port);
1831 
1832 	if (INVALID_SOCKET == iface->fd
1833 	    && INVALID_SOCKET == iface->bfd) {
1834 		msyslog(LOG_ERR, "unable to create socket on %s (%d) for %s#%d",
1835 			iface->name,
1836 			iface->ifnum,
1837 			stoa((&iface->sin)),
1838 			port);
1839 		delete_interface(iface);
1840 		return NULL;
1841 	}
1842 
1843 	/*
1844 	 * Blacklist our own addresses, no use talking to ourself
1845 	 */
1846 	SET_HOSTMASK(&resmask, AF(&iface->sin));
1847 	hack_restrict(RESTRICT_FLAGS, &iface->sin, &resmask,
1848 		      RESM_NTPONLY | RESM_INTERFACE, RES_IGNORE);
1849 
1850 	/*
1851 	 * set globals with the first found
1852 	 * loopback interface of the appropriate class
1853 	 */
1854 	if (NULL == loopback_interface && AF_INET == iface->family
1855 	    && (INT_LOOPBACK & iface->flags))
1856 		loopback_interface = iface;
1857 
1858 	/*
1859 	 * put into our interface list
1860 	 */
1861 	add_addr_to_list(&iface->sin, iface);
1862 	add_interface(iface);
1863 
1864 	DPRINT_INTERFACE(2, (iface, "created ", "\n"));
1865 	return iface;
1866 }
1867 
1868 
1869 #ifdef SO_EXCLUSIVEADDRUSE
1870 static void
1871 set_excladdruse(
1872 	SOCKET fd
1873 	)
1874 {
1875 	int one = 1;
1876 	int failed;
1877 #ifdef SYS_WINNT
1878 	DWORD err;
1879 #endif
1880 
1881 	failed = setsockopt(fd, SOL_SOCKET, SO_EXCLUSIVEADDRUSE,
1882 			    (char *)&one, sizeof(one));
1883 
1884 	if (!failed)
1885 		return;
1886 
1887 #ifdef SYS_WINNT
1888 	/*
1889 	 * Prior to Windows XP setting SO_EXCLUSIVEADDRUSE can fail with
1890 	 * error WSAINVAL depending on service pack level and whether
1891 	 * the user account is in the Administrators group.  Do not
1892 	 * complain if it fails that way on versions prior to XP (5.1).
1893 	 */
1894 	err = GetLastError();
1895 
1896 	if (isc_win32os_versioncheck(5, 1, 0, 0) < 0	/* < 5.1/XP */
1897 	    && WSAEINVAL == err)
1898 		return;
1899 
1900 	SetLastError(err);
1901 #endif
1902 	msyslog(LOG_ERR,
1903 		"setsockopt(%d, SO_EXCLUSIVEADDRUSE, on): %m",
1904 		(int)fd);
1905 }
1906 #endif  /* SO_EXCLUSIVEADDRUSE */
1907 
1908 
1909 /*
1910  * set_reuseaddr() - set/clear REUSEADDR on all sockets
1911  *			NB possible hole - should we be doing this on broadcast
1912  *			fd's also?
1913  */
1914 static void
1915 set_reuseaddr(
1916 	int flag
1917 	)
1918 {
1919 #ifndef SO_EXCLUSIVEADDRUSE
1920 	struct interface *interf;
1921 
1922 	for (interf = inter_list;
1923 	     interf != NULL;
1924 	     interf = interf->link) {
1925 
1926 		if (interf->flags & INT_WILDCARD)
1927 			continue;
1928 
1929 		/*
1930 		 * if interf->fd  is INVALID_SOCKET, we might have a adapter
1931 		 * configured but not present
1932 		 */
1933 		DPRINTF(4, ("setting SO_REUSEADDR on %.16s@%s to %s\n",
1934 			    interf->name, stoa(&interf->sin),
1935 			    flag ? "on" : "off"));
1936 
1937 		if (interf->fd != INVALID_SOCKET) {
1938 			if (setsockopt(interf->fd, SOL_SOCKET,
1939 					SO_REUSEADDR, (char *)&flag,
1940 					sizeof(flag))) {
1941 				msyslog(LOG_ERR, "set_reuseaddr: setsockopt(SO_REUSEADDR, %s) failed: %m", flag ? "on" : "off");
1942 			}
1943 		}
1944 	}
1945 #endif /* ! SO_EXCLUSIVEADDRUSE */
1946 }
1947 
1948 /*
1949  * This is just a wrapper around an internal function so we can
1950  * make other changes as necessary later on
1951  */
1952 void
1953 enable_broadcast(
1954 	struct interface *	iface,
1955 	sockaddr_u *		baddr
1956 	)
1957 {
1958 #ifdef OPEN_BCAST_SOCKET
1959 	socket_broadcast_enable(iface, iface->fd, baddr);
1960 #endif
1961 }
1962 
1963 #ifdef OPEN_BCAST_SOCKET
1964 /*
1965  * Enable a broadcast address to a given socket
1966  * The socket is in the inter_list all we need to do is enable
1967  * broadcasting. It is not this function's job to select the socket
1968  */
1969 static isc_boolean_t
1970 socket_broadcast_enable(
1971 	struct interface *	iface,
1972 	SOCKET			fd,
1973 	sockaddr_u *		baddr
1974 	)
1975 {
1976 #ifdef SO_BROADCAST
1977 	int on = 1;
1978 
1979 	if (IS_IPV4(baddr)) {
1980 		/* if this interface can support broadcast, set SO_BROADCAST */
1981 		if (setsockopt(fd, SOL_SOCKET, SO_BROADCAST,
1982 			       (char *)&on, sizeof(on)))
1983 			msyslog(LOG_ERR,
1984 				"setsockopt(SO_BROADCAST) enable failure on address %s: %m",
1985 				stoa(baddr));
1986 		else
1987 			DPRINTF(2, ("Broadcast enabled on socket %d for address %s\n",
1988 				    fd, stoa(baddr)));
1989 	}
1990 	iface->flags |= INT_BCASTOPEN;
1991 	broadcast_client_enabled = ISC_TRUE;
1992 	return ISC_TRUE;
1993 #else
1994 	return ISC_FALSE;
1995 #endif /* SO_BROADCAST */
1996 }
1997 
1998 /*
1999  * Remove a broadcast address from a given socket
2000  * The socket is in the inter_list all we need to do is disable
2001  * broadcasting. It is not this function's job to select the socket
2002  */
2003 static isc_boolean_t
2004 socket_broadcast_disable(
2005 	struct interface *	iface,
2006 	sockaddr_u *		baddr
2007 	)
2008 {
2009 #ifdef SO_BROADCAST
2010 	int off = 0;	/* This seems to be OK as an int */
2011 
2012 	if (IS_IPV4(baddr) && setsockopt(iface->fd, SOL_SOCKET,
2013 	    SO_BROADCAST, (char *)&off, sizeof(off)))
2014 		msyslog(LOG_ERR,
2015 			"setsockopt(SO_BROADCAST) disable failure on address %s: %m",
2016 			stoa(baddr));
2017 
2018 	iface->flags &= ~INT_BCASTOPEN;
2019 	broadcast_client_enabled = ISC_FALSE;
2020 	return ISC_TRUE;
2021 #else
2022 	return ISC_FALSE;
2023 #endif /* SO_BROADCAST */
2024 }
2025 
2026 #endif /* OPEN_BCAST_SOCKET */
2027 
2028 /*
2029  * return the broadcast client flag value
2030  */
2031 isc_boolean_t
2032 get_broadcastclient_flag(void)
2033 {
2034 	return (broadcast_client_enabled);
2035 }
2036 /*
2037  * Check to see if the address is a multicast address
2038  */
2039 static isc_boolean_t
2040 addr_ismulticast(
2041 	sockaddr_u *maddr
2042 	)
2043 {
2044 	isc_boolean_t result;
2045 
2046 #ifndef INCLUDE_IPV6_MULTICAST_SUPPORT
2047 	/*
2048 	 * If we don't have IPV6 support any IPV6 addr is not multicast
2049 	 */
2050 	if (IS_IPV6(maddr))
2051 		result = ISC_FALSE;
2052 	else
2053 #endif
2054 		result = IS_MCAST(maddr);
2055 
2056 	if (!result)
2057 		DPRINTF(4, ("address %s is not multicast\n",
2058 			    stoa(maddr)));
2059 
2060 	return result;
2061 }
2062 
2063 /*
2064  * Multicast servers need to set the appropriate Multicast interface
2065  * socket option in order for it to know which interface to use for
2066  * send the multicast packet.
2067  */
2068 void
2069 enable_multicast_if(
2070 	struct interface *	iface,
2071 	sockaddr_u *		maddr
2072 	)
2073 {
2074 #ifdef MCAST
2075 	TYPEOF_IP_MULTICAST_LOOP off = 0;
2076 
2077 	NTP_REQUIRE(AF(maddr) == AF(&iface->sin));
2078 
2079 	switch (AF(&iface->sin)) {
2080 
2081 	case AF_INET:
2082 		if (setsockopt(iface->fd, IPPROTO_IP, IP_MULTICAST_IF,
2083 			       (void *)&NSRCADR(&iface->sin),
2084 			       sizeof(NSRCADR(&iface->sin)))) {
2085 
2086 			msyslog(LOG_ERR,
2087 				"setsockopt IP_MULTICAST_IF failed: %m on socket %d, addr %s for multicast address %s",
2088 				iface->fd, stoa(&iface->sin),
2089 				stoa(maddr));
2090 			return;
2091 		}
2092 #ifdef IP_MULTICAST_LOOP
2093 		/*
2094 		 * Don't send back to itself, but allow failure to set
2095 		 */
2096 		if (setsockopt(iface->fd, IPPROTO_IP,
2097 			       IP_MULTICAST_LOOP,
2098 			       SETSOCKOPT_ARG_CAST &off,
2099 			       sizeof(off))) {
2100 
2101 			msyslog(LOG_ERR,
2102 				"setsockopt IP_MULTICAST_LOOP failed: %m on socket %d, addr %s for multicast address %s",
2103 				iface->fd, stoa(&iface->sin),
2104 				stoa(maddr));
2105 		}
2106 #endif
2107 		DPRINTF(4, ("Added IPv4 multicast interface on socket %d, addr %s for multicast address %s\n",
2108 			    iface->fd, stoa(&iface->sin),
2109 			    stoa(maddr)));
2110 		break;
2111 
2112 	case AF_INET6:
2113 #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
2114 		if (setsockopt(iface->fd, IPPROTO_IPV6,
2115 			       IPV6_MULTICAST_IF,
2116 			       (char *)&iface->scopeid,
2117 			       sizeof(iface->scopeid))) {
2118 
2119 			msyslog(LOG_ERR,
2120 				"setsockopt IPV6_MULTICAST_IF failed: %m on socket %d, addr %s, scope %d for multicast address %s",
2121 				iface->fd, stoa(&iface->sin),
2122 				iface->scopeid, stoa(maddr));
2123 			return;
2124 		}
2125 #ifdef IPV6_MULTICAST_LOOP
2126 		/*
2127 		 * Don't send back to itself, but allow failure to set
2128 		 */
2129 		if (setsockopt(iface->fd, IPPROTO_IPV6,
2130 			       IPV6_MULTICAST_LOOP,
2131 			       (char *) &off, sizeof(off))) {
2132 
2133 			msyslog(LOG_ERR,
2134 				"setsockopt IP_MULTICAST_LOOP failed: %m on socket %d, addr %s for multicast address %s",
2135 				iface->fd, stoa(&iface->sin),
2136 				stoa(maddr));
2137 		}
2138 #endif
2139 		DPRINTF(4, ("Added IPv6 multicast interface on socket %d, addr %s, scope %d for multicast address %s\n",
2140 			    iface->fd,  stoa(&iface->sin),
2141 			    iface->scopeid, stoa(maddr)));
2142 		break;
2143 #else
2144 		return;
2145 #endif	/* INCLUDE_IPV6_MULTICAST_SUPPORT */
2146 	}
2147 	return;
2148 #endif
2149 }
2150 
2151 /*
2152  * Add a multicast address to a given socket
2153  * The socket is in the inter_list all we need to do is enable
2154  * multicasting. It is not this function's job to select the socket
2155  */
2156 #ifdef MCAST
2157 static isc_boolean_t
2158 socket_multicast_enable(
2159 	struct interface *	iface,
2160 	int			lscope,
2161 	sockaddr_u *		maddr
2162 	)
2163 {
2164 	struct ip_mreq		mreq;
2165 #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
2166 	struct ipv6_mreq	mreq6;
2167 #endif
2168 
2169 	if (find_addr_in_list(maddr) != NULL) {
2170 		DPRINTF(4, ("socket_multicast_enable(%s): already enabled\n",
2171 			    stoa(maddr)));
2172 		return ISC_TRUE;
2173 	}
2174 
2175 	switch (AF(maddr)) {
2176 
2177 	case AF_INET:
2178 		memset(&mreq, 0, sizeof(mreq));
2179 		mreq.imr_multiaddr = SOCK_ADDR4(maddr);
2180 		mreq.imr_interface.s_addr = htonl(INADDR_ANY);
2181 		if (setsockopt(iface->fd,
2182 			       IPPROTO_IP,
2183 			       IP_ADD_MEMBERSHIP,
2184 			       (char *)&mreq,
2185 			       sizeof(mreq))) {
2186 			msyslog(LOG_ERR,
2187 				"setsockopt IP_ADD_MEMBERSHIP failed: %m on socket %d, addr %s for %x / %x (%s)",
2188 				iface->fd, stoa(&iface->sin),
2189 				mreq.imr_multiaddr.s_addr,
2190 				mreq.imr_interface.s_addr,
2191 				stoa(maddr));
2192 			return ISC_FALSE;
2193 		}
2194 		DPRINTF(4, ("Added IPv4 multicast membership on socket %d, addr %s for %x / %x (%s)\n",
2195 			    iface->fd, stoa(&iface->sin),
2196 			    mreq.imr_multiaddr.s_addr,
2197 			    mreq.imr_interface.s_addr, stoa(maddr)));
2198 		break;
2199 
2200 	case AF_INET6:
2201 #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
2202 		/*
2203 		 * Enable reception of multicast packets.
2204 		 * If the address is link-local we can get the
2205 		 * interface index from the scope id. Don't do this
2206 		 * for other types of multicast addresses. For now let
2207 		 * the kernel figure it out.
2208 		 */
2209 		memset(&mreq6, 0, sizeof(mreq6));
2210 		mreq6.ipv6mr_multiaddr = SOCK_ADDR6(maddr);
2211 		mreq6.ipv6mr_interface = lscope;
2212 
2213 		if (setsockopt(iface->fd, IPPROTO_IPV6,
2214 			       IPV6_JOIN_GROUP, (char *)&mreq6,
2215 			       sizeof(mreq6))) {
2216 			msyslog(LOG_ERR,
2217 				"setsockopt IPV6_JOIN_GROUP failed: %m on socket %d, addr %s for interface %d (%s)",
2218 				iface->fd, stoa(&iface->sin),
2219 				mreq6.ipv6mr_interface, stoa(maddr));
2220 			return ISC_FALSE;
2221 		}
2222 		DPRINTF(4, ("Added IPv6 multicast group on socket %d, addr %s for interface %d(%s)\n",
2223 			    iface->fd, stoa(&iface->sin),
2224 			    mreq6.ipv6mr_interface, stoa(maddr)));
2225 #else
2226 		return ISC_FALSE;
2227 #endif	/* INCLUDE_IPV6_MULTICAST_SUPPORT */
2228 	}
2229 	iface->flags |= INT_MCASTOPEN;
2230 	iface->num_mcast++;
2231 	add_addr_to_list(maddr, iface);
2232 	return ISC_TRUE;
2233 }
2234 
2235 /*
2236  * Remove a multicast address from a given socket
2237  * The socket is in the inter_list all we need to do is disable
2238  * multicasting. It is not this function's job to select the socket
2239  */
2240 static isc_boolean_t
2241 socket_multicast_disable(
2242 	struct interface *	iface,
2243 	sockaddr_u *		maddr
2244 	)
2245 {
2246 #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
2247 	struct ipv6_mreq mreq6;
2248 #endif
2249 	struct ip_mreq mreq;
2250 
2251 	memset(&mreq, 0, sizeof(mreq));
2252 
2253 	if (find_addr_in_list(maddr) == NULL) {
2254 		DPRINTF(4, ("socket_multicast_disable(%s): not found\n",
2255 			    stoa(maddr)));
2256 		return ISC_TRUE;
2257 	}
2258 
2259 	switch (AF(maddr)) {
2260 
2261 	case AF_INET:
2262 		mreq.imr_multiaddr = SOCK_ADDR4(maddr);
2263 		mreq.imr_interface = SOCK_ADDR4(&iface->sin);
2264 		if (setsockopt(iface->fd, IPPROTO_IP,
2265 			       IP_DROP_MEMBERSHIP, (char *)&mreq,
2266 			       sizeof(mreq))) {
2267 
2268 			msyslog(LOG_ERR,
2269 				"setsockopt IP_DROP_MEMBERSHIP failed: %m on socket %d, addr %s for %x / %x (%s)",
2270 				iface->fd, stoa(&iface->sin),
2271 				SRCADR(maddr), SRCADR(&iface->sin),
2272 				stoa(maddr));
2273 			return ISC_FALSE;
2274 		}
2275 		break;
2276 	case AF_INET6:
2277 #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
2278 		/*
2279 		 * Disable reception of multicast packets
2280 		 * If the address is link-local we can get the
2281 		 * interface index from the scope id.  Don't do this
2282 		 * for other types of multicast addresses. For now let
2283 		 * the kernel figure it out.
2284 		 */
2285 		mreq6.ipv6mr_multiaddr = SOCK_ADDR6(maddr);
2286 		mreq6.ipv6mr_interface = iface->scopeid;
2287 
2288 		if (setsockopt(iface->fd, IPPROTO_IPV6,
2289 			       IPV6_LEAVE_GROUP, (char *)&mreq6,
2290 			       sizeof(mreq6))) {
2291 
2292 			msyslog(LOG_ERR,
2293 				"setsockopt IPV6_LEAVE_GROUP failure: %m on socket %d, addr %s for %d (%s)",
2294 				iface->fd, stoa(&iface->sin),
2295 				iface->scopeid, stoa(maddr));
2296 			return ISC_FALSE;
2297 		}
2298 		break;
2299 #else
2300 		return ISC_FALSE;
2301 #endif	/* INCLUDE_IPV6_MULTICAST_SUPPORT */
2302 	}
2303 
2304 	iface->num_mcast--;
2305 	if (!iface->num_mcast)
2306 		iface->flags &= ~INT_MCASTOPEN;
2307 
2308 	return ISC_TRUE;
2309 }
2310 #endif	/* MCAST */
2311 
2312 /*
2313  * io_setbclient - open the broadcast client sockets
2314  */
2315 void
2316 io_setbclient(void)
2317 {
2318 #ifdef OPEN_BCAST_SOCKET
2319 	struct interface *	interf;
2320 	int			nif;
2321 	isc_boolean_t		jstatus;
2322 	SOCKET			fd;
2323 
2324 	nif = 0;
2325 	set_reuseaddr(1);
2326 
2327 	for (interf = inter_list;
2328 	     interf != NULL;
2329 	     interf = interf->link) {
2330 
2331 		if (interf->flags & (INT_WILDCARD | INT_LOOPBACK))
2332 			continue;
2333 
2334 		/* use only allowed addresses */
2335 		if (interf->ignore_packets)
2336 			continue;
2337 
2338 
2339 		/* Need a broadcast-capable interface */
2340 		if (!(interf->flags & INT_BROADCAST))
2341 			continue;
2342 
2343 		/* Only IPv4 addresses are valid for broadcast */
2344 		NTP_REQUIRE(IS_IPV4(&interf->sin));
2345 
2346 		/* Do we already have the broadcast address open? */
2347 		if (interf->flags & INT_BCASTOPEN) {
2348 			/*
2349 			 * account for already open interfaces to avoid
2350 			 * misleading warning below
2351 			 */
2352 			nif++;
2353 			continue;
2354 		}
2355 
2356 		/*
2357 		 * Try to open the broadcast address
2358 		 */
2359 		interf->family = AF_INET;
2360 		interf->bfd = open_socket(&interf->bcast, 1, 0, interf);
2361 
2362 		/*
2363 		 * If we succeeded then we use it otherwise enable
2364 		 * broadcast on the interface address
2365 		 */
2366 		if (interf->bfd != INVALID_SOCKET) {
2367 			fd = interf->bfd;
2368 			jstatus = ISC_TRUE;
2369 		} else {
2370 			fd = interf->fd;
2371 			jstatus = socket_broadcast_enable(interf, fd,
2372 					&interf->sin);
2373 		}
2374 
2375 		/* Enable Broadcast on socket */
2376 		if (jstatus) {
2377 			nif++;
2378 			msyslog(LOG_INFO,
2379 				"io_setbclient: Opened broadcast client on interface #%d %s",
2380 				interf->ifnum, interf->name);
2381 			interf->addr_refid = addr2refid(&interf->sin);
2382 		}
2383 	}
2384 	set_reuseaddr(0);
2385 	if (nif > 0)
2386 		DPRINTF(1, ("io_setbclient: Opened broadcast clients\n"));
2387 	else if (!nif)
2388 		msyslog(LOG_ERR,
2389 			"Unable to listen for broadcasts, no broadcast interfaces available");
2390 #else
2391 	msyslog(LOG_ERR,
2392 		"io_setbclient: Broadcast Client disabled by build");
2393 #endif	/* OPEN_BCAST_SOCKET */
2394 }
2395 
2396 /*
2397  * io_unsetbclient - close the broadcast client sockets
2398  */
2399 void
2400 io_unsetbclient(void)
2401 {
2402 	struct interface *interf;
2403 
2404 	for (interf = inter_list;
2405 	     NULL != interf;
2406 	     interf = interf->link)
2407 	{
2408 		if (interf->flags & INT_WILDCARD)
2409 			continue;
2410 
2411 		if (!(interf->flags & INT_BCASTOPEN))
2412 			continue;
2413 
2414 		socket_broadcast_disable(interf, &interf->sin);
2415 	}
2416 }
2417 
2418 /*
2419  * io_multicast_add() - add multicast group address
2420  */
2421 void
2422 io_multicast_add(
2423 	sockaddr_u *addr
2424 	)
2425 {
2426 #ifdef MCAST
2427 	struct interface *interface;
2428 #ifndef MULTICAST_NONEWSOCKET
2429 	struct interface *iface;
2430 #endif
2431 	int lscope = 0;
2432 
2433 	/*
2434 	 * Check to see if this is a multicast address
2435 	 */
2436 	if (!addr_ismulticast(addr))
2437 		return;
2438 
2439 	/* If we already have it we can just return */
2440 	if (NULL != find_flagged_addr_in_list(addr, INT_MCASTOPEN)) {
2441 		msyslog(LOG_INFO,
2442 			"Duplicate request found for multicast address %s",
2443 			stoa(addr));
2444 		return;
2445 	}
2446 
2447 #ifndef MULTICAST_NONEWSOCKET
2448 	interface = new_interface(NULL);
2449 
2450 	/*
2451 	 * Open a new socket for the multicast address
2452 	 */
2453 	interface->family =
2454 		AF(&interface->sin) =
2455 		AF(&interface->mask) = AF(addr);
2456 	SET_PORT(&interface->sin, NTP_PORT);
2457 	SET_ONESMASK(&interface->mask);
2458 
2459 	switch(AF(addr)) {
2460 
2461 	case AF_INET:
2462 		NSRCADR(&interface->sin) = NSRCADR(addr);
2463 		break;
2464 
2465 	case AF_INET6:
2466 #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
2467 		SET_ADDR6N(&interface->sin, SOCK_ADDR6(addr));
2468 		lscope = SCOPE(addr);
2469 		SET_SCOPE(&interface->sin, lscope);
2470 #endif
2471 		iface = findlocalcastinterface(addr);
2472 		if (iface != NULL)
2473 			DPRINTF(4, ("Found interface #%d %s, scope %d for address %s\n",
2474 				    iface->ifnum, iface->name, lscope,
2475 				    stoa(addr)));
2476 	}
2477 
2478 	set_reuseaddr(1);
2479 	interface->bfd = INVALID_SOCKET;
2480 	interface->fd = open_socket(&interface->sin, INT_MULTICAST, 0,
2481 				    interface);
2482 
2483 	if (interface->fd != INVALID_SOCKET) {
2484 		interface->bfd = INVALID_SOCKET;
2485 		interface->ignore_packets = ISC_FALSE;
2486 		interface->flags |= INT_MCASTIF;
2487 
2488 		strncpy(interface->name, "multicast",
2489 			sizeof(interface->name));
2490 		DPRINT_INTERFACE(2, (interface, "multicast add ", "\n"));
2491 		/*
2492 		 * socket_multicast_enable() will add this address to
2493 		 * the addresslist
2494 		 */
2495 		add_interface(interface);
2496 		list_if_listening(interface);
2497 	} else {
2498 		/* bind failed, re-use wildcard interface */
2499 		delete_interface(interface);
2500 
2501 		if (IS_IPV4(addr))
2502 			interface = wildipv4;
2503 		else if (IS_IPV6(addr))
2504 			interface = wildipv6;
2505 		else
2506 			interface = NULL;
2507 
2508 		if (interface != NULL) {
2509 			/* HACK ! -- stuff in an address */
2510 			/* because we don't bind addr? DH */
2511 			interface->bcast = *addr;
2512 			msyslog(LOG_ERR,
2513 				"multicast address %s using wildcard interface #%d %s",
2514 				 stoa(addr), interface->ifnum,
2515 				 interface->name);
2516 		} else {
2517 			msyslog(LOG_ERR,
2518 				"No multicast socket available to use for address %s",
2519 				stoa(addr));
2520 			return;
2521 		}
2522 	}
2523 #else
2524 	/*
2525 	 * For the case where we can't use a separate socket
2526 	 */
2527 	interface = findlocalcastinterface(addr);
2528 	/*
2529 	 * If we don't have a valid socket, just return
2530 	 */
2531 	if (NULL == interface) {
2532 		msyslog(LOG_ERR,
2533 			"Can not add multicast address %s: no multicast interface found",
2534 			stoa(addr));
2535 		return;
2536 	}
2537 
2538 #endif
2539 	if (socket_multicast_enable(interface, lscope, addr))
2540 		msyslog(LOG_INFO,
2541 			"Added Multicast Listener %s on interface #%d %s",
2542 			stoa(addr), interface->ifnum, interface->name);
2543 	else
2544 		msyslog(LOG_ERR, "Failed to add Multicast Listener %s",
2545 			stoa(addr));
2546 #else /* MCAST */
2547 	msyslog(LOG_ERR,
2548 		"Can not add multicast address %s: no multicast support",
2549 		stoa(addr));
2550 #endif /* MCAST */
2551 	return;
2552 }
2553 
2554 
2555 /*
2556  * io_multicast_del() - delete multicast group address
2557  */
2558 void
2559 io_multicast_del(
2560 	sockaddr_u *	addr
2561 	)
2562 {
2563 #ifdef MCAST
2564 	struct interface *iface;
2565 
2566 	/*
2567 	 * Check to see if this is a multicast address
2568 	 */
2569 	if (!addr_ismulticast(addr)) {
2570 		msyslog(LOG_ERR, "invalid multicast address %s",
2571 			stoa(addr));
2572 		return;
2573 	}
2574 
2575 	/*
2576 	 * Disable reception of multicast packets
2577 	 */
2578 	while ((iface = find_flagged_addr_in_list(addr, INT_MCASTOPEN))
2579 	       != NULL)
2580 		socket_multicast_disable(iface, addr);
2581 
2582 	delete_addr_from_list(addr);
2583 
2584 #else /* not MCAST */
2585 	msyslog(LOG_ERR,
2586 		"Can not delete multicast address %s: no multicast support",
2587 		stoa(addr));
2588 #endif /* not MCAST */
2589 }
2590 
2591 
2592 /*
2593  * init_nonblocking_io() - set up descriptor to be non blocking
2594  */
2595 static void init_nonblocking_io(
2596 	SOCKET fd
2597 	)
2598 {
2599 	/*
2600 	 * set non-blocking,
2601 	 */
2602 
2603 #ifdef USE_FIONBIO
2604 	/* in vxWorks we use FIONBIO, but the others are defined for old systems, so
2605 	 * all hell breaks loose if we leave them defined
2606 	 */
2607 #undef O_NONBLOCK
2608 #undef FNDELAY
2609 #undef O_NDELAY
2610 #endif
2611 
2612 #if defined(O_NONBLOCK) /* POSIX */
2613 	if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0) {
2614 		msyslog(LOG_ERR,
2615 			"fcntl(O_NONBLOCK) fails on fd #%d: %m", fd);
2616 		exit(1);
2617 	}
2618 #elif defined(FNDELAY)
2619 	if (fcntl(fd, F_SETFL, FNDELAY) < 0) {
2620 		msyslog(LOG_ERR, "fcntl(FNDELAY) fails on fd #%d: %m",
2621 			fd);
2622 		exit(1);
2623 	}
2624 #elif defined(O_NDELAY) /* generally the same as FNDELAY */
2625 	if (fcntl(fd, F_SETFL, O_NDELAY) < 0) {
2626 		msyslog(LOG_ERR, "fcntl(O_NDELAY) fails on fd #%d: %m",
2627 			fd);
2628 		exit(1);
2629 	}
2630 #elif defined(FIONBIO)
2631 	{
2632 		int on = 1;
2633 
2634 		if (ioctl(fd, FIONBIO, &on) < 0) {
2635 			msyslog(LOG_ERR,
2636 				"ioctl(FIONBIO) fails on fd #%d: %m",
2637 				fd);
2638 			exit(1);
2639 		}
2640 	}
2641 #elif defined(FIOSNBIO)
2642 	if (ioctl(fd, FIOSNBIO, &on) < 0) {
2643 		msyslog(LOG_ERR,
2644 			"ioctl(FIOSNBIO) fails on fd #%d: %m", fd);
2645 		exit(1);
2646 	}
2647 #else
2648 # include "Bletch: Need non-blocking I/O!"
2649 #endif
2650 }
2651 
2652 /*
2653  * open_socket - open a socket, returning the file descriptor
2654  */
2655 
2656 static SOCKET
2657 open_socket(
2658 	sockaddr_u *		addr,
2659 	int			bcast,
2660 	int			turn_off_reuse,
2661 	struct interface *	interf
2662 	)
2663 {
2664 	SOCKET	fd;
2665 	int	errval;
2666 	char	scopetext[16];
2667 	/*
2668 	 * int is OK for REUSEADR per
2669 	 * http://www.kohala.com/start/mcast.api.txt
2670 	 */
2671 	int	on = 1;
2672 	int	off = 0;
2673 
2674 	if (IS_IPV6(addr) && !ipv6_works)
2675 		return INVALID_SOCKET;
2676 
2677 	/* create a datagram (UDP) socket */
2678 	fd = socket(AF(addr), SOCK_DGRAM, 0);
2679 	if (INVALID_SOCKET == fd) {
2680 #ifndef SYS_WINNT
2681 		errval = errno;
2682 #else
2683 		errval = WSAGetLastError();
2684 #endif
2685 		msyslog(LOG_ERR,
2686 			"socket(AF_INET%s, SOCK_DGRAM, 0) failed on address %s: %m",
2687 			IS_IPV6(addr) ? "6" : "", stoa(addr));
2688 
2689 		if (errval == EPROTONOSUPPORT ||
2690 		    errval == EAFNOSUPPORT ||
2691 		    errval == EPFNOSUPPORT)
2692 			return (INVALID_SOCKET);
2693 
2694 		errno = errval;
2695 		msyslog(LOG_ERR,
2696 			"unexpected socket() error %m code %d (not EPROTONOSUPPORT nor EAFNOSUPPORT nor EPFNOSUPPORT) - exiting",
2697 			errno);
2698 		exit(1);
2699 	}
2700 
2701 #ifdef SYS_WINNT
2702 	connection_reset_fix(fd, addr);
2703 #endif
2704 	/*
2705 	 * Fixup the file descriptor for some systems
2706 	 * See bug #530 for details of the issue.
2707 	 */
2708 	fd = move_fd(fd);
2709 
2710 	/*
2711 	 * set SO_REUSEADDR since we will be binding the same port
2712 	 * number on each interface according to turn_off_reuse.
2713 	 * This is undesirable on Windows versions starting with
2714 	 * Windows XP (numeric version 5.1).
2715 	 */
2716 #ifdef SYS_WINNT
2717 	if (isc_win32os_versioncheck(5, 1, 0, 0) < 0)  /* before 5.1 */
2718 #endif
2719 		if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
2720 			       (char *)((turn_off_reuse)
2721 					    ? &off
2722 					    : &on),
2723 			       sizeof(on))) {
2724 
2725 			msyslog(LOG_ERR,
2726 				"setsockopt SO_REUSEADDR %s fails for address %s: %m",
2727 				(turn_off_reuse)
2728 				    ? "off"
2729 				    : "on",
2730 				stoa(addr));
2731 			closesocket(fd);
2732 			return INVALID_SOCKET;
2733 		}
2734 #ifdef SO_EXCLUSIVEADDRUSE
2735 	/*
2736 	 * setting SO_EXCLUSIVEADDRUSE on the wildcard we open
2737 	 * first will cause more specific binds to fail.
2738 	 */
2739 	if (!(interf->flags & INT_WILDCARD))
2740 		set_excladdruse(fd);
2741 #endif
2742 
2743 	/*
2744 	 * IPv4 specific options go here
2745 	 */
2746 	if (IS_IPV4(addr)) {
2747 #if defined(HAVE_IPTOS_SUPPORT)
2748 		if (setsockopt(fd, IPPROTO_IP, IP_TOS, (char *)&qos,
2749 			       sizeof(qos)))
2750 			msyslog(LOG_ERR,
2751 				"setsockopt IP_TOS (%02x) fails on address %s: %m",
2752 				qos, stoa(addr));
2753 #endif /* HAVE_IPTOS_SUPPORT */
2754 		if (bcast)
2755 			socket_broadcast_enable(interf, fd, addr);
2756 	}
2757 
2758 	/*
2759 	 * IPv6 specific options go here
2760 	 */
2761 	if (IS_IPV6(addr)) {
2762 #if defined(IPV6_V6ONLY)
2763 		if (isc_net_probe_ipv6only() == ISC_R_SUCCESS
2764 		    && setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY,
2765 		    (char*)&on, sizeof(on)))
2766 			msyslog(LOG_ERR,
2767 				"setsockopt IPV6_V6ONLY on fails on address %s: %m",
2768 				stoa(addr));
2769 #endif /* IPV6_V6ONLY */
2770 #if defined(IPV6_BINDV6ONLY)
2771 		if (setsockopt(fd, IPPROTO_IPV6, IPV6_BINDV6ONLY,
2772 		    (char*)&on, sizeof(on)))
2773 			msyslog(LOG_ERR,
2774 				"setsockopt IPV6_BINDV6ONLY on fails on address %s: %m",
2775 				stoa(addr));
2776 #endif /* IPV6_BINDV6ONLY */
2777 	}
2778 
2779 #ifdef OS_NEEDS_REUSEADDR_FOR_IFADDRBIND
2780 	/*
2781 	 * some OSes don't allow binding to more specific
2782 	 * addresses if a wildcard address already bound
2783 	 * to the port and SO_REUSEADDR is not set
2784 	 */
2785 	if (!is_wildcard_addr(addr))
2786 		set_wildcard_reuse(AF(addr), 1);
2787 #endif
2788 
2789 	/*
2790 	 * bind the local address.
2791 	 */
2792 	errval = bind(fd, &addr->sa, SOCKLEN(addr));
2793 
2794 #ifdef OS_NEEDS_REUSEADDR_FOR_IFADDRBIND
2795 	if (!is_wildcard_addr(addr))
2796 		set_wildcard_reuse(AF(addr), 0);
2797 #endif
2798 
2799 	if (errval < 0) {
2800 		/*
2801 		 * Don't log this under all conditions
2802 		 */
2803 		if (turn_off_reuse == 0
2804 #ifdef DEBUG
2805 		    || debug > 1
2806 #endif
2807 		    ) {
2808 			if (SCOPE(addr))
2809 				snprintf(scopetext, sizeof(scopetext),
2810 					 "%%%d", SCOPE(addr));
2811 			else
2812 				scopetext[0] = 0;
2813 
2814 			msyslog(LOG_ERR,
2815 				"bind(%d) AF_INET%s %s%s#%d%s flags 0x%x failed: %m",
2816 				fd, IS_IPV6(addr) ? "6" : "",
2817 				stoa(addr), scopetext, SRCPORT(addr),
2818 				IS_MCAST(addr) ? " (multicast)" : "",
2819 				interf->flags);
2820 		}
2821 
2822 		closesocket(fd);
2823 
2824 		return INVALID_SOCKET;
2825 	}
2826 
2827 #ifdef HAVE_TIMESTAMP
2828 	{
2829 		if (setsockopt(fd, SOL_SOCKET, SO_TIMESTAMP,
2830 			       (char*)&on, sizeof(on)))
2831 			msyslog(LOG_DEBUG,
2832 				"setsockopt SO_TIMESTAMP on fails on address %s: %m",
2833 				stoa(addr));
2834 		else
2835 			DPRINTF(4, ("setsockopt SO_TIMESTAMP enabled on fd %d address %s\n",
2836 				    fd, stoa(addr)));
2837 	}
2838 #endif
2839 	DPRINTF(4, ("bind(%d) AF_INET%s, addr %s%%%d#%d, flags 0x%x\n",
2840 		   fd, IS_IPV6(addr) ? "6" : "", stoa(addr),
2841 		   SCOPE(addr), SRCPORT(addr), interf->flags));
2842 
2843 	init_nonblocking_io(fd);
2844 
2845 #ifdef HAVE_SIGNALED_IO
2846 	init_socket_sig(fd);
2847 #endif /* not HAVE_SIGNALED_IO */
2848 
2849 	add_fd_to_list(fd, FD_TYPE_SOCKET);
2850 
2851 #if !defined(SYS_WINNT) && !defined(VMS)
2852 	DPRINTF(4, ("flags for fd %d: 0x%x\n", fd,
2853 		    fcntl(fd, F_GETFL, 0)));
2854 #endif /* SYS_WINNT || VMS */
2855 
2856 #if defined (HAVE_IO_COMPLETION_PORT)
2857 /*
2858  * Add the socket to the completion port
2859  */
2860 	if (io_completion_port_add_socket(fd, interf)) {
2861 		msyslog(LOG_ERR, "unable to set up io completion port - EXITING");
2862 		exit(1);
2863 	}
2864 #endif
2865 	return fd;
2866 }
2867 
2868 /* XXX ELIMINATE sendpkt similar in ntpq.c, ntpdc.c, ntp_io.c, ntptrace.c */
2869 /*
2870  * sendpkt - send a packet to the specified destination. Maintain a
2871  * send error cache so that only the first consecutive error for a
2872  * destination is logged.
2873  */
2874 void
2875 sendpkt(
2876 	sockaddr_u *dest,
2877 	struct interface *inter,
2878 	int ttl,
2879 	struct pkt *pkt,
2880 	int len
2881 	)
2882 {
2883 	int cc;
2884 
2885 	if (NULL == inter) {
2886 		/*
2887 		 * unbound peer - drop request and wait for better
2888 		 * network conditions
2889 		 */
2890 		DPRINTF(2, ("%ssendpkt(dst=%s, ttl=%d, len=%d): no interface - IGNORED\n",
2891 			    (ttl > 0) ? "\tMCAST\t***** " : "",
2892 			    stoa(dest), ttl, len));
2893 		return;
2894 	}
2895 
2896 	DPRINTF(2, ("%ssendpkt(%d, dst=%s, src=%s, ttl=%d, len=%d)\n",
2897 		    (ttl > 0) ? "\tMCAST\t***** " : "",
2898 		    inter->fd, stoa(dest), stoa(&inter->sin),
2899 		    ttl, len));
2900 
2901 #ifdef MCAST
2902 	/*
2903 	 * for the moment we use the bcast option to set multicast ttl
2904 	 */
2905 	if (ttl > 0 && ttl != inter->last_ttl) {
2906 		/*
2907 		 * set the multicast ttl for outgoing packets
2908 		 */
2909 		int	rtc;
2910 		u_char	cttl;
2911 #ifdef INCLUDE_IPV6_SUPPORT
2912 		u_int	uttl;
2913 #endif
2914 
2915 		switch (AF(&inter->sin)) {
2916 
2917 		case AF_INET :
2918 			cttl = (u_char)ttl;
2919 			rtc = setsockopt(inter->fd, IPPROTO_IP,
2920 					 IP_MULTICAST_TTL,
2921 					 (void *)&cttl, sizeof(cttl));
2922 			break;
2923 
2924 #ifdef INCLUDE_IPV6_SUPPORT
2925 		case AF_INET6 :
2926 			uttl = (u_int)ttl;
2927 			rtc = setsockopt(inter->fd, IPPROTO_IPV6,
2928 					 IPV6_MULTICAST_HOPS,
2929 					 (void *)&uttl, sizeof(uttl));
2930 			break;
2931 #endif /* INCLUDE_IPV6_SUPPORT */
2932 
2933 		default:	/* just NOP if not supported */
2934 			DPRINTF(1, ("sendpkt unknown AF %d",
2935 				    AF(&inter->sin)));
2936 			rtc = 0;
2937 		}
2938 
2939 		if (!rtc)
2940 			inter->last_ttl = ttl;
2941 		else
2942 			msyslog(LOG_ERR,
2943 				"setsockopt IP_MULTICAST_TTL/IPV6_MULTICAST_HOPS fails on address %s: %m",
2944 				stoa(&inter->sin));
2945 	}
2946 
2947 #endif /* MCAST */
2948 
2949 #if defined(HAVE_IO_COMPLETION_PORT)
2950 	cc = io_completion_port_sendto(inter, pkt, len, dest);
2951 	if (cc != ERROR_SUCCESS) {
2952 #else
2953 #ifdef SIM
2954 	cc = simulate_server(dest, inter, pkt);
2955 #else /* SIM */
2956 	cc = sendto(inter->fd, (char *)pkt, (unsigned int)len, 0,
2957 		    (struct sockaddr *)dest, SOCKLEN(dest));
2958 #endif /* SIM */
2959 	if (cc == -1) {
2960 #endif
2961 		inter->notsent++;
2962 		packets_notsent++;
2963 	} else	{
2964 		inter->sent++;
2965 		packets_sent++;
2966 	}
2967 }
2968 
2969 
2970 #if !defined(HAVE_IO_COMPLETION_PORT)
2971 /*
2972  * fdbits - generate ascii representation of fd_set (FAU debug support)
2973  * HFDF format - highest fd first.
2974  */
2975 static char *
2976 fdbits(
2977 	int count,
2978 	fd_set *set
2979 	)
2980 {
2981 	static char buffer[256];
2982 	char * buf = buffer;
2983 
2984 	count = min(count,  255);
2985 
2986 	while (count >= 0) {
2987 		*buf++ = FD_ISSET(count, set) ? '#' : '-';
2988 		count--;
2989 	}
2990 	*buf = '\0';
2991 
2992 	return buffer;
2993 }
2994 
2995 /*
2996  * Routine to read the refclock packets for a specific interface
2997  * Return the number of bytes read. That way we know if we should
2998  * read it again or go on to the next one if no bytes returned
2999  */
3000 static inline int
3001 read_refclock_packet(SOCKET fd, struct refclockio *rp, l_fp ts)
3002 {
3003 	int i;
3004 	int buflen;
3005 	register struct recvbuf *rb;
3006 
3007 	rb = get_free_recv_buffer();
3008 
3009 	if (NULL == rb) {
3010 		/*
3011 		 * No buffer space available - just drop the packet
3012 		 */
3013 		char buf[RX_BUFF_SIZE];
3014 
3015 		buflen = read(fd, buf, sizeof buf);
3016 		packets_dropped++;
3017 		return (buflen);
3018 	}
3019 
3020 	i = (rp->datalen == 0
3021 	     || rp->datalen > (int)sizeof(rb->recv_space))
3022 		? (int)sizeof(rb->recv_space)
3023 		: rp->datalen;
3024 	buflen = read(fd, (char *)&rb->recv_space, (unsigned)i);
3025 
3026 	if (buflen < 0) {
3027 		if (errno != EINTR && errno != EAGAIN)
3028 			msyslog(LOG_ERR, "clock read fd %d: %m", fd);
3029 		freerecvbuf(rb);
3030 		return (buflen);
3031 	}
3032 
3033 	/*
3034 	 * Got one. Mark how and when it got here,
3035 	 * put it on the full list and do bookkeeping.
3036 	 */
3037 	rb->recv_length = buflen;
3038 	rb->recv_srcclock = rp->srcclock;
3039 	rb->dstadr = 0;
3040 	rb->fd = fd;
3041 	rb->recv_time = ts;
3042 	rb->receiver = rp->clock_recv;
3043 
3044 	if (rp->io_input) {
3045 		/*
3046 		 * have direct input routine for refclocks
3047 		 */
3048 		if (rp->io_input(rb) == 0) {
3049 			/*
3050 			 * data was consumed - nothing to pass up
3051 			 * into block input machine
3052 			 */
3053 			freerecvbuf(rb);
3054 			return (buflen);
3055 		}
3056 	}
3057 
3058 	add_full_recv_buffer(rb);
3059 
3060 	rp->recvcount++;
3061 	packets_received++;
3062 	return (buflen);
3063 }
3064 
3065 
3066 #ifdef HAVE_TIMESTAMP
3067 /*
3068  * extract timestamps from control message buffer
3069  */
3070 static l_fp
3071 fetch_timestamp(
3072 	struct recvbuf *	rb,
3073 	struct msghdr *		msghdr,
3074 	l_fp			ts
3075 	)
3076 {
3077 #ifdef USE_TIMESTAMP_CMSG
3078 	struct cmsghdr *cmsghdr;
3079 
3080 	cmsghdr = CMSG_FIRSTHDR(msghdr);
3081 	while (cmsghdr != NULL) {
3082 		switch (cmsghdr->cmsg_type)
3083 		{
3084 		case SCM_TIMESTAMP:
3085 		{
3086 			struct timeval *tvp;
3087 			double dtemp;
3088 			l_fp nts;
3089 
3090 			tvp = (struct timeval *)CMSG_DATA(cmsghdr);
3091 			DPRINTF(4, ("fetch_timestamp: system network time stamp: %lld.%06ld\n",
3092 				    (long long)tvp->tv_sec, (long)tvp->tv_usec));
3093 			nts.l_i = tvp->tv_sec + JAN_1970;
3094 			dtemp = (tvp->tv_usec
3095 				 + (ntp_random() * 2. / FRAC)) / 1e6;
3096 			nts.l_uf = (u_int32)(dtemp * FRAC);
3097 #ifdef DEBUG_TIMING
3098 			{
3099 				l_fp dts;
3100 
3101 				dts = ts;
3102 				L_SUB(&dts, &nts);
3103 				collect_timing(rb,
3104 					       "input processing delay",
3105 					       1, &dts);
3106 				DPRINTF(4, ("fetch_timestamp: timestamp delta: %s (incl. prec fuzz)\n",
3107 					    lfptoa(&dts, 9)));
3108 			}
3109 #endif
3110 			ts = nts;  /* network time stamp */
3111 			break;
3112 		}
3113 		default:
3114 			DPRINTF(4, ("fetch_timestamp: skipping control message 0x%x\n",
3115 				    cmsghdr->cmsg_type));
3116 		}
3117 		cmsghdr = CMSG_NXTHDR(msghdr, cmsghdr);
3118 	}
3119 #endif
3120 	return ts;
3121 }
3122 #endif
3123 
3124 
3125 /*
3126  * Routine to read the network NTP packets for a specific interface
3127  * Return the number of bytes read. That way we know if we should
3128  * read it again or go on to the next one if no bytes returned
3129  */
3130 static inline int
3131 read_network_packet(
3132 	SOCKET			fd,
3133 	struct interface *	itf,
3134 	l_fp			ts
3135 	)
3136 {
3137 	GETSOCKNAME_SOCKLEN_TYPE fromlen;
3138 	int buflen;
3139 	register struct recvbuf *rb;
3140 #ifdef HAVE_TIMESTAMP
3141 	struct msghdr msghdr;
3142 	struct iovec iovec;
3143 	char control[TIMESTAMP_CTLMSGBUF_SIZE];
3144 #endif
3145 
3146 	/*
3147 	 * Get a buffer and read the frame.  If we
3148 	 * haven't got a buffer, or this is received
3149 	 * on a disallowed socket, just dump the
3150 	 * packet.
3151 	 */
3152 
3153 	rb = get_free_recv_buffer();
3154 	if (NULL == rb || itf->ignore_packets) {
3155 		char buf[RX_BUFF_SIZE];
3156 		sockaddr_u from;
3157 
3158 		if (rb != NULL)
3159 			freerecvbuf(rb);
3160 
3161 		fromlen = sizeof(from);
3162 		buflen = recvfrom(fd, buf, sizeof(buf), 0,
3163 				  &from.sa, &fromlen);
3164 		DPRINTF(4, ("%s on (%lu) fd=%d from %s\n",
3165 			(itf->ignore_packets)
3166 			    ? "ignore"
3167 			    : "drop",
3168 			free_recvbuffs(), fd, stoa(&from)));
3169 		if (itf->ignore_packets)
3170 			packets_ignored++;
3171 		else
3172 			packets_dropped++;
3173 		return (buflen);
3174 	}
3175 
3176 	fromlen = sizeof(rb->recv_srcadr);
3177 
3178 #ifndef HAVE_TIMESTAMP
3179 	rb->recv_length = recvfrom(fd, (char *)&rb->recv_space,
3180 				   sizeof(rb->recv_space), 0,
3181 				   &rb->recv_srcadr.sa, &fromlen);
3182 #else
3183 	iovec.iov_base        = &rb->recv_space;
3184 	iovec.iov_len         = sizeof(rb->recv_space);
3185 	msghdr.msg_name       = &rb->recv_srcadr;
3186 	msghdr.msg_namelen    = fromlen;
3187 	msghdr.msg_iov        = &iovec;
3188 	msghdr.msg_iovlen     = 1;
3189 	msghdr.msg_control    = (void *)&control;
3190 	msghdr.msg_controllen = sizeof(control);
3191 	msghdr.msg_flags      = 0;
3192 	rb->recv_length       = recvmsg(fd, &msghdr, 0);
3193 #endif
3194 
3195 	buflen = rb->recv_length;
3196 
3197 	if (buflen == 0 || (buflen == -1 &&
3198 	    (EWOULDBLOCK == errno
3199 #ifdef EAGAIN
3200 	     || EAGAIN == errno
3201 #endif
3202 	     ))) {
3203 		freerecvbuf(rb);
3204 		return (buflen);
3205 	} else if (buflen < 0) {
3206 		msyslog(LOG_ERR, "recvfrom(%s) fd=%d: %m",
3207 			stoa(&rb->recv_srcadr), fd);
3208 		DPRINTF(5, ("read_network_packet: fd=%d dropped (bad recvfrom)\n",
3209 			    fd));
3210 		freerecvbuf(rb);
3211 		return (buflen);
3212 	}
3213 
3214 	DPRINTF(3, ("read_network_packet: fd=%d length %d from %s\n",
3215 		    fd, buflen, stoa(&rb->recv_srcadr)));
3216 
3217 	/*
3218 	 * Got one.  Mark how and when it got here,
3219 	 * put it on the full list and do bookkeeping.
3220 	 */
3221 	rb->dstadr = itf;
3222 	rb->fd = fd;
3223 #ifdef HAVE_TIMESTAMP
3224 	/* pick up a network time stamp if possible */
3225 	ts = fetch_timestamp(rb, &msghdr, ts);
3226 #endif
3227 	rb->recv_time = ts;
3228 	rb->receiver = receive;
3229 
3230 	add_full_recv_buffer(rb);
3231 
3232 	itf->received++;
3233 	packets_received++;
3234 	return (buflen);
3235 }
3236 
3237 
3238 /*
3239  * input_handler - receive packets asynchronously
3240  */
3241 void
3242 input_handler(
3243 	l_fp *cts
3244 	)
3245 {
3246 	int buflen;
3247 	int n;
3248 	int doing;
3249 	SOCKET fd;
3250 	struct timeval tvzero;
3251 	l_fp ts;		/* Timestamp at BOselect() gob */
3252 #ifdef DEBUG_TIMING
3253 	l_fp ts_e;		/* Timestamp at EOselect() gob */
3254 #endif
3255 	fd_set fds;
3256 	int select_count = 0;
3257 	struct interface *interface;
3258 #if defined(HAS_ROUTING_SOCKET)
3259 	struct asyncio_reader *asyncio_reader;
3260 #endif
3261 
3262 	handler_calls++;
3263 
3264 	/*
3265 	 * If we have something to do, freeze a timestamp.
3266 	 * See below for the other cases (nothing left to do or error)
3267 	 */
3268 	ts = *cts;
3269 
3270 	/*
3271 	 * Do a poll to see who has data
3272 	 */
3273 
3274 	fds = activefds;
3275 	tvzero.tv_sec = tvzero.tv_usec = 0;
3276 
3277 	n = select(maxactivefd + 1, &fds, (fd_set *)0, (fd_set *)0,
3278 		   &tvzero);
3279 
3280 	/*
3281 	 * If there are no packets waiting just return
3282 	 */
3283 	if (n < 0) {
3284 		int err = errno;
3285 		/*
3286 		 * extended FAU debugging output
3287 		 */
3288 		if (err != EINTR)
3289 			msyslog(LOG_ERR,
3290 				"select(%d, %s, 0L, 0L, &0.0) error: %m",
3291 				maxactivefd + 1,
3292 				fdbits(maxactivefd, &activefds));
3293 		if (err == EBADF) {
3294 			int j, b;
3295 			fds = activefds;
3296 			for (j = 0; j <= maxactivefd; j++)
3297 				if ((FD_ISSET(j, &fds)
3298 				    && (read(j, &b, 0) == -1)))
3299 					msyslog(LOG_ERR,
3300 						"Bad file descriptor %d",
3301 						j);
3302 		}
3303 		return;
3304 	}
3305 	else if (n == 0)
3306 		return;
3307 
3308 	++handler_pkts;
3309 
3310 #ifdef REFCLOCK
3311 	/*
3312 	 * Check out the reference clocks first, if any
3313 	 */
3314 
3315 	if (refio != NULL) {
3316 		register struct refclockio *rp;
3317 
3318 		for (rp = refio; rp != NULL; rp = rp->next) {
3319 			fd = rp->fd;
3320 
3321 			if (FD_ISSET(fd, &fds))
3322 				do {
3323 					++select_count;
3324 					buflen = read_refclock_packet(
3325 							fd, rp, ts);
3326 				} while (buflen > 0);
3327 		}
3328 	}
3329 #endif /* REFCLOCK */
3330 
3331 	/*
3332 	 * Loop through the interfaces looking for data to read.
3333 	 */
3334 	for (interface = inter_list;
3335 	     interface != NULL;
3336 	     interface = interface->link) {
3337 
3338 		for (doing = 0; (doing < 2); doing++) {
3339 			if (!doing)
3340 				fd = interface->fd;
3341 			else {
3342 				if (!(interface->flags & INT_BCASTOPEN))
3343 					break;
3344 				fd = interface->bfd;
3345 			}
3346 			if (fd < 0)
3347 				continue;
3348 			if (FD_ISSET(fd, &fds))
3349 				do {
3350 					++select_count;
3351 					buflen = read_network_packet(
3352 							fd, interface,
3353 							ts);
3354 				} while (buflen > 0);
3355 			/* Check more interfaces */
3356 		}
3357 	}
3358 
3359 #ifdef HAS_ROUTING_SOCKET
3360 	/*
3361 	 * scan list of asyncio readers - currently only used for routing sockets
3362 	 */
3363 	asyncio_reader = asyncio_reader_list;
3364 
3365 	while (asyncio_reader != NULL) {
3366 		struct asyncio_reader *next = asyncio_reader->link;
3367 
3368 		if (FD_ISSET(asyncio_reader->fd, &fds)) {
3369 			++select_count;
3370 			(asyncio_reader->receiver)(asyncio_reader);
3371 		}
3372 
3373 		asyncio_reader = next;
3374 	}
3375 #endif /* HAS_ROUTING_SOCKET */
3376 
3377 	/*
3378 	 * Done everything from that select.
3379 	 */
3380 
3381 	/*
3382 	 * If nothing to do, just return.
3383 	 * If an error occurred, complain and return.
3384 	 */
3385 	if (select_count == 0) { /* We really had nothing to do */
3386 #ifdef DEBUG
3387 		if (debug)
3388 			msyslog(LOG_DEBUG, "input_handler: select() returned 0");
3389 #endif
3390 		return;
3391 	}
3392 	/* We've done our work */
3393 #ifdef DEBUG_TIMING
3394 	get_systime(&ts_e);
3395 	/*
3396 	 * (ts_e - ts) is the amount of time we spent
3397 	 * processing this gob of file descriptors.  Log
3398 	 * it.
3399 	 */
3400 	L_SUB(&ts_e, &ts);
3401 	collect_timing(NULL, "input handler", 1, &ts_e);
3402 	if (debug > 3)
3403 		msyslog(LOG_DEBUG,
3404 			"input_handler: Processed a gob of fd's in %s msec",
3405 			lfptoms(&ts_e, 6));
3406 #endif
3407 	/* just bail. */
3408 	return;
3409 }
3410 #endif
3411 
3412 /*
3413  * findinterface - find local interface corresponding to address
3414  */
3415 struct interface *
3416 findinterface(
3417 	sockaddr_u *addr
3418 	)
3419 {
3420 	struct interface *iface;
3421 
3422 	iface = findlocalinterface(addr, INT_WILDCARD, 0);
3423 
3424 	if (NULL == iface) {
3425 		DPRINTF(4, ("Found no interface for address %s - returning wildcard\n",
3426 			    stoa(addr)));
3427 
3428 		iface = ANY_INTERFACE_CHOOSE(addr);
3429 	} else
3430 		DPRINTF(4, ("Found interface #%d %s for address %s\n",
3431 			    iface->ifnum, iface->name, stoa(addr)));
3432 
3433 	return iface;
3434 }
3435 
3436 /*
3437  * findlocalinterface - find local interface corresponding to addr,
3438  * which does not have any of flags set.  If bast is nonzero, addr is
3439  * a broadcast address.
3440  *
3441  * This code attempts to find the local sending address for an outgoing
3442  * address by connecting a new socket to destinationaddress:NTP_PORT
3443  * and reading the sockname of the resulting connect.
3444  * the complicated sequence simulates the routing table lookup
3445  * for to first hop without duplicating any of the routing logic into
3446  * ntpd. preferably we would have used an API call - but its not there -
3447  * so this is the best we can do here short of duplicating to entire routing
3448  * logic in ntpd which would be a silly and really unportable thing to do.
3449  *
3450  */
3451 static struct interface *
3452 findlocalinterface(
3453 	sockaddr_u *	addr,
3454 	int		flags,
3455 	int		bcast
3456 	)
3457 {
3458 	GETSOCKNAME_SOCKLEN_TYPE	sockaddrlen;
3459 	struct interface *		iface;
3460 	sockaddr_u			saddr;
3461 	SOCKET				s;
3462 	int				rtn;
3463 	int				on;
3464 
3465 	DPRINTF(4, ("Finding interface for addr %s in list of addresses\n",
3466 		    stoa(addr)));
3467 
3468 	s = socket(AF(addr), SOCK_DGRAM, 0);
3469 	if (INVALID_SOCKET == s)
3470 		return NULL;
3471 
3472 	/*
3473 	 * If we are looking for broadcast interface we need to set this
3474 	 * socket to allow broadcast
3475 	 */
3476 	if (bcast) {
3477 		on = 1;
3478 		setsockopt(s, SOL_SOCKET, SO_BROADCAST,
3479 			   (char *)&on, sizeof(on));
3480 	}
3481 
3482 	rtn = connect(s, &addr->sa, SOCKLEN(addr));
3483 	if (SOCKET_ERROR == rtn) {
3484 		closesocket(s);
3485 		return NULL;
3486 	}
3487 
3488 	sockaddrlen = sizeof(saddr);
3489 	rtn = getsockname(s, &saddr.sa, &sockaddrlen);
3490 	closesocket(s);
3491 
3492 	if (SOCKET_ERROR == rtn)
3493 		return NULL;
3494 
3495 	DPRINTF(4, ("findlocalinterface: kernel maps %s to %s\n",
3496 		    stoa(addr), stoa(&saddr)));
3497 
3498 	iface = getinterface(&saddr, flags);
3499 
3500 	/*
3501 	 * if we didn't find an exact match on saddr check for an
3502 	 * interface on the same subnet as saddr.  This handles the
3503 	 * case of the address suggested by the kernel being
3504 	 * excluded by the user's -I and -L options to ntpd, when
3505 	 * another address is enabled on the same subnet.
3506 	 * See http://bugs.ntp.org/1184 for more detail.
3507 	 */
3508 	if (NULL == iface || iface->ignore_packets)
3509 		iface = getsamenetinterface(&saddr, flags);
3510 
3511 	/* Don't use an interface which will ignore replies */
3512 	if (iface != NULL && iface->ignore_packets)
3513 		iface = NULL;
3514 
3515 	return iface;
3516 }
3517 
3518 
3519 /*
3520  * fetch an interface structure the matches the
3521  * address and has the given flags NOT set
3522  */
3523 static struct interface *
3524 getinterface(
3525 	sockaddr_u *	addr,
3526 	int		flags
3527 	)
3528 {
3529 	struct interface *iface;
3530 
3531 	iface = find_addr_in_list(addr);
3532 
3533 	if (iface != NULL && (iface->flags & flags))
3534 		iface = NULL;
3535 
3536 	return iface;
3537 }
3538 
3539 
3540 /*
3541  * fetch an interface structure with a local address on the same subnet
3542  * as addr which has the given flags NOT set
3543  */
3544 static struct interface *
3545 getsamenetinterface(
3546 	sockaddr_u *	addr,
3547 	int		flags
3548 	)
3549 {
3550 	struct interface *iface;
3551 
3552 	iface = find_samenet_addr_in_list(addr);
3553 
3554 	if (iface != NULL && (iface->flags & flags))
3555 		iface = NULL;
3556 
3557 	return iface;
3558 }
3559 
3560 
3561 /*
3562  * findlocalcastinterface - find local *cast interface for addr
3563  */
3564 static struct interface *
3565 findlocalcastinterface(
3566 	sockaddr_u *	addr
3567 	)
3568 {
3569 	struct interface *	iface;
3570 	struct interface *	nif;
3571 #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
3572 	isc_boolean_t		want_linklocal;
3573 #endif
3574 
3575 	NTP_REQUIRE(addr_ismulticast(addr));
3576 
3577 	/*
3578 	 * see how kernel maps the mcast address
3579 	 */
3580 	nif = findlocalinterface(addr, 0, 0);
3581 
3582 	if (nif != NULL && !nif->ignore_packets) {
3583 		DPRINTF(2, ("findlocalcastinterface: kernel recommends interface #%d %s for %s\n",
3584 			    nif->ifnum, nif->name, stoa(addr)));
3585 		return nif;
3586 	}
3587 
3588 #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
3589 	want_linklocal = (IN6_IS_ADDR_MC_LINKLOCAL(PSOCK_ADDR6(addr))
3590 		       || IN6_IS_ADDR_MC_SITELOCAL(PSOCK_ADDR6(addr)));
3591 #endif
3592 
3593 	for (iface = inter_list;
3594 	     iface != NULL;
3595 	     iface = iface->link)
3596 	  {
3597 		/* use only allowed addresses */
3598 		if (iface->ignore_packets)
3599 			continue;
3600 
3601 		/* Skip the loopback and wildcard addresses */
3602 		if (iface->flags & (INT_LOOPBACK | INT_WILDCARD))
3603 			continue;
3604 
3605 		/* Skip if different family */
3606 		if (AF(&iface->sin) != AF(addr))
3607 			continue;
3608 
3609 		/* Is it multicast capable? */
3610 		if (!(iface->flags & INT_MULTICAST))
3611 			continue;
3612 
3613 #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
3614 		if (want_linklocal && IS_IPV6(&iface->sin) &&
3615 		    IN6_IS_ADDR_LINKLOCAL(PSOCK_ADDR6(&iface->sin))) {
3616 			nif = iface;
3617 			break;
3618 		}
3619 		/* If we want a linklocal address, skip */
3620 		if (want_linklocal)
3621 			continue;
3622 #endif
3623 		nif = iface;
3624 		break;
3625 	}	/* for loop over interfaces */
3626 
3627 	if (nif != NULL)
3628 		DPRINTF(3, ("findlocalcastinterface: found interface #%d %s for %s\n",
3629 			    nif->ifnum, nif->name, stoa(addr)));
3630 	else
3631 		DPRINTF(3, ("findlocalcastinterface: no interface found for %s\n",
3632 			    stoa(addr)));
3633 	return nif;
3634 }
3635 
3636 
3637 /*
3638  * findbcastinter - find broadcast interface corresponding to address
3639  */
3640 struct interface *
3641 findbcastinter(
3642 	sockaddr_u *addr
3643 	)
3644 {
3645 #if !defined(MPE) && (defined(SIOCGIFCONF) || defined(SYS_WINNT))
3646 	struct interface *iface;
3647 
3648 
3649 	DPRINTF(4, ("Finding broadcast/multicast interface for addr %s in list of addresses\n",
3650 		    stoa(addr)));
3651 
3652 	iface = findlocalinterface(addr, INT_LOOPBACK | INT_WILDCARD,
3653 				   1);
3654 	if (iface != NULL) {
3655 		DPRINTF(4, ("Easily found bcast-/mcast- interface index #%d %s\n",
3656 			    iface->ifnum, iface->name));
3657 		return iface;
3658 	}
3659 
3660 	/*
3661 	 * plan B - try to find something reasonable in our lists in
3662 	 * case kernel lookup doesn't help
3663 	 */
3664 	for (iface = inter_list;
3665 	     iface != NULL;
3666 	     iface = iface->link)
3667 	{
3668 		if (iface->flags & INT_WILDCARD)
3669 			continue;
3670 
3671 		/* Don't bother with ignored interfaces */
3672 		if (iface->ignore_packets)
3673 			continue;
3674 
3675 		/*
3676 		 * First look if this is the correct family
3677 		 */
3678 		if(AF(&iface->sin) != AF(addr))
3679 			continue;
3680 
3681 		/* Skip the loopback addresses */
3682 		if (iface->flags & INT_LOOPBACK)
3683 			continue;
3684 
3685 		/*
3686 		 * If we are looking to match a multicast address and
3687 		 * this interface is one...
3688 		 */
3689 		if (addr_ismulticast(addr)
3690 		    && (iface->flags & INT_MULTICAST)) {
3691 #ifdef INCLUDE_IPV6_SUPPORT
3692 			/*
3693 			 * ...it is the winner unless we're looking for
3694 			 * an interface to use for link-local multicast
3695 			 * and its address is not link-local.
3696 			 */
3697 			if (IS_IPV6(addr)
3698 			    && IN6_IS_ADDR_MC_LINKLOCAL(PSOCK_ADDR6(addr))
3699 			    && !IN6_IS_ADDR_LINKLOCAL(PSOCK_ADDR6(&iface->sin)))
3700 				continue;
3701 #endif
3702 			break;
3703 		}
3704 
3705 		/*
3706 		 * We match only those interfaces marked as
3707 		 * broadcastable and either the explicit broadcast
3708 		 * address or the network portion of the IP address.
3709 		 * Sloppy.
3710 		 */
3711 		if (IS_IPV4(addr)) {
3712 			if (SOCK_EQ(&iface->bcast, addr))
3713 				break;
3714 
3715 			if ((NSRCADR(&iface->sin) & NSRCADR(&iface->mask))
3716 			    == (NSRCADR(addr)	  & NSRCADR(&iface->mask)))
3717 				break;
3718 		}
3719 #ifdef INCLUDE_IPV6_SUPPORT
3720 		else if(IS_IPV6(addr)) {
3721 			if (SOCK_EQ(&iface->bcast, addr))
3722 				break;
3723 
3724 			if (SOCK_EQ(netof(&iface->sin), netof(addr)))
3725 				break;
3726 		}
3727 #endif
3728 	}
3729 #endif /* SIOCGIFCONF */
3730 	if (NULL == iface) {
3731 		DPRINTF(4, ("No bcast interface found for %s\n",
3732 			    stoa(addr)));
3733 		iface = ANY_INTERFACE_CHOOSE(addr);
3734 	} else
3735 		DPRINTF(4, ("Found bcast-/mcast- interface index #%d %s\n",
3736 			    iface->ifnum, iface->name));
3737 	return iface;
3738 }
3739 
3740 
3741 /*
3742  * io_clr_stats - clear I/O module statistics
3743  */
3744 void
3745 io_clr_stats(void)
3746 {
3747 	packets_dropped = 0;
3748 	packets_ignored = 0;
3749 	packets_received = 0;
3750 	packets_sent = 0;
3751 	packets_notsent = 0;
3752 
3753 	handler_calls = 0;
3754 	handler_pkts = 0;
3755 	io_timereset = current_time;
3756 }
3757 
3758 
3759 #ifdef REFCLOCK
3760 /*
3761  * io_addclock - add a reference clock to the list and arrange that we
3762  *				 get SIGIO interrupts from it.
3763  */
3764 int
3765 io_addclock(
3766 	struct refclockio *rio
3767 	)
3768 {
3769 	BLOCKIO();
3770 
3771 	/*
3772 	 * Stuff the I/O structure in the list and mark the descriptor
3773 	 * in use.  There is a harmless (I hope) race condition here.
3774 	 */
3775 	rio->next = refio;
3776 
3777 # ifdef HAVE_SIGNALED_IO
3778 	if (init_clock_sig(rio)) {
3779 		UNBLOCKIO();
3780 		return 0;
3781 	}
3782 # elif defined(HAVE_IO_COMPLETION_PORT)
3783 	if (io_completion_port_add_clock_io(rio)) {
3784 		UNBLOCKIO();
3785 		return 0;
3786 	}
3787 # endif
3788 
3789 	/*
3790 	 * enqueue
3791 	 */
3792 	refio = rio;
3793 
3794 	/*
3795 	 * register fd
3796 	 */
3797 	add_fd_to_list(rio->fd, FD_TYPE_FILE);
3798 
3799 	UNBLOCKIO();
3800 	return 1;
3801 }
3802 
3803 /*
3804  * io_closeclock - close the clock in the I/O structure given
3805  */
3806 void
3807 io_closeclock(
3808 	struct refclockio *rio
3809 	)
3810 {
3811 	register struct refclockio *rp;
3812 
3813 	BLOCKIO();
3814 
3815 	/*
3816 	 * Remove structure from the list
3817 	 */
3818 	if (refio == rio)
3819 		refio = rio->next;
3820 	else {
3821 		for (rp = refio; rp != NULL; rp = rp->next)
3822 			if (rp->next == rio) {
3823 				rp->next = rio->next;
3824 				break;
3825 			}
3826 
3827 		if (NULL == rp) {
3828 			UNBLOCKIO();
3829 			return;
3830 		}
3831 	}
3832 
3833 	/*
3834 	 * Close the descriptor.
3835 	 */
3836 	close_and_delete_fd_from_list(rio->fd);
3837 	UNBLOCKIO();
3838 }
3839 #endif	/* REFCLOCK */
3840 
3841 /*
3842  * On NT a SOCKET is an unsigned int so we cannot possibly keep it in
3843  * an array. So we use one of the ISC_LIST functions to hold the
3844  * socket value and use that when we want to enumerate it.
3845  *
3846  * This routine is called by the forked intres child process to close
3847  * all open sockets.  On Windows there's no need as intres runs in
3848  * the same process as a thread.
3849  */
3850 #ifndef SYS_WINNT
3851 void
3852 kill_asyncio(int startfd)
3853 {
3854 	BLOCKIO();
3855 
3856 	/*
3857 	 * In the child process we do not maintain activefds and
3858 	 * maxactivefd.  Zeroing maxactivefd disables code which
3859 	 * maintains it in close_and_delete_fd_from_list().
3860 	 */
3861 	maxactivefd = 0;
3862 
3863 	while (fd_list != NULL)
3864 		close_and_delete_fd_from_list(fd_list->fd);
3865 
3866 	UNBLOCKIO();
3867 }
3868 #endif	/* !SYS_WINNT */
3869 
3870 /*
3871  * Add and delete functions for the list of open sockets
3872  */
3873 static void
3874 add_fd_to_list(
3875 	SOCKET fd,
3876 	enum desc_type type
3877 	)
3878 {
3879 	vsock_t *lsock = emalloc(sizeof(*lsock));
3880 
3881 	lsock->fd = fd;
3882 	lsock->type = type;
3883 
3884 	LINK_SLIST(fd_list, lsock, link);
3885 	/*
3886 	 * I/O Completion Ports don't care about the select and FD_SET
3887 	 */
3888 #ifndef HAVE_IO_COMPLETION_PORT
3889 	if (fd < 0 || fd >= FD_SETSIZE) {
3890 		msyslog(LOG_ERR,
3891 			"Too many sockets in use, FD_SETSIZE %d exceeded",
3892 			FD_SETSIZE);
3893 		exit(1);
3894 	}
3895 	/*
3896 	 * keep activefds in sync
3897 	 */
3898 	maxactivefd = max(fd, maxactivefd);
3899 
3900 	FD_SET(fd, &activefds);
3901 #endif
3902 }
3903 
3904 static void
3905 close_and_delete_fd_from_list(
3906 	SOCKET fd
3907 	)
3908 {
3909 	vsock_t *lsock;
3910 
3911 	UNLINK_EXPR_SLIST(lsock, fd_list, fd ==
3912 	    UNLINK_EXPR_SLIST_CURRENT()->fd, link, vsock_t);
3913 
3914 	if (lsock != NULL) {
3915 		switch (lsock->type) {
3916 		case FD_TYPE_SOCKET:
3917 			closesocket(lsock->fd);
3918 			break;
3919 
3920 		case FD_TYPE_FILE:
3921 			close(lsock->fd);
3922 			break;
3923 
3924 		default:
3925 			msyslog(LOG_ERR,
3926 				"internal error - illegal descriptor type %d - EXITING",
3927 				(int)lsock->type);
3928 			exit(1);
3929 		}
3930 
3931 		free(lsock);
3932 		/*
3933 		 * I/O Completion Ports don't care about select and fd_set
3934 		 */
3935 #ifndef HAVE_IO_COMPLETION_PORT
3936 		/*
3937 		 * remove from activefds
3938 		 */
3939 		FD_CLR(fd, &activefds);
3940 
3941 		if (fd == maxactivefd && maxactivefd) {
3942 			int i;
3943 			NTP_INSIST(maxactivefd - 1 < FD_SETSIZE);
3944 			for (i = maxactivefd - 1; i >= 0; i--)
3945 				if (FD_ISSET(i, &activefds)) {
3946 					maxactivefd = i;
3947 					break;
3948 				}
3949 			NTP_INSIST(fd != maxactivefd);
3950 		}
3951 #endif
3952 	}
3953 }
3954 
3955 static void
3956 add_addr_to_list(
3957 	sockaddr_u *addr,
3958 	struct interface *interface
3959 	)
3960 {
3961 	remaddr_t *laddr;
3962 
3963 #ifdef DEBUG
3964 	if (find_addr_in_list(addr) == NULL) {
3965 #endif
3966 		/* not there yet - add to list */
3967 		laddr = emalloc(sizeof(*laddr));
3968 		memcpy(&laddr->addr, addr, sizeof(laddr->addr));
3969 		laddr->interface = interface;
3970 
3971 		LINK_SLIST(remoteaddr_list, laddr, link);
3972 
3973 		DPRINTF(4, ("Added addr %s to list of addresses\n",
3974 			    stoa(addr)));
3975 #ifdef DEBUG
3976 	} else
3977 		DPRINTF(4, ("WARNING: Attempt to add duplicate addr %s to address list\n",
3978 			    stoa(addr)));
3979 #endif
3980 }
3981 
3982 
3983 static void
3984 delete_addr_from_list(
3985 	sockaddr_u *addr
3986 	)
3987 {
3988 	remaddr_t *unlinked;
3989 
3990 	UNLINK_EXPR_SLIST(unlinked, remoteaddr_list, SOCK_EQ(addr,
3991 		&(UNLINK_EXPR_SLIST_CURRENT()->addr)), link, remaddr_t);
3992 
3993 	if (unlinked != NULL) {
3994 		DPRINTF(4, ("Deleted addr %s from list of addresses\n",
3995 			stoa(addr)));
3996 		free(unlinked);
3997 	}
3998 }
3999 
4000 
4001 static void
4002 delete_interface_from_list(
4003 	struct interface *iface
4004 	)
4005 {
4006 	remaddr_t *unlinked;
4007 
4008 	do {
4009 		UNLINK_EXPR_SLIST(unlinked, remoteaddr_list, iface ==
4010 		    UNLINK_EXPR_SLIST_CURRENT()->interface, link,
4011 		    remaddr_t);
4012 
4013 		if (unlinked != NULL) {
4014 			DPRINTF(4, ("Deleted addr %s for interface #%d %s from list of addresses\n",
4015 				stoa(&unlinked->addr), iface->ifnum,
4016 				iface->name));
4017 			if (addr_ismulticast(&unlinked->addr))
4018 				/* find a new interface to use */
4019 				io_multicast_add(&unlinked->addr);
4020 			free(unlinked);
4021 		}
4022 	} while (unlinked != NULL);
4023 }
4024 
4025 
4026 static struct interface *
4027 find_addr_in_list(
4028 	sockaddr_u *addr
4029 	)
4030 {
4031 	remaddr_t *entry;
4032 
4033 	DPRINTF(4, ("Searching for addr %s in list of addresses - ",
4034 		    stoa(addr)));
4035 
4036 	for (entry = remoteaddr_list;
4037 	     entry != NULL;
4038 	     entry = entry->link)
4039 		if (SOCK_EQ(&entry->addr, addr)) {
4040 			DPRINTF(4, ("FOUND\n"));
4041 			return entry->interface;
4042 		}
4043 
4044 	DPRINTF(4, ("NOT FOUND\n"));
4045 	return NULL;
4046 }
4047 
4048 static inline isc_boolean_t
4049 same_network_v4(
4050 	struct sockaddr_in *addr1,
4051 	struct sockaddr_in *mask,
4052 	struct sockaddr_in *addr2
4053 	)
4054 {
4055 	return (addr1->sin_addr.s_addr & mask->sin_addr.s_addr)
4056 	       == (addr2->sin_addr.s_addr & mask->sin_addr.s_addr);
4057 }
4058 
4059 #ifdef INCLUDE_IPV6_SUPPORT
4060 static inline isc_boolean_t
4061 same_network_v6(
4062 	struct sockaddr_in6 *addr1,
4063 	struct sockaddr_in6 *mask,
4064 	struct sockaddr_in6 *addr2
4065 	)
4066 {
4067 	size_t i;
4068 
4069 	for (i = 0;
4070 	     i < sizeof(addr1->sin6_addr.s6_addr) /
4071 	         sizeof(addr1->sin6_addr.s6_addr[0]);
4072 	     i++)
4073 
4074 		if ((addr1->sin6_addr.s6_addr[i] &
4075 		     mask->sin6_addr.s6_addr[i])
4076 		    !=
4077 		    (addr2->sin6_addr.s6_addr[i] &
4078 		     mask->sin6_addr.s6_addr[i]))
4079 
4080 			return ISC_FALSE;
4081 
4082 	return ISC_TRUE;
4083 }
4084 #endif	/* INCLUDE_IPV6_SUPPORT */
4085 
4086 
4087 static isc_boolean_t
4088 same_network(
4089 	sockaddr_u *a1,
4090 	sockaddr_u *mask,
4091 	sockaddr_u *a2
4092 	)
4093 {
4094 	isc_boolean_t sn;
4095 
4096 	if (AF(a1) != AF(a2))
4097 		sn = ISC_FALSE;
4098 	else if (IS_IPV4(a1))
4099 		sn = same_network_v4(&a1->sa4, &mask->sa4, &a2->sa4);
4100 #ifdef INCLUDE_IPV6_SUPPORT
4101 	else if (IS_IPV6(a1))
4102 		sn = same_network_v6(&a1->sa6, &mask->sa6, &a2->sa6);
4103 #endif
4104 	else
4105 		sn = ISC_FALSE;
4106 
4107 	return sn;
4108 }
4109 
4110 /*
4111  * Find an address in the list on the same network as addr which is not
4112  * addr.
4113  */
4114 static struct interface *
4115 find_samenet_addr_in_list(
4116 	sockaddr_u *addr
4117 	)
4118 {
4119 	remaddr_t *entry;
4120 
4121 	DPRINTF(4, ("Searching for addr with same subnet as %s in list of addresses - ",
4122 		    stoa(addr)));
4123 
4124 	for (entry = remoteaddr_list;
4125 	     entry != NULL;
4126 	     entry = entry->link)
4127 
4128 		if (!SOCK_EQ(addr, &entry->addr)
4129 		    && same_network(&entry->addr,
4130 				    &entry->interface->mask,
4131 				    addr)) {
4132 			DPRINTF(4, ("FOUND\n"));
4133 			return entry->interface;
4134 		}
4135 
4136 	DPRINTF(4, ("NOT FOUND\n"));
4137 	return NULL;
4138 }
4139 
4140 
4141 /*
4142  * Find the given address with the all given flags set in the list
4143  */
4144 static struct interface *
4145 find_flagged_addr_in_list(
4146 	sockaddr_u *	addr,
4147 	int		flags
4148 	)
4149 {
4150 	remaddr_t *entry;
4151 
4152 	DPRINTF(4, ("Finding addr %s with flags %d in list: ",
4153 		    stoa(addr), flags));
4154 
4155 	for (entry = remoteaddr_list;
4156 	     entry != NULL;
4157 	     entry = entry->link)
4158 
4159 		if (SOCK_EQ(&entry->addr, addr)
4160 		    && (entry->interface->flags & flags) == flags) {
4161 
4162 			DPRINTF(4, ("FOUND\n"));
4163 			return entry->interface;
4164 		}
4165 
4166 	DPRINTF(4, ("NOT FOUND\n"));
4167 	return NULL;
4168 }
4169 
4170 
4171 #ifdef HAS_ROUTING_SOCKET
4172 # ifndef UPDATE_GRACE
4173 #  define UPDATE_GRACE	2	/* wait UPDATE_GRACE seconds before scanning */
4174 # endif
4175 
4176 static void
4177 process_routing_msgs(struct asyncio_reader *reader)
4178 {
4179 	char buffer[5120];
4180 	int cnt, msg_type;
4181 #ifdef HAVE_RTNETLINK
4182 	struct nlmsghdr *nh;
4183 #else
4184 	struct rt_msghdr *rtm;
4185 	char *p;
4186 #endif
4187 
4188 	if (disable_dynamic_updates) {
4189 		/*
4190 		 * discard ourselves if we are not needed any more
4191 		 * usually happens when running unprivileged
4192 		 */
4193 		remove_asyncio_reader(reader);
4194 		delete_asyncio_reader(reader);
4195 		return;
4196 	}
4197 
4198 	cnt = read(reader->fd, buffer, sizeof(buffer));
4199 
4200 	if (cnt < 0) {
4201 		msyslog(LOG_ERR,
4202 			"i/o error on routing socket %m - disabling");
4203 		remove_asyncio_reader(reader);
4204 		delete_asyncio_reader(reader);
4205 		return;
4206 	}
4207 
4208 	/*
4209 	 * process routing message
4210 	 */
4211 #ifdef HAVE_RTNETLINK
4212 	for (nh = (struct nlmsghdr *)buffer;
4213 	     NLMSG_OK(nh, cnt);
4214 	     nh = NLMSG_NEXT(nh, cnt)) {
4215 		msg_type = nh->nlmsg_type;
4216 #else
4217 	for (p = buffer;
4218 	     (p + sizeof(struct rt_msghdr)) <= (buffer + cnt);
4219 	     p += rtm->rtm_msglen) {
4220 		rtm = (struct rt_msghdr *)p;
4221 		if (rtm->rtm_version != RTM_VERSION) {
4222 			msyslog(LOG_ERR,
4223 				"version mismatch (got %d - expected %d) on routing socket - disabling",
4224 				rtm->rtm_version, RTM_VERSION);
4225 
4226 			remove_asyncio_reader(reader);
4227 			delete_asyncio_reader(reader);
4228 			return;
4229 		}
4230 		msg_type = rtm->rtm_type;
4231 #endif
4232 		switch (msg_type) {
4233 #ifdef RTM_NEWADDR
4234 		case RTM_NEWADDR:
4235 #endif
4236 #ifdef RTM_DELADDR
4237 		case RTM_DELADDR:
4238 #endif
4239 #ifdef RTM_ADD
4240 		case RTM_ADD:
4241 #endif
4242 #ifdef RTM_DELETE
4243 		case RTM_DELETE:
4244 #endif
4245 #ifdef RTM_REDIRECT
4246 		case RTM_REDIRECT:
4247 #endif
4248 #ifdef RTM_CHANGE
4249 		case RTM_CHANGE:
4250 #endif
4251 #ifdef RTM_LOSING
4252 		case RTM_LOSING:
4253 #endif
4254 #ifdef RTM_IFINFO
4255 		case RTM_IFINFO:
4256 #endif
4257 #ifdef RTM_IFANNOUNCE
4258 		case RTM_IFANNOUNCE:
4259 #endif
4260 #ifdef RTM_NEWLINK
4261 		case RTM_NEWLINK:
4262 #endif
4263 #ifdef RTM_DELLINK
4264 		case RTM_DELLINK:
4265 #endif
4266 #ifdef RTM_NEWROUTE
4267 		case RTM_NEWROUTE:
4268 #endif
4269 #ifdef RTM_DELROUTE
4270 		case RTM_DELROUTE:
4271 #endif
4272 			/*
4273 			 * we are keen on new and deleted addresses and
4274 			 * if an interface goes up and down or routing
4275 			 * changes
4276 			 */
4277 			DPRINTF(3, ("routing message op = %d: scheduling interface update\n",
4278 				    msg_type));
4279 			timer_interfacetimeout(current_time + UPDATE_GRACE);
4280 			break;
4281 #ifdef HAVE_RTNETLINK
4282 		case NLMSG_DONE:
4283 			/* end of multipart message */
4284 			return;
4285 #endif
4286 		default:
4287 			/*
4288 			 * the rest doesn't bother us.
4289 			 */
4290 			DPRINTF(4, ("routing message op = %d: ignored\n",
4291 				    msg_type));
4292 			break;
4293 		}
4294 	}
4295 }
4296 
4297 /*
4298  * set up routing notifications
4299  */
4300 static void
4301 init_async_notifications()
4302 {
4303 	struct asyncio_reader *reader;
4304 #ifdef HAVE_RTNETLINK
4305 	int fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
4306 	struct sockaddr_nl sa;
4307 #else
4308 	int fd = socket(PF_ROUTE, SOCK_RAW, 0);
4309 #endif
4310 	if (fd < 0) {
4311 		msyslog(LOG_ERR,
4312 			"unable to open routing socket (%m) - using polled interface update");
4313 		return;
4314 	}
4315 
4316 	fd = move_fd(fd);
4317 #ifdef HAVE_RTNETLINK
4318 	memset(&sa, 0, sizeof(sa));
4319 	sa.nl_family = PF_NETLINK;
4320 	sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_IFADDR
4321 		       | RTMGRP_IPV6_IFADDR | RTMGRP_IPV4_ROUTE
4322 		       | RTMGRP_IPV4_MROUTE | RTMGRP_IPV6_ROUTE
4323 		       | RTMGRP_IPV6_MROUTE;
4324 	if (bind(fd, (struct sockaddr *)&sa, sizeof(sa)) < 0) {
4325 		msyslog(LOG_ERR,
4326 			"bind failed on routing socket (%m) - using polled interface update");
4327 		return;
4328 	}
4329 #endif
4330 	init_nonblocking_io(fd);
4331 #if defined(HAVE_SIGNALED_IO)
4332 	init_socket_sig(fd);
4333 #endif /* HAVE_SIGNALED_IO */
4334 
4335 	reader = new_asyncio_reader();
4336 
4337 	reader->fd = fd;
4338 	reader->receiver = process_routing_msgs;
4339 
4340 	add_asyncio_reader(reader, FD_TYPE_SOCKET);
4341 	msyslog(LOG_INFO,
4342 		"Listening on routing socket on fd #%d for interface updates",
4343 		fd);
4344 }
4345 #else
4346 /* HAS_ROUTING_SOCKET not defined */
4347 static void
4348 init_async_notifications(void)
4349 {
4350 }
4351 #endif
4352